Optimizing diode thickness for thin-film solid state thermal neutrondetectorsJohn W. Murphy, George R. Kunnen, Israel Mejia, Manuel A. Quevedo-Lopez, David Allee et al. Citation: Appl. Phys. Lett. 101, 143506 (2012); doi: 10.1063/1.4757292 View online: http://dx.doi.org/10.1063/1.4757292 View Table of Contents: http://apl.aip.org/resource/1/APPLAB/v101/i14 Published by the American Institute of Physics. Related ArticlesUltrafast semiconductor x-ray detector Appl. Phys. Lett. 101, 031107 (2012) X-ray bang-time and fusion reaction history at picosecond resolution using RadOptic detection Rev. Sci. Instrum. 83, 10D307 (2012) Self-powered micro-structured solid state neutron detector with very low leakage current and high efficiency Appl. Phys. Lett. 100, 243507 (2012) Position and energy-resolved particle detection using phonon-mediated microwave kinetic inductance detectors Appl. Phys. Lett. 100, 232601 (2012) Time response of Cd0.9Zn0.1Te crystals under transient and pulsed irradiation AIP Advances 2, 012162 (2012) Additional information on Appl. Phys. Lett.Journal Homepage: http://apl.aip.org/ Journal Information: http://apl.aip.org/about/about_the_journal Top downloads: http://apl.aip.org/features/most_downloaded Information for Authors: http://apl.aip.org/authors

Downloaded 30 Oct 2012 to 129.110.241.14. Redistribution subject to AIP license or copyright; see http://apl.aip.org/about/rights_and_permissions

Optimizing diode thickness for thin-film solid state thermal neutrondetectors

John W. Murphy,1,a),b) George R. Kunnen,2,b) Israel Mejia,1 Manuel A. Quevedo-Lopez,1

David Allee,2 and Bruce Gnade1

1Department of Materials and Science, University of Texas at Dallas, Richardson, Texas 75080, USA2Flexible Display Center at Arizona State University, Tempe, Arizona 85284, USA

(Received 22 August 2012; accepted 20 September 2012; published online 4 October 2012)

In this work, we investigate the optimal thickness of a semiconductor diode for thin-film solid state

thermal neutron detectors. We evaluate several diode materials, Si, CdTe, GaAs, C (diamond), and

ZnO, and two neutron converter materials, 10B and 6LiF. Investigating a coplanar diode/converter

geometry, we determine the minimum semiconductor thickness needed to achieve maximum neutron

detection efficiency. By keeping the semiconductor thickness to a minimum, gamma rejection is kept

as high as possible. In this way, we optimize detector performance for different thin-film semiconductor

materials. VC 2012 American Institute of Physics. [http://dx.doi.org/10.1063/1.4757292]

A solid state semiconductor neutron detector typically

consists of two stacked layers: a neutron converter film and a

semiconductor diode. The principle of operation is that

neutrons absorbed in the converter film generate charged

particles, which are subsequently detected by the ionization

produced in the semiconductor diode; Figure 1 shows a

diagram illustrating the concept. There are several reports in

the literature optimizing the converter layer thickness.

McGregor et al. published a definitive work on designing

architectures for thin-film coated semiconductor thermal

neutron detectors in 2003; however, one aspect of the design

which is not considered is the thickness of the semiconductor

diode itself.1 Many of the more recent reports simulate perfo-

rated or textured semiconductor diodes, used to increase

neutron detection efficiency.2–4

Single crystal thin films of Si in both silicon-on-insula-

tor (SOI) and free standing forms can be purchased. It is well

known that using a thinner semiconductor film will reduce

interference from gamma interactions. This has been demon-

strated using SOI wafers.5,6 However, to date there has been

no systematic study of the optimum thickness of the semi-

conductor diode, even for the simplest coplanar arrange-

ments of semiconductor and converter. One exception is a

1981 report which discusses the optimal thickness of both

converter and semiconductor for a theoretical LiF:polyacety-

lene based device.7 There is also a recently awarded United

States patent which describes the optimum thickness of the

semiconductor portion of a solid state neutron detector as

10–30 lm, and although they do not describe how they arrive

at this thickness range, it is in excellent agreement with our

results.8

Other than silicon, there are recent reports of additional

high quality, single crystal semiconductor films that have

been or might conceivably be utilized as thermal neutron

detectors. For instance, Almaviva et al. fabricated chemical

vapor deposited diamond neutron detectors in the B-doped/

intrinsic diamond/Al configuration with an intrinsic layer

thickness of 25–100 lm.9 Kaidashev et al. report ZnO films

grown by pulsed-laser deposition on sapphire substrates

with a thickness >1 lm. They achieve a Hall mobility of

100 cm2V�1s�1 with a carrier concentration of only

7.5� 1015 cm�3.10 Single-crystal GaAs films with a 2 lm

thickness for solar cells have been prepared using epitaxial

growth and lift-off in which the growth substrate can be

recycled and used for subsequent epitaxial depositions.11

Finally, there is a report of epitaxial growth of CdTe with a

thickness of 200 lm on a GaAs substrate. Although the

thickness is quite high, the authors report the film is sensitive

to alpha particles.12

We utilized the publicly available Monte Carlo based

computer modeling programs MCNP5 and MCNPX v2.7.0 to

simulate neutron, charged particle and photon/electron trans-

port. Our primary tool within these programs was the pulse

height tally, which allows easy determination of the energy

deposited in a material volume by incident charged particles

and photons.13,14 This allowed us to model both the effi-

ciency of thin film semiconductor neutron detectors, and

evaluate the susceptibility of the detectors to background

gamma interference. Densities, atomic weights, and isotopic

ratios used for simulations were taken from the 89th edition

of CRC Handbook.15

We first determined the optimal thickness of both con-

verter materials, since this is largely independent of the

choice of material or thickness for the diode. Converter layer

thickness optimization was conducted through a series of

FIG. 1. Schematic drawing showing principle of operation for a solid-state

thermal neutron detector using a 10B (left) or 6LiF (right) converter layer, and

the approximate range for each of the charged particle reaction products.

a)Author to whom correspondence should be addressed. Electronic mail:

johnwmurphy@utdallas.edu.b)John W. Murphy and George R. Kunnen contributed equally to this work.

0003-6951/2012/101(14)/143506/5/$30.00 VC 2012 American Institute of Physics101, 143506-1

APPLIED PHYSICS LETTERS 101, 143506 (2012)

Downloaded 30 Oct 2012 to 129.110.241.14. Redistribution subject to AIP license or copyright; see http://apl.aip.org/about/rights_and_permissions

approximately 60 MCNPX simulations for both converter

materials, using 5� 107 particle histories for each converter

thickness. We evaluate intrinsic neutron detection efficiency

for a normally incident beam of thermal neutrons (0.026 eV),

as described by McGregor and Shultis.16 Using the pulse

tally feature in the latest version of MCNPX (v2.7.0), it is pos-

sible to directly simulate the energy deposited in the detector

volume by the alpha, triton, and 7Li ions generated by ther-

mal neutron capture reactions in the converter layer, and

thus predict the neutron detection efficiency of different con-

verter/diode arrangements. Analog or explicit capture was

used in these simulations, and the neutron capture ion algo-

rithm (NCIA) was used to produce correlated charged parti-

cle products from neutron capture reactions. The total

number of pulses generated, however, does not represent the

true or effective efficiency of such a device, because in prac-

tice a lower-level discriminator (LLD) must be implemented

to separate actual neutron events from background induced

pulses. We evaluated a range of LLD values from 300 to

1800 keV.

We then used the optimal thickness for each converter

layer using a 300 keV LLD to study the minimum thickness

required for each semiconductor material to collect a certain

quantity of energy from either of the charged particle prod-

ucts, i.e., absorb enough energy from the charged particle to

produce a “detectable” pulse. In order to accomplish this, we

ran a series of 610 simulations with MCNPX, with each simula-

tion having an increasing thickness of the semiconductor

diode layer for the five different semiconductors and

two converting layers. The reason for choosing the specific

LLD setting of 300 keV is to closely follow McGregor’s

experiments described in Ref. 20.

In order to account for the detector gamma rejection

capability, we simulated the effect of different diode thick-

nesses while also varying the gamma ray energy over a broad

range. Similar to the optimizations for charged particle

detection, these simulations made use of the pulse height

tally in order to calculate intrinsic gamma-neutron detection

efficiency (eint,cn). The efficiency eint,cn is described using the

following equation:17

eint;cn ¼Number of photons that produce neutron counts

Number of photons incident upon the detector:

(1)

Ideally, this quantity should be zero—ensuring that no

gamma exposure triggers “false-positive” neutron counts.

The standard specification for radiation portal monitors,

however, requires eint,cn� 10�6. According to specifications

laid out by the Pacific Northwest National Laboratory

(PNNL), eint,cn should be measured with either a 192Ir or60Co gamma source.17,18 For our simulations, however, it

was beneficial to observe the gamma response over a larger

energy range of gamma exposure. For these simulations, a

standard LLD setting of 300 keV was chosen.

We also compare the gamma rejection performance of

thinner semiconductor layers, down to 3 lm, at two specific

gamma ray energies. We chose to evaluate the response of

the detector due to 511 keV and 1.46 MeV gammas because

these two energies are reported to be the two most frequently

FIG. 2. Intrinsic thermal neutron detection efficiency for 6LiF and 10B

converter films as a function of thickness and LLD. Thickness is shown on

the x-axis. LLD is designated by the small numbers near the curves.

FIG. 3. Detector efficiency as a function of semiconductor diode thickness; shown to the left is a 2.8 lm 10B converter layer (a), and to the right, a 26 lm 6LiF

layer (b).

143506-2 Murphy et al. Appl. Phys. Lett. 101, 143506 (2012)

Downloaded 30 Oct 2012 to 129.110.241.14. Redistribution subject to AIP license or copyright; see http://apl.aip.org/about/rights_and_permissions

encountered gamma-ray energies in an extremely long term

gamma background.19

The resulting intrinsic neutron detection efficiencies

predicted for the two converters on 500 lm of Si as a func-

tion of thickness and LLD are plotted in Figure 2. These

results closely agree with the calculations performed by

McGregor et al., in which similar curves for a LLD of

300 keV are shown.20 With optimal conversion layer thick-

nesses chosen for both 10B and 6LiF, it was possible to then

model the different diode materials and thicknesses.

It should be noted that no “dead” or contact layer was

used in these simulations. Such a layer, usually consisting of

�100 nm of metal and/or heavily doped semiconductor

between the converter layer and intrinsic semiconductor,

would reduce the energy of the incident charged particles.

Consider the charged particle products of a 2.8 lm 10B con-

verter layer. Using MCNPX, we find that normally incident

alpha and 7Li ions with a kinetic energy of 0.3-1.5 MeV lose

on average �30 and 50 keV, respectively, as they pass

through 100 nm of Al. This would reduce the efficiency of

the detector from about 4.6% to 4.2% by effectively raising

the LLD to �350 keV. For comparison, a 1 MeV triton will

lose only �10 keV going through 100 nm Al, showing that a6LiF converter will suffer comparatively less from this

effect. We chose not to implement a contact layer for two

reasons. First, the choice of contact material will depend on

the semiconductor. Second, the geometry of the detector

may result in the device having a non-uniform “dead” layer.

The neutron detectors demonstrated on SOI wafers in Ref. 6

did not use the traditional “sandwich” diode structure,

instead they used interdigitated finger contacts where both

electrodes are on the same side of the semiconductor layer.

In this case, parts of the “dead” layer are made up of the

heavily doped Si contacts but other regions of the “dead”

layer are composed of 500 nm of protective SiO2.

Figure 3 shows the predicted detector efficiency as a

function of thickness for the five selected semiconductor

materials with a 2.8 lm 10B and a 26 lm 6LiF converter

layer. Each of the curves represents a different LLD setting

for each material. As the diode increases in thickness, all of

the curves level off at the maximum efficiency for that LLD

setting and converter layer. Interestingly, Si and CdTe

require almost exactly the same thickness to achieve maxi-

mum efficiency, especially for 10B, indicating they have sim-

ilar stopping power for charged particles despite their large

difference in atomic number and bulk density. Diamond

requires the least thickness, approximately 50% less than Si

to absorb the same amount of energy, ZnO is close to dia-

mond, and GaAs falls in the middle. For instance, it requires

approximately 1.3 lm of either Si or CdTe to achieve

maximum efficiency for a 300 keV LLD setting and 2.8 lm10B film, while the thicknesses required for diamond, ZnO,

and GaAs would be about 0.6, 0.7, and 1.0 lm, respectively.

Results of the simulated intrinsic gamma–neutron

detection efficiency illustrate the maximum allowable diode

thicknesses for various materials. This value is found by

determining the intersection of the 3D pulse height curve

with the Z¼ 10�6 plane (see Figure 4). The lowest value of

diode thickness within this intersection represents the

maximum allowable diode thickness for that particular

material—according to the eint,cn� 10�6 specification. These

results are summarized in Table I.

In general, the added thickness of the 6LiF layer necessi-

tates a lower maximum diode thickness in order to fulfill the

eint,cn� 10�6 requirement. This presents an advantage of

using 10B over 6Li in a single conversion layer detection

FIG. 4. Gamma response as a function of diode thickness and incident gamma energy. This shows simulation results for the 10B and Si detector stack (a) and

the 10B and CdTe detector stack (b). The LLD for this simulation is 300 keV.

TABLE I. Maximum allowable diode thickness such that: eint,cn� 10�6 for

LLD¼ 300 keV (gamma energies: 0.3–1.5 MeV).

Diode

material

2.8 lm 10B

converter

26.5 lm 6LiF

converter

C (diamond) 60 lm 49 lm

Si 86 lm 74 lm

ZnO 36 lm 31 lm

GaAs 38 lm 30 lm

CdTe 28 lm 27 lm

143506-3 Murphy et al. Appl. Phys. Lett. 101, 143506 (2012)

Downloaded 30 Oct 2012 to 129.110.241.14. Redistribution subject to AIP license or copyright; see http://apl.aip.org/about/rights_and_permissions

approach. This is largely due to the thicker layer of 6LiF ena-

bling more electrons to be generated, as when compared to

the detectors with the 10B conversion layer. These electrons

are generated primarily due to Compton scattering. The very

small change in the CdTe thicknesses originates primarily

from the diode material enabling greater amounts of photo-

electric absorptions due to gamma rays with energies close

to the LLD setting.

Also observed in the 3D plots in Figure 4 is a valley for

all diode thicknesses, corresponding to gamma energies

around 450 keV. This valley, which can be observed for all

five materials, corresponds to the minimum gamma energy

required to generate Compton recoil electrons with adequate

energy to deposit past the LLD energy bin. The non-zero

pulses recorded at lower gamma energies near the LLD bin

are due to electrons generated from only photoelectric

absorption. The approximate location of this valley can be

derived using Eq. (2), which Knoll defines for the Compton

edge electron energy Ee� jh¼p, due to the photon energy hvwith a photon scattering angle h.21

LLD energy � Ee� jh¼p ¼ hv2hv=m0c2

1þ 2hv=m0c2

� �: (2)

With the electron rest mass energy m0c2 being about

511 keV, and solving for hv with a LLD setting of 300 keV,

we determine hv � 464:9 keV. This value corresponds very

closely to the gamma energy value observed in the valleys in

the simulated 3D surface plots.

Intrinsic gamma detection efficiencies corresponding to

the specific gamma energies and diode thicknesses are

reported in Table II. It is clear from these results that Si pro-

vides better gamma rejection as compared to the other four

materials. However, it is also clear that when considering all

five materials with diode thicknesses less than about 30 lm,

there is a rapid decrease in gamma detection efficiencies,

such that for thicknesses of <10 lm, the response is so low

that it is nearly zero. Here, it should also be noted that for

diode thicknesses of �10 lm, it is difficult to attain non-zero

tallies past the LLD energy bin. Even while running 1� 109

particle histories along with the forced collision variance

reduction technique, tallies of only very small values were

recorded. The very high gamma rejection rates are a strong

motivation for developing thin-film diode charged particle

detectors.

From these simulations, we have provided a range of

diode thicknesses and materials from which to design an

optimal neutron detector based on a single planar conversion

layer. Diamond and ZnO render the greatest stopping power

for charged particles, typically requiring only half the thick-

ness of Si to achieve maximum intrinsic thermal neutron

detection efficiency. We find that Si possesses the best

gamma rejection capability for a given thickness, even

greater than that of diamond; this is not surprising given that

diamond has the highest atomic density of any material and

actually has a greater electron density than silicon.22 How-

ever, for thicknesses of �10 lm, all of the materials are

essentially gamma “blind” with a LLD setting of 300 keV or

greater.

We would like to acknowledge the support of the United

States Department of Homeland Security and the National

Science Foundation, Grant No. ECCS-11139986.

1D. S. McGregor, M. D. Hammig, Y.-H. Yang, H. K. Gersch, and R. T.

Klann, Nucl. Instrum. Methods Phys. Res. A 500, 272 (2003).2J. K. Shultis, “Efficiencies of coated and perforated semiconductor neutron

detectors,” IEEE Trans. Nucl. Sci. 53, 1659 (2006).3C. J. Solomon, J. K. Shultis, W. J. McNeil, T. C. Unruh, B. B. Rice,

and D. S. McGregor, Nucl. Instrum. Methods Phys. Res. A 580, 326

(2007).4A. M. Conway, T. F. Wang, N. Deo, C. L. Cheung, and R. J. Nikolic,

IEEE Trans. Nucl. Sci. 56, 2802 (2009).5B. F. Phlips, F. J. Kub, E. I. Novikova, E. A. Wulf, and C. Fitzgerald,

Nucl. Instrum. Methods Phys. Res. A 579, 173 (2007).6M. Subramanian, B. Phlips, and F. Kub, in Nuclear Science SymposiumConference Record (NSS/MIC), Orlando, United States, 24 October–1November 2009 (IEEE, 2009), pp. 1306–1309.

7K. A. Kress, IEEE Trans. Nucl. Sci. 28, 807 (1981).8R. Orava and T. Schulman, “Ultra thin neutron detector, method for manu-

facturing the neutron detector and neutron imaging apparatus,” U.S. patent

US 8,022,369 B2, Washington, DC (2011).9S. Almaviva, M. Marinelli, E. Milani, G. Prestopino, A. Tucciarone,

C. Verona, G. Verona-Rinati, M. Angelone, D. Lattanzi, M. Pillon, R. M.

Montereali, and M. A. Vincenti, J. Appl. Phys. 103, 054501 (2008).10E. M. Kaidashev, M. Lorenz, H. von Wenckstern, A. Rahm, H.-C.

Semmelhack, K.-H. Han, G. Benndorf, C. Bundesmann, H. Hochmuth,

and M. Grundmann, Appl. Phys. Lett. 82, 3901 (2003).11G. J. Bauhuis, P. Mulder, E. J. Haverkamp, J. J. Schermer, E. Bongers,

G. Oomen, W. K€ostler, and G. Strobl, Prog. Photovoltaics 18, 155

(2010).12Q. Jiang, A. W. Brinkman, B. J. Cantwell, J. T. Mullins, F. Dierre,

A. Basu, P. Veeramani, and P. Sellin, J. Electron. Mater. 38, 1548

(2009).13“MCNPX user’s manual v2.7.0,” edited by D. B. Pelowitz, Los Alamos

National Laboratory Report No. LA-CP-11-00438, Los Alamos, NM,

2011.

TABLE II. Intrinsic gamma efficiencies for selected gamma energies and diode thicknesses (LLD¼ 300 keV and a 2.8 lm 10B conversion layer).

Thickness

(lm)

c energy

(keV) C (diamond) Si ZnO GaAs CdTe

3 511 0 0 0 0 0

1460 0 0 8� 10�14 0 9� 10�14

10 511 0 0 1� 10�12 0 6� 10�13

1460 3� 10�13 1� 10�13 8� 10�12 2� 10�12 9� 10�12

30 511 4� 10�9 6� 10�11 3� 10�7 1� 10�7 6� 10�7

1460 1� 10�8 1� 10�10 2� 10�7 6� 10�8 2� 10�7

100 511 1� 10�5 3� 10�6 2� 10�4 2� 10�4 4� 10�4

1460 1� 10�5 2� 10�6 9� 10�5 6� 10�5 1� 10�4

300 511 7� 10�4 3� 10�4 2� 10�3 2� 10�3 4� 10�3

1460 4� 10�4 2� 10�4 2� 10�3 2� 10�3 3� 10�3

143506-4 Murphy et al. Appl. Phys. Lett. 101, 143506 (2012)

Downloaded 30 Oct 2012 to 129.110.241.14. Redistribution subject to AIP license or copyright; see http://apl.aip.org/about/rights_and_permissions

14X-5 Monte Carlo Team, “MCNP–A general Monte Carlo N-particle trans-

port code, version 5–Vol. II: User’s guide,” Los Alamos National Labora-

tory Report No. LA-CP-03-0245, Los Alamos, NM, 2008.15CRC Handbook of Chemistry and Physics, 89th ed., edited by D. R. Lide

(CRC, Boca Raton, 2008).16D. S. McGregor and J. K. Shultis, Nucl. Instrum. Methods Phys. Res. A

632, 167 (2011).17R. T. Kouzes, J. R. Ely, A. T. Lintereur, and D. L. Stephens, “Neutron de-

tector gamma insensitivity criteria,” Pacific Northwest National Labora-

tory Report No. PNNL-18903, Richland, WA, 2009.

18Pacific Northwest National Laboratory, “Specifications for radiation portal

monitor systems revision 6.7,” Pacific Northwest National Laboratory

Report No. PNNL-14716, Richland, WA, 2003.19P. Bossew, Appl. Radiat. Isot. 62, 635 (2004).20D. S. McGregor, R. T. Klann, H. K. Gersch, and Y.-H. Yang, Nucl. Ins-

trum. Methods Phys. Res. A 466, 126 (2001).21G. F. Knoll, Radiation Detection and Measurement, 3rd ed. (John Wiley

& Sons, Boston, 2000), pp. 310–311.22G. J. Schmid, J. A. Koch, R. A. Lerche, and M. J. Moran, Nucl. Instrum.

Methods Phys. Res. A 527, 554 (2004).

143506-5 Murphy et al. Appl. Phys. Lett. 101, 143506 (2012)

Downloaded 30 Oct 2012 to 129.110.241.14. Redistribution subject to AIP license or copyright; see http://apl.aip.org/about/rights_and_permissions