Direct thermal neutron detection by the 2D semiconductor 6LiInP2Se6

Article abstract of DOI:10.1038/s41586-019-1886-8

Highly efficient neutron detectors are critical in many sectors, including national security^1,2, medicine³, crystallography⁴ and astronomy⁵. The main neutron detection technologies currently used involve ³He-gas-filled proportional counters⁶ and light scintillators⁷ for thermalized neutrons. Semiconductors could provide the next generation of neutron detectors because their advantages could make them competitive with or superior to existing detectors. In particular, solids with a high concentration of high-neutron-capture nuclides (such as ⁶Li, ¹⁰B) could be used to develop smaller detectors with high intrinsic efficiencies. However, no promising materials have been reported so far for the construction of direct-conversion semiconductor detectors. Here we report on the semiconductor LiInP₂Se₆ and demonstrate its potential as a candidate material for the direct detection of thermal neutrons at room temperature. This compound has a good thermal-neutron-capture cross-section, a suitable bandgap (2.06 electronvolts) and a favourable electronic band structure for efficient electron charge transport. We used α particles from an ²⁴¹Am source as a proxy for the neutron-capture reaction and determined that the compact two-dimensional (2D) LiInP₂Se₆ detectors resolved the full-energy peak with an energy resolution of 13.9 per cent. Direct neutron detection from a moderated Pu–Be source was achieved using ⁶Li-enriched (95 per cent) LiInP₂Se₆ detectors with full-peak resolution. We anticipate that these results will spark interest in this field and enable the replacement of ³He counters by semiconductor-based neutron detectors.

Full text of DOI:10.1038/s41586-019-1886-8

Article
Direct thermal neutron detection by the 2D
6
semiconductor LiInP Se
2
6
1
,5
1,5
1,2
3
3
https://doi.org/10.1038/s41586-019-1886-8
Received: 28 April 2019
Daniel G. Chica , Yihui He , Kyle M. McCall , Duck Young Chung , Rahmi O. Pak ,
1 2 4 2
Giancarlo Trimarchi , Zhifu Liu , Patrick M. De Lurgio , Bruce W. Wessels &
Mercouri G. Kanatzidis
1,3
*
Accepted: 8 October 2019
Published online: 15 January 2020
Highly efcient neutron detectors are critical in many sectors, including national
1,2
3
4
5
security , medicine , crystallography and astronomy . The main neutron detection
3
6
technologies currently used involve He-gas-ꢀlled proportional counters and light
7
scintillators ꢁor thermalized neutrons. Semiconductors could provide the next
generation oꢁ neutron detectors because their advantages could make them
competitive with or superior to existing detectors. In particular, solids with a high
6
10
concentration oꢁ high-neutron-capture nuclides (such as Li, B) could be used to
develop smaller detectors with high intrinsic efciencies. However, no promising
materials have been reported so ꢁar ꢁor the construction oꢁ direct-conversion
semiconductor detectors. Here we report on the semiconductor LiInP Se and
2
6
demonstrate its potential as a candidate material ꢁor the direct detection oꢁ thermal
neutrons at room temperature. This compound has a good thermal-neutron-capture
cross-section, a suitable bandgap (2.06 electronvolts) and a ꢁavourable electronic
band structure ꢁor efcient electron charge transport. We used α particles ꢁrom an
241
Am source as a proxy ꢁor the neutron-capture reaction and determined that the
compact two-dimensional (2D) LiInP Se detectors resolved the ꢁull-energy peak with
2
6
an energy resolution oꢁ 13.9 per cent. Direct neutron detection ꢁrom a moderated Pu–
6
Be source was achieved using Li-enriched (95 per cent) LiInP Se detectors with ꢁull-
2
6
peak resolution. We anticipate that these results will spark interest in this ꢀeld and
3
enable the replacement oꢁ He counters by semiconductor-based neutron detectors.
1
2
Direct neutron detection presents a tremendous challenge because direct-conversion detectors . The thermal-neutron detection eꢁꢁi-
8
neutrons interact weakly with most matter . Neutron detectors exploit ciency ꢁor indirect-conversion semiconductors reaches a maximum oꢁ
13
the properties oꢁ certain nuclides that decay into highly energetic about 40% . Direct-conversion semiconductors use a single material
charged ꢁragments upon capture oꢁ a neutron, providing easily detect ꢁor both neutron capture and charge collection, enabling a simpler
-
9
3
10
able signals through ionization products . In practice, only He, B and detector geometry, with intrinsic thermal-neutron detection eꢁꢁi-
Li combine high neutron-capture cross-sections with detectable decay ciencies approaching 100%. The semiconductors, however, must be
6
products, and these isotopes have been incorporated into various extremely pure and with low carrier trapping because neutron ꢁluxes
detector architectures. The detector oꢁ choice ꢁor several decades has can be very low, creating very small numbers oꢁ excited charges. Mate-
3
14–16
been the He-gas-ꢁilled proportional counter. However, in the past two rials used ꢁor direct conversion include LiInSe2
and hexagonal-BN
3
17
decades, the stockpile oꢁ He has been greatly depleted with no viable (h-BN) , both oꢁ which suꢁꢁer ꢁrom severe drawbacks that limit their
means ꢁor increasing production to meet demand . Thereꢁore, many perꢁormance. Despite these challenges, we demonstrate here the devel-
alternative technologies have attracted interest, including BF -ꢁilled opment and outstanding neutron detection capabilities oꢁ the layered
tubes, B-lined tubes, LiF-based scintillators, lithium-loaded plastic semiconductor LiInP Se .
6
10
3
10
6
2
6
7
6
10
scintillators , LiF-ꢁilled micro-structured semiconductor detectors
The selenophosphate compound LiInP Se is a 2D layered semicon-
2 6
and neutron-sensitive semiconductors. These technologies have advan- ductor that oꢁꢁers the requisite properties needed to achieve direct
1
1
tages and disadvantages , but so ꢁar no widespread replacement ꢁor thermal neutron detection. Bulk LiInP Se was synthesized phase pure
2
6
3
He tubes has been ꢁound.
(see Extended Data Fig. 1a–c) via a slightly oꢁꢁ-stoichiometric solid-state
6
10
Li- and B-containing semiconductors are emerging technolo- reaction at 750ꢂ°C (see Methods). LiInP Se crystallizes in the trigonal
2
6
gies that promise highly eꢁꢁicient detectors because the concentra- space group P31c with lattice parameters a = b = 6.3975(9) Å,
tion oꢁ neutron-absorbing isotopes is much greater in solids. There c = 13.351(3) Å (estimated standard deviations in parentheses),
are two classes oꢁ semiconductor neutron detectors: indirect- and α= β= 90° and γ= 120° (details oꢁ the reꢁinement and structure can be
1
2
3
Department of Chemistry, Northwestern University, Evanston, IL, USA. Department of Materials Science and Engineering, Northwestern University, Evanston, IL, USA. Materials Science
4
5
Division, Argonne National Laboratory, Argonne, IL, USA. Strategic Security Sciences Division, Argonne National Laboratory, Argonne, IL, USA. These authors contributed equally: Daniel G.
Chica, Yihui He. *e-mail: m-kanatzidis@northwestern.edu
346 | Nature | Vol 577 | 16 January 2020

a
b
a
b
Li
In
P
Se s
Se p
P s
b3
A
4
L
H
Se
12
Γ
2
b1
b2
P p
M
K
5
.328 Å
In s
In p
VBM
CBM
8
4
0
0
b
c
–
–
2
a
b
a
c
4Γ
M
K
Γ
A
L
H
A LM KH
c
d
–10
–5
0
5
Wavevector
E – EVBM (eV)
LiInP2Se6
295 K, 25 mW
33 nm Raman
c
d
In-Se-like
6
12.5 K, 2 mW PL
Cumulative ꢀt
6
5
4
3
2
1
0
0
0
0
0
0
0
Eg = 2.06 eV
P2Se6-like
20
1
1
0
3
0
2
P2Se6-like
1
2
3
Energy (eV)
1
00
200
300
400
)
500
–
1
Wavenumber (cm
500 750 1,000 1,250 1,500
Wavelength (nm)
1.4
1.6
1.8
2.0
2.2
Energy (eV)
Fig. 1 | Structural properties of LiInP
× 2 unit cell looking down the c axis (a) and a single unit cell looking down the
a–b plane, showing layered structure. c, Source side oꢁ the reaction vessel
containing plate-like crystals oꢁ CVT-grown LiInP Se . d, Raman spectrum oꢁ
LiInP Se acquired at room temperature, showing characteristic vibrational
modes associated with In–Se bonds and P Se units. a.u., arbitrary units.
2 6 2 6
Se . a, b, Crystal structure oꢁ LiInP Se : a
2
Fig. 2 | Band structure and optical properties of LiInP
a, b, Electronic band structure (a) and density oꢁ states (number oꢁ states per
unit cell volume per electronvolt) (b) ꢁor LiInP Se calculated using the
PBE+vdW ꢁunctional on a ꢁully relaxed structure. The inset in b shows the
Brillouin zone. CBM, conduction band minimum; VBM, valence band
2 6
Se single crystals.
2
6
2
6
2
6
2
6
maximum; E, energy; E
LiInP Se . The inset shows the corresponding Tauc plot. α, absorbance; E
bandgap energy. d, PL spectrum oꢁ LiInP Se obtained at 12.5 K and 2 mW, with
three Gaussian ꢁit peaks. The inset shows an image oꢁ the PL emission.
F
, Fermi energy. c, UV-Vis-NIR transmission spectrum oꢁ
2
6
g
,
2
6
ꢁ
ound in Extended Data Tables 1, 2). LiInP Se comprises polyhedral
2 6
layers separated by van der Waals gaps. Scanning electron microscopy
images show a clear layered nature ꢁor this compound (Extended Data
Fig. 2c–d). Each individual layer has the structure oꢁ CdI , where Li, In LiInP Se was at the source side. The crystals produced by the transport
2
2
6
and P dimers each occupy one-third oꢁ the octahedral Cd sites in an reactions without this ꢁirst step are smaller and oꢁten intergrown with
2
ordered manner and Se occupies the I site, as illustrated in Fig. 1a. The each other owing to multiple nucleations (Extended Data Fig. 1e). The
stacking oꢁ these layers ꢁollows an ABAB sequence such that the InSe6 reversal oꢁ the hot and cold sides at the end oꢁ the reaction limits the
polyhedra align whereas the LiSe and P Se polyhedra alternate along deposition oꢁ the gas phase species onto the as-grown crystal surꢁaces,
6
2
6
the c axis (Fig. 1b). This structure is typical among metal selenophos- which is necessary ꢁor high-quality, clean surꢁaces that permit good
+
3+
18,19
phates with the general ꢁormula M M P Se (M, metal; reꢁs.
) and electrical contact with the electrodes. The X-ray powder diꢁꢁraction
2
6
derives ꢁrom a larger layered ꢁamily based on divalent metal cations pattern oꢁ ground CVT-grown LiInP Se crystals matches the simulated
2
6
2+
2
20
M
P Se (reꢁ. ). Raman spectroscopy at an excitation oꢁ 633 nm shows pattern and exhibits identical thermal behaviour to the bulk material
2 6
several well-resolved peaks consistent with the vibrational modes oꢁ (Extended Data Figs. 1ꢁ, h, 2b), conꢁirming the purity oꢁ the transported
4−
the InSe octahedra and the [P Se ] unit seen in Fig. 1d (see Extended LiInP Se single crystals.
6
2
6
2
6
Data Fig. 3 ꢁor details).
Density ꢁunctional theory calculations oꢁ the electronic band struc-
The melting behaviour oꢁ LiInP Se was determined through diꢁ- ture were carried out using the PBE+vdW ꢁunctional, revealing LiInP Se
2
6
2
6
ꢁerential thermal analysis (DTA), giving melting and crystallization to be an indirect-gap semiconductor with a bandgap transition
points oꢁ 717ꢂ°C and 695ꢂ°C, respectively (Extended Data Fig. 2a). The ꢁrom near the Κ point at the valence band maximum to the Γ point at
compound melts congruently, as a single thermal event is observed the conduction band minimum ꢁor a bandgap oꢁ 0.94 eV (Fig. 2a). The
upon heating or cooling, and the powder diꢁꢁraction pattern aꢁter DTA conduction band is composed oꢁ hybridized In sand Se pstates associ
-
shows only LiInP Se present (Extended Data Fig. 1g) . ated with the empty In 5s lone pair, whereas the valence band is made
2
6
We then developed a more eꢁꢁective approach to grow detector- up oꢁ Se p states, as seen in Fig. 2b. The conduction band has a high
⁎
e
grade, large-size LiInP Se single crystals using the chemical vapour curvature, corresponding to a low eꢁꢁective electron mass (
m
) oꢁ 0.16me
2
6
transport (CVT) method. The temperature oꢁ the hot zone (660ꢂ°C) and 0.30m ꢁor the in- and out-oꢁ-plane directions, respectively (m ,
e
e
was below the melting point because otherwise the atmosphere would electron mass). The valence band maximum lies along the out-oꢁ-plane
saturate with volatile components and inhibit the transport oꢁ the other direction in reciprocal space (K to H), and the band in this direction is
constituents. The cold zone was set to 560ꢂ°C, creating a driving ꢁorce very ꢁlat, so the hole eꢁꢁective mass in the out-oꢁ-plane direction will
ꢁor the transport oꢁ material ꢁrom the hot to the cold zone. The initial be immense. Accordingly, we expect that the electron mobility should
transport reactions oꢁ LiInP Se were carried out without a transport- be vastly superior in LiInP Se , especially given that the planar nature
2
6
2
6
ing agent and produced tiny LiInP Se crystals with a low yield (about oꢁ the material ꢁorces devices to operate in the out-oꢁ-plane direction.
2
6
1
mg out oꢁ ~0.7 g oꢁ starting material) even aꢁter a ꢁew days oꢁ transport
Ultraviolet–visible–near inꢁrared (UV-Vis-NIR) transmission and
owing to decomposition into amorphous P Se (Extended Data Fig. 1d). reꢁlection measurements on a LiInP Se single crystal were used to
x
y
2
6
This issue was resolved by the use oꢁ iodine as a transporting agent, measure the bandgap, which was determined to be 2.06 eV (Fig. 2c).
as the ꢁormation oꢁ InI and LiI enables the transport oꢁ each metal Photoluminescence (PL) measurements on LiInP Se at low temperature
3
2
6
2
1
component to balance the vapour transport . Iodine-assisted CVT (12.5 K) under an excitation oꢁ 405 nm showed red–orange emission ꢁrom
enabled the growth oꢁ large single LiInP Se crystals with an area oꢁ the as-grown surꢁace oꢁ the single crystal, with two broad emission bands
about 1 cm and thickness oꢁ ~0.05–1 mm ꢁor the week-long reactions observed with peak maxima near 1.73 eV and 2.05 eV (Fig. 2d). Excitation
in Fig. 1c. To reduce the number oꢁ nucleation sites and improve the intensity and temperature-dependent PL measurements support the
yield oꢁ large crystals, we reversed the zones ꢁor 24 h to ensure that all assignment oꢁ the 1.73 eV peak to donor–acceptor pair recombination
2
6
2
Nature | Vol 577 | 16 January 2020 | 347

Article
a
105
b
a
b
8
0
7
00 V
Unshielded strong source
Cd-shielded strong source
Unshielded weak source
04
800
1
1
1
1
1
600 V
03
60
40
20
0
6
4
2
00
00
00
5
00 V
00 V
02
01
Photocurrent
Dark current
4
0.95 s–1 cm–2
300 V
200 V
0
.20 s–1 cm–2
00
0
.12 s–1 cm–2
1
00 V
12,000
Channel number
1
1
0–
–
100 –50
0
50
100
0
6,000
1 inch
Voltage (V)
0
32
64
96
128
Channel number
c
104
d
50
7
00 V
700 V
6
Fig. 4 | Thermal-neutron spectra resolved by LiInP
2
Se
6
devices. a, Binned
2
1
nA
03
02
01
00
241Am
6
1
1
1
1
pulse-height spectrum oꢁ a LiInP Se device (shown in b) exposed to a
2
6
nA
0
moderated Pu–Be source without shielding (red) and with Cd shielding
(0.125 inch thick; black) and to a source ~5 times weaker (grey) ꢁor 300 V bias,
2
5
3
μs shaping time and 30-min measurements. The counts ꢁrom channel 16 and
–
1 kV
0
1 kV
above correspond to neutron-capture events. The count rate ꢁor the
Background
−1
−2
2 6
5
7
2 6
unshielded measurement was 0.95 s cm . b, The 7 × 7 mm LiInP Se device
Co
(90 μm thickness) used ꢁor the measurements shown in a.
0
0
300 600 900 1,200
Channel number
0
6,000
12,000
Channel number
can be seen in Fig. 3b. Here the channel number reꢁlects the charge
collection eꢁꢁiciency (CCE), which is the ratio oꢁ collected and gener-
ated charge ꢁrom a single event (ꢁor example, neutron capture), and
under constant bias can be considered proportional to the energy oꢁ
the incident particle. The peak ꢁrom the interaction oꢁ α particles shiꢁts
to higher channel numbers with increasing voltage (Extended Data
Fig. 5), as higher CCE can be achieved with higher applied electric ꢁield.
The energy resolution ꢁor the ꢁull-energy peak at 700 V is 13.9% ꢁor the
LiInP Se detector, conꢁirming that LiInP Se can accurately resolve the
Fig. 3 | Electrical characteristics and pulse-height spectra of LiInP
devices irradiated by α-particle and γ-ray sources. a, Dark current and
photocurrent (absolute values) as a ꢁunction oꢁ voltage (ꢁrom −100 V to 100 V)
or CVT-grown LiInP Se (~0.3 mm thickness) with gold electrodes illuminated
by ambient light (~0.2 mW cm ). b, Pulse-height spectra oꢁ α particles ꢁrom
2
Se
6
ꢁ
2
6
−2
2
41
Am ꢁor a LiInP
various voltages, collected ꢁor 60 s each. c, Pulse-height spectrum oꢁ a
LiInP Se detector (~0.5 mm thickness) under a bias oꢁ 700 V illuminated by
γ-rays ꢁrom Am and Co, and without a source. The counts ꢁor Co γ-rays and
2 6
Se device (~120 μm thickness) ꢁor electron collection at
2
6
2
41
57
57
2
6
2
6
ꢁor the natural background are negligible, as shown. The spectra were collected
charged particles generated by neutron interaction. To our knowledge,
ꢁor 200 s each. d, Pulse-height spectrum at 700 V with the same conditions as in
241
this is the highest energy resolution reported ꢁor Am α particles
b. Inset, Dark current measured with the voltage varying ꢁrom 0 to 1 kV, to −1 kV,
and to 0 V under the same conditions as in a. The contact area on the crystals in
ꢁ
or any direct-conversion thermal-neutron semiconductor detector,
15
surpassing the 23.3% resolution oꢁ LiInSe or the non-resolution oꢁ
2
2
a–d was 3 × 3 mm . The shaping time in b–d was 1 μs.
the ꢁull-energy peak in h-BN due to poor CCE.
These results are reproducible ꢁor diꢁꢁerent growth sizes, and the
and oꢁ the 2.05 eV peak to a ꢁree-to-bound transition; a detailed discus- α-particle response oꢁ other LiInP Se devices with crystals grown under
2
6
sion may be ꢁound in Methods section ‘Notes on Extended Data Fig. 3’. identical conditions can be seen in Extended Data Fig. 6a–h. For hole
Despite these deꢁect-related peaks, a strong PL signal ꢁrom an indirect- collection in a typical sample, a shoulder, instead oꢁ a ꢁull-energy peak,
gap material is an indication oꢁ good optical quality.
is usually observed, implying relatively poor CCE oꢁ holes. Nevertheless,
241
LiInP Se detector devices were ꢁabricated ꢁrom large CVT-grown several devices were able to resolve the Am peak ꢁor both holes and
2
6
single crystals using evaporated planar gold contacts. Pristine LiInP Se6 electrons. This is because the high electric ꢁield that this material can
2
−1
devices exhibit very low dark current, oꢁ the order oꢁ picoamperes withstand (at least 60 kV cm ) allows the generated carriers to be swept
(
1
Fig. 3a), corresponding to a high electrical resistivity oꢁ around out to the contacts beꢁore trapping or recombination can occur, result-
0 Ω cm. The dark-current plot shows a non-zero current at 0 V, ing in a spectroscopic response even ꢁor the relatively low-mobility
1
3
though the dark current exhibits linearity in this range (Extended Data holes. For a constant applied voltage across various samples, the CCE
Fig. 4a). When larger voltages are applied a diode-like current–voltage is higher ꢁor thinner samples because oꢁ the larger external electric
curve is observed, whereas the low dark current is maintained at 2 nA ꢁield. These results conclusively demonstrate that LiInP Se devices can
2
41
6
2
at 1,000 V (Fig. 3d), which is more than suꢁꢁicient ꢁor eꢁꢁective α- and achieve a high-resolution spectroscopic response to Am α particles.
6
neutron-detection measurements. This behaviour might originate ꢁrom
Isotopically enriched lithium (95% Li) was used to grow LiInP Se
2
6
6
6
blocked/injected current ꢁrom the electrodes under a large electric crystals (hereaꢁter labelled LiInP Se ) with maximal Li to maximize
2
6
ꢁ
ield oꢁ up to tens oꢁ kilovolts per centimetre. When exposed to ambient the thermal-neutron-capture cross-section. The α-particle response oꢁ
−2
6
light (~0.2 mW cm ), photogenerated charge carriers in the LiInP Se
detector increase the current by three to ꢁour orders oꢁ magnitude with ꢁull resolution oꢁ the Am peak ꢁor electron collection (Extended
(
the LiInP Se devices is similar to that oꢁ their unenriched analogues,
2
6
2
6
241
6
Fig. 3a). This is an exceptional photoresponse that indicates ꢁavour- Data Fig. 7a–e). Hole collection oꢁ LiInP Se devices also demonstrates
2
6
able charge transport properties in LiInP Se , as the photogenerated some level oꢁ α-particle response.
2
6
6
carriers eꢁꢁiciently traverse the detector thickness.
The thermal-neutron response oꢁ the 95% Li-enriched detectors was
As a direct-conversion semiconductor based on lithium, neutrons tested at room temperature using a moderated and very weak Pu–Be
(
n) are captured in LiInP Se through the ꢁollowing nuclear reaction: source, which generated roughly 75 neutrons per second per square
2
6
6
4
3
n+ Li = He + H (energy Q= 4.78 MeV). Thus, the interaction oꢁ α parti
cles produced by the decay oꢁ Am (Q= 5.486 MeV) mimics the energy sus the thickness oꢁ LiInP Se is shown in Extended Data Fig. 8. The
deposition oꢁ a neutron-capture event at the detector surꢁace.
-
centimetre. The absorption probability ꢁor thermalized neutrons ver-
241
6
2
6
2
device had an active area oꢁ 7 × 7 mm and a thickness oꢁ ~90 μm, and
The pulse-height spectrum oꢁ a LiInP Se device with a thickness oꢁ was subjected to a bias oꢁ 300 V. The binned pulse-height spectrum
2
6
241
6
1
20 μm irradiated by 5.486 MeV Am α particles ꢁor electron collection oꢁ a LiInP Se device under neutron irradiation shows a deꢁined peak
2 6
348 | Nature | Vol 577 | 16 January 2020

1
2
.
.
Detection of Radioactive Materials at Borders (International Atomic Energy Agency, 2002).
Runkle, R. C., Bernstein, A. & Vanier, P. E. Securing special nuclear material: recent
advances in neutron detection and their role in nonproliferation. J. Appl. Phys. 108, 111101
(2010).
Barth, R. F., Coderre, J. A., Vicente, M. G. H. & Blue, T. E. Boron neutron capture therapy of
cancer: current status and future prospects. Clin. Cancer Res. 11, 3987–4002 (2005).
Wollan, E. O. & Shull, C. G. The diffraction of neutrons by crystalline powders. Phys. Rev.
73, 830–841 (1948).
that is clearly resolved ꢁrom the background, demonstrating success-
ul direct detection oꢁ neutrons (Fig. 4a). The pulse-height spectrum
ꢁ
shows a broad ꢁull-energy peak corresponding to the neutron-capture
reaction close to the dark-current noise background edge. When cad-
mium shielding (0.125-inch-thick; 1 inch = 2.54 cm) was placed in ꢁront
oꢁ the device to block most oꢁ the thermalized neutrons, the count rate
was reduced by 87%. This reduction stems ꢁrom the high absorption oꢁ
3
.
4
.
5
.
Ruffolo, D. et al. Monitoring short-term cosmic-ray spectral variations using neutron
monitor time-delay measurements. Astrophys. J. 817, 38–50 (2016).
22
neutrons below the cadmium cutoꢁꢁ energy (0.5 eV; reꢁ. ). The energy
resolution oꢁ this neutron-capture peak was 69.7%. This response is
very promising at this early stage.
3
6
.
Kouzes, R. T. The He Supply Problem. Report No. PNNL-18388 (Paciꢀic Northwest National
Laboratory, 2009).
7
.
Breukers, R. D., Bartle, C. M. & Edgar, A. Transparent lithium loaded plastic scintillators for
thermal neutron detection. Nucl. Instrum. Methods Phys. Res. A 701, 58–61 (2013).
Dunning, J. R., Pegram, G. B., Fink, G. A. & Mitchell, D. P. Interaction of neutrons with
matter. Phys. Rev. 48, 265–280 (1935).
An ideal neutron detector should be insensitive to γ-rays to ensure
that ꢁalse neutron counts are not generated ꢁrom incident γ radiation.
The ability oꢁ the LiInP Se device to discriminate against γ-rays was
8
.
9
.
Knoll, G. F. in Radiation Detection and Measurement 505–535 (John Wiley & Sons, 2000).
2
6
10. Ochs, T. R. et al. Present status of the microstructured semiconductor neutron detector-
57
evaluated on a thicker detector (~0.5 mm) exposed to a 0.2-mCi Co
γ-radiation source (a photon ꢁlux multiple orders oꢁ magnitude higher
than the neutron source discussed earlier). Despite the higher ꢁlux, the
pulse-height spectra oꢁ the ambient background and under direct γ
irradiation shown in Fig. 3c are essentially identical, demonstrating
the negligible γ sensitivity oꢁ this material.
based direct helium-3 replacement. IEEE Trans. Nucl. Sci. 64, 1846–1850 (2017).
11. Simpson, A. P., Jones, S., Clapham, M. J. & Mcelhaney, S. A. A review of neutron detection
technology alternatives to helium-3 for safeguards applications. In INMM 52nd Annual
Meeting Vol. 8 (2011).
2. Caruso, A. N. The physics of solid-state neutron detector materials and geometries.
J. Phys. Condens. Matter 22, 443201 (2010).
3. Shultis, J. K. & Mcgregor, D. S. Design and performance considerations for perforated
semiconductor thermal-neutron detectors. Nucl. Instrum. Methods Phys. Res. A 606,
608–636 (2009).
14. Tupitsyn, E. et al. Lithium containing chalcogenide single crystals for neutron detection.
J. Cryst. Growth 393, 23–27 (2014).
5. Guo, L. et al. Stoichiometric effects on the photoelectric properties of LiInSe crystals for
2
neutron detection. Cryst. Growth Des. 18, 2864–2870 (2018).
16. Hamm, D. S. et al. Semiconducting lithium indium diselenide: charge-carrier properties
1
1
Inconclusion, LiInP Se istheꢁirst, toourknowledge, directthermal-neu-
2
6
6
tron semiconductor detector with intrinsic Li to demonstrate a resolved
ull-energy neutron peak above the background. Further improvements
ꢁ
1
in the neutron detection eꢁꢁiciency could be achieved by the growth oꢁ
larger single crystals, such as through the Bridgman growth method. Fur-
241
and the impacts of high ꢀlux thermal neutron irradiation. Appl. Phys. Lett. 112, 242104
thermore, spectroscopic Am α-particle detection is achieved with 13.9%
energy resolution. The excellent semiconducting properties and radiation
response achieved here demonstrate that LiInP Se has the potential to
(2018).
1
7. Maity, A., Grenadier, S. J., Li, J., Lin, J. Y. & Jiang, H. X. High-efꢀiciency and high-sensitivity
thermal neutron detectors based on hexagonal BN epilayers. In Proc. SPIE Hard X-Ray,
Gamma-Ray, and Neutron Detector Physics XIX, 10392 (2017).
2
6
transꢁorm neutron detection technology, which relies almost exclusively
on scintillation paradigms, and usher in the semiconductor modality ꢁor
direct-conversion thermal-neutron detection.
I
III
I
III
1
8. Pfeiff, R. & Kniep, R. Quaternary selenodiphosphates(IV): M M [P
2
Se
6
], (M = Cu, Ag; M
=
Cr, Al, Ga. J. Alloys Compd. 186, 111–133 (1992).
19. Seidlmayer, S. Strukturchemische Untersuchungen an Hexachalkogenohypodiphosphaten
und Verwandten Verbindungen. PhD Thesis, University of Regensburg (2009).
2
0. Wang, F. et al. New frontiers on van der Waals layered metal phosphorous
trichalcogenides. Adv. Funct. Mater. 28, 1–24 (2018).
Online content
21. Binnewies, M., Glaum, R., Schmidt, M. & Schmidt, P. in Chemical Vapor Transport
Reactions 360–369 (De Gruyter, 2012).
Any methods, additional reꢁerences, Nature Research reporting sum-
maries, source data, extended data, supplementary inꢁormation,
acknowledgements, peer review inꢁormation; details oꢁ author con-
tributions and competing interests; and statements oꢁ data and code
availability are available at https://doi.org/10.1038/s41586-019-1886-8.
22. Knoll, G. F. Radiation Detection and Measurement 757–775 (John Wiley & Sons, 2000).
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in
published maps and institutional afꢀiliations.
© The Author(s), under exclusive licence to Springer Nature Limited 2020
Nature | Vol 577 | 16 January 2020 | 349

Article
Methods
CVT
Chemical reagents
Preꢁormed material ꢁrom the bulk synthesis was used as the source
Lithium (99.9% Sigma Aldrich), indium (99.99% American Elements), ꢁor the CVT growth oꢁ LiInP Se . In a typical synthesis, approximately
2
6
phosphorus (99.999% Alꢁa Aesar), selenium (99.99% American Ele- 0.5–5 g oꢁ source material was used, accompanied by 16–30 mg oꢁ I —the
2
ments) and iodine (99.999% Sigma Aldrich) were used as received. amount oꢁ I used depends on the inner diameter oꢁ the ꢁused-silica
2
For the material used in the detectors, high-purity indium (99.9999% tube (11–16 mm). The tube was sealed to a length oꢁ 17 cm while the
Alꢁa Aesar) and selenium (99.999% Alꢁa Aesar) were used. Isotopically bottom was submerged in liquid nitrogen to prevent loss oꢁ iodine.
6
enriched lithium (95% Li) was used to synthesize materials ꢁor the The tube was then placed in a two-zone ꢁurnace and subjected to the
neutron detection measurements.
ꢁollowing heating proꢁile: source zone: heat to 560ꢂ°C over 6 h, hold
the temperature ꢁor 12 h, heat to 660ꢂ°C over 6 h, hold ꢁor 168 h and
then cool to ambient temperature over 6 h; sink zone: heat to 660ꢂ°C
Synthesis of Li Se
2
2
3,24
Li Se was synthesized using a modiꢁied literature procedure . In an over 6 h, hold ꢁor 12 h, heat to 560ꢂ°C over 6 h, hold ꢁor 170 h and then
2
argon-ꢁilled glovebox, lithium metal was removed ꢁrom mineral oil cool to ambient temperature over 10 h. Aꢁter the transport process,
and the oxidized edges were cut oꢁꢁ. The lithium metal, selenium and a deep-red plates were ꢁound on the sink side, and the source material
stir bar were added into a 500-mL three-neck ꢁlask. The reaction vessel was consumed. The vacuum oꢁ the tube was careꢁully breached by
was attached to a custom apparatus that allowed an ammonia stream slowly cutting into the tube at the source side with a diamond saw to
to condense liquid ammonia into the vessel under ꢁlowing argon. In make a pinhole opening. This step was necessary to prevent damage
a typical reaction, ~300 mL oꢁ ammonia was condensed using a cold to the thinner crystals ꢁrom the inꢁlux oꢁ air.
ꢁ
inger containing an acetone–dry-ice mixture. The reaction continued
under reꢁlux and stirring until the ammonia colour changed ꢁrom deep X-ray diffraction
blue to clear. Aꢁter the completion oꢁ the reaction, the liquid ammonia A Rigaku Miniꢁlex600 X-ray diꢁꢁraction system equipped with a Dtex
was boiled oꢁꢁ and the vessel material was placed under vacuum over- silicon 1-D detector was used to obtain powder X-ray diꢁꢁraction pat-
night. The material was stored in a N glovebox. In a typical reaction, terns. Cu Kα radiation (wavelength λ = 1.5406 Å) was produced using
2
1
5 g oꢁ material was synthesized with a Li:Se ratio oꢁ 2:1.002. The slight a voltage oꢁ 40 kV and a current oꢁ 15 mA and ꢁiltered with a graphite
excess oꢁ selenium was to ensure that the reaction goes to completion. monochromator and a Kβ ꢁoil ꢁilter. A zero-background silicon holder
was used. Crystals suitable ꢁor single-crystal diꢁꢁraction were obtained
Synthesis of Li1.03In
using CVT reactions. Single-crystal X-ray diꢁꢁraction was perꢁormed
In an argon-ꢁilled glovebox, In (2.535 g) and Li (0.147 g) chunks were using a STOE IPDS II diꢁꢁractometer operating at 50 kV and 40 mA with
loaded into a glassy carbon crucible with a lid. This crucible was then Mo Kα radiation (λ= 0.71073 Å) with a graphite monochromatizer. Data
placed in a ꢁused quartz silica tube and the tube was ꢁlame-sealed under reduction, integration and absorption correction were perꢁormed
−3
a pressure oꢁ ~2 × 10 mbar. The tube was placed in a vertical ꢁurnace using the X-area soꢁtware package. The program XPREP was used to
and subjected to the ꢁollowing heating proꢁile: heat to 700ꢂ°C over determine the space group and prepare the data ꢁor the structure solu-
1
0 h, soak ꢁor 2 h, and then cool to ambient temperature. The ingot tion. Structure solutions using intrinsic phasing and reꢁinements were
2
5
26
was extracted and stored in a nitrogen box.
obtained using SHELXT and SHELXL , respectively. OLEX2 was used
27
28
as a GUI . VESTA was used to produce crystal structure ꢁigures .
Synthesis of P Se
2
5
For P Se , red phosphorus pieces (1.356 g) and selenium shot (8.644 g) Optical property characterization
2
5
were loaded into a 13-mm ꢁused-silica tube in a stoichiometric ratio. UV-Vis-NIR transmission and reꢁlection were measured on a PerkinElmer
−3
The tube was ꢁlame-sealed under vacuum (~2 × 10 mbar). The reagents LAMBDA 1050 UV/Vis spectrophotometer in the range 1,500–250 nm
were physically mixed by shaking the tube. The tubes were placed in on a single crystal oꢁ LiInP Se at room temperature. These were used
2
6
a tilted tube ꢁurnace so that the tops oꢁ the tubes were at the centre to obtain the absorption and thereby determine the bandgap using a
oꢁ the ꢁurnace. The ꢁollowing temperature proꢁile was used: heat to linear ꢁit oꢁ the Tauc plot. Raman spectroscopy was conducted on a
5
00ꢂ°C over 12 h, hold the temperature ꢁor 72 h and then ꢁurnace-cool single crystal oꢁ LiInP Se at room temperature using a conꢁocal Horiba
2 6
to ambient temperature. The tubes were opened in a nitrogen-ꢁilled LabRAM HR Evolution spectrometer with an excitation wavelength oꢁ
glovebox ꢁor storage.
633 nm and a laser power oꢁ 25 mW.
Room-temperature PL was measured in the range 475–1,000 nm
with the same spectrometer using an above-bandgap excitation wave-
Synthesis of bulk LiInP Se6
2
Owing to the air-sensitive nature oꢁ Li Se, the reagents were manipu- length oꢁ 473 nm. The laser power (25 mW) was varied in the range
2
lated in a nitrogen-ꢁilled glovebox. For the bulk synthesis oꢁ LiInP Se , 0.25–12.5 mW using neutral density ꢁilters. Laser powers above 1.25 mW
2
6
Li Se (0.437 g), In (1.030 g), P Se (4.181 g) and Se (0.351 g) were loaded led to visible damage to the sample, and below this threshold no PL
2
2
5
into a carbon-coated, 13-mm ꢁused-silica tube at a molar ratio oꢁ was observed. Above this point, a broad band appears ꢁrom 1.4 eV to
Li:In:P:Se = 1.05:1:2.04:6.12. Alternatively, Li1.03In (2.500 g), P Se (9.643 g) 2 eV, probably due to an impurity phase ꢁormed upon decomposition
2
5
and Se (1.667 g) could be used as the starting reagents, and were used ꢁor oꢁ the sample.
the synthesis oꢁ detector-grade crystals. Synthesis oꢁ chalcogenides con-
The temperature and excitation intensity dependence oꢁ the PL were
taining Li typically require crucibles that can withstand chemical attack measured on a LiInP Se6 single crystal in a closed-cycle He cryostat
2
14
ꢁ
rom Li, such as glassy carbon or pyrolytic BN . In the present synthesis, (SHI Cryogenics DE-202). The single crystal was mounted on the stage
the tubes were carbon-coated to prevent glass attack. The tubes were using Apiezon N grease. The excitation source was a continuous-wave
−3
ꢁ
lame-sealed under vacuum (~2 × 10 mbar). The tubes and thermocouple diode laser (OBIS 405 nm LX; beam diameter 0.8 mm) with the power
were placed in the same manner as in the reaction oꢁ P Se . The ꢁollowing digitally adjusted in the range oꢁ 0.5–21 mW. A 405-nm bandpass ꢁilter
2
5
optimized heating proꢁile was used: heat to 750ꢂ°C over 10 h, hold the (ꢁull-width at halꢁ-maximum 10 nm; Thorlabs) screened the laser light
temperature ꢁor 24 h, cool to 350ꢂ°C over 12 h and then ꢁurnace-cool to ꢁocused onto the sample, and the emitted light was ꢁiltered by a 450-nm
ambient temperature. The tubes were opened in air, revealing an ingot long-pass ꢁilter (Thorlabs) beꢁore reaching the monochromator
6
comprised oꢁ deep red plates. The synthesis oꢁ bulk LiInP Se was per- entrance slit (width 200 μm). A chopper was used at 710 Hz to increase
2
6
6
ꢁ
ormed using Li1 In, P Se and Se as the starting reagents.
the signal-to-noise ratio. PL spectra were resolved by a 500M SPEX
.03
2
5

grating monochromator with a Hamamatsu R928 photomultiplier respectively. A dual 16K-input multichannel analyser (model ASPEC-
tube biased at 500 V, which was read into a SR810 lock-in ampliꢁier 927) and MAESTRO-32 soꢁtware were used to process the ampliꢁied
(
Stanꢁord Research Systems). PL spectra were collected over the range signal and display the response spectrum, respectively. The Hecht
−1
oꢁ 530–900 nm at a rate oꢁ 1 nm s .
equation ꢁor a single carrier type is a common way to estimate the
2
3
4
(μτ)eV 
−d


mobility–liꢁetime product μτ (reꢁ. ) as CCE =
1 − exp(
)
,
2

(μτ)eV 
DTA
d

For the determination oꢁ the thermal properties oꢁ LiInP Se , a Netzch where dis the thickness oꢁ the semiconducting region through which
2
6
Simultaneous Thermal Analysis (STA 449F3) instrument was employed. the charge carrier travels and Vis the applied voltage. The charge col-
Approximately 25 mg oꢁ the sample was sealed in a ꢁused-silica ampoule. lection eꢁꢁiciency (CCE) reꢁers to the ꢁraction oꢁ charge that is collected
An alumina ampoule oꢁ similar mass was used as a reꢁerence. The heat- versus the generated charge.
−1
ing and cooling rates were 10 K min .
Neutron detection
Band structure
The same setup used to measure the pulse-height spectrum oꢁ α
The band structure calculations were perꢁormed using VASP (Vienna ab particles was used to determine the neutron response. The source
3
initio simulation package) within the ꢁramework oꢁ density ꢁunctional oꢁ neutrons used was a polyethylene-moderated PuBe source. A He
theory. The symmetry-unconstrained lattice parameters and atom proportional tube was used to determine the neutron ꢁlux oꢁ the source
coordinates were relaxed to the local total energy minimum, starting so that the absolute eꢁꢁiciency oꢁ the devices could be calculated. The
6
ꢁ
rom the structure that we measured experimentally. The structure
LiInP Se device was placed at the same distance away ꢁrom the source
2 6
3
optimization was perꢁormed by applying the Perdew–Burke–Ernzerhoꢁ as the He detector. Collection times were 30 min. The number oꢁ chan-
exchange and correlation ꢁunctional with the van der Waals extension nels was reduced ꢁrom 16,384 to 128 by summing the counts ꢁrom 128
PBE+vdW) using the optB88-vdW ꢁunctional . The electronic-band channels into a single channel.
29
(
dispersions were calculated ꢁor the ꢁully relaxed structures using the
special k points and the high-symmetry path in the Brillouin zone
Notes on Extended Data Fig. 1
30
deꢁined by Setyawan and Curtarolo . The analysis oꢁ the data generated Note for bulk synthesis and chemical stability. The optimized synthe
by the ab initio calculations was carried out with Python scripts with sis oꢁ LiInP Se involves a relatively small excess oꢁ Li/P/Se to synthesize
extensive use oꢁ the Atomic Simulation Environment (ASE) and the phase-pure material. When a stoichiometric ratio oꢁ reagents was used,
-
2
6
3
1
32
Python Materials Genomics (pymatgen) materials analysis libraries
.
secondary phases were present (see Extended Data Fig. 1c). The need
The eꢁꢁective masses were calculated using the set oꢁ python scripts ꢁor additional P and Se may stem ꢁrom the imbalanced vapour pressure
33
available at the emc repository .
oꢁ diꢁꢁerent vapour components, which probably induces a P/Se-rich
atmosphere, leaving the melt deꢁicient in P/Se. For a ꢁew reactions,
even when the carbon coat peeled oꢁꢁ the silica wall, there were no
Scanning electron microscopy
A Hitachi 3400 scanning electron microscope was used to image the signs oꢁ glass attack, suggesting that the as-ꢁormed LiInP
microstructure. The accelerating voltage and probe current were set to does not react with silica. Bulk LiInP Se shows good stability under
0 keV and 70 mA, respectively. Energy-dispersive X-ray spectroscopy ambient atmospheric conditions. However, when the ingot is ꢁinely
EDS) was perꢁormed using a PGT energy-dispersive X-ray analyser. ground, the resulting powder turns ꢁrom bright orange to dark brown
Aztec soꢁtware ꢁrom Oxꢁord Instruments was used to analyse the EDS aꢁter several minutes oꢁ air exposure. Both the ingot and powder were
Se material
2
6
2
6
2
(
data.
very sensitive to liquid water and appeared to exꢁoliate aꢁter several
seconds in water, potentially ꢁrom water intercalating into the layers.
Device fabrication
Planar-type devices were ꢁabricated using single crystals grown by CVT Gas phase equilibrium in the CVT reaction. Because this vapour
without cutting or polishing, owing to the high surꢁace quality oꢁ the transport experiment involves a quinary system when considering the
as-grown crystals. Translucent and ꢁlat crystals that were visibly deꢁect- iodine, the gas–solid phase equilibrium is quite complex. For a given
2
ꢁ
(
ree were selected. Gold electrodes with a typical area oꢁ 3 × 3 mm
material to transport, every element present in the material must have
unless otherwise stated) and thickness oꢁ ~70 nm were deposited on some species that ꢁavours the gas phase. Typically, metal selenides can
−1
both sides oꢁ the crystal by thermal evaporation at a rate oꢁ ~1 Å s using be transported using iodine through the ꢁormation oꢁ the gas phase spe-
n+
21
an SQC-310 Inꢁicon deposition controller. Cu wires were attached to cies, M I
, and Se
dimers . Phosphorus is relatively volatile and trans-
n
2
35
the gold electrodes using graphite paste, allowing the crystals to be ports well through the ꢁormation oꢁ either P tetramers or PI (reꢁ. ).
4
3
3
6
connected to the external circuit used ꢁor property measurements.
By contrast, the transport oꢁ alkali metals is more diꢁꢁicult as alkali
metal compounds typically have low vapour pressures at moderate
Electrical properties and α-particle response
temperatures and thus transport poorly. Li is probably transported
o
The electrical resistivity was measured using a Keithley 6517B elec- as LiI, given its relatively low melting point oꢁ 469 C and boiling point
o
trometer. For dark-current measurements, the device was enclosed in oꢁ 1,171 C. For crystals grown without employing the reversal oꢁ zones
a light-tight metal box. The photoconductivity was measured using at the end oꢁ the growth, EDS mapping (Extended Data Fig. 2e) oꢁ the
−2
ambient light (0.2 mW cm 400–800 nm) as the source oꢁ excitation. surꢁace demonstrated a growth ꢁront rich in iodine, demonstrating the
,
Current–voltage characteristic curves ꢁrom 100 V to −100 V were used involvement oꢁ iodine in the transport oꢁ LiInP
to calculate the electrical resistivity. To enable collection oꢁ the
Se6.
2
Notes on Extended Data Fig. 3
α-particle spectrum, planar-type detectors with evaporated gold con-
tacts were used ꢁor pulse-height spectroscopy. The α-particle source Room-temperature PL measurements on a LiInP
Se
employed was Am, with a typical α-particle kinetic energy oꢁ showed no response to excitation at 473 nm except upon decomposi-
.486 MeV. The detectors were enclosed in a light-tight metal box with tion at high laser powers, when a broad band centred at 1.7 eV appears
single crystal
2
6
2
41
5
the source ~3 mm away ꢁrom the top electrode. The device was con- (Extended Data Fig. 3). This lack oꢁ PL response may be due to the indi-
nected to an eV-550 preampliꢁier and a voltage source that applied a rect gap oꢁ the compound, which reduces the probability oꢁ radiative
voltage oꢁ ±20–1,000 V to the bottom electrode. An ORTEC ampliꢁier recombination due to the momentum diꢁꢁerence between the band
(
model 572A) was used to ampliꢁy and shape the signal ꢁrom the pre- edge states, requiring phonons to mediate the transition. Thus, low-
ampliꢁier. The gain and shaping time were set to 100–200 and 0.5–2 μs, temperature PL was used to determine the optical quality oꢁ these

Article
crystals and investigate whether any deꢁects were present. As observed levels in the middle oꢁ the bandgap) to the ‘shallower’ peaks 2 and 3 that
in Fig. 2d, LiInP Se exhibits two broad PL peaks at low temperature emit ꢁrom energy levels closer to the band edge. DAP recombination
2
6
(12.5 K), which can be modelled by three Gaussian peaks. A two-peak ꢁit occurs between a donor and an acceptor band, both oꢁ which must lie
was attempted, but it was ꢁound to require the use oꢁ an exponentially deep enough within the bandgap to avoid being ionized, whereas a ꢁree-
37
modiꢁied Gaussian peak as described in reꢁ. to adequately model the to-bound transition is between a single deꢁect band and the opposite
asymmetry oꢁ the higher energy band. However, the use oꢁ such a peak band edge, so it is only as deep as the deꢁect band. At 1.73 eV, peak 1 lies
is not justiꢁied here because the source oꢁ emission is not the result oꢁ a well below the band edge (2.06 eV at room temperature) while peaks 2
distribution oꢁ phonon-related peaks. Hence, these emission bands are and 3 are much closer to the bandgap at 1.98 eV and 2.06 eV, respectively.
best ꢁitted using three Gaussian peaks, where peaks 2 and 3 correspond Together with the diꢁꢁerence in temperature-dependent behaviour,
to the same broad band so they are expected to show similar behaviour. this supports the assignment oꢁ peak 1 to DAP recombination and oꢁ
To ꢁurther characterize the PL emission at 12.5 K, the PL dependence peaks 2 and 3 to a ꢁree-to-bound transition. The PL measurements are
on the excitation intensity was tested by increasing the laser power summarized in Extended Data Fig. 3ꢁ, along with these tentative peak
ꢁ
rom 0.5 mW to 21 mW, resulting in corresponding enhancements in assignments. These results reveal the presence oꢁ at least two deꢁect lev-
the emission intensity and a slight blueshiꢁt in the peak maxima ꢁor all els, which should be identiꢁied and removed. Nevertheless, the presence
three peaks (Extended Data Fig. 3a). oꢁ a strong PL signal ꢁrom an indirect-gap material is an indication oꢁ
The PL intensity Ihas a power-law dependence on the laser power Lin good optical quality and the removal oꢁ these deꢁects promises ꢁurther
k
38
the ꢁorm I ∝ L with the behaviour governed by the exponent k (reꢁ. ). improvements in the detector perꢁormance oꢁ LiInP Se .
2
6
Excitonic emission shows superlinear behaviour with increasing exci-
tation intensity, corresponding to a coeꢁꢁicient oꢁ 1 < k < 2, whereas
Data availability
ꢁ
ree-to-bound and donor–acceptor pair (DAP) recombination pro-
cesses have a power-law coeꢁꢁicient kbelow 1. This coeꢁꢁicient can be The data that support the ꢁindings oꢁ this study are available ꢁrom the
derived ꢁrom the slope oꢁ a plot oꢁ logI versus logL ꢁor each peak, as corresponding author upon request.
shown in Extended Data Fig. 3b. Each oꢁ the three peaks observed in
LiInP Se has a power-law coeꢁꢁicient below 1 (Extended Data Fig. 3b),
23. Feher, F. Handbuch der Präparativen Anorganischen Chemie (F. Enke, 1954).
24. Aitken, J. A., Cowen, J. A. & Kanatzidis, M. G. Metamagnetic transition in EuSe : a new,
2
2
6
so these emissions are a result oꢁ either ꢁree-to-bound transitions or
DAP recombination. We note the essentially identical behaviour oꢁ
peaks 2 and 3; these peaks correspond to the same band and thus
have the same power-law coeꢁꢁicient oꢁ 0.91 ± 0.04. The temperature
dependence oꢁ PL at 2 mW demonstrates that the higher-energy emis-
sion quenches ꢁirst, with peaks 2 and 3 merging at 70 K and vanishing
by 90 K, and peak 1 persisting until about 150 K (Extended Data Fig. 3c).
The energies oꢁ peaks 2 and 3 blueshiꢁt slightly as the temperature
rises, whereas peak 1 redshiꢁts dramatically between 12.5 K and 150 K
metastable binary rare-earth polychalcogenide. Chem. Mater. 10, 3928–3935 (1998).
5. Sheldrick, G. M. SHELXT – integrated space-group and crystal- structure determination.
Acta Crystallogr. A 71, 3–8 (2014).
2
2
2
6. Sheldrick, G. M. Crystal structure reꢀinement with SHELXL. Acta Crystallogr. C 71, 3–8
(2015).
7. Dolomanov, O. V., Bourhis, L. J., Gildea, R., Howard, J. A. K. & Puschmann, H. OLEX2: a
complete structure solution, reꢀinement and analysis program. J. Appl. Crystallogr. 42,
3
39–341 (2009).
2
8. Momma, K. & Izumi, F. VESTA 3 for three-dimensional visualization of crystal, volumetric
and morphology data. J. Appl. Crystallogr. 44, 1272–1276 (2011).
29. Klimeš, J., Bowler, D. R. & Michaelides, A. Chemical accuracy for the van der Waals
density functional. J. Phys. Condens. Matter 22, 022201 (2010).
0. Setyawan, W. & Curtarolo, S. High-throughput electronic band structure calculations:
(
Extended Data Fig. 3d). This redshiꢁt causes the low-energy tail oꢁ
3
peak 1 to ꢁall past the response limit oꢁ the photomultiplier tube
challenges and tools. Comput. Mater. Sci. 49, 299–312 (2010).
(
ꢁ
1.46 eV), producing an artiꢁicial asymmetry; thus, above 55 K the
it ignores data below 1.48 eV to accurately reproduce the Gaussian
peak shape.
PL quenching occurs when a nonradiative recombination process
competes ꢁor the photogenerated carriers involved in the emission. The
temperature dependence oꢁ this quenching behaviour can be modelled
33. Fonari, A. & Sutton, C. Effective mass calculator https://github.com/afonari/emc (2012).
3
3
3
4. Owens, A. & Peacock, A. Compound semiconductor radiation detectors. Nucl. Instrum.
Methods Phys. Res. A 531, 18–37 (2004).
5. Binnewies, M., Glaum, R., Schmidt, M. & Schmidt, P. in Chemical Vapor Transport
Reactions 452–457 (De Gruyter, 2012).
6. Binnewies, M., Glaum, R., Schmidt, M. & Schmidt, P. in Chemical Vapor Transport
Reactions 165–166 (De Gruyter, 2012).
39
using an Arrhenius plot, ꢁollowing the ꢁrequently used expression :
−
1


E 
a
37. McCall, K. M., Stoumpos, C. C., Kostina, S. S., Kanatzidis, M. G. & Wessels, B. W. Strong
electron–phonon coupling and self-trapped excitons in the defect halide perovskites
I(T) = I0 1 + a exp −


(1)

kT



A
3 2 9
M I (A = Cs, Rb; M = Bi, Sb). Chem. Mater. 29, 4129–4145 (2017).
3
8. Schmidt, T., Lischka, K. & Zulehner, W. Excitation-power dependence of the near-band-
where I0 is the PL intensity at 0 K, a is a constant, k is the Boltzmann
edge photoluminescence of semiconductors. Phys. Rev. B 45, 8989–8994 (1992).
constant and E is the activation energy corresponding to the compet-
39. Pankove, J. I. Optical processes in semiconductors (Dover, 1971).
a
4
4
4
0. Pelant, I. & Valenta, J. Luminescence Spectroscopy of Semiconductors (Oxford Univ.
ing nonradiative recombination. This model corresponds to the sim-
ple case oꢁ a single nonradiative transition, and adequately describes
the Arrhenius plots oꢁ the integrated PL intensity versus the inverse
temperature, as indicated by the solid lines in Extended Data Fig. 3e.
The activation energies oꢁ the nonradiative quenching processes are
determined to be 40.0 ± 6.6 meV, 15.2 ± 6.2 meV and 16.8 ± 3.4 meV ꢁor
peaks 1, 2 and 3, respectively. Again, the behaviours oꢁ peaks 2 and 3
are quite similar, indicating that they should have the same emission
mechanism.
Press, 2012).
1. Veale, M. C. et al. X-ray spectroscopy and charge transport properties of CdZnTe grown
by the vertical Bridgman method. Nucl. Instrum. Methods Phys. Res. A 576, 90–94 (2007).
2. Watanabe, S. et al. Stacked CdTe gamma-ray detector and its application to a range
ꢀinder. Nucl. Instrum. Methods Phys. Res. A 505, 118–121 (2003).
Acknowledgements The exploratory synthesis and materials characterization work was
supported by the National Science Foundation through grant DMR-1708254. The device
fabrication and neutron response measurements were supported by Laboratory Directed
Research and Development (LDRD) funding from Argonne National Laboratory, provided by
the Director, Ofꢀice of Science of the US Department of Energy under contract number DE-
AC02-06CH11357. PL measurements were supported by the Murphy Fellowship from
Northwestern University. This work made use of the SPID and EPIC facilities of Northwestern
University’s NUANCE Center, which has received support from the Soft and Hybrid
Nanotechnology Experimental (SHyNE) Resource (NSF ECCS-1542205); the MRSEC
programme (NSF DMR-1720139) at the Materials Research Center; the International Institute for
Nanotechnology (IIN); the Keck Foundation; and the State of Illinois, through IIN. This work
used the Northwestern University’s Keck Biophysics Facility, which is funded by a Cancer
Center Support Grant (NCI CA060553). This work made use of IMSERC at Northwestern
University, which has received support from the Soft and Hybrid Nanotechnology
Experimental (SHyNE) Resource (NSF ECCS-1542205), the State of Illinois and IIN.
The power-law coeꢁꢁicients oꢁ all three peaks are consistent with
either DAP recombination or ꢁree-to-bound transitions, whereas the
energy oꢁ each peak blueshiꢁts with increasing excitation intensity, as
40
expected ꢁor DAP recombination . However, the temperature-depend-
ent behaviour oꢁ peak 1 is markedly diꢁꢁerent ꢁrom that oꢁ peaks 2 and 3,
with diꢁꢁerent energy shiꢁts as the temperature increases and distinct
quenching processes ꢁor the two bands. To assign these accurately, we
compare the relatively ‘deep’ energy oꢁ peak 1 (coming ꢁrom energy

Author contributions The idea for this project was conceived by M.G.K., D.Y.C., D.G.C. and Y.H.
The manuscript was written by D.G.C., Y.H. and M.G.K. with input from all authors. D.G.C.
developed the synthetic and crystal growth methods. The crystal structure and physical
properties, excluding the optical properties, were measured by D.G.C. The band structure and
effective-mass calculations were conducted by G.T. Optical transmission measurements were
performed by K.M.M. and Z.L. K.M.M. measured the variable-temperature and power-
dependent PL and the Raman spectra. Y.H. and D.G.C. fabricated the devices and performed
Competing interests The authors declare the U.S. patent application 62/701,144.
Additional information
Correspondence and requests for materials should be addressed to M.G.K.
Peer review information Nature thanks Paul Sellin and Mariya Zhuravleva for their contribution
to the peer review of this work.
Reprints and permissions information is available at http://www.nature.com/reprints.
241
57
pulse-height spectrometry of the Am and Co sources. Neutron measurements were
performed by R.O.P., Y.H., D.G.C., D.Y.C. and P.M.D.L. M.G.K. and B.W.W. supervised the project.

Article
Extended Data Fig. 1 | X-ray powder diffraction patterns of LiInP
2
Se
6
reꢁlections with contributions ꢁrom the hand kdirections can be seen and
match well with the simulated pattern (b). c, Experimental versus simulated
powder diꢁꢁraction pattern oꢁ LiInP
Se obtained using a stoichiometric
2 6
amount oꢁ reagents. The unknown secondary phases are marked by asterisks.
d, Sink side oꢁ the reaction tube used ꢁor CVT with no iodine charged into the
tube. e, Sink side oꢁ the reaction tube used ꢁor CVT with iodine as the
transporting agent. This reaction did not employ an initial reverse transport
prepared under different conditions. a, b, Experimental versus simulated
powder diꢁꢁraction pattern oꢁ LiInP Se scaled to the largest peak (a) and scaled
by a ꢁactor oꢁ 20 (b). The simulated pattern has a March–Dollase parameter oꢁ
.3. The X-ray powder diꢁꢁraction pattern oꢁ LiInP Se in a shows a considerable
2
6
0
2
6
preꢁerence to orientation along ⟨0 0 1⟩, that is, the peaks with the largest
intensity correspond to the (0 0 2), (0 0 4) and (0 0 8) planes. This preꢁerred
orientation stems ꢁrom the layered nature oꢁ this compound, which causes the
layers to lie parallel to the sample holder. The simulated pattern with a March–
Dollase parameter oꢁ 0.3 was the reꢁerence pattern used to compare with the
experimental pattern, and accurately accounts ꢁor the sample’s preꢁerred
orientation. When the patterns are zoomed-in closer to the baseline, the
step. f, Experimental versus simulated powder diꢁꢁraction pattern oꢁ LiInP
grown using CVT. The simulated pattern has a March–Dollase coeꢁꢁicient oꢁ
0.3. g, Experimental powder diꢁꢁraction pattern oꢁ bulk LiInP Se beꢁore (red)
and aꢁter (black) DTA up to 760ꢂ°C. h, Experimental diꢁꢁraction pattern oꢁ
LiInP Se grown by CVT beꢁore (red) and aꢁter (black) DTA up to 760ꢂ°C.
2 6
Se
2
6
2
6

Extended Data Fig. 2 | Thermal analysis, morphology and compositional
analysis for LiInP Se . a, b, DTA measurement oꢁ bulk-synthesized (a) and CVT-
grown (b) LiInP Se up to 760ꢂ°C. The vertical axis is a measure oꢁ the voltage
diꢁꢁerence oꢁ the thermocouples on the sample and the alumina standard as the
temperature is varied; thus, the spikes correspond to a transꢁer oꢁ heat ꢁrom the
(c) and a cross-section oꢁ the layers (d). The lamellar crystal morphology oꢁ the
bulk ingot can be clearly seen in the images. A large ꢁlake with a smooth surꢁace
was selected to examine the surꢁace oꢁ the basal plane. c shows a typical image
oꢁ the smooth basal plane. The composition obtained ꢁrom EDS ꢁrom both
2
6
2
6
orientations yielded the semiquantitative ꢁormula In1.1P2.0 Se6.0, which is
sample to the environment and vice versa. The melting point oꢁ LiInP
2
Se
6
is
consistent with the expected ratio oꢁ 1:2:6 determined ꢁrom the crystal
structure. The lithium content could not be determined ꢁrom EDS because oꢁ
14
considerably lower than that oꢁ LiInSe
2
(~915ꢂ°C) , which yields several practical
beneꢁits, including the reduction oꢁ thermally induced deꢁects and improved
stoichiometry ꢁrom reduced evaporative losses oꢁ the volatile Li, P and Se.
c, d, Scanning electron microscopy images oꢁ bulk LiInP Se in the basal plane
2 6
2 6
the insensitivity oꢁ EDS to lithium. e, EDS mapping oꢁ the surꢁace oꢁ LiInP Se
grown using CVT without reverse transport. The growth ꢁront oꢁ the layer
appears rich in iodine.

Article
Extended Data Fig. 3 | PL characterization of LiInP
2
Se
6
. a, Dependence oꢁ the
temperature. e, Arrhenius plots oꢁ integrated PL intensity versus inverse
PL spectra oꢁ LiInP
integrated PL intensity versus power ꢁor each peak. c, Temperature
dependence oꢁ the PL spectra oꢁ LiInP Se at 2 mW. d, PL peak energy versus
2
Se
6
at 12.5 K on laser power. b, Power-law ꢁits oꢁ the
temperature. The solid lines are the least-squares best ꢁit ꢁor each peak
according to Eq. (1). f, Tabulated PL peak properties ꢁor LiInP Se at 12.5 K.
2 6
See Methods ꢁor ꢁurther discussion.
2
6

Extended Data Fig. 4 | Electrical properties and spectroscopic response of
LiInP Se device. a, Current–voltage plot (ꢁrom –100 V to 100 V) oꢁ a pristine
LiInP Se device showing linear behaviour. b, Pulse-height spectra oꢁ α
particles ꢁrom an Am source at 700 V with a Gaussian ꢁit (blue curve).
thickness) exposed to a moderated Pu–Be source while unshielded under 300 V
bias ꢁor 30 min, and the corresponding Gaussian ꢁit (blue curve). The tables in
b, c show the Gaussian equation used to ꢁit the ꢁull-energy peak and values oꢁ
2
6
2
6
2
41
the peak parameters. y
0 c
, vertical oꢁꢁset; x , centre; w, width; A, area; FWHM, ꢁull-
2
6
2 6
c, Binned pulse-height spectrum oꢁ a 7 × 7 mm LiInP Se device (90 μm
width at halꢁ-maximum.

Article
2 6
Extended Data Fig. 5 | Channel number versus voltage for LiInP Se devices
collecting electrons. The ꢁigure oꢁ merit ꢁor charge transport in
semiconductor radiation detectors is the mobility–liꢁetime product, μτ. For
reliable neutron detection, μτmust be suꢁꢁicient so that counts ꢁrom neutron
capture are at a suꢁꢁiciently high channel number to be distinct ꢁrom the
background to allow accurate discrimination. Here the channel number is
plotted versus voltage ꢁor a typical device, and surprisingly the curve exhibits
two slopes in diꢁꢁerent voltage ranges. The Hecht equation cannot model these
data because the model assumes linear behaviour at low voltages, which then
saturate towards a maximum channel number as the CCE reaches unity. The
observed behaviour has the opposite concavity to the model, with a slight rise
in slope as voltage increases. This change in slope can be explained by the
reduction oꢁ the eꢁꢁective voltage ꢁrom an internal polarization antiparallel to
the applied ꢁield. Similarly atypical behaviour has been observed in the
α-particle response oꢁ cadmium zinc telluride at T = 200 K, where this eꢁꢁect
41
was attributed to polarization ꢁrom deep-level traps
.

Extended Data Fig. 6 | Pulse-height spectra of four other samples. a–h, Pulse-
height spectra are shown ꢁor electron (a–d) and hole (e–h) collection. The
growth conditions oꢁ these samples were the same as those used ꢁor the
samples that gave the spectra shown in the main-text ꢁigures. For hole
collection, the spectra had lower channel numbers than the electron-collection
spectra, indicating lower CCE ꢁor hole transport.

Article
6
Extended Data Fig. 7 | Pulse-height spectrum for Li-enriched LiInP
2
Se
6
that the peak channel number and count rate per second were constant over
time. h, Collection oꢁ α-particle spectra by detector 1 in f with a pulse generator.
i, Collection oꢁ α-particle spectra by detector 2 in f with a pulse generator. We
note that the positions oꢁ the pulser peak in h and i are the same and the
corresponding energy resolution is (3.5 ± 0.1)%, indicating the same
electronics noise during collection.
6
devices. a–e, Data are shown ꢁor Li-enriched LiInP
2
Se
6
devices with areas oꢁ
2
2
2
7
× 7 mm (a, b), 5 × 5 mm (c) and 3 × 3 mm (d, e) . Devices with larger electrode
areas had lower energy resolutions ꢁor electron collection, possibly owing to
property variation among diꢁꢁerent samples. f, As-made LiInP Se detectors
with a diameter oꢁ 2 mm (‘1’) and 5 mm (‘2’) on the same single crystal. g, Time-
dependent α-particle spectra collection by LiInP Se detector 2 in f. We note
2
6
2
6

6
Extended Data Fig. 8 | Absorption probability versus LiInP
2
Se
6
thickness
for thermalized neutrons. In addition to charge transport properties, the
neutron-capture cross-section oꢁ LiInP Se must be considered, as this directly
2
6
correlates with the intrinsic maximum oꢁ the detector eꢁꢁiciency. A linear
attenuation coeꢁꢁicient can be used to quantiꢁy the percentage oꢁ incident
neutrons absorbed in a given material through the equation:
−
lα
ꢁ
raction oꢁ neutrons captured (%) = (1 – e ) × 10%, where l is the thickness oꢁ the
active region and α is the linear attenuation coeꢁꢁicient, which is calculated
using the capture cross-section oꢁ thermalized neutrons ꢁor each element and
its molar density. The calculated mass attenuation coeꢁꢁicient ꢁor thermalized
−1
−1
6
neutrons is 5.1 cm and 1.4 cm ꢁor LiInP
2
Se
6
ꢁully enriched in Li and natural-
abundance Li, respectively. A detector with a thickness oꢁ about 9 mm would be
able to absorb 99% oꢁ the incident neutrons. Here we achieved the successꢁul
growth oꢁ crystals with thickness oꢁ ~1 mm, which would absorb ~40% oꢁ
incident neutrons. Increases in eꢁꢁiciency would be achievable through scaled-
42
up growth oꢁ thicker crystals or stacking oꢁ several thinner detectors . When
1
15
In (the most common isotope in natural indium) absorbs a neutron, the
nuclide produces a γ-ray instead oꢁ a highly energetic charged particle. Thus,
about 20% oꢁ the neutrons absorbed do not contribute to the signal, which sets
6
the maximum theoretical detector eꢁꢁiciency oꢁ LiInP
2
Se
6
to approximately
8
0%.

Article
Extended Data Table 1 | Crystal data and structure reꢀinement for LiInP Se at 293 K
2
6

4
3
2
Extended Data Table 2 | Atomic coordinates (×10 ), equivalent isotropic and anisotropic displacement parameters (×10 Å ),
selected bond lengths (Å) and angles (°) for LiInP Se at 293(2) K, with estimated standard deviations in parentheses
2
6