phys. stat. sol. (c) 2, No. 5, 1592 – 1605 (2005) / DOI 10.1002/pssc.200460840
Neutron detection with cryogenics and semiconductors
Zane W. Bell*1, D. A. Carpenter2, S. S. Cristy2, V. E. Lamberti2, Arnold Burger3,
Brian F. Woodfield4, Thomas Niedermayr5, I. Dragos Hau5, Simon E. Labov5,
Stephan Friedrich5, W. Geoffrey West6, Kenneth R. Pohl7, and Lodewijk van den Berg7
1
2
3
4
5
6
7
Oak Ridge National Laboratory, 1 Bethel Valley Rd., Oak Ridge, TN 37831-6010, USA
Y-12 National Security Complex, Bear Creek Rd., Oak Ridge, TN 37831-8084, USA
Fisk University, 1000 Seventeenth Ave. North, Nashville, TN 37208-3051, USA
Brigham Young University, Department of Chemistry and Biochemistry, Provo, UT 84602, USA
Lawrence Livermore National Laboratory, 7000 East Ave. L-270, Livermore, CA 94550, USA
University of Michigan, 2355 Bonisteel Blvd., Ann Arbor, MI 48109-2104, USA
Constellation Technology Corporation, 7887 Bryan Dairy Rd. Suite 100, Largo, FL 33777-1452, USA
Received 8 September 2004, accepted 10 February 2005
Published online 10 March 2005
PACS 29.30.Hs, 29.40.Vj, 29.40.Wk
The common methods of neutron detection are reviewed with special attention paid to the application of
cryogenics and semiconductors to the problem. The authors’ work with LiF- and boron-based cryogenic
instruments is described as well as the use of CdTe and HgI2 for direct detection of neutrons.
© 2005 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
1 Introduction Advances in materials and methods have enabled the detection of radiation by
means today that would have seemed, to pioneers in the field a century ago, like science fiction. Improvements in technology have resulted, for gamma ray detection, in high-purity germanium operating at
77 K and providing 0.1% energy resolution above 1 MeV, more than an order of magnitude improvement over what was (and still is) achievable by scintillators. However, operating below 1 K, cryogenic
calorimeters have been used in X-ray astronomy, in the search for dark matter, and more recently in
gamma ray spectroscopy, and have achieved better than 70 eV resolution at 60 keV [1], a factor of 4 to 5
improvement over what can be achieved by germanium at that energy. Meanwhile, at the other end of the
temperature spectrum, the development of new, wide band-gap semiconductors has sparked research in
room temperature gamma ray detectors and has held out the hope of 1 – 2% resolution and freedom from
cryogenics [2, 3].
With such results being reported from the X- and gamma ray world it is natural to examine the possibilities for neutron detection. A cryogenic neutron detector would operate by detecting the heat pulses
caused by neutron capture and scattering, while a semiconducting detector would detect the nuclear reaction products from a sensitizer (for example, fission fragments detected in a 235U-coated Si diode) or from
some constituent of the semiconductor.
In the following sections, the common methods of neutron detection are described and their deficiencies with respect to neutron spectroscopy at energies above thermal (0.025 eV) are outlined. Published
work on neutron-detecting cryogenic calorimeters will be reviewed and work by the present authors on
boron- and lithium-based instruments will be discussed. Turning to semiconductors, we review work
with coated and native (uncoated) semiconductor, including Cd1–xZnxTe (CZT) and HgI2, as applied to
neutron detection. Results obtained by the authors with HgI2 will be shown.
*
Corresponding author: e-mail: bellzw@ornl.gov, Phone: +1 865 574 6120, Fax: +1 865 576 8380
© 2005 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
phys. stat. sol. (c) 2, No. 5 (2005) / www.pss-c.com
1593
2 Review of neutron detection techniques Since neutrons are uncharged, they are always detected
via their nuclear interactions, examples of which are (n, f), (n, γ), (n, α), and (n, n’). In all cases it is the
interaction of energetic charged reaction products and/or electromagnetic radiation with a detecting medium that produces the sensible signal, typically, but not exclusively, some combination of electric
charge or scintillation light.
Not all reaction mechanisms lend themselves to spectroscopy. That is, they are useful as the basis of a
detector, but cannot easily be used to determine the energy of the incident neutrons. Reactions that produce fission or the emission of energetic gamma rays result in incomplete registration of the energy of
the incident neutron because reaction products escape the detector. For example, fission is accompanied
by the emission of 6 – 8 MeV in prompt and delayed fast neutrons and a similar amount of energy as
gamma rays, implying that a fission detector always underestimates the total energy released in the reaction. In addition, fission detectors are not sensitive to the variation of fission Q-value with the atomic
numbers and masses of the fission product pair.
2.1 Devices based on charged particles Neutron reactions that result in charged particles include
fission, recoils (protons and heavier particles), and conversion electrons. The neutron-sensitive layer may
be realized as a converter in close proximity to a charged particle detector, or the detecting medium itself
(as in the case of 3He). Included in this category of device are scintillators which convert the energy of
the charged particle to light. A discussion of detectors that rely on capture leading to alpha or beta activity is deferred to the following section.
An example of a coated detector is given by Litovchenko et al. [4]. In this instance, a 0.15 mm thick
film of uranium oxide, enriched to 99.92% 235U, was deposited on an aluminium planchet, and placed in
contact with a silicon surface-barrier detector. Both alpha particles from the natural decay of uranium
and neutron-induced fission fragments were detected by the 100 mm2 silicon diode.
The thermal neutron fission cross section of 235U is 585 barns [5], making their mean free path in a
layer of oxide (not specified by Litovchenko, but perhaps U3O8) approximately 1.1 mm in fully dense
material. Consequently, thermal neutrons are absorbed almost uniformly throughout the layer, although
only fission fragments born within 10 micrometers of the diode have a chance of reaching it. These fragments, traversing a continuum of paths, reach the diode with a continuum of energies, and create charge
within the silicon in proportion to their energy. The electric field within the diode structure collects this
charge and external electronics converts it to a voltage pulse whose amplitude is indicative of the fragment’s energy. Because of conservation of momentum, only one of the two fission fragments from each
fission can enter the diode.
Uranium is not the sole converter material that has been used and silicon is not the only charged particle detector that has been demonstrated. Filho et al. [6] describe a silicon diode covered with boron or
polyethylene. In the case of boron, alpha particles and lithium ions are produced upon absorption of
thermal neutrons, while the polyethylene generates protons from (n, p) scattering. In a conceptually
similar arrangement, McGregor et al. replaces the silicon diode with GaAs [7], and adds texture to the
surface of a GaAs device to improve the efficiency of detection of ions from the 10B(n, α) reaction [8].
Aoyama et al. [9] report a device based on a Gd foil viewed by a silicon diode. The diode was not
sensitive to capture gamma rays or resulting X-rays, but, because of the 25 micrometer thickness of the
foil, could not distinguish between conversion electrons from the foil (a consequence of the de-excitation
of 158Gd proceeding via low-energy E2 transitions) and photo- or Compton electrons. Consequently,
Aoyama added a tin foil viewed by a second diode by which to estimate the photon response of the Gd
foil.
Coated detectors suffer from the necessity for sensitivity to be balanced against the range (<5 micrometers for the heavy ions, up to 1 mm for energetic protons) of the reaction products, and ultimately
the range of the particles limits the maximum sensitivity. These detectors are necessarily inefficient because the thickness of the converter layer must be sufficiently thin that particles can exit it with sufficient
energy to be registered. For a fission-fragment device, typical thickness of the uranium layer is only a
© 2005 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
1594
Z. W. Bell et al.: Neutron detection with cryogenics and semiconductors
few micrometers (metal equivalent), making the devices approximately 0.3% efficient. Litovchenko’s
detector was quoted at 0.055%, implying an oxide layer containing less uranium than 1 micrometer of
pure metal. However, McGregor reports detection efficiency near 4% with a textured boron-coated
diode, and predicts efficiencies approaching 10% to be possible.
If the neutron-sensitive material is part of the detecting medium itself, issues of range of reaction
products disappear since they are created and completely absorbed in the detecting medium. Such is the
case for 3He- and 10BF3-filled proportional counters. In these devices, the charged particles are generated
in the gas as a result of the capture of neutrons, and the ionization created along their paths is collected.
Since the energy shared by the charged particles is the sum of the Q-value of the capture reaction and the
energy of the incident neutron, measurement of the charge is indicative of the energy of the incident
neutron. Similarly, boron, lithium, or gadolinium may be incorporated in scintillators in which the energetic charged particles generate light in proportion to the energy deposited by the particles.
Knitel et al. [10] report results for a number inorganic scintillators containing various combinations of
Li, B, and Gd. In most cases, the light yields are disappointingly low (< 104 scintillation photons/MeV,
electron equivalent) making them unattractive as detector materials. However, cerium activated YBO3,
GdBO3, and Li6Gd(BO3)3 are reported to be reasonably bright and emit at 390 and 415 nm.
Organic scintillators consist (generally) of a hydrocarbon plastic, such as polystyrene or polyvinyltoluene, or liquid, such as benzene, xylene, or toluene, into which are dissolved one or more phosphors.
Silicone rubber [11, 12] has also been demonstrated to be a suitable matrix. Since the phosphors are
themselves molecules containing phenyl rings (such as diphenyl oxazole and anthracene), it is necessary
for the hydrocarbon matrix also to contain rings so that the energy deposited by ionization can hop from
ring to ring until it finally transfers to a phosphor molecule. Typically, the excited singlet to ground state
singlet transition has a time constant of a few nanoseconds and the de-excitation results in light. The
process is described by Birks [13], Förster [14], and Scholes [15].
Since the number of scintillation photons is approximately proportional to the amount of energy deposited in the scintillator, and the range of even MeV protons is only a millimeter, the energy of the
recoiling proton can be recovered from a measurement of the amount of light from the event. However,
(n, p) scattering is isotropic in the center of momentum frame, and, for a neutron with energy En, this
leads to a uniform distribution of recoil proton energies from 0 (grazing collision) to En (head-on collision). Consequently, it is necessary to unfold a response function from the measured spectrum to obtain
the spectrum of incident neutrons.
The unfolding process is a stripping procedure, and it accumulates errors in the low-energy portion of
the spectrum. The problem is compounded by the fact that recoil protons produce less light (by a factor
of 4 – 8, depending on the matrix, phosphors, and energy) than do electrons. Since the noise floor of an
uncooled photomultiplier tube is often near 10 keV (electron equivalent), neutrons with energy much
below 200 keV do not produce a pulse-height spectrum that can be analyzed with any certainty.
Recently, Robertson et al. [16], have reported the fabrication of a diode from 270 nm boron carbide
layer deposited on n-type Si. They exposed the device to a nuclear reactor, acquired pulse-height spectra,
and ascribed features in the spectra to the reaction products of the 10B(n, α)7Li reaction. In subsequent
work [17], they report the fabrication of an all-boron-carbide device made by depositing n-type boron
carbide onto a p-type boron carbide layer previously deposited on an aluminium substrate, and, on exposure to a 120 n/cm2/s thermal neutron flux, they observed a current of approximately 18 fA. Unfortunately, the authors give no indication that this current is consistent with the incident flux, nor any explanation of how they observe such small currents in the presence of reported DC bias currents several orders of magnitude larger. Further compounding the issue is the failure of the authors to perform simple
experiments to separate the neutron and gamma responses of the devices. It is also possible that some of
the observed spectral features are a consequence of fast neutron interactions. This is not to say the devices do not perform as described, but rather that there are some questions regarding the interpretation of
the data.
© 2005 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
phys. stat. sol. (c) 2, No. 5 (2005) / www.pss-c.com
1595
2.2 Devices based on capture gamma rays and induced radioactivity Neutron capture usually
results in the formation of a nucleus in a highly excited state. This state may decay by the emission of
charged particles as described above, or, as is more likely, by the emission of a cascade of gamma rays,
conceptually similar to the cascade of X-rays emitted when an ion captures an electron. In addition, it is
possible for the transmuted nucleus to be unstable and subsequently decay by alpha, beta, or positron
emission, or electron capture followed by the emission of more gamma rays and/or X-rays. Examples of
these reactions are 113Cd(n, γ)114Cd (stable product) and 197Au(n, γ)198Au (unstable product) The deexcitation can also produce conversion electrons, as in the case of 158Gd, described above.
The similarities between devices that detect either capture gamma rays or gamma rays resulting from
induced radioactivity from the product are such that it is sufficient to consider them together. In both
cases, a gamma detector is placed in proximity to the neutron sensitive material and the resulting pulseheight spectrum is analyzed for evidence of the signature gamma ray or rays.
Bell [18] describes the use of the 478 keV gamma ray that is emitted in the 10B(n, α)7Li* reaction. The
use of boron as a converter generating charged particles has been described above, however if the converter is placed close to an efficient gamma ray detector, such as CdWO4, the gamma ray can be used as
a signature of the presence of thermal neutrons. Although this is a significantly less efficient method than
direct detection of charged particles, the limitations imposed by the fabrication considerations of semiconductor detectors are removed by the use of scintillator that is readily fabricated in large volume and
surface area. Consequently, in a flux-starved environment, higher count rates may be achieved at lower
total cost because of the larger sensitive volume of a scintillator. Other materials with high neutron capture cross section that produce an analyzable spectrum are Cd and Hg, both of which will be discussed
below.
As the name implies, activation foil analysis relies on the detection of induced radioactivity in a detecting medium. Generally, this technique is not useful for single event counting, but rather is useful only
to determine the presence of a flux. The method is hindered by the time needed to activate the foil (long
half-life foil implies long time to reach saturation), the loss of activity resulting from delays in reaching
the counting apparatus, and from self-absorption effects when the induced radioactivity manifests itself
as alpha particles, low-energy beta particles, or X-rays.
Johnson [19] reported the detection of induced radioactivity in a CdS crystal serving as both neutron
absorber and detector. Subsequent to irradiation by 14 MeV neutrons (1011-1017 n/cm2), the decays of
115
Cd and 32P were observed. CdS is inappropriate for thermal neutron activation detection since the sulphur cross section is very small, and the dominant thermal neutron absorber among the cadmium isotopes, 113Cd, leads to a stable product.
Some activation foil materials, such as rare earth elements, can be incorporated into scintillators, and
Bell et al. [20] show a method for doing so. Figure 1 shows a spectrum obtained immediately after the
irradiation of plastic scintillator containing the tributylphosphate complex of Dy(NO3)3. The counting
time was 400 seconds starting within 20 seconds after the end of neutron irradiation. The scintillator
sample was irradiated for 10 minutes to saturate the activity of the 108 keV excited state of 165Dy.
Fig. 1 Spectrum of activated Dy
contained within plastic scintillator.
The peak near 126 keV is from the 75
second decay of the 108 keV first
165
excited state of Dy and is caused by
the ejection of L-shell electrons.
© 2005 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
1596
Z. W. Bell et al.: Neutron detection with cryogenics and semiconductors
Since the Dy is uniformly distributed within the scintillator, the low energy electrons generated in the
decay deposit all their energy within the scintillator, making a self-absorption correction unnecessary.
Activation foils can be used for spectroscopy by taking advantage of the thresholds for various reactions. Knoll [21] provides a list of candidates for threshold activation analysis but a perusal of the
threshold energies and the magnitude of the relevant cross sections quickly persuade the reader that precise spectroscopy is not feasible by this technique.
2.4 Devices based on radiation damage Detectors in this category are the track-etch devices and
bubble dosimeters. The track-etch devices use an insulating material that is damaged by the passage of
ionizing radiation, which must be heavy particles. Tracks are “developed” by etching the detecting medium in an acid or base and take advantage of the fact that undamaged material is removed more slowly
than damaged material. However, if the angle of incidence of the radiation is too great, then the etch rate
perpendicular to the surface will exceed the rate along the track and the track will not be enhanced. Neutrons are detected with these devices by placing a hydrogenous radiator in front of the track-etch material
to detect recoil protons, by making the detector with a plastic and using the hydrogen constituent, or by
placing the track-etch material near fissionable material, or by incorporating fissionable material in the
track-etch medium to detect fission fragments. Although the use of fissionable materials in this application does not lend itself to spectroscopy, measuring the distribution of the lengths of proton recoil tracks
provides analogous information to that which is obtained from hydrogenous scintillators.
Bubble dosimeters operate on the principle that radiation damage in a superheated fluid provides nucleation sites for bubble formation. The number of bubbles that form is a function of the incident neutron
energy and flux, and the temperature and pressure under which the fluid is kept. In addition, the threshold energy for bubble formation depends on the temperature and pressure of the fluid. By preparing a set
of dosimeters with varying degrees of superheating, crude spectroscopy can be performed.
3 Cryogenic detectors Cryogenic detectors operate by detecting the heat pulse generated by the
interaction of a neutron with the absorber. For a material with heat capacity C, the temperature rise due
to the absorption of energy ∆E is given by
∆T =
∆E
C
.
(1)
The heat capacity of the absorber must be adjusted by a combination of the proper material and operating
temperature so that the absorption of the interaction energy yields a detectable temperature rise that remains within the operational envelope of the detector. From knowledge of C, and the measurement of
∆T, the absorbed energy may be recovered via Eq. (1).
The resolution of a cryogenic detector, in contrast to a solid-state detector, does not depend on the
energy deposited in the absorber. Moseley et al. [22], have shown that in the simplest case of a single
absorber/detector thermal unit the distribution of pulses corresponding to monoenergetic radiation has
theoretical full width at half maximum given by
FWHM = 2.35ξ ( k B T ) 2 (C / k B ) ,
(2)
with ξ being a factor numerically between 1 and 2 that depends on the absorber and temperature sensor.
Equation (2) implies that for insulators following the Debye model, FWHM decreases as T5/2 at temperatures below 1 K, making low operating temperature highly desirable.
What nuclear reactions are appropriate for a cryogenic spectroscopic detector? From the discussion in
Section 2, the most appropriate reactions are those producing only charged particles, few or no gamma
rays, and having high absorption cross section. In such cases, the energy of the neutron is recovered from
Eq. (1) by subtracting the Q-value of the capture reaction from the calculated energy deposition. Examination of the nuclear data (Ref. [5], for example) narrows the candidate absorbers to 3He, 10B, and 6Li,
and, of these, 3He is not practical because it is an inert gas at room temperature and a mechanically stable
absorber would be extremely difficult to fabricate.
© 2005 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
phys. stat. sol. (c) 2, No. 5 (2005) / www.pss-c.com
1597
De Marcillac et al. [23], reports the detection of alpha particles from a variety of sources (these are
used because their range in LiF is only an order of magnitude smaller than the range of thermal neutrons)
and from thermal neutron capture in a 2-gram LiF crystal held at 80 mK. Temperature was sensed with
thermistors, and the authors report 16 keV resolution at 4.8 MeV deposited energy. They did not perform
measurements with fast neutrons, but, based on calibration with alpha particles and thermal neutrons,
they concluded that the response of the crystal was different when 4.78 MeV was deposited by an alpha
and triton rather than by a single alpha particle.
Silver et al. [24] report results obtained with a series of 6LiF crystals ranging in mass from approximately 80 mg to 2.6 g, with the best resolution of 39 keV obtained with the 80 mg crystal operating at
323 mK. They exposed the 2.6 g crystal to neutrons and were able to observe pulses caused by thermal,
3.996, 5.167, and 7.223 MeV neutrons, the high-energy neutrons having been obtained from a D-D accelerator source. The temperature sensor was a neutron transmutation doped p-type Ge thermistor.
Richardson et al. [25, 26], and Chowduri et al. [27] describe the use of Li alloys with Pb and Mg,
respectively, for use as flux monitors. In these particular instances, the goal of the work was to develop a
device for precise measurement (0.1% accuracy) of thermal and sub-thermal fluxes. Using 6LiPb alloy,
Richardson et al., first demonstrated a dual-compensated calorimeter (temperature of neutron absorber
and heat bath independently controlled) with a 5.7 cm diameter by 6.3 mm thick absorber sensitive to a
thermal current of 1.3 × 106 n/s, and in their subsequent paper, estimate the sensitivity of an improved
design to be a factor of 1300 n/s. Chowduri investigated 6Li0.76Mg0.24 in place of the LiPb alloy and reported sensitivity to neutron currents of 105 n/s.
Takahashi et al. [28], propose a detector based on superconducting MgB2 formed in a thin serpentine
strip on a sapphire substrate. The basic scheme is to monitor the resistance of the MgB2 strip while it is
irradiated by neutrons. With strips having cross sections of 1 µm2, the absorption of the reaction products
from the 10B(n, α) would be expected to cause sufficient local heating for a small volume of the strip to
become normal-conducting. However, the authors indicated that they were only in the preliminary design
stages of the project and did not present count rate or pulse-shape data.
Nakamura et al. [29], report that they have fabricated a neutron detector from a 0.3 mm thick Li2B4O7
single crystal made from natural materials and read out with superconducting tunnel junctions. The crystal was held at 450 mK and exposed to both gamma and neutron sources. As expected, the device
showed little efficiency for gamma rays, but, oddly, did not show peaks corresponding to capture by
lithium and boron. In another paper [30], Nakamura et al., immersed an In-Sb semiconductor in 3He gas
held at 4.2 K to detect reaction products from the 3He(n, t) reaction. Although operated at cryogenic
temperatures, the device is more properly classified as a coated semiconductor detector similar in operation to the coated diodes described in Section 2.1.
The present authors have recently reported [31] initial results with a bolometer read out by a transition
edge sensor (TES), a very sensitive thermometer held between the normal and superconducting state,
whose current is sensed by a SQUID. A small rise in temperature results in a large change in TES resistance, which in turn results in a large change in current in the bias circuit. This changing current passes
through a coil, and changes the magnetic field sensed by the SQUID, which is in the negative feedback
loop of an amplifier, and permits the measurement of small currents in low impedance circuits.
The structure of the absorber is shown in Fig. 2, below; the reader should be aware that this is a proofof-principle device intended to demonstrate a TES/SQUID instrument with a boron-containing absorber.
The absorber is glued to the TES which has been deposited directly onto a silicon nitride membrane
grown on a silicon wafer. After the silicon nitride is deposited, the silicon wafer is etched away to reduce
the thermal conductivity between the TES and the silicon wafer (which is coupled to the cold bath). The
dimensions of the absorber are approximately 1 mm2 by 0.25 mm thick. The refrigerator is described by
Friedrich et al. [32].
The TES is a Mo/Cu multilayer with R-T characteristics as shown in Fig. 3. Noting that the width of
the transition region is only a few tenths of a mK, and knowing that the energy deposited in a boron- or
lithium-based absorber is 2 – 5 MeV + En, it is easily seen that the absorber’s heat capacity must be ap© 2005 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
1598
Z. W. Bell et al.: Neutron detection with cryogenics and semiconductors
proximately 10 pJ/mK (60 MeV/mK) for the TES to remain in the linear portion of the transition region.
In addition, the thinness of the Si3N4 membrane requires the absorber to have a mass of only a few mg.
Among the boron compounds, TiB2, a normal conductor, has the required characteristics.
Fig. 2 Schematic of arrangement of neutron
absorber and Mo/Cu TES on Si3N4 membrane.
The drawing is not to scale.
0.40
Fig. 3 Variation of resistance of TES as a
function of temperature near the transition.
Note that the transition occurs in only a few
tenths millikelvin, making the sensitivity
approximately 1000 Ω/K in the transition
region.
0.35
Resistance (Ω)
0.30
0.25
0.20
0.15
0.10
0.05
0.00
0.144
0.146
0.148
0.15
0.152
Temperature (K)
Fig. 4 Heat capacity of TiB2
as measured in this work
and compared to earlier
measurements. The inset
shows the data from 500
mK to 8 K in greater detail.
© 2005 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
phys. stat. sol. (c) 2, No. 5 (2005) / www.pss-c.com
1599
The heat capacity of TiB2, as measured in this work between 500 mK and 100 K and compared to data
from the literature [33, 34], is shown in Fig. 4. The low temperature data is well fit by the Debye model
alone or with the addition of a Schottky anomaly ascribable either to the boron or to the presence of
impurities. Extrapolating the Debye fit to 100 mK (the desired operating temperature), results in a predicted heat capacity of 0.100 mJ/K/mol. If a Schottky anomaly is hypothesized, then the two models
(boron alone or impurities) lead to predicted heat capacities of 0.435 and 0.408 mJ/K/mol at 100 mK,
respectively. In the impurity model, an entirely plausible concentration of 450 ppm fit the data.
300
Fig. 5 Spectrum of pulses from TiB2
absorber irradiated with thermal neutrons.
The two peaks are from disintegrations
leaving Li in its ground state and first
excited state.
Counts
250
200
150
100
50
0
2.0
2.2
2.4
2.6
Energy (MeV)
2.8
3.0
Figure 5 shows a spectrum obtained from a thermalized 252Cf source using a TES/SQUID bolometer
with a 4 mg (~12 pJ/mK) TiB2 absorber in the configuration of Fig. 2. The two peaks have areas in the
ratio of 14.86:1, exactly what would be expected from the 10B(n, α) reaction at thermal energies. The
width of the 2.31 MeV peak is 10.5 keV while that of the 2.79 MeV peak is 5.5 keV, the discrepancy
being caused by the emission of a 478 keV gamma ray from the decay of excited 7Li nuclei in flight.
This introduces an uncertainty in deposited energy of 9.8 keV, and if the instrument’s resolution is assumed to be 5.5 keV, then when the 9.8 keV is added in quadrature, the observed resolution is obtained.
While a small absorber is sufficient for thermal neutrons, the 1/v dependence of the boron and lithium
cross sections demand that absorbers with masses of approximately 10 grams be used. Consequently, a
holder to accommodate absorbers with volumes of 15 cm3 has been designed, and is shown in Fig. 6.
Fig. 6 Large-absorber holder with 0.8 cm3 natLiF crystal. The crystal is supported on thermally insulating balls mounted on screws,
two of which are visible at the left and one is visible under the
crystal.
In this configuration, the arrangement of Fig. 2 was inverted and the weak thermal link between the
TES and cold bath was realized with thin wires. It should be noted that while the holder is at 100 mK, the
absorber’s supports are thermally insulating so that heat generated by the absorption of neutrons must
flow through the glue to the TES and then through the contact wires that are maintained at the temperature of the bath. A spectrum obtained in this configuration is shown in Fig. 7, below. The resolution seen
in this spectrum is approximately 45 keV and is thought to be caused positional dependence of the response. The low energy continuum was caused by inelastic scattering of neutrons, and scattering of
gamma rays.
© 2005 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
1600
Z. W. Bell et al.: Neutron detection with cryogenics and semiconductors
90
80
70
Counts
60
Fig. 7 Spectrum of obtained with 0.8 cm3
LiF crystal in holder of Fig. 6. Note the
high-energy tail in the peak indicative of
the presence non-thermal neutrons.
nat
50
40
30
20
10
0
0.0
1.0
2.0
3.0
4.0
5.0
6.0
Energy (MeV)
Which will make the better absorber, 10B or 6Li? The advantages of 6Li are the higher Q-value of the
capture reaction (4.78 v. 2.79 MeV) and the fact that there is only a single available nuclear channel
(there are two capture reactions in boron, with Doppler broadening in one; see Fig.5). The higher Qvalue is important because capture events will be offset by the Q-value from scattering events and
gamma ray interactions. However, of lithium compounds, only LiF is suitable for use because of its relative inertness and the relatively high density of lithium (0.064 atoms/b-cm), while there is a myriad of
air- and water-stable borides, not to mention elemental 10B with a density of 0.13 atoms/b-cm. In addition, inelastic neutron scattering from fluorine (the threshold for which is 115 keV) and from 6Li itself
(important above 2.5 MeV) result in escaping gamma rays and neutrons whose now lowered energy
makes them more likely to be captured. Unfortunately, capture of these neutrons distorts the spectrum if
their coincident gamma rays escape, making the quality of the spectrum from a lithium absorber dependent, then, on the fate of inelastic gammas. In contrast to this, the effects of inelastic scattering in elemental 10B become important only above 4.4 MeV. Therefore, in addition to using Li absorbers, we are investigating the possibility of fusing elemental boron according to the procedure of Holcombe et al. [35],
no mean feat, considering that the melting point is approximately 2100 C. It should be noted that elastic
scattering followed by neutron capture does not distort the spectrum because the ion’s recoil energy is
completely absorbed in the detector.
4 Semiconductor detectors For the purposes of this paper, the term “semiconductor detector” refers
to devices in which the semiconducting material absorbs or interacts with the incident neutron rather than
merely detecting reaction products produced externally to the device. That is, coated semiconductors
(discussed in Section 2.1, above) will be omitted.
Silicon is useful as a semiconductor detector owing to the 28Si(n, p) and 28Si(n, α) reactions, and to the
state of the art of manufacturing, enabling the fabrication of devices with surface areas of hundreds of
square millimetres. Unfortunately, since the above-mentioned reactions have small cross sections (peak
values are ~300 mb [36]), silicon diodes typically are not more than a few mm thick, and the reactions
have thresholds near 3 and 4 MeV, respectively, silicon finds application in high flux, high energy environments. The typical interaction rate in 1 mm can be calculated to be approximately 0.1% at 12 MeV.
Miller and Kavanagh [37] report the use of lithium-drifted silicon diodes with incident neutrons in the
energy range 6 – 17 MeV. They were able to identify peaks in the pulse height spectrum from alpha
particles leaving 25Mg in its ground state and first four excited states, 28Al in its ground state and first
eight excited states (although not all were resolved), and 26Mg in its ground and first excited states. The
latter species arises from the 29Si(n, α) reaction; 29Si making up 4.7% of the natural. From knowledge of
the various cross sections and resolution of the detector, and assuming that all (except for a 31 keV
gamma from the first excited state of 28Al) gamma rays escape, the pulse height spectrum can be un© 2005 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
phys. stat. sol. (c) 2, No. 5 (2005) / www.pss-c.com
1601
folded to recover the incident spectrum. Dési [38] using the same technique used Si(Li) detectors and
silicon diodes to measure fast fluxes in a reactor.
Chung and Chen [39] have measured neutron fluxes with high purity Ge detectors usually used for
gamma ray spectroscopy. The technique takes advantage of the 73Ge(n, γ), 72,74Ge(n, n’γ) reactions which
produce gamma rays at 596 and 691 keV. The authors report that in a 10% detector (58 cm3), a count rate
of 0.5 event/s in the 596 photopeak at an incident flux of 10 n/cm2/s. However, since this technique
measures the gamma ray energies, the recoil energy of the Ge nucleus is lost except for inelastic scattering leading to Doppler broadening of photopeaks.
Diamond as a semiconductor detector has been exploited by Kozlov et al. [40], Kovalchuk, Trotsik,
and Kovalchuk [41], Bergonzo et al. [42], and Schmid et al. [43]. In these instances, the authors applied
an electric field across a diamond stones or CVD films to collect the charge generated by (n, α) or (n, n)
reactions. The very recent work of Schmid et al., shows exceptional results with resolution of 2.9% at 14
MeV. Below the (n, α) threshold, however, a diamond detector relies on elastic and inelastic scattering
from carbon and so faces the same difficulties that proton recoil detectors at these energies: The spectrum must be unfolded to recover spectroscopic information.
Doty and collaborators have investigated Li2B4O7, and BaB4O7 [44], hexagonal boron nitride [45], and
organic semiconductors [46] for neutron detectors. Semiconducting detectors based on boron carbide
have already been discussed in Section 2.1. References [43] and [44] describe devices intended for detection only, while Ref. [45] describes thin film devices that, unless loaded with some material like lithium
or boron, will not have sufficient efficiency for spectroscopy applications. In addition, the spectrum
developed by an organic semiconductor will require unfolding as described in Section 2.1. Organic
semiconductors are still in their infancy with respect to their use as radiation detectors.
CdZnTe and CdTe have also been used for thermal neutron detection. Siffert et al. [47], Vradii et al.
[48], Fasasi et al. [49], and McGregor et al. [50] have all used the capture gamma rays from the
113
Cd(n, γ)114Cd reaction as a signature of the presence of neutrons, and show that the 558 keV emission
is easily seen in small detectors. Vradii concludes from observed count rates that the efficiency of the
CdTe devices is comparable to that of an unspecified 3He tube, when the entire pulse-height spectrum is
used. Fasasi, however, reports efficiency of about 5%, based on the counts in the entire spectrum, 0.5%
when only the 96 keV gamma is used, 0.38% when the 558 keV line is measured with a 2 mm thick
detector, and 0.28% when that line is measured with a 1 mm thick detector. McGregor reports 3.7%
efficiency for the 558 keV gamma ray with a 3 mm thick CdZnTe detector. It is not obvious that all these
measurements are in agreement.
The cadmium reaction may not be the best choice for semiconductor detectors based on capture
gamma rays. The de-excitation of 114Cd contains hundreds of gamma rays between 60 keV and 9 MeV,
with 6 (558, 661, 805, 756, 725, and 96 keV) having probabilities of emission exceeding 1%. Since these
gamma rays are emitted in cascades, the detector sees them as simultaneous, and sums the energy from
each deposited in it. Since neutron capture occurs within 1 mm of the surface of CZT, and the interactions of all the gamma rays are independent of each other, and there are a large number of individual
gamma rays, the total amount of energy deposited by each event must resemble a continuum more than a
spectrum of discrete peaks, serving to decrease the efficiency as measured by examining the photopeak.
If a detector constituent emits only one gamma ray, however, or if the other gamma rays are invisible to
the detector, then the signature gamma ray will stand out from the continuum.
Beyerle and Hull [51], and Bell, Pohl, and van den Berg [52] have exploited this property of the
199
Hg(n, γ) reaction to detect thermal neutrons. 199Hg comprises 16.9% of natural mercury, but has a capture cross section of 2200 b [53]. Since the dominant capture gamma ray from this reaction has energy
368 keV and occurs in 81% of events and the vast majority of others have energies in excess of 1.5 MeV
[54], a thin HgI2 detector would be expected to generate a 368 keV photopeak broadened by the small
amount of energy deposited by the other gamma rays. This is exactly what was observed in Refs. [50]
and [51].
Bell, Pohl, and van den Berg modified the arrangement of Beyerle and Hull by adding a 100% absorptive (at 0.0253 eV) 10B cladding to the HgI2 detector. The proximity of the boron to the detector ensured
© 2005 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
1602
Z. W. Bell et al.: Neutron detection with cryogenics and semiconductors
that the single 478 keV gamma ray from the 10B(n, α) reaction had a high probability of entering the
HgI2 crystal, and both gamma rays were observed. It should be noted that since the energies of these
gamma rays is lower than the main ones from Cd, the probability of gamma ray detection is higher in an
HgI2-based device.
Both Cd- and Hg-based devices are thermal neutron detectors; no spectral information for the neutrons
can be obtained from measurement of the gamma rays. However, consider an HgI2 detector, with a boron
cladding, that has been packed in polyethylene moderator, as shown in Fig. 8. Monte Carlo modelling
indicates that crude spectroscopic information over a broad energy range can be extracted from an analysis of the two capture gamma rays.
Fig. 8 Schematic of HgI2 detector packed in
polyethylene. The HgI2 is the topmost
horizontal structure; the boron cladding is
not visible, but is on the top surface of the
HgI2. The vertical structures are the printed
circuit boards and high-voltage connector
of the detector. The vertical lines represent
the 1, 2, 4, and 8 cm boundaries of polyethylene moderator. There are 4 cm of polyethylene behind. Neutrons, at constant
fluence, are incident from the region above.
Essentially, this configuration is expected to behave like the Bonner spheres and long counter described by Knoll [55]. Figure 9 shows the effects of increasing wall thickness on the calculated photopeak areas as a function of incident neutron energy. The dramatic increase of the number of Hg photopeak (368 keV) events with increasing moderator thickness is indicative of the increase in the number
of returning neutrons, while the only modest increase in boron events (478 keV) indicates that the HgI2
layer is not 100 % absorptive. Figure 9 also shows the response, as measured by the ratio of the 478 keV
peak to the 368 keV peak, is approximately constant from approximately 10 eV to 10 MeV for the thinnest walls.
Fig. 9 Calculated [56] locus of points defined by the photopeak heights at 368 keV
and 478 keV for a boron-clad 2.6 mm thick
HgI2 detector in the configuration of Fig. 8.
Each labelled line indicates the thickness of
the polyethylene side walls. The points along
each line represent, clockwise from the top,
0.025 eV, 0.1 eV, 1 eV, 10 eV, 100 eV, 1
keV, 10 keV, 100 keV, 1 MeV, and 10 MeV.
What does neutron detection with gamma-detecting semiconductors gain us? First, neutron detection
comes with little or no additional cost: neutrons are treated just like any other gamma ray. Second, the
combined device is still a semiconductor, and as such is more compact than two individual detectors. As
© 2005 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
phys. stat. sol. (c) 2, No. 5 (2005) / www.pss-c.com
1603
such, these kinds of devices are likely to find application in hand-held instrumentation and spaceflight,
where power consumption, mass, and volume are at a premium.
Neutron-gamma discrimination in the Cd- and Hg- devices described above is accomplished by pulseheight analysis and its effectiveness depends on the resolution of the detector, and the incident gamma
spectrum. The use of multiple independent gamma rays (such as by the addition of boron or other external absorbers) decreases the probability of false positive identification of neutrons because the multiple
gamma rays must appear in specific relation to each other.
5 Conclusions The future of semiconductor neutron detectors is reasonably bright, although in the
near term it will probably be limited to thermal neutron detection. There is active research in boronbased materials whose theoretical thermal neutron efficiencies can approach 100%. However, the small
volumes of these devices will make it necessary to construct large-area detectors by tiling. Using present-day technologies, such detectors will be financially impractical because of the cost of over US
$1000/cm3 for finished material. Similar considerations apply to organic semiconductors.
Spectroscopy with semiconductor devices will probably not be economically feasible except with low
efficiency in polyethylene-clad silicon (or other mass-produced semiconductor), boron-based materials,
or with thermal detectors in Bonner sphere configurations. In the last case, however, the semiconductor
is replacing a much less expensive scintillator or 3He tube and is not a particularly economical substitution. However, boron-based semiconductors offer the possibility of resolution similar to that obtained
with gamma rays (few per cent in room-temperature devices), but further research is needed to discover
new materials amenable to mass production.
Cryogenic detectors can be expected to find applications in laboratory or other controlled settings for
the detection of fast neutrons. The resolution offered by these devices is expected to enable the identification of neutron sources via analysis of the observed spectrum and perhaps the source’s chemical form
and identity of shielding material by observation of absorption and transmission resonances. Although
not likely to become field instruments, the devices can be made transportable, and improvements in refrigeration techniques, such as the use of mechanical coolers in place of liquid nitrogen, pulse tubes in
place of liquid helium, and a 2-stage adiabatic demagnitization refrigerator to bring the absorber to 0.1
K, may well further that end.
From a materials point of view, although the 6Li(n, α) reaction has a higher Q-value, the chemical
instability (with respect to oxygen and water) of many lithium-containing compounds, the hazards associated with their handling, and their relatively low lithium content make boron-rich superconductors,
insulators and semiconductors attractive. However, since many boron-rich materials (such as MgB2) are
often not isotropic, not easily worked because they do not melt congruently, or, if they do, they do so
only at high temperatures (making their single crystals or polycrystals difficult to grow, and leaving
sintering the only alternative), it is not obvious that a satisfactorily large boron absorber is realizable.
Nevertheless, since sintering is often possible, and the temperature sensor can be glued to the absorber
with seemingly good results, the immense variety of boron-rich solids makes the search for suitable
materials among them an exciting prospect.
Acknowledgments This work was supported by the U. S. Department of Energy, Office of Nonproliferation
Research and Engineering (NA-22). The Oak Ridge National Laboratory is managed and operated for the U. S.
Department of Energy by UT-Battelle, LLC, under contract DE-AC05-00OR22725. The Y-12 National Security
Complex is managed and operated for the U. S. Department of Energy by BWXT Y-12, LLC, under contract DEAC05-00OR22800. Lawrence Livermore National Laboratory is managed and operated for the U. S. Department of
Energy by the University of California under contract W-7405-Eng-48.
References
[1] M. F. Cunningham et al., Appl. Phys. Lett. 81, 159 (2002).
[2] J. E. Baciak and Z. He, Nucl. Instrum. Methods Phys. Res. A 505, 191 (2003).
© 2005 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
1604
Z. W. Bell et al.: Neutron detection with cryogenics and semiconductors
[3] A. Shor, Y. Eisen, and I. Mardor, IEEE Trans. Nucl. Sci. 51, 1204 (2004).
[4] P. G. Litovchenko et. al., Nucl. Instrum. Methods Phys. Res. A 518, 423 (2004).
[5] F. William Walker, Josef R. Parrington, and Frank Feiner, Nuclides and Isotopes, Fourteenth Edition (General
Electric Company, San Jose, 1989), p. 92.
[6] T. M. Filho et al., Nucl. Instrum. Methods Phys. Res. A 458, 441 (2001).
[7] D. S. McGregor et al., Nucl. Instrum. Methods Phys. Res. A 380, 271 (1996).
[8] D. S. McGregor et al., IEEE Trans. Nucl. Sci. 49, 1999 (2002).
[9] T. Aoyama et al., Nucl. Instrum. Methods Phys. Res. A 314, 590 (1992).
[10] M. J. Knitel et al., Nucl. Instrum. Methods Phys. Res. A 443, 364 (2000).
[11] Z. W. Bell, G. M. Brown, C. H. Ho, and F. V. Sloop, Jr., Proc. SPIE, 4784, 150 (2002).
[12] M. Bowen et al., IEEE Trans. Nucl. Sci. 36, No. 1, 562 (1989).
[13] J. B. Birks, The Theory and Practice of Scintillation Counting (Pergamon, Oxford, 1964), chapters 3 and 6.
[14] Th. Förster, Ann. Phys. 2, 55 (1948).
[15] G. D. Scholes, Annu. Rev. Phys. Chem. 54, 57 (2003).
[16] B. W. Robertson et al., Appl. Phys. Lett. 80, 3644 (2002).
[17] A. N. Caruso et al., J. Phys.: Condens. Matter 16, L139 (2004).
[18] Z. W. Bell, U.S. Patent 6,011,266, 4 January 2000.
[19] R. T. Johnson, Jr., Nucl. Instrum. Meth. 77, 189 (1970).
[20] Z. W. Bell et al., U. S. Patent 6,544,442, 8 April 2003.
[21] G. F. Knoll, Radiation Detection and Measurement, Third Edition (Wiley, New York, 2000), pp. 748-751.
[22] S. H. Moseley, J. C. Mather, and D. McCammon, J. Appl. Phys. 56, 1257 (1984).
[23] P. de Marcillac et al., Nucl. Instrum. Methods Phys. Res. A 337, 95 (1993).
[24] C. S. Silver et al., Nucl. Instrum. Methods Phys. Res. A 485, 615 (2002).
[25] J. M. Richardson et al., Nucl. Instrum. Methods Phys. Res. A 306, 291 (1991).
[26] J. M. Richardson et al., IEEE Trans. Nucl. Sci. 45, 550 (1998).
[27] Z. Chowduri et al., Rev. Sci. Instrum. 74, 4280 (2003).
[28] K. Takahashi et al., Physica C 392-396, 1501 (2003).
[29] T. Nakamura et al., Nucl. Instrum. Methods Phys. Res. A 520, 67 (2004).
[30] T. Nakamura et al., Nucl. Instrum. Methods Phys. Res. A 520, 76 (2004).
[31] T. Niedermayr et al., Nucl. Instrum. Methods Phys. Res. A 520, 70 (2004).
[32] S. Friedrich et al., Nucl. Instrum. Methods Phys. Res. A 467-468, 1117 (2001).
[33] J. Castaing, R. Caudron, G. Toupance, and P. Costa, Solid State Commun. 7, 1453 (1969).
[34] E. F. Westrum and G. A. Clay, J. Chem. Thermo. 10, 629 (1978).
[35] C. E. Holcombe et al., High Temp. Sci. 5, 349 (1973).
[36] M. B. Chadwick and P. G. Young, ENDF/B-VI Mod 2 Evaluation, July 1997, available on-line from the National Nuclear Data Center, Brookhaven National Laboratory, www.nndc.bnl.gov.
[37] R. G. Miller and R. W. Kavanagh, Nucl. Instrum. Meth. 48, 13 (1967).
[38] S. Dési, Nucl. Instrum. Meth. 140, 467 (1977).
[39] C. Chung and Y. R. Chen, Nucl. Instrum. Methods Phys. Res. A 301, 328 (1991).
[40] S. F. Kozlov, E. A. Konorova, and I. A. Kuznetsov, IEEE Trans. Nucl. Sci. NS-24, 235 (1977).
[41] V. D. Kovalchuk, V. I. Trotsik, and V. D. Kovalchuk, Nucl. Instrum. Methods Phys. Res. A 351, 590 (1994).
[42] P. Bergonzo et al., Nucl. Instrum. Methods Phys. Res. A 476, 694 (2002).
[43] G. J. Schmid, J. A. Koch, R. A. Lerche, and M. J. Moran, Nucl. Instrum. Methods Phys. Res. A 527, 554
(2004).
[44] F. P. Doty, I. Zwieback, and W. Ruderman, U. S. Patent 6,388,260, 14 May 2002.
[45] F. P. Doty, U. S. Patent 6,727,504, 27 April 2004.
[46] F. P. Doty, C. M. Munoz, D. H. Morse, D. A. Chinn, and A. J. Anatolak, paper 5541-10, presented at SPIE
Conference 5541, Penetrating Radiation Systems and Applications VI, Denver, Colorado, 4-5 August 2004.
[47] P. M. Siffert et al., U. S. Patent 3,999,071, 21 December 1976.
[48] A. G. Vradii, Atomnaya Energiya 42, 64 (1977).
[49] M. Fasasi, M. Jung, P. Siffert, and C. Teissier, Radiat. Prot. Dosim. 23, 429 (1988)
[50] D. S. McGregor, J. T. Lindsay, and R. W. Olsen, Nucl. Instrum. Methods Phys. Res. A 381, 498 (1996).
[51] A. G. Beyerle and K. L. Hull, Nucl. Instrum. Methods Phys. Res. A 256, 377 (1980).
[52] Z. W. Bell, K. R. Pohl, and L. van den Berg, IEEE Trans. Nucl. Sci. 51, 1163 (2004).
[53] F. William Walker, Josef R. Parrington, and Frank Feiner, op. cit., p. 42.
© 2005 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
phys. stat. sol. (c) 2, No. 5 (2005) / www.pss-c.com
1605
[54] J. K. Tuli, Thermal Neutron Capture Gamma Rays, available on-line from the National Nuclear Data Center,
Brookhaven National Laboratory, www.nndc.bnl.gov.
[55] G. F. Knoll, op. cit., pp. 539 – 545.
[56] MCNP – A General Monte Carlo N-Particle Transport Code, Version 5, available from the Radiation Safety Information Computational Center, Oak Ridge National Laboratory, Oak Ridge, TN, USA.
© 2005 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim