Advances in Biochemical Engineering/

38 Biotechnology

Managing Editor: A. Fiechter

Lignocellulosic
Materials
With Contributionsby
L.R.Lynd,St.Marsili-Libelli,E Parisi

With 49 Figures and 18 Tables

l~i Springer-Veflag
Berlin Heidelberg NewYork
London Paris Tokyo
ISBN 3-540-50163-0 Springer-Verlag Berlin Heidelberg New York
ISBN 0-387-50163-0 Springer-Verlag New York Berlin Heidelberg

This work is subject to copyright. All rights are reserved, whether the whole or part of
the material is concerned, specifically the rights of translation, reprinting, re-use of illustra-
tions, recitation, broadcasting, reproduction on microfilms or in other ways, and storage
in data banks. Duplicatiou of this publication or parts thereof is only permitted under
the provisions of the German Copyright Law of September 9, 1965, in its version of June 24,
1985, and a copyright fee must always be paid. Violations fall under the prosecution act
of the German Copyright Law.
© by Springer-Verlag Berlin - Heidelberg 1989
Library of Congress Catalog Coard Number 72-152360
The use of registered names, trademarks, etc. in this publication does not imply, even
in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.

Bookbinding: Liideritz & Bauer, Berlin

2152/3020-543210
Managing Editor

Professor Dr. A. Fiechter

Institut f/Jr Biotechnologie, Eidgen6ssische Technische Hochschule
ETH -- H6nggerberg, CH-8093 Ztlrich

Editorial Board

Prof. Dr. S. Aiba Department of Fermentation Technology, Faculty of

Engineering, Osaka University, Yamada-Kan~i, Suita-Shi,
Osaka 565, Japan
Prof. Dr. H. R. Bungay Rensselaer Polytechnic Institute, Dept. of Chem. and
Environment. Engineering, Troy, NY 12180-3590/USA
Prof. Dr. Ch. L. Cooney Massachusetts Institute of Technology,
Department of Chemical Engineering,
Cambridge, Massachusetts 02139/USA
Prof. Dr. A. L. Demain Massachusetts Institute of Technology, Dept. of
Nutrition & Food Sc., Room 56-125
Cambridge, Massachusetts 02139/USA
Prof. Dr. S. Fukui Dept. of Industrial Chemistry, Faculty of
Engineering, Sakyo-Ku, Kyoto 606, Japan
Prof. Dr. K. Kiesl&h Gesellschaft fiir Biotechnologie, Forschung mbH,
Mascheroder Weg 1, D-3300 Braunschweig
Prof. Dr. A. M. Klibanov Massachusetts Institute of Technology, Dept. of Applied
Biological Sciences, Cambridge, Massachusetts 02139/USA
Prof. Dr. R. M. Lafferty Techn. Hochschule Graz, Institut ftir Biochem. Technol.,
Schl6gelgasse 9, A-8010 Graz
Prof. Dr. S. B. Primrose General Manager, Molecular Biology Division,
Amersham International plc., White Lion Road Amersham,
Buckinghamshire HP7 9LL, England
Pros Dr. H. ~L Rehm Westf. Wilhelms Universitfit, lnstitut fiir Mikrobiologie,
Corrensstr. 3, D-4400 M/inster
Prof. Dr. P. L. Rogers School of Biological Technology, The University of
New South Wales. P.O. Box 1,
Kensington, New South Wales, Australia 2033
Prof. Dr. H. Sahm lnstitut ffir Biotechnologie, Kernforschungsanlage
Jtilich, D-5170 Jtilich
Prof. Dr. K. Schiigerl lnstitut f/Jr Technische Chemie, Universitgt Hannover,
Callinstrage 3, D-3000 Hannover
Prof. Dr. S. Suzuki Tokyo Institute of Technology,
Nagatsuta Campus, Res. Lab. of Resources Utilization,
4259, Nagatsuta, Midori-ku, Yokohama 227/Japan
Prof. Dr. G. T. Tsao Director, Lab. of Renewable Resources Eng., A. A. Potter
Eng. Center, Purdue University, West Lafayette,
IN 47907/USA
Dr. K. Venkat Corporate Director Science and Technology, H. J. Heinz
Company U,S. Steel Building, P.O. Box 57, Pittsburgh,
PA 15230/USA
ProL Dr. E.-L. Winnacker Universit~it Miinchen, Institut f. Biochemie, Karlsstr. 23,
D-8000 Mfinchen 2
Table of Contents

Production of Ethanol from LignoceHulosic

Materials Using Thermophilic Bacteria:
Critical Evaluation of Potential and Review
L. R. Lynd . . . . . . . . . . . . . . . . . . . . .

Advances in Lignocellulosics Hydrolysis

and in the Utilization of the Hydrolyzates
F. Parisi . . . . . . . . . . . . . . . . . . . . . . . 53

Modelling, Identification and Control of

the Activated Sludge Process
St. Marsili-Libelli . . . . . . . . . . . . . . . . . . . 89

Author Index Volumes 1-38 . . . . . . . . . . . . . . 149

Production of Ethanol from Lignocellulosic Materials Using
Thermophilic Bacteria: Critical Evaluation of Potential
and Review

Lee Rybeck Lynd

Thayer School of Engineering, Dartmouth C o l l e g e H a n o v e r N H 03755, U S A

1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2 Overview of Ethanol Production from Biomass . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.l Resource Aspects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.1.t Petroleum Supply and Demand . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.1.2 Substrate Availability, Composition. and Potential Ethanol Yield . . . . . . . . . . . . . . 5
2.1.3 Environmental Impact . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1l
2.2 Technological Aspects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.2.1 Current Ethanol Production and Utilization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.2.2 Production Technology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.2.3 Properties of Ethanol as a Fuel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.2.4 Energetic Considerations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . I8
2.2.5 End-Use . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3 Potential of Thermophilic Bacteria for Ethanol Production . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.1 Identification of Distinguishing Features . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.2 Evaluation of Distinguishing Features . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.2.1 General Impact . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.2.2 Basis for Economic Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.2.3 Economic Impact . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.3 Comparison with Other Ethanol Production Processes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
4 Progress Toward Realization of the Potential of Thermophilic Bacteria for Ethanol Production 36
4.1 Cellulase Production and Activity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
4.2 Utilization of Pentose Sugars . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .............. 38
4.3 Ethanol Tolerance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
4.4 End-Product Metabolism and Ethanol Yields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
5 Concluding Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
6 Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
7 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

Resource and technological aspects of ethanol production are considered. Conversion of lignocellu-
losic substrates to ethanol via thermophilic bacteria is then addressed, with particular emphasis on
evaluation from an engineering perspective.
The biological conversion of lignoccllulosic materials to ethanol is a versatile process which can be
used in various applications for replacing or improving petroleum products, treating wastes, or reduc-
ing air pollution. Petroleum replacement can be in relation to neat fuels, fuel additives, or raw materials.
Waste treatment applications may be either for wastes which require treatment (e.g. municipal solid
waste) or wastes which do not (e.g. many forestry and agricultural residues). Biological treatment of
solid wastes with concomitant ethanol production may become attractive in that solid wastes re-
present less expensive substrates than those usually considered for ethanol production. In addition,
the potential energetic yield of ethanol production is about twice that of electricity generation, and
electricity and ethanol have comparable value per unit energy.
Estimated recoverable oil reserves represent a resource approximately 75 times the current annual
consumption on a world-wide basis. However, some countries are in a particularly poor position with

Advances in Biochemical Engineering/

Biotechnology, Vol 38
Managing Editor: A. Fiechter
Sprlnger-Ver/~g Rerlin Heidelberg 1989
2 L.R. Lynd

regard to petroleum supply and demand. For example the U.S. estimated recoverable oil reserves
represent approximately 15 times the current annual consttmption. The annual ethanol production
potential in the U.S. achievablewithin 20 years is estimated at 1.3 x 1013MJ based on a compilation of
estimates for the rates of production and availability of various biomass materials. Relative contri-
butions to this potential are: 41% for wastes, 39 % for energy-devoted forestry, and 19 % for energy-
devoted agriculture. Notably only 6 % of the total ethanol production potential is derived from corn.
Pentose sugars represent 28 % of the total potential with hexose sugars the remainder. Ethanol can
displace gasoline at a ratio of about 1: 1.3 on an energetic basis, thus 1.3 x 1013MJ of ethanol can
displace about 1.7 x 1013MJ of gasoline. The U.S. ethanol production potential of 1.3 x 10x3 MJ,
or 1.7 x 1013MJ of displaced gasoline,can be compared to the yearly U.S. consumption of 7.5 x 1013MJ
for energy of all kinds, 2 x 10la MJ for the transportation sector, and 1.2 x 1013MJ for gasoline.
Four distinguishing features of thermophilic bacteria for ethanol production in comparison to
yeast systems are identified. These include the advantages of pentose utilization and in situ cellulase
production and cellulose utilization, and the disadvantages of low ethanol tolerance and low ethanol
yield. Many frequently-cited advantages are not considered to be of great significancefrom an economic
viewpoint, including facilitated product recovery and high conversion rates. The economic impacts
of the distinguishing features of thermophiles for ethanol production are evaluated relative to a base-
case process for ethanol production consisting of pretreated hardwood hydrolysis using Trichoderma
reesei cellulase followed by conversion of soluble hexose sugars by yeast and reaction of xylose to
furfural. Relative to the base case, the impact of in situ cellulase production and substrate hydrolysis
is to lower the ethanol sellingcost by 37~o, and the impact of pentose utilization is to lower the cost by
23 %. These two features together increase the ethanol yield per unit wood substrate by 47 % over the
base case. The increased cost of ethanol separation at low concentrations appears to be relatively
small if energy-efficientprocesses are used, however such processes have not yet been implemented
on a large scale. High ethanol yields must be obtained if thermophilic ethanol production is to be
practiced on a significant scale.
Research results pertaining to the distinguishing features of thermophiles for ethanol production
are reviewed. Critical research areas are proposed for closing the large gap between the potential of
thermophilic bacteria for ethanol production and that which has been experimentally realized to
date. These include process-oriented studies utilizing potentially realistic substrates and conditions,
and both biological and engineering approaches to increasing ethanol yields.

1 Introduction
Widespread recognition of the finite nature of the world's petroleum resources has
led to examination of alternative sources of materials and energy. One such alternative
is biological ethanol production from renewably-produced products of photosyn-
thesis. The most a b u n d a n t products of photosynthesis, also considered to be the
most a b u n d a n t renewable natural resource available to h u m a n d k i n d , are lignocellu-
losic materials 1). The composition and structure of lignocellulosic materials have
been reviewed l j, and the composition, availability and ethanol production potential
are considered in Sect. 2.1.2.
Anaerobic bacteria with optimal growth temperatures of > 60 ~ have frequently
been proposed for the production of ethanol from lignocellulose. Over the last ten
years, rather intense research programs h a v e b e e n directed toward ethanol production
using thermophilic bacteria in perhaps a dozen laboratories world-wide. Recent and
comprehensive reviews are available which consider the cellulase enzyme complex of
Clostridium thermocellum 2~ the genetics and biochemistry of Clostridium including
thermophilic species 3), aspects of the general and applied physiology of thermophilic
bacteria 4, 5, 6.7, s, 9,10), thermophilic cellulolytic bacteria 11,12), and thermophilic
conversion of cellulose to ethanol 13,14). M a n y of these reviews contain exhaustive
lists of thermophiles and their properties, as well as detailed discussions of biochemi-
stry and physiology relating to thermophiles and thermophilic ethanol production.
Production of Ethanol from LignocellulosicMaterials Using ThermophilicBacteria 3

Typically a thorough consideration of biologically-related topics is followed by

more cursory reference to engineering-related issues.
There is no attempt made here to duplicate the many excellent biologically-oriented
reviews of thermophilic ethanol production. Instead this paper considers engineering
topics first, and in detail, followed by a focused review of biologically-related topics.
Engineering topics addressed include an overview of ethanol production from bio-
mass, Sect. 2, which focuses on lignocellulosic materials and includes both resource
and technological aspects. The potential of thermophilic bacteria for ethanol pro-
duction in comparison to yeast-based processes is then considered in detail from an
engineering and economic perspective in Sect. 3. Section 4 considers progress toward
realization of the potential of thermophilic bacteria for ethanol production, emphasiz-
ing research areas judged to be of particular applied significance based on consi-
derations addressed in previous sections. Section 5 offers concluding remarks. This
paper is intended to be an interim progress report on the use of thermophilic bacteria
for ethanol production from lignocellulose. Both review of what has been done and
exposition of what could be done are important parts of this task.
Consideration of several topics, in particular ethanol production potential, will
focus on the United States. This focus is convenient because it allows more specific
statements to be made about issues such as substrate availability and economics. In
addition, the situation with respect to petroleum supply and demand is particularly
acute in the U.S., as discussed in Sect. 2.1.1. Because of this, the potential of biomass
resources to replace petroleum in the U.S. has frequently been considered, albeit
with different conclusions, and abundant data are available.
The rate of oil consumption in the U.S. is approximately 27 % of the world-wide
rate, and the highest of any country in the world [15]. However, the quantity of ethanol
which may be derived from biomass in the U.S. is representative of that in other
countries on a per capita basis. Vergara and Pimentel [16] give values for the annual
photosynthetic energy (MJ x 103) fixed per capita of 250 for the U.S., 250 for Sweden,
38 for India, 465 for Brazil, and 536 for Sudan at 1976 population levels. The total
solar energy fixed by photosynthesis worldwide is approximately 1 x 1015 MJ a-1 i7),
which may be compared to the 1980 world-wide annual demand for petroleum~
1.26 x 1014 MJ, and for all forms of energy, 2.9 x 1014 MJ a -1 is). At the 1976 world
population of about 4 billion people, the world-wide solar energy fixed per capita
was 250 x 103 MJ per capita. In 1987 this value is 200 x 103 MJ per capita.

2 Overview of Ethanol Production from Biomass

2.1 Resource Aspects

2.1.1 Petroleum Supply and Demand
A paper written in 1987 on production of ethanol from renewable resources may
seem out of place. After all, oil sells for under 20 U.S.$ per barrel, and at times has
been less than 1/3 of its real price in 1980. Though the fact of finite petroleum reserves is
certainly felt with less immediacy today in economic terms, the long-range picture
has changed very little in resource terms.
Conversion of biomass to ethanol has received attention as a means of replacing
energy and materials presently derived from oil. The long-term motivation for re-
4 L.R. Lynd

placing oil is the finite nature of its supply, whereas biomass is renewably available.
There is little doubt that finding and developing the world's as yet unrecovered oil
reserves will be progressively more difficult and costly 18). Thus in the shorter term
before oil supplies near exhaustion, the price of oil may be expected to rise, and bio-
mass feedstocks may become more competitive. The issues of limited oil supply and
the prospect of an improvement in the relative economics of biomass feedstocks
compared to oil are addressed in the paragraphs below.
In a comprehensive study published in 1982, Grathwohl is) presents estimates for
the worldwide total amount of recoverable oil at 2.4 to 3.6 • 10~1 metric tons, based
on 40 ~ oil recovery during drilling. As of January 1979, ~ 15 ~ of this quantity had
already been recovered, and 25 ~ represented proven reserves. In 1980 the worldwide
oil consumption was 3.0 x 109 t a -1 is), or about 1 ~o of the total recoverable oil per
year. Thus in 1987, approximately 23 ~ of the total recoverable oil has already been
used. Because of the uneven geographical distribution of petroleum reserves, the
situation is more acute in some countries than others. For example, the U.S. had used
45 ~o of its total estimated recoverable oil by the beginning of 1979 a5), and 64 ~o by the
beginning of 1987 15,19). Data cited by Grathwohl for the beginning of 1979 puts the
estimated total amount of recoverable oil remaining in the U.S. at 19.7 x 109 t.
Deducting oil consumption since 1979 2o) results in a value of 13.1 • 109 t for the oil
remaining at the beginning of 1987. This quantity represents 15 times the annual U.S.
oil consumption in 1986 of 0.87 x 109 t. The U.S. Geological Survey (USGS) re-
cently estimated the undiscovered recoverable oil in the U.S. at 4.9 x 109t za), an
amount equal to 5.6 times the 1986 annual consumption.
Most estimates of future oil consumption for both the U.S. and the world show a

60

50

40

:30

20

10

0 I I I I I I I

1955 1960 1965 1970 1 9 7 5 1980 1 9 8 5 1990

Fig. 1. Historical view of estimated worldwide oil reserves, and oil price, consumption rate, and prices
of ethanol precursors for the U.S. --@--, Estimates of recoverable oil reserves from xs, (values
plotted are 10-13 • actual values in kg a-l); -- ~--, oil price from z2) (1984 U.S. $ barrel 1); _ • ,
oil consumption from ls,2z) (0.5 x 10-12 x MJ a-l); --A--, ethylene price from z3) (0.5 x 1984 U.S.
$ kg-1); _ I - - , price for delivered green mixed hardwood chips from 24.25) (1984 U.S. $ t-1)
Production of Ethanol from LignocellulosicMaterials Using ThermophilicBacteria 5

small increase in the rate of consumption up to the year 2000, with a decrease in
consumption beginning shortly thereafter ~5.22~. The U.S. Department of Energy 22)
estimates that the real price of oil in the U.S. will fall slightly by 1990, and increase by
nearly 3-fold in the 20 years thereafter.
Figure 1 presents an historical view of estimates for the size of the world oil resource~
and oil price and utilization rate in the United States. United States price histories
are also presented for ethylene, the petroleum-derived precursor for ethanol synthesis,
and delivered green hardwood chips, representative of one of the primary lignocellu-
losic ethanol precursors considered herein. It may be seen that estimates of the total
recoverable oil reserve generally increased anti! I970, but remained nearly constant
between 1970 and 1980. U.S. petroleum consumption relative to 1960 had increased
by 74 ~o in 1973, and had decreased to 56 ~ above 1960 consumption by 1980 15,221
The price of oil, given in real U.S.$ 2z), increased sharply in the mid 1970's and again
in the late 1970's, and then fell sharply in the mid 1980's. These price changes do not
reflect changes in the size of the estimated oil resource, but rather factors such as
OPEC policy, and over-compensation in the development of oil pumping capacity.
The real price of ethylene 23) follows the price of oil, from which it is derived, fairly
closely. This is not true for the price of wood, which has a relatively stable history in
the U.S. and has not increased in real terms since 1960 24.25). Thus if oil prices increase
as expected, it appears likely that lignocellulosic substrates such as wood will become
more attractive sources of raw materials.

2.1.2 Substrate Availability, Composition, and Potential Ethanol Yield

In 1981, 4.6 ~ of U.S. refinery output was devoted to chemical production, whereas
44.8 ~ was used for gasoline, and 87 ~o for fuels of all kinds 17~.In light of this difference
in material requirements, several studies have concluded that large-scale utilization
of biomass for chemical production, with feedstocks including but not limited to
ethanol, is a realistic possibility, whereas fuel production is not realistic z6.27,2s~
However others differ in their evaluation of the potential contribution of biomass to
meet fuel needs 29,3o.31) For example, Ferchak and Pye 30) concluded that the
theoretical limit of potential biomass resources for production of fuel ethanol in the
U.S. exceeds foreseeable demand. It may be noted that the amount of ethanol required
to mix with all gasoline used at 10 }s ethanol is comparable te the total refinery output
devoted to chemical production.
Faced with this difference in evaluation of the biomass resource, it is useful to
consider estimates of the size of this resource. Table 1 presents estimates for the avail-
ability of biomass in the U.S. from waste materials, forestry, and agricultural crops
from 8 different studies. The three estimates for total waste are fairly close at 856 to
1081 • 106 t a -1. In general, forest and agricultural residues necessary to maintain
soil fertility are not included in estimates for total or collectable wastes. The two
estimates for collectable waste are somewhat more disparate at 502 and 864 million
tons a -~. Obviously estimates for the fraction of a waste which is collectable depend
on the value attached to the waste materials, which is usually unspecified. The forest
production (t a -1) available for energy production is variously estimated at 182 and
245 in the late 1970's, and 280 to 560 and 909 in the year 2000. The higher of the two
estimates for forest production in the year 2000 is based on a greater role for high
productivity energy-devoted silviculture. The availability of agricultural crops,
Table 1. Estimates for the availability of biomass in the U.S. from waste materials, forestry, and agricultural crops (millions of dry t a-1), ~,

Young Jeffries Ng et al. Ferchak Goldstien OTA Vergara Humphrey

et al. 1983 1983 and Pye 1981 1980 and Pimentel et al.
1986 1981 1979 1977

Waste Materials
Agricultural
wastes + residues
Total 350 736 382 323 381 430 364
Collectable 364 289 2532 75
Forestry and mill
Total 348 209 76 55
Collectable 91 159 155 253 b
Urban wastes
Total 70 1512 146 123 136
Collectable 182 48 1052
Manure
Total 236 182 255 182
Collectable 227 10 127
Sewage
Total 13 14
Other 260 105
Total 1251 >909 1081 856
Total collectable 864 502

Forestry
Total c
late 70's 379 409 616-1639
2000 (assumes intensive 2727
management)
Available for energy production
late 70's 182 245 7z
2000 909 a 280-560
Agricultural crops
Corn O
1980, with by-product 69
utilization 0
2000, total m
69-1575
Forage Grasses
mid 80s 91-182
2000 0-414 f
Year o f data 1982 1978-1981 1977-1979 1976-1979 ~1976 1976-1978 1975-1978 ~1975
0

a Data are from 1 6 , 2 7 , 3 0 , 3 1 , 3 2 , 3 3 , 3 4 , 3 5 ) .

b values are for the year 2000; 0~
c Estimates are based on commercial forest land available in excess o f the requirements of the forest products industry; estimates also do not include wastes
from the forest products industry;
a this value assumes intensive m a n a g e m e n t a n d is based on large privately-owned land tracts that meet site, climate, a n d precipitation requirements for deciduous
species silviculture and are available for energy production. This a m o u n t o f land is equal to 1/3 the total forest area estimated available for energy production,
and is also equal to 5.3 % o f the land area in the contiguous United States; g~
the m i n i m u m value is for the case where no agricultural land is available for energy crops and production from corn with by-product utilization is at 1980 levels; 3.
g~
r full development o f grass production would proclude corn production above the level where there is d e m a n d for process by-products

-]

g~

o
8 L.R. Lynd

discussed in more detail below, is estimated at 69 x 106 t a -1 for corn with process
by-products used for animal feed, plus either an additional 0 to 88 x 106 t of corn, or
0 to 414 x 106 t of forage grass. The studies considered in Table 1 which give values
for the mid 1980's and the year 2000 were done in 1980 and 1981. The allocation of
land, research, and manufacturing resources to ethanol production during the 1980's
has been less than envisioned by these studies. Until this situation changes, the data
in Table 1 for the year 2000 are probably more realistically regarded as biomass
availability attainable in a 20 year time span, recognizing that these estimates become
more approximate with time.
Ethanol production from sugar- and starch rich agricultural crops may be considered
to compete with food production 36), is generally considered to have a less favorable
energy balance compared to lignocellulosic substrates (see Sect. 2.2.4), and has the
greatest potential for unfavorable environmental impact (Sect. 2.1.3). However,
some agricultural crops are primarily used for animal feed in many countries. For
example, over 80~o of the total U.S. corn crop is used for feed, with 55-60~ used
domestically, whereas approximately 10~o is used for human consumption 37)
Moreover, the starch fraction of the corn crop can be used for alcohol production
while residues and/or processing by-products still retain considerable feed value 31,3v)
Thus it has been suggested that ethanol can be made from corn ultimately used for
animal feed, with relatively small incremental resource demands 30, 38). In the case
where by-products from ethanol production are used for feed, corn is the most at-
tractive of the grain and sugar crops for ethanol production 31, 37).
There are however limits to the demand for the feed materials available as by-
products from corn-derived ethanol production. In a comprehensive study, the U.S.
Office of Technology Assessment 31) cited values for the point where by-product
utilization will drop at 8 to 11 x 109 L a -1 ethanol for distillers' grain, a by-product
of ethanol production based on corn dry-milling 37), and as much as 26 x 109 L a -1
for corn gluten meal, a by-product of ethanol production based on wet-milling 37).
26 x 1 0 9 L of ethanol correspond to about 69 x 106 t of corn abailable for ethanol
production, or about 70~o of the total amount of corn used for animal feed in the
U.S. in the early eighties 37)
Forage grasses represent agricultural crops with a more favorable energy balance
and lower potential for unfavorable environmental impact in comparison with corn al)
Moreover grass utilization for ethanol production makes use of the entire plant, and
does not depend on by-product utilization. The OTA study 31) foresees a much larger
role for grasses than for corn among agricultural crops used for energy production.
However, this study cautions that there is considerable uncertainty regarding the
availability of agricultural land for energy crops, with between 0 and 26 • 106 ha
available in the year 2000.
Table 2 presents data on the composition of representative cellulosic materials.
Pentans contribute significantly to the total degradable carbohydrate in these ma-
terials. It is notable that the degradable carbohydrate content of urban waste is
approximately 60~o of the dry weight, and that the degradable carbohydrate content of
wood, forage grass and corn are comparable when pentans are considered.
The data in Tables 1 and 2 are used to arrive at estimated values for the carbohy-
drate composition, annual rates of production, and potential ethanol production
from biomass materials achievable within a 20-year time span, presented in Table 3.
In arriving at the values in Table 3, an effort was made to choose intermediate values
Table 2. Composition of respresentative cellulosic materials"

Dry weight
o

Hardwood Softwood Wheat straw Corn stalks Forage grass Corn Manure Urban
Waste
Hexan
Glucan 50 46 35 36.5 72 10-18
O
Galactan 0.8 1.4 0.7 1.1
Mannan 2.5 11.2 0.4 0.6
Total 53.3 58.6 36.1 38.2 42 72 25 37.6
Pentan O

Xylan 17.5 5.7 19.0 17.2 g

Arabanan 0.5 1.0 4.4 2.1 o
Total 18.0 6.7 23.4 19.3 30 16.0
Total degradable 79.7 72.7 66.7 64.4 80.8 80.0 27.2 60.0 3:
Lignh~ 21.0 29.0 14 7.0 6~10 9.5

Data on hardwood, softwood, wheat straw, and corn stalks from 3,~. Data on Corn from 397 data on manure from 4.0,41). Data on forage grasses is estimated
from x.42.43) Any estimate of the composition of municipal waste must be approximate. The carbohydrate compositions shown are calculated from a "typical
waste" based on data from 44,45). The typical waste contains 16.6 ~ newspaper, 24.9 ~ waste paper and cardboard, 16.5 ~ yard waste, 11 ~o food waste, and qQ

2 ~ wood. The cellulose hemicellulose, and lignin content of each fraction (based mainly on data from ~)) is assumed to be: waste papier and cardboard, 65/15/12; g
newspaper, 45/31/20; yard waste, 45/25/15; food waste, 40/25/2; wood, 55/13/24;
b calculation of degradable carbohydrate considers the water of hydrolysis. Thus degradable carbohydrate = (180/162)*hexan + (150/132)*pentan. Calculation
of the degradable carbohydrate in manure is based on 80 ~ carbohydrate in fiber, treating all fiber as hexan
10 L.R. Lynd

Table 3. Estimated annual ethanol production potential for biomass materials achievable within a
20 year timespan

Material Compositiona Collectableb Degradable Potential~

production carbohydrate ethanol
~o Hexan ~ Pentan
(millions of t a -1)

Wastes
Ag. 37 21 273 177 74.9
Forestry 55 17 182 146 61.8
Urban 37.6 16 91 55 23.3
Manure 20 5 136 38 16.1
Other 40 10 91 51 21.6
Totald wastes 38.5 15.4 773 467 198
Forestry
53.3 18.0 545 435 184
Agricultural
Grasses 42 30 182 147 62.2
Corn (with 72 69 55.2 25.9
by-products)
Total 1569 1104 470

a Values based on Table 2,

b values are the author's estimates based on data in Table 1,
c calculated based on: (degradable carbohydrate)* (0.9 degradable carbohydrate utilization)* (0.47
ethanol/utilized carbohydrate), except for corn where complete carbohydrate utilization is assumed,
d totals are weighted averages in the case of ~ hexan and ~ pentan

between the m a x i m u m and m i n i m u m estimates given. Biomass proponents could

argue, doubtless with some validity, that these estimates are less than the ultimate
practical potential. Biomass detractors could argue, also with validity, that a society
requires a strong motivation to devote the resources required to produce alcohol at
the levels in Table 3. The total estimated potential ethanol production from collectable
wastes is 198 x 106 t a - a , with the order o f ethanol production potential: agriculture >
forestry > u r b a n and other > manure. The contribution of forestry other than
wastes of the forest products industry is estimated at 184x 106t a -1, representing
the annual yield from 545 x 106t o f w o o d with the composition o f hardwood.
Forage grasses could add an additional 62.2 M t o f ethanol, while the ethanol potential
from corn ultimately used as an animal feed is estimated at 26 M t assuming wet-
milling processing.
The total estimated annual ethanol production achievable within 20 years from
renewably-produced biomass materials in the U.S. is 470 x 106t a -1, representing
a b o u t 1.3 x 1013 MJ. The relative contributions o f wastes, forestry, and agriculture
are 42 ~ 39 ~o, and 19 ~o respectively. The contribution o f pentose sugars is a b o u t
28 ~ of the total; the contribution o f corn is 6 ~o o f the total. 1.3 x 1013 MJ of ethanol
m a y be c o m p a r e d to the annual usage of in the U.S. for transportation (2 x 1013 M J) 17),
gasoline (1.2 x 1013 MJ) 46) for gasoline, and energy of all kinds (7.5 x 1013 MJ) 17)
Based on the d a t a presented in this section, it is the author's conclusion that the
ethanol production potential of the biomass resource in the United States is significant
relative to the current d e m a n d for liquid t r a n s p o r t a t i o n fuel, and that this end use
for ethanol cannot be dismissed based on substrate supply considerations. Thus
Production of Ethanol from LignocellulosicMaterials Using ThermophilicBacteria 11

characteristics of ethanol for use as both a fuel and chemical feedstock will be consid-
ered in subsequent sections. It may be noted that transportation fuel is the energy-
related end-use for which ethanol is best suited. Most stationary applications offer
little incentive for the lost available energy accompanying ethanol production from
biomass substrates which could be burned directly. Future demand for transportation
fue! will be affected by variables such as fuel price and availability, demographic
changes, transportation fleet efficiency, and the efficiency of particular fuels 221. The
U.S. Department of Energy (DOE) has predicted that total transportation energy
consumption will decline somewhat through 1995 due primarily to increased efficiency
of passenger vehicles.
In a global context, the U.S. has disproportionately high oil and energy consumption
and representative photosynthetic energy fixed per capita as already noted in this
paper. In light of these considerations, the potentially significant contribution of
biomass-derived ethanol for liquid fuel requirements in the U.S. strongly suggests
that very significant potential exists in other countries as well.
The estimates in this section do not take into consideration processing losses,
energy inputs for substrate growth, harvest or transportation, nor the relative fuel
efficiency of ethanol and gasoline, all of which are considered in subsequent sections.
Obviously these factors will affect the net work which can be derived from biomass
fuels. Consideration of ethanol production potential separately from procurement,
conversion, and end use related factors is useful because the latter depend to a much
greater extent on the current or assumed state of several technologies than does the
former.

2.1.3 Environmental Impact

The environmental impacts of ethanol production can be divided into three classes :
those resulting from process waste streams; those resulting from the interplay between
substrate production and land resource considerations such as soil fertility and ero-
sion, and maintaining wildlife habitat; and those relating to ethanol utilization per se.
The former class does not appear to pose problems which cannot be addressed by
conventional waste treatment technology 31,47,48,49). The principal waste streams
are airborne emissions, suspended solids and BOD in wastewater, and solids in the
form of ash and insoluble salts, particularly arising from neutralization of acid.
Several process designs burn residual organics and sludge from wastewater treatment,
thereby providing process energy and eliminating organic solid waste 5o, 51)
Land resource issues have sharply differing potential to be environmental problems
depending on the ethanol feedstock considered. According to the U.S. Office of
Technology Assessment (OTA) 31), the potential for environmental damage associated
with various feedstocks is as follows. Wood and food-processing wastes, animal
wastes and collected logging waste have no significant potential. Grasses should have
few significant adverse impacts for most applications. Crop and logging residues have
some potential for harm if mismanaged, and speculative potential for long-term
damage to productivity because of loss of soil organic matter. Other wood sources
have high potential but theoretically can be managed. Grain and sugar crops have the
highest potefatial. The OTA 31) concluded that biomass has the potential to be an
energy source that has few significant environmental problems and some important
environmental benefits. For a number of reasons however, a vigorous expansion of
12 L.R. Lynd

bioenergy still may cause serious environmental damage because of poorly managed
feedstock supplies and inadequately controlled conversion technologies. Also, some
uncertainties remain about the long-term effects of intensive biomass harvests on soil
fertility.
Somewhat different considerations arise for tropical regions, particularly with
respect to starch and sugar crops. Matsuda and Kubota 52) have investigated the
feasibility of fuel alcohol production in Southeast Asia. These authors concluded
that increased land use to cultivate crops for ethanol production using field agriculture
would cause large scale destruction of the ecosystem in tropical regions.
According to Grathwohl 15) it has definitely been established by measurement that
the atmosphere is currently being enriched in CO2 at the rate of 1.3 ppm per year.
Increased CO2 levels have prompted concern over climatic warming and its effects sa,54).
A very significant environmental advantage of ethanol production is that the photo-
synthesis-ethanol production-ethanol combustion cycle has no net CO2 generation
provided that it is driven by fuels derived from photosynthesis. A second factor relating
to ethanol utilization is the substitution of ethanol for gasoline to reduce air pollution.
Use of ethanol as a gasoline additive to reduce air pollution is receiving attention in
urban areas of the U.S. having air quality problems related to smog formation per-
sonal communication, R. Datta, Michigan Biotechnology Institute, and 55)

2.2 Technological Aspects

2.2.1 Current Ethanol Production and Utilization
Presently, ethanol is produced either by using yeast, or by catalytic hydration of
ethylene. The technology for bioethanol production is discussed in Sect. 2.2.2. Aspects
of ethanol production in the U.S. have recently been considered by Venkatasubra-
manian and Kiem 37), Hacking 56), and Murtagh 57). In 1977, processes using yeast
accounted for 26 % of the total U.S. industrial ethanol production, with the remainder
made from ethylene. Between 1977 and 1982, U.S. ethanol production shifted dra-
matically in favor of the bioproduct 56,s8). During this period, ethylene-derived
ethanol production stayed constant at ~ 908 x 106 L a-1, while bioethanol production
increased from 318 to 2271 x 106 L a -1. A similar trend toward bioethanol has oc-
curred in the European Community 59). The development of fuel ethanol production
in the United States has been stimulated by substantial tax incentives 37,57). Tax
incentives are largely responsible for the difference between the recent prices of
synthetic ethanol (~0.44 $ L -1 anhydrous) as compared to bioethanol (~0.32 $ L -1
anhydrous). Actual production costs for ethylene- and bioethanol are more nearly
equal than these prices would indicate, and both are strongly influenced by the cost
of substrate (see Sect. 3.3). Demand for ethylene by the chemical process industry
has been near capacity levels in the U.S. in 1987 6o).
Presently synthetically-produced ethanol is used in the U.S. for synthesis of acetal-
dehyde and acetic acid (43 %), cosmetics and pharmaceuticals (28 %), cleaning pre-
parations and solvents (16 %), and coatings (13 %) s6). Ethanol for fuel use is currently
produced at the level of about 3 x 1 0 9 L a -1 57), and accounts for thebulk of bio-
logical production. Originally fuel ethanol was developed as a gasoline extender, but
today is utilized primarily as an octane enhancer.
Production of Ethanol from LignocellulosicMaterials Using ThermophilicBacteria 13

2.2.2 Production Technology

The technology for biological production of ethanol from biomass can be divided
into 6 steps : 1) substrate growth, harvesting and transport, 2) substrate pretreatment,
3) substrate hydrolysis, 4) biological conversion, 5) product recovery, including by-
products, and 6) waste treatment. Some steps can be combined or eliminated depend-
ing on the process. The emphasis here will be on ethanol production technology
relevant to lignoeellulose utilization, including both technology in use and in the
research stage.
Substrate growth and harvesting have been considered in detail for wood by Smith
and Corcoran 61) and also by Ferchak and Pye 62). Energy requirements for growth,
harvest, and transportation of lignocellulosic substrates and related equipment are
discussed in Sect. 2.2.4.
Insoluble substrates for ethanol production must usually undergo some pretreat-
ment prior to biological conversion to, at least, provide enough surface area for en-
zymes to obtain access to the substrate. Pretreatment processes for lignocellulosic
materials are the subject of reviews by Datta 63), Wilke et al. 64) Grethlein 65), and
Dale 66). For lignocellulosic materials, the lignin/hemicellulose/cellulose matrix
has low porosity and is therefore resistant to attack by enzymatic hydrolysis 67)
Most successful pretreatments rely on removing either hemicellulose or lignin to
create a material with porosity sufficient to allow significant access for enzymatic
attack. Using pore-size measurements, Weimer and Weston 68) determined that a
minimum pore dimension of approximately 43 A is required for hydrolysis to be
catalyzed by cellulases of both Trichoderma reesei and C. therrnocellum. Cellulose
crystaUinity has an effect on reaction rates, but only becomes a factor once the enzyme
has access to the cellulose fiber. Allowing this access is thought to be the primary
function of pretreatment 69, 70)
As reviewed by Dale 66) and Grethlein 65), a variety of pretreatments have been
studied. Some increase in the rate and yield of enzymatic hydrolysis has been obtained
using acid, base, ammonia (subcritical and supercritical), solvents, heat, explosive
decompression, and combinations of these. Several pretreatments allow subsequent
enzymatic hydrolysis yields of >90 ~ theoretical. Capital and operating costs for
lignocellulose pretreatments such as dilute acid hydrolysis and steam explosion are
a small fraction of total ethanol production costs 51, 70, 71). Thus the importance of the
choice of pretreatment system is primarily the effect of the chosen technology on other
process steps 65, 7o). An important limitation of many pretreatment technologies is
their ineffectiveness against softwood substrates 67,69,72) However, progress is
being made in this area 73)
Hydrolysis of substrates prior to biological conversion can be achieved by pre-
treatment followed by enzymatic hydrolysis, or by acid hydrolysis. The practicality
of acid hydrolysis is hindered by glucose yields < 60 ~ at low acid concentrations,
and high costs for acid recovery at high acid concentrations 72,74). Enzymatic hydroly-
sis achieves high substrate conversion yields, but cellulase production is very expensive.
Wright et al. 7m, calculate that the combined costs of cellulase production and
enzymatic hydrolysis are responsible for 46 ~ of total capital investment and contri-
bute 0.28 $ L -1 to the cost of producing ethanol from wood. Furthermore cellulase
production requires substrate, about 12--15 ~ of the total 51, 7o), which could other-
14 L.R. Lynd

wise be used to produce ethanol. Enzymatic hydrolysis has been reviewed by Wilke
et al. 64) Ladisch et al. 74), and Mardsen and Gray 7s). Acid hydrolysis is discussed
in reviews by Grethlein 76), Ladisch et al. 77), and Wright and Power 78)
Acid hydrolysis gives rise to sugar degradation products in amounts approximately
equal to undegraded sugars 79, 80). Furfural, a by-product of xylose degradation, could
be used as a chemical feedstock, and has received attention as a by-product to improve
the economics of acid hydrolysis-based ethanol production 5o, 811.The price of furfural
will depend on its volume of production 81). It has been estimated 50) that the price of
furfural at the level of production resulting from 15-20 large-scale acid hydrolysis-
based plants would be 22-33 cents kg- 1, or roughly half the price of ethanol. The
yield of furfural from xylose during acid hydrolysis of biomass, ~38 ~o on a mass
basis 50), is lower than the ethanol yields which could be obtained from xylose via
biological conversion. In addition there are added costs for capital, utilities, and raw
materials associated with furfural production. Considering these factors, it appears
that xylose conversion into ethanol is more attractive than conversion into furfural.
Enzymatic hydrolysis and biological conversion need not necessarily be separated,
and combining these steps offers the potential for relieving end-product inhibition by
products of enzymatic hydrolysis, and lowering capital costs. Simultaneous sacchari-
fication and biological conversion using thermophilic bacteria is the subject of Chap-
ters 4 and 5 in this report. This approach has also been studied using ethanol-producing
mesophiles in combination with added cellulase, as reviewed by Wilke et al. 64)
Both biological and engineering aspects of the biological conversion step of ethanol
production are the subjects of a large body of literature (see 58,64-,82,83,84,85) for
reviews). Ethanol production per se is relatively inexpensive. Thus, like pretreatment,
the choice of process system is based primarily on the impact on other process steps.
Considering the entire process, the most important consequences of the choice of
the biological conversion system are the range of substrates which can be utilized, the
ethanol yield, and the ethanol concentration produced, all primarily properties of
the organism and not bioreactor config-uration. Much research has been devoted to
increasing volumetric productivity. Approaches include techniques to keep cells in
the bioreactor, principally by immobilization 86) and recycle 84); and/or to keep
the ethanol concentration low by removal via solvent extraction, either with 87, 88) or
without 89, 9o) separation of solvent and aqueous phases by a membrane, or removing
an enriched vapor stream 64.Ss'91,9z). Batch process systems presently used for
ethanol production in the U.S. typically have volumetric ethanol productivities of
about 3 g L -1 h -1 37). Continuous stir-tank reactors offer somewhat higher produc-
tivities than batch systems s4), and high productivity systems employing immobilized
cells or ethanol removal can achieve productivities of 100 g L-1 h-1 37, 91) Though
it has been claimed that high productivity processes offer the potential for large
reductions in costs related to biological conversion for yeast-based systems 6r 8s),
very few continuous processes are in use commercially 37.84)
Ethanol production is normally carried out by one of several species of yeast. The
bacterium Zymomonas mobilis has also been considered (see 93,9r for reviews), as
have thermophilic bacteria (see Sects. 3 and 4). Both the yeasts normally employed
for ethanol production and Z. mobilis grow only on mono- and di-saccharides, and
both can produce ethanol at concentrations up to about 10 ~o s9). The merits of these
organisms for ethanol production have been compared sg,9s~,. Z. mobilis does not
Production of Ethanol from LignocellulosicMaterials Using Thermophilic Bacteria 15

use pentoses. Of the yeasts that do use pentoses, none do so with rates and ethanol
yields which are either comparable to those on hexoses or sufficient to allow a practical
process lO, 7o). Efficient conversion of pentoses to ethanol is an active field of research
(see 34.42, 96) for reviews).
Recovery of ethanol from spent medium is normally accomplished by distillation,
though alternative processes have been proposed, many with lower energy require-
ments then conventional distillation 97,98,99), A variety of techniques have been propos-
ed for dehydration 62,1oo); molecular sieve dehydration appears to be particularly
promising 57, lol). Current practice for energy-efficient distillation of ethanol is based
on vapor recompression heat pumps operating between the overhead vapor and the
reboiler, and also heat integration between columns operating at different pres-
sures 102,103). Busche lO3)reports an equivalent heat requirement (q + 3*w) of 4.24 MJ
L-1 for producing essentially pure ethanol from a 10 wt. ~ feed using state of the
art distillation techniques.
Extractive distillation offers the dual advantages of low reflux ratios, and therefore
low energy requirements, and also elimination of the azeotrope. Ethanol recovery by
extractive distillation has been studied by Barba et al. lO4) and also by Schmitt and
Vogelpohl lo5). Both studies concluded that this technique offered the potential of
significant energy savings. Solvent extractive agents are effective lo6), however their
use involves very large column temperature drops and so does not favor heat inte-
gration between columns. Lynd and Grethlein lo7) have designed an ethanol distil-
lation process specifically for separating ethanol from dilute broths. This process uses
intermediafe heat pumps with optimal sidestream return, IHOSR lOS,109), in con-
junction with extractive distillation. A flow sheet for a modified version of this process
is presented in Fig. 2. Capital costs for this process and the impact on the ethanol
selling price of substituting this process for conventional distillation are presented
in Sect. 3.2.3. In the modified version condensation of the overhead vapor from the
extractive column provides heat for the evaporator and the stripping column, which
are operated at a lower pressure. The mhin advantage of the modified version over the
original process design is that the stripping column is the low temperature column
instead of the high temperature column. This property decreases or eliminates heat
required for preheating the feed to the column temperature. In addition, the tempe-
ratures in the stripping column can be made low enough that cells may pass through
the column, which was not practical in the previous design.
The energy requirements (q + 3*w) for the IHOSR/extractive process are shown
in Fig. 3 as a function of feed ethanol concentration assuming a saturated liquid feed.
Also shown in Fig. 3 are energy requirements for conventional distillation. It may be
seen that the energy requirement of the IHOSR/extractive design remains relatively
flat down to concentrations as low as 1 wt. ~ ethanol, whereas that for conventional
distillation increases sharply with decreasing concentration.
General economic issues related to ethanol production have been reviewed pri-
marily for production from corn 37, s6,11o, 111), and for production from wood 70)
Some features of ethanol production economics are of importance to subsequent
sections in this paper. The cost of production of bioethanol is dominated by substrate
costs and capital-related costs. In the manufacturing cost summary presented by
Venkatasubramanian and Kiem 3~) for ethanol production from corn, the substrate
cost, including by-product credits, represents 46 ~o of manufacturing costs, capital-
 _yLL
~] STRIPPING
EVAPORATOR /TOWER
VACUUM-JE]-I
BOOSTER

REBOILER

~4

DRUM DRYER

:)LIDS FEEDER RECYCLETO

BIOREACTOR
OL MIX TANK

Fig. 2. Flow sheet for ethanol separation using IHOSR/extractive distillation (modified from lo7) t-
Production of Ethanol from LignocellulosicMaterials Using ThermophilicBacteria 17

T_.j 2O

(~ 15

E+
~
o v

-I- czr
(1)
or" 0
0 1 2 :3 4 5 6 7 8 9 10
Feed Ethenol Concentration(wt% )

Fig. 3. Energy requirements (q + 3*w) as a function of ethanol concentration for anhydrous ethanol
production using conventional distillation - - ~ - , and IHOSR/extractive distillation --41~--

related fixed expenses represent 33 ~, and variable expenses including utilities re-
present 21 Vo. A detailed economic summary of ethanol production from hardwood
using yeast and enzymatic hydrolysis and also an economic summary of yeast-based
ethanol production from corn are presented in Sect. 3.3. Waste treatment for ethanol
production processes was briefly considered in Sect. 2.1.3.

2.2.3 Properties of Ethanol as a Fuel

The debate over fuel ethanol has focused on issues of substrate supply, production
economics, and net energy balance. In contrast with the widely differing schools of
thought on these matters, it is generally agreed that alcohols make excellent fuels for
spark-ignited engines 31,112,113). A few considerations are relevant here.
The energy density of ethanol is 65 ~o that of gasoline (based on low heating values,
used throughout). However, several factors, described below mitigate this difference
both for use of ethanol in gasoline mixtures and as a neat fuel. Moreover, it is not
clear that energy density should be an overriding consideration when evaluating
alternative fuels. The energy yield, work done or distance traveled per unit energy,
would seem to be at least as important.
Ethanol has a higher octane rating, (R + M)/2 = 102, than does gasoline, and
causes a disproportionate increase in octane when blended with gasoline. The blending
(R + M)/2 octane rating of ethanol is 119. This may be compared to values for other
octane-enhancers: 120 for methanol, 97.5 for tert-butanol, and 108 for methyl tert-
butyl ether 113). Methanol boosts octane as effectively as ethanol, but blending
methanol with gasoline results in greater difficulty with phase separation, materials
incompatability, and vapor lock m,za3). Estimates for the volumetric milage for
ethanol blended 10 ~owith gasoline vary from 96 ~ to 107 ~ relative to gasoline 31,114)
An intermediate value considering the range of estimates is equivalent volumetric
milage 114).
Burning ethanol as a neat fuel requires carburetor modification and provision for
cold starts, as well as minor changes in engine materials. These are not difficult
18 L.R. Lynd

problems in a new system, but make retrofitting somewhat awkward 113). Automobiles
designed for burning hydrous ethanol have been marketed in Brazil by Fiat, Ford,
Volkswagen, G.M. and Chrysler 62). Engines designed for ethanol can have higher
compression ratios and so use leaner fuel mixtures, thus improving engine efficiency.
Octane rating increases with increasing amounts of water for neat ethanol fuel n2)
Estimates for the increase in volumetric ethanol consumption relative to gasoline
consumption, based on the same work done, are 10-20~o 112) and 15-25~o i161 for
hydrous ethanol, and 25 ~o for anhydrous ethanol 112). The 10-25 ~ greater volumetric
consumption and 35 ~o lower energy density of hydrous ethanol relative to gasoline
imply a 2 3 ~ 0 ~ energy yield for the ethanol fuel x. At an intermediate value
of 17.5 ~ greater volumetric consumption the increase in energy yield is 31 ~o.
The higher energy yield of ethanol compared to gasoline has an important impli-
cation for consideration of replacing gasoline by ethanol. That is, the ratio of gasoline
displaced per unit ethanol is greater than 1, and approximately 1.3. Thus to replace
the 1982 U.S. annual gasoline usage of about 1.2 x 1013 MJ would require about
9.2 x 1012 MJ of ethanol.

2.2.4 Energetic Considerations

The energy balance for ethanol production is of most relevance for the use of ethanol
as a fuel. A major point of contention in the debate over the efficacy of fuel ethanol
production from starch- and sugar-rich agricultural crops has been whether net
production of useful energy accompanies such processes. This subject is dealt with in
detail, and with differing conclusions, in papers by Sama nT), Yorifuji ns), John-
son 119), and Parisi 114), as well as earlier papers.
Studies considering cellulosic substrates are generally in agreement that the energy
balance tbr ethanol production from feedstocks is favorable, and more so than for
ethanol production from starch or sugar crops 62, n4-, 120). Several studies 62.70,121)
including this one (Sects. 3.2.3, 3.3) have concluded that essentially all of the at-plant
energy requirements, including electricity, for producing ethanol from lignocellulosic
substrates such as wood and agricultural wastes are available from burning process
residues, notably lignin. If ethanol production from biomass were to occur on a scale
sufficient to make a significant contribution to total transportation energy require-
ments, it is very likely that markets for lignin-derived aromatic chemicals would be
saturated and most residual lignin would be burned to provide process heat require-
ments and electricity. Depending on the availability of substitute fuels, lignin might well
be more valuable as a feedstock than as a process fuel for ethanol production on a
smaller scale 122. 123,124-)
The energy requirements for harvesting, chipping and transporting cellulosic
materials prior to processing must also be considered. Table 4 presents data for these
off-site operations for a self-planted, unfertilized fuel wood case, and for a high-
productivity, short-rotation, fertilized wood fuel crop. The energy expended on these

1 The calculation is as follows: (work/volume)etUa.oj(work/volume)g,s= 0.800 to 0.909; (volume/

energy)ea~,nol/(Volume/energy)gas= 1.54. The product of these two terms has units of (work/ener-
gy)e~ano](work/energy)ga~and is equal to 1.23 to 1.40.
Production of Ethanol from Lignocellulosic Materials Using Thennophilic Bacteria 19

Table 4. Energy requirements for growing, harvesting, chipping, and transporting fuel wood (MJ t 1)a

Item Self-planted, no fertilizer High-productivity,

(shipped as chips)b short-rotation fertilized
wood fuel cropc

Silviculture 640.7
Harvest 67.1 71.9
Prep. + chipping i66.9 140.0
Haulingd 182.2 182.2
Other 56.6 56.6
Total 472.8 1091.4 (919.3)
~ Ethanol energy, 8.6 19.8 (16.7)
only hexan used e
'~/oEthanol energy, 6.4 14.8 (12.5)
hexan + pentan used

a All values include the energy devoted to equipment manufacture and maintenance as well as listed
operations ;
b data from 61~;
~ data from 61); data in parenthesis from 6z); adding in the same transportation allowance used in ~1~;
d for a 50 mile average radius, the same value used in sly;
e the composition of hardwood in Table 2 is used, along with 90 ~o substrate utilization, and 47 ~o
ethanol yield

operations relative to the energy in the ethanol which could be produced from the
transported substrate depends on assumptions about the fraction o f substrate which
m a y be converted. If only hexan is used, 8.6 ~o of the ethanol energy is used in the
self-planted case, and 16.7 19.8~o are used in the high-productivity case. If both
hexan and pentan are used, these values become 6.4 ~o and 12.5 14.8 ~o.
Parisi 11~ has considered energy requirements for lignocellulosic substrates other
than wood. The only significant raw material-related energy requirements for ma-
terials not specifically grown for ethanol production (e.g. agricultural residues and
municipal solid waste) are for collection and transportation. In the case of M S W ,
these can be practically zero. Transportation energy requirements for agricultural
residues such as straw and corn stover are approximately 1.8 times those for wood
(see Table 4).
In contrast to starch- and sugar-rich agricultural crops, the situation with respect to
the "energy balance" for ethanol production from lignocellulosic substrates appears
to be quite clear cut: Processing energy requirements including electricity can be met
or nearly met using substrate-derived process residues with little or no supplemental
biomass and no supplemental liquid fuel. Pre-processing energy requirements re-
present a liquid fuel requirement on the order of 10 ~o o f the energy value of the ethanol.

2.2.5 End Use

Ethanol production may alternatively be viewed as either a process yielding a chemical

feedstock, a transportation fuel (for either blending or use neat), or a waste treatment
process with ethanol as a valuable by-product. The level of conceptual acceptance for
these uses of ethanol is p r o b a b l y in the order given.
20 L.R. Lynd

Production of ethanol from biomass for use as a chemical feedstock has been
favored over use as a fuel because of the differences in the amount of materials for
these two uses, and consideration of the available biomass supply (see Sect. 2.1.2).
Modification of the U.S. chemical industry to accomodate biologically-produced
materials instead of traditional petroleum-derived feedstocks has been considered
in many studies 26,27,28,56,122-126), with ethanol typically playing a key role.
In addition to the match of material demands and available biomass, the value of
ethanol used for chemical and fuel applications must be considered. According to
Greek 127), the alternative fuel value of any hydrocarbon is its lowest value as a
chemical feedstock. This general statement appears to apply in the case of ethanol,
as shown below. On a volumetric basis, and considering only energetic aspects,
ethanol used as a neat fuel should be worth approximately 0.85 x the price of gasoline
(based on 17.5 % greater volumetric consumption, see Sect. 2.2.3). The higher octane
rating of ethanol relative to gasoline can be taken advantage of in ethanol-gasoline
blends by either leaving gas quality unchanged and increasing engine compression
ratios and efficiency, or by decreasing gas quality to obtain a fuel suitable for use
in gasoline engines 31). Considering the latter option, the OTA 31) estimates that the
value of ethanol on a volumetric basis when blended with gasoline is 1.7 times the
crude oil price. Estimates for the refinery energy savings per gallon of ethanol blended
10 % in gasoline vary widely from 8 to 63 MJ 31). An intermediate value for refinery
savings considering the range of estimates is 0.4 L crude oil per L ethanol, or 0.25 L
gasoline equivalent per L ethanol 31). Using this value and equivalent volumetric
milage for a 10 % ethanol/gasoline blend (Sect. 3.2.3) the value of ethanol as a gasoline
extender is 1.25 times the value of gasoline. Ifa wholesale gasoline price of 0.21 $ L -j
is assumed, then the value of ethanol as a neat fuel is about 0. l 8 $ L-1, and the value
as an octane enhancer is 0.26 $ L -1. These values may be compared to the January
1987 price for synthetically-produced ethanol, 0.44 $ L -1, which reflects its value
as a solvent and chemical intermediate. This simplistic analysis makes the order in
which biomass-derived ethanol will achieve market penetration in the absence of
tax incentives seem clear: chemicals before octane enhancer before neat fuel.
To consider ethanol production as a waste treatment process, the substrate must be
one which is a waste disposal problem. While this is generally not the case for agri-
cultural or logging residues, it is for municipal solid waste (MSW). Economics for
alternative methods of disposal for solid wastes are highly site-specific. Recent trends
in the U.S. include rapidly increasing landfill costs, increased acceptance and utili-
zation of source separation and recycling, and consideration of electricity generation
via incineration. Obviously separation of the degradable fraction raises the cost of
MSW over the negative cost of tipping the unseparated waste. However, even at the
price presently paid for separated recyclable paper and cardboard, highly variable
but generally about 15-30 $ t-1, the carbohydrate fraction of MSW appears to be a
very competitive substrate for ethanol production compared to, for example, hard-
wood chips. Moreover, the price paid for separated paper and cardboard waste does
not reflect the avoided costs of disposing of these materials.
The potential yield of ethanol-from MSW is estimated at 318 L t-1 based on 60 %
degradable carbohydrate content (Table 2), 90 % carbohydrate utilization, and 47 %
bioconversion yield. This represents a 25.4 % yield by weight, and a 60.6 % energy
yield based on a total organic fraction (degradable and non-degradable) of 80 % with
Production of Ethanol from Lignocellulosic Materials Using Thermophilic Bacteria 21

a m e a n heating value of 13,900 kJ kg 1 organics. Electricity g e n e r a t i o n w o u l d be

likely to have a n overall efficiency of __<30 % based o n the total organic fraction given
the m o i s t u r e c o n t e n t o f M S W . The values o f e t h a n o l a n d electricity are similar on a
per u n i t energy basis at a p p r o x i m a t e l y $16 per 106 kJ. T h u s f r o m energetic a n d p r o d u c t
value considerations, e t h a n o l p r o d u c t i o n appears to be a n attractive alternative to
i n c i n e r a t i o n a n d electricity generation.

3 Potential of Thermophilic Bacteria for Ethanol Production

3.1 Identification of Distinguishing Features

Table 5 presents several features of thermophilic bacteria which have been cited as
either advantages or disadvantages in the literature. Reference is made to both general
features and features discussed primarily with respect to ethanol production. When
considered in the context of ethanol production from lignocellulose, some of these
factors are of more importance than others, and some factors do not apply. It is
clear that thermophilic ethanol production must compete with yeast-based processes
in order to find significant utilization, thus comparison is made primarily with ethanol
production using yeast in the following discussion.
By far the most important of the advantages of thermophilic bacteria listed in
Table 5 for ethanol production from lignocellulose is advantage 1, the wide range of
substrates utilized. This general characteristic can be divided into two specific features
possessed by thermophiles collectively, but not by any single described species:

Table 5. Advantages and disadvantages of thermophilic bacteria as biocatalysts

Refs.
Advantages
1 Wide range of substrates metabolized (particularly pentoses 4.5.8,9.59.1281

and insoluble carbohydrates)

2 Benificial physical properties of growth medium at high temperatures 4,5,8,9,128)

(reduced viscosity and surface tension, increased diffusion rates and

substrate solubility)
3 High reaction rates 4. 5 , 8 . 9 . 128)
4 Economical bioreactor cooling 4, 5 . 8 , 9 . 1 2 8 }

5 Low risk of contamination and growth of pathogens 4 . 5 , 8 . 9 . 5 9 . 128~

6 Production of thermostable enzymes 4, 5.8, 9, 128)

7 Low cell yields and high product yields per unit substrate 8.9, 59, 128)
8 Low oxygen solubility 8, 9)

9 Increased value of metabolic heat 9)

10 Facilitated product recovery (primarily due to high vapor 4. 5 . 8 . 9 , 59, 1281

pressure of volatile compounds)
Disadvantages
1 Low ethanol tolerance 8. 591

2 Production of organic acids in addition to ethanol 8.59)

3 Lower substrate tolerance S)

4 Greater stress on biotechnological hardware 8.129)

5 Sensitivity to inhibitors 130)

6 Complex growth factors required in relatively high amounts 131)

7 Limited fundamental knowledge of physiology and biochemistry 8.9.59,128)

22 L.R. Lynd

pentose utilization and cellulase production. Pentose sugars represent approximately

one-fourth of the total fermentable carbohydrate in lignocellulosic substrates (Sect.
2.1.2). In light of the dominant place of substrate costs in overall process economics
(Sects. 2.2.2), and that many substrate-related costs are the same with and without
pentose utilization, e.g. growth, harvest, transportation, and pretreatment, the
importance of pentose utilization is clear. The high cost of cellulase production and
enzymatic hydrolysis for ethanol production from wood using yeast has already been
alluded to (Sect. 2.2.2), and will be demonstrated further in Sect. 3.2.3.
Beneficial physical properties of growth medium at high temperatures, can be
expected to allow reduced energy requirements for mixing, and perhaps other small
advantages, but are unlikely to have a significant effect on overall production costs.
Advantage 3, high reaction rates, may not apply to a significant extent in the case
of ethanol production. Mesophilic bacteria hydrolyze cellulose at approximately
the same rate as the thermophile Clostridium thermocellum 132,133), Specific growth
rates reported for thermophiles such as Clostridium thermohydrosulfuricum, C. ther-
mocellum, C. thermosaccharolyticum, and Thermoanaerobacter ethanolicus on soluble
substrates are generally between 0.4 and 0.6 h -1 8,128), which may be compared to
0.43 h-1 to 0.46 h -1 for the yeast Saccharomyces cerevisiae growing on glucose 85,134)
The specific rates of ethanol production by thermophiles are not substantially higher
than those for either yeast of the bacterium Z. mobilis 6, 8). In a more general con-
sideration of the growth rates of various thermo-classes of bacteria, Sonnleitner and
Fiechter 6) concluded that maximum growth rates for thermophiles can in general be
expected to be higher than for a comparable mesophilic organism. The Arrhenius
activation energy relating the rates of comparable organisms from different thermo-
classes is however smaller than for a single organism. It is not clear whether the
difference between this point of view and that suggested by the available data on
thermophilic and mesophilic ethanol production is due to a peculiarity of ethanol
production and/or cellulose hydrolysis, or to incomplete understanding of thermo-
philic ethanol producing bacteria.
Bioreactor cooling is not a significant cost in the production of ethanol using
yeast 39, 51), and thus advantage 4 is not of great importance.
Low risk of contamination and growth of pathogens, has also been cited as a
feature of yeast-based processes s, 59), because of low pH and high ethanol concentra-
tion. The susceptability of thermophilic and yeast-based ethanol production to contam-
ination has not been compared to the author's knowledge.
Thermostable enzymes could probably be a valuable by-product of thermophilic
ethanol production, especially on a small scale. For example, thermostable starch-
hydrolyzing enzymes could find extensive use in the manufacture of corn-derived
sweeteners (reviewed in 10)). In the case of very large-scale ethanol production, e.g. to
meet fuel demands, the demand for thermostable enzymes is very likely to become
insignificant.
Low cell yields and high product yields based on higher cell maintenance require-
ments in thermophiles have been proposed, but not proven for thermophiles general-
ly 5). In a detailed study of C. thermohydrosulfuricum, Lacis and Lawford 135)concluded
that there was no significant difference between the energy required for cellular
maintenance in this thermophile and values for this parameter generally reported for
mesophiles. To the author's knowledge, there is no correspondingly strong evidence
Production of Ethanol from LignocellulosicMaterials Using ThermophilicBacteria 23

that favors the interpretation that thermophiles have higher maintenance requirements.
In any case, the cell yield is clearly of secondary importance compared to the synthesis
of organic acids, discussed below, in affecting the ethanol yields of thermophilic
processes.
At the scale and cell concentrations required for a practical process it is very un-
likely that the lower oxygen solubility at elevated temperatures, advantage 8, would
be of any consequence.
Calorimetric studies of cellulose utilization by anaerobic bacteria have reported the
production of metabolic heat at a level which corresponds to about 1.27 MJ L-~ ~36),
assuming ethanol is the only soluble product formed. This quantity of heat, assuming
that it all is recovered, represents approximately 1/4 of distillation energy require-
ments and 1/6 of total energy requirements for production of ethanol from corn
using yeast; total energy requirements for ethanol production from corn using
yeast account for about 13~o of production costs 3v). Thus the potential economic
impact of metabolic heat is likely to be on the order of a few percent of production
costs. Use of metabolic heat is restricted to process heat requirements at temperatures
less than the temperature of bioconversion. These include preheat of various process
streams, and also product recovery in cases where this step is carried out at suitably
low temperatures, e.g. reduced pressure distillation. In some situations, the contri-
bution of metabolic heat may be significant, and should not be discounted. At the
least, metabolic heat can contribute to maintaining the bioreactor at thermophilic
temperatures, for which a counter-current heat exchanger heating the bioreactor feed
is also helpful.
Facilitated product recovery is often cited as an advantage of thermophilic bacteria
for ethanol production (see Table 5). This factor will be an advantage of thermophilic
systems compared to mesophilic systems at the same ethanol concentration, and a
distinct advantage for thermophilic systems if continuous ethanol removal is employ-
ed. However, it is the author's evaluation that facilitated product recovery has been
over-rated as an advantage of thermophilic systems, particularly when comparing
thermophiles to yeast. Contrary to some reports 137),there is no particular significance
to the fact that some thermophilic ethanol producers will grow at temperatures near
the boiling point of ethanol, since aqueous solutions of a few % ethanol boil at
> 95 ~ It is true that operation at a higher temperature may decrease the temperature
difference between the fermentor and distillation system, which is normally operated
at > atmospheric pressure. The result of this may be only to change the size of the
preheat heat exchanger since many process streams, such as the stripping section
bottoms, are often available to provide preheat. At most the preheat energy require-
ment will be saved. The energy required for preheat can be significant at the low ethanol
concentrations processes using thermophiles may have to operate at, but this energy
requirement is a small fraction of distillation energy requirements for processes using
yeast. It is also true that the pressure for boiling spent medium at the conversion
temperature is higher for thermophilic systems than mesophilic systems, but the
pressures are still tow in most cases. For example the pressure must be ~0.2 atm to
boil an ethanol-water'mixture at a few % ethanol at 60 ~ This may be compared to
~0.09 atm for operation ~t 37 ~ Because of the higher pressures which may be
used, continuous ethanol refn~val at thermophilic conversion temperatures should
involve lower capital costs, which is fortunate because ethanol removal may be
24 L.R. Lynd

necessary to obtain reasonable productivites. However energy is not saved by virture

of the difference in pressures for continuous ethanol removal. The latent heats are
slightly larger at low pressure and thermophilic temperatures, and the relative volatil-
ity of ethanol is slightly lower than at normal distillation temperatures both negative
factors for distillation energy requirements. The "gas stripping ethanol rectifier"
discussed by Slapack et al. 8~, and also other similar gas stripping processes for con-
tinuously removing ethanol during thermophilic ethanol production 6,7~, reduce
subsequent distillation energy requirements, as claimed, but will generally have an un-
acceptably large energy requirement for vapor evaporation in the gas stripping step 1.
An important disadvantage of thermophilic bacteria relative to yeast is low ethanol
tolerance. Low ethanol tolerance is undesirable because it increases the cost of product
recovery. In addition, low ethanol tolerance limits allowable substrate concentrations,
thus decreasing productivity and increasing bioreactor costs. As pointed out by
Maiorella et al. 85~, low feed substrate concentrations also have economic penalties
other than bioreactor costs, such as costs related to sterilization, evaporation, drying,
and waste treatment.
In general the available data appears to be consistent with the assumption of similar
ethanol 15roduction rates as a function of substrate and cell concentrations for both
yeast and thermophiles using soluble substrates. It may also be reasonable to assume
that a similar function describes the degree of ethanol inhibition in relation to the
maximum ethanol concentration for both systems. For kinetically identical systems,
volumetric bioreactor productivity will be directly related to the maximum ethanol
concentration tolerated. Yeast and thermophilic bacteria appear to differ by a factor
of approximately 3 with respect to this key parameter (see Sect. 4.3). Thus volumetric
productivity for thermophilic and yeast-based ethanol production systems in the
same kind of bioreactor may be expected to differ by about 3-fold. Volumetric pro-
ductivity for in situ thermophilic lignocellulose utilization is likely to be less than for
thermophilic utilization of soluble substrates by a further factor of 3 due to lower
growth rates.
With the exception of ethanol recovery costs, the other limitations and costs
associated with low ethanol tolerance can in principle be aleviated by continuously
removing ethanol from the bioreactor, and so uncoupling substrate and product
concentrations. The incentive to utilize continuous ethanol removal with thermophilic
ethanol production systems appears to be strong, especially since a favorable impact
on ethanol yields may also be achieved (see Sect. 4.4). Economically- and physio-
logically-acceptable methods of continuous ethanol removal have not been de-
monstrated for thermophilic systems even on a laboratory scale (see Sect. 5).
The fraction of substrate converted to ethanol, which is decreased by organic acid
production, is a key economic factor because of the dominance of substrate costs.

1 For example: Ethanol and water have essentiallyequal latent heats on a molar basis at about 40.6 kJ
g mole-1. If an inert gas is bubbled through a 2 wt. ,% ethanol aqueous solution (liquid ethanol
mole fraction, Xethanol, = 0.00792) to continuously remove ethanol, the liquid which may be
obtained by condensing the inert gas/ethanol/water mixture will have mole fraction Xeth~o~= 0.086,
with the remainder water. To recover 1 mole of ethanol thus requires evaporation of 11.6 moles of
liquid, corresponding to a heat requirement of 472 kJ g mol- 1 ethanol. This heat requirement re-
presents about 38% of the combustion energy of the ethanol (1231 kJ g mole 1), and subsequent
distillation to complete ethanol separation will require further heat.
Production of Ethanol from LignocellulosicMaterialsUsing ThermophilicBacteria 25

Engineering aspects of ethanol tolerance and yield will be discussed further in Sect.
3.2.3, biological aspects will be discussed in Sects. 4.3 and 4.4.
Low substrate tolerance would not be expected to be a factor in continuous culture,
though it does limit the utility of the batch processes. In addition to substrate tolerance,
other factors favoring continuous systems over batch systems for thermophilic ethanol
production are the generally higher productivity of continuous systems 37, 8~), and
the relative ease of ethanol removal.
Sensitive to inhibitors, particularly arising from pretreatment, and also a possible
requirement for more elaborate and costly growth medium, could be significant
limitations for thermophilic ethanol production if they were shown to be widespread
fundamentally-based characteristics. However, optimization of medium formulation
and substrate pretreatment to minimize by-product inhibition are in very preliminary
stages with respect to thermophilic systems. Moreover, similar problems have been
solved for other systems. Thus it is a distinct possibility that disadvantages 5 and 6,
along with others, may be mitigated by addressing disadvantage 7, limited fundamental
knowledge of the physiology and biochemistry of thermophilic ethanol-producing
bacteria.
In summary, it is the author's evaluation that in comparison to ethanol production
using yeast, the two most important advantages of thermophilic bacteria for ethanol
production from lignocellulosic materials are pentose utilization and cellulase pro-
duction. The two most important disadvantages are low ethanol tolerance and organic
acid production.
An interesting observation is that neither pentose utilization nor cellulase production
are associated with growth at high temperatures per se. Thus obtaining non-thermo-
philic organisms with these properties is an endeavor with some justification. Several
studies have demonstrated some degree of expression of genes coding for one or
more cellulase components from mesophilic organisms in a convenient host 13s,139,140),
and preliminary studies have also been carried out on cellulase expression in yeasts 141)
and on gene transfer systems for Zymomonas mobilis 142). However, obtaining co-
ordinated expression of foreign genes coding for synthesis and secretion of a multi-
enzyme complex, together with transport and catabolism of resulting soluble sub-
states is an ambitious undertaking. By contrast the achievements of genetic enginer-
ing to date typically involve expression, often without secretion, of a single gene 143)
A second observation about the advantages of thermophilic bacteria for ethanol
production is that neither pentose utilization nor cellulase production is a significant
factor for starch and/or hexose-rich substrates 'such as corn and juice from sugar cane.
In contrast to cellulase, amylase is relatively inexpensive to produce 39).

3.2 Evaluation of Distinguishing Features

Having identified the important distinguishing features of thermophilic bacteria for

ethanol production from lignocellulose, it is desirable to evaluate these features. This
task is made more difficult by the fact that there have to date been no experimental
studies of thermophilic bacteria with an economically acceptable combination of
substrate and medium composition, bioreactor productivity, ethanol concentration,
26 L.R. Lynd

and ethanol yield. Progress toward realization of the goal of economic ethanol pro-
duction using thermophilic bacteria is considered in Sect. 4.
The approach taken here is to consider a detailed process design and economic
analysis for ethanol production from lignocellulose via enzymatic hydrolysis and
yeast, and then to present the economic impact of process changes associated with the
distinguishing features of thermophiles. The purpose of this exercise is to determine
if and to what extent these features impact strategic cost factors in the context of
overall process economics. In so doing the relative importance of the distinguishing
features of thermophiles and the potential economics of thermophile-based and other
ethanol production processes can be compared.
A detailed plant design and economic evaluation of an enzymatic hydrolysis-based
plant producing 94.6 x 106 L a- 1 of ethanol via yeast is used as a base case. This design,
by Chem Systems Inc., Tarrytown NY 51), utilizes dilute acid hydrolysis for substrate
pretreatment, a process which has been shown to be effective in allowing high extents
of hydrolysis for both in vivo and in vitro thermophilic systems (Sect. 4.1). By-product
credits are taken for both furfural, produced from xylose and valued here at 33 cents
kg-1, and CO 2, valued at 4.4 cents kg-1. High pressure steam is expanded in a turbo-
generator to produce electricity, and the exhaust steam is used for process heat re-
quirements. Process residues and unprocessed wood amounting to 11 ~ of the total
wood used provide all process heat and 95 ~o of required electricity. Raw materials
required in addition to the primary substrate include medium chemicals, H2SO4,
lime, and cellobiase.

3.2.1 General Impact

The primary effects of the positive distinguishing features of thermophilic bacteria,

cellulase production and pentose utilization, are process simplification and an in-
crease in the fraction of substrate which could be converted to ethanol. A schematic
flow sheet of the Chem Systems design for ethanol production from wood is shown
in Fig. 4a. Important process steps include pretreatment, separation and neutrali-
zation, enzyme production, enzyme hydrolysis, sugar concentration, biological con-
version to ethanol, purification, furfural production, heat generation, waste treatment
and CO z recovery. Fig. 4b presents a flow sheet for a hypothetical thermophilic
process assuming that cellulase production and substrate hydrolysis occur in the
fermentor, and 5-carbon as well as 6-carbon sugars are converted to ethanol. It is
evident that there are fewer process steps and fewer divided flows, the latter often
representing costly solid-liquid separations.
The amount of degradable carbohydrate available from a given amount of wood
leaving the pretreatment section will be greater for the thermophilic design (Fig. 4b)
than for the mesophilic design (Fig. 4a). This difference is due to 2 factors: 1) that
no solids and accompanying soluble sugars are diverted to a separate cellulase pro-
duction step, and 2) that pentoses are used in the thermophilic case. The substrate
flow in the base-case design is presented in Fig. 5. All values represent fractions of
material per unit of original substrate using the substrate composition assumed for
the base case. Carbohydrate fractions are calculated on a monomer basis, that is
including the weight added by the water of hydrolysis, throughout. The yield impli-
cations of various process modifications are presented in Table 6. It may be seen that
Production of Ethanol from Lignocellulosic Materials Using Thermophilic Bacteria 27
WATER AIR
STiAM T AC{D WAiER CoiOH)2 NUTRIENTS
FEED. PPFTI::?FATI~AFiXITI.~I
SEPARATION A N LD W ENZYME:
A T ~
| ................. I I_NEUTRALZATION-~ PRODUCTION
[

R~RY CONcSSGTARRT
I0 N
]~ CELL (]OH

ETHANOL
PRODUCT
BIOLOGICAL
;OLIDS TO CONVERSION
GYPSTACK

I WOODr,----J jACID
~ I FURFURAL
WASTE ~_ HEAT -I ~
GENERATION] FURFURAL BY-PROD.
"TREATMENT j" PRODUCT ON ~"

a, PROCESSWATER STEAM STEAM Ca(OH)2

STEAM WATERACID WATER Ca(OH)2

-- PRETREATMENT iL~NEUTRALIZATION

J#C02
~ BY-PRODUCT
I
CO2
RECOVERY

[
I ETHANOL
- PRODUCT
BIOLO GICA~L--I~PURI FICATION
GYPSTAcKSOLIDS
TO t.ui~v~_r~bUl~ I

f
f
I WASTE h UTILITIES
"qTREATM ENTI-

b PROCESSWATER STEAM
Fig. 4. Comparison of flow sheets for ethanol production from wood using yeast and enzymatic
hydrolysis, Fig. 4a, and thermophilic bacteria, Fig. 4b. Fig. 4a is essentially as presented in st~
28 L.R. Lynd

Wood
'0.6122 total hexan I
0.5244 cellulose
0.1875 pentan
~. retreatment
10.0085 furfural/C()z]
0.1140 hexose1
0.4982 eellulose
0.1261 pentose
0.05:31 pentan

Solids /
~ ~
Separation
ds

Liquid
().4982 cellulose 0.0759 hexose
0.0.380 hexose 0,0840 pentose
0,0531 pentan
0,0420 pentose
Enzyme Enzymatic
x"~kHydr~
Production
0"0598 cellulosel [0.0438 cellulosel
0.0046 hexose [0.5059 hexose ]
0.0064 pentan I [0.1677 pentose I
0.0050 pentose I / ~

Fig. 5. Substrate flow in the Chem Systemsst) wood to ethanol process design employing yeast and
enzymatic hydrolysis. All values are on a monomer basis (adjusted throughout for the water of hy-
drolysis) per unit substrate as originally received

eliminating the cellulase production step increases the degradable carbohydrate

per unit substrate by 11.6 %, utilization of pentoses (while still making cellulase)
results in a 33.3 % increase, and both eliminating cellulase production and utilizing
pentoses increases the degradable carbohydrate yield by 47.1%.

3.2.2 Basis for Economic Analysis

In modifying the base-case process to reflect the distinguishing characteristics of

thermophiles, the size of the equipment for many process steps are changed. A power
law is used to adjust the cost of this equipment in relation to changes in size. Power
law exponents are derived from costs given by Badger so) for ethanol plants with
different production capacities (the Chem Systems study did not give alternative
capacities). Raw materials, by-product credits, and utility-related operating costs are
assumed to be proportional to the rate of utilization or production, as in the Chem
Systems study.
Several process steps perform the same functions in the base case and in the thermo-
philic case. These include wood handling, pretreatment, heat generation, waste
treatment, and CO 2 recovery. These steps are only changed due to changes in capacity
Production of Ethanol from Lignocellulosic Materials Using Thermophilic Bacteria 29

Table 6. Fate of carbohydrate species in the chem systems plant design for ethanol production from
wood using yeast and enzymatic hydrolysis

Hexose:
0.6122 0.5039 to bioreactor (Hf)
0.0598 cellulose to cellulase production (He, c)
0.0046 hexose to cellulase production (He,h)
0.0438 to unreacted cellulose (H u)
0.6121
Pentose:
0.1875 0.1677 to bioreactor (Pf)
0.0064 pentan to cellulase production (Po,,)
0.0050 pentose to cellulase production (Po, s)
0.0083 to furfural/CO 2 (Pd)
0.1874
Yield implications of process modifications
I. Eliminate cellulase production:
He + He,0 x0.9 + Hr
Yield multiplyer (Y.M) = = 1.116
Hf
II. Use pentoses:
Hf q- Pf
Y.M.- - 1.333
Hf
III. I and II:
Y.M. = Hf + Hc.ox0.9 + He. h + Pf 4- Po,,, + P~,s 1.471
Hf

All values are on a monomer basis (adjusted throughout for the water of hydrolysis) per unit sub-
strate as originally received
Values from Fig. 5 as given by 5~)

in the thermophilic case. Changed cost due to capacity only is also assumed for the bio-
logical conversion section. A bioreactor designed for thermophilic ethanol production
would probably be somewhat different from the types used in the base case because
solids are separated prior to bioconversion and are used directly when thermophiles
are used, and also because of other factors. However, the base-case bioreactor system
has a residence time of 18 h, which is similar to but somewhat greater than the 12 to
16 h residence times required to achieve > 85 % utilization of pretreated mixed hard-
wood by C. thermocellum in continuous culture, as discussed in Sect. 4.1. Given these
considerations, the lack of data upon which to base a design for the thermophilic
system, and the generally non-critical nature of bioreactor costs, the cost of the
biological conversion section is assumed to be the same, with adjustments for changes
in capacity, for the thermophilic case and for the base case. The bioreactor is operatod
at atmospheric pressure in both cases.
In this study a bioreactor ethanol concentration of 1.5 % is assumed. This value has
consistently been produced by both batch and continuous thermophilic cultures (see
Sect. 4.3 for further discussion of ethanol tolerance). Liquor from the bioreactor is
allowed to flow to the distillation system, stripped, and pumped back to the fermentor,
all at 60 ~ Low ethanol concentrations are thus maintained in the bioreactor though
the substrate concentration leaving the pretreatment reactor is 16.5 wt. % solids in
both the base and thermophilic cases.
30 L.R. Lynd

Conventional ethanol distillation is very unlikely to be satisfactory for separating

ethanol at 1.5 wt. ~ because of high steam requirements. Therefore the IHOSR/
extractive distillation process is used (see Sect. 2.2.2). The size and cost of equipment
for this process is based wherever possible on the design parameters and costs of
similar equipment in the Chem Sytems design. Standard cost correlations and costing
procedures are used 14< 14s)
The Chem Systems study assumes 90 ~ enzymatic hydrolysis yield using cellulase
from Trichoderma reesei acting on a wood feed consisting of 57 ~o aspen, 20 ~o maple,
and 23 ~o other woods after pretreatment at 200 ~ for 12 s with 0.75 ~o H2SO4. Data
are presented in Sect. 4.1 which show that in vitro and in vivo yields from hydrolysis
using Clostridium therrnocellurn cellulase are consistently comparable to hydrolysis
yields using T. reesei cellulase. Thus the same hydrolysis yield is used in the thermo-
philic case as in the base case.
A major assumption is that high substrate concentrations can be converted by the
thermophilic system with the same utilization of degradable carbohydrate observed
at low substrate concentrations. While this would be expected given sufficient knowl-
edge of nutritional requirements, it has not been demonstrated. Also, no change is made
in the cost of chemicals for medium supplements (nitrogen source etc.) compared to
the base case. Finally, it is assumed that there is no inhibition of growth or hydrolysis
by pretreatment by-products. The issues of utilization of high substrate concentra-
tions, nutrient supplement requirements, and by-product inhibition are some of the
perhaps less glamorous research areas which must be addressed in order to develop
thermophilic ethanol production. Making optimistic assumptions about these issues
is based on the fact that similar issues have been solved for other systems. It is also con-
sistent with the goal of this analysis, that is, to investigate the economic implications
of distinguishing features of thermophiles as biocatalysts for ethanol production.
Given a process with higher yields per unit substrate, one could either keep the
ethanol production the same and reduce the substrate flow, or keep the substrate flow
the same and increase the output. Since the major issues in determining plant capa-
city are related to the substrate (e.g. the available substrate supply and the distance it
must be transported) and not the product, it is elected to keep the flow of substrate
to the pretrea'tment reactor the same as in the base case, and to let the ethanol pro-
duction increase. This actually results in a ~ 11 ~ smaller total wood usage since the
base case design uses some wood fed directly to the boiler whereas the thermophilic
design does not.

3.2.3 Economic Impact

A detailed consideration of the economic impact of the distinguishing features of

thermophiles for ethanol production from lignocellulose has recently been completed
by the author 146). A summary of the results is presented here.
Table 7 presents capital costs for production of ethanol from wood using yeast and
enzymatic hydrolysis, the base case, as reported by Chem Systems sl). Also presented
in Table 7 is the thermophilic case representing the cumulative economic impact of
three of the four distinguishing features of thermophilic bacteria: pentose utilization,
cellulase production, and low ethanol tolerance. Production of organic acids is not
Production of Ethanol from Lignocellulosic Materials Using Thermophilic Bacteria 31

Table 7. Capital costs for production of ethanol from wood using yeast and enzymatic hydrolysis,
and the capital cost-related impact of cellulase production, pentose utilization, and operation at
1.5 wt. ~ ethanola

Item U.S.$ x 10 6

Base case Thermophilic case

1 Wood handling 8.10 7.74

2 Pretreatment 3.10 3.10
3 Sugar separation/neutralization 3.90 0.72
4 Cellulase production 6.38 0
5 Enzymatic hydrolysis 7.07 2.94
6 Sugar concentration 5.94 0
7 Biological conversion 4.21 4.87
8 Ethanol recovery 4.11 11.28
9 Utilities 28.86 25.29
10 By-product processing 10.97 6.73
11 Storage 1.93 1.68
12 Pollution control 2.86 2.58
Total direct 87.43 66.93
Indirect 24.58 15.00
Total fixed investment 112.0 81.93

" Base case in mid 1984 $ as presented by Chem System 51~,for a plant producing 94.6 x 10 6 L a -1
ethanol. The thermophilic case is in mid 1984 $ for a plant processing the same amount of wood
through the pretreatment section and producing 139 x 106 L a -1 ethanol. See text and 146) for
further details

considered in Table 7. The reasons for the differences in the capital costs for items in
the base and thermophilic cases are discussed below.
In situ cellulase production eliminates or greatly reduces costs for items 3, 4, 5,
and 6, all o f which are devoted to either production or efficient utilization of cellulase
in the base case. Costs for certain functions accomplished in these items are retained,
in particular for solids concentration. Solids concentration is accomplished in step 5
in the base case, and is accomplished following biological conversion and prior to
burning process residues in the thermophilic case. Pentose utilization eliminates
furfural production and processing, a part o f item 10, and associated storage and
waste treatment in items 11 and 12. Pentose utilization in combination with cellulase
p r o d u c t i o n increases the ethanol output by 47 ~ from 94.6 x 106 L a -1 to 143 x 106 L
a -x, as discussed in Sect. 3.1, with consequent increases in costs for biological con-
version and CO 2 recovery. The cost o f distillation, item 8, is increased by both the
increased ethanol output and also due to the use of IHOSR/extractive distillation to
recover ethanol at a concentration of 1.5 wt. ~o. Both the effect o f cellulase production
in eliminating portions of items 3 through 6, and also the substitution of the more
efficient IHOSR/extractive distillation for the conventional distillation used in the
base case contribute to reducing utility-related capital costs, item 9, in the thermophilic
case. O f the costs for distillation in the modified thermophilic case, 73 ~ are for two
kinds o f items: eight 1Oft diameter disc and doughnut-type reduced pressure stripping
columns, and a 3400 horsepower compressor which drives the heat p u m p cycle.
Compressor technology is relatively well-developed, and costs and performance m a y
32 L.R. Lynd

be predicted with relative certainty. There is more uncertainty about the cost and
performance of the stripping columns, principally because the influence of cells and
substrate residue on stripping efficiency and the VLE relationship have not been
adequately studied.
Table 8 presents production costs for the base case and the thermophilic case.
W o o d costs are different because excess wood is required to meet process steam re-
quirements in the base case, but not in the thermophilic case. Acid and lime require-
ments are lower in the thermophilic case because furfural production is eliminated.
Utility-related operating costs are lower for the same reasons capital costs for utilities
are lower. 95 ~ of electricity requirements are generated on-site in the base case, and
93 ~o in the thermophilic case. By-product credits for CO2 are increased in the thermo-
philic case due to the increased carbohydrate metabolized, but are eliminated for
furfural. Labor costs are lower in the thermophilic base case because they are more
closely related to the n u m b e r of process steps than to production capacity. Mainten-
ance and taxes are related to capital costs, overhead is related to labor and capital costs.
A capital recovery factor based on a 15 ~ before taxes return on investment over
10 years for the case of 100 ~ equity financing is used in Table 8 to convert capital

Table 8. Production costs for producing ethanol from wood using yeast and enzymatic hydrolysis,
and the production cost-related impact of cellulase production, pentose utilization, and operation
at 1.5 wt. ~o ethanoP

Item U.S.$ x 10 6

Base c a s e Thermophilic case

Raw materials
Wood 15.06 13.39
Acid 1.56 0.86
Lime 0.67 0.36
Cellobiase 1.35 0
Medium components 1.61 2.37
By-product credits
Furfural (33 cts kg-i) --5.49 0
C Q (6.2 cts kg -i) --4.56 --6.71
Operating costs
Labor 1.69 1.26
Maintenance 3.98 2.58
Purchased power 0.34 0.40
Labor, maintenance and 2 overhead for utilities
Generated power 1.29 1.10
Boiler feed water 0.22 0.15
Cooling water 1.07 0.75
Steam 2.05 1.50
Overhead and taxes 6.12 4.29
Total production costs 26.96 22.30
Capital recovery (15 ~ R.O.I, 10 years) 22.32 16.33
Ethanol production (L a- 1) 94.6 139
Selling price (U.S.$ L -1) 52.1 27.8

" All costs in mid 1984 dollars. Base case and thermophilic case as for Table 6;
b labor, overhead, and maintenance are charged separately for utilities, which are considered off-site,
and process equipment, which is considered on-site
Production of Ethanol from LignocellulosicMaterials Using ThermophilicBacteria 33

costs into an annual operating cost. Using this procedure, the selling prices for ethanol,
including return on capital investment, are 0.52 $ L - 1 for the base case, and 0.28 $ L - 1
for the thermophilic case. All costs are in mid 1984 U.S. dollars.
Relative to the base case selling price of 0.52 $ L -1, the impact of cellulase pro-
duction considered individually is to reduce the selling cost by 19 cts L-1 ethanol, or
37 ~o. The impact of pentose utilization relative to the base case selling price is to
reduce the selling price by 12 cts L -1 ethanol, or 23 ~. The influence of these factors
on the final selling price is not additive. The contribution of distillation to the ethanol
selling price, considering capital and operating cost factors including utilities is 2.9 cts
L - ~for the base case, 5.1 cts L - 1 for the thermophilic case at the same 94.6 x 106 L a- 1
capacity, and 4.3 cts L -1 for the thermophilic case at the increased capacity as con-
sidered above. Thus the added cost of operating at 1.5 ~ ethanol with continuous
ethanol removal is relatively small.
Reductions in the ethanol selling price due to cellulase production and pentose
utillzation presented above are supported by economic data presented by Wright
et al. 70), for a plant producing ethanol from hardwood using yeast and enzymatic
hydrolysis. The design analyzed by these workers differs from the Chem Systems
design considered thus far in that furfural is not produced and steam explosion is
used as a pretreatment. According to Wright et al. 70), cellulase production and en-
zymatic hydrolysis together are responsible for approximately 40 ~ of total ethanol
production costs, and conversion of pentoses to ethanol would reduce the selling
price by 34 ~. This value for the impact of pentose utilization is higher than that
presented above, presumably reflecting the benefit of furfural production in the Chem
Systems design.
Consideration of the impact of lower than theoretical ethanol yield involves an
assessment of the value or disposal cost of by-products. Both acetic and lactic acids,
the main products other than ethanol produced by many thermophilic bacteria, have
values at least that of ethanol, and can be produced in higher yield from sugars because
of their higher molecular weights. However, recovering organic acids by distillation
from dilute solution is considerably more difficult then recovering ethanol because
the boiling point of acetic acid is higher than water, so the water must be boiled away
from the acetic acid instead of vice-versa 103). Separation technologies other than
distillation, such as solvent extraction are not established and are also costly lo3).
Dupont investigated biological acetic acid production in the early 1980's and conclud-
ed it was uneconomical, largely because of separation issues [Dr. Thomas Ng, Dupont,
personal communication]. Similarly, processes based on non-biological production
of fuels from organic acids 147) have also been abandoned.
Both acetic and lactic acid make excellent substrates for methane production,
however methane is less valuable than ethanol as both a chemical and a fuel, and is
produced in 50 ~ lower mass yield. Thus even if methane digestion had no production
costs, a substantial portion of the value of ethanol would be lost for every unit of organic
acids converted to methane. Methane production from soluble organics was considered
in the Chem Systems design as an alternative to multieffect evaporation to concentrate
the organics followed by combustion; the two options were roughly equivalent in cost.
If acetic or lactic acid is co-produced with ethanol in a wood-based plant, the eco-
nomics for both products will suffer because there will be equipment specifically
devoted to each product, produced on a smaller scale than if it were the only product.
34 L.R. Lynd

and the size of the plant is limited by the availability of substrate. Moreover, co-
production of ethanol and organic acids is likely to face severe problems of unequal
demand if ethanol is produced on a scale able to contribute significantiy to fuel re-
quirements.
In light of the considerations above, the assumption that organic acids can be dis-
posed of with no net revenue or cost may be realistic for many cases, particularly
when large-scale ethanol production is considered. In the case of large scale production,
there is a very great incentive to achieve high ethanol yields in thermophilic processes.
In particular, the discussion above points to the greater sensitivity of process economics
to ethanol yield than to ethanol tolerance. For example an 8 0 ~ decrease in ethanol
concentration relative to the base case brought about a 2.2 cts L-a increase in the
ethanol selling price. This same increase would be expected to arise from a roughly
8 ~ decrease in the ethanol yields.
Based on the consideration of the distinguishing features of thermophilic bacteria
presented above, it is concluded that the two key advantages of thermophilic bacteria,
cellulase production and pentose utilization, potentially have large impacts on overall
process economics. Specifically they reduce the ethanol selling price for ethanol
production from wood using enzymatic hydrolysis and yeast by about a factor of 2.
Approximately the same cost reduction is realized in comparison to acid hydrolysis-
based processes, since production costs are similar for acid- and enzymatically-
catalyzed cellulose hydrolysis 5o). At least for the case where ethanol is continuously
removed, ethanol tolerance does not appear to be as important a limitation for thermo-
philic ethanol production as has often been claimed. The capital costs for conven-
tional distillation are a relatively small fraction of total capital costs in yeast-based
ethanol production. Thus a process which can more efficiently separate ethanol at
low concentrations can be substituted in the thermophilic case with a relatively small
cost penalty. The IHOSR/extractive distillation process considered above appears
to be one such process. High ethanol yields appear to be a requirement for practical
thermophilic ethanol production at high production levels. At lower levels of pro-
duction, significant production of organic acids could be acceptable only if progress
were made in the areas of organic acid recovery and/or conversion to fuels.

3.3 Comparison with Other Ethanol Production Processes

Since wood-based processes are in the research stage and are presently uneconomical,
it is of interest to compare the potential economics for ethanol production from wood
using thermophilic bacteria with ethanol production processes presently in use.
Table 9 compares economics for thermophilic ethanol production from wood, based
on the analysis above and assuming high ethanol yields, with economics for ethanol
production from corn using yeast, based on the design by Katzen et al. 39), and also
for production from ethylene using data presented by Hacking 56). For the sake of
this comparison, no by-product credit for CO 2 is taken for thermophilic production
from wood, and maintenance, overhead, taxes and insurance are calculated on a
common basis. When CO 2 is not recovered, all process fuel and electricity require-
ments are satisfied from burning process residues in the thermophilic case. The ethanol
selling prices presented in Table 9 for the three processes are essentially equal. This
Production of Ethanol from Lignocellulosic Materials Using Thermophilic Bacteria 35

Table 9. Comparison of the potential economics of ethanol produced from wood using thermophilic
bacteria with economics for ethanol produced from corn using yeast and from ethylene by Chemical
synthesis (mid 1984 U.S. $)

I II III
Wood/thermophities Corn/yeast b Ethylene/catalyst ~
(potential, no CO 2
recovered)a

Capacity (L a 1) 139x 106 139>(106 189x 106

Capital cost ($ x 106) 74.4 70.6 57.8
Production cost (cts L-1)
Substrate 9.62 22.88 16.57
Other raw materials 2.59 2.27 1.29
Labor, maintenance, overhead. 6.12 6.01 4.39
taxes and insurance d
Fuel and electricitye 0 1.74 4.52
By-product credits 0 -- 13.21 0
Totalproduction cost 18.33 19.69 26.7
Capital recovery 10.65 10.09 6.08
(15 ~ R.O.I., 10 years)
Selling price 29.0 29.8 32.9

Costs are based on the modified thermophilic case considered in Tables 7 and 8, but with no re-
covery of CO 2;
b capital, labor, and utilities costs are based on the design of Katzen and Associates 39). Adjustment
was made for scale by interpolatingbetween costs given for plants with capacities of 94.6 x 106 L a - 1
and 189.3 x 106 L a -1. Adjustment is also made for the year of the cost estimates from the end of
1978 to mid 1984 using standard indices for capital and coal. A price of 0.06 $ kWh -1 is used for
purchased electricity as in the Chem Systems study 51). The price of the distillers' dried grain by-
product is based on data from 3v). A corn price of 3 $ bushel 1 (0.12 $ kg-1);
c capital costs and materials other than ethylene are based on values presented by Hacking 56) taken
from data of Flannery and Steinschneider, adjusted from first quarter 1981 to mid 1984 using a
6 ~ per year growth rate. The price of ethylene is taken to be 33 cents kg-i. Utilities are from per-
sonal communication with Genaro Maffia, Arco Oil;
d these items are charged for on a common basis as follows: maintenance, 4 ~ of capital costs;
overhead, 0.5 x (labor + maintenance); taxes and insurance, 2 % of capital costs;
costs are for fuel and electricity for I and II, and for fuel, electricity and also other utility-related
costs, such as feed water conditioning, for III

suggests that if e t h a n o l yields c o m p a r a b l e to yeasts can be achieved by thermophiles,

a n d o t h e r a s s u m e d characteristics o f t h e r m o p h i l e s , e.g. utilization o f high substrate
c o n c e n t r a t i o n s , are also experimentally d e m o n s t r a t e d , t h e n t h e r m o p h i l i c e t h a n o l
p r o d u c t i o n f r o m w o o d can be expected to be c o m p e t i t i v e with yeast-based p r o d u c t i o n
f r o m c o r n and synthetic e t h a n o l p r o d u c t i o n f r o m ethylene. I f the high yields and
o t h e r a s s u m e d characteristics were realized by t h e r m o p h i l i c systems, then features
o f t h e r m o p h i l e s ofter than the distinguishing features considered thus far m i g h t be-
c o m e i m p o r t a n t in d e t e r m i n i n g the competitiveness o f the processes considered in
T a b l e 9.
T h e similarity o f the cost structure for t h e r m o p h i l i c e t h a n o l p r o d u c t i o n f r o m w o o d
and yeast-based e t h a n o l p r o d u c t i o n f r o m c o r n as p r e s e n t e d in Table 9 is striking. In
p a r t i c u l a r the net cost o f substrate and capital costs are very similar. Clearly the
c o r n / y e a s t case is uninteresting in the absence o f b y - p r o d u c t credits. O f the three
36 L.R. Lynd

processes, the selling price of ethanol produced from ethylene is the most sensitive
to raw material costs and the least influenced by capital costs. It may be noted that
in 1982 the real price of ethylene was nearly twice the 33 cts kg -~ value assumed
in Table 9. Factors such as operation at ethanol concentrations greater than 1.5 ~o
and utilization of a less expensive substrate, for example municipal solid wastes,
would make the thermophilic process more competitive.

4 Progress Toward Realization of the Potential of Thermophilic

Bacteria for Ethanol Production

Having identified cellulase production, pentose utilization, ethanol tolerance, and

ethanol yield as the important distinguishing features of thermophiles for ethanol
production (Sect. 3.1), research results in each of these areas are reviewed and eva-
luated.

4.1 Cellulase Production and Activity

The production of cellulase by thermophilic bacteria is one of their most important

attributes for ethanol production (Sect. 3). Knowledge of cellulase production and
activity in thermophilic bacteria is essentially restricted to Clostridium thermocellum.
The cellulase activity of C. thermocellum is predominantly cell-associated in expo-
nential phase, and cell-free in stationary phase 148.149). Glucose and cellobiose are
the main products of cellulose hydrolysis 150,151). In addition to cellulose hydrolysis,
C. thermocellum cellulase also catalyzes xylan hydrolysis 150). Cellulase or other
extracellular enzymes of C. thermocellum also catalyze the hydrolysis of pectin 152)
Johnson et al. 151) have demonstrated that the cellulase system of C. thermocellum
catalyzes complete hydrolysis of purified crystalline cellulosic substrates such as
Avicel, cotton, and filter paper. Rates of cellulose hydrolysis are comparable for
reaction mixtures from C. therrnocellum 27405 and Trichoderma reesei QM414. The
specific activity of C. thermocellum cellulase is much greater than that of T. reesei.
Hydrolysis of crystalline cellulose requires Ca ++ and is inhibited by oxygen 151),
and also by cellobiose 153). Cellulase activity on crystalline substrates has character-
istics distinct from activity on noncrystalline substrates with respect to both product
inhibition 153), and thermal deactivation 151). The formation of cellulase with activity
toward crystalline substrates is repressed by cellobiose, and derepressed by Avice1154).
At least two protein components have been found to be required for crystalline
cellulase activity by Wu and Demain 155), though many more components are present
156,2,157). Recently a single protein produced by C. thermocellum apparently with both
endo- and exo-glucanase activity has been isolated 158). A cellulose-binding, cellulase-
containing complex produced by this organism has been isolated by Lamed et al. 159)
which is responsible for most of the activity observed in culture broths and has proper-
ties consistent with those of unpurified extracellular protein 16o~.This complex, termed
a cellulosome, is composed of at least 14 distinct polypeptides with a combined mole-
cular weight of approximately 2.1 MDa. During exponential growth on cellulose, the
cellulosome is anchored to the cell surface, and allows adherence of the cell to the
Production of Ethanol from Lignocellulosic Materials Using Thermophilic Bacteria 37

insoluble cellulosic substrate. The rate o f cellulose hydrolysis by intact cells is signi-
ficantly higher than that o f the cell-free eellulase system 2).
Coughlan et al. 16~) have reported that C. thermocellum cellulase can be resolved
into 2 m a j o r complexes with diameters o f 210 A and 610 A and molecular weights of
4.2 and 102 M D a respectively; other reports have been consistent with these very
large dimensions ~s6,159). These values may be c o m p a r e d to a representative diameter
of 51 A 1) and component molecular weights of < 100 k D a 162) for T. reesei cellulase.
In spite o f this great difference in size, cellulase produced by C. thermocellum is
roughly as effective as cellulase produced by T. reesei in attacking microporous sub-
strates which have progressively smaller surface area accessible to larger molecules ~63).
Furthermore, the m i n i m u m pore dimension required for access by C. thermocellum
cellulase is essentially the same as that required by the cellulase o f T. reesei, about
43 A 687. The apparent contradiction between the reported sizes o f these cellulase
systems and the pore size required for substrate accessability raises questions about
structure-function relationships in the cellulase enzyme complex o f C. thermocellum.
The most i m p o r t a n t practical question regarding C. thermocellum cellulase is its
effectiveness against complex, lignin-containing cellulosic substrates. D a t a are
available on the cellulase activity toward lignocellulose from in vivo studies 13o. ~64,
~65,166. a67, ~6s), however the properties and limitations of the enzyme are not easily
distinguished from those o f the organism in these studies.
A direct in vitro evaluation o f the effectiveness o f C. thermocellum cellulase on
lignocellulosic substrates is possible based on recent work by the author and collea-

Table 10. Comparison of hydrolysis yields for various systems including C. thermocellum or T. reesei

Source of Substrate I Pretreatment Solids Enzyme Reaction Yield5 Refs.

cellnlase conditions Concen- Loading3 Time4
tration2
Temp. [H2SO4] Time
(~ (~) (s) (gL -1) (IUg -~) (h) (%)

A. In vitro
C. thermocellum Hardwood 220 t.0 9.0 2.5 36.0 13 95.0 169)
T. reesei Hardwood 220 1.22 7.2 14.0 24.3 24 85.6 17o)
C. thermocellum Whitepine 220 1.0 10.0 0.52 7.0 47 43.0 169)
T. reesei White pine 220 1.08 7.8 15.0 11.3 48 26.8 17o~
T. reesei Popular 200 0.5 .7.9 18.14 27.6 24 100 17~)
B. In vivo
C. thermoeellum Hardwood 220 1.0 9.0 4.14 7.0 12 86 169)
C. thermocellum Hardwood 220 1.0 9.0 5.48 2.74 16 88.6 169)

" "Hardwood" is 90 ~ birch, 10 % maple;

calculated for data from 17o)using values given for unpretreated solids assuming all hemieellulose
and 10 % of lignin are solubilized;
r calculated based on free cellulase activity for continuous cultures. Activity for C. thermoeellum
is measured on Avicel, and activity of T. reesei cellulase is measured on filter paper. However, the
activity of both C. thermocellum and T. reesei cellulases are about equal on Avicel and filter pa-
per ls1~;
d reaction time: time of experiment for batch experiments; residence time for continuous experiments:
~ all yields are based on the glucose content of the solids before reaction
38 L.R. Lynd

gues. In vitro and in vivo hydrolysis data are presented in Table 10, and compared
with data for the cellulase of Trichoderma reesei. In vitro hydrolysis yields at compar-
able enzyme loadings and reaction times are somewheat higher for the C. thermo-
cellum cellulase system than for T. reesei cellulase acting on both pretreated mixed
hardwood and pretreated white pine. The substrate concentrations are however
lower for the results with C. thermocellum. The difference between the yields on pine
and mixed hardwood attests to the difficulty of effectively pretreating softwood pre-
viously mentioned (Sect. 2.2.2). Poplar hydrolysis by T. reesei cellulase is included
in Table 10 to illustrate how much more easily this substrate is hydrolyzed by this
system than is mixed hardwood. Hydrolysis of poplar using C. thermocellum cellulase
has not been studied. Yields from in vivo cellulose hydrolysis by C. thermoeellum
are slightly lower than in vitro yields using C. thermocellum cellulase, and are essential-
ly equal to the yields obtained with T. reesei cellulase at comparable reaction times.
The lower apparent enzyme loading, based on free cellulase, in the in vivo studies may
be indicative of most of the enzyme being bound to cells or substrate, and/or a greater
effectiveness of the enzyme in the presence of cells.
The data in Table 10 suggest that high extents of substrate utilization should be
attainable with thermophilic systems and pretreated hardwood substrates. De-
monstration of comparable performance at high substrate concentrations represents
an important goal for future applied research in this area. Dilute-acid pretreatment
of softwood substrates appears unlikely to be a successful pretreatment for this
enzyme system, as with others. Testing C. thermocellum cellulase activity against
substrates pretreated by means other than dilute acid hydrolysis would be informative.

4.2 Utilization of Pentose Sugars

The ability of thermophiles to consume a very broad range of carbohydrates, in-

cluding pentose sugars and their polymers, is well established. This ability is also a
key advantage of thermophiles for production of ethanol from lignocellulose. In
particular, the simultaneous utilization of the hexose and pentose sugars present in
biomass is desirable from a practical point of view. Thermophiles such as Clostridium
thermohydrosulfuricum, C. thermosaccharolyticum, whose taxonomic status has been
questioned 10), and Thermoanaerobacter ethanlicus utilize xylose and other pentose
sugars at rates comparable to hexoses 8,10). Notably, all of these species are not
cellulolytic, and the cellulolytic C. thermocellum, does not utilize pentoses 11). Thus
simultaneous utilization of pentose and hexose sugars as present in pretreated bio-
mass by described species of thermophilic bacteria requires that both a cellulolytic
and pentose-utilizing organism be present. Cocultures have been studied with C.
thermocellum paired with C. thermosaccharolyticum 130,16s), C. thermohydrolsulfuri-
c u m 164,168,172), and T. ethanolicus 7,172). The role of the non-cellulolytic bacterium

in these co-cultures, reviewed by Carreira and Ljungdahl 7), includes hexose as well
as pentose utilization. This is suggested by marked effects of the presence of the non-
cellulolytic organism in cultures grown on pentose-free substrates, and also the
general tendency for end-product profiles in co-cultures to more nearly reflect that
of the non-cellulolytic bacterium.
Production of Ethanol from LignocellulosicMaterials Using ThermophilicBacteria 39

Simultaneous utilization of hexose and pentose sugars has been investigated to a

limited degree in batch culture, and not at all in continuous culture. Slaff and Humph-
rey ~73) report diauxic utilization of glucose in preference to xylose or cellobiose for
C. thermohydrosutfuricum, with xylose and cellobiose consumed simultaneously in the
absence of glucose. Carreira et al. ~74) found that glucose and xylose are used si-
multaneously by T. ethanolicus, but that glucose is used in preference to starch. It
may be noted that C. thermocellum uses cellobiose in preference to glucose 175)
Limited data are available on pentose and hexose utilization during growth on in-
soluble substrates. Avgerinos and Wang 130) reported 85--90 % utilization of both
glucose and xylose at the end of batch conversion of solvent-extracted corn stover,
indicating that both carbohydrate species were eventually used to a large extent.

4.3 Ethanol Tolerance

The ethanol tolerance of thermophilic bacteria is an important applied characteristic

which, along with ethanol yield, is usually seen as the greatest barrier to the commercial
utilization of thermophiles for ethanol production ~4, 166,176). Slapack et al. 8) and
Lovitt et al. 10) have reviewed strain development to increase ethanol tolerance in
thermophilic bacteria, and Slapack et al. 8) and Lovitt et al. ~0) and Rogers 3) have
reviewed mechanisms of ethanol tolerance and sensitivity.
The mechanisms of ethanol tolerance and intolerance in thermophiles may be dif-
ferent for different species. Increasing ethanol resistance with decreasing temperature
has been reported for both C. thermocellum 177) and C. thermohydrosulfuricum 178)
This observation was used to support an explanation for ethanol inhibition in terms
of membrane fluidity in the former study, but r~ot in the latter. Slapack et al. 8) discuss
membrane fluidity effects in detail. Herrero et al. 179)attribute ethanol intolerance to
a blockage in glycolysis, possibly in response to ethanol-induced changes in the
membrane 180,18~). However, Lovitt et al. a0) concluded that ethanol inhibition was
not due to either of these effects but is due to regulatory phenomena. It may be noted
that Lovitt et al.'s interpretation of ethanol sensitivity is based in part on greater
sensitivity to ethanol than to other solvents (methanol and acetone). Though the
tolerance of C. thermocellum to acetone and methanol has not to the authors knowledge
been published, this organism does exhibit different solvent sensitivity, with sensi-
tivity to ethanol greater than to either propanol or butanol xsz)
Lovitt et al. 10) have recently proposed a mechanism for ethanol inhibition in C.
thermohydroIsulfuricum, which also is consistent with the sensitivity of this organism
to inhibition by hydrogen. In the presence of inhibitory concentrations of either
ethanol or HE, the NADH/NAD ratio increases by a factor of about 2.5 in wild-type
cells, but not in an ethanol-resistant mutant. Addition of acetone provides an acceptor
for electrons carried by NADH, and relieves inhibition. These effects have been
attributed to inhibition of glycolysis, in particular glyceraldehyde dehydrogenase, due
to over reduction of the cellular pyridine nucleotide pool lo)
Though the ethanol-tolerance of wild-type thermophilic bacteria is typically < 1%
ethanol, numerous workers have selected ethanol-tolerant strains [see s)]. Mutants
with quite high ethanol tolerance have been reported. Examples include a strain of
T. ethanolicus able to tolerate 10 ~ ethanol 7), C. thermocellum $7, which can tolerate
40 L.R. Lynd

6 % ethanol with < 50 % inhibition of growth (the basis is not clear) 14),and C. thermo-
hydrosulfuricum 39EA 178),which grows well at 8 % ethanol at 45 ~ and 5 % ethanol
at 60 ~
There is a substantial difference between reports of the concentrations of ethanol
tolerated by thermophiles and the concentrations of ethanol they produce. In a
Masters thesis completed in 1982, Kim 183) observed production of 60 g L -1 ethanol
by C. thermosaccharolyticum growing on xylose in fed-batch culture. However, more
than half this concentration was produced after the cessation of exponential growth,
and ethanol production occurred over a period of nearly 200 h. Writing in 1983,
Carreira et al. 174) state that the highest concentration of ethanol reported to be
produced by thermophilic anaerobes from xylose, glucose or starch is ~ 3 % v/v
ethanol. C. thermoeellum is generally considered to be less ethanol tolerant than the
frequently-studied non-cellulolytic thermophiles 7,10). Ethanol concentrations of
1.5-1.7 % have consistently been produced by batch co-cultures of C. thermocellum
and C. thermosaccharolyticum (personal communication, A. Demain, MIT). The
recent and comprehensive reviews of Slapack et al. 8) and Lovitt et al. lo) mention
no values higher than 3 4 % ethanol produced by thermophiles in batch culture, and
it is not stated what fraction of this is formed after growth ceases. These values may
be compared to the maximum endogenously-produced ethanol concentration allow-
ing cell growth in yeast, which is about 9 wt. ~o 85). In reviewing strain development
work at MIT in 1986, Mistry 184) states that available strains of either C. thermo-
cellum or C. thermosaccharolyticum are unable to grow rapidly in the presence of
ethanol concentrations >20 g L -1, and describes a difference of 2-fold in the in-
hibition of growth by exogenous and endogenous ethanol. Mistry also reports rather
severe p . ~ c m s with revertance for an ethanol tolerant mutant of C. thermocellum
strain $7. The highest ethanol concentration known to the author produced by ther-
mophiles in continuous culture is 2 to 2.4 wt. %, reported by Kim 183) for C. thermo-
saccharolyticum. In work to be published with the same organism, the author and
colleagues have observed values of 1.5 to 1.7 wt. % ethanol in continuous cultures.
Continuous production of ~ 1% ethanol has been reported for C. thermosaccharo-
lyticum 18~), and C. thermohydrosulfuricum 185)
Within the thermophilic ethanol tolerance literature, reported ethanol tolerance
is generally highest for exogenously added ethanol, intermediate for endogenously-
produced ethanol formed in batch culture, often with extensive ethanol formation
after cessation of growth, and lowest for ethanol produced by exponentially-growing
cells. Distinguishing between these measures of ethanol tolerance is important be-
cause typical values for the various measures differ by a factor of at least 3, and perhaps
5. Realistically interpreting ethanol tolerance and production data for thermophilic
bacteria is made more difficult by the fact that experiments in which high ethanol
concentrations are obtained are often not presented in detail but are referred to as
unpublished results or personal communication. Instability of ethanol-tolerant strains
further complicates evaluation 184).
The concentration of endogenously-produced ethanol is a more relevant measure
of ethanol production capability from a practical perspective than is exogenously-
tolerated ethanol. In some cases relative tolerance to exogenous ethanol may indicate
a potential not yet achieved in studies on produced ethanol. However, it is well es-
tablished that yeast is more tolerant to exogenously added ethanol than to endogenous
Production of Ethanol from LignocellulosicMaterialsUsing ThermophilicBacteria 41

ethanol 186); the same appears to be true of thermophiles 184).The ethanol concentrati-
on which can be achieved in free-cell continuous culture is that produced and tolerat-
ed by exponentially-growing cells. Higher ethanol concentrations have been observed
and are likely to be achieved in batch culture. However, batch culture systems for
ethanol production using thermophilic bacteria can be expected to have lower pro-
ductivity than continuous systems, and are not easily coupled with continuous ethanol
removal (see Sect. 3.1). Cell immobilization and cell recycle systems may allow
relatively high ethanol concentrations to be produced while enjoying some of the
advantages of continuous systems. These systems are generally less easily implimented
when insoluble substrates are employed. Differential residence times for liquid and
insoluble substrates in a continuous system are relatively easy to achieve, and may
provide a means of retaining cells attached to the substrate.

4.4 End-Product Metabolism and Ethanol Yields

Metabolic control and manipulation of end-product yields, and the selection of

mutants with high ethanol yields are reviewed by Slapack et al. 8), Rogers 3), and
Lovitt et al. 10). The branched pathway leading to ethanol, acetic acid, and lactic
acid is shown in Fig. 6. Considerations pertaining to end-product metabolism are
similar for hexose and pentoses utilization because end-products are formed via
pyruvate from both classes of substrates 3). The carbon intermediates in this pathway
are generally the same for different species of thermophilic anaerobes, however the
electron carriers and also demonstrated reversibility for particular reactions varies
between species and even strains 187,188). It is theoretically possible for ethanol,
acetic acid or lactic acid to be the sole soluble product in that the pathways leading to all
of these products result in balanced production and consumption of electron carriers
and net generation of ATP. For production of acetate only, the NADH produced by
glycolysis could be reoxidized with the concomitant reduction of ferredoxin and sub-
sequent production of hydrogen gas.
Two classes of regulatory mechanisms may be involved in thermophilic endproduct
metabolism: those which regulate end-product synthesis in response to the flux of
intermediates in the pathways leading to particular end-products, and those which
respond to the concentration-of the end-products. These mechanisms are not mutually
exclusive, and may both have a role. The relative importance of these mechanisms is
important not only scientifically, but also practically. Concentration-dependent
mechanisms should allow higher ethanol yields by manipulation of end-product
concentrations in the bioreactor environment, for example by ethanol removal, whereas
flux-dependent mechanisms should not allow this.
In a classic paper, Thauer et al. 189)review end-product control of anaerobic meta-
bolism. They explain end-product distributions in terms of regulation of metabolic
efficiency by controlling the flux of intermediates in pathways leading to products
with different ATP yields. Mistry 190) has recently employed intrinsically-based
mechanisms in explaining the end-product yields of C. thermosaccharolyticum.
Control of the rate of reactions in enzyme-catalyzed metabolic pathways is different
for reversible reactions than for essentially irreversible reactions 191). The rates of
42 L.R. Lynd

CELLULOSE
G LUCOSE ADP+Pi
NAD(P)H+hr ~Dr
NAD(P)e / ~--------~.~DH+Hr
LACTATEe+Hr ~ // _ PYRUVATE~+H
~, | -ATP H2
CoASH~ I~ /
Fdo• ~,,_~/~ 2Hr
..~f---~'NAD(P)H+He
C02 " - - - - - ~ l ~ Fdred~ L.......,,..NAD(P)e
AC ETYLC / ~ ETALDEH~E~ANOL

CoASH = ~ NAD e i~ NADip)e

NADH+H e NAD(P)H+H e
ACETYLPHOSPHATE
ADP ~
ATP ~
AC ETATEe+He
Fig. 6.Typicalend-productmetabolismof thermophilicethanol-producingbateria data from 7,187,188)
Reactions shown as reversiblehave been demonstrated to be reversiblein at least one thermophilic
species

reversible reactions are controlled by substrate and product concentrations. The rates
for essentially irreversible reactions are controlled by enzyme activity, and are fre-
quently subject to allosteric control. As shown in Fig. 6, the majority of the reactions
involved with end-product metabolism in thermophiles have been shown to be re-
versible in at least some organisms. Many, and for some organisms all, of the re-
actions for interconversion of the electron carriers NAD, NADP, and ferredoxin
have been shown to be reversible. Thus the concentration of end-products can be
expected to exert an effect on the pattern of end-products formed. Several examples
of such effects are discussed below.
According to Slapack et al. m, the most important factors in regulating end-product
ratios in thermophilic anaerobes are alcohol dehydrogenase, ferredoxin: NAD(P)
oxidoreductases, and FDP-activated lactate dehydrogenase. The role of hydrogen is
discussed by Rogers 3). Generally, increasing hydrogen partial pressure increases
ethanol yield, though the sensitivity to hydrogen depends on the activity of ferrodoxin:
NAD(P) oxidoreductase. Dramatic increases in ethanol yield due to increased hydro-
gen partial pressure have been reported for C. thermohydrosulfuricum 195), C. thermo-
saccharolyticum 18~), and C. thermocellum 192). All of these reports describe super-
saturation of the growth medium with hydrogen, and changes in the ethanol yield
by variables which would be expected to effect the degree of hydrogen supersaturation,
e.g. gas sparging and stirring rate. Lovitt et al. lo) discuss the differences in response
Production of Ethanol from LignocellulosicMaterials Using ThermophilicBacteria 43

to hydrogen amoung thermophiles and possible mechanistic explanations for these

differences.
Mass-action effects involving soluble end-products have also been implicated in
determining end-product ratios. Herrero-Molina 182) found that addition of each of
the fermentation products of C. thermocelIum (ethanol, acetate, and lactate) results in
decreased synthesis of the added product and increased synthesis of the other products.
Altered end-product ratios due to selective addition or removal of a particular product
has also been reported in other systems involving thermophilic ethanol producing
bacteria 193,194-)
As discussed in Sect. 3.1, the added costs associated with operating at low ethanol
concentration may be limited to ethanol separation per se by employing continuous
ethanol removal. In addition, ethanol removal may be expected to increase ethanol
yield if concentration-dependent control mechanisms are operative. Only preliminary
lab-scale studies of continuous ethanol removal from thermophilic systems have been
undertaken, some of them without proper regard for energy requirements. In addition
to careful consideration of energy requirements, it is also important to evaluate the
physiological impact of proposed ethanol removal techniques. Though ethanol
removal may have the potential to increase ethan01 yields, the opposite effect has
been observed. Sundquist et al. lS~)found that ethanol removal via a reduced pressure
flash vessel decreased ethanol yields due to reduced hydrogen partial pressure. In
reduced-pressure ethanol removal systems, minimization of the time spent in the
ethanol removal apparatus is likely to be critical.
Increasing iron concentration has been found to promote ethanol production in
C. thermosaccharolyticum 195), a result explained in terms of limitation of ferredoxin
synthesis and thus pyruvate dehydrogenase activity. However, Ljungdahl et al. 196)
report using iron limitation to obtain a strain of T. ethanolieus with high ethanol
yields. Growth rate has also been implicated in controlling ethanol yield. Zeikus et
al. 128) and Ben-Bassat et al. 194) report high ethanol yields at low growth rates with
both Thermoanaerobium brockii and C. thermohydrosulfuricum. Avgerinos and
Wang 13o) were able to obtain high ethanol yields only at high growth rates using
C. thermosaccharolyticum and C. thermocellum in both mono- and co-culture.
Thermophilic bacteria, including both wild-type and mutant strains, have been
reported with high ethanol yields. For example, an ethanol yield of 1.95 moles mole- 1
glucose has been reported for T. ethanolicus JW200 174), and a yield of 1.9 moles
mole -1 glucose has been reported for C. thermohydrosulfuricum 39E 197). (These
yields are both higher than would be expected based on incorporation of 10 ~ of the
substrate into cell material, which may indicate uncoupled growth and substrate
consumption). However, substantially lower yields have also been obtained using
both of these species, including the same strains 172,173,198). Unfortunately, ethanol
yields appear to be rather sensitive to growth conditions. In particular, several studies
have found ethanol yields to be higher on laboratory substrates than on heterogeneous
substrates derived from native biomass 13o,172). Personal communication with several
investigators suggest that high-yielding strains do best in the hands of the investigators
that isolated them. This observation also applies to ethanol tolerant mutants. Mutants
with ratios of ethanol to acidic end-products of > 5 : 1 have been obtained for both
C. thermocellum (strain S-7, 14~) and C. thermosaccharolyticum (strains HG6 and
HG8, 13o.199,200)). However, neither of these strains perform so well in the hands of
44 L.R. Lynd

the author. DI:. Demain's group at MIT found the reduced acetate production re-
ported for a mutant strain of C. thermosaccharolyticum 2Ol) to be an unstable char-
acteristic.
From a mechanistic point of view, it is very clear that concentration-dependent
metabolic control mechanisms are operative in thermophilic ethanol-producing
bacteria. The situation is not so clear with respect to flux-dependent mechanisms. At
the very least, the "set point" in terms of efficiency of energetic coupling appears
to be very flexible in light of the drastically different product yields produced by
different thermophilic species and strains. A definitive work on ethanol removal from
thermophilic cultures might be expected to shed further light on this question, but
has yet to appear in the literature.
From a practical point of view, the goal of obtaining stable thermophilic strains
producing ethanol at high yields and relatively high concentrations from practical
substrates has not been entirely achieved to the author's knowledge. Development of
cellulolytic strains with these properties would appear to be a particular priority for
research.

5 Concluding Remarks

Notwithstanding short term economic factors and trends in the attention of scientists
and policy makers, the prospect of decreasing and ultimately vanishing supplies of
oil is a real problem today for the same reasons it was a real problem five years ago.
Moreover this problem is not one confined to the distant future. In countries such as
the United States, where demand for oil is particularly high relative to reserves, large
increases in the real oil price have been forecast within the next 15 years 22). Strategies
aimed at maintaining oil reserves, boosting domestic production, and maintaining
good relationships with oil-rich countries can buffer the effects of oil exhaustion in
oil-poor countries. However, developing alternatives to oil is the only real solution.
Biological production of ethanol is worthy of attention today as a route to partially
replacing oil for the same reasons this process received considerable attention five
years ago. Ethanol is a versatile chemical feedstock, and an excellent fuel which can
displace gasoline on a better than 1 : 1 basis in energetic terms. Furthermore biological
systems are well-known for their desirable catalytic features including high selectivity
and reaction under mild conditions 202). However, in order to make a significant
contribution to replacing oil, ethanol production systems must convert a plentiful
feedstock in a manner which is acceptable from economic, environmental, and ener-
getic points of view. From this perspective, production of ethanol from starch- or
sugar- rich substrates, as is presently practiced, has limited potential. Efforts to
develop processes for ethanol production from substrates other than sugar- and
starch-rich agricultural crops have not been successful to date, though a decade of
intensive research has provided many insights of both fundamental and applied
value.
The quantity of fermentable carbohydrate available for ethanol production from
lignocellulosic substrates in the United States appears to be sufficiently large that
conversion of this material into ethanol to replace liquid transportation fuels is a
Production of Ethanol from LignocellulosicMaterialsUsing ThermophilicBacteria 45

realistic possibility in this country. Biomass supplies are on average larger relative to
oil consumption irl countries other than the U.S. Compared to corn, the potential
ethanol production from lignocellulosic substrates in the U.S. is greater by a factor
on the order of 20. At ethanol production levels above that which can support the
corn by-product market as an animal feed, lignocellulosic substrates potentially have
superior features to corn from energetic and environmental viewpoints, and essentially
do not compete with food production. In addition to these advantages, lignocellulosic
substrates can be distinguished from starch- and sugar-rich substrates by their rela-
tively high content of insoluble polymers containing [3-1inked glucose (cellulose) and
pentoses (found in hemicellulose). The difficulty in economically converting these
components has been primarily responsible for the incomplete success to date in
efforts to develop practical processes for ethanol production from lignocellulose.
Very little research on thermophilic ethanol production has been reported under
conditions which approach being practical. However, the value of distinguishing
features of thermophilic bacteria for ethanol production can be assessed in economic
terms in order that the potential of these organisms can be more clearly defined, and
priority research areas identified. Using this approach, this study concludes that
thermophilic bacteria offer the potential for a large, e.g. two-fold, reduction in
ethanol production costs compared to processes based on available technology
using enzymatic or acid hydrolysis and yeast. These yeast-based processes may of
course be improved 7o, 7s~,however they are likely to have several more process steps,
and other things being equal higher cost, in the absence of breakthroughs in substrate
utilization capability. The primary potential advantages of thermophilic bacteria
compared to yeast for conversion of lignocellulosic substrates arise from their ability
to utilize cellulose and pentoses, and not from process advantages conferred by
thermophily. Thus thermophilic bacteria are particularly suited to lignocellulosic
substrates rather than starch- and sugar-rich substrates which can be made convertable
by yeasts at relatively low cost.
In evaluating thermophilic bacteria for ethanol production, it has frequently been
assumed that ethanol tolerance comparable to yeast is required. Research aimed at
strain development to achieve such ethanol tolerance has been unsuccessful to date.
The requirement for high ethanol tolerance is likely to apply if separation technologies
which are designed for the ethanol concentrations tolerated by yeasts are employed.
One general conclusion of the present study is that product recovery technologies
which have higher capital costs but low energy requirements for separation of dilute
ethanol solutions can have small cost penalties in terms of overall process economics.
In particular, these costs are small relative to the reduction of total costs resulting
from the key advantages of thermophiles for ethanol production from lignocellulose.
Engineering-based approaches appear able to lower the threshold ethanol con-
centration for a practical process into the range of concentrations produced by actively
growing thermophilic cultures. It is very likely that approaches to dealing with the
problem of low ethanol yields will be based to a much larger extent on biochemical
understanding of the metabolism of thermophilic bacteria. Both strain development,
using a variety of screening and selective techniques, and manipulation of environ-
mental conditions have been somewhat successful in obtaining high ethanol yields
under some conditions. However, stability of high ethanol yielding strains has been
a problem with respect to both repeatability of results and sensitivity to growth
46 L.R. Lynd

conditions. Moreover, few strains with high ethanol yield have been tested under
repeated sub-culturing or continuous culture.
In general, process-oriented studies of thermophilic ethanol production are scarce
in the literature, and much needed if development of this technology is to proceed.
Particularly lacking are studies in continuous culture, studies at high concentrations
of economically interesting feedstocks containing a combination of substrates and
also potentially inhibitory compounds, and ethanol removal studies.
The material and analysis presented in this paper lead the author to conclude that
there is a large gap between the potential of thermophilic bacteria for ethanol pro-
duction, even at relatively low ethanol concentrations, and that which has been
experimentally demonstrated. Process studies as described above and research to
obtain stable thermophilic cultures with high ethanol yields under realistic conditions
would appear to be particularly important topics for study in order to close this gap.

6 Acknowlegements
I am grateful to the many people who contributed to this review by generously making
available their reference collections, original papers, and ideas. These include Drs.
H. W. Blanch, A. O. Converse, C. L. Cooney, R. Datta, A. L. Demain, H. E. Greth-
lein, C. High, L. G. Ljungdahl, L. Mednick, T. Peterson, K. Venkatasubramanian,
R. H. Wolkin, and J. G. Zeikus, and also R. Chamberlin, C. Leach, K. Levinson,
and G. Maffia. I thank C. Johnson and D. Call for donating their time and skill in
the preparation of figures. Also I thank Drs. Converse, Wolkin, S. C. Lynd, and also
J. Byrnes, for critically reading and proof-reading the manuscript. Finally I thank
Drs. A. E. Balber, H. E. Grethlein, N. J. Poole, and J. G. Zeikus for inspiration
and encouragement.

7 References
1. Cowling EB, Kirk TK (1976) Biotechnol. Bioeng. Symp. 6:95
2. Lamed R, Bayer EA (In press) The cellulosome of Clostridium thermocellum. In: Laskin AI (ed)
Advances in applied microbiology. Academic, NY, vol 33
3. Rogers P (1986) Genetics and biochemistry of Clostridium relevant to development of ferment-
ation processes. In: Laskin AI (ed) Advances in applied microbiology. Academic, New York,
vol 31 p l
4. Hartley BS, Payton MA 1,1983) Biochem. Soc. Symp. 48:133
5. Sonnleitner B (1983) Biotechnology of thermophilic bacteria growth, products and appli-
cation. In : Fiechter A (ed) Advances in biochemical engineering/Biotechnology. Springer, Berlin
Heidelberg New York, vol 28 p 69
6. Sonnleitner B, Fiechter A. (1983) Trends Biotechnol. 1(3): 74
7. Carreira, LH, Ljungdahl LG (1984) Production of ethanol from biomass using anaerobic ther-
mophilic bacteria. In: Wise DL (ed) Liquid fuel developments (CRC series in biotechnology)
CRC, Boca Raton FL
8. Slapack GE, Russel, I, Stewart GG (1987) Thermophilic microbes in ethanol production. CRC,
Boca Raton FL
9. Wiegel J, Ljundahl LG (1986) CRC Crit. Rev. Biotechnol. 3(1): 39
10. Lovitt, RW, Kim, BH, Shen G-J, Zeikus JG (In press) Solvent production by microorganisms
(to be published in CRC Critical Reviews)
Production of Ethanol from Lignocellulosic Materials Using Thermophilic Bacteria 47

11. Duong T-VC, Johnson EA, Demain AL (1983) Thermophilic anaerobic cellulolytic bacteria,
In: Wiseman A (ed) Topics in enzyme and fermentation biotechnology, vol 7 p 156 Halsted Press,
Hopwood N.Y.
12. Ljungdahl LG, Eriksson KE (1985)Ecology of microbial cellulose degradation. In: Marshall KC
(ed) Advances in microbial physiology, vol 8 p 237 Academic Press, N.Y.
13. Wiegel J (1982) Experentia 82:151
14. Wang DIC, Averginos GC, Biocic L Fang SD, Fang HY (1983) Philos. Trans. R. Soc. London
B300: 323
15. Grathwohl M (1982) World energy supply, de Gruyter, Berlin
16. Vergara W, Pimentel D (1978) A study of the energy potential of fuels from biomass in five
countries. In: Energy from biomass and wastes. Symposium, 14-18 May, Washington DC.
Available from The Institute, Chicago
17. Owsley DC, Bloomfield JJ (1985) Chemtech 15(2): 94
18. Exxon Corporation (1981) World energy outlook. Corporate Planning and Public Affairs Dep-
artment, New York
19. National Petroleum News Fatbook Issues, 1984-1987
20. American Petroleum Institute (1988) Basic petroleum data book -- Petroleum Industry Statistics,
8(1), Washington
21. Mast RE, Dolton GL, Crovelli RA, Powers RB, Charpentier RR, Root DH, Attanasi ED
(1988) Estimates of undiscovered recoverable oil and gas resources for the onshore and state
offshore areas of the United States. In: USGS Program and Abstract on Mineral and Energy
Resources, V.E. McKelvey Forum, Denver CO, Abstract in the U.S. Geolgical Survey Circular
1025
22. U.S. Department of Energy (1985), National energy policy plan projections. Office of Planning
and Analysis; DOE/PE-0029/3
23. _L Chemical marketing reporter. (1960 through 1986) Schnell Publishing, New York
24. Peterson T Wisconsin forest extension price reviews. Cooperative Extension Program, U.S.D.A.,
Madison, Wisconsin (April 1981 through November 1986)
25. Rudderman FK (1980) Pacific northwest production prices employment and trade. In: North-
west Forest Industries. Pacific Northwest Forest and Range Experimental Station, Portland
OR
26. Palsson BO, Fathi-Afshar S, Rudd DF, Lightfoot EN (i981) Science 213:513
27. Ng, TK, Busche RM, McDonald CC, Hardy RWF (1983) Science 219(4585): 733
28. Busche RM (1985) Biotech. Prog. 1(3): I65
29. Hall DO (1979) Fuel 57(6): 322
30. Ferchak JD, Pye EK (1981) Solar Energy 26:9
31. Energy from biological processes, volume II -- technical and environmental analyses. Office
of Technology Assessment, Congress of the United States, Washington DC (1980)
32. Humphrey AE, Moreira A, Armiger W, Zabriske D (1977) Biotechnol. Bioeng. Symp. 7:45
33. Goldstein IS (198 l) Biomass availability and utility for chemicals. In: Goldstein, IS (ed) Organic
chemicals from biomass. CRC Press, Boca Raton, FL, p 1
34. Jeffries TW (1983) Utilization of xylose by bacteria, yeasts, and fungi. In: Fiechter A (ed) Ad-
vances in biochemical engineering/Biotechnology. Springer, Berlin Heidelberg New York,
vol 27 p 1
35. Young J, Griffin E, Russell J (1986) Biomass 10:9
36. Levinson A (1982) Resource Man. Optim. 2(2): 99
37. Venkatasubramanian K, Kiem C (1985) Starch and energy: technology and economics of fuel
alcohol production. In: van Beynum GMA, Roels JA (eds) Starch conversion technology.
Marcel Dekker, New York, p 143
38. Lipinsky ES (1978) Science 199:644
39. Katzen R et al. (1978) Grain motor fuel alcohol technical and economic assessment. Available
from NTIS, Springfield VA; HCP/J6639-01
40. Bellamy WD (1975) Conversion of insoluble agricultural wastes to SCP by thermophilic micro-
organisms. In: Tannebaum SR, Wang DIC (eds) Single cell protein II. MIT Press, Carelton,
M A p 263
41. Stephens HR, Heichel GH (1975) Biotechnol. Bioeng. Syrup. 5:27
48 L.R. Lynd

42. Gong C-S, Chen LF, Flickinger MC, Tsoa GT (1981) Conversion of hemicellulose carbohy-
drates. In: Fiechter A (ed) Advances in biochemical engineering/Biotechnology. Springer,
Berlin Heidelberg New York, rot 20 p 93
43. Morrison FB (1956) Feeds and feeding -- a handbook for the student and stockman. Morrison,
Ithaca NY
44. Tchobanaglous GH, Theisen H, Eliassen R (1977) Solid wastes: engineering principles and
management issues. McGraw-Hill, New York
45. Herman MF, Othmer DF, Overberger CG, Seaborg T (eds) Kirk-Othmer encyclopedia of
chemical technology, 3rd edn, Wiley, New York (1978)
46. Gaines EL, Karpuk M (1987) Fermentation of lignocellulosic feedstocks: product markets and
values. In: Klass DL (ed) Energy from biomass and wastes X, Institute of Gas Technology,
Chicago
47. Maiorella BL, Blanch HW, Wilke CR (1983) Proc Biochem. 18(4): 5
48. Waits ED, Elmore JL (1983) Environ. Int. 9:325
49. Loehr RC, Sengupta M (1985) Environ. Sanit. Rev. 16:1
50. Badger Engineers Inc. (1984) Economic feasibility study of an acid-based ethanol plant. SERI,
Golden, CO; ZX-3-03096-2
51. Chem Systems Inc. (1984) Economic feasibility study of an enzymatic hydrolysis-based ethanol
plant with prehydrolysis pretreatment. SERI, Golden, CO; XX-0-03097-2
52. Matsuda S, Kubota H (1984) Biomass 4:161-182
53. National Research Council, carbon dioxide assessment committee (1983) Changing climate.
National Academy Press, Washington, DC
54. Hileman B (1984) Environ. Sci. Technol. 18(2): 45A
55. Rinehart S (1988) Stations start selling high-oxygen fuels. Colorado Daily, 94(284): 1
56. Hacking AJ (1986) Economic aspects of biotechnology. Cambridge studies in biotechnology 3.
Cambridge University Press, Cambridge
57. Murtagh JE (1986) Process Biochem. 21(2): 61
58. Keim CR (1983) Enzyme Microb. Technol. 5:103
59. Esser K, Karsch T (1984) Process Biochem. 19(3): 116
60. Greek BF (1987) Chem. Eng. News 65(6): 9
61. Smith N, Corcoran TJ (1981) Wood production energetics: an analysis for fuel applications. In:
Klass DL (ed) Biomass as a nonfossil fuel source. ACS, Washington, DC (Symposium series
No. 144), p 433
62. Ferchak JD, Pye EK (1981) Solar Energy 26:17
63. Datta R (1981) Process Biochem. 16(4): 16
64. Wilke CR, Maiorella B, Sciamanna A, Tangnu K, Wiley D, Wong H (1983) Enzymatic hydrolysis
of cellulose -- theory and applications. Noyes Data Corp, Park Ridge, NJ
65. Grethlein H (1984) Biotech. Adv. 2:43
66. Dale BE (1985) Cellulose pretreatments: technology and techniques. In: Tsoa GT (ed) Annual
reports on fermentation processes, vol 8 p 299
67. Grethlein H (1985) Bio/Technol. 3(2): 155
68. Weimer PJ, Weston WM (1985) Biotechnol. Bioeng. 27:1540
69. Grethlein H, Converse AO (1985) Understanding how pretreatment increases the rate of en-
zymatic hydrolysis of wood. Presented at: 190th meeting of the ACS
70. Wright JD, Power AJ, Douglas LJ (1986) Biotechnol. Bioeng. Symp. 17:285
71. Allen DC, Grethlein HE, Converse AO (1984) Solar Energy 33(2): 175
72. Weimer PJ, Chou Y-CT, Weston WM, Chase DB (1986) Biotechnol. Bioeng. Syrup. 17:5
73. Preprints from: Symposium on the pretreatment of lignocellulosic materials. 23-27 June 1986,
Graz, Austria. Forest Research Institute, Rotura, New Zealand
74. Ladisch MR, Lin KW, Voloch M, Tsoa GT (1983) Enzyme Microb. Technol. 5:82
75. Mardsen WL, Gray PP (1986) CRC Crit. Rev. Biotechnol. 3(3): 235
76. Grethlein HE Acid hydrolysis review. Presented at: Conference on anaerobic digestion and
carbohydrate hydrolysis of wastes. 8-10 May 1984, Luxembourg, Commission of the European
Communities
77. Ladisch MR, Tsoa GT (1986) Enzyme Microb. Technol. 8:66
Production of Ethanol from Lignocellulosic Materials Using Thermophilic Bacteria 49

78. Wright JD, Power AJ (1987) Comparative technicial evaluation of acid hydrolysis processes for
conversion of cellulose to ethanol. In: Klass, DL (ed) Energy from biomass and wastes X.. Else-
vier, Essex
79. Smith PC, Grethlein HE, Converse AO (1982) Solar Energy 28(1): 41
80. Kwarteng K (1983) Kinetics of acid hydrolysis of hardwood in a continuous plug flow reactor.
Ph.D. Thesis, Thayer School of Engineering, Hanover, NH
81. Parker S, Calnon M, Feinberg D, Power A, Weiss L (1983) The value of furfural/ethanol co-
production from acid hydrolysis processes. SERI, Golden. CO; TR-231-2000
82. Esser K, Schmidt U (1982) Process Biochem. 17(3): 46
83. Faust U, Prave P, Schlingmann M (1983) Process Biochem. 18(3): 31
84. Guidoboni GE (1984) Enzyme Microb. Technol. 6:194
85. Maiorella BL, Blanch HW, Wilke CR (1984) Biotechnol. Bioeng. 26:1003
86. Kolot FB (1984) Process Biochem. 19(1): 7
87. Mulder MHV, Smolders CA (1986) Process Biochem. 21(2): 35
88. Hoffman H, Scheper T, Schugerl K, Schmidt W (1987) Chem. Eng. J. (Lausanne) 34:B13
89. Matsumura M, Markl H (1984) Appl. Microbiol. Biotechnol. 20:371
90. Crabbe PG, Tse CW, Munro PA (1986) Biotechnol. Bioeng. 28:939
91. Cysewski GR, Wilke CR (1977) Biotechnol. Bioeng. 19:1125
92. Ghose TK, Roychoudhury PK, Ghose P (1984) Biotechnol. Bioeng. 26:377
93. Rogers PL, Lee K J, Skotnicki ML, Tribe DL (1982) Ethanol production by Zymomonas mobilis
In: Fiechter A (ed) Advances in biochemical engineering/Biotechnology. Springer, Berlin
Heidelberg New York, vol 23 p 37
94. Montencourt BS (1985) Zyrnomonas, a unique genus of bacteria. In: Demain AL, Solomon N
(eds) Biology of industrial microorganisms. Cummings, Menlo Park, p 261
95. Karsch T, Stahl U, Esser K (1983) Eur. J. Appl. Microbiol. Biotechnol. 18:387
96. Magee, RJ, Kosaric N (1985) Bioconversion of hemicellulosics. In: Fiechter A (ed) Advances in
biochemical engineering/Biotechnology. Springer, Berlin Heidelberg, New York, vol 32 p 61
97. Hartline FF (1979) Science 206:41
98. Parkinson G (1981) Chem. Eng. 88(11): 29
99. Choudhury JP, Ghose P, Guha PK (1985) Biotechnol. Bioeng. 27:1081
i00. Essien D, Pyle DL (1983) Process Biochem. 18(4): 31
101. Garg DR, Ausikaitis JP (1983) Chem. Eng. Prog. 79(4): 60
102. Katzen R, Ackley WR, Moon, GD, Messick JR, Brush BF, Kaupisch KF (1981) Low energy
distillation systems. In: Klass DL, Emert GH (eds) Fuels and chemicals from biomass. Ann
Arbor Science, Ann Arbor
103. Busche RM (1984) Biotechnol. Bioeng. Syrup. No. 13:597
104. Barba D, Brandani V, Di Giacomo G (1985) Chem. Eng. Sci. 50(12): 2287
105. Schmitt D, Vogelpohl A (1983) Sep. Sci. Technol. 18(6): 547
106. Lee F-M, Pahl RH (1985) Ind. Eng. Chem. Process Des. Dev. 24:168
107. Lynd LR, Grethlein HE (1984) Chem. Eng. Prog. 81 : 59
108. Grethlein HE, Lynd LH (1986) U.S. Patent No. 4, 626, 321
109. Lynd LR, Grethlein HE (1986) AIChE J. 32(8): 1347
110. Martin SR (1982) Chem. Eng. NY 377:50-53
1l 1. Lyons TP (1983) Proc. Biochem. 18(2): 18
112. Kampen WH (1980) Hydrocarbon Process. 59(2): 72
113. Parker HW (1982) Mech. Eng. 104(5): 54
114. Parisi F (1983) Energy balances for ethanol as a fuel. In: Fiechter A (ed) Advances in biochemical
engineering/Biotechnology. Springer, Berlin Heidelberg New York, vol 28 p 41
115. Pimentel LS (1980) Biotechnol. Bioeng. 22:1989
116. Rothman H, Greenshields R, Calle FR (1983) The alcohol economy: fuel ethanol and the Bra-
zilian experience. Francis Pinter, London
117. Sama DA (1981) Hydrocarbon Process. 60(7): 89
118. Yorifuji T (1981) Energy Dev. Jpn. 3:195
119. Johnson MA (1983) Energy 8(3): 225
120. Krochta JM (1979) Energy analysis for ethanol from biomass. Second international conference
on energy use management. Pergamon, New York, p 1956
50 L.R. Lynd

121. Cooney CL, Mistry FR (1982) Analysis of direct fermentation of lignocellulose to ethanol.
Presented at: 184th meeting of the ACS
122. Huibers DTA, Jones MW (1980) Can. J. Chem. Eng. 58:718
I23. Graft GM (1982)Chem. Eng. NY. 89(26): 25
124. Janshekar H, Fiechter A (1983) Lignin: biosynthesis, application, and biodegradation. In:
Fiechter A (ed) Advances in biochemical engineering/Biotechnology. Springer, Berlin Heidel-
berg New York, vol 27 p 119
125. Clements LD, Beck SR, Heintz C (1983) Chem. Eng. Prog. 79(11): 59
126. Shah RB, Clausen EC, Gaddy JL (1984) Chem. Eng. Prog, 80(1): 76
127. Greek BF (1984) Chem. Eng. News 62(11): 17
128. Zeikus JG, Ben-Bassat ANgTK, Lamed R (1981) Thermophilic ethanol fermentations. In:
Hollaender A (ed) Trends in the biology of fermentations for chemicals and fuels. Plenum,
New York, p 441
129. Sonnleitner B, Cometta S, Fiechter A (1982) Biotechnol. Bioeng. 24:2597
130. Avgerinos GC, Wang DIC (1983) Biotechnol. Bioeng. 25:67
131. Sonnleitner B, Fiechter A, Giovanni F (1984) Appl. Microbiol. Biotechnol. 19:326
132. Leschine SB, Canale-Parola E (1983) Appl. Environ. Microbiol. 46(3): 728
133. Sleat R, Mah RA, Robinson R (1984) Appl. Environ. Microbiol. 48(1): 88
134. Lawford GR, Lavers BH, Good D, Charley R, Fein J, Lawford HG (1982) Zymomonas ethanol
fermentations: biochemistry and bioengineering. Presented at: International symposium on
ethanol from biomass, 13-15 Oct 1982, Winnipeg, p 482
135. Lacis L, Lawford HG (1985) J. Bacteriol. 163(3): 1275
136. Fardeau M-L, Plasse F, Belaich J-P (1980) European J. Appl. Microbiol. 10:133
137. Wiegel J (1982) Experientia 38:151
138. Barras F, Boyer MH, Chambost JP, Chippaux M (1984) Mol. Gen. Genet. 197(3): 513
139. Gilkes NG, Langsford ML, Kilburn DG, Miller RC, Warren RAJ (1984) J. Biol. Chem. 259(16):
10455
140. Kotoujansky A, Diolez A, Boccara M, Bertheau Y, Andro T, Coleno A (1985) EMBO J. 4(3):
781
141. Skipper N, Sutherland M, Davies RW, Kilburn D, Miller RC, Warren A, Wong R (1985)
Science 230(4728): 958
142. Shalita FP, Yablonsky MD, Dooley MM, Bucholz, SE, Kahrs SK, Murphy-Holland K, Eveleigh
DE (1987) Genetic engineering of bacteria for alcohol fuel production. In: Klass D L (ed) Energy
from biomass and wastes X, Elsevier, Essex, p 907
143. Imanaka T (1986) Application of recombinant DNA technology to the production of useful
biomaterials. In: Fiechter A (ed) Advances in biochemical engineering/Biotechnology. Springer,
Berlin Heidelberg New York, vol 33 p 1
144. Guthrie KM (1969) Chem. Eng. NY. 76(6): 114
145. Peters MS, Timmerhaus KD (1969) Plant design and economics for chemical engineers, 2nd
edn, McGraw-Hill, New York
146. Lynd LR (1987) Production of ethariol from lignocellulosic materials using thermophilic bacteria.
DE thesis, Thayer School of Engineering, Hanover, NH
147. Levy PF, Sanderson JE, Ashare E, Wise, DL, Molyneaux MS (1980) Liquid fuels production
from biomass. Report for DOE/SERI contract no. AC02-77ET20050; DOE/ET/20050-T4
148. Bayer EA, Kenig R, Lamed RL (1983) J. Bacteriot. 156(3): 818
149. Ljungdahl LG, Pettersson B, Ericksson KE, Wiegel J. (1983) Curr. Microbiol. 9:195
150. Ng TK, Zeikus JG (1981) Appl. Environ. Microbiol. 42(2): 231
151. Johnson EA, Sakajoh M, Halliwell G, Madia A, Demain AL (1982) Appl. Environ. Microbiol.
43(5): 1125
152. Spinnler HE, gavigne B, Blachere H (1986) Appl. Microbiol. Biotechnol. 23:434
153. Johnson EA, Reese, ET, Demain AL (1982) J. Appl. Biochem. 4:64
154. Johnson EA, Bouchot F, Demain AL (1985) J. Gen. Microbiol. 131:2303
155. Wu D, Demain AL (1986) Abstracts of the annueal meeting of the ASM, p 73
156. Wu D, Demain AL (1985) Abstracts of the annual meeting of the ASM, p 248
157. Hon-Nami K, Coughlan MP, Hon-Nami H, Ljungdahl LG (1986) Arch Microbiol. 145:13
158. Afeyan N (1987) A mechanistic study of the Clostridium thermocellum cellulase system. PhD
Thesis, MIT
Production of Ethanol from Lignocellulosic Materials Using Thermophilic Bacteria 51

159. Lamed RL, Setter E, Bayer EA (1983) J. Bacteriol. 156(2): 828

160. Lamed RL, Kenig R, Setter EA (1985) Enzyme Microb. Technol. 7:37
161. Coughlan MP, Hon-Nami K, Hon-Nami H, Ljungdahl LG, Paulin JJ, Rigsby WE (1985)
Biochem. Biophys. Res. Comm. 130(2):904
162. Bisaria VS, Ghose TK (1981) Enzyme Microb. Technol. 3:90
163. Lynd LR, Grethlein HE (1987)Biotechnol. Bioeng. 29:92
164. Ng TK, Ben-Bassat A, Zeikus JG (1981) Appl. Environ. Microbiol. 41:1337
165. Saddler JN, Chan MK-H (1982) Eur. J. Appl. Microbiol. Biotechnol. 16:99
166. Kundu S, Ghose TK, Mukhopadhyay SN (1983) Biotechnol. Bioeng. 25:1109
167. Khan AW, Asther M, Giuliano C (1984) J. Ferment. Technol. 62(4): 335
i68. Saddler JN, Chan MK-H (1984) Can. J. Microbiol. 30:2123
169. No. 141, and Wolkin, Lynd and Grethlein, manuscript in preparation.
170. Grethlein HE, Allen D C, Converse AO (1984) Biotechnol. Bioeng. 25:1498
171. Knappert D, Grethlein HE, Converse A (198t) Biotechnol. Bioeng. Symp. 11:66
172. Hon-Nami K, Coughlan MP, Hon-Nami H, Carriera LH, Ljungdahl LG (1985) Biotechnol.
Bioeng. Symp. 15:191
173. Slaff GF, Humphrey AE (1981) Diauxic growth of C. thermohydrosulfuricum. Presented at:
182nd meeting of the ACS
174. Carreira LH, Wiegel J, Ljungdahl LG (1983) Biotechnol. Bioeng. Symp. 13:183
175. Ng TK, Zeikus JG (1982) J. Bacteriol. 150(3): 1391
176. Slater GJ, Wakelin WS (1985) Thermophilic ethanol fermentation: an engineering assessment.
NTIS, Springfield, VA; PB85-169142
177. Herrero AA, Gomez RF (1980) Appl. Environ. Microbiol. 40(3): 57i
178. Lovitt RW, Longin R, Zeikus JG (1984) Appl. Environ. Microbiol. 48(1): 171
179. Herrero AA, Gomez RF, Roberts MF (1985) J. Biol. Chem. 260(12): 7442
180. Herrero AA, Gomez RF, Roberts MF (1982) Biochim. Biophys. Acta 693:195 (1982)
181. Curatolo W, Kanodia S, Roberts MF (1983) Biochim. Biophys. Acta 734:336
182. Herrero-Molina AA (1981) The physiology of Clostridium thermocellum in relation to its energy
metabolism. PhD Thesis, MIT, Cambridge
183. Kim S (1982) Microbial production of ethanol by Clostridium thermosaccharolyticum. MS
Thesis. MIT, Cambridge
184. Mistry FR (1986) Ethanol Production by Clostriditon thermosaccharolyticum in a continuous
cell recycle system. PhD Thesis, MIT, Cambridge
185. Sundquist JA, Blanch HW, Wilke CR (1986) Ethanol production with Clostridium thermo-
hydrosulfuricum. Presented at: 192nd meeting of the ACS
186. van Uden N (1985) Ethanol toxicity and ethanol tolerance in yeasts. In: Tsao G (ed) Annual
reports on fermentation processes, vol 8 p 11
187. Lamed R, Zeikus JG (1980) J. Bacteriol. 144(2): 569
188. Hyun HH, Shen G-J, Zeikus JG (1985) J. Bacteriol. 164(3): 1153
189. Thauer RK, Jungermann K, Dekker K (1977) Bacteriol. Rev. 41(1): 100
190. Mistry FR No. 178, and manuscript submitted for publication
191. Krebs H (1969) The role of equilibria in the regulation of metabolism. In: Horeker B L, Stadt-
man ER (eds) Current topics in cellular regulation, Academic, New York, vol 1 p 45
192. Su T, Lamed R, Lobos J, Brennan M, Smith J, Tabor D, Brooks R (1981) Bioconversion of
plant biomass to ethanol. Final report for DOE subcontract no. XR-9-8271-1. SERI, Golden,
CO
193. Weimer PJ, Zeikus JG (1977) Appl. Environ. Microbiol. 33(2): 289
194. Ben-Bassat A, Lamed R, Zeikus JG (1981) J. Bacteriol. 146:192
195. Mistry F, Cooney CL (1985) Ethanol production by Clostridium thermosaccharolyticum in a
continuous culture cell-recycle system. Presented at: 190th meeting of the ACS
!96. Ljungdahl LG, Bryant F, Carriera L, Saiki T, Wiegel J (1981) Some aspects of thermophilic
and extreme thermophilic anaerobic microorganisms. In: Hollaender A (ed) Trends in the
biology of fermentation for chemicals and fuels. Plenum, New York p 397
197. Zeikus JG, Ben-Bassat A, Hegge P (1980) J. Bacteriol. 143:432
i98. Ward PJ, Matharasan R (1986) The effect of controlled redox potential on the growth and ener-
getics of Thermoanaerobacter ethanolicus. Presented at: 192nd meeting of the ACS
199. Wang DIC, Dalal R (1986) U.S. Patent no. 4,568,644
52 L.R. Lynd

200. Avgerinos GC (1982) Direct conversion of cellulosic biomass to ethanol by mixed culture fer-
mentation of Clostridium therrnocellum and Clostridium therrnosaccharolyticum. Ph.D. Thesis,
MIT, Cambridge
201. Rothstein DM (1986) J. Bacteriol. 165(1): 319
202. Bailey JE, Ollis DF (1977) Biochemical Engineering Fundamentals. McGraw-Hill, New York
Advances in Lignocellulosics Hydrolysis
and in the Utilization of the Hydrolyzates

Federico Parisi
I s t i t u t o di S c i e n z a e T e c n o l o g i a d e l l ' I n g e g n e r i a C h i m i c a d e l l ' U n i v e r s i t / t , v i a a l l ' O p e r a
P i a 15, I 1 6 1 0 0 G e n o v a , Italy

1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
2 Nature and Structure o f Lignocellulosics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
3 Acid Hydrolysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
3.1 History . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
3.2 Theory of Acid Hydrolysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
3.3 The State of the Art . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
4 Enzymatic Hydrolysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
4.1 The Trichoderma Route . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
4.2 Other Cellulolytic Microorganisms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
4.3 The Pretreatment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
4.4 Enzyme Production and Hydrolysis Processes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
5 The Utilization of Hydrolyzates and By-products . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
5.1 Ethanol Production . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
5.1.1 Ethanol from Pentoses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
5.2 Furfural Production and Utilization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
5.3 Other Uses of the C 5 Hydrolyzates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
5.4 The Utilization of Lignin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
5.5 Other Utilizations of Hydrolyzates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
6 Ethanol Production Economics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
7 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
8 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80

The debate if acid or enzymatic hydrolysis of lignocellulosics will prevail over the other in the near
future is still open. Different types of acid hydrolysis (use of concentrated acids, or diluted
acids, and in this case use of extreme temperatures, or attempts to realize semi-continuous or
continuous processes) are described. Advantages and inconveniences are described for each case.
However, only a limited margin for improvement is left to acid hydrolysis, compared to the
enzymatic one. New microorganisms, new strains, and genetic engineering are actually improving
classic enzymatic processes. Simultaneous hydrolysis and utilization of produced sugars will sub-
stantially modify current perplexities. A survey of the present trends is given.
In any case, the utilization of lignocellulosics hydrolysis will be of commercial interest only if
hemicellulose hydrolyzates and lignin find profitable employment. Considerable effort is being made
in this direction by both the scientific and the technical community.

Advances in Biochemical Engmeering/

Biotechnology, Vol. 38
Managing Editor:A. Ficchter
r~Springer-Vertag Berlm Heidelberg 1989
54

1 Introduction

The literature on the use of lignocellulosic biomass, especially on the hydrolysis

and biotransformation of cellulose and hemicellulose to ethanol and the utilization
of the products and byproducts, is extremely abundant. This wealth is justified by the
enormous quantity of this biomass available. The available quantity of lignocellulosic
biomass in 1972 was estimated to be one hundred billion (1011) tons per year 11. The
current value may very well be higher.
Many articles from this series are devoted to the theme of exploiting lignocellulosic
biomass 1-20). The specific topics treated are hydrolysis and its products utilization,
with emphasis on the large-scale production of ethanol 21). A complete survey of the
most recent literature is beyond the scope of this article: here we will summarize the
present trends and the progress being made toward industrial-scale applications.

2 Nature and Structure of Lignocellulosics

Lignocellulosic nature and structure have been treated in numerous publications,

including issues in this same series 6,9,10.19) Only the fundamentals are summarized
here.
Lignocellulosic materials are composed of cellulose, hemicellulose, and lignin.
The respective quantities in the different species are indicated in Table 1. Cellulose,

Table 1. Average composition of lignocellulosics

Species Cellulose Hemicelluloses Lignin

Conifers 40-50 O/~o 20-30 o/ 25-35 %

Deciduous trees 40-50 % 30-40 % 15-20 ~
Cane bagasse 40 % 30 % 20 o/
Corn cobs 45 % 35 o/ 15 o/
/o
Corn stalks 35 o/ 25 ~ 35 %
Wheat straw 30 % 50 % 15 o/
/o

a polymer of [3-glucose, has a high degree of polymerization (200 to 2000 kDa) and
crystallinity. Hemicelluloses are polymers of pentoses (xylose, mainly, and arabinose
and ribose), hexoses (as glucose, mannose and galactose), and uronic acids. Lignin
is made up of phenylpropane units, methoxylated and linked in various ways. It acts
as a binder among the fibers in the lignocellulosic materials; thus, wood can be
considered a natural example of composite material.
The high crystallinity of the cellulosic material is the first, fundamental obstacle
to its hydrolysis, whatever the method used. Also, the lignin and hemicellulose
shield the cellulose from enzymatic attack. In order to make an economically com-
petitive process, it is necessary to convert each of these three major components into
saleable products.
Advances in LignocellulosicsHydrolysisand in the Utilization of the Hydrolyzates 55

3 Acid Hydrolysis
3.1 History
Braconnot, in 1812, first attempted to hydrolyze cellulose using concentrated sul-
furic acid. In 1913, Willst/itter hydrolyzed cellulose with fuming hydrochloric acid
and hydrolysis with diluted sulfuric acid was first performed by Simonsen in 1888.
Only at the beginning of the twentieth century, however, was industrial-scale
hydrolysis considered for the production of sugar solutions. A plant using the diluted
sulfuric acid process of Ewen and TomIison zz,~3~ was built in the USA in 1910.
Several generations of process improvement followed, cuhninating in the "Scholler"
24-27) and "Madison" zs) percolation processes. A Scholler-type plant used sulfuric
acid at 0.6 % at a maximum temperature of 184 ~ and produced a solution at about
3 O//oof glucose, with a yield of about 50 % on cellulose. The Madison process had
cellulose conversions of 50 o/and sugar concentrations of 4 to 5 %, under the same
conditions of temperature and concentration of the acid 29)
A commercial scale concentrated acid process using hydrochloric acid, was con-
structed by Bergius in 1925 and described later 30). Research has also been carried
out with concentrated sulfuric acid in Japan, and in the USA, and with hydrofluoric
acid in Germany 31.32)
The primary advantages of acid hydrolysis are its rapid rate and simplicity. Con-
centrated acid hydrolysis has the advantage of high yields (virtually 100%o), but
suffers from high acid consumption and/or recovery costs. Dilute acid hydrolysis is
extremely simple, but has low yields and produces large amounts of degradation
products which inhibit microbial activity_

3.2 Theory of Acid Hydrolysis

The proposed hydrolysis mechanism is shown in Fig. l, where Cell means a long
chain of 13-glucose units 33-36). At the beginning of hydrolysis, the glycosidic oxygen
is quickly protonated. The rate-limiting step is the flexure of the glucose molecule
from the chair to the semiplanar configuration, accompanied by the elimination of the
Cell residue from the glucose unit, The next steps are the rapid addition of water
and the fast regeneration of the proton. The rotational energy required in the ring
flexure seems to be the rate-controlling factor in hydrolysis, and the slow hydrolysis
rate of cellulose is explained by the rigidity of the glucose rings held tightly in the
crystal structure determined by the hydrogen bonding between hydroxyl groups
and hydrogen atoms of adjacent chains. The hydrolysis rates of amorphous cellulose
and of hemicelluloses are much faster because the amorphous nature does not hinder
the flexure of the ring. The hydrolysis of amorphous pentosans and hexosans can
thus be carried out at low temperature and low acid concentration (e.g. 1% concen-
tration and 150 ~ but to break the chemically resistant crystalline cellulose it is
necessary to use either dilute acid with temperatures in excess of 180 ~ or to use
concentrated acids at lower temperatures.
While the hydrolysis of pentosans is rapid, they are also easily decomposed to
furfural and tars. Table 2 shows the values of the rate constants, in minutes, at two
56 F. Parisi

OH OH OH
H~/CH2OH ,o\ o. I

~ fast
/~"c"2~ ~ c"2~

~ -'.| 5 .'o -ce..OH I I

U ,,o. ".o-L,'o "
0 Cell ?\|
Cell H
014 OH
CH2OH
fast ~, ~fas l,
,o

I oH
/
H H
Fig. 1. Hydrolysis of cellulose via the cyclic carbonium-oxide ion [from aT)]

different acid c o n c e n t r a t i o n s a n d at v a r i o u s t e m p e r a t u r e s 33-36) a s s u m i n g t h a t

hydrolysis a n d d e c o m p o s i t i o n are m o d e l e d as first-order h o m o g e n e o u s reactions.
T h e table shows t h a t higher t e m p e r a t u r e s c a n p r o v i d e slightly i m p r o v e d ratios
o f hydrolysis to d e c o m p o s i t i o n .
T a b l e 3 (same references and a s s u m p t i o n s as for T a b l e 2) lists the rate c o n s t a n t s for
the hydrolysis o f cellulose and the glucose d e c o m p o s i t i o n to h y d r o x y m e t h y l f u r f u r a l
w i t h dilute sulfuric acid. It can be seen that, theoretically, the m a x i m u m a d v a n t a g e ,

Table 2. Rate constants of hydrolysis of xylans and of xylose decomposition (K in rain-1)

H2SO 4 1% H2SO4 0.6%

Temp. ~ Kxyl Kdr 0 K~rJKne 0 Kxyl Kdeg Kxy/Kae0

120 2.6 x 10 .2 2.4 • 10 - 3 10.8 1.4 x 10 -2 1.7 x 10 .3 8.2

150 3.2x 10 -1 2.7x 10 .2 11.8 1.8x 10 -1 1.9x 10 .2 9.5
180 2.8 2.3x 10 -1 12.2 1.6 1.6x 10 -1 10.0

Table 3. Rate constants of hydrolysis of cellulose and of glucose decomposition (K in min-1)

H 2 S O 4 1 o/ HzSO,,0.6%

Temp. ~ Kql. Kaeg Kr Kol. Kdeo Kgl./Kdeo

180 1.02• 10 -1 3.34• 10 -1 0.305 6.0• 10 -2 2.49• 10 -1 0.241

210 1.30 1.40 0.929 7.6• 10 -1 1.05 0.723
240 12.10 5.00 2.42 7.14 3.73 1.91
Advances in Lignocellulosics Hydrolysis and in the Utilization of the Hydrolyzates 57

100

1"2s~ 1
1900C 180 ~
8O

6o

o 40

20

10 20 30 40 50 ~Os
Reaction time

Fig. 2. Xylose yield from the acid hydrolysis of xylan as a function of reaction time and temperature
[from37~]

160~
~100

ii 8o

g
o 60
-fi

0
0 10 20 30 40 50 60s
Reaction time

Fig. 3. Acid hydrolysis of cellulose as a function of reaction time and temperature [from 37)]

corresponding to the maximum value of the ratio Koz,/Kaeo, is reached at the highest
temperatures and acid concentrations.
The yield in xylose from xylosan and the yield in glucose from cellulose are given
in Figs. 2 and 3 respectively, at different times and for different temperatures with
an acid concentration of 1 ~ 37)
As can be seen, high yields in dilute acid hydrolysis can be achieved either with
58 F. Pansi

high temperature and/or acid concentration, or by removing the sugars from the
reactor before they have a chance to decompose. This last condition may be realized,
for instance, by using percolation reactors. However, the water used in washing the
sugars from the reactor causes dilute sugar solutions, which result in high capital
and energy costs.
In cellulose hydrolysis by concentrated acids (i.e. concentrated sulfuric acid,
fuming or gaseous HC1, liquid or gaseous HF), the acid disrupts the lattice of
crystalline cellulose by breaking the hydrogen bonds between adjacent cellulose
chains, with evolution of heat. The cellulose becomes amorphous in character and is
easily hydrolyzable at low temperature with high yields and practically no by-
products.

3.3 The State of the Art

Comparative studies of the various systems for acid hydrolysis have compared these
in terms of yield and costs. These studies show the importance of maximizing yield,
and minimizing the formation of sugar degradation products, which may inhibit the
bioconversion of sugars to ethanol 38). In order to accomplish this, one can use
strong acids at moderate temperature, or dilute sulfuric acid at high temperatures or in
percolation type reactors.
With the choice of a strong acid, the problem of its recovery is paramount. The
original Bergius process (with concentrated HC1) has given rise to a series of possible
improvements, aimed at making the process continuous 39) and at making the acid
recovery easier 4~ The process, however, is still basically the same: the acid is
fed in at 4 1 ~ , exits from the hydrolysis at about 30~o, and must then be reconcen-
trated. The hydrolysis takes place at about 35 ~ for 1 h, with a yield of essentially
100 o/ O/ o f biodegradable sugars. Post-hydrolysis with diluted acid may be necessary to
complete the hydrolysis of some oligomers. The high cost of the acid recovery
system in the liquid-phase HC1 hydrolysis process is caused by the high volume of
liquid to be processed, and the extremely corrosive nature of HC1, which requires
the use of expensive alloys as Hastelloy or Monel, glass-lined steel, and graphite
for the heat exchangers.
In order to reduce acid recovery costs, the use of gaseous HC1 has been sug-
gested 42-44). For such a process, wood chips dried to less than 10 % moisture are
impregnated with gaseous HC1 at low temperature. The heat is removed by circulating
cold HCI from a self-refrigerating expansion system. The impregnation is carried
out at high pressure (2000 to 2200 kPa) in a fluidized bed. The self-refrigeration
system acts through compression of the HC1 at the previously stated pressure and
subsequent expansion. Although the cost of such a plant is only two thirds of the cost
of a liquid phase HC1 system, this is offset by higher losses of acid. No plant adopting
this system has been constructed on an industrial scale.
The use of HF in liquid 31,32,45 -49) or gaseous form s0,51) would also have a yield
of essentially 100 ~o. However, because hydrofluoric acid costs five times as much as
hydrochloric acid, even small losses are economically significant. Losses will occur
due to the difficulty of desorbing HF from the carbohydrate produced and by loss
Advances in LignocellulosicsHydrolysisand in the Utilization of the Hydrolyzates 59

as an HF/H20 azeotrope in the recovery section. Moreover, even small losses from
the plant could be extremely dangerous because of the high toxicity of the acid.
The temperature in the hydrolyzer is about 23 ~ and the ratio of HF to ligno-
cellulosic is 2.7 to 1 by weight. The reactor can be carbon steel. Duration of hydrolysis
is about 20 rain. HF is recovered by repeatedly heating and flashing of the products.
The post-hydrolysis, which could be theoretically realized with the H F - - H 2 0 azeo-
trope, must in fact use diluted sulfuric acid to avoid the presence of the fluoride ion in
the hydrolyzate, since this ion is toxic to most microorganisms. The cost of the plant,
which may at first glance seem comparatively modest because of the low cost ma-
terials employed, is actually much higher because of the requirement for perfect
containment of the HF and because a refrigeration plant needs to remove the heat
of reaction.
The use of concentrated sulfuric acid 52-59) which is considerably less expensive
than hydrochloric and hydrofluoric acids ($ 0.09 kg- 1 vs $ 0.26 and $1.35 respectively)
makes recovery less important. The acid is neutralized with lime (a demonstration
recovery plant working with membranes in Japan revealed the cumbersomeness of
the process). Still, the gypsum formed during neutralization is formed in great quan-
tities (e. g. twice the weight of the produced ethanol), and the cost of the acid
employed and the lime necessary to neutralize it ends up being not less than that
expected for a plant based on liquid HCI, but exceeds it by almost 50 0~. Hox~ever,
the capital investment is considerably less than that of a liquid HC1 process (half as
much) because the acid can be handled in fiberglass and plastic equipment. The
processing scheme, as realized in a pilot plant of the Tennessee Valley Authority
(Muscle Shoals, Alabama) 60,61) is shown in Fig. 4. It is worth noting, in particular,
that prehydrolysis of hemicelluloses is carried out with the acidified product from the
cellulose hydrolysis. Solids, once washed, centrifuged, and dried at 85 '-C to 10~i;
moisture, are mixed with sulfuric acid and the hydrolysis of cellulose is carried out

Stover C5+C6 Sugars

LGr,~ H M,x H,st.y

t r

J
r Dewater
1 fi
Acid

~ D ry----'H 2ndHyd "- Lignin

Fig. 4. Simplifiedschemeof hydrolysisof tignocellulosicswith concentrated sulfuricacid (Tennessee

Valley Authority)
60 F. Parisi

in 30 min at 140 ~ This scheme minimizes the formation of inhibitors from hemi-
celluloses.
Alternative processing schemes using high temperature dilute sulfuric acid and
various plug flow designs have been advanced which require acid concentrations of
0.5 to 1.5 ~o and temperatures between 180 and 240 ~ 6z-65). Hydrolysis durations
range from a few minutes to a few seconds, in inverse relation to temperature. If the
reaction is carried out in a single, high temperature step, the pentose sugars are
efficiently converted to furfural. However, glucose yields do not exceed 55 ~/ and
large amounts of degradation products are formed, which inhibit the bioconversion
to ethanol. The process, though, has the economic advantage of treating relatively
high concentrations of lignocellulosics and thus obtaining hydrolyzates with com-
paratively high glucose concentration 66), although there are important mechanical
problems in moving mixtures of lignocellulosics and sulfuric acid with more than
20 O/~odry matter.
Saving the xylose for a subsequent transformation requires the process to be divided
into two steps. The first step is the prehydrolysis of hemicelluloses by dilute sulfuric
acid at moderate temperatures or by acetic acid (1%) or calcium phosphate at 205 ~
67). The second step consists of the conditions normally adopted for a simplified plug
flow process 67 - 73).
For several years now, just because of problems caused by other processes, many
proposals have been made for improvements of the Scholler process to improve
yields and sugar concentrations s5.74-81). The acid concentration has been raised
to 2 ~/o or even 4 o/, and the stages of hydrolysis of hemicelluloses and of cellulose
have been separated, each to be performed at different temperatures (140 to 160 ~
and 160 to 180 ~ and over, respectively). Correspondingly, residence times have
been reduced: for instance, from the 3.5 h of the original Scholler process to 20 to
90 min. In this way, yield rises to 70 to 80 % and glucose concentration may reach
12 to 14% if a suitable recirculation of the hydrolyzate is performed. As a general
rule, the trend is to keep the hydrolyzates of hemicelluloses separated from those of
cellulose, either with the purpose of using the xylose in a different way, or because
the bioconversion of glucose and xylose to ethanol (or to any other product) may be
best carried out separately. A remedy to the aggressivity of dilute sulfuric acid at high
temperatures is the use of Monel linings for the prehydrolysis and the use of Monel
or Zircalloy or tantalum linings for the hydrolysis stage. Shortening residence times
obviously reduces the reactors volume and, consequently, lowers their cost.
A particularly interesting process based on the Scholler process that is under
study is the progressing-batch dilute-acid hydrolysis process, which combines the
simplicity of operation of the Scholler process with the advantages of the counter-
current operation 82-83). The continuous counter-current reactor is attractive on
paper, but when using lignocellulosics, it is a very complicated operation. One
solution is to simulate a continuous process by means of several batch reactors in
series, as shown in Fig. 5. The liquid enters reactor 6, proceeds towards the left,
and exits from reactor 2. The hydrolyzate is flashed to quench the hydrolysis and
degradation reactions. After 15 to 30 min, the reactor stages are changed: fresh
liquid enters reactor 5 and exits from reactor 1, reactor 7 is filled with fresh
lignocellulosic, and residual solids in reactor 6 are discharged. The net effect is
that fresh solids enter the reactor train at the left and spent solids are discharged
 Prehydrolysis Hydrolysis Hot P r o c e s s >
Water
- - 180oc

=_

Dump

0..
Condenser

S t e a m , Furfural

Flash
Tank 0
o

Product, Water, Sugar, Acid

Fig. 5. Progressing batch reactor schematic [from 82)] 0..

62 F. ParlSl

1 0 0 ~ Prehydrolysis: 150~
80~- \ Hydrolysis: 183~
........ ~ 0"7% H2S04

~- 40 /Crystalline

o1~--, i~ i I i-"------4~------4~.
0 20 40 60 80 100 12:0 140
Solids Residence Time(rain)
Solid Entrance Liquid Entrance
Liquid Exit Solid Exit

Fig. 6. Progressing batch reactor. Yields as a function of residence time [from 82)]

after leaving the last hydrolysis reactor on the right. The first reactor in the series
is used as a prehydrolysis reactor, and the remaining four reactors are used as
hydrolysis reactors.
It is calculated that in this way it should be possible to attain sugar yields of 80 ~4
and to increase concentration 20 % over that obtained in a single percolation reactor.
The design conditions for the reactor system are a solid residence time of 130 min
and a liquid residence time of 45 rain (prehydrolysis at 150 ~ for 30 minutes and
hydrolysis at 183 ~C for 100 min) with an acid concentration in the liquid phase of
0.7~0. Roughly, yields as a function of time are expected as reported in Fig. 6.
A true counter-current process is being studied 81), using a reactor commonly
employed by the pulp and paper industry, in which the material fills a chain of small
basins moving counter-current to the acid solution. No final results are available at
present.
Various techniques s4) to overcome inhibitor formation in acid hydrolysis processes
and to increase yields have been proposed. On toxicity for yeast in aerobic and
anaerobic conditions, see also Ltiers et al. 8s). If acid hydrolysis is performed in
separate stages, the first stage contains more pentoses and fewer hexoses; the second
stage contains more hexoses and few pentoses. Despite the presence of pentoses, for
ethanol production no special techniques are generally required for the hydrolyzate
of the second stage, which is usually biodegraded directly by S. cerevisiae,
S. uvarum, or Zymomonas mobilis.
According to Beck and Strickland 8o,86), who have exhaustively researched the
subject, a warm treatment with Ca(OH)2 at ph 5.5 with the addition of sulphite, or
an overliming at pH 10, allows better yields in ethanol production, particularly in
the presence of hemicellulose hydrolyzates. However, designing clean-up treatments
to prepare the hydrolyzates for biotransformation is poorly understood, and is largely
Advances in LignocellulosicsHydrolysisand in the Utilization of the Hydrolyzates 63

a matter of trial and error. Purification procedures may be even more necessary
when processes with immobilized cells 8o, 87. as) or loop bioreactors are used.

4 Enzymatic Hydrolysis
4.1 The Trichoderma Route

Enzymatic hydrolysis has very high yields, because cellulase enzymes only catalyze
hydrolysis reactions, and not sugar degradation reactions. However, relevant research
has been conducted mainly over the past decade, and considerable improvement is
necessary before it is economically competitive. The most important research issues
are: improving the performance and understanding of pretreatment; producing less
expensive and more effective enzymes, and developing hydrolysis processes with
greater yields, product concentrations, and rates.
In 1950, E. T. Reese isolated and described 89) a fungus, Trichoderma viride,
subsequently classified as T. reesei, capable of producing a suitable active cellulase.
Many articles published in this series 1-5.7-1o), concerned in particular with the
nature of the enzyme and the kinetics of enzymatic hydrolysis, should be consulted
for more detailed information. As the literature concerned has already reached
several thousand papers, it is beyond the scope of this work to supply a com-
prehensive list of references that would anyway tend to be incomplete. Reference is
made here only to publications and results that appear to be most pertinent.
The enzyme cellulase system is a mixture of endo-f3-1,4-glucanglucanhydrolases
(EC 3,2,1,4), exo-~-l,4-glucancellobiohydrolases (EC 3,2,1,91), and a [3-glucosidase
(EC 3,2,1,21). The enzyme system has been found in numerous mesophilic and
thermophilic fungi, mesophilic and thermophilic bacteria, and some Actinomycetes.
The first enzyme acts randomly on the interior of the cellulose polymer to generate
new chain ends and is competitively inhibited by end products. The second enzyme
catalyzes the cleavage of a cellobiose unit from the non-reducing end of the cellulose
chain and appears to be end-product inhibited. The last component splits cellobiose
and other oligomers to glucose and is also end-product inhibited by a non-
competitive mechanism 3.9o). Several microorganisms produce also other hydrolytic
enzymes such as endo-l,4-~-xylanase, [3-xylosidase, ~-l,3-arabinosidase, ~-l,2-glucu-
ronosidase, mannanase and acetylesterase 91). Spray-drying methods for cellulase
preservation have been studied: xylanase is destroyed and [3-glucosidase is largely
inactivated by such a process 92)
The composite nature of cellulase and the inhibition phenomena connected to it
explain one of the major trends of present research on enzymatic hydrolysis of
cellulose. Using the wild strain of Trichoderma, glucose product concentrations were
limited to 2.0-2.5 ~o. Such a concentration could be acceptable for acetone-butanol,
but not for industrial ethanol production, as the cost of recovery of ethanol from
a 1% medium is prohibitive and the product inhibition from ethanol is well over
1%. Clearly, the most interesting possibility was to produce a mutant Trichoderma
that could supply a high quantity of [3-glucosidase. Research work with this objective
has been carried out, particularly in the USA at Rutgers University, the US Army
Natick Laboratories (MA) and Genencor; as well as in Europe, in Finland 93) in
64 F. Parisi

1.25

1.00
'S

"•t C30

0.75

with
s u p p l e m e n t a l j3G

0.50 I I I I
50 60 70 80 90 100%
Y i e l d on C e l l u l o s e

Fig. 7. Cost of ethanol if different enzymes are used for cellulose hydrolysis [from 9v~]

France and Denmark (Novo Industri). The productivity of cellulase producing fungi
has been increased by almost two orders of magnitude.
At the same time attempts have been made to increase the specific activity of
the system. The activity is a function of enzyme and substrate composition. As a
general rule, an I.U. is the quantity of enzyme necessary to produce glucose at a rate
of 1 gmol min-1. The activity of endoglucanase is usually measured on carboxy-
methylcellulose and the activity of glucosidase on cellobiose. The most prevalent
assay of the total system is to measure its activity against filter paper (in this
case, one inserts in I.U. the letters F.P.). The specific activity of cellulase has been
increased by only a factor of two.
The addition of [3-glucosidase produced from other sources, e.g. Aspergillus niger
or wentii 94-96), has been tested. Figure 7 shows the variation of ethanol cost
using different enzymes and an enzyme added with 13-glucosidase. The final concen-
tration when [3-glucosidase has been added may reach 8 ~ / o f glucose versus 4.3 %,
which is attainable, for example, with the enzyme Rut C30 alone 97)
[3-Glucosidase is also subject to product inhibition; therefore an excess of it is not
as useful as the continuous removal of glucose, e.g. by bioconversion 98,. Inhibition
from ethanol can be considered less important than that from glucose, even at equal
concentration. Processes in which the hydrolysis and alcoholic "fermentation" are
carried out simultaneously are referred as "simultaneous saccharification and fermen-
tation" (SSF) or "combined hydrolysis and fermentation" (CHF).
Because the optimum temperature for cellulase ranges from 45 to 50 ~ while
that of S. eerevisiae from 30 to 32 ~ a large number of tests have been carried out
with various microorganisms as Sehyzosaccharomyces pombe 98), Candida brassicae
98-101), Candida lusitaniae 1~176 Candida acidothermophilum 1~ Zymomonas
mobilis aos, 106), a thermoresistant mutant of Saccharomyces cerevisiae 103, lOS) and
Brettanomyces spp. 107-109). The most recent conclusions lo9) seem to indicate that
the most thermotolerant microorganisms present greater process difficulties, while
Advances in LignocellulosicsHydrolysisand in the Utilization of the Hydrolyzates 65

several organisms perform well at 36 to 40 ~ Brettanomyces clausenii, which has to be

used with the thermotolerant mutants of S. cerevisiae, allows cellobiose to be exploited
directly, further reducing inhibition.
The production of the enzyme presents several problems. Production on a
suitable lignocellulosic substrate is an aerobic process, the productivity of which is
around 50 IU L - ~ h-~. After a residence time of 13 d, a titre of 15 IU m L - ~ is
reached. The long residence time causes high capital costs 97) and increases the risk
of bioreactor contamination. Hydrolysis requires 3 d, employing 15 to 20 IU of
enzyme per g of solids. After this time, about 70 to 80 % of the cellulose and 100 %
of the xylans are hydrolyzed laO). Several attempts to speed up and increase the
enzyme production have been made, using, for instance, lactose as a carbon source
in a fed-batch mode under permanent sugar limitation. Under these conditions
it is possible to arrive at 22IU mE -193'111'112). Lactose (of whey) is also a
cheaper carbohydrate than relatively pure cellulose (see Sect. 4.4). However, cellulose,
possibly through its hydrolysis intermediates, ~hould normally be added as an
inducer H3-115)

4.2 Other Cellulolytic Microorganisms

Numerous microorganisms have been tested for their ability to produce cellulase.
With no intention to be exhaustive, mention will be made here of fungi such as
Chrysosporium lignorum 116), of Fusarium solani 117-120) and F. oxysporum 121-123),
of Neocallimastix frontalis lz4), of Penicillium funiculosum 125) and P. pinophilum 126,
127), of Chaetomium cellulolyticum 113-a15,a2s), and, among thermophilic fungi,
Tularomyces spp. 129-131), and Sporothricum cellulolyticum and S. cellulophilum 132~
Among mesophilic bacteria, Acetovibrio cellulolyticus and Clostridium cellulolyticum
133,134-); among thermophilic bacteria, Clostridium thermocellum 135-139); among
Actynomycetes, Micromonospora calcae, Pseudonocardia thermophila and Thermo-
monospora spp. ~4o- ~43)
CI. thermocellum is now attracting the most interest. According to Halliwell et al. t44),
this bacterium is endowed with the most powerful cellulasic system. Intensive genetic
engineering work is under way at the Massachusetts Institute of Technology and
elsewhere 139,144-a4-6). This Clostridium works at about 60 ~ and is also capable of
converting glucose and cellobiose (but not pentoses) to ethanol. Therefore, a co-
culture with CI. saccharolyticum has been suggested in order to use the pentoses
simultaneously. The co-culture speeds up the process by eliminating oligosaccharides"
after 48 h, the conversion percentage is approximately three times higher than in a
single culture of CI. thermocellum 147). However, the co-culture is not devoid of
inconveniences, since ethanol, acetic, lactic, and butyric acids form simultaneously.
Some authors have, thus, suggested the co-culture with Cl. thermohydrosulphuri-
cure 13s), which does not produce isobutyric acid, or with Thermoanaerobacter etha-
nolicus and its mutants ~21). The reason for resorting to these thermophilic micro-
organisms is to take the maximum advantage of the thermophilic cellulase from
C1. thermocellum.
A therrnophilic fungus, Talaromyces emersonii, is the object of intensive research
because it is apparently capable of producing a high quantity of 13-glucosidase when
66 F. Parisi

cultivated on lactose. Similarly, work is being carried on Fusarium oxysporum, which

is, however, mesophilic and very sensitive to the inhibitors that may form during a
pretreatment by steam explosion. This fungus supplies a high percentage of xylitol
with ethanol 122). TWO other recently isolated fungi are Penicillium pinophilum,
a mutant of which seems able to give in culture 10 I.U. mL -1, but with a high 13-glu-
cosidase content 126), and secondly Neocallimastixfrontalis, which has yielded (in 72 h
and with an enzyme concentration of 0.370 IU mL-1) 100 % of solubilization versus
the 38 o/~ogiven by T. reesei Rut C-30 under the same conditions 124)
The Actynomycetes group is an interesting group of cellulolytic microorganisms,
which are easier to grow than fungi and whose strain improvement by the application
of recombinant DNA technology is more tenable. Work is under way, as the
most recent references reported above show, but no satisfactory conclusions have been
reached to date.
For the sake of completeness, we must mention an attempt to transform enzymatic
hydrolysis from a heterogeneous catalytic process into a homogeneous process. This
is realized by transforming cellulose into acetylcellulose (16.2 to 16.4% of acetyl
groups) and attacking it with the enzyme of Pestalotiopsis westerdijkii. Enzyme
production from this fungus is completed in 5-8 d; the attack lasts 3 to 5 hours and
can be followed by an inoculum with S. cerevisiae. Only 0.7 g of enzyme per kg of
acetylcellulose is required, versus 15 g of enzyme from T. reesei per kg of cellulose 148)
No comparative evaluation of this process has been made.

4.3 T h e Pretreatment

Lignocellulosics are rather resistant to enzymatic hydrolysis unless a suitable pretreat-

ment is used. The surface area available for the enzyme-substrate interaction will
be influenced by pore size and shielding effect by hemiceUuloses and lignin. The
crystalline structure excludes water molecules as well as any larger molecule and
thus reduces available surface area. For hydrolysis of cellulose to occur, the enzyme
must bind to the surface of the cellulose molecule to catalyze the reaction.
Cellulase enzymes have a molecular weight of 30 to 60 kDa and an ellipsoidal shape
with major and minor dimensions of roughly 30 and 200 A 149.150) Conse-
quently, only 20 % of the pore volume is accessible to cellulase molecules, and the
number of accessible pores can be increased if hemicellulose and lignin are re-
moved.
Pretreatment can be physical or chemical. Physical pretreatments include various
forms of milling, shredding, and mulching. Chemical pretreatments enhance enzymatic
susceptibility by removing the shielding effect of lignin, reducing crystallinity, and
increasing cellulose solubility or swelling. Many of these processes have been
examined in former volumes of this series 6,9, lo). At present, the three pretreatment
processes that attract most attention are organosolv, which takes advantage of the
dissolving properties of water-alcohol (methanol, ethanol) mixtures, steam treatment
at 180 to 240 ~ for 1 to 30 rain 151-155), and dilute acid prehydrolysis.
Steam-explosion is widely described in the literature lS3-155). It should be pointed
out that the vitreous transition temperature is approximately 125 ~ for lignin,
165 ~ for xylans, and 234 ~ for cellulose. It is important to reach this last
 MSW ;>
in
Steam
Holding
Bin
High = SO 2
Pressure (optional)
Ram o
f0
e-

C> Orifice 1
:2
,<
Screw Feeder I N

--...jl S c r ew
Discharge
~/I
"--./1
~--./I

Valve """J o
t
Pretreated MSW o
Fig. 8. The stake pretreatment process (schematic)
U:

o-,
68 F. Parisi

temperature in a short time (45 to 60 s at most) to avoid decomposition phenomena

that become very evident at 260 ~ for cellulose as well.
The Iotech steam-explosion process is a batch process very similar to that for maso-
nite production. An explosion gun is charged with biomass, and steam is injected
until the desired cook temperature is attained (245 to 250 ~ After the required cook-
ing time (about 5 s), the pressure is quickly reduced, explosively discharging the content
in a cyclone. Some Water and degradation products are flashed in the cyclone.
The Stake process is a continuous steam-explosion process in which the lignocellu-
losic is transferred by a screwfeeder to the reactor, where the feedstock is compacted
by a high-pressure ram through an orifice and dropped onto an Archimedean
screw, where high pressure and temperature come from direct contact with steam.
The letdown is accomplished through a ball valve, which causes the lignocellulosic
to explode into fine particles (Fig. 8). Temperatures between 220 and 235 ~
are recommended 156). Lower temperatures and longer residence time (190 to 200 ~
and 10 rain.) are also being investigated.
Recents studies 247,248) have shown that the explosion is not necessary to increase
hydrolysis rates. On this ground and on the continuous extraction of gaseous and
soluble products is based the RApid Steam Hydrolysis (RASH) process 249,250)
Steam treatments release uronic acids and acetyl groups in the form of acetic
acid, which carry out a hydrolysis of hemicelluloses. A complete hydrolysis can be
achieved by introducing small quantities of mineral acids or SO2. Weak prehydrolysis
pretreatment, whatever the chemical or mechanical operation, is useful for two
reasons: it separates the hydrolyzate of hemicelluloses from that of cellulose and
even increases the surface area available to the enzyme by removing the shielding layer
of hemicelluloses 1575
Dilute acid pretreatment is also a prehydrolysis of the hemicellulose-lignin matrix.
In this process, dilute acid is added to the treatment to increase the ratio of
xylan hydrolysis to xylose degradation. While conversions of xylan to xylose are on
the order of 30 to 50~o in steam treatments (with the bulk of the scylan converted
to furfural and degradation products) xylose yields are of the order of 80 to 90 o/ /O
for systems in which dilute sulfuric acid is used 157-16o)
Organosolv pretreatment adds an organic solvent (methanol or ethanol) to the pre-
treatment reactor to dissolve and remove the lignin fraction. In the pretreatment
reactor the internal lignin-hemicellulose bonds are broken and both fractions
solubilized, while cellulose remains as a solid. Many combinations of solvent concen-
tration, acid type and concentration, temperature and time are possible. After
leaving the reactor, the organic fraction is removed by evaporation and recycled to the
reactor. Without an organic fraction in the liquid phase, lignin precipitates and can be
removed by filtration or centrifugation. Thus this process separates the feedstock into
a solid cellulose residue which is easily digestible, a solid lignin which has undergone
few condensation reactions, and a liquid stream containing xylose 161)
Another process, based on treatment with 12 ~ NaOH for 4 h at 80 ~ is suggested
by Cunningham 162). Other processes adopted by the pulp and paper industry are not
suitable for commercial application in this case. Processes for enzymatic solubiliza-
tion of lignin 19.1635 are also very interesting from a scientific point of view but are
actually too expensive for commercial application.
Advances in LignocellulosicsHydrolysisand in the Utilization of the Hydrolyzates 69

4.4 Enzyme Production and Hydrolysis Processes

Here we shall consider T. reesei because more realistic data are available and it has
potential for improvement.
As previously mentioned, enzyme production takes approximately 13 days and
represents the single most costly section of the plant. The best results are obtained
in a fed-batch system 15o) where the substrate is slowly added to the bioreactor,
reaching a final loading of 150 g of biomass per litre. The inoculum is prepared in
a first-stage bioreactor having a working volume of 10% of the volume of the
production stage, which has a working volume of 10 O//oof the daily production. At
the end, the final mixture is filtered through a rotatory filter and stored, The
remaining solids (about 18 g L -1 of mycelium and 7.5 g L -1 of cellulose) are sent
to byproduct recovery (as animal feed) or burnt. Given these conditions up to
100 IU L -1 h -1 can be obtained on steam-exploded agricultural residues and 50 IU
on aspen wood. It is also reported in the literature 1H) that it is possible to raise
productivity to 400 IU L - 1 h - 1 with a particular mutant of T. reesei and lactose as the
carbon source, but with steam-exploded aspen wood or rayon cellulose as inducer.
It seems now that it is possible to use lactose alone and that it could be possible to
use the hemicelluloses hydrolyzate as 80 ~ of the carbon source together with 20 ~o
of rayon pulp as inducer 164).
The production ofcellulase by an SSF ("solid-state fermentation") 165)and a recycle
process with a two phase system and ultrafiltration 1661have been also proposed.
The most important parameters in the hydrolysis section (yield, concentration,
hydrolysis duration, and required enzyme loading) are strictly interrelated, once the
pretreatment and the nature of the enzyme are selected. Yields are higher in more
dilute systems, where inhibition is minimized and where increasing the enzyme loading
can, to a limited extent, overcome inhibition and increase yield and product concen-
tration. Finally, longer reaction times also make higher yield and concentration
possible. However, this is not a simple optimization problem.
When hydrolysis and biological conversion are conducted separately, hydrolysis
is carried out at 45 ~ and pH 5. By adding fresh feed over a number of hours after the
hydrolysis begins, high substrate loadings can be achieved and mixing problems
avoided. Enzyme loadings are in the range of 20 to 50 IU g-1 of cellulose.
Figure 9 97) shows the effect of enzyme loading on the cost of a final product
such as ethanol. From the curve, given the high cost of the enzyme, higher savings
are attained with enzyme loadings of approximately 12 IU g-1 of cellulose. After
this point, it appears that the available sites on the cellulose surface are practically
saturated, and adding more enzyme (e.g. the 20 IU quoted before) is useful only because
it provides more 13-glucosidase.
Recovery of the the enzyme is one method of minimizing the effect of the
high enzyme cost. Physical separation methods such as ultrafiltration are impractical
because of high costs 167) The most promising techniques make use of the high
affinity o f cellulase for cellulose. This can be exploited by a process of counter-
current absorption of the endo- and exo-enzyme components released at the end of
the hydrolysis on fresh feed. As a consequence, the recovered enzyme still needs to
be supplemented with [3-glucosidase. Also, a fraction of the enzyme is degraded during
hydrolysis and another part adheres to lignin and non-hydrolyzed cellulose. The
70 F. Paris1

1.0

T
0.9

uJ
0.8

o
0
J
o.71 I ~ _ _ I
0 5 10 15 20
IFPU ~ solids
L L L .]
8 16 24 32
IFPU g- cellulose

Fig. 9. Effect of enzyme loading on the cost of ethanol [from 97)]

major problem with such processes is that they greatly increase the risk of contamina-
tion. Immobilization of the [3-glucosidase would reduce the amount which needs to be
made up 168-171)
Energy consumption of mixing during enzyme production and hydrolysis has not
been widely discussed in the literature. Mixing to enhance oxygen transfer during
enzyme production is one of the major costs; mixing requirements during hydrolysis
are not well defined, but it appears that the major function of such stirring is to
prevent high local super concentrations which inhibit enzyme activity. The power
input per unit volume in hydrolysis reactors could be quite low.
If ethanol is the desired final product, simultaneous saccharification and biocon-
version ("fermentation") SSF process can alleviate many of the limitations of enzyma-
tic hydrolysis. SSF processes require less enzyme (7 to 15 IU g-1 of cellulose) because
inhibition is reduced. Also final product concentrations are higher (4~o ethanol), and
yields are also improved.

5 The Utilization of Hydrolyzates and By-products

5.1 EthanolProduction
Problems connected with ethanol production from cellulose and various hydrolyzates
were quoted and discussed in the preceding sections and the more pertinent
literature was quoted too. The treatment of the subject will not be repeated or
summarized here. However, see also Sect. 6.
Advances in LignocellulosicsHydrolysisand in the Utilization of the Hydrolyzates 71

5.1.1 Ethanol from Pentoses

Biodegradation of a medium rich in pentoses (as in the first stage hydrolyzate of an
acid hydrolysis process, or the hydrolyzate from a pretreatment process before
enzymatic hydrolysis) is performed either with microorganisms capable of using pen-
toses directly (e.g. Paehysolen tannophilus, Piehia stipitis, Candida shehatae, Clostri-
dium thermosaecharolyticum) or by first transforming xylose into xylulose using
xylulose-isomerase and then S. eerevisiae.
P. tannophilus has been studied in depth 15.16.167,172-177); in the presence of rela-
tively modest quantities of glucose it degrades the glucose rapidly and the xylose
more slowly, as shown in Fig. 10 8o). However, appreciable quantities of xylitol are
simultaneously formed. In anaerobic conditions, from five molecules of xylose, four
molecules of xylitol and one of ethanol are formed; in partially aerobic conditions
(0.025 vvm), 30 g L -1 of xylose yield 10 g L -1 of ethanol and 9 g L -1 of xylitol 9o).
The slow utilization of xylose in anaerobic conditions, at least for ethanol produc-
tion, results from an inbalance in production and consumption of NADH in the
overall conversion of xylose to ethanol. For this reason, Alexander 178), by adding
100 to 200 mM L -1 of acetone, increased ethanol yields from xylose from 29.0 to
between 38.7 and 45.0 % by weight, respectively. Yields in xylitol were correspondingly
reduced from 19.0 to between 16.5 and 6.5 %.
An observation by Beck et al. 8o) concerning P. tannophilus is particularly
interesting. When glucose and xylose are present in equivalent amounts (as when
hydrolysis is performed in a single stage), glucose is consumed in preference to
xylose; xylose and ethanol are then consumed simultaneously (see Fig. 11). This
confirms that it is worth keeping the hydrolyzates C5 and C6 separate.
More recent studies have considered Pichia stipitis and Candida shehatae for
xylose utilisation 179-181) Yields by weight on xylose, which are approximately 30 o/
" /O

by weight with the original strains, attain values of 40 to 43 O/~o181-183) and up to

46 ~/184) in 48 h with selected strains. These values are very close to the theoretical.

40
total sugars

"7
.J 3 0

-6

=~ 2 0

g
m lo
/ " f " ' ~ ethanol
,/
/

0 i [ I i
0 2 4 6 8 10 12d

Fig. 10. Bioconversionto ethanol of sugars of a hemicellulosehydrolyzate with Pachysolen tanno-

philus. The C5/C6 sugars ratio is very high [8o~]
72 F. Parisi

40

\
-~ total sugars
,, 30

-s
~ 2o
xvlose

g
g~ ~o - ...""'"--" ethanol " ~ ~
I. I ~.

0 I I I I I
0 2 4 6 8 10 12d

Fig. 11. Bioconversion to ethanol of sugars of a lignocellulosic hydrolyzate containing about the same
quantity of C 5 and C 6 sugars with Pachysolen tannophilus [80)]

These microorganisms suffer considerably from ethanol inhibition and maximum

ethanol concentrations are of the order of 3 ~ 185). The two microorganisms only
assimilate L-arabinose 186)
Recently, tests on continuous processes have been carried out as well 186-188);
comparison work has been performed by Grba et al. 189), Dellweg et al. ,90) and Rizzi
et al. 191.192)
Despite higher yields, Fusarium oxysporum appears not to be so interesting,
because its working times extend to approximately 6 d 193,194)
Although common yeasts cannot directly give ethanol from xylose, many yeasts,
including S. cerevisiae, can give ethanol from xylulose with high efficiency 195). The
major drawbacks of a process using xylose-isomerase to isomerize xylose to xylulose
and then converting xylulose to ethanol with S. cerev&iae, is the high cost of the
large quantity of enzyme required, and the difference in pH optima between the
yeast (pH 5) and the enzyme (pH 7). Nevertheless a large amount of work is being
carried out in this field, beginning with the overproduction of the E. coli xylose iso-
merase in E. coli and B. subtilis by DNA recombinant techniques and continuing with
cloning and expression of the xylose isomerase gene in yeast 196-2ol)

5.2 Furfural Production and Utilization

Furfural is a byproduct of acid hydrolysis of pentosans and can be easily recovered
and purified. If furfural were to be used as fuel in the plant, the purification phase
could be eliminated.
Current uses of furfural are limited to that as a solvent (especially in the oil
industry), to the production of plastic materials (urea-furfural), and to that of
tetrahydrofuran and furfuryl alcohol.
Studies on other methods of furfural utilization are not new: numerous works
by Reppe and coworkers cover this topic. Many more researchers searched for
Advances in LignocellulosicsHydrolysisand in the Utilizationof the Hydrolyzates 73

methods to obtain from furfural products such as butadiene, styrene, vinylfuran,

adipic acid and adiponitrile, hexamethylenediamine, 1-butanol, lubricants and
plastics. All these processes, some of which also had industrial application, have
been abandoned because of the high cost of furfural. Should furfural become the
by-product of very large-scale production, such as that of ethanol as a fuel, interest
in its use could revive. For its uses, furfural should often be converted into furan
(by elimination of the aldehyde group) and tetrahydrofuran (by subsequent catalytic
hydrogenation).
From tetrahydrofuran it is possible, for instance, to produce 1,3-butadiene by
dehydration-dehydrogenation 203). About 1.8 t of furfural yield 1 t of butadiene,
which is commonly used to produce acrylonitrile-butadiene-styrene (ABS) resins
and styrene-butadiene (SBR) rubber.
Styrene, in turn, can be produced from 1,3-butadiene through a Diels-Alder reac-
tion yielding 4-vinyl cyclohexene, which is subsequently passed to styrene 204)
Approximately 1.9 t of furfural give 1 t of styrene. Styrene can be employed to
produce polystyrene in all its forms, ABS resins and SBR rubber. Attention could
be given to 2-vinyl furan: the furanic homologue of styrene, although its synthesis
still presents some difficulty in practice, could be certainly cheaper than styrene.
Also the production of adipic acid (and hexamethylenediamine) for nylon-66,
through a Reppe synthesis 206,207)and of maleic anhydride 208) for which respectively
1 and 1.4 t of furfural per t are needed, could become interesting with low furfural
costs.

5.3 Other Uses of the C5 Hydrolyzates

Still in the field of C5 hydrolyzates, the production of furanic polyols and xylitol
from xylose is of interest. Furanic polyols are obtained from xylose by reaction
in a hydro-alcoholic medium having active methylenic groups. They can be used both
as intermediate products for fine chemicals and pharmaceutical chemistry and in
the formulation of new polyurethanes presenting high thermal stability.
Xylitol is a non-cariogenic sweetener, the metabolism of which does not require
insulin. However, for this last characteristic, the world consumption is not expected
to exceed 100000 t a -1 209). Xylitol is obtained by catalytic hydrogenation of the
xylose sirup and by its subsequent purification. The microbial reduction encounters
less favour; yet it could be joined with ethanol production (see Sect. 5.1.1).

5.4 The Utilization of Lignin

For the production of ethanol from lignocellulosics to be economically competitive,
a large part of the lignins should be used in higher value applications than as process
fuel.
Much of the experimental work for the chemical utilization of lignin has been
performed with lignins from the paper and pulp industry. Lignin from hydrolytic
processes can be very different, depending on the process chosen. The best lignins
may be those obtained through steam explosion or organosolv pretreatment for
enzymatic hydrolysis 210). The best lignins from acid processes are definitely those
from halogen acid processes.
74 1-. Parl~l

For many applications, see in this series the exhaustive paper by Janshekar and
Fiechter 19)
The production of vanillin and syringic aldehyde by alkaline hydrolysis and
oxydation with air is well known, but syringic aldehyde has no commercial appli-
cations and the vanillin market is not so large compared to the fuel-ethanol market.
The hydrocracking and dealkylation of lignins give a 38 o/yield/o of monophenols
and a 80/jo yield of higher substituted phenolic compounds and catechol 2 1 1 - 2 1 3 )
After dealkylation, the monophenols yield 24% phenol, 13 o/benzene, 22.5 % light,
mostly gaseous hydrocarbons, and 22% of a heavy liquid 2~2). According to
Goheen 214), a hydrocracking process would be interesting only if the yield of
phenols reached 50 o/. A different approach would be to convert the phenols into
methyl aryl and methyl substituted aryl ethers instead of separating the different
phenols. These ethers could be used with good results as gasoline extenders and octane
enhancers 215)
Many other possible uses are described by Janshekar and Fiechter 19). Here we
intend to add only some complementary information.
An interesting suggestion by Hsu and Glaser 2 1 6 - 2 1 8 ) about plastics is that lignin
can be reacted with maleic anhydride and then with propylene oxide, to obtain a
lignin-polyester-polyether that can react with diisocyanates to give polyurethane
foams, cast films, adhesives, and coatings. A negative aspect of the use of lignin
for plastics is the color: a recent US patent 219) obtains lignin of a much less deep
hue.
In addition, lignin can be used as a raw material for phenol-formaldehyde and
urea-formaldehyde resin type resins and adhesives and also as an extender, mainly
for plywood and laminates production [for a general review, see Nimz 22o) and
Gillespie 221)].
Comparisons between Kraft lignin and bioconversion lignin are made by Muller
222.223)

Because alcoholic and phenolic groups are simultaneously present, lignin can be
ethoxylated to form a water-soluble polymer: the amount of ethoxylation and the
choice of the lignin source (steam explosion, organosolv or acid hydrolysis) can
give the desired degree of water solubility 224). This could open to lignin the field of
non-ionic surfactants, with some analogy with those derived from nonylphenol,
which are much more valuable than anionic surfactants as lignosulfonates, even
though they are marketed in smaller amounts.
Finally, an interesting use of lignin from hydrolysis processes could be that of the
reinforcement of rubber in place of carbon black 225-227).

5.5 Other Utilizations of Hydrolyzates

Acetone-butanol production by anaerobic conversion of sugars is very well known.
The same microorganism (Clostridium acetobutylicum) can utilize glucose and xylose
and some work is still being done, particularly in France and in the U.S. to improve
such a process. The concentration of resulting solvents (acetone-butanol-ethanol)
cannot exceed 16 to 18 g L-~ because of the inhibitory effect of butanol. Some genetic
engineering work has allowed conversions to reach 23 to 24 g L -1. This value,
however, is too low for this process to be commercially acceptable. References 228-238)
Advances in Lignocellulosics Hydrolysis and in the Utilization of the Hydrolyzates 75

supply recent information on this topic. For the utilization of xylose in particular,
see 18). Soni 2397 gives information about the use of C1. saccharoperbutylacetonicmn.
Production of single cell protein (SCP) is also well known: Schneider 24o) fully
describes production from pentoses. Lastly, the production of 2,3-butanediol by
Klebsiella pneumoniae from pentoses could be of some interest and is described by
Jansen and Tsao in this collection 17)
Possibly, the production of biopolymers can be realized using lignocellulosic hydro-
lyzates with Alcaligenes eutrophus and/or other microorganisms, but there is, at
present, no information on the matter.

6 Ethanol Production Economics

The cost of ethanol production by the various acid processes discussed earlier is
shown in Table 4. This table assumes that only the glucose is converted to ethanol.
The most important observation is the similarity between the two dilute acid hydro-
lysis processes (plug flow and progressing batch) and the three concentrated acid
hydrolysis processes (concentrated sulfuric, hydrochloric and hydrofluoric acid).
The dilute acid processes have high feedstock costs because of the yield losses due to
sugar degradation, but have low acid and base costs (because sulfuric acid is in-
expensive, and only small amounts are used). Utility costs are low because the lignin
and xylose, which are not utilized for ethanol production, are burned and provide
a surplus of steam and electricity. Capital charges are high because a large and ex-
pensive processing plant is needed to convert the very large quantities of lignocellulosic
(large quantities are necessary because of the low yield). Also important is the
difference in selling price between the case in which furfural is sold as a chemical,
and that in which furfural is burned at its fuel value to produce process heat. This
points out the necessity of high process yields (fraction of the total feedstock converted
to saleable product). Similar reductions in selling price would be achieved if the xylan
fraction were reacted to xylose and converted to ethanol.
The concentrated acid processes also show similarities to each other. Because they
convert virtually all the cellulose to utilizable sugars, their feedstock costs are lower.

Table 4. Data on production of ethanol from lignocellulosics

Hardwood consumption Process Operating costs +

tt 1 Capital charges $ t-

7.6 Percolation 220

6.1 Progressing batch 181
8.2 Plug flow 204
8.2 id. (furfural sold) 33
5.3 HzSO4 conc. 330
5.4 HC1 liquid 302
4.9 HF liquid 326
4.6 Enzymatic SHF (*) 450
3.2 Enzymatic SSF (*) ~ 250

(*)-Hemicellulose used for ethanol production

 M o s t like ly e s t i m a t e of lignin Lignin market by value
m ark e t volume in the year 2000 in in m i l l i o n s of 1987 dollars
million t o n s

'Carbon
Bla'ck ~
~,...:!~i:A,o.o,t ~ ! - ~ ; : ~
":::.:- ,oo , ---:'..'.:

Fig. 12. Projected lignin market in the U.S, for the year 2000 [from 21oj]
Advances in LignocellulosicsHydrolysisand in the Utilization of the Hydrolyzates 77

However, because they use expensive acids such as HC1 or HF (some of which is
inevitably lost due to neutralization by ash, or losses from the system) or because they
use large amounts of an inexpensive acid ( H 2 8 0 4 ) , the costs for acid consumption
are quite high. The capital costs are similar to those for dilute acid processes. Although
the overall plant size for concentrated acid, processes is smaller (higher yields mean
less material has to be processed), the capital charges are still very large (due to the
large expense involved in acid recovery or drying the biomass prior to the reaction).
The cost of ethanol production by these processes is quite high, to the order of
$ 0.43 to $ 0.53 kg -1 ($1.30 to 1.60 US gallon -1) when xylose is not converted to a
saleable product. Although some opportunities exist for improving the basic acid
hydrolysis technology, these opportunities are limited because these processes have
received years of development. The most important improvement which can be
achieved in these systems is to introduce a xylose utilization process (e.g. for ethanol
production). This will decrease the feedstock costs, dramatically reduce the pro-
cessing plant size and capital investment, and decrease the amount of energy and labor
needed to produce a given amount of ethanol. Further improvements would be
possible if the lignin were also used beneficially (Fig. 12). Opportunities for this are
greatest with the halogen acid processes (which do not greatly modify the lignin),
somewhat less with the high temperature dilute acid process, and lowest with the
progressing batch and concentrated sulfuric acid processes which extensively modify
the lignin. For the economic analysis of xylose utilization see 251)
Cost of production summaries for the separate hydrolysis and bioconversion (SHF)
and simultaneous saccharification and bioconversion (SSF) processes are shown in
Figs. 13 and 14. In the SHF the major costs are feedstock, energy and capital charges,
enzyme production, and the offsite systems for environmental control and utility
(steam and electricity) production. The feedstock shown in near the lower limit, as
it accounts for only the sugars actually converted to ethanol. A hidden feedstock cost
is that of energy. Approximately $ 0.20 -1 kg ($ 0.60 US gallon -~) is converted into

30
Co8t of : [ ] Energy [ ] Operation
25
"T I~J Feedstock I~ Capital
-~ 20
Total Production Cost 0.905kg -1
15
o

uJ

!fN ~.~_ (~ .~_ g

,~
g
7
7
a .-

2 "~

Fig. 13. Breakdown of ethanol production costs by process area for the SHF process [from977]
78 F. Parisi

20! Cost of: []Capital [ ] Operation

[ ] Energy []Feedstock
15- Total Production
o

10-
N Cost : 0 . 605 kg-1

,7, 5
I
[

o
-~ E o:
o o E.s
69
= ~ =
,~ ~ o
s w

Fig. 14. Breakdown of ethanol production costs by process area for the SSF process [from 24m]

steam and electricity to run the process instead of into useful products. This is a con-
sequence of the low process yield (only 70~ of the cellulose, and none of the xylose
or lignin are converted into saleable products with current technology). Capital
charges are high, because a large processing plant is necessary to convert large volumes
at low yields, and because of the large equipment sizes needed to process the dilute
product streams coming from the hydrolysis reactor. Enzyme production costs are
high because large amounts of enzyme are used to help overcome the effects of end
product inhibition by cellobiose and glucose. Finally, the cost of cleaning up the waste
streams and generating the steam and electricity is high because of the large volumes
which must be processed (anything which is not converted into a saleable product
must be processed either through the anaerobic digestion section or the boiler/steam
generator).
The cost of production for SSF process is considerably lower. All cost reductions
have their root in the reduction of end product inhibition due to the continuous
removal of glucose and cellobiose by bioconversion. First, the yield is improved
because the reaction is not stopped by the build up of sugars. Approximately 90/~
of the cellulose is hydrolyzed in a SSF process vs 70 ~ in SHF. Also, product concen-
trations are roughly twice as high in SSF. This reduces the size, capital cost, and energy
consumption of the downstream process by a factor of almost two. Energy costs are
also reduced because considerably more ethanol is produced while the amount of
material processed is reduced. The enzyme production costs are reduced by almost
a factor of five, because enzyme loading can be reduced to one seventh of that
necessary when inhibition by glucose must be overcome. Finally, because less of the
initial feed must be processed in the waste treatment and boiler plants, the cost of
these auxiliary sections are reduced. Taken together, these process improvements
reduce the predicted price of ethanol from 0.90 to $ 0.60 kg 1 (from 2.70 to $1.80 US
gallon 1). Unlike acid hydrolysis, the technology of enzymatic hydrolysis is relatively
new, and there is considerable room for improvement in each of the process areas.
The interaction between the pretreatment and hydrolysis rate is still not completely
Advances in Lignocellulosics Hydrolysis and in the Utilization of the Hydrolyzates 79

understood. Better understanding of this relationship should allow higher hydrolysis

rates to be achieved, as rates for the cellulose hydrolysis are still approximately two
orders of magnitude below those of starch hydrolysis. Also, current pretreatment
methods degrade up 50 % of the C 5 fraction. This loss directly reduces the potential
for ethanol production by xylose bioconversion and also produces degradation pro-
ducts which inhibit biotransformation.
Enzyme production shows considerable room for improving the process. Current
designs use solid substrates, or substrates with limited supplies such as lactose.
Research is needed to produce enzymes at high rates from lignocellulosic derived
sugars such as glucose or xylose. A factor of three improvement in productivity should
be achievable with suitable mutants. Similarly, cellulase is, as it was said, a mixture
of three different types of individual enzymes and the total activity of the enzyme is
dependent on the relative proportions of the three components in the mixture. It is
not yet certain what proportions are correct for an enzyme mixture used in biomass
processing, although it appears that mixtures with greater quantities of f3-glucosidase
are preferred. Optimization of these ratios should lower the total amount of
enzyme which is needed, and improve the rate and yield of the reaction. The hydrolysis
reaction can be improved by better matching the reaction conditions to those needed
by the enzyme and yeast. Also the amount of energy needed to stir the reactor may
be considerably less than is currently assumed. Continuous removal of the ethanol
product should reduce inhibition of both the enzyme and the yeast, improving both
rate and yield.
Finally, as in the case of acid hydrolysis, use of the other two major fractions
(hemicellulose and lignin) would bring about the greatest reductions in ethanol cost.
The cumulative effect of these improvements is shown in Fig. 15.

1.00 -

0.75
"5

0.50

0.25
I
o
SHF SHF SSF SSF SSF
Rut C30 Genencor and Xylose Lignin
[] Feedstock [ ] SHF, SSF Fermentation and
Credit Xylose
[ ] Enzyme Production [ ] Offsites Credit
9 Distillation [ ] Pretreatment

Fig. 15. Comparison of costs by process area for SHF and SSF process and cost reduction of SSF
process if xylose and or lignin are sold [from 246)]
80 F. Parisi

7 Conclusions
Biological processes for the conversion of lignocellulosics to ethanol are attractive
because of their potentially high efficiency. However, this means that the process must
be configured to use all the major fractions o f the feedstock: cellulose, hemicelluloses,
lignin. The cellulose and hemicelluloses can be broken down to sugars by either acid
or enzymatic hydrolysis processes. Acid processes are more developed, but enzymatic
processes have the greater potential for improved performance. In the past decade,
xylose has gone from being regarded as recalcitrant to ethanol to a point where 70 %
of the theoretical yield can be achieved at reasonable ethanol concentrations. Finally,
methods have been identified for the conversion of lignins to liquid fuels.
Using advanced enzymatic hydrolysis processes in conjunction with xylose bio-
conversion, ethanol cost of approximately $ 0.33 to 0.34 kg -1 ($1.00 US gallon -1)
should be achievable. Addition of lignin processing steps should further reduce this
cost.

This report could not have been prepared without the invaluable advice of John
D. Wright, of the Solar Energy Research Institute, Golden, Colorado. His help is
gratefully acknowledged.

8 References
1. Reese, ET, Mandels M, Weiss AH (1972) in: Ghose TK, Fiechter A (eds) Springer, Berlin
Heidelberg New York, p 181 (Advances in biochemical engineering, vol 2)
2. Ghose TK, Das K (1972) in: Ghose TK, Fiechter A (eds) Springer, Berlin Heidelberg
New York, p 55 (Advances in biochemical Engineering, vol 1)
3. Enari TM, Markkanen P (1977) in: Ghose TK, Fiechter A (eds) Springer, Berlin Heidelberg
New York, p 1 (Advances in biochemical engineering, vol 5)
4. Linko M (1977) in: Ghose TK, Fiechter A (eds) Springer, Berlin Heidelberg New York,
p 25 (Advances in biochemical engineering, vol 5)
5. Ghose TK (1977) in: Ghose TK, Fiechter A (eds) Springer, Berlin Heidelberg New York, p 39
(Advances in biochemical engineering, vol 6)
6. Fan LT, Lee Y-H, Beardmore DH (1980) in: Fiechter A (ed) Springer, Berlin Heidelberg
New York, p 101 (Advances in biochemical engineering, vol 14)
7. Lee Y-H, Fan LT (1980) in: Fiechter A (ed) Springer, Berlin Heidelberg New York, p 101
(Advances in biochemical engineering, vol 17)
8. Lee Y-H, Fan LT, Fan LS (1980) in: Fiechter A (ed) Springer, Berlin Heidelberg New York,
p 131 (Advances in biochemical Engineering, vol 17)
9. Chang MM, Chou TYC, Tsao GT (1981) in: Fiechter A (ed) Springer, Berlin Heidelberg
New York, p 15 (Advances in biochemical engineering, vol 20)
10. Fan LT, Lee Y-H, Gharpuray MM (1982) in: Fiechter A (ed) Springer, Berlin Heidelberg
New York, p 155 (Advances in biochemical engineering, vol 23)
11. Gong Ch-Sh et al. (1981) in: Fiechter A (ed) Springer, Berlin Heidelberg New York, p 93
(Advances in biochemical engineering, vol 20)
12. Kosaric N, Duvniak Z, Stewart GG (1981) in: Fiechter A (ed) Springer, Berlin Heidelberg
New York, p 119 (Advances in biochemical engineering, vol 20)
13. Jeffries TW (1983) in: Fiechter A (ed) Springer, Berlin Heidelberg New York, p 1 (Advances
in biochemical engineering/biotechnology, vol 27)
14. McCracken LD, Gong Ch-Sh (1983) in: Fiechter A (ed) Springer, Berlin Heidelberg New York,
p 33 (Advances in biochemical engineering/biotechnology, vol 27)
15. Schneider H et al. (1983) in: Fiechter A (ed) Springer, Berlin Heidelberg New York, p 57
(Advances in biochemical engineering/biotechnology, vol 27)
Advances in Lignocellulosics Hydrolysis and in the Utilization of the Hydrolyzates 8|

16. Kurtzmann CP (1983) in: Fiechter A (ed) Springer, Berlin Heidelberg New York, p. 73 (Advances
in biochemical engineering/biotechnology,vo127)
17. Jansen NB, Tsao GT (1983) in: Fiechter A (ed) Springer, Berlin Heidelberg New York, p 85
(Advances in biochemical engineering/biotechnology,vol 27)
18. Volesky B, Szczesny T (1983) in: Fiechter A (ed) Springer, Berlin Heidelberg New York,
p 101 (Advances in biochemical engineering/biotechnology, vol 27)
19. Janshekar H, Fiechter A (1983) in: Fiechter A (ed) Springer, Berlin Heidelberg New York,
p 119 (Advances in biochemical engineering/biotechnology,vol 27)
20. Magee RJ, Kosaric N (1985) in: Fiechter A (ed) Springer, Berlin Heidelberg New York, p 61
(Advances in biochemical engineering/biotechnology,vol 32)
21. Franzidis JP, Porteous A (1981) in: Klass DL, Emert GH (eds) Fuels from biomass and
wastes. Ann Arbor Science, Kent, p 267
22. Demuth R von (1913) Z. angew. Chem. 26:786
23. Kressman FW (1922) US Dpt of Agriculture Bull. 983
24. Scholler H US Patents 1 641 771 (1927); 2 083 347 (1934); 2 083 348 (1934); 2 188 192 (1937);
2 188 193 (1937)
25. Scholler H Brevets Fran~ais 706 678 (1930); 799 358 (1936)
26. Scholler H (1936) Chem.-Ztg 60:293
27. Scholler H (1939) DRP 676 967
28. Harris EE, Beglinger E (1946) Ind. Eng. Chem. 38 : 890
29. Harris EE et al. (1945) Ind. Eng. Chem. 37:12
30. Bergius F (1937) Ind. Eng. Chem. 29:247
31. Fredenhangen K, Helferich B (1927) DRP 560 535 to IG Farbenindustrie
32. Helferich B, B6ttger S (1929) Liebigs Ann. Chem. 476:510
33. Saeman JF (1945) Ind. Eng. Chem. 37:53
34. Root DF, Saeman JF, Harris JH (1959) Forest Products J. 9(5): 158
35. Fagan RD et al. (1971) Environm. Sc. Technol. 5:545
36. Grethlein HE (1983) Acid hydrolysis of cellulosic biomass. A Progress Report. SERI Sub-
contract No XX-2-02-02152-1. NH: Dartmouth College, Hanover USA
37. Wright JD, d'Agincourt CG (1984) Evaluation of sulfuric acid hydrolysis processes for alcohol
fuel production. SERI/TR-231-2074. Solar Energy Research Institute. Golden CO
38. Wright JD, Power AJ (1986) Comparative technical evaluation of acid hydrolysis processes
for conversion of cellulose to alcohol. Presented at the Conference on Energy from Biomass,
Institute of Gas Technology, Washington DC
39. Regnault A e t al. (1978) DRP 2 814 067 to Battelle Memorial Institute
40. Foster AV, Martz LE, Leng DE (1980) US Patent 4 237 110 to Dow Chemical Co
41. Sklarewitz ML, Goldstein IS (1983) Recycle of hydrochloric acid in a wood hydrolysis plant
by membrane technology. Presented at the Interamerican Congress of Chem. Eng., Santiago,
Chile
42. Kusama J (1979) Chemical Econom. and Eng. Rev. 11(6): 32
43. Antonoplis RA, Blanch HW, Wilke CR (1983) High pressure HC1 conversion of cellulose to
glucose. LBL-1422 Lawrence Berkeley Labs, Berkeley CA. Available from NTIS, Springfield
VA
44. Antonoplis RA et al. (1983) Biotechnol. Bioeng. 25:2757
45. Defaye D, Gadelle G, Pedersen C (1981) Degradation of cellulose with hydrogen fluoride. In:
Palz W, Chartier P, Hall DO (eds) Energy from biomass. Proceedings of the 1st ECC Con-
ference London, Applied Science, London, p 292
46. Bentsen T (1983) Hydrolysis of carbohydrates in straw using hydrogen fluoride pretreatment.
In: Strub A, Chartier P, Schleser J (eds) Energy from biomass. Proceedings of the 2nd ECC
Conference Berlin, p 1103. Applied Science, London
47. Downey K et al (1983) HF hydrolysis of wood for ethanol production. Presented at Industrial
Energy Forum 83, September 19, Nashville TN
48. Fredenhangen K, Cadenbach G (1933) Angew. Chem. 46:113
48. Selke SM, Hawley M, Lamport DTA (1983) Wood and Agricultural Residues 329
50. Franz R et al. (1983) Lignocellulose saccharification by HF. In: Strub A, Chattier P, Schleser J
(eds) Energy from biomass. Proceedings of the 2nd ECC Conference, Berlin, Applied Science,
London, p 873
82 F. Parlsi

51. Ostrovski CM, Aitken J, Free D (1984) New developments in fuel ethanol production by
gaseous anhydrous hydrofluoric acid hydrolysis of hardwood. Presented at Bioenergy 84,
Gathenburg Sweden
52. Dunning JW, Lathrop EC (1945) Ind. Eng. Chem. 37:24
53. Dunning JW, Lathrop EC (1948) US Patent 2 450 586
54. US Dpt of Interior (1951) Liquid fuels from agricultural residues. Rpt of Investigations 4772-
Part III. US Dpt of Interior, Washington DC
55. Wenzl HFJ (1970) in: The chemical technology &wood, Academic New York, p. 32
56. Ackerson M, Ziobro M, Gaddy GL (1981) Biotechnol. Bioeng. Symp. 11:103
57. Tsao GT et al. (1982) Process Biochem. (IX-X): 34
58. Wright JD (1983) Design and evaluation of low-temperature, concentrated acid hydrolysis
process. SERI/TR-231-1913. Solar Energy Research Institute, Golden, CO
59. Barrier JW et al. (w.d.) Experimental production of ethanol from agricultural cellulosic
materials using low temperature acid hydrolysis. Tennessee Valley Authority, Muscle Shoals
AL
60. Badger PC et al. (1984) in: Preprints of the 6th International Symposium on Alcohol Fuel
Technology, Ottawa, Canada, vol 2 p 100
61. Farina GE, Barrier JW, Forsythe ML (1986) in: Proceedings of the 7th International Symposium
on Alcohol Fuel Technology, Technip, Paris, p 44
62. Brenner W e t al. (1979) Radiat. Phys. Chem. 14:299
63. Thompson DR, Grethlein HE (1979) Ind. Eng. Chem. Prod. Res. Dev. 18:166
64. Church JA, Wooldridge D (1981) Ind. Eng. Chem. Prod. Res. Dev. 20:378
65. Teng KF, Mutharasan R, Grossmann ED (1983) Paper presented at the AIChE Annual
Meeting Oct. 30-Nov. 3, Washington DC
66. Grethlein HE (1978) Biotechnol. Bioeng. 20:503
67. Sharman DK (1984) Two-step process for the selective production of fermentable sugars and
ethanol from biomass residues. In: Prepr. of the 6th International Symposium on Alcohol Fuel
Technology, Ottawa, Canada, vol. 2 p 205
68. Horwath J, Mutharasan R, Grossmann ED (1983) Biotechnol. Bioeng. 25:19
69. Wright JD (1983) High temperature acid hydrolysis of cellulose for alcohol production. SERI/
TP-231-2058. Solar Energy Research Institute, Golden CO
70. Badger Engineers Inc (1984) Economic feasibility study of an acid based ethanol plant. SERI
Subcontract No ZX-3-030-96-2. Badger Eng. Inc., 1 Broadway, Cambridge MA
71. Ritter FJ (1984) US Patent 427453
72. Singh A, Das K, Sharma DK (1984) Ind. Eng. Chem. Prod. Res. Dev. 23:257
73. Harris JF et al. (1985) Two-stage sulfuric acid hydrolysis of wood. USDA Forest Products
Labs, Madison WI
74. Hokanson AE, Katzen R (1978) Chem. Eng. Progr. 74:67
75. Mendelsohn HR, Wettstein P (1981) Chem. Eng. 62
76. Burton RJ (1982) The New Zealand wood hydrolysis process. In: Preprints of the Ethanol
from Biomass Conference, Winnipeg Canada
77. Mackie K, Deverell K, Callander I (1982) Aspects of wood hydrolysis via the dilute sulfuric
acid process. Preprints of the Ethanol from Biomass Conference, Winnipeg Canada
78. Uprichard JM, Burton RJ (1982) Ethanol from wood. Preprints of the 5th International Sym-
posium on Alcohol Fuel Technology, Auckland New Zealand, p 317
79. Carvalho Neto CC, Kling SH (1983) Pilot plant acid studies using Eucalyptus paniculata.
Presented at the Interamerican Congress of Chem. Eng., Santiago Chile
80. Beck M J, Strickland RC (1984) Biomass 6:101
81. Raymond B (1986) in: Proceedings of the 7th International Symposium on Alcohol Fuel
Technology, Technip, Paris, p 39
82. Wright JD, Bergeron PW0 Wendene PJ (1985) The progressing batch hydrolysis reactor.
SERI/TP-232-2803. Solar Energy Research Institute, Golden CO
83. Bergeron PW, Wright JD, Werdene PJ (1986) Biotechnol. Bioeng. Syrup. 17:33
84. Leonard RH, Hainj GJ (1945) Ind. Eng. Chem. 37:390
85. Liiers H et al. (1937) Z. Spiritusind. 60:7
86. Strickland RC, Beck MJ (1985) Effective pretreatments and neutralization methods for ethanol
production from acid-catalyzed hardwood hydrolysates using Pachysolen tannophilus. Presented
at the 9th Symposium on Energy from Biomass and Wood Wastes, Lake Buena Vista FL
Advances in Lignocellulosics Hydrolysis and in the Utilization of the Hydrolyzates 83

87. Converti A e t al. (1985) Biotechnol. Bioeng. 27:1108

88. Del Borghi M et al (1985) Biotechnol. Bioeng. 27:761
89. Reese ET, Sin RGH, Levinson HS (1950) J. Bacteriol. 59:485
90. Bruinenberg PM et al. (1983) J. Appl. Microbiol. Biotechnol. 18:278
9l. Poutanen K, Pouls J, Linko M (1987) in: Proceedings of the 4th European Congress on Bio-
technology, Elsevier, Amsterdam, vol 3 p 90
92. Himmel M (1986) Biotechnol. Bioeng. Syrup. 17:413
93. Bailey MJ. Nevalainen KMH (1981) Enz. Microbiol. Technol. 3:153
94. Ghose TK, Ghosh P (1978) J. Appl. Chem. Biotechnol. 28:309
95. Woodward J, Wiseman A (1982) Enz. Microbiol. Technol. 4:73
96. Alfani F et al. (1983) Biodegradation of native cellulose. In: Palz W, Coombs J, Hall DO (eds)
Energy from biomass. Proceedings of the 3rd ECC Conference Venice, Elsevier, Amsterdam,
p 989
97. Wright JD, Power AJ, Douglas LJ (1986) Biotechnol. Bioeng. Symp. 17:285
98. Takagi M et al. (1977) A method for production of alcohol directly from cellulose using
cellulase and yeast. In: Proceedings of the Bioconversion Symposium Delhi, ITT, New Delhi,
p 551
99~ Rivers DB, Emert GH (1980) in: Preprints of the Bioenergy 80 World Congress and Exposi-
tion Atlanta GA, p 157
!00. Pemberton MS, Brown RD Jr, Emert GH (1980) Can. J. Chem. Eng. 58:273
101. Becker DK, Blotkamp PJ, Emert GH (1981) Pilot scale conversion of cellulose to ethanol.
In: Klass DL, Emert GH (eds) Ann Arbor Science, Ann Arbor MI, p 375 (Fuels from Biomass
and Wastes)
102. Gonde P e t al. (1984) Appl. Environm. Microbiol. 48:265
103. Lastick SM (1984) Simultaneous saccharification and fermentation of cellulose. In: Biotech 84,
p 277
104. Ghosh P (1984) Biotechnol. Bioeng. 26:377
105. Spang!er DJ, Emert GH (1986) Biotechnol. Bioeng. 28:115
106. Favela Torres E, Baratti JC (1987) Appl. Microbiol. Biotechnol. 27:121
107. Gonde P e t al. (1984) in: Proceedings of the 3rd European Congress on Biotechnology, Verlag
Chemic, Weinheim, vol 2 p 15
108. Wyman CE et al. (1986) Biotechnol. Bioeng. Syrup. 17:221
109. Spindler DD et al. (1987) Thermotolerant yeast for simultaneous saccharification and fer-
mentation of cellulose to ethanol. Presented at the 9th Symposium on Biotechnology for Fuels
and Chemicals, Boulder Colorado. To be published in Biotechnol. Bioeng. Symp.
110~ Dekker RFH, Wallis AF (1983) Biotechnol. Bioeng. 25:3027
111. Foody E (w. d.) Scale-up testing of Iotech's Enzyme Production Process using chemical grade
feedstock and the RL-P37 Organism. Contractor's Final Report to Energy, Mines and
Resources, Canada, Contract No 0250.23216-3-6264
112. Pourqui6 J (1987) Lignocellulosic biomass valorisation. Scale-up of the enzymatic process.
In: Grassi G et al. (eds) Biomass for energy and industry. Proceedings of the 4th ECC Con-
ference Orl6ans. Elseyier, London, p 341
113. F~ihnrich P, Irrgang K (1981) Biotechnol. Lett. 3:201
I14. F/ihnrich P. lrrgang K (1982) Biotechnol. Lett. 4: 519, 775
115. Leisola MSA et al. (1985) Biotechnol. Bioeng. 27:1389
116. Eriksson KE, Petterson B (1971) Biodeterioration of materials. Applied Science, London,
vol 2 p 116
117. Wood TM (1969) Biochem. J. 113:457
118. Wood TM, McCrae SI (1977) Carbohydr. Res. 57:117
119. Madan N, Sood P (1979) Microbiol. Lett. 12:109
120. Targonski Z, Szajer C (1979) Biotechnol. Lett. 1: 75, 439
121. White WL et al. (1948) Mycologia 40:34
122. Veng PP, Gong CS (1982) Enz. Microbiol. Technol. 4:169
123. Macris BJ, Kekos K, Evangelidou X (1987) Enhanced cellulase activity ofFusarium oxysporum
grown on straw for ethanol production. In: Grassi Get al. (eds) Biomass for energy and industry.
Proceedings of the 4th ECC Conference Orl6ans, Elsevier, London, p 699
124. Wood TM, Wilson CA (1987) The tureen of a sheep: a new source of cellulase for producing
84 F. Parisi

fermentable glucose from cellulosic wastes. In: Grassi G e t al. (eds) Biomass for energy and
industry. Proceedings of the 4th ECC Conference Orl6ans, Elsevier, London, p 727
125. Salby K, Maitland GC (1971) Bioehem. J. 104:716
126. Wood TM et al. (1987) Maximizing the production of fermentable soluble sugars from straw
using enzymes synthesized by the Fungus Penicillium pinophilum. In: G. Grassiet al. (eds)
Biomass for energy and industry. Proceedings of the 4th ECC Conference Orl6ans, Elsevier,
London, p 732
127. Wood TM, Brown JA, McCrae SI (1987) in: Proceedings of the 4th European Congress on
Biotechnology, Elsevier, Amsterdam, vol 2 p 335
128. Chahal DS, Hawksworth DL (1976) Mycologia 68:600
129. McHale A, Coughlan MP (1980) FEBS Lett. 117:318
130. Jain S, Tiraby G (1987) Separation and characterisatization of the cellulolytic components of a
thermophilic Fungus: Talaromyces sp. CL-240. In: G. Grassi et al. (eds) Biomass for energy
and industry. Proceedings of the 4th ECC Conference Orl6ans, Elsevier, London, p 355
131. McHale AP, Morrison J, McCarthy U (1987) Production ofcellulase by Talaromyees emersonii
CBS 814.70 and a mutant UV7 during growth on lactose. In: G. Grassi et al. (eds) Biomass
for energy and industry. Proceedings of the 4th ECC Conference Orl6ans, Elsevier, London,
p 704
132. Durand H, Soucaille P, Tiraby G (1984) Enz. Microbiol. Technol. 6:175
133. Saddler JN, Khan AW (1980) Can. J. Microbiol. 26:760
134. Armstrong DW, Brown DA, Martin GM (1982) in: Proceedings of the 4th. Bioenergy Research
and Development Seminar. Winnipeg, Canada
135. Cooney CL et al. (1978) Biotechnol. Bioeng. Symp. 8:103
136. Brooks R et al. (1979) in: 3rd Annual Biomass Energy System Proceedings. SERI, Golden CO
137. Shinmyo A, Garcia-Martinez V, Demain AL (1979) J. Appl. Biochem. 1:202
138. Ng TK, Ben Bassat A, Zeikus JG (1981) Appl. Environm. Microbiol. 4t : 1337
139. Schwarz W H e t al. (1987)Appl. Microbiol. Biotechnol. 27:50
140. Holtzapple M, Humphrey AE, Sye EK (1980) in: Preprints of the 6th International Fermen-
tation Symposium, London, Canada, p 87
141. Phillips JA, Humphrey AE (1980) in: Preprints of the 2nd International Symposium on Bio-
conversion and Biochemical Engineering Delhi. ITT, New Delhi
142. Malfait M, Godden B, Pennincks M (1984) Ann. Microbiol. 135B: 79
143. McCarthy AJ, Ball AS (1987) Lignocellulose degradation by Aetynomycetes. In: Grassi G e t al.
(eds) Biomass for energy and industry. Proceedings of the 4th ECC Conference Orleans,
Elsevier, London, p 351
144. Halliwell G, Phillips T (1987) Synthesis and activity of cellulase produced by strains of the
thermophilic, anaerobic, cellulolytic bacterium Clostridium thermocellum. In: Grassi G e t al.
(eds) Biomass for energy and industry. Proceedings of the 4th ECC Conference Orl6ans,
Elsevier, London, p 709
145. B6guin Pet al. (1987) Conversion of cellulose into ethanol by Clostridium thermocellum: Genetic
engineering cellulase. In: G. Grassi et al. (eds) Biomass for energy and industry. Proceedings
of the 4th ECC Conference Orl6ans, Elsevier, London, p 346
146. Faure E et al. (1987) Caract&isation du syst+me cellulolytique du Clostridium thermocellum.
In: G. Grassi et al. (eds) Biomass for energy and industry. Proceedings of the 4th ECC
Conference Orleans, Elsevier, London, p 717
147. Blotkamp PJ et al. (1978) AIChE Symp. Series 74 (181): 85
148. Downing KM, Ho CH, Zabriskie DW (1987) Biotechnol. Bioeng. 29:1086
149. Cowling EG, Kirk TK (1976) Biotechnol. Bioeng. Symp. 6:59
150. Hendy NW, Wilke CR, Blanch H (1984) Enz. Microbiol. Technol. 6:73
151. Taylor JD (1981) in: Palz W, Chartier P, Hall DO (eds) Energy from biomass. Proceedings of
the 1st ECC Conference London. Applied Science, London, p 330
152. Sinner M, Sehreier M, Ballweg A (1983) in: Strub A, Chartier P, Schleser J (eds) Energy from
biomass. Proceedings of the 2rid ECC Conference Berlin, Applied Science, London, p 984
153. Wayman M (1980) in: Proceedings of the 4tb International Symposium on Alcohol Fuel
Technology. Guaruj~, Brazil, vol 1 p 79
154. Delong EA (1981) Can. Patent 1 096 374
Advances in Lignocellulosics Hydrolysis and in the Utilization of the Hydrolyzates 85

155. Foody E (w.d.) Optimization of steam explosion pretreatment. Final Report to US Department
of Energy (Contract No DE-AC-02-79-ET 23050)
156. Pourqui6 J, Glinkmans G (1986) in: Proceedings of the 7th International Symposium on Al-
cohol Fuel Technology, Technip, Paris, p 54
157. Grohmann K, Torget R, Himmel M (1985) Biotechnol. Bioeng. Symp. 15:59
158. Grethlein HE, Allen DC, Converse AO (1984) Biotechnol. Bioeng. 26:1498
159. Torget R et al. (1988) Initial design and parametric evaluation of a dilute acid pretreatment
process for aspen wood chips. Appl. Biochem. Biotechnol. (in press)
160. Brownell HH, Yu EKC, Saddler JN (1986) Biotechnol. Bioeng. 28:792
161. Sarkanen KV (1980) in: Acid catalyzed delignification of lignocellulosics in organic solvents,
Academic, New York, vol 2 p 127
162. Cunningham RL, Carr ME, Bagby MO (1985) Biotechnol. Bioeng. Symp. 15:17
163. Leisola MSA, Fiechter A (1985) Adv. Biotechnol. Processes 5:59
164. Sinitsyn AP, Bougay HR, Clesceir LS (1983) Biotechnol. Bioeng. 25:1393
165. Chahal DS (1986) in: Proceedings of the 7th International Symposium on Alcohol Fuel
Technology, Technip, Paris, p 39
166. Tjerneld et al. (1985) Biotechnol. Bioeng. Symp. 15:419
167. Tan LUL et al. (1986) Appl. Microbiol. Biotechnol. 25: 250, 256
168. Cantarella M e t al. (1979) Biochem. J. 179:15
169. Alfani et al. (1986) in: Magnien E (ed) Biomolecular engineering in the European Community,
Martinus Hjihoff, Dordrecht, Netherland, p 143
170. Spasov SD (1987) Efficient hydrolysis of natural lignocellulosics materials by a Trichoderma
viride cellulase immobilized on various soluble polymers. In: Proceedings of the 4th European
Congress on Biotechnology, Elsevier, Amsterdam, vol 2 p 101
171. Cantarella M et al. (1987) A comparative study on cellulase immobilization in synthetic and
natural gels. In: Proceedings of the 4th European Congress on Biotechnology, Elsevier,
Amsterdam, vol 2 p 112
172. Boidin J, Adzet JM (1957) Bull. Soc. Mycol. France 73:331
173. Schneider H (1981) Biotechnol. Lett. 3:89
174. Slininger PJ et al. (1982) Biotechnol. Bioeng. 24:371
175. Detroy RW et al. (1982) Biotechnol. Bioeng. 24:1105
176. Dekker RFH et al. (1982) Biotechnol. Lett. 4:411
177. Chung IS, Lee Y-Y, Beck MJ (1986) Biotechnol. Bioeng. Symp. 17:391
178. Alexander NJ (1986) Appl. Microbiol. Biotechnol. 25:203
179. du Preez JC, van der Waalt JP (1983) Biotechnol. Lett. 5:537
180. du Preez JC, Prior BA, Monteiro MT (1984) Appl. Microbiol. Biotechnol. 19:261
181. Toivola A e t al. (1984) Appl. Environm. Microbiol. 47:221
182. Jeffries TW (1985)Biotechnol. Bioeng. Syrup. 15:149
183. Wayman M, Tsuyuki ST (1985) Biotechnol. Bioeng. Symp. 15:167
184. Wayman M e t al. (1986) in: Proceedings of the 7th International Symposium on Alcohol Fuel
Technology, Technip, Paris, p 597
185. Lucas C, van Uden N (1985) J. Basic Microbiol. 25:547
186. van Zyl C, Prior BA, du Preez JC (1987) Production of ethanol from sugar cane bagasse
hemicellulose hydrolysate by Pichia stipitis. Presented at the 9th Symposium on Biotechnology for
Fuels and Chemicals, Boulder CO. To be published in Biotechnol. Bioeng. Symp.
187. Rosnick CS, Chung IS, Lee Y-Y (1987) High-cell continuous fermentation of xylose to ethanol
by Pichia stipitis using cell-recycled bioreactor. Presented at the 9th Symposium on Biotechno-
togy for Fuels and Chemicals, Boulder Colorado. To be published in Biotechnol. Bioeng.
Symp.
188. Alexander MA, Jeffries TW (1987) Effect of ethanol on continuous fermentation of xylose by
Candida shehatae. Presented at the 9th Symposium on Biotechnology for Fuels and Chemicals,
Boulder Colorado. To be published in Biotechnol. Bioeng. Syrup.
189. Grba S, Besic S, Ban SN (1987). In: Proceedings ofthe 4th European Congress on Biotechnology.
Elsevier, Amsterdam, vol 3 p 393
190. Dellweg H et al. (1984) Biotechnol. Lett. 6:395
191. Rizzi M e t al. (1987) in: Proceedings of the 4th European Congress on Biotechnology, Elsevier,
Amsterdam, vol 3 p 415
86 F. Parl.~J

192. Torrie JP, Wilson JJ (1987) Evaluation of fermenting yeasts and fungi using sugar substrates.
Presented at the 9th Symposium on Biotechnology for Fuels and Chemicals, Boulder, Colorado.
To be published in Biotechnol. Bioeng. Syrup.
193. Snikho ML, Enari TM (1981) Biotechnol. Lett, 3:273
194. Snikho ML, Suomalainen I, Enari TM (1983) Biotechnol. Lett. 5:525
195. Tsao GT et al. (1981) Biotechnol. Bioeng.-Symp. i1:315
196. Ho NWY et al. (1983) Biotechnol. Bioeng. Symp. 13:245
197. Huang JJ, Ho NWY (1983) Biochem. Biophys. Res. Comm. 126:1154
198. Ho NWY (1985) Improvement of yeast xylose fermentation and utilization via genetic engineer-
ing. In: Biochemical Conversion Program Semi-Annual Review Meeting, SERI/CP-231-2726
Solar Energy Research Institute, Golden CO, p 295
199. Ueng PP et al. (1985) Biotechnol. Lett. 7:153
200. Lastick SM et al. (1986) BiotechnoL Lett. 8:1
201. Jeyaseelan K, Singh P (1987) in: Proceedings of the 4th European Congress on Biotechnology,
Elsevier, Amsterdam, vol 1 p 394
202. Lastick SM et al. (1985) Xylose fermentation project. In: Biochemical Conversion Program
Semi Annual Review Meeting, SERI/CP-231-2726. Solar Energy Research Institute, Golden
CO, p 243
203. Hasche RL (1945) Chem. Eng. News 23 (20): 1840
204. Kirschenbaum I (1978) Butadiene. In: Kirk-Othmer Encyclopedia of Chemical Technology.
Wiley, New York, vol 4 p 313
205. Delmas M, Gaset A (1986) Chimie et utilisation industrielle des pentoses et d~rivds. Presented
at the Symposium International sur l'Utilisation non Alimentaire du BI6 et du Mais, APRIA,
Paris
206. BIOS (British Intelligence Objective Subcommittee) (1948) No 351
207. Reppe W (1953) Liebigs Ann. Chem. 582:87
208. Mile NA, Walsh WL (1935) J. Am. Chem. Soc. 57:1389
209. Parisi F (1987) Production of ethanol from biomass and its socio-economic impact. Presented
at the 4th European Congress on Biotechnology, Amsterdam
210. Chum HL et al. (1985) The economic contribution of lignins to ethanol production from biomass.
SERI/TR-231-2488. Solar Energy Research Institute, Golden CO
211. Parkhurst H J, Huibers DTA Jr., Jones MW (1980) in: ACS Symp. Series, Div. Petroleum
Chemistry, American Chemical Society, Washington DC, p 657
212. Gendler GL, Huibers DTA Jr, Parkhurst HJ (1983) in: Soltes E (ed) Wood and agriculture
residues: Research on use for feed, fuels and chemicals, Academic, New York, p 391
213. Coughlin RW et al. (1984) In: Wise DC (ed) Bioconversion) systems, CRC Press, Boca Raton
FL, p 41
214. Goheen DW (1971) in: Sarkanen KV, Ludwig CH (eds) Lignins: occurrence, formation, struc-
ture and reactions. Wiley, New York, p 797
215. Singeman GM (1980) Methyl aryl ethers from coal liquids as gasoline extenders and octane
improvers. Gulf Research and Development Co, Pittsburgh PA
216. Hsu OH-H, Glasser WG (1975) Appl. Polymer Syrup. 28:297
217. Hsu OH-H, Glasser WG (1976) Wood Science 9(2): 97
218. Saraf VP, Glasser WG (1984) J. Appl. Polymer Sci. 29:1831
219. Dilling P, Sargent P (1984) US Patent 4 454 066
220. Nimz HH (1983) in: Pizzi A (ed) Wood adhesives, Dekker, New York, p 248
221. Gillespie RH (ed) (1984) Adhesives for wood. Noyes Publications, Park Ridge NJ
222. Muller PC, Glasser WG (1984) J. of Adhesion 17:157
223. Muller PC, Kelley SS, Glasser WG (1984) J. of Adhesion 17:185
224. Glasser WG, Wu LC, Selin J F (1983) in: Soltes E (ed) Wood and agriculture residues: Research
on use for feed, fuels and chemicals Academic, New York, p 149
225. Sirianni AF, Puddington IE (1972) Rubber World 165(6): 40
226. Dimitri MS (1976) US Patent 3 991 022
227. Sirianni AF, Puddington IE (1976) US Patent 3 984 362
228. Weizmann C (1912) British Patent 4 845
229. Andersch W, Bahl A, Gottschalk G (1982) in: Proceedings 5th Symposium Technische Mikro-
biologic Berlin, Institut ffir G~irungsgewerbe und Biotechnologie, Berlin FRG, p 177
Advances in Lignocellulosics Hydrolysis and in the Utilization of the Hydrolyzates $7

230. Vandecasteele JP, Pourqui6 J (1984) in : Preprints of the 6th International Symposium on Alcohol
Fuel Technology, Ottawa, Canada, vol 2 p 227
231. Wayman M, Husted GR, Santangelo JD (1984) in: Preprints of the 6th International Sym-
posium on Alcohol Fuel Technology, Ottawa, Canada, vol 2 p 234.
232. Votruba J, Voleski B (1984) in: Proceedings of the 3rd European Congress on Biotechnology,
Verlag Chemic, Weinheim FRG, vol 2 p 301
233. Larsson M, Holst O, Mattiasson B (1984) in: Proceedings of the 3rd European Congress on
Biotechnology, Verlag Chemic, Weinheim FRG, vol 2 p 313
234. Dadgar AM, Foutch GL (1985) Biotechnol. Bioeng. Symp. 15:611
235. Vandecasteele JP et al. (1986) in: Proceedings of the 7th International Symposium on Alcohol
Fuel Technology, Technip, Paris, p 24
236. Pierrot Pet al. (1987) in: Proceedings of the 4th European Congress on Biotechnology, Elsevier,
Amsterdam, vol 1 p 258
237. Soni BK, Soucaille P, Goma G (1987) in: Proceedings of the 4th European Congress on Bio-
technology, Elsevier, Amsterdam, vol 3 p 335
238. Fick M, Engasser GM (1987) in: Proceedings of the 4th European Congress on Biotechnology,
Elsevier, Amsterdam, vol 3 p 509
239. Soni BK et al. (1986) Bioteehnol. Bioeng. Symp. 17:591
240. Schneider H et al. (1981) Pentose fermentation by yeasts. In: Stewart GG, Rennel I (eds)
Current developments in yeast research. Advances in Biotechnology, Pergamon, Toronto,
p81
241. Chum HL et al. (1985) Evaluation of pretreatment of biomass for enzymatic hydrolysis of
cellulose. SERI/TR-231-2183. Solar Energy Research Institute, Golden CO
242. Wright JD, Power AJ (1985) Biotechnol. Bioeng. Symp. 15:511
243. Parker S e t al. (1983) The value of furfural/ethanol coproduction from acid hydrolysis pro-
cesses. SERI/TR-231-2000. Solar Energy Research Institute, Golden CO
244. Kosaric Net al. (1983) Ethanol fermentation. In: Dellweg H (ed) Biotechnology, Verlag Chemic,
Weinheim FRG, vol 3 p 309
245. Isaacs SH (1984) Ethanol production by enzymatic hydrolysis. SERI/TR-231-2093. Solar
Energy Research Institute, Golden CO
246. Wright JD, Wyman CE, Grohman K (1987) Simultaneous saccharification and fermentation of
lignocellulose: Process evaluation. Presented at the 9th Symposium on Biotechnology for Fuels
and Chemicals, Boulder CO. To be published in Biotechnol. Bioeng. Symp.

References added in proofs:

247. Brownbell HH, Saddler JN (1987) Biotechnol. Bioeng. 29:228

248. Wong KY et al. (1988) Biotechnol. Bioeng. 30:447
249. Wasson L e t al. (1988) The effect of time and temperature on rapid steam hydrolysis (RASH).
Presented at the 10th Symposium on Biotechnology for Fuels and Chemicals, Gatlinburg TN.
To be published in Biotechnol. Bioeng. Symp.
250. Schultz TP, Rughani J, McGinnis JD (1988) A comparison of the pretreatment of sweetgum and
white oak by steam explosion and RASH processes. Presented at the 10th Symposium on Bio-
technology for Fuels and Chemicals, Gatlinburg TN. To be published in Biotechnol. Bioeng.
Symp.
251. Hinman ND et al. (1988) Xylose fermentation: an economic analysis. Presented at the 10th
Symposium on Biotechnology for Fuels and Chemicals, Gatlinburg TN. To be published in
Biotechnol. Bioeng. Syrup.
Modelling, Identification and Control
of the Activated Sludge Process

Stefano Marsili-Libelli
Department of Systems and Computers Engineering, University of Florence, Via
S. M a r t a , 3-50139 F i r e n z e , I t a l y

1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
2 Reduced-Order Dynamic Modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
2.1 Simplified Microbial Kinetics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
2.2 Dissolved Oxygen Kinetics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
2.3 Simplified Nitrification Kinetics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
2.4 Sedimentation Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
2.5 Continuous-Flow Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
3 Model Identification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
3.1 SML Model Identifiability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
3.2 Parametric Sensitivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
3.3 Practical Model Identification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
3.4 On-line Parameter Identification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
3.5 On-line Estimation of Bioactivities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
3.6 On-line Estimation of Process Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
4 Process Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
4.1 Process Performance Indicators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
4.2 Conventional Control Strategies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
4.2.1 PID Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
4.2.2 Approximate Optimal Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
4.3 Activated Sludge Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
4.3.1 Dissolved Oxygen Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
4.3.2 Sludge Recycle Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
4.3.3 SCOUR Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
4.4 Adaptive Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
4.4.1 Structure of the Dissolved Oxygen Adaptive Controllers . . . . . . . . . . . . . . . . . . . . . . . 138
4.4.2 Performance of Self-tuning Controllers . . . . . . . . . . ' . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
6 List of Symbols . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
7 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146

Practical experience shows that the efficiency and reliability of the activated sludge wastewater treat-
ment process can be significantly improved if its time-varying nature is taken into account in the design
of automatic control systems. This paper focuses on the dynamic aspects of the activated sludge
treatment process and assesses the operational improvements brought about by automatic control.
The three aspects of modelling, identification and control are reviewed and their interdependence is
stressed. After introducing a simplified model for the pollutant/biomass interaction, the estimation
and control problems are addressed and practical algorithms are discussed. The improvements brought
about by such algorithms are clearly demonstrated.

Advancesin BiochemicalEngineering/
Biotechnology,Vol. 38
ManagingEditor: A. Fiechter
9 Springer-VerlagBerlinHeidelberg 1989
 90 S. Marsili-Libelli

1 Introduction
Biological wastewater treatment by means of activated sludge processes has been
an expanding business for years and this may convey the notion of a well established
engineering practice. Its widespread use might imply that such a technical expertise
9has been reached that any further significant advancement is unlikely to occur and
that plant operation is exactly predictable and satisfactory.
A closer look at scores of operating records, though, reveals a dramatically different
picture. Traditionally, sanitary engineers have put a premium on plant design rather
than operation, and there are now countless reports of gross process failures caused
by inadequate operational management. This, together with increasing energy costs
and stricter environmental standards, has stressed the need for reliable automatic
control.
In fact, the day-to-day operational practice has shown that relying on the steady-
state design assumptions could be dangerously misleading given the inherently un-
certain process dynamics and the large, unmeasurable variations of operating condi-
tions. Regrettably, several factors have prevented the application of computer control
to activated sludge treatment plants, compared with other biotechnological processes.
The difficulty in obtaining reliable process measurements ranks highest among them,
and the ensuing uncertainty in assessing the plant efficiency has surely discouraged
any endeavour to improve it.
These considerations demonstrate that a deep division exists between design practice
and process performance, but also indicates that the high operationalstandards now
demanded can only be attained by combining sanitary engineering skills with a thor-
ough knowledge of process dynamics and the application of advanced control strate-
gies.
Since the pioneering research of Andrews 1-6) and Olsson 7, 8) the scientific interest
in the dynamic aspects of wastewater treatment processes has increased and several
initiatives took place to assess the research needs in this area. References 9-11) collect
papers presented at specialized scientific meetings to assess the impact and potential
benefits of taking into account the dynamic nature of the problem in a comprehensive
time-varying water quality management scheme.
Mathematical modelling, identification and real-time control of this biotechnologi-
cal process represent a challenging area of endeavour for biotechnologists and control
engineers alike, In fact, nonlinear process kinetics, time-varying parameters, and the
lack of directly measurable process variables, all call for new and imaginative engineer-
ing solutions.
This paper reviews the research developments over the last decade concerning the
three aspects of modelling, identification, and control of the activated sludge waste-
water treatment process demonstrating how these aspects are tightly interconnected.
Though mathematical modelling of microbiological systems is now a well established
discip!ine (see e.g. 12.13)) and many descriptive mathematical models are available,
few of these results can be applied to process control because of their complexity.
As an alternative, this paper advocates the use of simplified models, which can be
easily included in estimation and control algorithms arid implemented on small,
dedicated microcomputers.
The scope of the paper is evenly distributed among three main sections, each devoted
Modelling, Identificationand Control of the ActivatedSludge Process 91

with equal emphasis to modelling, identification and control of the process to stress
their interdependence. In Sect. 2, a simplified model, developed by the author t4, ~5)
is introduced. It is shown that this model compares well with the widely used Monod 56)
kinetics, and can explain satisfactorily the interaction between pollutant and the
specialized biomass. Section 3 describes the identification of this model, both theo-
retically and practically, with reference to numerical and experimental problems; the
on-line estimation of relevant process quantities is also considered and practical
estimation algorithms are presented and discussed. Section 4 focuses on the design
of control laws for this specific process and reviews several possible control strategies
to cope with typical problems encountered in practice.

2 Reduced-Order Dynamic Modelling

Wastewater processing by means of activated sludge is undoubtedly the most wide-

spread sewage treatment practice. The activated sludge which forms the basis of
the biological process is in fact a mixture of several microorganisms which under
proper environmental conditions act as tiny bioreactors and transform the biode-
gradable pollutant, which constitutes their growth substrate, into protoplasm with
energy being supplied by dissolved oxygen and carbon dioxide being released. The
difficulty of modelling the whole chain of bioreactions lies in the complexity and
variability of the microbial colony, with each species exhibiting a peculiar behaviour
and interacting with all the others. Many mathematical models already exist, which
take into account the full structure of the microbial dynamics and give a detailed pic-
ture of the bioreactions developing in the oxidation stage. In this area, credit ought
to be given to Andrews 1-3), Busby and Andrews 4~, and Olsson 7) for their pioneering
work.
Hystorically, the first aspect to receive modelling attention was the biodegradation
of carbonaceous substrate, globally measured by the "biochemical oxygen demand"
(BOD), representing the amount of oxygen that the microorganisms require to meta-
bolize the substrate. Later, the oxidation of nitrogenous wastes was considered and
the first model describing the dynamics of nitrification was proposed by Poduska and
Andrews 17). Then the implications of the dissolved oxygen (DO) dynamics began to
emerge. The coupling nature of DO and its potential role for control were highlighted
by Olsson and Andrews 18) and Stenstrom and Andrews 59). Presently, the detailed
modelling of the interactions between the microbial colony and the complex substrate
is well developed and very detailed, speculative models are available 20-25). Given
these premises and a much wider literature than that just quoted, one might wonder
if there is really a~need for yet another model. The answer lies in the complexity of
these literature models, developed for speculative purpose, which make them un-
suitable l\~r the mathematical manipulations required in most control applications.
To fill the gap between modelling accuracy and control needs, a simplified model
suitable for control applications was derived by the author 24, ~5) and is now briefly
reviewed. It describes the dynamics of the main four process phases:
a) Biodegradation of carbonaceous BOD ;
b) nitrification;
92 S. Marsili-Libelli

c) dissolved oxygen utilization;

d) sludge sedimentation.
Before introducing the complete continuous-flow process model, the dynamics of
each subprocess is first derived for a batch situation.

2.1 Simplified Microbial Kinetics

This section summarizes the basic ideas underlying the development of a simplified
kinetics involving substrate and activated sludge. The core of the model is represented
by the interaction between carbonaceous BOD and the sludge biomass, which is
globally modelled as "Mixed Liquor Suspended Solids" (MLSS). The kinetics
that will be used in the sequel may be regarded as a simplified version of the well-
known Monod dynamics of the substrate/biomass {S, X} pair, which is repeated here
for convenience:

1
Substrate" dS/dt = -- ~ g(S) X ; (1)

biomass: dX/dt = g(S) X -- KdX (2)

where the specific growth rate is defined as

ftS
~(s) - (3)
Ks+S

with fi representing the maximum growth rate and K s the half-maximum velocity
constant. In contrast with the above Monod Eqs. (1)-(3), in the model proposed here
the substrate/biomass interaction is viewed as an ecological prey/predator pair and
modelled according to a modified Volterra-Leslie logistic equation (Maynard-Smith
26)). The resulting model, which will be referred to as SML, is now introduced. The
two differential equations describing the dynamics of substrate and biomass are the
following:

Substrate: dS/dt = --KbSX ; (4)

biomass: dX/dt = KoSX -- KmX2/S (5)

Where the positive constants {Kb, K , Kin} are kinetic parameters thus defined:
K b = substrate transformation rate coefficient (rag 1-1 h 1);
K c = biomass synthesis rate coefficient (rag 1-1 h-1);
K m-= biomass endogenous metabolism rate coefficient (h-l).
The basic assumptions under which the model was derived are the followi~ng:
1) The substrate uptake rate is proportional to the product of substrate and biomass
(second-order kinetics).
Modelling, Identification and Control of the Activated Sludge Process 93

2) The biomass growth rate is proportional to the substrate uptake rate, hence it has
the same form as the substrate decay. In addition, a negative restraint term has
been added to account for diminished growth due to substrate limitation. This
term, expressed by the quadratic term X2/S, represents the effect of endogenous
metabolism and depends on substrate concentration.
Some qualitative features of the SML model (4) (5) can now be easily established.
Consider the critical biomass X* given by the parabola

X* = S2Kc/Km - (6)

It can be seen that Eq. (6) divides the first quadrant of the {S, X} plane in two regions,
each with a clear biological meaning. Figure 1 a shows that Eq. (6) partitions the
{S, X} plane into a synthesis region (zone B) and an endogenous metabolism region
(zone A). For any given substrate S, if X < X* the substrate/biomass system lies
in the B zone, where synthesis is the prevailing metabolism and dX/dt > 0. Since
dS/dt is always negative, any bioreaction originating in this region will tend towards
the boundary (6), which is intersected with zero derivative (dX/dS = 0). Beyond this
boundary the region with prevailing endogenous metabolism (X > X*) is entered:
both variables have negative derivatives and any bioreaction tends to the origin, which

80
A / a
/

60

4o
X

2O

0
0 10 20 30 40
& S (rag [-11

80 A ~ B

6O
'.._,
Fig. 1 a and b. {S, X}-plane evolution of
E 40 batch models. The boundary between the
X
synthesis (Zone B) and endogenous metha-
P5 bolism (Zone A) regions is represented by
20 the dashed curve: a SML model Eqs. (4),
(5): the boundary is given by Eq. (6).
b Monod model Eqs. (1) (3): the boundary
Pl I P2 i r P4 is given by Eq. (7)
0
0 10 20 30 40
s (rag 111
94 S. Marsili-Llbelli

is reached in an infinite time. Hence the model does not exhibit oscillations, which is
an undesirable feature of the classical Leslie equation. In fact, any evolution originat-
ing in the first quadrant, with the exception of the X axis (S = 0), is bound to remain
in this quadrant and will tend to the origin without ever reaching it. On the other hand,
it cannot interesect either the S axis, since all the bioreactions tend to leave it, or the
X axis as then S would tend to zero increasing the second term in (5) and dX/dt would
tend to - - oe. Therefore, the X axis is an asymptote for all the possible batch evolutions.
Figure 1 b shows the equivalent trajectories produced by the Monod model (1)-(3),
which exhibits a similar state space partitioning between synthesis and endogeneous
metabolism. In this case the two regions are divided by a vertical line which depends
only on the model parameters

S*- KsKa (7)

fi - - K d

There are several aspects which seem to favour model (4), (5) over the Monod kinetics.
From the structural point of view, the endogenous metabolism is included in the basic
kinetics and it is not required, as with Monod, to introduce an additional decay term
for which the linear choice (--KdX) is questionable. Furthermore, the SML model
is linear in the parameters, which simplifies the numerical calibration as will be describ-
ed in Sect. 3.

2.2 Dissolved Oxygen Kinetics

Aerobic bioprocessing of organic matter requires oxygen as a primary source of energy.

It has been shown that this can become a limiting factor 27), but if it is plentiful enough
to sustain growth then oxygen utilization is stoichiometrically related to biomass
synthesis and endogenous metabolism. A key process variable, the "oxygen uptake
rate" (OUR), can be defined on the basis of rate Eqs. (4), (5) as the sum of synthetic
and maintenance activities

O U R = 7sKbSX -t- TeKmX2/S (8)

where 7s, 7e are oxygen utilization coefficients. Equation (8) represents the total
depletion rate of oxygen in the batch, which is in turn resupplied through diffusion
from the gaseous phase. In fact, oxygen diffusion rate into the liquid is proportional
to the difference between the dissolved oxygen concentration C and its saturation
value Cs,t, i.e. KLa(Csa t - - C) where KLa represents the oxygen mass transfer rate
coefficient into the liquid phase. The complete dissolved oxygen balance is then

dC/dt = KLa(C, t C) - - O U R (9)

To gain further insight into the DO dynamics (9) and to simplify the subsequent
parameter identification, a further relationship among DO balance parameters is
now derived. Equations (4), (5) are solved for the terms SX and X2/S and substituted
into Eqs. (8), (9). Integrating over the whole duration of the batch (theoretically from
Modelling, Identification and Control of the Activated Sludge Process 95

0 to oo) with initial conditions {So, Xo} and under the assumption that the dissolved
oxygen concentration is kept constant (dC/dt = 0), yields Coo, the total amount of
oxygen needed to metabolize the initial substrate So and for self-oxidation of the
initial biomass Xo

0 0
C~o : ~ KLa(Csa, - C) dt -(Ys + 7eKc/Kb) f dS - 7e f dX {10)
O So Xo

In practice the substrate oxidation process takes a finite time, hence the left-hand-
side integral is botmded. Reworking the integration limits of the right-hand-side
integrals and solving yields

C = (Ts + 7r So + y Xo (11)

If So and Xo are expressed in mg 1 1 of oxygen and if the term ~eXO includes all the
self-oxidizing processes, then the total utilized oxygen C ~ must equal the sum of the.
initial substrate So and the self-consumption term 7eXO. Therefore, the following
must hold

C = So + 7eXo (12)

Equating Eqs. (11) and (12), the following relation among parameters is obtained

Ys + YeKo/Kb = 1. (13)

The interdependence of Ys and Ye is important in view of parameter calibration. In

fact, as a consequence of Eq. (13) either 7s or 7e only need to be estimated.

2.3 Simplified Nitrification Kinetics

It is well known that nitrification occurs in two successive oxidation stages, with nitrite
as an intermediate product. The complete chain of reactions was modelled by Poduska
and Andrews av), though more recently single-step nitrification models were proposed
by Gujer and Erni 2s) based on the notion that once the intermediate nitrite stage has
been reached, the reaction proceeds until complete oxidation is achieved. Also, from
the system-theoretical point of view, the N H 4 ~ NO 2 stage is the slowest and therefore
controls the overall reaction rate. Hence, the following single-stage model is pro-
posed as)

Ammonium-N: dSam/dt - ~tn X ' (14)

Ya " '

Nitrate-N: dNa/dt = --dSam/dt ; (15)

Nitrifiers: dX/dt = ~l~nX n - - KnX n (16)

96 S. Marsili-Libclli

where the rate expression gn takes the following composite form:

SomC
~tn = 0n (Kam q_ Sa) (K ~ q_ C) (17)

Stenstrom and Poduska 27) have demonstrated that the nitrification rate depends on
both the amount of substrate and of dissolved oxygen available. It should also be
noticed that the yield factor Ya has not the same meaning of a conversion factor as
in the case of carbonaceous BOD, since the nitrifiers take up a very limited amount
of nitrogen from the ammonia pool, and act instead mainly as catalysts.
As far as oxygen utilization is concerned, the nitrification can be included in the
DO balance already determined in Sect. 2.2 by simply adding a third term to the
O U R expression (8) of the form

Tn.nXn/Y a (18)

This factor is to represent the oxygen consumption for ammonium-N oxidation.

Since the nitrifying bacteria mainly act as catalysts and do not use oxygen directly,
a maintenance term is not included. Hence the full O U R expression for both carbon-
aceous and nitrogenous oxidation is the following

OURc~ = TsKbSX q- TeKmX2/S q- 7ngnXn/Ya (19)

2.4 Sedimentation Dynamics

This section of the model is not directly concerned with biochemical reactions, but
rather considers the double function of the secondary settler: separating the bioflocs
from the liquid in order to produce a solid-free effluent, and thickening the biomass
at the bottom of the settler to be recycled back into the aerator. Of these two actions,
thickening is surely the most important and complex, whereas clarification may be
regarded as a consequence of the former. In fact, although a clarification failure has
an immediate impact on the process management, it is always the result of a thickening
failure which would take longer to detect but whose implications are surely far more
catastrophic.
Compared with the biochemical aspects of the process, sedimentation has received
relatively less attention, and though the basic theory of flocculant suspensions is
well established 29-31) the resulting partial differential equation models have often
been neglected in favour of heuristic or empirical algebraic rules 32). Sedimentation
is a mass transfer process which can be modelled according to the solid flux theory 29.
31) to describe the subsidence of suspended solids through layers of differing concen-
trations. Consider the settler schematic diagram of Fig. 2 and the following diffusion
equation

~X ~F ~F ~X
(20)
St -- ~z - ~X x ~z
Modelling, Identification and Control of the Activated Sludge Process 97

II xo
X X~
F*

h~b~ Fig. 2. Diagram of the secondary

| settler, showing the feed point at height
,~ ~ ~ and the sludge discontinuity at height
..t .J_
hu between the clarification zone with
~ concentration X and the compaction
Xr Xr zone with concentration Xb

where F is the biomass solids flux of concentration X at height z from the thickener
bottom. If Eq. (20) is approximated by finite differences, the following holds

~X Fi - - Fo
m
(21)
at Az

Equation (21) states that the rate o f change of biomass concentration X through a
horizontal layer of thickness Az is equal to the difference between the incoming (Fi)
and outgoing (Fo) fluxes divided by Az. The motion of the bioflocs generating the solid
flux is governed by two main forces: settling subsidence due to gravity and "bulk"
flow due to sludge withdrawal from the bottom. Therefore the total flux can be express-
ed as

F=F(X,u)--Xv +Xu (22)

where v and u are respectively the settling and bulk velocity of the bioflocs. Several
analytical expressions have been proposed for the settling velocity, and the exponential
and power-law approximations are the most widely used

exponential 3a). v = n exp (--aX) " (23.1)

power law 32). v = nX" ; (23.2)

a
inverse power law 33): v - (23.3)
b +X"

where the settling parameters {n, a} have different values and meaning in each of the
three Eqs. (23). Severin et al. 9o) have demonstrated with a series of experiments that
the values of these parameters are very stable, even over a period of many seasons of
operations. A family of solid flux curves obtained with Eq. (23.1) is shown in Fig. 3
as a function of solids concentration X and with bulk velocity u as a parameter. This
can be expressed in terms of recycle and waste ratios r and w, assuming total solids
capture at the thickener bottom

r+w
u = Q ~K- (24)
98 S. Marsili-Libelli

2.5 u:O.5

2.0 z F , ~ ~ u=0.4
&---
'E 1.5 u=0.3
"T

C~

LL
1.0 ~r\ ~ u:0.2

0.5 / u:O.1

0 -' / , , i ~X* 'Xr , ,

0 1 2 3 4 5
X (kg m-31
Fig. 3. Solid flux curves with bulk velocityu as a parameter, showing how the limiting quantities F*
and X* and underflow concentration Xr are graphically related

where A is the thickener cross-section and Q the process flow rate. Figure 3 also shows
that given a bulk velocity u, the underflow concentration X r is uniquely determined
by the solid flux established in the thickener. In fact the minimum flux F*(u) represents
the maximum rate of solids transmission through the liquid/solid interface for a given
clarifier geometry and sludge settling behaviour 29-33). Excess solids which cannot
be transferred to the bottom accumulate above the interface and eventually migrate
upwards, leading to a clarification failure. In this sense clarification may be regarded
as a consequential aspect of thickening. Conversely, the underflow solids concentra-
tion is not a dynamic variable but depends on the limiting flux and available mass
above the interface. If the thickener is critically loaded, the total limiting flux AF*(u)
passing through the thickening zone equals the underflow withdrawal. Equating the
two fluxes, the underflow concentration is obtained

AF*(u) (25)
Xr -- Q(r + w)

Figure 3 shows how X and F*(u) can be determined graphically, whereas for numerical
computations the po'wer law approximation (23.2) can be used. Expressing the total
flux as
F(X, u) = nX a + Xu (26)

with n > 0 and a < 0, and letting ~F/~X = 0, the limiting quantities X* and F* are
determined

X*(u) = exp
{1 (u)} In - (27)

F*(u) = n e X P { a ~ _ a l l n ( - ~ ) } + uexp {a_~ll in ( - ~ ) } (28)

Modelling, Identificationand Control of the Activated Sludge Process 99

Now a solid mass balance around the thickener can be set up considering that the solids
loading rate depends on the incoming concentration X and that the solids depletion
rate from the underflow is governed by the limiting flux only

dM/dt = Q(1 + r) X - AF*(u) (29)

where M is the total mass of solids in the thickener and the underflow concentration
is determined through Eq. (25), as shown in Fig. 3.
A slightly different approach has been proposed by Stehfest 33) resulting in a simple
ordinary differential equation to describe the settler dynamics, and particularly that
of the interface separating the clarified water from the thickened sludge (sludge
blanket). Assuming again the settler scheme of Fig. 2, where the feed point and the
position of the sludge blanket are shown, Stehfest reworked the original partial dif-
ferential Eq. (20), taking into account the concentration above (X) and below (Xb)
the sludge blanket, which is in fact a concentration discontinuity. Integrating Eq. (20)
in the spatial variable z yields

Zb

d I X dz = F(Xb) -- F(Xa)
dt (30)
Za

If a discontinuity exists at an intermediate height zo e (Za, Zb) then Eq. (30) yields

dzo/dt = _ F ( X ) F(Xb) (31)

X a -- X b

From Eqs. (30) and (31) a mass balance above and below the sludge blanket can be
written under the assumption that the two zones above and below the discontinuity
have homogeneous concentrations X a and X b. After a number of intermediate mani-
pulations, mainly due to continuity constraints and the propagation of rarefaction
waves from the clarifier bottom, the following lumped-parameter model is obtained:
Solids concentration above the interface (X) :

1
dXa/dt - r _ hb [Fi(X ) -- F(Xa) ] - (32)

solids concentration below the interface ( X b ) :

X b -- X a }
dXu/dt = hbb
1 {F(X~) - F(Xo) X - Z Xa [F(X~) - F(X_)] 9 (33)

sludge blanket height (hb)"

dhb/dt = F(Xa) -- F(X_) (34)

X_ -- X a
100 S. Marsili-Libelli

12. 6.0-

10- 5.5-

8- C s.o:
..~ 6 .
r

4. :2 4.0-

2 84 3.5-
Calibration Validation Calibration Validation
i I i i i i i i i i i i i i i i i i i ii i ii i i i i

12 24 12 2l, 12 12 24 12 24 12
Time (h) Time (h)

Fig. 4. Experimental performance of Stehfest model 33~ Eqs. (32)-(34). The sludge blanket height
(hb) and underflow sludge concentration (Xr) are shown. The data in the first 24-h period are used for
calibration and those in the following 24-h for validation. Reproduced by permission of the Institute
of Measurement and Control on behalf of Stehfest (33)

where
X = min (Xb, Xta)
X ~ = max (Xb, X , )
= feed height above clarifier bottom (m).
Model (32)-(34) was practically calibrated with measurements from municipal
wastewater treatment plants and the agreement between experimental data and model
predictions are shown in Fig. 4.
So far the thickening function of the settler has been considered. As far as clarifica-
tion is concerned, empirical relations have been used almost invariably 31, 32) but
recently a dynamic double black-box approach was proposed by Olsson and Chapman
34) to model the effluent suspended solids. It is based on the notion that the clarifica-
tion response to influent step inputs varies depending on the sign of the step. In fact,
changes in effluent suspended solids following a step increase were rapid and possibly
with some overshoot, whereas a step decrease produced an exponential decline in
turbidity. Two separate linear models were adopted, one for each direction of the
flow rate change:
Positive flow-rate change:

d 2 X / d t 2 + a 1 d X J d t + a2X ~ = b 1 dQ/dt + bzQ ; (35.1)

negative flow-rate change."

dXe/dt + alX e = b t Q , (35.2)

where X e is the concentration suspended solids in the clarifier effluent and {a 1, a2,
b,, b2} are model parameters that can be adjusted on-line during process operation.
Combining the two Eqs. (35.1) and (35.2) a good agreement between observed and
predicted effluent suspended solids has been achieved by the authors, as shown in
Fig. 5 from 34).
Modelling, Identification and Control of the Activated Sludge Process 101

60

2.=_ ~5

0 1.5 3.0 4.5 6.0

Time (h}
oJ 150

ck

E 100
03
t- "(3

~o 50
LLI
o 1'.s 3'.o 4'.s ~.o
Time (h)

Fig. 5. Double linear approximate clarifier model Eqs. (35) : flow rate disturbance performance 347
Published by the International Association on Water pollution Control (IAWPRC) in conjunction
with Pergamon Press

2.5 Continuous-Flow Model

Having considered the batch kinetics of the main process streams, the complete
continuous-flow model can be easily obtained adding the appropriate input-output
transport terms. Hereafter, the bioreactor is considered to be perfectly mixed so that
the concentration of each component is spatially homogeneous. The opposite to the
completely mixed arrangement is the plug flow reactor 82), where no transversal mixing
takes place. Practical plants lie in between these two extremes, being closer to either
kind depending on tank geometry and flow patterns.
Figure 6 depicts the process scheme of a completely-mixed activated sludge treat-
ment composed of an aerator of volume V and a secondary settler with cross section A.
The recycle and waste streams expressed as fractions of the process flow rate Q are
also shown. S i is the incoming organic load usually expressed as BOD, whereas the
biomass in the aerator is referred to as mixed liquor suspended solids (MLSS).

Q, Si Aerotor (l+rlQ Settler (1-w)Q

{s,x,c}

1
L D
rQ wQ

Fig. 6. Schematic diagram of a continuous-flow completely mixed activated sludge process

102 s. Marsili-Libelli

A straightforward hydraulic mass b a l a n c e yields the necessary t r a n s p o r t terms to

s u p p l e m e n t the kinetics already established by Eqs. (4), (5), (9) whereas the settler
mass Balance o f Sect. 2.4 was already derived for a c o n t i n u o u s - f l o w ~rocess. Then,
the c o n t i n u o u s - f l o w m o d e l comprises the following e q u a t i o n s :
a) Differential equations."

BOD : dS/dt = - - K B S X - - q(1 + r) S + qS i ; (36)

MLSS: d X / d t = KcSX - - KmX2/S - - q(1 + r) X + qrXr (37)

DO: d C / d t = KLa(C at - - C) - - O U R - - q(1 + r) C + qCi ; (38)

settler M A S S : d M / d t = Q(1 + r) X - AF*(u). (39)

b) Auxiliary equations."

AF*(u)
Underflow MLSS: X - Q(r + w) ' (40)

oxygen u p t a k e rate" O U R = 7sKbSX -}- 7eKmX2/S (41)

where q = Q/V is the d i l u t i o n rate a n d C i the i n c o m i n g dissolved oxygen. It should

be stressed that all the above Eqs. (36) (41) deal with c o n c e n t r a t i o n s , with the excep-
tion of Eq. (39), which represents a global b a l a n c e a n d hence involves global quantities
such as clarifier cross section A a n d total i n c o m i n g flow Q. E q u a t i o n s (36)--(41)
constitute the basic d y n a m i c model for c a r b o n a c e o u s B O D o x i d a t i o n from which
e s t i m a t i o n a n d c o n t r o l algorithms are n o w derived. The typical n u m e r i c a l values of
the p a r a m e t e r s a p p e a r i n g in these m o d e l e q u a t i o n s are shown in T a b l e 1. T h e y were

Table 1. SML model 14,15) parameter ranges

Parameter Value Units Ref.

Kb 5.2x10-4-7.5• -4 mg 11h-1 14,5z)

Ko 4.7 - 10 - 4 - 1.5x10 -3 mg-llh-1 14,52)
K m- 1.2 + 10 -5 + 5.2x 10-4 h -1 14, szl
15, 52)
Ys 0.51 + 0.71
15, 52)
y~ 0.24 + 0.36
Kea 0.14 + 0.27 h-1 is)
17)
Ya 0.05 + 0.067 --
Jin 0.04 + 0.058 h-1 17)
Ko 0.0025 + 0.004 h-i iv)
Kam 9.5 + 11.5 mgl l h - 1 17)
K~ 0.5 + 1.5 mgl - l h 1 277
Yn 0.33 + 0.46 _ 15)
31 33)
n 3.76 + 5.2 --
31-33)
a --2.0 +--2.25 --
Modelling, Identification and Control of the Activated Sludge Process 103

obtained either from literature value or from specific experiments, as described in the
following section.

3 Model Identification
Identifying a dynamic model entails the determination of the numerical values of its
parameters so that the model response reproduces as closely as possible the available
experimental data. The most widely used "discrepancy measure" between model
output and experimental data is the sum of squared differences

N
Esx - ~ [S(i) - Sexp(i)] 2 + [X(i) - Xexp(i)]2 (42)
i-I

where {Sexp(i), Xexp(i) ; i= 1, 2 . . . . . N} are experimental data of substrate and bio-

mass. It must be stressed that Esx is a function of model parameters and as such its
actual form depends on the model equations being used.
The determination of the parameter values of microbial kinetics has represented
a challenging endeavour ever since such mathematical models were first introduced.
Nonlinear equations, such as M o n o d kinetics Eqs. (1)-(3), together with data scarcity
and inaccuracy, generated new identification problems and ad hoc estimation proce-
dures were devised to deal with these difficulties, as described by early contributions
such as Heineken et al. 35~and Naito et al. 36) This section reviews most of the research
developed in the last decade for the identification of activated sludge kinetic para-
meters. However, before undertaking the actual parameter identification it is necessary
to assess whether the model equations are such that their parameters can be identified
from experimental data. This intrinsic model property is termed identifiability and
its assessment is a necessary prerequisite to any practical parameter identification
endeavour.

3.1 SML Model Identifiability

The difficulties encountered in the identification of the Monod dynamics have been
extensively reported in the literature 35-42). They arise mainly from the close correla-
tion between (t and Ks, as pointed out particularly by Holmberg and Ranta 42). In
fact it can be shown that the estimation error Esx defined by Eq. (42), when combined
with the M o n o d kinetics (1)-(3), produces a narrow and elongated "valley" around
the minimum corresponding to the best fitting model parameters. An awkward shape
of the estimation error E x is likely to cause severe numerical problems to the identifica-
tion algorithm.
Since the SML model (4), (5), (9) was derived independently of the Monod kinetics
and has a quite different structure, it should not be expected to encounter the same
kind of difficulties and the results obtained for the Monod model cannot be transferred
directly into this context. Therefore a new identifiability analysis is now carried out
in two steps : first, a general test is performed to assess the model identifiability from
a theoretical point of view, then the more practical aspects are considered.
104 S. Marsili-Libelli

For the first part, the basic notation and definitions introduced by Di Stefano and
Cobelli 43) are used throughout. Consider a dynamic system described by the vector
differential equations

dZ(t, P)/dt = f(Z(t, P), t, P) (43)

Y(t, P) = g(Z(t, P), t, T) (44)

with initial conditions

Z o = Z(t o, P) (45)

where Z ~/~n, y e/~q, P e R p are the state, output, and parameter vectors respectively,
and f(.) and g(.) are vector-valued nonlinear functions of consistent dimensions. The
system is said to be parameter identifiable on the interval (0, T) if there exists a unique
solution to the parameter vector P from Eqs. (43)-(45). The theoretical identifiability,
sometimes referred to as structural identifiability, can be assessed through the Poh-
jampalo test 44), which states that the system (43)-(45) is identifiable over the interval
(0, T) if there exists a unique solution to the set of equations

g(k)(z(t0, P)) = Zk(0 ) k = 0, 1, 2,... (46)

where g(k) is the k-th derivative of the vector function g(.) and Zk(0 ) is the k-th derivative
of the state vector Z evaluated at t = 0. This test assumes perfect continuous measure-
ments along the interval (0, T) and in principle relates the identifiability to the recon-
structibility of the initial state Z o-
Therefore it does not take into account the state evolution in the interval (0, T)
and consequently it disregards any situation in which the state variables and/or the
available data m a y be such as to render the identification less reliable. Thus, the
Pohjampalo test should be regarded just as a preliminary stage of identifiability assess-
ment, before applying more realistic tests which take into account the actual experi-
mental setting. Assuming that all the state vector Z = [S, X, C] T is accessible (i.e.
Y = Z and f = g) and indicating with {So, Xo, Co} the state variables at t = 0 and
with {$1, X1, C1, Xz} their first and second derivatives at t = 0, the following quanti-
ties are obtained

S 1 = --KbSoX o (47.1)

X 1 = KcSoX o - - KmX2/So (47.2)

C 1 = KLa(Cs, t - - Co) -- ~tsKbSoX 0 "~eKm-Xo2/So (47.3)

• = otSl• + s0• Km/2•

: / • 474t
According to the Pohjampalo test, the model (4), (5), (9) is identifiable if the para-
meters {Kb, Kc, Kin, 7e} can be obtained by solving Eqs. (47). The K b parameter can
Modelling, Identification and Control of the Activated Sludge Process 105

be obtained independently solving Eq. (47.1) and then 7e is obtained from Eq. (47.3),
making use of Eq. (47.2)

Ca -- KLa(Csat -- Co) + KbS~176 (48)

7~ = Xl

The two remaining parameters Kr and K m can be obtained from the linear system
Eq. (47.2) and Eq. (47.4) provided that the matrix

XoS0 X2So
(49)
D = XoS1 + X1S0 _ 2XlXoSo82- X2S1

is non-singular. It is easy to show that singularity occurs if and only if

2 X~ - XI (50)
SO S~

Condition (50) will never hold in a batch situation. In fact, whereas X o and SO are
always positive, the derivatives at t = 0 have opposite signs since the substrate can
only decrease (S 1 < 0) whereas the biomass can only increase (X 1 > 0). Since the
two members of(50) can never have the same sign, this equality never holds and there-
fore matrix D cannot become singular, thus assuring a unique solution for Ko and K m.
The above reasoning proves that all the parameters {Kb, Kc, Km, 7e} are identifiable
whenever SO > 0 and X o > 0, which is the only realistic situation.

3.2 P a r a m e t r i c Sensitivity

As already stated, the Pohjampalo test is a minimal guarantee that the system is indeed
identifiable. More practical considerations are now required to ensure practical iden-
tification with limited experimental data. Sensitivity analysis provides a quantitative
assessment of the extent to which parameter errors influence model response. Consider
again the nonlinear vector dynamic system Eq. (43)

dZ/dt = f(Z, P, t) (51)

where Z e R n is the state vector and P = {p~; i = 1, ..., p} e R p is the parameter

vector. The sensitivity S~z with respect to the i-th parameter p~ is then defined as the
incremental variation of the state vector Z caused by an incremental variation of the
i-th parameter Pi' namely

~Z
Si = -- (52)
z ~Pi
106 S. Marsili-Libelli

A sensitivity system can be associated to the given system (51) simply by applying the
definition (52) to yield the following sensitivity-generating system with respect to Pl

(53)

where the terms in parentheses are computed along the trajectory with nominal para-
meter values pl "). These results, mainly due to Perkins 45), are useful for practical
identifiability assessment. In fact the evolution of the sensitivity system (53) provides
information on the practical identifiability of the system from a given set of measure-
ment. In fact, as pointed out by Holmberg 40) and later by Vialas et al. 46) the sensi-
tivities associated to the Monod kinetics (1)-(3) show that this model takes full advan-
tage of the information contained in the data only during a short part of a typical

30 " ~ - N ~++++++_,

,:G
,_t.*:>_- 15
C a.. %§

m El SKb

-30 " ~ '

0 5 10 15
a Time (h)
100
~ " ~+K
+++~+++++~ ~ "++++++++++-
75

§
+'~ 50
+ ?.,~,

Cg 25
~~ 0

~, ~ -25
SKb
m o -50
-75 I 9
0 5 10 15
b Time (h)
Fig. 7a and b. Trajectory sensitivityof the SML batch model: a Substrate and related sensitivities,
b biomass and related sensitivities
Modelling, Identificationand Control of the ActivatedSludgeProcess 107

batch experiment and that large estimation errors are generated. The trajectory
sensitivity analysis is now applied to the reduced-order model (4), (5) in order to assess
its practical identifiability. The system matrices appearing in the sensitivity system
(53) are the following

-KbX -KbS
af/aZ = K~X + KmX2/S 2 KcS - 2KmX/S (54)

af/ap = -SXo sxO _ 2 2 / s (55)

The trajectory sensitivities obtained by substitution of matrices (54), (55) into Eq. (53)
are shown in Figs. 7a and 7b and can be compared with the corresponding Monod
curves of Figs. 8a and 8b. By inspection two main differences can be noticed:

30 ++++++++++++++++
20 - ++++++~Sy

-10! \ /
~ -2ol \ /
"6 o -30

- 0 5 10 1;
a, Time (hi

3O Sf~
7~
+E"
2+.>_- 20
T"~
E-~ 10

.2 -~ 0
SKd
-1C i t -- I

0 5 10 15
b Time (h)

Fig. 8 a and b. Trajectorysensitivityof the Monod batch model: a Substrate and related sensitivities.
b biomass and related sensitivities
108 S. Marsili-Libelli

1) In the SML model the sensitivities of S are not all proportional, as in the Monod
case, thus implying limited parametric interdependence. As noted by Holmberg 4o),
it is the intercorrelation of ~t and Ks that makes their identification difficult. With
the SML model, this problem does not arise.
2) The biomass sensitivities of the SML model do not exhibit a sharp peak as with
the Monod kinetics 40,46). Therefore this model is expected to take better advantage
of the information provided by the entire data set.
It can be concluded that numerical calibration of model (4), (5) from batch data is
feasible and numerically reliable. This conclusion is independent of the particular
numerical method being used since it refers solely to the structural properties of the
model.

3.3 Practical Model Identification

The preceding Sects. 3.1 and 3.2 have assessed the SML model identifiability, in other
words it was demonstrated that the structure of the batch model Eqs. (4), (5) is such
that their parameters can be determined on the basis of a suitable number of experi-
mental data. Now, the numerical accuracy of the estimates will be assessed using
different numerical methods. In particular, it is important to check how the model
parameters interrelations affect the estimation reliability. The first attempt to estimate
the parameters of the SML continuous-flow model was proposed by the author 14)
using the error equation approach and taking advantage of the fact that this model,
unlike the Monod kinetics, is linear in the parameters.
Let {Sexp(i), Xexp(i); i = 1, 2,..., N} be a collection of substrate and biomass
experimental data. If they are substituted in the process Eqs. (36), (37) the following
set of equations are obtained
BOD."

dSexp(i)/dt = --KbSexp(i) X xp(i) -- q(1 + r) S xp(i) + qSi + el(i) (56)

MLSS:

dXcxp(i)/dt = KcSexp(i)Xexp(i) -- KmXZxp(i)/Sexp(i)

q(1 + r) Xexp(i ) -}- qrX r + ~2(i) (57)
for i = 1,2 . . . . . N

where g~(i) and az(i) are the equation errors induced by imperfect matching between
the experimental data and the model equations. Supposing that the plant operating
factors {q, r, Xr, Si} are known, the best fit of experimental data is obtained with the
parameter set {l~b, I~c, I~m} which minimizes the sum of the squared equation errors
~2(i) and g2(i). The parameter identification then gives rise to the following optimiza-
tion problem:

N
min ~ ~2(i) + ~2(i) (58)
Kb, Kc, Km i= 1
Modelling, Identificationand Control of the ActivatedSludge Process 109

or, equivalently,

N
min ~ ([(dS~xp(i)/dt)- (dS/dt)] 2 + [(dXoxp(i)/dt)- (dX/dt)] 2} (59)
Kb, Kc, Km i= 1

In principle this approach seeks to minimize the sum of quadratic errors between the
experimental and model derivatives rather than the process variables S and X. Taking
advantage of the parametric linearity of Eqs. (56), (57), the optimization problem
(59) is in fact a linear least-square problem and therefore can be solved analytically
through the intermediate quantities
N
A~ = ~ S~o(i ) X2x.(i) (60)
i=l
N
A,_ = ~ X~xl,(l)/S~xpO)
4 9 2 . (61)
i=1
N
A3 = ~, X2~p(i) (62)
i=l

N
Y1 = ~ [Sexp(i) Xexp(i) (dSexp(i)/dt)
i=1
-t- q(l + r) 2 . Sexp(i) -
Xexp(l) qXrXexv(i) Sexp(i)] (63)

Y 2 = i =~l
{ ~Xe3xp(i)
q(1 +r)S~xp(i [ q r X r q- ( d X e x p ( i ) / d t ) ]
X:xti)
Sexp(i ) J (64)

then the following estimates are obtained:

N
I~b = ~ [qSiSexp(i) Xexp(i) - q(l + r) S2xp(i) X~xp(i)
i=l

-(dSoxp(i)/dt) Soxgi) Xoxgi)] A; 1 (65)

I(, = (Y1A2 -- A3Y2) x (A1A2 -- A2) -1 (66)

K m = (Y2A* -- A3Y1) x (A1A3 A2) - a (67)

This procedure, though conceptually very simple, conceals some pitfalls. First, the
equation errors q(i) and 82(i ) are nonlinear functions of the data errors and therefore
can produce biased estimates. Secondly, the derivatives of the experimental quantities
dSexp(i)/dt and Xexp(i)/dt appearing in the estimator Eqs. (60)-(67) should be approxi-
mated with a suitable numerical technique to reduce the error amplification typical
of numerical derivatives. In this application, the raw data were smoothed using least-
square approximating splines 47) and both the order of the spline and their knots
locations were used as smoothing parameters to reduce the experimental errors. This
technique, described in details in a,), gives adequate results as shown in Fig. 9 where
a set of 17 hourly BOD data was used to calibrate the model. Several sets ofexperimen-
110 S. Marsili-Libelli

30
.o-_~

..- -- --.o.. "o

20
131
E
co 10
..... experimental
- - simulation
, I , I I F ! I i I i I

2 3 4 5 6 7 8 9
T i m e (hi

30

,' , , ,, ,, /

v'
E
u~ 10
..... experimental
- - simulation

0 I , q , r , I , lrl , 113 , 115 , I

3 5 7 9 17
b T i m e (h)

Fig. 9a and b. SML model response to a varying BOD input. Model calibrated using the estimator
Eqs. (60)-(67) with splines approximation of the derivative i4): a Municipal medium-scale plant,
b pilot-plant

Table 2. SML model parameter obtained from

operating plants 14)

Parameter Value Units

Municipal medium-scale plant:

Kb 7.413x 10 .4 mg -1 1 h -1
Kc 1.482x10 3 mg 11h-1
Km 5.213 • 10 .4 h -1
Pilot plant:
Kb 6.134x 10 -4 mg -I 1 h -1
K 1.935x10 -3 mg l l h -1
K m 5.023 X 1 0 - 4 h -1

tal d a t a were t a k e n f r o m a pilot-plant o p e r a t e d by the m u n i c i p a l i t y o f M o d e n a and

f r o m several m e d i u m - s c a l e c o m p l e t e l y m i x e d plants o p e r a t e d by the m u n i c i p a l i t y
o f F l o r e n c e . T h e p a r a m e t e r c a l i b r a t i o n results are s u m m a r i z e d in T a b l e 2. Their
values c o n f i r m that the efficiency o f the p l a n t is p r o p o r t i o n a l to its size. In fact, recall-
ing that K b and K c are related to the ability o f the b i o m a s s to degrade o r g a n i c substrate,
the m e d i u m - s c a l e plants exhibit a higher v a l u e o f this p a r a m e t e r . O n the o t h e r hand,
efficiency c o m p u t a t i o n carried out on the s a m e plants s h o w e d that while the efficiency
 Modelling, identification and Control of the Activated Sludge Process 111

of the pilot plant was only 80 %, that of the medium-size plants was above 90 %.
Further, the ratio Kb/K ~ is both conceptually and numerically very close to the "Yield
Factor" Y used in the M o n o d kinetics to indicate the transformation efficiency of
substrate. F r o m the values of Table 2, the ratio Kb/K ~ is 0.5 for the medium-size
plants and only 0.33 for the pilot plant, again confirming that size affects treatment
efficiency.
The substrate/biomass continuous-flow model (36), (37) was obtained adding the
pertinent hydrauJic mass balance terms to the batch kinetics already specified as
Eqs. (4), (5). Since these terms can be estimated independently, it may be practical to
calibrate the kinetic part independently from the hydraulic balance, which can be
determined by other simple means. In other words, the kinetic parameters {Kb, Kc,
Km} can be determined by analyzing a suitable batch reaction carried out by the already
acclimated biomass, irrespective of the hydraulic operating conditions. This approach
represents a viable alternative to the previous one, as the minimization problem now
can be restated as follows

N
min E~ = ~ [S(i) - S~p(i)] z + [ X ( i ) - X~xv(i)]z (68)
{Kb, Kc, Kin} i= 1

with Eqs. (4), (5) as optimization constraints:

dS/dt = - - K b S X (4)

dX/dt = KcSX - - K X2/S (5)

As already pointed out, a typical difficulty in calibrating microbial growth models

is caused by the interrelation between kinetic parameters. This is particularly apparent
with the pair {fi, Ks} of the M o n o d kinetics, as thoroughly discussed by Holmberg
and Ranta 42l. In fact, if these two parameters are varied so as to obtain a constant
value E*x of the functional (68), then very narrow and elongated closed contours are
obtained in the {fi, Ks} plane. As a result, the optimal point situated at the center of

0.100
1.0
0.075
t 0.9
0.8
0.050
0.7
0.025
0.6

0 0.5 , , , , , , ,
0 o. s 0.10 0.15 0.20 0.25 o. 0 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4
a Kb - = b K b ~

Fig. 10a and b. Error level contours of the SML model as a function o f K b and Kc : a E~xfrom Eq. (68)
with both variables S and X available, b Er from Eq. (69) with only dissolved oxygen C available
112 s. Marsili-Libelli

this narrow "valley" is reached with great difficulty by numerical optimization al-
gorithms, which may converge poorly, produce unreliable estimates or fail altogether.
It is therefore appropriate to study the shape of the cost functional (68) when connected
to model (4), (5) and ensure that such a situation is avoided. Figure 10a depicts the
level contours of Esx as a function of {Kb, Ko} parameters, showing that the elongation,
caused by the interrelation between parameters, though present, is quite acceptable.
Further, to avoid the numerical derivative problem non-gradient methods were used
such as the flexible polyhedron algorithm 48,49) in which the basic search procedure
has been improved by optimizing the pattern search adjustments so). This enables the
algorithm to cope successfully with the "narrow valley" problem resulting in a highly
efficient search. In fact, the modified algorithm can adjust the search parameters to
the shape of the functional, directing the search along the trough main direction and
selecting a projection length which is optimal for each particular step. As a practical
result, if a mixed liquor sample is taken out of the plant oxidation tank and its oxygen
consumption progress is analyzed in a stand-alone batch, the substrate/biomass
interaction will be the same as in the previous continuous-flow arrangement, since the
microbial colony is already acclimated to that particular substrate, but with obvious
practical advantages.
The identification problem becomes more difficult if not all the biological quanti-
ties are experimentally available. A practical case is when only dissolved oxygen data
are available instead of substrate and biomass data. Then, the dissolved oxygen dy-
namics (9) has to be included in the model and consequently the optimization problem
(68) becomes

N
min Ec = ~ [C(i)- C~xp(i)]2 (69)
{Kb, Kc. Km, Ye} i= 1

with the following model as optimization constraint:

dS/dt = --KbSX (4)

dX/dt = K SX -- KmX2/S (5)

dC/dt = ~, KbSX + YeKmX2/S (70)

Figure 10b shows that this new arrangement produces a narrower trough in the E
contour portrait. Now, the identification problem (69) is that of fitting model (4), (5),
(70) with only oxygen measurements available. Although identification is numerically
more difficult, nonetheless the structural properties previously demonstrated still
hold and the problem (69) has indeed a practical solution. This is precisely the situation
encountered when determining the BOD content of a sample through the well-known
respiration test, as described for example by Bhatla and Gaudy 51). This kind of curve
(usually termed "exertion curve") exhibits two peculiar features: the incubation
delay, caused by the initially slow growth of bacteria, and the "plateau" when the
microbial colony becomes mature and endogenous metabolism prevails over synthesis.
The pertinent model for oxygen utilization is precisely Eq. (70) as shown in 52), because
in this kind of experiment no additional oxygen is supplied throughout the test and
Modelling, Identification and Control of the Activated Sludge Process 113

350

~;
300 It

250

'--" 200
E
150
E3
0 to set
m 100
det (4, 5,91
50 onod model

0 I
0 25 50 75 100 125
Time (hi

Fig. 11. Fitting batch oxygen respiration data 51) with the Monod model Eqs. (1), (2), (71) (thick line)
and SML reduced-order model Eqs. (4), (5), (70) (thin line)s2)

therefore the diffusion term KLa(Cs,t -- C) of Eq. (9) must be omitted. In other words,
the measured quantity is indeed the integral of Eq. (70) over the test period (usually
lasting five days). Both the SML model (4), (5), (70) and the classical Monod model
were calibrated using the experimental BOD respiration data from 51). For this
purpose, the Monod model was rewritten as

1
dS/dt = y ~t(s) x (1)

dX/dt = g(S) X -- KdX (2)

1
dC/dt = Ys~g(S) X + 7 K a x (71)

where g(S) is again provided by Eq. (3). Figure 11, where the two model responses are
compared, shows that the SML model (4), (5), (70) fits the experimental data better
than the Monod model (1), (2), (71). The numerical values of the model parameters
are grouped in Table 3. It can be noticed that the Monod model fails to reproduce the
plateau adequately and that the initial incubation delay is grossly missed. As a general
remark, it appears that the sigmoid response of the Monod kinetics has poor flexibility
and it is not amenable to fitting the respiration curve properly. This is also confirmed
by the massive estimation error in the respiration coefficients 7s and 7e, which are
almost unidentifiable. Some other parameters too show a considerable estimation
inaccuracy, notably K s. This result agrees with Holmberg ~o) who reported estimation
experiments in which the standard deviation of the estimated K s is of the same order
of magnitude as the parameter itself. This test further demonstrates the difficulties
in using the Monod kinetics and the advantage of the simpler SML model.
114 S. Marsili-Libelli

Table 3. BOD exertion curve model parameters 52)

Parameter Range of values Units

l) Monod model:
12 0.343 _+ 0.082 h -1
K~ 15.696 + 5.232 mg 1-1
Y 0.583 + 0.151 --
Ka 0.2,83 + 0.707 h -1
7s 0.547 + 2.293 --

7e 0.777 _+ 3.238 --

So 298.03 + 43.5 mg 1-1

Xo 1.63 _+ 0.89 mg 1-x
2) SML model:
Kb (7.261 __+0.658)x 10-4 mg-X1 h -I
Kc (10.784 + 0.287)x 10-* mg -1 1 h -1
Km (8.973 + 2.569) x I 0 - 4 h -1
7s 0.511 _ 0.076 --

u 0.241 + 0.026 --
So 299 __+ 14.3 mg 1-1
Xo 3.3 ___0.54 mg 1-1

3.4 On-line Parameter Identification

So far off-line methods have been considered for the identification o f the microbial
kinetics. F o r m a n y biotechnological control applications, though, it is i m p o r t a n t to
obtain real-time p a r a m e t e r estimates as the process unravels. The most widely used
a p p r o a c h is based on the "extended K a l m a n filter" ( E K F ) a p p r o a c h coupled with
the M o n o d kinetics (1)-(3). The general structure o f the E K F algorithm is described
in a number o f outstanding books, notably Jazwinski 53) a n d Gelb s,), and an early
biotechnical application was reported by Svrcek et al. 52). The application o f the E K F
algorithm to the microbial kinetics is particularly relevant because both the process
variables (substrate and biomass) and the related parameters {ft, Ks, Kd, Y} are usually
u n k n o w n a n d / o r time-varying. The E K F a p p r o a c h considers the u n k n o w n parameters
as additional process variables and seeks to estimates the extended vector (hence the
E K F name) of variables by filtering the differences between model outputs and ex-
perimental measurements as soon as these become available. The applicatio n of
E K F to the S M L model 56) is now considered. Let the extended state vector be defined
as
= [S, X, Kb, Ko, Kmt (72)

It consists o f the two process variables (substrate and biomass) plus the three kinetic
parameters Kb, Ko, K ~ I f the system parameters are assumed to be constant in time,
their dynamics can be expressed as d{i/dt = 0 with i = 3, 4, 5. Using the new notation
(72) in Eqs. (4), (5) yields

d ~ t / d t = --~t~2~a --,q(1 + r ) ~ t + qS~ (73)

d~2/dt = ~.~ -- q(1 + r) ~2 - - ~ 5 ~ 2 / ~ i + rqXr (74)
d~3/dt = d ~ / d t = d~5/dt = 0 (75)
Modelling, Identification and Control of the Activated Sludge Process 115

C o n s i d e r i n g that the u p d a t i n g m e a s u r e m e n t is available at discrete intervals h, two

kinds o f recursive equations are needed to mechanize the c o n t i n u o u s - - discrete filter:
1) E x t r a p o l a t i o n o f system Eqs. (73)-(75) between m e a s u r e m e n t s at time t - h
and t. D u r i n g this interval no external i n f o r m a t i o n is available a n d both the system
variables a n d the associated c o v a r i a n c e m a t r i x need to be e x t r a p o l a t e d . The
quantities ~i(~ ] t - - h) d e n o t e the i-th estimated variable at time t - - h =< z < t
based on the m e a s u r e m e n t s up to time t - - h, whereas the system d y n a m i c s f(.) is
linearized for c o m p u t i n g the c o v a r i a n c e m a t r i x i.e. A = 8f/8~.
Filter extrapolation:

d~i(z It - - h)/dt = f(~(z It - - h), Si(~)] ; (76)

covariance matrix extrapolation ."

dG(z It - - h)/dt = A G ( z ] t - - h) + G(z It - - h) A T + W (77)

2) U p d a t i n g o f filter quantities (state variables and covariance matrix) are p e r f o r m e d

as soon as the new m e a s u r e m e n t becomes available at time t.

Filter update:

K t = G ( t I t - - h) cX(cG(t I t - - h) e x + R) -~ (78)

~i(t ] t) = ~i(t I t - - h) + KT(mt - - c~i(t I t - - h)) " (79)

covariance matrix update:

C ( t l t ) = G ( t ] t - - h) - - K r e G ( t l t - - h) (80)

1.5 ~:c
xlO -3

1.0

0.5 Kr~

0 l I l l
0 2.5 5.0 7.5 10.0
Time (h)

Fig. 12. Extended Kalman filtering of the SML model. Substrate and biomass data from the municipal
medium scale plant were used to estimate the three parameters {Kb, K , Kin}. The dashed lines indicate
the parameter values obtained with the off-line algorithm Eqs. (60)-(67)
116 S. Marsili-Libelli

where m t is the t-th measurement and Kt is the corresponding Kalman gain used to
update the filter. Likewise, G(t I t - - h) and G(t [ t) represent the covariance matrix
before and after the updating of the t-th measurement. Lastly, W and R are covariance
matrices depending on the statistical characteristics of the data, and e is a vector
relating the filter state variables to the output (measured) variable. The algorithm
(73)-(80) was applied to the SML model to estimate the three system parameters
{Kb, Kc, Kin-} in addition to the two system variables {S, X} using the same experimen-
tal data set from the municipal medium scale plant from which the estimates of Table 2
were obtained. The estimation results are shown in Fig. 12. It can be seen that the EKF
algorithm converges to the same values which were obtained with the off-line algorithm
Eqs. (60)-(67) with the difference that in this arrangement the measurements are
processed one at a time, hence this algorithm can be used for on-line process monitor-
ing. A further E K F application to the SML model is reported in 56) where dissolved
oxygen experimental data are used to estimate four system parameters {Kw Kc,
Kin, Kea } together with the three process variables {S, X, C}.
In this limited information experiment the E K F approach is still feasible, as discussed
in 56), although its estimation performance is slightly degraded. Fig. 13 shows the
estimation results for this case, in which an experimental set of 48 hourly DO data
was used. Though the convergence is slower, a suitable choice of the filter parameters
W and R produces stable and reliable estimates.

8.0 2.0-

7.5
t 1.6-___~
7.O ~ 1.6-
"4
t
O C)
6.5- x 1.4-
x

<:~ 6.0- 1.2-

i i i

o ' 28 o "&
Time (h) Time (h)
5.6 3.6

2.8

2.4
<~
2.0

, , ~ i ' ' ' "8

4
0 24 48 0 24
Time (h) Time (h)

Fig. 13. Extended Kalman filtering of 48 hourly dissolved oxygen data to estimate the four system
parameters {Kb, Ko, K m, K~a} s6~
Modelling, Identificationand Control of the ActivatedSludge Process 117

3.5 On-line Estimation of Bioactivities

It has already been shown that oxygen uptake rate (OUR) is a key process quantity,
from which valuable information on the entire process can be inferred. Then it is not
surprising that the estimation of this variable has received considerable attention. The
importance of OUR as an indicator of bioactivities was first assessed by Olsson and
Andrews 18) and Stenstrom and Andrews 29) and several results became available soon
after regarding the estimation of OUR and its companion parameter KLa, indicating
the ability of the mixed liquor to accept oxygen from the adjacent gaseous phase. Given
the dissolved oxygen dynamic balance

dC/dt = KLa(Csat -- C) -- OUR (9)

solving for OUR would appear a straightforward task, if it were not for the uncertainty
on KLa (either unknown or time-varying) and the numerical difficulty in obtaining
a reliable approximation of dC/dt. If either or both problems can be tackled inde-
pendently, then partial solutions to Eq. (9) can be found, as described by Holmberg 39,
42), Marsili-Libelli 57), Goto and Andrews 58), Howell and Sodipo 59). The common
endeavour is to recast the non linear estimation problem in the context of linear
filtering theory and apply meaningful results, such as the extended Kalman filter 53.
s4). Holmberg 39) addressed the problem thoroughly and proposed two alternative
approaches. If KLa is known through prior experiments and DO is controlled to a
given set-point, then Eq. (9) yields a straightforward solution. Conversely, if the DO
level is allowed to vary and if KLa is expressed as a function of the air flowrate, then a
recursive least-squares estimator can be set-up to estimate both OUR and KLa. How-
ever, expressing KLa as a function of the air flow produces biased estimates, unless
very special test conditions are met. Another recent result 58) is still based on a pre-
liminary off-line determination of KLa from which OUR is obtained through Eq. (9)
either with suitable numerical approximation of dC/dt or by time discretization.
Holmberg and Olsson 6o) describe a simultaneous estimation scheme for K~ a and
OUR based on a linear Kalman filter and taking advantage of the differing time scale
of the two variables: in fact the oxygen utilization may vary significantly in a matter
of hours, whereas Kia variations can be detected over several days. Moreover,
care is placed in approximating the derivative dC/dt. The DO dynamics can be put
in a simpler form, choosing the dissolved oxygen deficit (DOD) C(t) as a new variable

dD(t)/dt = --KLaD(t ) + OUR(t) + q(l + r) C(t) (81)

sampling Eq. (81) at h intervals, yields the discrete-time equivalent

D(t + h) = aD(t) + bOUR(t) + bq(1 + r) C(t) (82)

where

1
a=exp(--hKLa ) and b= (l--a) (83)
KLa
118 S. Marsili-Libe[h

provided that both OUR and C do not vary appreciably during the sampling interval
h. If KLa is known a priori, then Eq. (82) can be solved to yield OUR

OUR(t) = l{D(t + h) -- aD(t)} -- q(1 + r) C(t) (84)

Otherwise, if both OUR and KLa have to be estimated concurrently, the bilinear
nature of Eq. (82) prevents a successful application of the classical recursive least
squares. In fact, parameter b is a function of KLa and the term b OUR is the product
of the two estimated quantities. This difficulty, which has not been sufficiently high-
lighted in the literature, can be circumvented by expanding a from Eq. (83) in power
series, retaining only the linear term and substituting in b to obtain b _~ h. In this way
decoupling is achieved and recursive least squares can be applied to the approximate
system

D(t + h) = aD(t) + hOUR(t) + hq(1 + r)C(t) (85)

From Eq. (85) the estimation error equation can be determined as

z(t) = p(t) x f(t) + e(t) (86)

where e(t) = [ p - li(t)]~ f(t) is the error induced by parameter mismatch and the
following quantities are defined.

Parameters: p(t) x = [aOUR(t)] ; (87)

representers: f(t) = [D(t)h] T ; (88)

predicted output: z(t) = D(t + h) -- qh(1 + r) C(t) (89)

with the symbol (T) denoting transpose. The estimated parameter vector p(t) can be
updated according to the following recursive least squares scheme

lJ(t) = lJ(t -- h) + G(t) f(t) e(t) (90)

1 I G(t) fT(t) G(t) 1 (91)

G(t + h) -- L ~ ) G(t) - L~(t)/L2(t) + fT(t) G(t) f(t)

The two parameters Ll(t ) and Lz(t) appearing in the covariance matrix update Eq. (91)
are such that Ll(t)/L2(t ) = constant and Ll(t) is adjusted in such a way to make
t r a c e [G(t)] = constant. In this way, suggested by Landau and Lozano 61), G(t)
never vanishes thus ensuring that time-varying parameters will be tracked. This is
a crucial requirement because both parameters, but especially OUR, are time-varying
quantities. Several applications of the above scheme appeared in the literature already
quoted. However, the main pitfall of this approach is that the tracking capability of
the basic least squares estimator is inadequate to follow the short-term variations of
OUR with sufficient accuracy. Including a constant term (h) in the representer vector
Modelling, Identificationand Control of the Activated Sludge Process i i9

f(t) and pretending to estimate OUR as the corresponding time-varying parameter

is a very crude approach, resulting in biased estimates. Even worse, the error induced
by the wrong representer choice reflects on the constant parameter also, hence KLa
too is incorrectly estimated. A partial solution to this problem, as suggested by Clarke
and Gawthrop 62) is to select a sampling period h much shorter than the fundamental
period of OUR. Further, it is dOUR/dt which ultimately determines the tracking
limit, and therefore the amplitude too has to be taken into account.
A more robust approach for joint KLa and OUR estimation, based on expressing
OUR as a power series of time, can be developed along the guidelines suggested by
Xianya and Evans for tackling inaccessible disturbances 63). Let OUR be represented
by a time-series approximation

OUR = bo + bit (92)

where ,the coefficients {b0, bl} are constant. Then the estimation scheme Eqs. (90),
(91) can be used with the following parameter and representer vectors

pT = [a bo hi] (93)

f(t) = [D(t) ltl] T (94)

where the new time variable t~ is defined within one sampling interval h, as

t1 - - = t - ( k - 1 ) h for ( k - - 1 ) h__<t < kh (95)

where k is an iteration counter. To avoid large numbers in the representers vector and
to adjust the linear approximation as time progresses, the estimates are updated as
follows

bo(kh) = bo(kh -- h) + hb,(kh -- tl) (96)

b,(kh) = b~(kh -- tl) (97)
Between two adjacent resets at times (k -- 1) h and kh the inaccessible quantity can
be expressed by the linear approximation

Ol21R(t) = 6o[(k -- 1) hi+ t l b l [ ( k - 1)h] for (k - 1) h __<t < RT (98)

It might be argued that a higher order approximation including higher powers of t

would give a better approximation, but two considerations suggest to keep the order
low. First, increasing the order would implyestimating more parameters, resulting
in a greater computational burden. This would in turn require a longer resetting period
and therefore large numbers, likely to cause numerical problems. Figure 14 shows the
joint estimation of OUR and a time-varying KLa over a 24-h period using the SML mo-
del in connection with the above algorithm Eqs. (92)-(98) and Square-Root 64, 65) fac-
torization to guarantee positive definiteness of the covariance matrix G. The basic
sampling interval is 3.75 min and resetting occurred at every four samples, i.e. at
15 min intervals.
120 S, Marsili-Libelli

30
OUR

2
__ZK" 2:

2
I

0
0 6 12 18 2/, 30 36 42 1.8
Time lh)

Fig. 14. Performanceof the estimator Eq. (98)using the linear piecewiseapproximationEqs. (92)-(97)
for the reconstructionof the inaccessibleinput OUR and of a time-varyingoxygentransfer coefficient
KLa

3.6 On-line Estimation of Process Variables

Having established the necessary link between dissolved oxygen measurements and
OUR, the analysis can be taken even further with the design of a state estimator to
make for the unavailability of direct process measurements. Joint state and parameter
estimation has been pursued by a number of researchers using the augmented state
equations to derive asymptotically stable observers. The EKF approach can in fact
be viewed in this light. Conversely, Aborhey and Williamson 3s) derived a parameter
estimator for the Monod kinetics under the assumption that both substrate and bio-
mass measurements were available. Though of system-theoretic value, these results
are of course not directly applicable, as the availability of either or both state variables
is at present unrealistic. By contrast, Holmberg and Ranta 42) proposed an indirect
state estimator based on the oxygen uptake rate measurement and Monod kinetics,
expressing OUR as a linear combination of synthesis and maintenance metabolisms

OUR = ~g(S) X + ~ X (99)

where ~ and ~ are respiration coefficients similar to y~ and y~ in the OUR Eq. (8)
and ~t(S) X is the microbial growth function. If the microorganism kinetics is given
by the linear combination of growth and death processes

dX/dt = (NS) -- K d) X (100)

then solving Eq. (99) for ~(S) X and substituting in Eq. (100) yields a biomass estimator
based on OUR measurements
Modelling, Identificationand Control of the Activated Sludge Process 121

40

30
O~
E

2o

r~

~0

0 i i 1 112 i i I I ,
0 24
Time Ihl
5

c
\
O~

u~
o~
13 /
E /
.o_
m - - Irue Fig. 15a and b. Performance of the state
--- estimated estimator Eqs. (102), (103): a Estimated
I I I I I I I I I I substrate g, b estimated biomass R
12 24
b Time (h)

The above derivation was based on a linear decay term --KaX, and resulted in a
linear biomass estimator independent of the corresponding substrate dynamics.
A similar reasoning based on the SML model (4), (5), (9) can be carried out using the
pertinent OUR definition (8), repeated here for convenience

OUR = 7KbSX + 7eKmX2/S (8)

Solving for the growth t e r m K b S X and substituting in the microbial dynamics Eq. (5)
yields the substrate/biomass observer

dS/dt = --KbSI~ -- q(1 + r) g + qSi (102)

dX/dt = --Kin(1 + %/%)X2/S + O U R / % - q(1 + r)J~ + qrX r

(103)

Figure 15 shows the observer performance using DO data and assuming that OUR.
is directly available. If instead this variable is obtained from the algorithm just describ-
ed by Eqs. (92)-(98) the accuracy is only slightly decreased. A sensitivity analysis of
the observer Eqs. (102)-(103) is carried out in Refs. 57) and s6). It was found that the
biomass X is little affected by parameter inaccuracies, whereas the substrate S is
more sensitive to variations of Kb, which on the other hand is the easiest to obtain
122 S. Marsili-Libelli

experimentally or from literature data. Conversely Km, the most difficult to tackle
and the crucial one in the observer, has a much more limited influence.
As a concluding remark on this section, it can be stated that real-time parameter
identification and reconstruction of process variables play a key role in activated
sludge process control. In fact, given the inherent difficulty of obtaining direct process
information, computer processing of simple measurements seems the only practical
way to obtain an updated picture of the process development.

4 Process Control
The previous sections were concerned with modelling and identification of the activated
sludge treatment process. These topics, in addition to having a scientific value in their
own right, pave the way toward advanced process control needed to upgrade the per-
formance in terms of reliability, treatment efficiency, and energy conservation. To
achieve these goals the activated sludge process control system would have to over-
come the following obstacles:
a) large variations of input organic load;
b) variations of sludge inventory and hence of kinetics;
c) limited availability of on-line process measurements.
These three aspects have already been addressed to justify the development of process
models of limited complexity and their use for real-time estimation and control. It
is the purpose of this section to survey the trends in activated sludge process control
and to use the previous results for the design of efficient real-time controllers.

4.1 Process Performance Indicators

Most activated sludge plants operate in a time-varying environment and their operat-
ing conditions may be significantly different from the original design specifications.
This may reduce treatment efficiency unless the plant operation is continually adjusted.
Furthermore, operating costs have to be kept as low as possible without impairing
the average effluent quality. Automatic plant control can be used to improve the
operational performance and increase flexibility at low cost, but given the complex
plant dynamics and differing time horizons, care must be placed in avoiding control
actions which seek to improve the efficiency in the short term at the expenses of long-
term performance. This brings into focus the need to design control laws which guaran-
tee not only a consistent effluent quality, but also safeguard plant operation in the
long run. For example, such process failures as "sludge bulking" (i.e. poor sludge
settling quality), anoxic conditions or secondary settler denitrification may be caused
by excessive aeration, possibly put into action to bring about a temporary improve-
ment. Before undertaking any control law design, some meaningful biological process
indicators must be defined. The most widely used of such indicators were first introduc-
ed by Lawrence and McCarty 66). They are the food-to-mass ratio Fro:
24
f (Si - S) dt
substrate mass utilized in 24-h o
Fm = bacterial mass in the aerator VX (104)
Modelling, Identification and Control of the Activated Sludge Process 123

and the mean cell retention time, or sludge age, 0o

bacterial mass in the aerator VX 0X

0c = bacterial mass wasted daily wQX r wX r (105)

Where 0 = q-1 is the hydraulic retention time. These two quantities, being derived
from steady-state assumptions, are normally used for activated sludge plant design,
but are of little value during time-varying operation, when the importance of maintain-
ing a prescribed growth rate has been demonstrated by Garrett 67) and a number of
other investigators. The following relation exists between 0c and growth rate, if the
Monod kinetics is assumed,

0~ 1 = It(S) -- K a (106)

However, this cannot be used for real-time control since 0c is defined as an average
quantity, normally on a 24 h basis. Bisogni and Lawrence 6s) have shown experimen-
tally that growth rate as measured by 0c is functionally related to specific oxygen uptake
rate (SCOUR), defined as the ratio of oxygen uptake rate (OUR) to the biomass con-
centration in the aerator. To obtain a relationship between SCOUR and growth rate it
is not necessary to write a material balance around the process nor is it necessary to
assume steady-state conditions. SCOUR relates directly to growth rate as follows 19).

OUR (1 -- Y)
SCOUR - X - Y g(S) + 7eKa (107)

4.2 Conventional Control Strategies

The following section will revise the current developments in control design for the
activated sludge process and adapt some well-established control theory results such
as the proportional-integral-derivative (PID) controllers to this particular process.
Early literature contributions have applied the direct optimization methods to the
Monod kinetics obtaining a nonlinear optimal regulator 69-71). Though a very elegant
solution results, the complexity of the ensuing control law makes their implementation
difficult.

4.2.1 PID Control

As far as practical regulators are concerned, the proportional-integral-derivative
(PID) controller is the most widely used in wastewater engineering. Its properties and
design techniques are well known 72-74). The relationship between PID regulator
input e(t) = ysp(t) -- y(t) and its corresponding output u(t) is:

u(I) = Kp e(t)+~o ' i e(c~)dcy+ Td dde.l-]

t-~ (lo8)
124 s. Marsili-Libelli

The three coefficients {Kp, T i, Td} are usually referred to as proportional gain, reset
time, and lead time respectively. The inverse of the reset time T~ is referred to as the
reset rate and expresses the time after which the integral action duplicates the propor-
tional action. The derivative term provides a rough forecasting of error trend and
as such enables preemptive control action. The lead time T d is the time interval by
which the derivative action anticipates the effect of the proportional control.
When a plant is controlled by a digital computer, the control law Eq. (108) is usually
implemented in incremental discrete-time, with sampling interval h, form to avoid
numerical overflow. In this case the full values of the controlled variable y(t) and of
the set-point y~p(t) are used and the integration is performed outside the incremental
computation loop. Then the sampled PID is implemented in two steps :
Incremental control signal computation

6u(t)=Kp~ysp(t)-Kp 1 +T~+ y(t)

(T~)
+K v 1+2-~
Td
y(t-h)-Kp~-y(t-2h); (109)

full control signal computation

u(t) = u(t -- h) + 8u(t) (1lO)

4.2.2 Approximate Optimal Control
The efficiency of the PID controller Eqs. (109), (110) depends on a sensible selection
of the three coefficients { K , T i, Td}, Many design techniques exist for linear systems
72-74) which could be applied in this context if the process dynamics is linearized and
possibly complemented with some performance criterion. Let the process kinetics
be approximated by a set of linear discrete-time state and output equations :

State equations: x(t + h) = Hx(t) + gu(t) ; (111)

output equation: y(t) = ex(t) (111 a)

where the substrate (S), biomass (X), and dissolved oxygen (C) are assumed as state
variables, i.e. x = [S X C] r. The vector e depends on the selected output variable of
the process and h is the sampling interval. The performance index to complement the
linearized dynamics Eqs. (110), (111) may be defined as a weighted sum of the regulator
inputs e(t) and outputs u(t):

1-Ic Hc

F = ~ {y2(t) + Qu2(t))~ = ~ {xV(t) e rWex(t) + Qu2(t)J (112)

t=0 t=0

where H c is the control horizon, i.e. the number of control steps over which the con-
trol action is to be optimized, and W > 0 and Q > 0 are appropriate matrix and scalar
weights to place the proper emphasis either on control accuracy (W) or control effort
Modelling, Identificationand Control of the ActivatedSludge Process 125

(~). The controller parameters are to be determined in order to minimize the perform-
ance index Eq. (112). Several options are available at this stage:
a) A linear feedback is selected u(t) = Fx(t) and the matrix F is determined in an
optimal way from the Riccati equation associated with the F criterion and the
system Eqs. (110), (111) 72).
b) A controller with known structure is selected and its parameters are determined
by numerical means in order to minimize F, in much the same way as system para-
meters identification in Sect. 3.
It can be shown that between the two approaches there are deep similarities 75) and both
have been applied to activated sludge control problems 76-ao). Given the linearized
dynamics Eqs. (110), (111) and the quadratic performance index Eq. (112), to enhance
the controller robustness integral action is incorporated into the system introducing
the integral state 1"1defined as

q(t + h) = q(t) + y(t) (113)

The augmented system state vector is Z = [ x r l ] T and its dynamics is obtained aggregat-
ing the state equations Eqs. (110) and (113) to yield the following matrices

A* = ; b* = 9 e*=[-c 0] (114)
C

The resulting Linear-Quadratic-Integral (LQI) 72, 75) control law is then of the form

u(t) = --Flx(t ) - - fzrl(t) (115)

If an infinite control horizon is assumed (i.e. H, ~ oe), the matrix F = [F1, f2] can
be obtained through the backward recursion.

Fj = (Q + b*TPj+lb*) -1 b*Wp j + l A* (116)

Pj = A*Tpj+I(A* -- b ' F ) + W (117)

which under very mild conditions on {A*, b*, c*} converges to a non-negative-
definite limit P* > 0 and gain F* to be used in Eq. (115) representing the optimal
control law for the linearized system Eqs. (110), (111). The overall feedback structure
is depicted in Fig. 16a where the external set-point Ysp is included. The complete
closed-loop dynamics is then

Z(t + h) = A Z(t) + b Ysp (118)

y(t) = c*Z(t) (119)

with system matrices Ao and bc defined as follows

Ar H-gF1 -gf2J 9 ?
c 1 b~ = 1 (120)
126 S. Marsili-Libelli

Is_2p

rll~l
a,

Fig. 16a and b. Structure of the discrete-time linear-quadratic-integral control scheme: a State feed-
back control incorporating integral action through matrix F 1 and gain t"2,b equivalent output controller

It can be demonstrated 75) that the system Eqs. (118)-(120) can track a constant set-
point even in presence of deterministic unmeasurable disturbances such as the time-
varying organic load S r This controller, however, makes use of the entire augmented
state Z(t), which is not a realistic assumption, but can be substituted by an equivalent
output controller (Fig. 16b) using only dissolved oxygen measurements. The output
controller can be determined from the closed-loop matrices {Ac, b , e*} using the well-
known equivalence between system matrices and closed-loop transfer function ~2)

Go(z) Gr(z) Y
Go(z) = c * ( z I - - A ) -1 b = = -- (121)
1 + Gp(Z) Gr(Z) Ysp
where Gr(Z ) and Gp(z) are the regulator and process transfer functions respectively.
From Eq. (121) the controller Gr(z ) in terms of an input-output difference equation
can be obtained. For practical implementation it is more convenient to determine
the transfer function in the backward shift z -1 operator:

G c _ u(z) a,z 1 4_ ... --k anz -n

Gr(z 1) _ (1 4 ) Gp e(z) -- 1 + bl z-1 q- ... -[- bn z-n (122)

which yields the difference equation

u(t) = b l U ( t - - h) + ... + b u ( t - - nh) + a l e ( t - - h) + ...

+ ane(t - - nh) (123)
Modelling, Identificationand Control of the Activated Sludge Process 127

Equation (123) determines the controller output u(t) at time t based on past inputs
u ( t - h), u ( t - 2h) . . . . . u ( t - nh) and control errors e ( t - h), e ( t - 2h), ...,
e(t -- nh) up to n previous sampling intervals. This approach yields a satisfactory
performance in terms of control accuracy and has the advantage of including the
control effort in the cost functional Eq. (112). The resulting controller Eq. (123) is
very simple to implement and requires little additional hardware with respect to con-
ventional automation. It is also easy to adjust the coefficients in order to avoid nu-
merical problems. As an example, a successful implementation of the output controller
(123) on a very limited microprocessor using integer arithmetics is reported in 75)

4.3 Activated Sludge Control

Activated ~ludge wastewater treatment processes may vary greatly in design. The
relevant literature reflects this variety and a number of different control problems have
been considered, covering all aspects from dissolved oxygen and specific oxygen
utilization rate (SCOUR) control 18,19,39,77-v9, By) to more comprehensive strategies
involving sludge recycle ,8,80-86) and secondary settler management 88-90)

4.3.1 Dissolved Oxygen Control

For obvious reasons, DO control is by far the most studied aspect of the activated
sludge process automation. It has been shown that the pertinent dynamics is very
simple, at least so far as it is not required to structure the OUR term. Furthermore,
unless the DO level is abnormally low, this dynamics is independent of the basic
substrate-biomass interaction. A precise DO control has its own merits, assuring a
good sludge quality, making available the right amount of oxygen needed by a time-
varying biochemical demand, and avoiding undue energy expenditure. It is also a
necessary prerequisite for any long-term plant control policy. Of course, the control
objective depends on the kind of plant: in a complete-mix plant only the DO level
would be controlled, whereas in a longitudinal plug-flow reactor 91) the shape of the
DO profile ,8) along the tank should be controlled.
Biological oxidation of carbonaceous substrate occurs when the dissolved oxygen
concentration is above a threshold limit of about 1 mg 1-1. Since the oxygen utiliza-
tion depends on the incoming organic load, dissolved oxygen control is required to
maintain a constant DO value during these fluctuations. The relevant model is the
DO dynamics in the aerator already established by Eqs. (8), (9) in Sect. 1.

OUR = ~/sKbSX q- TeemX2/S ; (8)

dC/dt = KLa(C -- C ) -- OUR q(1 -I- r) C i (9)

The manipulated variable is the airflow U a and this quantity appears in Eq. (9) through
the oxygen transfer rate coefficient KLa. Practical experience has shown that KLa
varies with the process temperature

KLa(T) = KLa(To) 1.024 (x-xo) (124)

128 S. Marsili-Libelli

and a linear relationship exists between air flow and Kea 39, 87)

Kea = K ~ + K U (125)

For example, Holmberg found for the Suomenoja plant 39) Ko = - - 0 . 0 1 1 and
K a = 0.0018 i f U a is expressed as m 3 h -1 and Kea as h -1. It was also found that the
transfer efficiency is always less than that obtained with the same aeration equipment
in tap water. Substituting Eq. (125) in Eq. (9) and neglecting the natural re-aeration
K ~ yields

dC/dt = (Csa t - - C ) K a U a - - O U R - - q(1 + r) C i (126)

It should also be remembered that the dissolved oxygen saturation value Cs, t depends
on the temperature as follows

Cs, t = 14.161 - - 0.3943T +.0.007714T z ~ 0.0000646T 3 (127)

Equation (127) is valid for tap water. For domestic wastes the saturation value of the
sludge liquor should be multiplied by a factor of approximately 0.95. Moreover, if a
submerged aeration system is considered, the diffusers depth must be taken into ac-
courrt as this affects the oxygen solubility. Therefore the saturation value is determined
as follows

CsS2? = C s a t ( P b -[- H a - - v)/(760 - - v) (128)

where:
PD = Barometric pressure (mm Hg);
H a = Hydrostatic pressure at diffusers depth (mm Hg);
v = Liquor vapour pressure at temperature T (mm Hg)
Equations (8), (124)-(128) form the basic model for dissolved oxygen control in
a completely mixed activated sludge plant. A PID controller can be used to determine
the amount of air supplied to the system U s depending on the difference between the
required and actual DO levels. Hence, referring to Eqs. (108)-(110) the following
quantities are defined: e(t) = C p - - C and u(t) = U a. The main practical problem
encountered in dissolved oxygen control is that the great majority of operating plants
do not have any provision for varying the air flow rate continuously and the aeration
equipment is operated on an on-off basis. However, as the advantages of adjustable
air flow begin to emerge, more and more plants are being upgraded. Flexible aeration
systems comprise movable weirs to control the submergence of surface turbines or
variable DC motors driven blowers if bottom diffusers are used. However, the practical
air flow regulation conceals several engineering problems as any change in the air
flow affects the efficiency of the aeration system. In the bottom aeration case the
blowers/diffusers combination is designed for maximum efficiency at a gixen air
flow and would operate at a lower efficiency ifa different flow is required. Particularly,
at low flow the static pressure may stop the airflow completely even though the blowers
are running. On the other hand, the regulating range of the blowers may be too narrow
for the control requirements; therefore two or more partial control loops may be
Modelling, Identification and Control of the Activated Sludge Process 129

D.O.Meosurement
Settler (1-wlO
C~p
S Aerotor

rQ

Fig. 17. Aeration control scheme using a PID regulator to manipulate the air flow and a dissolved
oxygen probe in the aerator

12~

V"\L ,I
Kp =0.76
?~ Ti =0.07
Td =0.L5
0
o 3
Time(d)
Fig. 18. Performance of the previous PID control scheme. The integral action is very strong to cope
with the inaccessible OUR disturbance, whereas the derivative action is kept low to prevent high
frequency oscillations in the controller

needed. Aarinen et al. 77) used a reciprocating compressor with continuously control-
lable D C motor drive. Personal experience has shown that since bottom diffuser
require a substantial pressure to produce sufficiently fine bubbles, it is preferable to
use a constant high pressure supply and to use a control valve to obtain the required
flow. In the case of surface aeration, a limited degree of adjustment can be achieved
by varying the submersion of the turbines and/or the speed of the drives, although
usually the efficiency of the system depends critically on both factors, making the
surface aeration systems less amenable to variable operation. The following examples
assume that the plant is equipped with continuous air flow regulation.
Figure 17 shows a possible dissolved oxygen P I D control scheme to maintain a
set-point of 2 mg 1-1. Integral and derivative actions must be introduced to cope at
least partially with the inaccessible disturbance represented by the variable incoming
load. The simulation of Fig. 18 was obtained with the model of Sect. 2 and the incoming
BOD varied from 90 to over 500 mg 1- I during the 24-h period. The P I D coefficients
were tuned by trial-and-error. It can be seen that the controller delivers a massive
integral action (T i = 0.07). This is required by the presence of the inaccessible dis-
130 S. Marsili-LibeUi

Table 4. Operating conditions for optimal aeration control 79,8o)

Variable Symbol Operating value Units

BOD S 21 mg 1-1
Input load S~ 350 mg 1-1
MLSS X 2800 mg 1-1
Recycle MLSS X 9600 mg 1-~
DO C 2 mg 1-1
Input D.O. C~ 1 mg 1-1
Airflow U 12 Nm 3 h-1
Dilution rate q 0.14 h-I
Recycle rate r 0.50 --

Parameter Value Units

Kb 7.42 x 10-4 1mg -1 h -1

K 1.47x 10 -3 l m g - l h -1
Km 5.23 x 10 -4 h -1
K 0.2286 Nm -3

turbance represented by the oxygen uptake rate which varies the oxygen d e m a n d o f
the system. By contrast, the derivative action is kept rather limited (T d = 0.45) to
avoid excessive high frequency oscillation in the control signal that would be detri-
mental to the aeration equipment. It can also be seen that the air flow rate has fairly
large swings and that the regulation is imperfect at the beginning of the high load
hours (from 10:00 a.m. to 1:00 p.m.) when the process conditions undergo a drastic
change, which the P I D can a c c o m m o d a t e only in part.
By contrast, if the controller parameters were to be selected to satisfy some optimal
criterion, the a p p r o a c h o f Sect. 4.2.2 and the objective function Eq. (112) can be adopt-
ed. This a p p r o a c h has been applied to the pilot p l a n t whose d a t a were previously used
t o calibrate the S M L model (see Table 2). The selected operating p o i n t is reported in
Table 4 and the resulting continuous-time system matrices are

-2.3347 -0.0155 0.0 0.0

A*= 13.9830 -0.3212 0.0 b*= 0.0 (129)
-0.4184 - 7 . 8 6 x 10 .3 -2.954 1.52 x 10 -3

with eigenvalues )~1 -- --0.43556; )~z = --2.22043; )~3 = --2.954, which guarantee
a stable linear approximation. Assuming a sampling interval h = 15 min the following
discrete-time output regulator is determined

U ( t ) = 3 . 1 U a ( t - - h) + 3 . 5 3 U ( t - - 2h) + ! . 7 2 U a ( t - - 3h)
- -0.26Ua(t - - 4h) + 0.044Ua(t - - 5h) + 0.01U (t - - 6h)
+ 749.8e(t - - h) - - 1871e(t - - 2h) + 1749.1e(t - - 3h)
- - 761.4e(t - - 4h) + 155.5e(t - - 5h) - - 12.1e(t - - 6h) (130)

The application o f the approximate optimal regulator Eq. (130) to maintain a dissolved
oxygen concentration of 2 mg 1-1 is shown in Fig. 19 where it is compared with a
Modelling, Identification and Control of the Activated Sludge Process 131

----- Set-point
..... Constant a e r a t i o n
- - Controlled aeration
01 3- // /"~ \
E._E-. iI \,
C I
. l ~ / \

2. . . . . t --4--

I~ " I I l "

"\\ l I

0
m "I - ~ It i /

"l~ t l t t
t I i I
I i
- - ~ , i , , 1 t i ~ i

0 /., B 12 16 20 2/, 28
a Time (hl
20000J

I
i
- - Controlled aeration
I ..... Constant aeration
Ol . . . . . - , ,
0 Z, 8 12 16 20 2A 28
b Time (h)

Fig. 19 a and b. Optimal dissolved oxygen control using Eq. (130) (solid line) versus constant aeration
(dashed line) applied to the pilot plant model for which the kinetic parameters were previously estimat-
ed (see Table 2) so~

constant aeration system. It can be seen that the latter is not capable of supplying an
adequate amount of oxygen during the peak loading hours. Apart from the avoidance
of anoxic conditions, the energy saving obtained with the optimal controller are in
the order of 30 ~o. Similar energy savings have been reported in the literature for
automatic control applications.

4.3.2 Sludge Recycle Control

Sludge recycle is the second most obvious control action and can be used to obtain
certain biological conditions in the aerator 76, 8u or for sludge conditioning in the
secondary settler 88-90). The control objective may have been determined through
previous steady-state optimization, such as described by Lauria et al. 32), Keinath
et al. 927, and Craig et al. 93). Hence the control target may be a given food-to-mass
ratio F or sludge age 0o as defined by Eqs. (102), (103). To obtain this, Flanagan s2)
proposed a steady-state recycle/waste control based on a M o n a d kinetics process
model
132 S. Marsili-kibelli

1) Aeration tank."

g
substrate dS/dt = q(S i - - S) - - T( X ; (131)

biomass dX/dt = qrX r - - q(1 + r) X + (g - - Ks) X (132)

2) Secondary settler."

biomass balance Q(1 + r) X = Q(1 - w) X e + Q(r + w) X r (133)

where Xo is the biomass concentration in the effluent. At steady state dS/dt = dX/dt
= 0 and a mass balance at the secondary settler yields

(1 + r ) X = ( 1 - - w ) X o + ( r + w ) X r (134)

Combining Eq. (134) with the steady-state solution of Eq. (132) yields the waste flow
ratio to maintain a prescribed solids retention time 0 c

XVO] 1 _ X
w - (135)
X r -- X e

F r o m the same Eqs. (131)-(134) the required recycle fraction is determined

1 -- q(g -- Kd)
r = X (136)
X r -- X

Equations (135) and (136) define a steady-state sludge management policy. Since the
plant usually operates in a dynamic environment, it is necessary to complement these
rules with automatic controllers if the time-varying nature of the operation is to be
accounted for. The P I D control scheme of Fig. 20 can be used, assuming that sludge
density can be measured directly. Fig. 21 shows the performance of the P I D obtained
with trial-and-error controller tuning using the same daily-periodic influent load as
in the simulation of Fig. 18 and assuming an abrupt set-point change from Xsp = 2 g 1- t

MLSS Measurement

Settler

Aerator

Fig. 20. Biomass control scheme using a PID regulator and assuming that biomass information is
directly available
Modelling, Identificationand Control of the ActivatedSludge Process 133

iiiit
1.0

0.8
r

3000J 0.6
E 1,.,1
>,,,
U
2000 o.~. g:
Kp = 1.12
lOOO- Ti = 12.50 0.2
Td = 0.54

O"
o 3~
Time (d)

Fig. 21. Performanceof the previous biomass control scheme. Contrary to the previous DO control,
the integral action is small given the inherent "inertia" of the sludge dynamics. For the same reason
the derivativeaction, acting as a forecast, has been increased

t o X s p = 3 g 1-1. Contrary to the PID control of dissolved oxygen (Fig. 18) the reset
time is very long (T i = 12.5 h) indicating that the integral action is kept very low. In
fact the sludge dynamics has such a large inherent "inertia" that a strong integral
control would amplify its already sluggish response. Conversely, the short lead time
(T a = 0.54 h) indicates that derivative control is quite strong, as it must provide the
predictive action required by the slow dynamics. In fact, it can be seen that the recycle
diurnal pattern anticipates the daily fluctuation of the sludge concentration about the
required set-point value.
As with dissolved oxygen control, the sludge control law can be complemented with
an optimality index and its parameters determined accordingly. Several optimal control
results are available in the literature: Hamalainen et al. 76) proposed a state feedback
controller with a structure similar to Eq. (115) and a quadratic performance index as
Eq. (112). In addition, the feedforward action was used to forecast the diurnal influent
flow rate. Since the optimal state feedback law implies the availability of the entire
state vector, as shown by Fig. 16a, a linear observer was used to obtain an estimate
of BOD and biomass. The operational improvement of this solution was fully demon-
strated.
A different approach was pursued by Sincic and Bailey 83), Yeung et al. 84), and
Stehfest 8s~. Their common assumption is that the plant input has a daily periodicity
and the controller parameters are optimized to yield the best performance for that
particular periodic input. This approach, which makes heavy use of simulation and
numerical techniques, is well suited for plants dealing with domestic wastes, whose
inputs are known to be consistently periodic. The resulting control law is strictly open-
loop as it relies completely on the periodic assumption and cannot correct any depar-
ture from the assumed nominal pattern. Sincic and Bailey 83) consider the following
pertbrmance index:
T T
F DT -1 j ' Q ( t ) [ l +fit)] S(t) dt + ( 1 - D ) T -~f[S(t)-S~,,] 2dt (137)
0 0
134 s. Marsili-Libelh

where T is the optimization period usually assumed to be 24 h, and Say is the average
output BOD defined as

T
f Q(t) [1 + r(t)] S(t) dt
S.v = o (t38)
T
Q(t) [1 + r(t)] dt
0

The parameter D can be used to shift the emphasis of the optimization on either term
of Eq. (137). Setting D = 1 corresponds to minimizing the daily amount of substrate
in the effluent whereas D = 0 corresponds to minimizing the fluctuations around
S v, Equation (138) was later used by Yeung et al. 84) to define another performance
index

T
r = T - 1 j" [ S ( t ) - Sav] 2 d t (139)
0

The results so obtained indicate that this second approach yields better BOD removal
that a feedback controller. The third periodic control exercise was recently proposed
by Stehfest 85) assuming the daily average pollutant concentration as the performance
index

T
lr T -~ f (S + ~X) dt (140)
0

where the parameter ~ xs a weighting coefficient and the biomass X is included as an

additional source of organic substrate. In ss) a comparison is made with a constant F m
recycle control to which the optimal periodic solution is clearly superior. As a conclud-
ing remark on periodic control, it must be emphasized that these results are not directly
comparable because they were obtained with differing models, performance indexes,

Si BOD/Biomass
kinetics

DO kinetics
I r
OUR
Observer n:l
estimator
+OOR
S,X Recycle
Observer n, 2 store estimator contro[

Fig. 22. Combined estimation and control scheme Eqs. (141) (145) using dissolved oxygen measure-
ments to implement set-point sludge control
Modelling, Identification and Control of the Activated Sludge Process 135

and periodic input functions. The unavoidable weakness of the periodic approach is
its feedforward philosophy: if for any reason the actual plant operation were to depart
from the nominal pattern, the periodic controller would be incapable of any corrective
action.
So far the problem of measurements availability has not been considered. In the
case of periodic optimization this on-line information is immaterial, because the con-
trol strategy is determined off-line through iterative simulation and then implemented
on the actual plant. Conversely, in feedback schemes it was so far assumed that bio-
mass density could be directly measured by means of turbidimeters or ultrasonic
devices. However, given the cost, complexity, and limited reliability of these on-line
devices, it would be highly desirable to implement a sludge controller which makes use
of inexpensive and reliable process measurements. For this, the observer theory
developed in Sect. 3.6 can be used to obtain a controller based on dissolved oxygen
measurements alone. This is made possible by the state observer (102)-(103) which
relates the estimation of BOD and biomass to the oxygen uptake rate, which in turn
can be reconstructed from dissolved oxygen measurements according to Eqs. (85)
and (98). Mechanizing the double observer/controller scheme yields the aggregated
structure shown in Fig. 22 where the first observer reconstructs OUR from DO
measurements and supplies this information to the second observer which in turn
estimates the un@servable state variables S and X, providing the regulator with their
estimates S and X. The complete double observer/controller scheme is then:
Observer n. 1." OUR estimation

OUR(t) = bo(t--b) + t ~ f ~ t ( t - h ) with 0 < t ~ < h' (141)

observer n. 2." substrate/biomass estimation

dS/dt = --KbS?( -- q(1 + r) S + qSi, (142)

d)(/dt = -Kin(1 + 7ffTe) X2/~ ~_ OUR(t)/% - q(1 + r) X + qrXr ; (143)

P I D biomass controller."

h ( h
ar(t)= Kp T~i X~p(t)-Kp 1+~+ X(t)

+Kp 1 +2 X(t-h)-Kp h-X(t-2h)' (144)

r(t) = r(t -- h) + 6r(t) (145)

Observer n. 2 is implemented in continuous-time for accuracy reasons, whereas the

Observer n. 1 and the PID controller are discrete-time with sampling period h. The
structure and performance of the combined scheme is analyzed in 86) and Fig. 23
136 S. Marsili-Libelh

5.

3
x_ s p . . . . . . : ~ . . . . . . . . . . . . . . .

U3
m 2
.J

0 J I I 116 ; a
0 4 8 12 2 24
a Time (h)
1.0

0.6

0.2

-0.2 4 8 12 16 20 24
2 Time (h]
E
-o.6

b -1.0

1.00

0.75

(.3
>" 0.50
~J
Fig. 23 a--c. Performance of the com-
n~
bined estimation/control scheme of
0.25 Fig. 22 in maintaining a required
concentration X sp : a Sludge concen-
tration, b control error, e recycle
fraction
0 t i 112 ll6 I I
0 4 8 20 24
e Time (hi

shows its performance in the same conditions as previous schemes o f Figs. 15, 17, and
20 with a biomass set-point Xsp = 2.5 g 1-1. The P I D parameters are such that the
desired set-point is reached after about 6 h and kept almost constant thereafter in
spite of a variable influent organic load S i. This satisfactory control action is achieved
with an admissible control effort. In fact the recycle fraction r remains within the
allowable limits (0, 1) (Fig. 23c). In practice a recycle greater than 1, though not
representing a physical limit, should be avoided as it would imply a massive solids
removal from the thickener underflow and a consequent sludge depletion. A sensitivity
Modelling, Identificationand Control of the ActivatedSludge Process 137

analysis of the combined scheme of Fig. 22 carried out in 86)revealed that the estimator/'
controller arrangement is only moderately sensitive to modelling errors, with the most
sensitive parameter being Kca. Its sensitivity, however, is not so high to impair the
overall efficiency.

4.3.3 SCOUR Control

The relationship between microbial activity and oxygen uptake rate has already been
established through the specific oxygen uptake rate (SCOUR) defined as

OUR (1 -- Y)
SCOUR - X - ~ g(S) + 7eKa (107)

Since OUR is either directly measurable or can be calculated with the algorithm of
Eq. (98) in Sect. 3.5, SCOUR is an on-line control variable whose validity has been
demonstrated through computer simulation 19~ and practical experience sv). In fact,
SCOUR is a meaningful growth indicator in a dynamically changing environment,
whereas food-to-mass ratio Fro-or mean cell retention time 0o are not. Keeping SCOUR
close to a given value by manipulating recycle flow and/or contacting pattern provides
positive real-time control in presence of inflow disturbances. Each plant has an "opti-
mal" SCOUR value depending on the kind of incoming sewage and microbial in-
ventory. Once this value has been experimentally determined, a recycle control can
be set up to obtain a sludge concentration Xsp compatible with the selected set-point
SCOURsp

OUR(t) (146)
Xsp(t) -- SCOUR
sp

This value can be used as a set-point for a biomass PID control loop acting on the
recycle flow rate as already considered in Fig. 20. The dynamic requirements of this
loop are now more stringent since the PID is meant to track a variable Xsp(t) given by
Eq. (146), therefore the previous tuning cannot be carried over and new control
parameters have to be selected.
Yust et al. sT) use SCOUR to compare recycle versus contacting pattern control
policies. A step-feed and a completely mixed plants were run in parallel, a SCOUR
value of 15 mg g-1 h-1 (0 2 per MLSS per time) was selected for both and the related
sludge set-point Xsp(t) was determined via Eq. (146). Then the control parameters were
determined minimizing the following performance index

F = [Xsp(t) -- X(t)] z (147)

at each sampling interval, where X(t) is the sludge concentration in the most relevant
process section. On-line measurements for sludge density in each compartment and
oxygen uptake rate were provided. Figure 24 compares the real-time experimental
results obtained manipulating the recycle flow in the completely mixed plant and
redistributing the influent load in the step-feed plant. It can be seen that the latter
exhibits a greater flexibility and follows the selected SCOURsp set-point more closely.
 138 S. Marsili-LibelIi

40

• 35
"7

o~ 30
E CSTR system (B)
- 9. . . . . ./ . .
2s - .

.'" "'..
13- 2O
t,o 9 " Setpoint, ""..
cO
-- 15

o_ 10 Step feed system (A)

c7
~" 5
O
(.3
cO 0 I l I L | I |

17 19 21 23 1 3 5 7 9 11 13 1 17
Time (hi

Fig. 24. SCOUR control of a step-feed pilot plant obtained by variable redistribution of the influent
between two cascaded compartments. Reproduced by permission of The Institute of Measurement
and Control on behalf of Yust et aI. 87>

As a concluding remark, S C O U R control appears to be the most promising real-time

control strategy as it provides the sludge controller with a target concentration related
to the bioactivities in the process.

4.4 Adaptive Control

As already stated, the main difficulty in controlling biotechnological processes arises

from the variability o f kinetic parameters and the limited amount of on-line informa-
tion. Though many significant applications of adaptive controllers are reported in
the literature for specific biotechnological processes, very few results are as yet
available for the activated sludge wastewater treatment process. The design and
implementation of dissolved oxygen self-tuning controllers, the only available to
date, will now be considered.

4.4.1 Structure of the Dissolved Oxygen Adaptive Controllers

Adaptive control is now a well-established branch o f control theory (a thorough
survey on adaptive controllers is presented in 947). In this paper, the choice is restricted
to a self-tuning regulator 62, 9sl. Self-tuners can be divided in two broad categories:
explicit and implicit, depending on whether the parameters of the process or of the
controller are being estimated. For this application an implicit controller is considered,
given the presence of the oxygen uptake rate, which is now regarded as an inaccessible
deterministic disturbance. The literature on the applications of self-tuners to the
activated sludge process is still very scarce. Apart from purely academic exercises,
\I odelling, Identificationand Control of the ActivatedSludge Process 139

such as 96) in which OUR is unreasonably supposed constant, few other papers deal
realistically with the practical implementation of dissolved oxygen control 97.98~.
The self-tuning algorithm which is now introduced uses as little information as
possible regarding process dynamics. The only practical assumption is that the aera-
tion system is continuously adjustable. In other words, the plant is supposed to be
equipped with some actuating mechanism to change the air flow rate between the
allowable limits, such as movable weirs, variable submergence surface aerators or
variable speed submerged blowers. The controller described in this section is a self-
tuning version of the popular PID regulator, based on a modified algorithm proposed
by Cameron and Seborg 99) and adapted to this context with nonlinear process dy-
namics and an inaccessible variable (OUR). The specific problem addressed in this
section is set-point dissolved oxygen control, which was previously studied using
conventional techniques. The limits of the conventional approach are clearly related
to the inability of a predetermined controller to adjust to the changing operational
environment. As in Sect. 4.2.1, the aim of the controller is to keep the DO level at a
constant value Csp in spite of input organic load fluctuations and/or parametric
variations. The starting point for the controller design is again the sampled dissolved
oxygen deficit (DOD) balance

D(t + h) = aD(t) + bOUR(t) + bq(1 + r)C(t) (82)

where

1
a = exp (--hKLa) -~ 1 -- hKaU a and b = ~ (1 -- a) ~ h (148)

Decomposing Eq. (82) in the average and time-varying parts

Ua(t ) = U o + fi(t) (149)

D(t) = D O + a(t) (150)

C(t) = Co + ~(t) (151)

OUR(t) = OURo + 6~(t) (152)

yields the following approximate linear static and dynamic balances

static bahmce."

KaUoD o = OUR o + q(1 + r)C O 9 (153)

dynamic balance."
d(t § h) (1 + a ) d ( t ) - hKaD0fi(t) + h oer(t) + hq(1 + r) c(t) (154)

The linearized dynamic balance (154) is the starting point for the DO self-tuning con-
troller design. It should be underlined that the 6"~(t) term is supposed to be unknown,
I40 S. Marsili-Libelli

OUR

C +[~_Csat
DO kinetics
t g+ -Csp Dsp
-qh (I +r) Fi[ter

Self- tuner
yf
L.S. Estimotor

Uo D(-N PID Controller

Csp

Fig. 25. Structure of the self-tuning PID regulator implementing dissolved oxygen control

therefore the controller should incorporate integral action to offset this inaccessible
deterministic disturbance. Furthermore, the K a parameter is generally unknown or
time-varying, so there is a clear need for a self-tuning controller. The structure of the
adaptive PID controller is depicted in Fig. 25, where its two main constituents (least
squares estimator and PID regulator) are shown together with the inputs consisting
of the filtered dissolved oxygen deficit (yf) and the deterministic part qh(1 + r) ~(t).

4.4.2 Performance of the Self-tuning Controllers

The controller structure which produces the incremental air flow signal fi(t) has the
following form

fi(t)=t](t-h)+vIH~ fiyf(t-ih))i=l (155)

where:
D p is the dissolved oxygen deficit set-point: Dsp = Csa t - - Csp,
is the incremental manipulated variable (air flow rate),
yf is the filtered DOD signal according to the recursive filtering equation,
yf(t) = yf(t -- h) Pa + d(t) -- Dsp, (156)

fl are the PID controller coefficients,

v is a weighting coefficient,
H ~ is a set-point weight defined so as to assure steady-state zero tracking error, i.e.

1 N
H~ ~ fl (157)
1 + pdi=l

where Pa e (--1, 1) is the filter pole of Eq. (156),

Modelling, Identification and Control of the Activated Sludge Process 141

N is the sum of the orders of the process model and of the filter Eq. (156).
The coefficients fi are estimated on-line from past samples of the filtered output yf
using the following error equation

r = a(t) - - Qv(t) l~(t) - - qh(1 + r) e(t) (158)

where ~(t) is the estimation error and the vectors of the representers Q(t) and of the
estimated parameters l~(t) are defined as follows"

s = [yf(t) yf(t - - h ) ... yf(t N h ) ] a" (159)

0(t) ~-- [f0' q . . . . . fNIT (160)

The estimation procedure is a recursive least-squares algorithm similar to Eqs. (90),

(91) where the updating of the covariance matrix G is mechanized with the square-
root 64, 65) robust factorization method to avoid numerical degeneracy in the long run.
More details on the derivation of this specific algorithm, including numerical tests
and comparison with other self-tuning control laws can be found in loo). These tests
have shown that this regulator is fairly insensitive to the choice of the order N. Select-
ing N = 2, the result of Fig. 26 is obtained, where a variable and unknown K a has
been introduced. It can be seen that contrary to the deterministic algorithm of the
previous Sect. 4.2, the self-tuner can adjust the air flow in order to keep the DO level
as close as possible to the prescribed value DOsp in spite of the K a variations. It was
also observed that increasing N beyond 2 did not produce any appreciable improve-
ment. Conversely, the filter pole Pd and integral weight v had a great influence on the
performance. For the 48-h simulation of Fig. 26 the following values were selected:

Ua
C Clt)

Csp
0
E3

01
0 6 12
~ 18
~ 24
~ 3~0 3~6 42
r 48
,
Time (h)
Fig. 26. Performance of the dissolved oxygen set-point self-tuning scheme of Fig. 25 with a variable
and unknown KLa. The airflow adjusts to the changes of the transfer coefficient K a
142 S. Marsiti-LibcH~

Pd = 0.4 and v = 1.0. The effect of a time-varying K a on the controlled airflow U

is quite evident: during the second 24-h period the controller decreases the aeration
rate to adjust to an increased transfer efficiency K in order to maintain a given DO
level C sp .
As a concluding remark, the merits of this self-tuning PID scheme can be sum-
marized as follows:
1) The controller is independent of any steady-state quantity such as the average air
flowrate U o. Therefore, no error in computing this preliminary parameter does
impair the on-line operation.
2) The performance of the controller with a time-varying K a is quite acceptable, at
least with the rate of variation that is observed in practice.
3) The controller output is sufficiently smooth to avoid undue wear and tear of the
actuating mechanism such as electrical drives and blowers.
4) The parameter estimates are very stable and do not degrade as time progresses,
thus allowing unattended long-term operation.
It can be concluded that this PID self-tuning algorithm is suitable for the application
to activated sludge plants equipped with adjustable aeration systems.
As anticipated, a similar approach was pursued by Olsson et al. 98) in producing
the first engineering application of a dissolved oxygen self-tuning controller. The
control scheme, implemented in a completely mixed activated sludge plant at Kfippala
(Sweden), relies on the usual dissolved oxygen model from which the linear sampled
model is derived

C(t) = aC(t - - h) + bUa(t - - h) (161)

where the two unknown parameters {a, b} are estimated on-line with a least squares
algorithm. The inaccessible deterministic disturbance O U R is not included in the

Do co[3c.
Guide
l
vanes ManifoId .... Aerator r------q
/~lr flOW 1

~ ISU_re I ~-------

VGne
position ~ lair flow Self-tuner

F
o P

Valve/" h =15 min

position REF

Fig. 27. Experimental self-tuning control scheme at K~ippala plant showing the double control loop
(guide vane and air flow valve) to adjust the airflow to the required value 98L Published by the Inter-
national Association on Water pollution Control (IAWPRC) in conjunction with Pergamon Pres~
Modclhng, Identification and Control of the Activated Sludge Process 143

>, 0 2 /-., 6 8 10
x

~
m
?5 1
b
o
i

4
Time (d)
; 8 lo

Fig. 28 a and b. Experimental self-tuning control performance at K/ippala plant 98). Published by the
International Association on Water pollution Control (IAWPRC) in conjunction with Pergamon Press.
Dissolved oxygen concentrations with a manual adjustments, b self-tuning control

model and to compensate for this modelling error the controller has integral action.
In fact the following sampled controller is selected

Ua(t ) = Ua(t - - h) - - ka[C(t ) - - C(t h)] + k2h[Csp - - C(t)] (162)

corresponding to a continuous-time proportional-integral (PI) action. Figure 27

depicts the K/ippala controller implementation with three cascaded PI control loops
to regulate the air flow~ The self-tuner produces a reference air flow value Qref which
is compared with the actual flow Q. The error is used to position the valve, which
also depends on the pressure in the manifold and this second error is fed into the guide
vanes controller. Using three cascaded PI controller enhances both flexibility and
energy conservation, since the overall control action is aimed at minimizing the energy
losses as the flow is varied. Figure 28 from 9s) shows the performance of this arrange-
ment in keeping the dissolved oxygen constant over a 7-day period. The authors also
demonstrate that a constant DO level has improved the settleability properties of the
sludge in the final clarifier.

5 Conclusions
This paper has surveyed the research aovancements in mathematical modelling,
estimation and control of the activated sludge wastewater treatment process. The
presentation has focused on the structure and use of the SML reduced-order model
to describe the interaction between carbonaceous BOD and biomass, with several
other concurrent aspects of the process being included.
In Sect. 2, the simplified SML model has been introduced and its main structural
features discussed. It was shown that it compares well with the widely used Monod
kinetics, but has the advantage of being structurally simpler and numerically reliable.
In addition to the main substrate-biomass interaction, other process streams were
modelled, such as the oxidation of ammonium-nitrogen and the dynamics of the
secondary clarifier.
144 S. Marsill-Labclh

Section 3 of the paper was devoted to the numerical calibration of the model, first
dealing with its structural identifiability and then with the more practical aspect of
making the best use of the available experimental information. It was concluded that
the SML model is identifiable and can reproduce some well known experimental
results with greater accuracy than the Monod model. On-line estimation of bioactivi-
ties was considered next, proposing an algorithm to estimate the oxygen uptake rate,
a key process quantity.
The implementation of reliable control laws was considered in Sect. 4 and a review
of several practical controllers was presented, including the popular PID regulator
and a more advanced linear-quadratic-integral control. Some examples of practical
controller design in connection with dissolved oxygefi and sludge regulation were
also presented. Lastly, self-tuning control has been considered as this approach is
surely to draw more research work in the near future.
The aim of this paper was to demonstrate that modelling, estimation and control
techniques can indeed improve the reliability and efficiency of the activated sludge
process, whose capabilities have yet to be fully appreciated and exploited. The applica-
tion of these control engineering tools can make up for the lack of sophisticated in-
strumentation and demonstrates that plant operation can be significantly improved
by computer control with only a marginal increase over the total investment cost.
The future trend is undoubtedly in this direction. A wider use of advanced control
application is envisioned to obtain real-time process information, especially regarding
biological indicators such as oxygen uptake rate and to implement reliable control'
strategies.

6 List of Symbols
A secondary settler cross section [m2]
C dissolved oxygen concentration [mg 1-a]
Ci input dissolved oxygen concentration [mg 1-']
C sp dissolved oxygen set-point [mg 1-a]
Csat dissolved oxygen saturation value [mg 1-1]
D dissolved oxygen deficit [mg 1-1]
F total setting flux [kg m -2 h-1]
Fm food-to-mass ratio [kg g-1 d-l] (BOD per MLSS per time)
h sampling interval [h]
hb sludge blanket height [m]
K a air transfer coefficient [Nm -3]
Kb substrate decay rate (SML) [rag -11 h -1]
Kc biomass specific growth rate (SML) [rag-11 h-1]
Kd microbial decay rate (Monod) [h-1]
Km biomass endogeneous metabolism rate (SML) [h 1]
Kn nitrifiers decay rate [h -1]
Ko dissolved oxygen half-velocity constant [rag 1-1]
K half-velocity constant (Monod) [nag 1-1]
Kam ammonium-nitrogen half-velocity constant [rag 1-1]
KLa oxygen diffusion rate coefficient [h-1]
Modelling, Identification and Control of the Activated Sludge Process 145

M accumulated biomass in the settler [kg]

Na nitrate-nitrogen concentration [mg 1-1]
n, a batch flux settling curve parameters [--]
Q process hydraulic flow rate [m3 h-1]
q dilution rate [h -1]
r~ w recycle and waste flow fractions [--]
S substrate (BOD) concentration [mg L-1]
Si input substrate (BOD) concentration [mg 1-1]
Sam ammonium-nitrogen concentration [mg 1-1]
Sexp, Xexp, Cexp substrate, biomass, and DO measurements [rag 1-1]
t, t 1 time [tl]
Ua air flow rate [Nm3 h -1]
u settling sludge bulk settling velocity [m h-1]
X biomass concentration [mg 1-1]
X settler sludge density above sludge blanket [rag 1-1]
Xb settler sludge density below sludge blanket [rag 1-1]
Xe sludge concentration in the effluent [mg 1-1]
Xn nitrifying bacteria concentration [mg 1-1]
Xr recycle biomass concentration [mg 1-1]
X sp sludge set-point [g 1-1]
Y yield factor (Monod) [--]
Y. nitrifiers yield factor [--]
0 hydraulic retention time [h]
0c solids retention time (sludge age) [d]
it maximum specific growth rate (Monod) [h-1]
itn nitrifiers maximum specific growth rate [h-1]
7e oxygen utilization coefficient for endogenous metabolism [--]
7. oxygen utilization coefficient for nitrification [--]
7s oxygen utilization coefficient for synthetic activities [--]
feed height above settle bottom [m]

A cronyms:
BOD biochemical oxygen demand [mg 1-1]
DO dissolved oxygen [mg 1-1]
LS least squares estimation algorithm
MLSS mixed liquor suspended solids [mg 1-1], as a global measure of bio-
mass in the activated sludge plant
OUR oxygen uptake rate [mg 1-1 h -1]
SCOUR specific oxygen uptake rate [mg g-1 tl t]
SML substrate/biomass reduced-order model, after the author's initials

Superscripts:
estimated variables
^

incremental variable around a steady-state value

()~ transpose of a matrix 0
146 S. Marsili-Libelli

7 References
1. Andrews, J. F. : Water Research 6, 319 (1972)
2. Andrews, J. F. : ibid. 6, 575 (1972)
3. Andrews, J. F. : ibid. 8, 261 (1974)
4. Busby, J. B. and Andrews, J. F. : J. WPCF 47, 1055 (1975)
5. Andrews, J. F. : Development of control strategies for waste-water treatment plants, in: Progress
in Water Technology (eds. Andrews, J. F., Briggs, R. and Jenkins, S. H.). p. 233, Pergamon Press,
Oxford 1974
6. Andrews, J. F., Stenstrom, M. K. and Burr, H. O. : Prog. Wat. Tech. 8, 41 (1976)
7. Olsson, G. : Activated sludge dynamics I: Biological Models, Report 7511 (C), Lund Institute of
Technology, p. 88 (1975)
8. Olsson, G. : AIChE Symposium Series, 72, No. 159, 52 (1976)
9. Beck, M. B. (ed.) : Operational water quality management: beyond planning and design, Execu-
tive Report 7, International Institute of Applied System Analysis (IIASA), Laxemburg (Austria),
p. 74 (1981)
10. Marsili-Libelli, S. : Modelling and control of biological wastewater treatment, Trans. IMC 6,
No. 3, 115 (1984)
1l. Marsili-Libelli, S. : Second Florence Workshop on modelling and control of biological waste-
water treatment, Env. Tech. Letters, 6, 515 (1985)
12. Roels, A. J. : Energetics and kinetics in biotechnology, p. 330, Amsterdz~m. Elsevier Biomedical
Press 1983
13. Bazin, M. (ed.): Mathematics in microbiology, p. 307, London, Academic Press 1983
14. Marsili-Libelli, S. : Ecol. Mod. 9, 15 (1980)
15. Marsili-Libelli, S. : ibid. 24, 171 (1984)
16. Monod, J. : Ann. Rev. Microb. 3, 371 (1942)
17. Poduska, R. A. and Andrews, J. F. : J. Water Pollut. Control Fed. 47, 2599 (1975)
18. Olsson, G. and Andrews, J. F. : Water Research 12, 985 (1978)
19. Stenstrom, M. K. and Andrews, J. F.: J. Env. Eng. Div. ASCE 105, No. EE2,245 (1979)
20. Vavilin, V. A. : Biotech. Bioeng. 24, 8 (1982)
21. Vavilin, V. A. and Vasiliev, V. B. : ibid. 25, 1521 (1983)
22. Vavilin, V. A. and Vasiliev, V. B. : ibid. 26, 1042 (1984)
23. Vandevenne, L. and Eckenfelder, W. W. : Water Res. 14, 561 (1980)
24. Esener, A. A., Veerman, T., Roels, J. A. and Kossen, N. W. F. : Biotech. Bioeng. 24, 1749 (1982)
25. Henze, M. (ed) : Modelling of Biological Wastewater Treatment, p. 204, Water Sci. Tech. 18
(1986)
26. Maynard-Smith, J. : Models in ecology, p. 146, Cambridge, Cambridge University Press 1974
27. Stenstrom, M. K. and Poduska, R. A.: Water Res. 14, 643 (1980)
28. Gujer, W. and Erni, P. : Progr. Water Tech. 10, 391 (1978)
29. Dick, R. I.: J. San. Eng. Div. ASCE 96, No. SA2, 423 (1970)
30. Tracy, K. D. and Keinath, T. M.: AIChE Symposium Series (Water) 70, No. 136, 291 (1973)
31. Shin, B. S. and Dick, R. I. : J. Env. Eng. Div. ASCE 106, No. EE3, 505 (1980)
32. Lauria, D. T., Uunk, J. B. and Schaefer, J. K. : J. Env. Eng. Div. ASCE 103, No. EE4, 625
(1977)
33. Stehfest, H. : Trans. IMC 6, 160 (1984)
34. Olsson, G. and Chapman, D. : Modelling the Dynamics of Clarifier Behaviour in Activated
Sludge Systems, in: Instrumentation and Control of Water and Wastewater Treatment and
Transport Systems (ed. Drake, R. A. R.) p. 405, Oxford, Pergamon Press 1985
35. Heineken, F. G., Tsuchiya, H. M., and Aris, M.: Math. Biosci. 1, 115 (1967)
36. Naito, M., Takamatsu, T., Fan, L. T., and Le, E. S. : Biotech. Bioeng. 11,731 (1969)
37. Nihtila, M. and Virkkunen, J. : ibid. 19, 1831 (1977)
3g, Aborhey, S. and Williamson, D. : Automatica 14, 493 (1978)
39. Holmberg, A. : Int. J. System Sci. 12, 703 (198!)
40. Homberg, A. : Math. Biosci. 62, 23 (1982)
41. Solomon, B. O., Erickson, L. E., Hess, J. E., and Yang, S. S. : Biotech. Bioeng. 24, 633 (1982)
42. Holmberg, A. and Ranta, J.: Automatica 18, 181 (1982)
43. Di Stefano III, J. J. and Cobelli, C. : IEEE Trans. AC 25, 830 (1980)
Modelling, Identification and Control of the Activated Sludge Process 147

44. Pohjampalo, H. : Math. Biosci. 41, 21 (1978)

45. Perkins, W. R. : Sensitivity Analysis, in : Feedback Systems (ed.) Cruz, J. B., p. 324, New York,
McGraw-Hill (1972)
46. Vialas, C., Cheruy, A. and Gentil, S. : An Experimental Approach to Improve the Monod Model
Identification, in: Modelling and Control of Biotechnological Processes (ed. Johnson, A.),
p. 262, Oxford, Pergamon Press 1985
47. De Boor C. SIAM J. Numer. Anal. 14, 441 (1977)
48. Himmelblau, D. : Applied Nonlinear Programming, p. 498, New York, McGraw-Hill 1972
49. Kuester, J. L. and Mize, J. H. : Optimization Techniques with FORTRAN, p. 500, New York,
McGraw-Hill 1974
50. Marsili-Libelli, S. and Castelli, M. : Appl. Math. and Comp. 23, n. 4, 341 (1987)
51. Bhatla, N. A. and Gaudy, A. F.: J. WPCF 38, 1441 (1966)
52. Marsili-Libelli, S. : Envir. Tech. Lett. 7, 341 (1986)
53. Jazwinski, A. H. : Stochastic Processes and Filtering Theory, p. 386, New York, Academic Press
1970
54, Gelb, D.: Applied Optimal Estimation, p. 374, Cambridge, Mass. M.I.T. Press 1974
55. Svrcek, N. Y., Elliot, R. F., and Zajic, J. E. : Biotech. Bioeng. 16, 827 (1974)
56. Cook, S. and Marsili-Libelli, S.: Wat. Sci. Tech. 13, 737 (1981)
57, Marsili-Libelli, S. : On-line Estimation of Bioactivities in Activated Sludge Processes, in : Modell-
ing and Control of Biotechnical Processes (ed. Halme, A.), p. 316, Oxford, Pergamon Press
1983
58. Goto, M. and Andrews, J. F. : On-line Estimation of Oxygen Uptake Rate in the Activated Studge
Process, in: Instrumentation and Control of Water and Wastewater Treatment and Transport
Systems (ed. Drake, R. A. R.), p. 867, Oxford, Pergamon Press 1985
59. Howell, J. A. and Sodipo, B. O. : On-line Respirometry and Estimation of Aeration Efficiencies
in an Activated Sludge Aeration Basin from Dissolved Oxygen Measurements, in: Modelling
and Control of Biotechnological Processes (ed. Johnson, A.) 191, Oxford, Pergamon Press 1985
60. Holmberg, U. and Olsson, G. : Simultaneous On-line Estimation of Oxygen Transfer Rate and
Respiration Rate, ibid., 185
61. Landau, I. D. and Lozano, R. : Automatica 17, 593 (1981)
62. Clarke, D. W. and Gawthrop, P. J. : Proc. IEE 126, 633 (1979)
63. Xianya, X. and Evans, R. J.: ibid. 131, Pt. D, 81 (1984)
64. Peterka, V. : Kibernetika 11, 53 (1975)
65. Bierman, G. J. : Factorization Methods for Discrete Sequential Estimation, p. 241, New York,
Academic Press 1977
66. Lawrence, E. W. and McCarty, P. L.: J. San. Eng. Div., ASCE 96, n. SA3,757 (1970)
67. Garrett, M. T. : Sew. and Ind. Wastes 30, n, 3,253 (1958)
68. Bisogni, J..J. and Lawrence, A. W. : Water Res. 5, n. 9, 753 (1971)
69. D'Ans, G., Kokotovic, P. V., and Gottlieb, D. : IEEE Trans. AC 16, 341 (1971)
70. D'Ans, G., Gottlieb, D., and Kokotovic, P. V. : Automatica 8, 729 (1972)
71. Muzychenko, L. A., Macheva, L. A., Yakovleva, G. V. : Biotech. Bioeng. 4, 629 (1974)
72. Kwaakernaak, H. and Sivan, R.: Linear Optimal Control Systems, p. 575, New York, Wiley-
Interscience 1972
73. Stephanopoulos, G. : Chemical Process Control : an introduction to theory and practice, p. 696,
Prentice Hall Int., Englewood Cliffs, N.J. 1984
74. Astrom, K. J. and Wittenmark, B. : Computer Controller Systems : theory and practice, p. 430,
Prentice-Hall Int., Englewood Cliffs, N.J. 1984
75. Marsili-Libelli, S. : Int. J. of Control 33, 601 (1981)
76. Hamalainen, R. P., Halme, A., and Gyllenberg, A. : A Control Model for Activated Sludge
Treatment Process, 6th Triennial IFAC World Congress, ISA, Pittsburg 1975
77. Aarinen, R., Tirkkonen, J., and Halme, A.: Experiences on instrumentation and control of
activated sludge plants -- a microprocessor application, 7th Triennial IFAC World Congress,
Pergamon Press, p. 255, 1978
78. Angelbeck, D. I., Shah Alam, A. B.: Simulation of optimal control strategies in a dynamic
continuous flow activated sludge system, Proc. 30th Industrial Waste Conference Purdue Uni-
versity, Ann Arbor Science PuN. p. 159 (t977)
148 S. Marsili-Libelll

79. Marsili-Libelli, S.: Optimal aeration control for wastewater treatment, in: Computer Aided
Design of Control Systems (ed. M. A. Cuenod), p. 511, Pergamon Press, Oxford 1980
80. Marsili-Libelli, S.: Trans. IMC 6, No. 3, 146 (1984)
81. S6rensen, P. E.: Pilot-scale evaluation of control schemes for the activated sludge process,
Water Quality Institute, Danish Academy of Technical Sciences, Horsholm, Technical Report
N. 1, p. 249 (1979)
82. Flanagan, M. J. : AIChE Symposium Series 75, No. 190, 232 (1979)
83. Sincic, D. and Bailey, J. E. : Water Res. 12, 47 (1978)
84. Yeung, S. Y. S., Sincic, D. and Bailey, J. E. : Water Res. 14, 77 (1980)
85. Stehfest, H. : Env. Tech. Letters 6, 556 (1985)
86. Marsili-Libelli, S. : Water Sci. Tech. 16, 613 (1984)
87. Yust, L. J., Stephenson, J. P., and Murphy, K. L.: Trans. IMC 6, n. 3, 165 (1984)
88. Tsugara, H., Sekine, T., Fujimoto, E., and Matsui, S. : Prediction and control of resident sludge
in a final clarifier, in: Instrumentation and Control of Water and Wastewater Treatment and
Transport Systems (ed. Drake, R. A. R.), p. 653, Oxford, Pergamon Press 1985
89. Schlegel, S. : Control ofthe sludge gravity thickening process, ibid. p. 391, Oxford Pergamon Press
1985
90. Severin, B. F., Poduska, R. A., Fogler, S. P., and Abrahamsen, T. A. : Novel use of steady-state
solids flux concepts for on-line clarifier control, ibid. p. 397, Oxford, Pergamon Press 1985
91. Levenspiel, O. : Chemical Reaction Engineering, p. 578, New York, John Wiley and Sons 1972
92. Keinath, T. M., Ryckman, M. D., Dana, C. H., and Hofer, D. A. : J. Env. Eng. Div. ASCE 103,
No. EE5, 829 (1977)
93. Craig, E. W., Meredith, D. D., Middleton, A. C.: J. Env. Eng. Div. ASCE 104, No. EE6, 1101
(1978)
94. Harris, C. J. and Billings, S. A. : Self-tuning and adaptive control: theory and application, p. 333,
IEE Press, London 1981
95. Astrom, K. J. : Automatica 19, 471 (1983)
96. Yuh-ju Ko, K., McInnis, B. C., and Goodwin, G. C. : Automatica 18, 727 (1982)
97. Cheruy, A., Panzarella, L., and Denat, J. P. : Multimodel simulation and adaptive stochastic
control of an activated sludge process, in: 1st IFAC Workshop on Modelling and Control of
Biotechnical Processes (ed. Halme, A.), p. 127, Pergamon Press, Oxford 1983
98. Olsson. G., Rundqwist, L., Eriksson, L., and Hall, L. : Self-tuning control of the dissolved oxygen
concentration in activated sludge systems, in: Instrumentation and Control of Water and Waste-
water Treatment and Transport Systems (ed. Drake, R. A. R.), p. 473, Oxford, Pergamon Press
1985
99. Cameron, F. and Seborg, D. E. : Int. J. of Control 38, 401 (1983)
100. Marsili-Libelli, S., Giardi R., Lasagni, M. : Env. Tech. Letters 6, 576 (1985)
Author Index Volumes 1-38

Acosta Jr., D. see Smith, R. V. Vol. 5, p. 69

Acton, R. T., Lynn, J. D. : Description and Operation of a Large-Scale Mammalian Cell, Suspension
Culture Facility. Vol. 7, p. 85
Agrawal, P., Lim, H. C. : Analysis of Various Control Schemes for Continuous Bioreactors. Vol. 30,
p. 61
Aiba, S. : Growth Kinetics of Photosynthetics Microorganisms. Vol. 23, p. 85
Aiba, S., Nagatani, M. : Separation of Cells from Culture Media. Vol. 1, p. 31
Aiba, S. see Sudo, R. Vol. 29, p. 117
Aiba, S., Okabe, M. : A Complementary Approach to Scale-Up. Vol. 7, p. 111
Alfermann, A. W. see Reinhard, E. Vol. 16, p. 49
Anderson, L. A., Phillipson, J. D., Roberts, M. F. : Biosynthesis of Secondary Products by Cell
Cultures of Higher Plants. Vol. 31, p. 1
Arnaud, A. see Jallageas, J.-C. Vol. 14, p. 1
Arora, H. L., see Carioca, J. O. B. Vol. 20, p. 153
Asher, Z. see Kosaric, N. Vol. 32, p. 25
Atkinson, B., Daoud, I. S. : Microbial Flocs and Flocculation. Vol. 4, p. 41
Atkinson, B., Fowler, H. W. : The Significance of Microbial Film in Fermenters. Vol. 3, p. 221

Bagnarelli, P., Clementi, M. : Serum-Free Growth of Human Hepatoma Cells. Vol. 34, p. 85
Barford, J.-P., see Harbour, C. Vol. 37, p. 1
Barker, A. A., Somers, P. J. : Biotechnology of Immobilized Multienzyme Systems. Vol. 10, p. 27
Beardmore, D. H. see Fan, L. T. Vol. 14, p. 101
Bedetti, C., Cantafora, A. : Extraction and Purification of Arachidonic Acid Metabolites from Cell
Cultures. Vol. 35, p. 47
Belfort, G. see Heath, C. Vol. 34, p. 1
Beker, M. J., Rapoport, A. J. : Conservation of Yeasts by Dehydration. Vol. 35, p. 127
Bell, D. J., Hoare, M., Dunnill, P.: The Formation of Protein Precipitates and their Centrifugal
Recovery. Vol. 26, p. 1
Berlin, J., Sasse, F. : Selection and Screening Techniques for Plant Cell Cultures. Vol. 31, p. 99
Binder, H. see Wiesmann, U. Vol. 24, p. 119
Bjare, M. : Serum-Free Cultivation of Lymphoid Cells. Vol. 34, p. 95
Blanch, H. W., Dunn, I. J. : Modelling and Simulation in Biochemical Engineering. Vol. 3, p. 127
Blanch, H. W., see Moo-Young, M. Vol. 19, p. 1
Blanch, H. W., see Maiorella, B. Vol. 20, p. 43
Blenke, H. see Seipenbusch, R. Vol. 15, p. 1
Blenke, H.: Loop Reactors. Vol. 13, p. 121
Blumauerov6, M. see Hostalek, Z. Vol. 3, p. 13
Bdhme, P. see Kopperschl/iger, G. Vol. 25, p. 101
150 Author Index Volumes 1-38

Bottino, P. J. see Gamborg, O. L. Vol. 19, p. 239

Bowers, L. D., Carr, P. W. : Immobilized Enzymes in Analytical Chemistry. Vol. 15, p. 89
Brauer, H. : Power Consumption in Aerated Stirred Tank Reactor Systems. Vol. 13, p. 87
Brodelius, P. : Industrial Applications o f Immobilized Biocatalysts. Vol. 10, p. 75
Brosseau, J. D. see Zajic, J. E. Vol. 9, p. 57
Bryant, J.: The Characterization of Mixing in Fermenters. Vol. 5, p. 101
Buchholz, K. : Reaction Engineering Parameters for Immobilized Biocatalysts. Vol. 24, p. 39
Bungay, H. R. : Biochemical Engineering for Fuel Production in United States. Vol. 20, p. 1
Butler, M. : Growth Limitations in Microcarriers Cultures. Vol. 34, p. 57

Cantafora, A. see Bedetti, C. Vol. 35, p. 47

Chan, Y. K. see Schneider, H. Vol. 27, p. 57
Carioca, J. O. B., Arora, H. L., Khan, A. S. : Biomass Conversion~Program in Brazil. Vol. 20, p. 153
Cart, P. W. see Bowers, L. D. Vol. 15, p. 89
Chang, M. M., Chou, T. Y. C., Tsao, G. T. : Structure, Pretreatment, and Hydrolysis of Cellulose.
Vol. 20, p. 15
Charles, M. : Technical Aspects of the Rheological Properties o f Microbial Cultures. Vol. 8, p. 1
Chen, L. F., see Gong, Ch.-S. Vol. 20, p. 93
Chou, T. Y. C., see Chang, M. M. Vol. 20, p. 15
Cibo-Geigy/Lepetit: Seminar on Topics of Fermentation Microbiology. Vol. 3, p. 1
Claus, R. see Haferburg, D. Vol. 33, p. 53
Clementi, P. see Bagnarelli, P. Vol. 34, p. 85
Cogoli, A., Yschopp, A. : Biotechnology in Space Laboratories. Vol. 22, p. 1
Cooney, C. L. see Koplove, H. M. Vol. 12, p. 1
Costentino, G. P. see Kosaric, N. Vol. 32, p. 1

Daoud, L S. see Atkinson, B. Vol. 4, p. 41

Das, K. see Ghose, T. K. Vol. 1, p. 55
Davis, P. J. see Smith, R. V. Vol. 14, p. 61
Deekwer, W.-D. see Schumpe, A. Vol. 24, p. 1
Demain, A. L. : Overproduction of Microbial Metabolites and Enzymes due to Alteration of Regula-
tion. Vol. 1, p. 113
Doelle, H. W., Ewings, K. N., Hollywood, N. W. : Regulation of Glucose Metabolism in Bacterial
Systems. Vol. 23, p. 1
Dunn, L J. see Blanch, H. W. Vol. 3, p. 127
Dunnill, P. see Bell, D. J.Vol. 26, p. 1
Dm,njak, Z., see Kosaric, N. Vol. 20, p. 119
Duvnjak, Z. see Kosaric, N. Vol. 32, p. 1

Eckenfelder Jr., W. W., Goodman, B. L., Englande, A. J. : Scale-Up of Biological Wastewater Treat-
merit Reactors. Vo!. 2, p. 145
Einsele, A., Fiechter, A. : Liquid and Solid Hydrocarbons. Vol. 1, p. 169
Electricwala, A. see Griffiths, J. B. Vol. 34, p. 147
Enari, T. M., Markkanen, P. : Production of Cellulolytic Enzymes by FunN. Vol. 5, p. 1
Enatsu, T., Shinmyo, A. : In Vitro Synthesis of Enzymes. Physiological Aspects of Microbial Enzyme
Production Vol. 9, p. 111
Author Index Volumes 1-38 151

Engels, J., Uhlmann, E. : Gene Synthesis. Vol. 37, p. 73

Englande, A. J. see Eckenfelder Jr., W. W. Vol. 2, p. 145
Eriksson, K. E. : Swedish Developments in Biotechnology Bossed on Lignocellulose Materials. Vol. 20,
p. 193
Esser, K.: Some Aspects o f Basic Genetic Research on Fungi and Their Practical Implications.
Vol. 3, p. 69
Esser, K., Lang-Hinrichs, Ch. : Molecular Cloning in Heterologous Systems, Vol. 26, p. 143
Ewings, K. N. see Doelle, H. W. Vol. 23, p. 1

Faith, W. T., Neubeck, C. E., Reese, E. T. : Production and Application of Enzymes. Vol. 1, p. 77
Fan, L. S. see Lee, Y. H. Vol. 17, p. 131
Fan, L. T., Lee, Y.-H., Beardmore, D. H.: Major Chemical and Physical Features of Cellulosic
Materials as Snbstrates for Enzymatic Hydrolysis. Vol. 14, p. 101
Fan, L. T., Lee, Y.-H., Gharppuray, M. M. : The Nature of Lignocellulosics and Their Pretreatments
for Enzymatic Hydrolysis. Vol. 23, p. 155
Fan, L. T. see Lee, Y.-H. Vol. 17, p. 101 and p. 131
Faust, U., Sittig, W. : Methanol as Carbon Source for Biomass Production in a Loop Reactor. Vol. 17,
p. 63
Fiechter, A. : Physical and Chemical Parameters of Microbial Growth. Vol. 30, p. 7
Fiechter, A. see Einsele, A. Vol. 1, p. 169
Fiechter, A. see Janshekar, H. Vol. 27, p. 119
Finocchiaro, T., Olson, N. F., Richardson, Y. : Use of Immobilized Lactase in Milk Systems. Vol. 15,
p. 71
Flaschel, E. see Wandrey, C. Vol. 12, p. 147
Flaschel, E., Wandrey, Ch., Kula, M.-R. : Ultrafiltration for the Separation o f Biocatalysts. Vol. 26,
p. 73
Flickinger, M. C. see Gong, Ch.-S. Vol. 20. p. 93
Fowler, H. W. see Atkinson, B. Vol. 3, p. 221
Fukui, S., Tanaka, A. : Application of Biocatalysts Immobilized by Prepolymer Methods. Vol. 29,
p. 1
Fukui, S., Tanaka, A. : Metabolism of Alkanes by Yeasts. Vol. 19, p. 217
Fukui, S., Tanaka, A. : Production of Useful Compounds from Alkane Media in Japan, Vol. 17, p. 1

Galzy, P. see Jallageas, J.-C. Vol. 14, p. 1

Garnborg, O. L., Bottino, P. J. : Protoplasts in Genetic Modifications of Plants. Vol. 19, p. 239
Gaudy Jr., A. F., Gaudy, E. T. : Mixed Microbial Populations. Vol. 2, p. 97
Gaudy, E. T. see Gaudy Jr., A. F. Vol. 2, p. 97
Gharpuray, M. M. see Fan, L. T. Vol. 23, p. 155
Ghose, T. K., Das, K. : A Simplified Kinetic Approach to Celtulose-Cellulase System. Vol. 1, p. 55
Ghose, T. K. : Cellulase Biosynthesis and Hydrolysis of Cellulosic Substances. Vol. 6, p. 39
Gogotov, I. N. see Kondratieva, E. N. Vol. 28, p. 139
Gomez, R. F. : Nucleic Acid Damage in Thermal Inactivation of Vegetative Microorganisms. Vol. 5,
p. 49
Gong, Ch.-S. see McCracken, L. D. Vol. 27, p. 33
Gong, Ch.-S., Chen, L. F., Tsao, G. T., Flickinger, M. G. : Conversion of Hemicellulose Carbohydrates,
Vol. 20, p. 93
Goodman, B. L. see Eckenfelder Jr., W." W. Vol. 2, p. 145
152 Author Index Volumes I 3S

Graves, D. J., Wu, Y.-T. : The Rational Design of Affinity Chromatography Separation Processes.
Vol. 12, p. 219
Gr~ffiths, J. B., Electricwala, A. : Production of Tissue Plasminogen Activators from Animal Cells.
Vol. 34, p. 147
Gutschick, V. P. : Energetics of Microbial Fixation of Dinitrogen. Vol. 21, p. 109

Haferberg, D., Hommel, R., Claus, R., Kleber, H.-P. : Extracellular Microbial Lipids as Biosurfac-
tants. Vol. 33, p. 53
Hahlbrock, K., Schr6der, J., Vieregge, J. : Enzyme Regulation in Parsley and Soybean Cell Cultures,
Vol. 18, p. 39
Haltmeier, Th. : Biomass Utilization in Switzerland. Vol. 20, p. 189
Hampel, W.: Application of Microcomputers in the Study of Microbial Processes. Vol. 13, p. 1
Harbour, C., Barford, J.-P., Low, K.-S. : Process Development for Hybridoma Cells. Vol. 37, p. 1
Harder, A., Roels, J. A. : Application of Simple Structures Models in Bioengineering. Vol. 21, p. 55
Harrison, 1). E. F., Topiwala, H. H. : Transient and Oscillatory States of Continuous Culture. Vol. 3,
p. 167
Heath, C., Belfort, C. : Immobilization of Suspended Mammalian Cells: Analysis of Hollow Fiber
and Microcapsule Bioreactors. Vol. 34, p. 1
Hedman, P. see Janson, J.-C. Vol. 25, p. 43
Heinzle, E. : Mass Spectrometry for On-line Monitoring of Biotechnological Processes. Vol. 35, p. 1
Ho, Ch., Smith, M. D., Shanahan, J. F. : Carbon Dioxide Transfer in Biochemical Reactors. Vol. 35,
p. 83
Hoare, M. see Bell, D. J. Vol. 26, p. 1
Hofmann, E. see Kopperschlfiger, G. Vol. 25, p. 101
Hall6, J. see Nyeste, L. Vol. 26, p. 175
Hollywood, N. W. see Doelle, H. W. Vol. 23, p. 1
Hommel, R. see Haferburg, D. Vol. 33, p. 53
Hogi61ek, Z., Blumauerov~, M., Vanek, Z. : Genetic Problems of the Biosynthesis of Tetracycline
Antibiotics. Vol. 3, p. 13
Hu, G. Y. see Wang, P. J. Vol. 18, p. 61
Humphrey, A. E., see Rolz, G. E. Vol. 21, p. 1
Hustedt, H. see Kula, M.-R. Vol. 24, p. 73

Imanaka, T.: Application of Recombinant DNA Technology to the Production of Useful Bio-
materials. Vol. 33, p. 1
Inculet, L L see Zajic, J. E. Vol. 22, p. 51

Jack, T. R., Zajic, J. E. : The Immobilization of Whole Cells. Vol. 5, p. 125

Jallageas, J.-C., Arnaud, A., Galzy, P. : Bioconversions of Nitriles and Their Applications. Vol. 14, p. 1
Jang, C. M., Tsao, G. T.: Packed-Bed Adsorption Theories and Their Applications to Affinity
Chromatography. Vol. 25, p. 1
Jang, C. M., Tsao, G. T. : Affinity Chromatography. Vol. 25, p. 19
Jansen, N. B., Tsao, G. T. : Bioconversion of Pentoses to 2,3-Butanediol by Klebsiella pneumonia.
Vol. 27, p. 85
Janshekar, H., Fiechter, A. : Lignin Biosynthesis, Application, and Biodegradation. Vol. 27, p. 119
Janson, J.-C., Hedman, P. : Large-Scale Chromatography of Proteins. Vol. 25, p. 43
Jeffries, Th. W.: Utilization of Xylose by Bacteria, Yeasts, and Fungi. Vol. 27, p. 1
Jiu, J.: Microbial Reactions in Prostaglandin Chemistry, Vol. 17, p. 37
Author Index Volumes 1-38 153

Kamihara, T., Nakamura, I. : Regulation of Respiration and Its Related Metabolism by Vitamin B1
and Vitamin B6 in Saccharomyces Yeasts. Vol. 29, p. 35
Keenan, J. D. see Shieh, W. K. Vol. 33, p. 131
Khan, A. S., see Carioca, J. O. B. Vol. 20, p. 153
Kirnura, A. : Application of recDNA Techniques to the Production of ATP and Glutathione by the
"Syntechno System". Vol. 33, p. 29
King, C.-K. see Wang, S. S. Vol. 12, p. 119
King, P. J.: Plant Tissue Culture and the Cell Cycle, Vol. 18, p. 1
Kjaergaard, L. : The Redox Potential: Its Use and Control in Biotechnology. Vol. 7, p. 131
Kleber, H.-P. see Haferburg, D. Vol. 33, p. 53
Kleinstreuer, C., Poweigha, T. : Modeling and Simulation of Bioreactor Process Dynamics. Vol. 30,
p. 91
Kochba, J. see Spiegel-Roy, P. Vol. 16, p. 27
Kondratieva, E. N., Gogotov, I. N. : Production of Molecular Hydrogen in Microorganism. Vol. 28,
p. 139
Koplove, H. M., Cooney, C. L. : Enzyme Production During Transient Growth. Vol. 12, p. 1
Kopperschliiger, G., B6hme, H.-J., Hofmann, E. : Cibacron Blue F3G-A and Related Dyes as Ligands
in Affinity Chromatography. Vol. 25, p. 101
Kosaric, N., Asher, Y. : The Utilization of Cheese Whey and its Components. Vol. 32, p. 25
Kosaric, N., Duvnjak, Z., Stewart, G. G. : Fuel Ethanol from Biomass Production, Economics, and
Energy. Vol. 20, p. 119
Kosaric, N. see Magee, R. J. Vol. 32, p. 61
Kosaric, N., Wieczorek, A., Cosentino, G. P., Duvnjak, Z. : Industrial Processing and Products from
the Jerusalem Artichoke. Vol. 32, p. 1
Kosaric, N., Zajic, J. E. : Microbial Oxidation of Methane and Methanol. Vol. 3, p. 89
Kosaric, N. see Zajic, K. E. Vol. 9, p. 57
Kossen, N. W. F. see Metz~ B. Vol. 11, p. 103
Kristapsons, M. Z. see Viesturs, U. Vol. 21, p. 169
Kroner, K. H. see Kula, M.-R. Vol. 24, p. 73
Kula, M.-R. see Flaschel, E. Vol. 26, p. 73
Kula, M.-R., Kroner, K. H., Hustedt, H. : Purification of Enzymes by Liquid-Liquid Extraction.
Vol. 24, p. 73
Kurtzman, C. P. : Biology and Physiology of the D-Xylose Degrading Yeast Pachysolen tannophilus.
Vol. 27, p. 73

Lafferty, R. M. see Schlegel, H. G. Vol. 1, p. 143

Lambe, C. A. see Rosevear, A. Vol. 31, p. 37
Lang-Hinrichs, Ch. see Esser, K. Vol. 26, p. 143
Lee, K. J. see Rogers, P. L. Vol. 23, p. 37
Lee, Y.-H. see Fan, L. T. Vol. 14, p. 101
Lee, Y.-H. see Fan, L. T. Vol. 23, p. 155
Lee, Y.-H., Fan, L. T., Fan, L. S. : Kinetics of Hydrolysis of Insoluble Cellulose by Cellulase, Vol. 17,
p. 131
Lee, Y.-H., Fan, L. T. : Properties and Mode of Action of Cellulase, Vol. 17, p. 101
Lee, Y.-H., Tsao, G. T. : Dissolved Oxygen Electrodes. Vol. 13, p. 35
Lehmann, J. see Schiigerl, K. Vol. 8, p. 63
Levitans, E. S. see Viesturs, U. Vol. 21, p. 169
Lira, H. C. see Agrawal, P. Vol. 30. p. 61
154 Author Index Volumes 1-38

Lira, H. C. see Parulekar, S. J. Vol. 32, p. 207

Linko, M. : An Evaluation of Enzymatic Hydrolysis of Cellulosic Materials. Vot. 5, p. 25
Linko, M. : Biomass Conversion Program in Finland, Vol. 20, p. 163
Low, K.-S., see Harbour, C. Vol. 37, p. 1
Liicke, J. see Schiigerl, K. Vol. 7, p. 1
Liicke, J. see Schfigerl, K. Vol. 8, p. 63
Luong, J. H. T., Volesky, B. : Heat Evolution During the Microbial Process Estimation, Measurement,
and Application. Vol. 28, p. 1
Luttmann, R., Munack, A., Thoma, M. : Mathematical Modelling, Parameter Identification and
Adaptive Control o f Single Cell Protein Processes in Tower Loop Bioreactors. Vol. 32, p. 95
Lynd, L. R. : Production of Ethanol from Lianocellulosic Materials Using Thermophilic Bacteria:
Critical Evaluation of Potential and Review. Vol. 38, p. 1
Lynn, J. D. see Acton, R. T. Vol. 7, p. 85

MacLeod, A. J.: The Use of Plasma Protein Fractions as Medium Supplements for Animal Cell
Culture. Vol. 37, p. 41
Magee, R. J., Kosaric, N. : Bioconversion of Hemicellulosics. Vol. 32, p. 61
Maiorella, B,, Wilke, Ch. R., Blanch, H. W. : Alcohol Production and Recovery. Vol. 20, p. 43
M6lek, L : Present State and Perspectives of Biochemical Engineering. Vol. 3, p. 279
Maleszka, R. see Schneider, H. Vol. 27, p. 57
Mandels, M. : The Culture of Plant Cells. Vol. 2, p. 201
Mandels, M. see Reese, E. T. Vol. 2, p. 181
Mangold, H. K. see Radwan, S. S. Vol. 16, p. 109
Markkanen, P. see Enari, T. M. Vol. 5, p. 1
Marsili-Libelli, St. : Modelling, Identification and Control of the Activated Sludge Process. Vol. 38,
p. 89
Martin, J. F.: Control of Antibiotic Synthesis by Phosphate. Vol. 6, p. 105
Martin, P. see Zajic, J. E. Vol. 22, p. 51
McCracken, L. D., Gong, Ch.-Sh. : D-Xylose Metabolism by Mutant Strains o f Candida sp. Vol. 27,
p. 33
Misawa, M. : Production of Useful Plant Metabolites. Vol. 31, p. 59
Miura, Y. : Submerged Aerobic Fermentation. Vol. 4, p. 3
Miura, Y. : Mechanism of Liquid Hydrocarbon Uptake by Microorganisms and Growth Kinetics.
Vol. 9, p. 31
Messing, R. A. : Carriers for Immobilized Biologically Active Systems. Vol. 10, p. 51
Metz, B., Kossen, N. W. F., van Suijidam. J. C. : The Rheology of Mould Suspensions. Vol. 11, p. 103
Moo-Young, M., Blanch, H. W. : Design o f Biochemical Reactors Mass Transfer Criteria for Simple
and Complex Systems. Vol. 19. p. 1
Moo-Young, M. see Scharer, J. M. Vol. 11, p. 85
Morandi, M., Valeri, A. : Industrial Scale Production o f 13-Interferon. Vol. 37, p. 57
Munack, A. see Luttmann, R. Vol. 32, p. 95

Nagai, S. : Mass and Energy Balances for Microbial Growth Kinetics. Vol. I 1, p. 49
Nagatani, M. see Aiba, S. Vol. 1, p. 31
Nakarnura, L see Kamihara, T. Vol. 29, p. 35
Neubeek, C. E. see Faith, W. T. Vol. 1, p. 77
Neirinck, L. see Schneider, H. Vol. 27, p. 57
Author Index Volumes 1-38 155

Nyeste, L., P6cs, M., Sevella, B., Holt6, J. : Production of r~-Tryptophan by Microbial Processes,
Vol. 26, p. 175
Nyiri, L. K. : Application of Computers in Biochemical Engineering. Vol. 2, p. 49

O'Driscoll, K. F.: Gel Entrapped Enzymes. Vol. 4, p. 155

Oels, U. see Schiigerl, K. Vol. 7, p. 1
Okabe, M. see Aiba, S. Vol. 7, p. 111
Olson, N. F. see Finocchiaro, T. Vol. 15, p. 71

Pace, G. W., Righelato, C. R. : Productio~ of Extracellular Microbial. Vol. 15, p. 41

Parisi, F. : Energy Balances for Ethanol as a Fuel. Vol. 28, p. 41
Parisi, F. : Advances in Lignocellulosic Hydrolysis and in the Utilization of the Hydrolyzates. Vol. 38,
p. 53
Parulekar, S. J., Lim, H. C. : Modelling, Optimization and Control of Semi-Batch Bioreactors.
Vol. 32, p. 207
Pkcs, M. see Nyeste, L. Vol. 26, p. 175
Phitlipson, J. D. see Anderson, L. A. Vol. 31, p. 1
Pitcher Jr., W. H. : Design and Operation of Immobilized Enzyme Reactors. Vol. 10, p. 1
Potgieter, H. J.: Biomass Conversion Program in South Africa. Vol. 20, p. 181
Poweigha, T. see Kleinstreuer, C. Vol. 3i3, p. 91

Quicker, G. see Schumpe, A. Vol. 24, p. 1

Radtett, P. J. : The Use Baby Hamster Kidney (BHK) Suspension Cells for the Production o f Foot
and Mouth Disease Vaccines. Vol. 34, p. 129
Radwan, S. S., Mangold, H. K. : Biochemistry o f Lipids in Plan.t Cell Cultures. Vol. 16, p. 109
Rarnkrishna, D. : Statistical Models of Cell Populations. Vol. 11, p. 1
Rapoport, A. J. see Beker, M. J. Vol. 35, p. 127
Reese, E. T. see Faith, W. T. Vol. 1, p. 77
Reese, E. T., Mandels, M., Weiss, A. H. : Cellulose as a Novel Energy Source. Vol. 2, p. 181
~eh6dek, Z. : Ergot Alkaloids and Their Biosynthesis. Vol. 14, p. 33
Rehm, H.-J., Reiff, I. : Mechanism and Occurrence of Microbial Oxidation of Long-Chain Alkanes,
Vol. 19, p. 175
Reiff, L see Rehm, H.-J. Vol. 19, p. 175
Reinhard, E., Alfermann, A. W. : Biotransformation by Plant Cell Cultures. Vol. 16, p. 49
Reuven),, S. see Shahar, A. Vol. 34, p. 33
Richardson, T. see Finocchiaro, T. Vol. 15, p. 71
Righelato, R. C. see Pace, G. W. Vol. 15, p. 41
Roberts, M. F. see Anderson, L. A. Vol. 31, p. 1
Rods, J. A. see Harder, A. Vol. 21, p. 55
Rogers, P. L. : Computation in Biochemical Engineering. Vol. 4, p. 125
Rogers, P. L., Lee, K. J., Skotnicki, M. L., Tribe, D. E. : Ethanol Production by Zymomonas Mobilis.
Vol. 23, p. 37
Rolz, C., Humphrey, A. : Microbial Biomass from Renewables: Review of Alternatives. Vol. 21, p. 1
Rosazza, J. P. see Smith, R. V. Vol. 5, p. 69
Rosevear, A., Lambe, C. A.: Immobilized Plant Cells. Vol. 31, p. 37
156 Author Index Volumes 1-38

Sahm, H. : Anaerobic Wastewater Treatment. Vol. 29. p. 83

Sahm, H. : Metabolism of Methanol by Yeasts. Vol. 6, p. 77
Sahm, H.: Biomass Conversion Program of West Germany. Vol. 20, p. 173
Sasse, F. see Berlin, J. Vol. 31, p. 99
Scharer, J. M., Moo-Young, M. : Methane Generation by Anaerobic Digestion of Cellulose-Con-
taining Wastes. Vol. 11, p. 85
Schlegel, H. G., Lafferty, R. M. : The Production of Biomass from Hydrogen and Carbon Dioxide.
Vol. l, p. 143
Schmid, R. D. : Stabilized Soluble Enzymes. Vol. 12, p. 41
Schneider, H., Maleszka, R., Neirinck, L., Veliky, I. A., Chan, Y. K., Wang, P. Y. : Ethanol Produc-
tion from D-Xylose and Several Other Carbohydrates by Pachysolen tannophilus. Vol. 27, p. 57
Schr6der, J. see Hahlbrock, K. Vol. 18, p. 39
Schumpe, A., Quicker, G., Deckwer, W.-D. : Gas Solubilities in Microbial Culture Media. Vol. 24,
p. 1
Sehiigerl, K. : Oxygen Transfer Into Highly Viscous Media. Vol. 19, p. 71
Schiigerl, K.: Characterization and Performance of Single- and Multistage Tower Reactors with
Outer Loop for Cell Mass Production. Vol. 22, p. 93
Schiigerl, K., Oels, U., Lticke, J. : Bubble Column Bioreactors. Vol. 7, p. l
Schiigerl, K., Lficke, J., Lehmann, J., Wagner, F. : Application of Tower Bioreactors in Cell Mass
Production. Vol. 8, p. 63
Schwab, H. : Strain Improvement in Industrial Microorganisms by Recombinant DNA Techniques.
Vol. 37, p. 129
Seipenbusch, R., Blenke, H. : The Loop Reactor for Cultivating Yeast on n-Praffin Substrate. Vol. 15,
p. 1
Sevella, B. see Nyeste, L. Vol. 26, p. 17_
Shahar, .4., Reuveny, S. : Nerve and Muscle Cells on Microcarriers Culture. Vol. 34, p. 33
Shanahan, J. F. see Ho, Ch. S. Vol. 35, p. 83
Shieh, W. K., Keenan, J. D.: Fluidized Bed Biofilm Reactor for Wastewater Treatment. Vol. 33,
p. 131
Shimizu, S. see Yaman~, T. Vol. 30, p. t47
Shinmyo, A. see Enatsu, T. Vol. 9, p. 111
Sittig, W., see Faust, U. Vol. 17, p. 63
Skotnicki, M. L. see Rogers, P. L. Vol. 23, p. 37
Smith, M. D. see Ho, Ch. S. Vol. 35, p. 83
Smith, R. V., Acosta Jr., D., Rosazza, J. P. : Cellular and Microbial Models in the Investigation of
Mammalian Metabolism of Xenobiotics. Vol. 5, p. 69
Smith, R. V., Davis, P. J. : Induction of Xenobiotic Monooxygenases. Vol. 14, p. 61
Soda, K. see Yonaha, K. Vol. 33, p. 95
Solomon, B. : Starch Hydrolysis by Immobilized Enzymes. Industrial Application. Vol. 10, p. 131
Somers, P. J. see Barker, S. A. Vol. 10. p. 27
Sonnleitner, B.: Biotechnology of Thermophilic Bacteria: Growth, Products, and Application.
Vol. 28, p. 69
Spiegel-Roy, P., Kochba, J. : Embryogenesis in Citrus Tissue Cultures. Vol. 16, p. 27
Spier, R. E. : R6cent Developments in the Large Scale Cultivation of Animal Cells in Monolayers.
Vol. 14, p. 119
Stewart, G. G. see Kosaric, N. Vol. 20, p. 119
Stohs, S. J. : Metabolism of Steroids in Plant Tissue Cultures. Vol. 16, p. 85
Sudo, R., Aiba, S. : Role and Function of Protozoa in the Biological Treatment of Polluted Waters.
Vol. 29, p. 117
Author Index Volumes 1-38 157

Suijidam, van, J. C. see Metz, N. W. Col. 11, p. 103

Sureau, P. : Rabies Vaccine Production in Animal Cell Cultures. Vol. 34, p. 111
Szczesny, T. see Volesky, B. Vol. 27, p. 101

Taguchi, H. : The Nature of Fermentation Fluids. Vol. 1, p. 1

Tanaka, A. see Fukui, S. Vol. 17, p. l and Vol. 19, p. 217
Tanaka, A. see Fukui, S. Vol. 29, p. 1
Thoma, M. see Luttmann, R. Vol. 32, p. 95
Topiwala, H. H. see Harrison, D. E. F. Vol. 3, p. 167
Torma, A. E. : The Role of Thiobacillus Ferrooxidans in Hydrometallurgical Processes. Vol. 6, p. 1
Tran Than Van, K.: Control of Morphogenesis or What Shapes a Group of Cells? Vol. 18, p. 151
Tribe, D. E. see Rogers, P. L. Vol. 23, p. 37
Tsao, G. T. see Lee, Y. H. Vol. 13, p. 35
Tsao, G. T. see Chang, M. M. Vol. 20, p. 93
Tsao, G. T. see Jang, C.-M. Vol. 25, p. 1
Tsao, G. T. see Jang, C.-M. Vol. 25, p. 19
Tsao, G. T. see Jansen, N. B. Vol. 27, p. 85
Tschopp, A. see Cogoli, A. Vol. 22, p. 1

l~hlmann, E. see Engels, J. Vol. 37, p. 73

Ursprung, H. : Biotechnology: The New Change for Industry. Vol. 30, p. 3

Valeri, A. see Morandi, M. Vol. 37, p. 57

Vanek, Z. see Hostalek, Z. Vol. 3, p. 13
Veliky, L A. see Schneider, H. Vol. 27, p. 57
Vieregge, J. see Hahlbrock, K. Vol. 18, p. 39
Viesturs, U. E., Kristapsons, M. Z., Levitans, E. S., Foam in Microbiological Processes. Vol. 21, p. 169
Volesky, B., Szczesny, T. : Bacterial Conversion of Pentose Sugars to Acetone and Butanol. Vol. 27,
p. 101
Volesky, B. see Luong, J. H. T. Vol. 28, p. l

Wagner, F. see Schfigerl, K. Vol. 8, p. 63

Wandrey, Ch., Flaschel, E. : Process Development and Economic Aspects in Enzyme Engineering
Acylase L-Methionine System. Vol. 12, p. 147
Wandrey, Ch. see Flaschel, E. Vol. 26, p. 73
Wang, P. J., Hu, C. J. : Regeneration of Virus-Free Plants Through in Vitro Culture. Vol. 18, p. 61
Wang, P. Y. see Schneider, H. Vol. 27, p. 57
Wang, S. S., King, C.-K. : The Use of Coenzymes in Biochemical Reactors. Vol. 12, p. 119
Weiss, A. H. see Reese, E. T., Vol. 2, p. 181
Wieczorek, A. see Kosaric, N. Vol. 32, p. 1
Wilke, Ch. R., see Maiorella, B. Vol. 20, p. 43
Wil~'on, G. : Continuous Culture o f Plant Cells Using the Chemostat Principle. Vol. 16, p. 1
Wingard Jr., L. B. : Enzyme Engineering Col. 2, p. 1
Wiesmann, U., Binder, H. : Biomass Separation from Liquids by Sedimentation and Centrifugation.
Vol. 24, p. 119
158 Author Index Volumes 1 38

Withers, L. A. : Low Temperature Storage of Plant Tissue Cultures. Vol. 18, p. 101
Wu, Y.-T. see Graves, D. J. Vol. 12, p. 219

Yamada, Y. : Photosynthetic Potential of Plant Cell Cultures. Vol. 31, p. 89

Yamank, T., Shimizu, S. : Fed-batch Techniques in Microbial Processes. Vol. 30, p. 147
Yarovenko, V. L. : Theory and Practice of Continuous Cultivation of Microorganisms in Industrial
Alcoholic Processes. Vol. 9, p. 1
Yonaha, K., Soda, K. : Applications of Stereoselectivity of Enzymes: Synthesis of Optically Active
Amino Acids and ~-Hydroxy Acids, and Stereospecific Isotope-Labeling of Amino Acids, Amines
and Coenzymes. Vol. 33, p. 95

Zajic, J. E. see Kosaric, N. Vol. 3, p. 89

Zajic, J. E. see Jack, T. R. Vol. 5, p. 125
Zajic, J. E., Kosaric, N., Brosseau, J. D. : Microbial Production of Hydrogen. Vol. 9, p. 57
Zajic, J. E., Inculet, I. I., Martin, P. : Basic Concepts in Microbial Aerosols. Vol. 22, p. 5I
Zlokarnik, M. : Sorption Characteristics for Gas-Liquid Contacting in Mixing Vessels. Vol. 8, p. 133
Zlokarnik, M. : Scale-Up of Surface Aerators for Waste Water Treatment. Vol. 11, p. 157