Biophysical Chemistry 18 (1983) 73-87
Elsevier
BPC 00790
73
zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
Review
SIMPLIFYING
Wlodzimierz
PRINCIPLES
KLONOWSKI
FOR CHEMICAL
AND
ENZYME
REACTION
KINETICS
*
Received
26th January 1983
Accepted 14th April I983
Tihonov’s Theorems for systems of first-order ordinary diffrxential equations containing small Parameters in the derivativea.
which for m the mathematical foundation of the stendy-state approximation, are restated. A general procedure for simplifying
chemical and enzyme reaction kinetics. based on the difference of characteristic time scales. is presented. Korzuhin’s Theorem.
which makes it possible to approximate any kinetic sysrem by il closed chemical system. is also reported. The notions and
theorrms are illustrated with examples of Michnehs-Menten enzyme kinetics and of a simple autocatalytic system. Another
example illustrates how the differences in the rate constants of different elementary reactions may be exploited 10 simplify
reaction kinetics by using Tihonov’s Theorem. AI1 necessary mathematical notions are explained in the appendices. The most
simple formulation of Tihonov’s 1st Theorem ‘for beginners’ is also given.
zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLK
1. Introduction
In their recent paper, Kijima and Kijima [I]
stated in section 1 that “the steady-state assumption or steady-state treatment in chemical reaction
kinetics (. _ . ) has been used so far without proof.”
And once again in section 3 “_ _ . there has been no
general study on the condition when the steadystate approximation
holds even on the first-order
reaction.”
It seems that the authors are not familiar with
some important work of such authors as A.N.
Tihonov,
L.S. Pontryagin.
A.B. Vasil’eva,
V-F.
Bu:uzov. V.M. Volosov,
IS. Gradstein
and V.
Vazov.
Especially
in Tihonov’s
work [2], the
mathematical problems which are the very basis of
* Gn !eave from: Medical Research Center. Polish Academy of
Sciences. Bialobrreska 58. 02-32.5 Warsaw. Poland, and from
FacuItti des Sciences. I’Universitti de Kinshasa. B.P. 190.
Kinshasa XI, Zaire.
N.B.
In different sources the name of A.N. Tihonov is often
written as Tikhonov, Tichonov. Tichonoff. Tychonoff. etc.
the steady-state approximation
(i.e.. the theory of
systems of ordinary first-order differential
equations (SFO)
containing small parameters in the
derivatives) are considered in a very general
manner. The theorems he proved therein, called by
other Soviet authors Tihonov’s Thecrems. may be
applied to systems with reactions of any order and
with any number of components.
In appendix
A we reproduce
the English
summary of Tihonov’s
paper [2]. by J.L. Massera
from Mathematical Reviews [3].
It is interesting
to note that the theory of
ordinary differential equations with slowly varying
coefficients
(see. for example, ref. 4) is in some
sense equivalent to the theory of differential equations containing small parameters in the derivatives. For example, the equation
pdx/dr’=p(r’).r
(p”
by the transformation
(time)
I = 1*/p
I)
of the independent
(1)
variab!e
(2)
is transformed
dr/dr
into the form
zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
and its solution, determined by initial conditions:
= p(pr)x
(3)
.K(I”) =.x0. z(rO) = z0
(5)
where now the new coefficient.p(~f).
varies slowlywhere X=(X,
. . . . . x,), x”=(_Y~,___,x,~)
and zyxwvutsrqponm
f=
However. the formulation of Tihonov’s Theorems
(f,. . - . .f,,) are vectors in n-dimensional
space,
has so far been known to the author only in
whereas
z = ( ~t..._.z,~),
z”=(z~,___,z~~),
F=
Russian f2.5.6] and reported in Polish [7,8]. As far
f
&,
.
,
F,)
are
vectors
in
s-dimensional
space.
as the literature in EngIish is concerned, even in
Putting p = 0 in eq. 4. one obtains the degezzerute
the speciahst book by Mtirray [9]_ singtdar perKvsrenr
turbation systems are treated heuristically with
only a brief mention of Tihonov’s
rigorous proof
ds/dr=j(x.=,r):
_x(J~)=_~~
(63)
[2] and a reference to Vasil’eva’s work [lo]. The
z=+(_r.r)
(6b)
latter seems to be the only source in English (apart
from the above-mentioned
summary [3]) in which
where z = +(_K, I) is a root of the system of algeTihonov’s Theorem is form&ted
and prove (and
braic equations F(x,z,t)= 0 or
some similar. more sophisticated
mathematical
F;(X.-__r)=o
(1=1.....5)
(7)
cases are discussed. everything being treated in zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
a
rather complicated
manner): even there the more
The system of equations
general theorem proven
by Tihonov
in ref. 2
(Tihonov’s
2nd Theorem) is not reported at all.
d=/dr=F(s.-_.r):
=(r”)=z”
(8)
As the steady-state approximation is w%dely used
in scientific literature comeming.
e.g. chemical
in which both x and t are taken as parameters. is
relaxation (see ref. 11) and enzyme kinetics (see
called the adjoi ned sysr enz. Of course. the point
r.zf_ 12). we think that it would be interesting to
z = Q(S, r) is an isolated singular point (root) of
restate here Tihonov’s Theorems (without proofs).
the adjoined system. as all terms on the right-hand
We will follow Tihonov’s
original paper [2]. The
sides of eq. 8 are nultified at this point.
notions used there are well known to the specialist.
We shall assume further that all functions we
But_ for the convenience of our readers, we give in
use are continuous ones and that the differential
appendix B the definitions of al1 the mathematical
equations we consider have uniquely determined
terms used.
solutions.
In ref. 2. Tihonov proved two theorems. one for
The aim of the work [2]. i.e., the investigation of
an SFO. not necessarily autonomous (see appendix zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDC
the solution of the SF0 (eq. 4) with initial condiC). containing a smail parameter in some derivations (eq. 5) when p + 0 is summarized in
tives. and another for SF0 containing sevt ral small
parameters in the derivatives_ We shall call them
Tihonov’s
1st Theorem and Tihonov’s
2nd Theorem. respectiveIy. The 1st Theorem is in reality a
When p + 0 the solution of the original system
special c&se of the more comphcated
2nd The(eq. 4). with conditions (eq. 5). tends to the soluorem+ However. because of its greater simplicity
tion of the degenerate system (eq. 6) if:
and wide applicability
\ve give here also the 1st
(1) the root z= +(s, t) is the stabIe root of the
Theorem.
Moreover.
in appendix D vve provide
adjoined system:
possibly most simple formulation of the 1st The(2) the initial values z0 lie in the domain of
orem (according to ref. 5) ‘for beginners‘.
influence of the root z = Q(X_ I) for initiai values
2. Tihonov’s
1st Theorem
Consider an SFO
~l\/J~=/(.~.~.r)
pdz/dr=
F(x.z.r)
with one small parameter
IL:
(411)
(4h)
(*X0. rO).
This asymptotic equality remains valid for alI
times r for which the solution of the adjoined
system lies inside the stability domain D of the
root z = +(s_ r ).
W Klonolvski/SimpIifyi,tg
(For the definitions of an isolated, stable root
and of the domain of influence. according
to
Tihonov [2], see appendix B; for the conditions of
existence, uniqueness and stability of solutions of
an SF0 see appendix C.)
This theorem remains valid also if the right-hand
sides of eq. 4 depend continuously on the parameter CL,i.e., for the system
dx/dt=j(x,z,r,p)
pdz/dr
(4a')
= F( x, -_.I, p)
(4b’)
75 zyxwvutsrq
reaction kinetics
d_r/dr =~(x.=~",...,z~"'.I)
#"dz"'/dr=
(=a)
F"'(x.="'.....='m'.I)
(j=l.....(m
-I))
(12b)
=(m,=*(m,(l.=(I),___.=(m--I))
(12c)
with initial conditions
.r(rO) =X0;
Z(“(Yo)=+”
(i=l....,(m-I))
where rcrn’ = #D(“*‘(_Y,z(I). . __ , z(“‘the system of algebraic equations
z(““,
t) = 0, i.e.
(13)
-l), t) is a root of
i (l ), _ _ _,
F(“‘)(r ,
where f and zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
F are continuous
functions of their
F:_““(.r,=(” ,.... Z(m). t)=O
(/,=I__.._&)
arguments 161.
(14)
As the asymptotic procedure lowers the order
The system of equations
of the SFO, the initial conditions (eq. 5) generally
may not be fulfilled
by the solutions
of the
d;(““/dr = Ftm’(_y. Z(“_..._;(nr’, I);
~(““(0) = +“”
(1s)
asymptotic (degenerate)
system. The solution of
in which x, z(I),. . . , z(“‘-“,
t are taken as paramethe original system may be approximated
by the
ters, is called the adjoi ned syst em of the fi r st or der .
solution of the degenerate system for times t >> rd.
The system is termed doub& degezzerate if it is a
where t, = ]plnp] [6]. If the asymptotic system is of
singly degenerate system for a degenerate system
the second order, one may make a complete disof the first order. The degenerate system of the
cussion of its, based e.g., on the phase-plane methfirst order has (m - 1) small parameters @’ (i =
ods [S-S].
1,...,(m1)); by putting $“‘-“=O
in this system, one obtains a degenerate system of the second order and similarly the definitions of the other
3. zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
Tihonov’s
2nd Theorem
notions of the second order. Analogously,
one
Consider
now an SF0
with several small
defines degenerate systems of the k-th order and
parameters p”‘.
all other notions of the k-th order.
The behavior of the solution of the original
d.r/dr=f(x.=(".....=(m'.t)
(93)
system (eq. 9) with initial conditions. eq. 10, when
P(‘“.,jr”‘/&
= F’/‘(.~,=“‘,_._,;‘“‘.
t);
(j= l,...,m)(9b)
all p(j) + 0 are under the condition
eq. 11, is
summarized in
and
its solution determined
by the initial conditions
_r(rO)=.G’.
$I’(rO)=.+
(_/=l,_._,“I)
Tiizorzov’s 2xd Theorenz
(10)
When p(i) -+ 0 the solution of the whole original
and f=
where x = (xz ,..., x,), _x”=(x~,___,x~)
system (eq. 9), with the initial conditions (eq. IO),
( f, _ ___. f, ) are vectors in n-dimensional zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCB
space,
L(J) = ($‘,_
__, p),
=0(J) = (g(i),_ __,
tends to the solution of the degenerate (nz-times
whereas
degenerate) system if:
=.:(I)) and F(J)= (I?:‘),. __, ks:-‘)) are vectors in s,(1) the roots z(J) = +(-“, with the aid of which
dtmensional spaces. respectively (for i = 1,. . - , zzz)the degenerate system is defined, are stable roots
Tihonov investigated the solution of eq. 9 with
of adjoined equations of the j-th order for any i
the initial conditions, eq. 10, when all /L’*(I)-+0 in
such a way that
~~J+“/p’J’+~
Putting
(11)
in eq. 9 p (“I) = 0 , one obtains
deget zer at e syst em
of the fi r st or der :
the sitzg&
(1 zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLK
GjGrn);
(2) the initial values z’(”
are in the domain of
influence of the roots z(i) = +(I’ for initial values
(X",io~'~,...,Z~~'~-l~,
The
asymptotic
to).
equalities
remain
valid
for all
times 1. for which the solution of the totally degenerate system .r( f ). 3” (t). lies inside the stability
domains of the roots Z(J) = &I) for ah;.
be more than one group of reagents for which an
Mc L holds.
As far as the reagents for which the system is
open are concerned_ suitable flux terms. describing
an exchange
with the environment,
must be in4. Chemicals zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
systems. Korzuhin’s Theorem
cluded in the kinetic equations
in addition to the
terms arising from chemical reactions_ Such terms
Tihonov’s
1st and 2nd Theorems
have a direct
are usually introduced
in the form
application
in the simplification
of kinetic equaJ,=k-(R,.,,R)
tions_ From the law of mass action (LMA)
one
(18)
obtains
generally
an SF0
with right-hand
sides
where .I, is the flux of component
R (positive when
being polynomials
of orders not greater than the
R is being supplied to the system. negative when it
second. i.e.. the reactions are uni- and bimolecular.
is flowing out of the system),
R the actual conIn some chemical
reaction
models
trimolecular
centration
inside the system. and R,,,, the actual
reactions are also aswmed
(e.g... in the ‘Brusselaconcentration
outside the system, i.e.. in the entor’ model [13]). Lvhich leads to terms of the third
vironment.
Usually.
R_,, is assumed to be indeorder.
But.
for applying
Tihonov’s
Theorems.
pendently
controllable
(a so-called
control varianeither linearity nor positiveness
of variables.
nor
ble). A special case, but one which is very often
even autonomy
of the SF0
(see appendis
B) is
utilized. is to keep R,,, constant: zyxwvutsrqponmlkjihgfedcb
required.
R <,“I = RU
First of all. it is necessary
to take into account
(‘9)
the reagents for which the system is clored - for
R, is called the reservoir concentration
Then the
such reagents (or groups of reagents) a mass conflux term (eq. IS) contributes
50 the right-hand
servation
law (MCL)
holds. Biochemical
systems
side of the kinetic equation
for R a constant term
are very often closed
for some macromolecular
equal to
enzymes
and enzymatic
comcomponents_
e-g.. zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
-Jr- = X-R,,
piexes:
(20)
and a linear
Jr-
where
actua!
E,, is the total enzyme concentration.
E the
concentration
of free enzyme.
EC the actual
concentration
sunl.naiion
of
the c--th enzyme
is taken
over
complex
a11 complexes
and
the
present
in
For simplicity
here after we denote a
reagent and its molecular
concentration
by the
s:tmc s_vmhol. ;t capital
Latin
letter.
If one
difthe
system.
ferentiates
dE,‘dr’+
From
eq.
16. one
~dE<,w=O
obtains
(17)
17 it is seen that the kinetic equations for
for which the MCL exists (e.g..
for an enzyme
and its compleses)
are linearly
dependent.
So. if the SF0
has been primarily
written down for all reagents present in the system. one equation
(for E or one of EC ) must be
dropped
and replaced
by the algebraic
equation
expressing
the MCL (eq. 16). Of course. there may
a group
eq.
=
-
term
kR
(21)
R ,,“( may also change with time. but the time
dependence
R,,“,(t)
must be known. In such a case
the SF0
is no longer an autonomous
one. but
Tihonov’s
Theorems
still
may
be applied.
If there are IV reagents
in the system
under
consideration
and there are g groups for which
MCL are fulfilled (the h-th group composed
of IV,,
reagents. h = 1.. . _. g). then the system of kinetic
equations
to be solved is as follows:
s
.v
JR,/dr’=
k, i
c
k,‘R,+
c
,_I=
1-I
of reagents
kj’R,R,
I
.v
+
c
k,““‘R,R,R,,
,.l.nr
(i=l.....(rV-_R))
= I
(nl)
\;
R,
=
R:;‘-
2
R_
PI,-1
(i=(A-~+I).___.n-)
(h = I......?)
(72b)
where zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
Jz differential
equations have been replaced
(6) in elementary
reactions all stoichiometric
by algebraic ones of the type in eq. 16. Hereinafter
coefficients are equal to unity or zero. i.e.. even the
t’ denotes real time and R, the actual concentrareaction A --+ 2B is treated as a nonelementary one
tion of the i-th component; the summation over p,,
but as a result of a chain of elementary reactions.
denotes summation over all N,, components
be+ B: so the kinetic constants in
e.g., A+B+C,C
longing to the h-th group for which an MCL is
eq. 22 do not contain factors related to stoichiofulfilled and Rg” is the total concentration of this
metric coefficients.
since one assumes that in a
group of reagents. The RPh are present also on the
CCS only elementary reactions occur.
right-hand sides of eq. 22a as their summations are
Some authors (e.g.. see ref. 5) define a CCS
taken over all N components present in the sysusing only conditions
1-4. It is interesting
to
tem.
report here the theorem proven by Korzuhin (see
As usual. each kinetic constant (li,. k{. kj’.
refs. 5 and 6):
k:““) is taken as h aving a positive or negative sign
or being equal to zero if. respectively. R, is proKorzuhin s TJzeorem
duced. consumed or is not involved at all in the
given elementary step (i.e., in reaction with R,. R,.
It is always possible to construct a closed chemetc., or the flux term)_ The terms k, arise from the
ical system of kinetic equations. in which the beinflux (eq. 20) or from the decomposition
reactions
havior of some variables will coincide with any
of the zero-th order (e.g.. one may assume that the
desired accuracy and for any desired time period
reagent is among other things a substrate for an
with the behavior of a given system
enzyme and that the enzyme is saturated with the
dR,/dr’=‘k,(R
,._._. R,v)
(i=I
_____A’)
substrate so that the reaction rate is maximal and
where 9, are polynomials
of nonnegarive integer
Goes not depend on the concentration of the reapowers.
gent). The terms k:R, arise from the outflux (eq.
21) or from decomposition
reactions of the first
For an open system the MCL (eq. 23) is not
order_
An SF0 like eq. 22a is called a closed cilemical
fulfilled. But any system may always be ‘extended
qxrenz
(CCS)
if the following
conditions
are
to a closed system’ by introducing
coupled refulfilled [5]:
servoirs of the necessary reagents. Additional
variables have to be introduced to fulfill also other
(1) all R, are positive (concentrations):
conditions imposed on the system (eq. 22) to be a
(2) the right-hand sides are of order not greater
CCS. Korzuhin’s Theorem is. in some sense, the
than two (all ki”” and all higher terms are equal to
inverse of Tihonov’s
Theorems as it concerns the
zero. i.e., at most bimolecular interactions are preproblem
of construction
of a whole (‘original’)
sent):
SF0 for the given system treated as a degenerate
(3) no autocatalytic terms are present. i.e.. all
one. The SF0 eq. 22a. with conditions l-6. repreJ.z:.k:’ and k:’ possess negative signs or are equal to
sents a homogeneous CCS which is in general the
zero:
simplest possible nonlinear SFO. The importance
(4) the system is closed. i.e.. there is neither
of Korzuhin’s
Theorem
lies in the fact that it
influx (all k, = 0) nor outflux and so the mass
demonstrates the possibility of realization of any
conservation law
given behavior (in particular, of stable auto-oscillations) of a part of the reagents during any given
time period just in a homogeneous CCS if only the
is valid, where p, denotes the molecular mass of
number of variables (reagents) is sufficiently large.
In heterogeneous (e.g., compartmental)
systems
the i-th component;
(5) no elementary reactions of decomposition
some of the conditions
1-6 are no longer valid.
of the zero-th order are present, i.e.. in eq. 22 k:
therefore the SF0 representing such systems are
and k,J’ are nonnegative for all j # i. I == i:
more compiicated. Complex behavior may be ob-
served much more frequently in heterogeneous
systems than in homogeneous ones as the number
of variables may be then considerably smaller.
Korzuhin’s
algorithm
treating any ‘nonchemical’ term in the: SF0 as a complex chemical
reaction which unfolds in a chain of elementary
reactions - makes it possible to inspect quickly
and effectively the different model variants for any
given comptex ehemicai or biochemical system. At
the same time it shows one to understand better
such extremely important
phenomena as, for example. autocatalysis. We discuss later the application of Korzuhin’s Theorem to a simple autocataIytic model and its connections
with Tihonov’s
Theorems.
where now a11 kinetic
sion [time] - I
constants
have the dimen-
!c,= k,/R)Q’
(254
tcII - zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONML
k,,R;“‘/R:”
(25b)
zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJI
(2 5 c l
IQ 9 X .,,,Rj”R$ “/R!”
As a rule, the values for RIO’ it-( eqs. 24 are such
taken that the dimensionless
variables $- are between 0 and I and of the order of unity at most_
Then the most important term in any equation of
the form of eq. 24a is the one for which the
constant (K,. nr,,, K ~){, - - _ > ha s the greatest absolute
value, yi. Dividing
the i-th equation
by yi one
obtaines from eq. 24a
5. zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
General procedure to simplify kinetic equations
Small
parameters
appear
in (bio)chemicaI
kinetic systems in a natural way if one represents
variables in a normalized dimension1es.s form. To
do this one puts in eq. 22
R, = r,.:,
(31)
where 5 denotes
the dimensionless
normalized
concentration of the i-th component and RI’) is a
constant having the dimension of concentration.
As Rj”’ one may
take. for example.
the total
concentration of a group of reagents for which an
MCL is fuffilfed (as EC3in eq. 16 and RF’ in eq.
22b). the Michnelis constant for an enzyme. the
inhibition constant for an inhibitor. Constant concentration in the environment {reservoir vaiue. as
R,, in eq. 20). initiaL concentration
Rf. or concentration
z, that the given component
would
have in the steady
state. etc. Smali parameters
appear also in a kinetic system if some elcmenta~
reaction steps are rapid as compared to others.
Taking into account eq. 23, one obtains eq. 22
in the form
s
.v
dr,/dr’=
c
Y, +
“,,‘r
/-’
x
+
/,I
“,,r”,rr
Ic,,,“,T)i,~“,
c
J&
fh - t
(27)
denotes the i-th characteristic
constants
time and where al1
~:==h.,/Y‘
(%a)
“,I’ = $0,
(ZSb)
=;,I:,I==z
5/Z/
(zk)
7,
are now Less or much Iess than unity and only the
one for which x was selected for defining the time
q (eq. 27) is exactly equal to unity.
If we are interested in the behavior
of the
system in times of a certain specified order, say of
the order of Y-, we introduce the dimensionless
time variabIe I by taking T as the unit of time:
I = r’/T
(2 9 )
Then. changing the differentiation
with respect to
t’ by differcntintion
with respect to dimensionless
d
c zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
(ret.....
(N-g))
(243)
,J.“’ - t
.v*
$=I-
7; = t/u,
time. I
- I
s
+
where
d
z?
dr
dl’”
-i-_dr
1
T
d
dr
(30)
denoting
fr=(1~“-ni-r)
. . . . . IV)
(24b)
T,/T= “‘*
(3’)
and
eliminating
from
the right-hand
sides of eqs.
26 the variables r,v_,+ ,,, -. , r,\. by using eqs. M b,
one obtains finally the system
P
r?l,-d~/dr=o,t
c
P
a,,/5-+
c
a,,,‘/‘,+
whereP=(N-g).
In short-hand
hereinafter
sive classes go the equations
iiaving smaller and
smaller parameters
WI, (i.e., the kinetic equations
for variables for which characteristic
times T are
smaller and smaller than the time unit T). into the
j-th class (j = I,. . _, m) go s, equations.
The smali
parameters
(much less than unity) we w ill denote
further by p(i) or just by pcI-So
,+~‘=,~J(r
notation
,.._.. r,:r,+ ,,__.. r&,:,4/ _*. .._. / fp)
(1, = I..... s,:j=
J?z=:,= ‘t,(r ,‘.._. rp)
where
The remaining s equations
(S = p - 11) are classified into m classes in such a way that into succes-
(33)
I.....m)
( 3’b)
where
denotes zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
dif(37C)
,*-Ito the dimensionless
time
the dotted
variable
ferentiation
with respect
variable I and ‘k, denotes the right-hand
side of the
i-th ec,uation, eqs. 32.
If the time scale T we are interested
in is such
that some nli are much greater
than unity. say
these q with indexes
i=(pi
l).(p-+2).....P.
then. by dividing the corresponding
equation
by
III;, one finds that the time derivatives
of these
variables
&==“-p
r-1
and t.~“ ’ fulfill eqs. 1 I, i.e., they are small parameters of greater and greater degree. These groups of
variables are called quick variables - after a time
of order even much smaller than that of T they
reach their (quasi)stationary
values. In the following the quick. basic and reservoir variables w ill be
denoted by z. x and p. respectively.
The separation
of reservoir
variables
is often
r ,.._ ~_rp) zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
(i =(p+I)
. .._. P)
: =-+(
(34)
made to some extent automatically,
based on the
I
fact that the components
having very great conare close to zero. since I /tlz, -=+zI. They are caIIed zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGF
centrations
as compared
to others (that is exactly
cery slow or reservoir vuriuhks, i.e., changing only
why they are called ‘reservoir
variables’)
change
with characteristic
times T, z=- zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
T. For time duramuch slower than other components.
If we conl
tion of order 7they
remain practically
unchanged
and may be replaced by their initial values
sider the simplest
bimolecular
reaction
A I- X 5
B,
it is easy to see that in any moment the instantar,(O) = ry= A, (i = (p-f- l )___..P)
are
dX/ dt
a
rates
(351
transformation
neous
- exp( - t/tX) and d A/ dr a - exp( - r/ ~.~) where
(where ( P -p) = 4) and be treated as parameters.
time
lh = I/ x-P and r, = I/ kX are instantaneous
This is equivalent
to treating
the system as an
constants
for X and A, respectively.
If one now
open one 161. In this way eqs. 33 are reduced to
assumes
that A 3r) X then r,, B I_~. i.e., in any
h~,?,=*,(r ,.-... zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
rp;d,,,
. . . . . dP)
(i =l .....
p)
(36)
instant A ci-anges much slower than X. So, if one
assumes that some variables are reservoir ones, it
The system may now be further reduced
using
is equivafent
to treat them as parameters
- they
Tihonov’s
Theorems.
The equations
are classified
may be replaced
in eqs. 22 by their constant
in such a way that the first cIass contains all. say
(initial) values (cf. eqs. 35). The differential
equarr. equations
whose relative time scale parameter
tions for these variables may be dropped from the
#II, is of the order (equal to) unity (i.e., the kinetic
very beginning
before introducing
dimensionless
equations
for variables
for which characteristic
concentrations
and time variables
and separating
times q are of the same order as T) - these are so
basic and quick variables.
In such a case the
cahed slog or basic variables
constant values of these reservoir variables. treated
?,=+,(r ,..... r,;A,,, ._... d,,) (i ==l . . . . . n)
as parameters,
are often useful to define the time
(37a)
unit 1. and/or the small parame:ers $I) or are
included directly in kinetic constants.
The system eqs. 37. when taken together with
the necessary initial conditions, is identica1 with
eqs.9 and 10 (or with eqs. 4 and 5 for nr - 1) and
Tihonov’s
Theorem
may be used to simplify it.
The conditions of continuity and of uniqueness of
solutions (see appendix C) are of course fulfilled
for the SFO. eqs. 32). and thus also for eqs. 37.
whereas the stability of a solution (see appendices
B and C) must be verified in each particular case.
However. if the system eqs. 4b or 4b’ is linear in
quick variables = (this is the case when reactions
between
quick reagents. e.g.. between
enzyme
complexes. may be neglected). then the algebraic
system rqs. 7 is a linear one and therefore it has
the unique solution: the adjoined system eqs. 8 has
then the unique stationary root. This root is stable
if there are no bifurcations, i.e.. no reactions of
decomposition
of the type 2, -+ Z, + Z, of the
quick reagents [f;]. In such a case. before applying
Tihonov’s
1st Theorem
one must verify only a
condition
concerning
the initial values. namcIy.
vvhrther they lie in the domain
of infIuence
of the
root (rq. 6b).
If we are interested
in times of order 7;- we obtain
[I41
d.r/dr = -x-+(x--p)=+.\_=
fidz/dr
= .r - X-z -.x-z
with initial conditions
s(0) = I:
Z(O) = 0
(41c)
where
r = C/T,
are dimensionless variables
defined as follows
p = T, /l-,
= E&S,,:
R,,=(k_,+k2)/‘k,:
and the constants
are
0 = k,,‘k,S,,:
k=K,,/S,
(42b)
t’ denotes real time and K,, the Michaelis constant.
If one assumes S,, 3-> E. then T,. > T,. and p < I _
In such a case the enzyme is the f;lst component
whereas the substrate is the slow one. The degenerate system is here reduced to one equation obtained from eq. 41b. The root
?=S/(k+_L)
(43)
(here and in eq_ 44 s is treated as a parameter!)
obtained from eq. 41b when p = 0 is the stable
root of the adjoined system
6. EsampIes
d-_/dr
a s - x-z - x-_
(44)
AS the first example. let us take the simplest
since for any value of s the right-hand side and SO
enzymatic reaction - the mechanism of Michaelis
the
time derivative in eq. 44 is negative for : h ? (:
and Menten: zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
wiI1 therefore decrease until the derivative will be
equal to 0. i.e._ until = = 5) and positive for = < (in such a case : vvill increase uniil : = _T). When
the system starts from the initial value c(O) = 0. =
xvith four components:
substrate S. enzyme E.
will increase until :=?,
so the initial conditions
enzyme-substrate complrs (ES) and product I’. In
(eq. 41~) he in the domain of influence of the root
this mechanism no fluxes are t&err into account
= == -_.
and so there exist two MCL:
The assumptions of Tihonov’s
1st Theorem are
(39u)
&.i ( ES) = E(0) _ E,,
fulfilled and the steady-state approximation
may
ss
P;(ES)-=S(O)=S, zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
(39b)
be used: putting = = ? in eq. 41a. a single equation
with one vrariable A- is obtained
A,S ;t result, there are only t\vo independent
difds/dr = -p-V/( k i- I)
ferential kinetic equations. The characteristic time
6453
constants for substrate and snzyme are. respecwhich is easily integrable and gives the transcentivelv.
dental algebraic equation for s(r):
$ = l/X-,&,:
r, = I/x-,s,,
(10)
.Vi x-. In s = 1- pr
(46)
81
zyxwvutsrq
Returning to the drmensional variables one obconcentration, i.e., that they are reservoir variables
tains from eq. ~5 the we11known Michaelis-Menten
and may be treated as parameters (cf. section 5).
expression zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
The kinetic equations for 2,. Z, and Z, are
dS/dr’ =
- k,SE,,/(K,,
+ S)
(47)
dZ,/dr’ =
-x-,2,
dZ,/dr’ =
k,Z,
+ x-,~lZ~
- k ,BZ ,
i- x-,/I&
(5Oa)
The steady-state approximation
fails. however, for
dZ,/dr’=
k ,BZ z - k,.-tZ,
(SOC)
very short times (of the order I < I/p)_ One may
see that the initial conditions
(eq. 41~) can no
Writing eq. 50 in terms of dimensionless variables
longer be satisfied in a consistent manner (the
- =Z ,/M”
(i =1.2.3)
-,
authors of ref. 1 speak about induction per-iod.
(51)
T,). If we do not want to use the quasi-stationary
one will observe that the characteristic time conapproximation.
the solution of the system eq. 41,
stants for Z,. Z,, and Z, are equal
is very complicated. based on coordinate transfor?-,= l/x-,: T, = T, = I/k zB
(52)
mation of the rate equations and subsequent solution of an integral equation [ 14:.
Introducing dimensionless time
It has been demonstrated
that the steady-state
I = I./T =
k,Ar’
(53)
approximation
for the Michaelis-Menten
mechaand assuming that
nism may be applied either when E, < .S, (irrespective of the values of rate constants) zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
or when
X-,‘.t/k2L3
=p;
X-,/t/X-,
=p
(54)
the rate constants of decomposition
of the en(for simplicity we assume that both quotients have
zyme-substrate complex (i.e., k_ , and k,) are much
identical values) one finds that all 112,(cf. eqs. 32)
greater than the rate constant of its formation (i.e..
are equal to p and so eqs. 50 in dimensionless form
k,) even if the concentrations
of enzyme and
becomes
substrate are of the same order. E,, = .S,: the appliof
Tihonov’s
1st Theorem
to
the
cations
dz,/dr
= - z, + z1
(553)
Michaelis-Menten
mechanism in different cases
dz,/dr = z, - z1 C zz
(55b)
are discussed in refs. 5 and 6.
d=,/dr = zz - z3
(55c)
As the second example. let us consider the
simplest autocatalytic
process. describing an exOne may demonstrate
that when p -+ 0. the
ponential growth
solution z3( r) of the system. eqs. 55, coincides with
the solution X(I) of eq. 48 (cf. ref. 5). By changing
ds/dr = .r
(43)
the variable
- It is not a closed chemical system in the sense of
s = pz3
(56)
the definition
given in section 4. But. following
Korzuhin’s Theorem_ it is possibie to find a closed
one obtains from eqs. 5.5
chemical system. the behavior of which will coin(573)
d.r/dr = - .r + zz
cide with any desired accuracy with the behavior
pdz,/‘ dr=s-z,
(57b)
of eq. 48. One may show that this is the case. for
example. for the following system:
pdz,/dr
= .x f -_, - z1
(57c)
The
&I
B+Z,+Z,iC
adjoined
system is in this case
dz,/dT=x-z=,
zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
(49)
dz,/dr
= .r f z, - zz
*3
‘Zz
A+-B+Ct Z ,+Z 2t Z 3=11fO=con~i r l n~
where x is no longer a function of
parameter_ Putting p = 0, one obtains
the adjoined system
We assume that A and B are present at very high
:, =x;
z,
zz = 2x
(50b)
(58a)
(58b)
time but a
the root of
(59)
As eqs. 58 are a linear SFO, it is easy to integrate.
The result is:
the adjoined system, similar considerations
must
be made for all roots separately, as the coefficients
of the characteristic equation and consequently the
=,(z)=i,+(-_p--,)exp(--)
(boa)
Routh-Hurwitz
determinants depend on the value
=1(‘)=I,i[(.-:--E,)s~(=4-~z)]exp(--)
(60b)
of the root (cf. eqs. C7-ClO).
From the above considerations.
one may see
One sees immediately from eqs. 60 that the root.
that to sustain an autocatalytic process ‘inexhausrqs. 59, is stable: =I and z2 quickly tend to the
tible’ reservoirs of ‘building’ materials (A and B)
steady-state values 5, and ZZ, irrespective of their
for the reagents 2, and Z2 must exist. In reality
initial values z:’ and z:. i.e., all possible initial
the system, eqs. 49, modeIs an autocatalytic system
values zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
lie in the domain of influence of the root.
only for the times for which A and B are practiThe assumptions of Tihonov’s
1st Theorem are
caIly constant (of the order I < l/p). After a suffithus fulfilled and. when p - 0, solutions of the
cientIy long time any reservoir wiI1 in reaIity be
system_ eq. 57. tend to the solution of the degenerexhausted. One ought also to note that the autoate system
catalytic process gives a byproduct C.
dz/dr = -.r -cZ2
(613)
As the third example, let us consider a case
when small parameters arise in the system because
Z, = s
(6’b)
of the differences in rate constants (cf.. the remark
Z2= 2.r
(61c)
concerning
the applicability
of the steady-state
and so s (and also z;) behaves as if it were
approximation
to the Michaelis-Menten
mechaproduced in the autocatalytic process (eq. 48).
nism in this section)_ This will also demonstrate
To
demonstrate
that the assumptions
of
another method of separation of quick and basic
Tihonov’s Theorem are fulfilled, it is not necessary
(slow) variables. Let us assume that the reaction
to integrate
the adjoined
system.
In
more
L
R,tR,dR,
complicated cases this may be practically impossi(66)
ble. Instead. one may use the Routh-Hurwitz critetakes place in the system and that this reaction is a
rion for this purpose (cf. appendix
C). In the
quick one, i.e.
above example. the adjoined system (eqs. 58) has
the uniqtie root (the uniqueness of a root may be
k = l/p where
p-z 1
(67)
in general checked with the aid of LiFschitz condiThe reagents R,. R, and R, may participate also
tions. cf. appendix C). The characteristic equation
in slow reactions
(Le., with rate constants
much
(eq. C8) in this case is
&I
I
-1-x
,
0
-,-A
less than Jc). Assume
I
=0
(Q)
i.e.
that in appropriately
variables
the kinetics
3, = .q,( r. s )
XZ+2X+1=0
/>*,==I: h,=Z:
and further
defined
of the system
(683)
(63)
So xve have (eq. C9)
Q,=l:
dimensionless
are given by
h2=1
(b-1)
i,=
-~r,r2~G,(r_x)
F2=
-_S’,r+G~(r_.x)
(68b)
zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
I
D,=Ir
D1= zo
I
i.e.. they are all positive.
I, =I!
I
(65)
The Routh-Hunvitz
criterion shows that the root is stable and. as it is also
unique. any initial condi:ions
must lie in its domain of influence. If there is more than one root of
wherer=(r,.
‘;. ~-_~)and~=(s,_..._x,,):
theterms
G, and g, denote contributions of slow reactions_ I
denotes slow variables.
The SF0 in which the terms ?f the order less
than 1.1~ (e.g.. the contributions or slow reactions)
W. Kt ono~sk i /Si nt pIl i /yi l l g r eact i on k i net i cs
83
‘diagrams’ used by Hill [ 181 and other authors are
also labelled SFG and therefore the general theory
developed
for SFG may be applied to these diagrams [20].
Tihonov’s Theorems may be applied to systems
in which reactions of the second (and even higher)
order take place. If some rate constants are equal
to zero, i.e., some reaction steps are irreversible.
the fundamental
assumption
of ref. 1 is not
fulfilled. In contrast, Tihonov’s Theorems may be
(69x)
X,+1 = r, + r,
even easier to apply in such cases (because of the
x,+2 = r, - ‘2
(@b)
smaller number of terms) than in the case when all
reactions are reversible and no general assumpBy introducing
these integrals into eq. 63 and
tions about the number of components
in the
denoting r, by I, one obtains
system are necessary.
_?,Tg,(=._r)
(i =l .....n)
The fast components ‘forget’ their initial values.
These values are necessary only to check if the
(70a)
-% + I =G,(-_._x)+G,(~._x)
assumptions of Tihonov’s
Theorems are fulfilled
Kn+2 = G,(z,x)-G2(z_x)
(Le., to check if they are lying in the domain of
p?=
-z(z-X”LZ ;f~G,(z.X)
(‘Ob)
influence of the root). However,
sums of initial
where now X=(X ,,.-.. x,,, x,~, ,, x,,+z). The sysconcentrations of some groups of fast components
tem obtained is identical with eqs. 4’ and Tihonov’s
(e.g., the totai initial enzyme concentration
E,,)
1st Theorem may be applied if the assumptions
may determine some characteristic constants of
are fulfilled_ The above example illustrates how
the system, e.g. small parameters p(j)_
The obvious conclusion from Tihonov’s
Theothe differences
in rate constants of elementary
rems is that the steady-state approximation
works
reactions may be exploited
to simplify reaction
kinetics by using Tihonov’s Theorem.
well for times (much) greater than the characteristic time constant used as the unit of time (i.e., the
time constant of the components
which may be
7. zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
Concluding
remarks
thought to be slow in the time scale we are interThe method proposed by Kijima and Kijima [I]
ested in), however, not so great that the reagents
m_ay not be applied directly (without linearlizaassumed to be reservoir variables are exhausted_
tion) even to a Michaelis-Menten
mechanism (eq.
The fact that the characteristic time constant
41) because of the nonlinear term XI. Their method
for a given reagent is small, compared to the time
scale we are interested in, constitutes a criterion
may be classified between graph-theoretical
methfor fast equilibration
which is even more imods, however, the word ‘graph’ does not appear in
the article. Because of introducing two kinds of
portant in practice than the fact that some elemenarrows (edges or branches in graph-theoretical
tary reactions are relatively quick (i.e., some reacterminology)
their graphs correspond to so-called
tion rate constants are much greater than others).
colored graphs (two ‘colors’ of edges) and because
Only relations between characteristic time scales 7
each edge has given direction and ‘value (the rate
of different reagents and their relations to the time
constant) they belong to the class of graphs called
scale of observation we are interested in (which is
taken as the normalizing
unit of time T) are of
labeiled si gnal flow gr aphs (SFG)
or labelled directed graphs [15]. The theory of graphs, deimportance.
in section 5 we have subdivided all
veloped
primarily
for the analysis of electrical
the reagents into at least three classes, called renetworks, has been extensively used for simplifyservoir (very slow), y. basic (slow), x, and quick, z,
ing kinetic (bio)chemical
systems (see, for exam&=rL@&(r,_s.=. t,p)
(k + I....,q)
ple, refs. 16-19). It is interesting to note that the
(713)
are neglected, is called a t r uncat ed syst em. Usually.
the truncated system has certain linear integrals,
some of which represent conservation laws. These
integrals are slow variables for the original SF0
and taking them as new variables leads to separation of quick and slow variables_
It is easy to see that in the example under
consideration there exist two such lineary independent integrals of the truncated system
zyxwvutsr
_i,=/,(s._~.z.r.p) zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
(i =l ..__
.n)
r(r) of the adjoined system dz/d?- = F(_Y’, z. to),
(71b)
~,=+_.Lz.r _P)
(I=l .....s)
z(0) = z@, satisfies lim,,,
(‘1c)
The concentration hierarchy in the system is usualiy in agreement with the time hierarchy 161.i.e.
.r; < Y, < z,
I( T) = 9(x”,
to). Then
zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFE
the solution
of (1) through
the initial point
(72)
(x0. z”, r”) tends. as p + 0. to the solution of the
degenerate system dx/dr =f(x.
Q(x. f). f) as long
as the point (x. r) does not leave the region 3. A
similar result holds for systems containing several
small parameters p,_ ____p,,, tending to 0 in such a
way that p, c ,/p, + 0. J. L. Mas~eru (Montevideo)_
we have demonstrated.
if we are interested in
the behavior of the system in periods of the order
of the time scale characteristic for s,. then we put
Appendix
B
_YL= 0 (cf. eqs. 34 and 35) and we eliminate the
variables 2, using Tihonov’s Theorem. However_ if
B. I. I?fahetnatical
dejittitions
(from ref- 2)
we are interested in the transient phenomena (i.e..
in time interval 0 -Z t CCp), then the variables z,
The ttorttz of an s-dimensional
vector
o=
ought to be considered as basic ones and s, as
(F,..__. v~) will be denoted by ]o] and defined as
reservoir ones. Oppositely. if we are interested in
the evolution of the system (I 3-> 1). then the varifL‘I=
,+
(Et)
ablesr;, ought t<>be treated as basic ones and X, as
/-I
quick ones. ReaI systems geurrally have more than
three characteristic time scales (i.e.. some groups
(Euclidian norm)_
of quick variables. z:” (i = 1.. ___m) and then
The root z = +(x, I) of the adjoined system of
Tihonov’s
2nd Theorem may be a useful zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCB
matheequations F(x, z, r) = 0 will be called the i sohr ed
As
matical
tool.
it may be said that the mathematical foundations of the steady-state approximation have been well established four more than
30 years.
In conclusion.
Appendix
A
if there exists an E for which this system may
not be fuIfiIled by any other vector z’ = (z;_ ___ . z-z)
with the property
roof
I-_‘-+(_\‘_r)l’<
(=‘-+I)
The isolated singular point (root) .Z= (Z,_ __ . , f,)
of the acljoitted system
d:/ds=F(s.z.r);
Tihonov.
A.N. Systems of differential
equations
containing small parameters in the derivatives. Mat. Sbornik N.S. 31 (73). 575-586
( 1952). (Russian)
Consider
a system (1) d.ridr =/(_I-. z. I).
pd z/d1 = F(x. z. I). where s. f are pr-vectors. z. F
i)r-vectors. and f. F satisfy suitable regularity assumptions. Assume -_ = +(s. I ) is an isolated solution of F(_(s. z. I) = 0 vvhose points are (asymptotically) stable equilibrium
points of the adjoined
system d z/d7 = F(s. z. t) (s. t being considered
here as parameters. the point (s. P) belonging to a
hounded open region D). Assume that the inittal
conditions ( _I-“. z”. t”) are such that the solution
(W
=(r”)=z”
(B3)
(where x and r are taken as parameters) is called a
srohilir_s point if for any E there exists a 6(e) such
that trajectories starting in a point M belonging to
the 6(c) neighborhood
of the singular point tend
to this singular point without leaving the l neighborhood when r -+ so. i.e.
(1) the trajectory of any point ?=(z,__.__z,)
beronging to the 6(e) neighborhood
of the point 5
tends to 5 when r -* co:
Iim
7-x
Z(T)=:
if
I=(O)-?lc6(~)
(B3)
(2) trajectories of points from the 8(e) neighborhood of the point Z do not leave the e neighborhood of this point:
f:(r)--_f<r
foran~r.ifI=(O)-_l<S(r)
(=I
The isolated root z = Q(x, t) of the system
tinuous and if the Li pscJzi t z condi t i ons
F(x, Z, t) = 0 is called zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
st abl e
in some bounded
region D of the space (x, t), if for all points
IF(c=, , . . . . =,)-F(T ;
t ,..._, l J,)I -= nr . 2 I=, - o,I
(C3)
,--I
belonging to 5 the points z = Q(_K, t) are singular
stability points of the system (eq. B3).
are fulfilled for some A4 independent of z and o.
The domai n of influence of the stable root z =
Then z(r)
is a continuous function of the given
+(x, f) in the subspace x = constant, I = constant
value z0 = ~(7~). Each solution extends to the
is defined as a set of all points z” from which
boundary of D.
trajectories of the adjoined system (eq. B3) tend to
The Lipschitz conditions are satisfied, in partic= = +(x, I) when 7 --j cc.
ular. whenever F(z, T ) has bounded and continuThe root z(“‘)=
#“‘)(x.
z(‘),....P(“~-‘),
2) of the
ous derivatives, aF/ar {,
(I = 1, ___, s) in D.
adjoined system of the first order F(“‘)(_Y_
z(l ). _ _ _.
The root of an SF0
is a szabl e root if all
z”“‘, I) = 0 (where x, z(‘) ..__,_ -(“‘-‘).
t are treated
perturbations 6~ around this root diminish to zero
as parameters) is called an i sol at ed r oof of the first
when 7 * cc
or der .
if there exists such E that this system may
not be fulfilled by any other vector z’(“‘) for which
s=0
(C4)
5-x
,-_“““_.*o’“Z’, <c (=?Wi,,+!nl,)
(B6)
The root of the first order
L(“‘) = #“‘)(x.
-..(I) ._._,_ -(m-‘). t) is called a st abl e r oar in some
closed bounded region 5, if for all (s. I(‘)._ __,
:(“I-I),
f) of this region z(“‘)= r$(“‘) is a stable
singular point of the adjointed system.
To investigate if a root 5 of the autonomous SF0
(eq. C2) is stable, one linearizes the system around
Z- ; by putting
=,==,+s-_,
(I=1 . . . . . s)
(C5)
one obtains
Appendis
Cl.
C
Or her
d(&,)/dT=
wat henzat i cul
c a,,:, (I=l.....s)
1-I
defi ni t i ons
where the matrix A = l l a,,l j is given
An ordinary differential
equation of the first
order or a system of first-order ordinary differential equation (SFO)
dz/dr=F(z.i)
(~=~,.___.~,)
(Cl)
is called ar c;ononzous if the right-hand sides do not
depend explicitly on the independent variable r,
i.e.
” ,, =aF,/a:,I,_:
(l.j=
(CI)
where F and z are vectors of s-dimensional
(s=
1,2,...).
and
uni queness
of sol ul i ot zs
space
I . . . . . .r)
by
(C7)
The root z = 5 is stable if all eigenvalues h,, (I =
1,..., s). of the matrix A have negative real parts.
1, are the zeros of the characteristic equation
della,, - h&J= 0
when S,!= 0 for r-i,
d-_/di=F(=)
C_7_ Exi sr ence
(C6)
(cs)
6,‘=
1 for i=i
(Kronecker’s
delta).
The characteristic equation, after reso!ving the
determinant in eq. C8, is a polynomial of the s-th
order:
of SF0
b,~~+b,h’-‘+...+b_,_,h+h,=O
The system (eqs. Cl )
has a unique solution
z = ~(7) through every point (Z = z,,, 7 = ~“0)in the
(S
+
I)-dimensional
domain D if for (z, T) E D
functions F( =, T ) are single-valued, bounded. con-
cc91
zyxwvutsrq
Given
b, > 0, all roots X, of the characteristic
zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCB
polynomial
(all eigenvalues of the system eqs. C2
have negative real parts if and only if the determi-
are no others roots);
(b) the solution t,,. _. , Zs is a stable isolated
4, = ho zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
singular point of the adjoined system (i.e., any
D,=h,
perturbation throwing the system out of this point,
h, &I
diminishes with time to zero) for all values of
Dz =
I 4 h2 I
(x ,, _ . . , x,) which are treated here just as parameters and not as functions of time;
bt kl 0
(c) the initial values (zp,. . _, z,“) are in the doD, = 6,
b,
cc101 zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHG
6,
main of influence of the stable singular point of
6,
b.,
b,
the adjoined system (i.e., the system will evolve in
_
_
_
_
_
.
_
such a way that (z,, . . _, z,), when started from
_
_
.
(ZPI--.,iI O), will tend to (F,, . . . , z;);
0’
...
0
(d) the solutions of the whole system (eqs. Dl)
ho
b,
and of the adjoined system (eq. Dl b) are unique
hz
b,
_‘. (where b,,, = 0 if nz > s)
D,=
h! ________________
:_‘..
____.____..._____.......
and the right-hand sides (f;- and F,) are continuous
6,,_,
6,,_,
bZ<--3
...
b,
functions.
nanis
I I
all zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
positive (Routh-Hunvitz
criterion). This is
true only if all 6, and either all even-numbered zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFED
D,
Note added in proof (Received 20th June 1983)
or all odd-numbered
D, are positive
(LiCnardChipart test; [21]. zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
Professor Benno IHess has recently called the
author’s attention to the publication of Reich and
Sel’kov [22], in which the authors outline the simAppendis
D
plest case of using Tihonov’s
1st Theorem (however, without formulation
of the theorem itself)
Di. Tihonov’s Theorem for Begitmers
with one rapid and one slow variable.
are
Consider
a system of p first-order ordinary
differential equations. Let us assume that s of the
p equations have a small parameter p. multiplying
the time derivatives:
d.r,/dr=f,(.r
pdz,/dr
,.___._ r_.-_ ,.___. 2.)
= F,(x
,.___._ r,.
-,.._..
Z, )
(;=I
(I=
____. II)
(Dla)
I.....s)
(Dlb)
where (n +s)=p.
Eq. Dla is called the clegetlet-ate sy.ster?z. eq.
D 1b the adjoined system.
02.
Tihonov ‘s Theorem
The solution of the whole (original) system (eqs.
D 1) tends to the solution of the degenerate system
when p + 0 if the following conditions are fulfilled:
(a) the solution
-7, = +,(_x,, _ _ _ , x,,. =,. - _ . _
=,).___._,=~~(s,.___._Y,,.T,,___.=_~)
is an isolated
root of the algebraic system
F,(X ,..._._r,.-_ I.‘.‘. =,)=cl
(I=1 . . . . . s)
(D2)
Acknowledgements
The author is greatly indebted
to the MaxPlanck-Gesellschaft
zur Fiirderung
der Wissenschaften e-V. for the scholarship for staying in the
Max-Planck-Institut
fiir Biophysikalische
Chemie
in Giittingen and would like to express his gratitude to Professors L. De Mayer, M. Eigen and H.
Strehlow for the kind hospitality he received during his stay.
References
I H. Kijima and S. Kijima, Biophys.Chem. 16 (1982)
181.
2 A .N. Tihonov.
M at. Sb. N.S. 31 (73) (1952)
575 fin Russian).
3 J.L. M assera. M ath. Rev. 14 (1953)
1085.
4 Y.A. M itropolskir
Problemes de la theorie
oscillations
5 Yu.M .
non-stationnaires
Romanowski,N.V.
ski, M athematical
(i.e.. in the small neighborhood
of this root there
1975)
(in Russian).
modetling
asymptotique
(Gautiers-V illars,
St’epnanova
Paris.
des
1966).
and D.S. Tchemav-
in biophysics
(Nauka.
M oscow.
6
7
8
9
10
11
12
13
A.M. Zhabotinsky,
Koncentracionnye
kolebania
(Nauka,
Moscow, 1974) (in Russian).
W. Klonowski:
Thermodynamics
of dissipative structures
in
enzymatic
systems as an example of process modelling
in
biophysics.
IFTR Reports l/1978,
Warsaw (in Polish).
W. Klonowski,
Probl. Contemp.
Biophys..
5 (1980) 199.
(PWN. Warsaw) (in Poli -5).
J.D. Mur r ay:
Nonlinear-differential-equation
models in biology (Clarendon
Press. Oxford, 1977).
A.B. Vasil’eva. Russ. Math. Sure. 18 (1963) 13.
M. Eigen and L. de Maeyer
in: Techniques
of organic
chemistry.
vol. 8. part. II. 2nd edn.. eds. S.L. Friess. ES.
Lewis and A. Weissberger
(Wiley Interscience.
New York,
1963) p_ 895.
Ch. Waiter:
Steady-state
applications
in enzyme kinetics
(Ronald Press. New York. 1965).
G. Nicolis and 1. Prigogine:
Self-organisation
in non-equi-
librium systems (Wiley Interscience,
New York, 1977).
14 P.L. Philipson, Biophys. Chem. 16 (1982) 173.
15 L. Robichaud,
M. Buaver aqd J. robert, Signal flow graphs
(Prentice-Hall,
New York. I9c-2).
16 M.V. Vol’kenshtein
and B.N. G-ol’dshtein,
B&him.
Biophys. Acta 115 (1966) 471.
17 M.V. Vol’kenshrein.
Enzyme physics (Plenum
Press, New
York, 1969).
18 T.L. Hill, Free energy transduction
in biology (Academic
Press, New York. 1977).
19 J.S. Shiner, J. Theor. Biol. 99 (1982) 309.
20 W. Klonowski
and Z. Osiak. unpublished
results.
21 G.A. Kom and T.M. Korn: Mathematical
handbook
for
Scientists and Engineers (McGraw-Hill.
New York, 1968).
22 J.G. Reich and E.E. Sel’kov. Energy metabolism
of the cell
(Academic
Press. New York. 1981).