Defect Antiperovskite Compounds Hg3Q2I2 (Q = S, Se, and Te) for

Article pubs.acs.org/JACS

Defect Antiperovskite Compounds Hg3Q2I2 (Q = S, Se, and Te) for Room-Temperature Hard Radiation Detection Yihui He,† Oleg Y. Kontsevoi,∥ Constantinos C. Stoumpos,† Giancarlo G. Trimarchi,∥ Saiful M. Islam,† Zhifu Liu,‡ Svetlana S. Kostina,‡ Sanjib Das,‡ Joon-Il Kim,‡ Wenwen Lin,† Bruce W. Wessels,‡,§ and Mercouri G. Kanatzidis*,† †

Department of Chemistry, ‡Department of Materials Science and Engineering, §Department of Electrical Engineering and Computer Science, and ∥Department of Physics and Astronomy, Northwestern University, Evanston, Illinois 60208, United States S Supporting Information *

ABSTRACT: The high Z chalcohalides Hg3Q2I2 (Q = S, Se, and Te) can be regarded as of antiperovskite structure with ordered vacancies and are demonstrated to be very promising candidates for X- and γ-ray semiconductor detectors. Depending on Q, the ordering of the Hg vacancies in these defect antiperovskites varies and yields a rich family of distinct crystal structures ranging from zero-dimensional to three-dimensional, with a dramatic effect on the properties of each compound. All three Hg3Q2I2 compounds show very suitable optical, electrical, and good mechanical properties required for radiation detection at room temperature. These compounds possess a high density (>7 g/cm3) and wide bandgaps (>1.9 eV), showing great stopping power for hard radiation and high intrinsic electrical resistivity, over 1011 Ω cm. Large single crystals are grown using the vapor transport method, and each material shows excellent photo sensitivity under energetic photons. Detectors made from thin Hg3Q2I2 crystals show reasonable response under a series of radiation sources, including 241Am and 57Co radiation. The dimensionality of Hg−Q motifs (in terms of ordering patterns of Hg vacancies) has a strong influence on the conduction band structure, which gives the quasi one-dimensional Hg3Se2I2 a more prominently dispersive conduction band structure and leads to a low electron effective mass (0.20 m0). For Hg3Se2I2 detectors, spectroscopic resolution is achieved for both 241Am α particles (5.49 MeV) and 241Am γ-rays (59.5 keV), with full widths at half-maximum (FWHM, in percentage) of 19% and 50%, respectively. The carrier mobility-lifetime μτ product for Hg3Q2I2 detectors is achieved as 10−5−10−6 cm2/V. The electron mobility for Hg3Se2I2 is estimated as 104 ± 12 cm2/(V·s). On the basis of these results, Hg3Se2I2 is the most promising for roomtemperature radiation detection.

■

INTRODUCTION Hard radiation detection at room temperature is now becoming a universal concern. There are both enormous and incremental demands for such X- and γ-ray detectors in the science of astronomy and applications of industrial and medical imaging as well as nuclear safeguard and national security.1,2 Such radiation detectors are based on semiconductors and promise unparalleled capability for direct photoelectric conversion, with good spatial and energy resolution.3,4 Therefore, there is a longstanding interest in developing new semiconductors for X- and γ-ray detection at room temperature.5−8 Semiconductors must simultaneously satisfy several requirements to achieve hard radiation detection. For sufficient stopping power of high-energy photons, both a high density and high Z elements are required, such as the heavy metals Cd, Hg, In, Tl, Sn, Pb, Sb, and Bi.6 Large bandgaps, in the range of 1.5−2.5 eV, are also required to keep low intrinsic carrier concentration and maintain low leakage current during detector operation at room temperature.1 In addition, the number of carrier trapping centers must be extremely low.9 For conventional compound semiconductors, the general tendency of © 2017 American Chemical Society

decreasing bandgap Eg with increasing atomic number Z has severely confined the possibilities of new candidates.10,11 As a result, to date only a few binary or pseudobinary compounds have shown such functionality.12−15 However, they all suffer from substantial drawbacks, such as growth issues or detector polarization problems. CdTe and CdTe-based pseudobinary chalcogenides, such as CdZnTe (CZT), have been considered as the most promising room-temperature radiation detection materials; however, due to the high cost and growth issues (mainly microstructural defects), the yield and device applications are severely restricted.16,17 Because of either growth and processing difficulties, or long-term stability of the device, only a few alternative compounds, mainly metal halides, such as α-HgI2, TlBr, and PbI2, have been developed. These halides primarily suffer from their weak mechanical properties and detector polarization problems.18−20 To overcome the Eg−Z issue, the concepts of dimensional reduction (DR) and lattice hybridization (LH) hypothesis have Received: March 29, 2017 Published: May 15, 2017 7939

DOI: 10.1021/jacs.7b03174 J. Am. Chem. Soc. 2017, 139, 7939−7951

Article

Journal of the American Chemical Society

from the as-grown crystals using a Miniflex diffractometer (Cu Kα radiation, λ = 1.5418 Å) over the 2θ range of 3−80° with a step size of 0.02°. The diffractometer was operated at 45 kV/40 mA. Surface imaging and composition analysis by energy dispersive spectroscopy (EDS) on as-grown crystals was performed using a Hitachi S8030 scanning electron microscope (SEM) equipped with a PGT energy dispersive X-ray analyzer with an accelerating voltage of 15 kV and a 100 s accumulation time for data acquisition. Atomic force microscopy (AFM) imaging was done in tapping mode using a Bruker’s Dimension FastScan microsce in air, to analyze morphological structures of the as-grown surface with a scan rate of 4 Hz. AFM images were analyzed using Nanoscope version 9.0 software. Optical Characterization. To determine the optical bandgap of Hg3Q2I2, optical diffuse reflectance measurements were employed. The measurements were performed at room temperature using a Shimadzu UV-3600 double-beam, double-monochromator spectrophotometer. The finely ground crystalline sample was spread on a compacted substrate of BaSO4 powder, which served as a 100% reflectance standard. The generated reflectance versus wavelength data were converted to absorbance data using the Kubelka−Munk equation, so the optical bandgap could be estimated.25b−d The optical transmission spectrum was measured on a Lambda 1050 UV−vis−IR spectrophotometer in the range of 300−1500 nm. For photoluminescence (PL) spectroscopy measurements, the setup was identical to that reported previously.32 The samples were illuminated using a CW 405 nm semiconductor laser source (0.8 mm laser spot diameter) with power below 5 mW (1011 Ω cm), high specific density (>7 g/cm3), and good transport properties. They all have high photo response under highenergy photon/α-particle radiation. Furthermore, detailed firstprinciples electronic structure calculations and detector performance measurements confirm their strong potential as materials for X- and γ-ray detection.

■

EXPERIMENTAL SECTION

Materials. Chemicals in this work were used as obtained: (i) mercury metal, 99.9999%, Sigma-Aldrich; (ii) sulfur pieces, 99.999+%, Alfa Aesar; (iii) selenium shot, 99.999+%, Alfa Aesar; (iv) tellurium shot, 99.999+%, Alfa Aesar; (v) iodine lump, 99.999%, Alfa Aesar; (vi) mercury iodide, 99.999%, Alfa Aesar. Caution: Hg metal and its related compound are highly toxic, and great care should be taken with proper protective equipment in both synthesis and handling. Starting Material Synthesis. HgQ was synthesized using elemental mercury and corresponding Q with a stoichiometric 1:1 ratio. Because of the high vapor pressure of elemental mercury (∼15 atm) and the chalcogens at 823 K, a long quartz tube was used that extended beyond the hot zone of the furnace, providing a cold end for the elements to condense and reduce the pressure within the reaction tube. After the reaction, all of the compounds were transported to the cold end. Vapor Transport Growth of Single Crystals. Vapor transport has been successfully applied for the growth of high-quality large single crystals of the chalcohalides Hg3Q2I2. HgQ and α-HgI2 (∼2 g in total) were used as the starting materials with a molar ratio of 2:1. The detailed information for the temperature profile and vapor transport parameters is given in Figure S1 and Table S1. The indicated temperature parameters, including the hot and cold zone temperatures and the transport time, were explored extensively to find the ideal values for large single-crystal growth. To determine phase purity, powder X-ray diffraction (XRD) was performed on powder ground 7940

DOI: 10.1021/jacs.7b03174 J. Am. Chem. Soc. 2017, 139, 7939−7951

Article

Journal of the American Chemical Society

Figure 1. Diversity of crystal structures in Hg3Q2X2 (Q = S, Se, Te; X = Cl, Br, I) and schematic dimensional reduction (DR) of Hg−Q frameworks through lattice hybridization (LH) in this system. It should be noted that the crystal structures shown are all stable at ambient conditions. Row I shows the representative crystal structures of the Hg3Q2X2 compounds as these vary from 3D to 0D, where mercury atoms are depicted as red spheres, sulfur, selenium, and tellurium atoms are shown as the yellow, purple, and blue spheres, respectively, while bromine and iodine atoms are shown as green and blue-gray spheres, respectively. Row II illustrates the dimensional reduction of Hg−Q frameworks in this system, where mercury atoms are shown as red atoms and chalcogen elements are shown as gray spheres. Row III shows the structural evolution of the connectivity between the [QHg3]4+ pyramids for the basic cubic building block of the Hg3Q2X2 compounds as the dimensionality evolves from the 0D, to the 1D, to the 2D, and to the 3D crystal lattice. In this system, there includes Hg3Q2Cl2 (Q = S, Se, Te), and Hg3Te2X2 (X = Br, I) 3D structure, Hg3Se2Br2 2D layered monoclinic structure, Hg3Q2I2 (Q = S, Se) 1D orthorhombic structure that contains Hg−Q ribbons along the a axis, and Hg3S2Br2 0D monoclinic structure. detectors were prepared, where the C or Au electrodes were on parallel surfaces of the crystal. Carbon electrodes were prepared by depositing colloidal graphite paint, and Au contacts were deposited by e-beam evaporation. The γ-ray sources employed were a noncollimated 0.2 mCi 57Co 122 keV and 241Am 59.5 keV γ-ray source. In each measurement, the as-prepared detectors were placed in an enclosed shielding box connected to an eV-550 preamplifier. A positive bias varying from 10 to 500 V was applied on the bottom contact, while the γ-ray source was irradiated on the top cathode contact, as schemed in Figure S2. A SPEAR detector operated at a bias voltage of 500 V and equipped with a 5 × 5 × 5 mm3 CZT crystal was used as a reference. The signals from the preamplifier were further amplified and shaped by the ORTEC amplifier (model 572A) with a gain of 100− 500 and shaping time of 1−6 μs. The final signals were subsequently evaluated by a dual 16 K input multichannel analyzer (model ASPEC927) and read into the MAESTRO-32 software, which generated and displayed the response spectrum. The mobility was estimated using alpha pulse height spectroscopy, which was tested by the 241Am α particle source with a typical kinetic energy of 5.49 MeV. The distance of alpha source from the surface of the detector was less than 5 mm. Alpha particles usually have a typical characteristic decay distance of less than tens of micrometers in dense materials.36 For the mobility evaluation, signals from the preamplifier, instead of transferring to the ORTEC amplifier, were collected by using a homemade interface based on National Instruments software, which could capture the complete transient waveforms directly from the preamplifier with a maximum time resolution of 4 ns. Each transient waveform was analyzed to determine its rise time tr, which corresponded to the transit time between 10% and 90% of the amplitude of the transient pulses.

family in this system is Hg3Q2X2 (Q = S, Se, and Te; X = Cl, Br, and I), with numerous structural types.37−40 Figure 1 shows all of the known crystal structure types of the Hg 3 Q 2 X 2 compounds, which are thermodynamically stable at room temperature. Detailed crystallographic information is given in Table 1. Table 1. Crystal Structure and Physical Properties of Hg3Q2X2 (Q = S, Se, and Te; X = Cl, Br, and I) compound

crystal structure

density (g/cm3)

dimensionality of [Hg3Q2] frameworks

Hg3S2Cl241 Hg3Se2Cl242 Hg3Te2Cl242 Hg3S2Br243 Hg3Se2Br244 Hg3Te2Br242 Hg3S2I238 Hg3Se2I238 Hg3Te2I245

I213 I213 I213 C2/m C2/m I213 Imma Imma C2/c

6.83 7.42 7.59 7.11 7.57 7.78 7.04 7.38 7.58

3D 3D 3D 0D 2D 3D 1D 1D 3D

All Hg3Q2X2 compounds consist of the same [QHg3]4+ pyramidal building block, covalently linked with itself through corner-sharing to form Q−Hg−Q bridges (Figure 1). Depending on the orientation of the 2-coordinate Hg bridge, the crystal structure of the Hg3Q2X2 compounds varies from zerodimensional (0D) cubes (Q = S, X = Br) to one-dimensional (1D) chains (Q = S, X = I) to two-dimensional(2D) sheets (Q = Se, X = Br) and finally three-dimensional (3D) networks (Q = Te, X = I). The halide ions do not directly influence the

■

RESULTS AND DISCUSSION Crystal Structure. The crystal chemistry in the ternary Hg−Q−X system is remarkably diverse. The most important 7941

DOI: 10.1021/jacs.7b03174 J. Am. Chem. Soc. 2017, 139, 7939−7951

Article

Journal of the American Chemical Society

Figure 2. Structure evolution in Hg3Q2X2 (Q = S, Se, Te; X = Cl, Br, I) based on the BaTiO3-type perovskite structure. The building block X[QHg1.5] is modified from the unit cell of BaTiO3 (P4/mmm, ICSD-27968),47 where three Hg atoms were removed and Q atom is trigonal pyramidal coordinated and Hg atoms are always found on neighboring three surfaces. Through operating the building block X[QHg1.5] by translation and rotation, different prototype structures with 0D-3D Hg−Q motifs can be obtained.

in the connectivity of the lattice is caused by further rotation of two of the eight [QHg3]4+ pyramids to form linear dihedral angles (δ = 180°, the pyramids are pointing inward the cubic cavity), thus allowing for the expansion of the lattice in two dimensions. The most common structural type, the 3D 3 2+ framework, is found in the Hg3Q2Cl2 (Q = S, ∞ [Hg3Q2] Se, Te) and Hg3Te2X2 (X = Br, I) derivatives. In the 3D framework, all of the [QHg3]4+ pyramids have a nearestneighbor dihedral angle of δ ≈ 120° forming an open framework, which consists of defective [Hg7Q6]2+ cubes leaving a single [QHg3]4+ pyramid completely detached from the rest of the cube. An alternative structural description is based on the topology of the atoms (Figure 2). Assuming that the [QHg3]4+ pyramids are part of incomplete [QHg3□3]4+ octahedra (where □ denotes a vacancy), the crystal structure of the Hg3Q2X2 compounds can be described as a defective antiperovskite with an idealized formula “Hg6Q2X2” in which 50% of the Hg atoms are missing to produce the Hg3□3Q2X2 composition (see Figure 2). A similar structure is also adopted in defect perovskite K2Sn2O348 and recently reported defect antiperovskite Fe2SeO,31 where 1/2 of anions and 1/3 of cations are missing, respectively. The structure evolution of [Hg3Q2] motifs plays an important role in determining the physical properties of these materials as indicated in the following discussion. To illustrate the factors that determine the stable structure type for a given Hg3Q2I2 compound, we investigated the energy stability of the Hg3Q2I2 compounds assuming the Hg-vacancy ordered patterns that define the structure types. To this end, we calculated the stability of each Hg3Q2I2 compound in all five possible crystal structures. These structure types taken as prototypes are listed in Table S2. We generated the structure for the total energy minimization of the Hg3Q2I2 compounds in each of the structure types by replacing in the corresponding prototype solids with different Q. We then fully relaxed both the symmetry-unconstrained unit-cell parameters and the atom positions of these structures to the DFT (local) total energy minimum. In these DFT structural optimizations, we used the Perdew−Burke−Ernzerhof gradient-corrected exchange and

crystal structure because they participate in the chemical bonding only loosely forming weak Hg···X bonds (for Hg2+ the ionic radii are 0.69 and 0.96 Å for linear and tetrahedral coordination,46 respectively). Their main functions in the structure are (i) to fill the voids left in the [QHg3]4+ frameworks, (ii) to template the directionality of the covalent framework through size restrictions, and (iii) to counterbalance the excess positive charge. However, the dimensionality of the lattice is indirectly affected by the nature of the halide ions due to steric effects (crystal packing) and electronic effects (electronegativity). The combination of lighter halogens (Cl) and heavier chalcogens (Te), that is, Hg3Q2Cl2 and Hg3Te2X2, invariably led to 3D [Hg3Q2] frameworks, whereas the combinations of heavier halides and the lighter chalcogenides produce lower dimensionality structures. Irrespective of the dimensionality, however, the high density of the materials is always retained higher than 6.8 g/cm3 (see Table 1). The simplest structural type is seen in the crystal structure of Hg3S2Br2, which consists of discrete molecular [Hg12S8]8+ cubes arising from the fusion of eight [SHg3]4+ pyramids so that the eight vertices are occupied by the chalcogenide atoms. The bromide ions occupy the center of the cubes as well as the space between the cubes. These molecular cubes can conceptually polymerize to form higher dimensionality ∞ z [Hg3Q2]2+ frameworks. The determining factor of the dimensionality of the lattice is the relative connectivity of the [QHg3]4+ pyramids, which can be described by the dihedral angle between the basal planes of the pyramids (Figure 1). In the case of the 0D molecular cubes of α-Hg3S2Br2, the dihedral angle is δ ≈ 60° for each nearest neighbor [SHg3]4+ pyramid. The structure can convert to a 1D ∞1[Hg3Q2]2+ chain when the top four [QHg3]4+pyramids rotate by ∼60° (δ ≈ 120°) to form bridges that connect the Hg-deficient [Hg10S8]4+ cubes along the crystallographic a-axis (Imma space group). This structural modality is found in Hg3Q2I2 (Q = S, Se) presumably imposed by the large size of the iodide anions. A further rearrangement of the Hg pyramids leads to the 2D structure found in Hg3Se2Br2. The puckered ∞2[Hg3Se2]2+ layers consist of interconnected Hg-deficient [Hg11Se8]6+ cubes growing along the crystallographic ab plane (C2/m space group). The change 7942

DOI: 10.1021/jacs.7b03174 J. Am. Chem. Soc. 2017, 139, 7939−7951

Article

Journal of the American Chemical Society correlation functional.35,49 The ladder diagram of Figure 3 shows the total energies in eV per formula unit (fu) obtained by

about 95.5°. Therefore, in Hg3Te2I2 it is the 3D structure S1 instead of the 1D structure S4 that exhibits Hg−Q−Hg angles closer to 90°. This indicates that for Hg3Te2I2, the structure S1 affords a better accommodation of the strain in the Hg−Q framework than structure S4. Crystal Growth and Characterization. The ternary phase diagrams of Hg3Q2I2 indicate an incongruently melting behavior,50 inhibiting crystal growth from their stoichiometric melt. Therefore, we employed vapor transport to grow single crystals, as this method is also performs self-purification during transport.51 It has been employed in growing compounds with high melting points, semiconductors that are inclined to dissociate at their melting temperatures, and those suffering from phase-transitions during cooling, such as ZnSe, ZnS, and α-HgI2, III nitrides (GaN), oxides (ZnO), and so forth.19,52−56 In general, the temperature for vapor transport takes place at lower temperature than the melting point, thus reducing possible contamination from the crucible. The solid−vapor interfaces exhibit higher interfacial morphological stability during growth than do solid−liquid interfaces,57 which suppresses the formation of secondary phases during growth and leads to better quality of the crystals. Here, we reproducibly grew large single crystals of Hg3Q2I2 through vapor transport (Figure 4). A typical vapor transport system consists of a closed silica tube with starting materials (HgQ and α-HgI2) placed at one end as shown in Figure S1a. The equilibrium species in the gas phase at transport temperature presumably include Hg(g), Qn (g), and HgI2(g), I2(g), etc.58

Figure 3. Ladder diagram showing the total energies of the Hg3Q2I2 compounds calculated for the structure types listed in Table S2. The cell-internal and lattice vector degrees of freedom were fully relaxed to the nearest local total-energy minimum. The atom coordinates of the prototype solid associated with each structure were used as a starting point for each local total-energy minimization.

full relaxation of Hg3S2I2, Hg3Se2I2, and Hg3Te2I2 in each of the five structure types. Here, we arbitrarily set the lowest energy minimum among the calculated structure types as the zero energy level. We find that the lowest-energy structure type for each Hg3Q2I2 system corresponds to the experimentally observed structure: the 1D structure type S4 in the case of Hg3S2I2 and Hg3Se2I2, and the 3D structure type S1 in the case of Hg3Te2I2, showing an excellent agreement between theory and experiment (Figure 3). The deviation of the fully relaxed and structures from the respective ideal cubic vacancy-defective antiperovskite configurations is the result of competing structural effects: the formation of Hg−I bonds and Hg−Q bonds with optimal length, and the deformation of the bond angles in the Hg− chalcogen network from the ideal cubic values. The interplay among these effects is subtle as it is indicated by the fact that the five structure types in each Hg3Q2I2 system are distributed within an energy interval smaller than 0.05 eV/fu and therefore are close in energy to each other. We find that the structure S1, which is above S4, the minimum energy structure of Hg3S2I2, moves down in energy in the case of Hg3Se2I2 and finally becomes the most stable structure type for Hg3Te2I2. We inspected the relaxed structures of the S1 and S4 type in search of changes in the bonding network that could produce this switch in energy ordering. The minimum I−Hg bond distance remains constant within the Hg3Q2I2 series at 3.05 Å in structure S1 and at 3.10 Å in structure S4. The increase of the atomic radius of the chalcogen species going from S to Se to Te produces an expansion of the Hg−Q bonds. This expansion occurs along with a decrease of the Hg−Q−Hg bonding angles that get in closer to the 90° value of the ideal defective cubic antiperovskite structure. A smaller deviation from the 90° value results in a lower energy cost associated with the accommodation of strain in the Hg−Q framework, and the corresponding crystal structure becomes more stable. In the fully relaxed Hg3S2I2, we find that the Hg−S−Hg angles vary between 100.2° and 100.6° in structure S4 and between 98.6° and 101.4° in structure S1. In Hg3Te2I2, the Hg−Te−Hg angles vary between 96.5° and 97.2° in S4, while in structure S1 they range from 90.6° and 101.5° with the majority of angles being

Figure 4. Single-crystalline wafers of Hg3S2I2 (a), Hg3Se2I2 (b), and Hg3Te2I2 (c) grown through vapor transport method. The scale bars correspond to 5 mm. (d) Electronic absorption spectrum obtained from diffuse reflectance measurement on ground Hg3Q2I2 crystals. (e) Absorption spectrum of a thin Hg3Se2I2 single crystal with dimension of ∼5 × 5 × 0.1 mm3. The inset shows linear analysis of the α1/2 versus energy plot used to determine the bandgap. 7943

DOI: 10.1021/jacs.7b03174 J. Am. Chem. Soc. 2017, 139, 7939−7951

Article

Journal of the American Chemical Society

Figure 5. SEM and AFM images on the surfaces of Hg3Se2I2 crystals in (a)−(c) and Hg3Te2I2 crystals in (e) and (f) showing flat steps and terrace morphology. SEM images in (a) and (e) showed the feature on the cleavage surface. AFM images of the as-grown crystal surface (b), (c), and (f) also show terraces on the surface. The indices of the basal (0k0) reflections observed of an as-grown crystal surface using XRD for Hg3Se2I2 and Hg3Te2I2 in (g). The lattice planes of (0k0) in Hg3Se2I2 and (hh0) in Hg3Te2I2 are shown in (d) and (h).

intercept with the x-axis (Figure 4e). The Eg values obtained through these two methods are in good agreement with each other. The bandgaps of Hg3Q2I2 are smaller than the bromides and chlorides, while Hg3Q2X2 (Q = S, Se, Te; X = Cl, Br) tends to be larger than 2.50 eV.39,60,61 A smaller bandgap is preferable for better intrinsic energy resolution, due to a smaller intrinsic electron−hole pair generation energy ε.6 Under SEM and AFM examinations, we observed parallel strips of stepwise terraces near the edge of the cleaved surfaces of Hg3S2I2, Hg3Se2I2, and Hg3Te2I2 (Figure 5). The step size was about 0.1−10 μm. Using atomic force microscopy (AFM) studies, the typical terrace morphology was found on all of the as-grown samples. The terrace step distance for Hg3Se2I2 was 0.1−0.5 μm, and ≥1 μm for Hg3Te2I2. XRD patterns (Figure 5g) indicated that the surface of the as-grown Hg3Se2I2 crystal was comprised of (0k0) planes. This terrace height of the terraces shown in Figure 5c was measured by AFM for Hg3Se2I2 to be about 0.75−0.96 nm, and is in good agreement with the spacings of the basal (0k0) planes of 0.969 nm. Because of the strong covalent bonding character of Hg−Se bonds, the crystals tend to grow along the direction of [Hg3Se2] ribbons (Figure 5d). A similar terrace feature was also confirmed on the as-grown surface of Hg3Te2I2 samples. The index of crystallographic plane of the as-grown surface of Hg3Te2I2 is (hh0) as determined by the well-defined (hh0) family of Bragg reflections observed of the surface of the crystal by XRD pattern. The measured terrace height of the Hg3Te2I2 surface was between 0.70 and 0.97 nm (Figure 5f), which corresponds to the minimum spacing of ∼0.968 nm in the crystal plane of (hh0). The width of the terraces was about 1 μm, which was much larger than that in Hg3Se2I2 crystals. Optical Properties. The photoluminescent (PL) properties of Hg3Se2I2 single crystals at 15 K were examined to investigate any radiative defects. The PL spectrum of the as-grown Hg3Se2I2 crystal exhibited three overlapping peaks in the 1.4− 1.9 eV range, and two additional overlapping peaks in the 1.9−

According to partial pressure measurements of Hg3Te2I2, the most dominant of the gaseous species present are HgI2(g) and Hg(g).58 Thus, it is reasonable to deduce that the major limitation of the vapor transport efficiency is the lower vapor pressure of chalcogens. The effect of additional iodine (I2) as the transport agent was investigated. This method was very effective for Hg3Te2I2 with 10−100 mg of excess I2 during transport. By forming volatile tellurium iodides (such as TeI and TeI458), Te atoms are effectively transported to the cold end where they are incorporated into the single crystalline Hg3Te2I2 phase. The preferred morphology of as-grown crystals for all three materials is plate-like (Figure 4). The single crystals of Hg3S2I2, Hg3Se2I2, and Hg3Te2I2 appear transparent yellow, bright red, and gray-red in color, respectively. Under optical microscopy, there is no evidence of intergrown domains, or secondary phases. The powder X-ray diffraction (XRD) pattern in Figure S3 showed all samples are phase pure. In particular, Hg3S2I2 crystallizes into orthorhombic structure in space group Imma, with a = 9.7992(8) Å, b = 18.703(3) Å, and c = 9.4622(7) Å.38 Hg3Se2I2 is isostructural with Hg3S2I2, with a = 9.7660(9) Å, b = 19.381(3) Å, and c = 9.6332(9) Å.38 Hg3Te2I2 crystallizes into monoclinic crystal structure in space group C2/c, with a = 14.22(4) Å, b = 9.70(3), and c = 14.34(2) Å, β = 79.9(2)°.58 In addition, there is no phase transition of Hg3Se2I2 in the range of ∼110−573 K.38 The experimental bandgaps for Hg3Q2I2 are 2.25, 2.12, and 1.93 eV, respectively, decreasing from S to Te (Figure 4d). The extension of the outermost s and p orbitals of the chalcogen atoms in the series S, Se, and Te accounts for the systematic decrease in the experimental band gap. In the Hg3Q2I2 system, the [Hg3Q2]2+ frameworks are mainly covalent, while the bonding for Hg−I is mostly ionic. Eg is usually determined by the weakest Hg−I bond rather than Hg−Q because it has the minimal interband separation.59 For Hg3Se2I2, a UV−visible transmission spectrum was obtained (Figure 4e). For materials with indirect band gaps, a linear analysis of the plot of α1/2 versus energy gives the bandgap of 2.15 eV for Hg3Se2I2 as an 7944

DOI: 10.1021/jacs.7b03174 J. Am. Chem. Soc. 2017, 139, 7939−7951

Article

Journal of the American Chemical Society

dependence of laser power on PL intensity in Hg3Se2I2. When k is in the range of 1 < k < 2, the emission is attributed to bound exciton emission. When 0 < k < 1, the emission is believed to arise from free-to-bound radiative recombination such as free electron and neutral acceptor recombination, free hole and neutral donor recombination, or donor−acceptor pair (DAP) recombination.62 The intensity of the peaks increased with increasing laser power. The corresponding values of k for each peak are shown in Figure 6b. From the power dependence on the peak intensity, we attribute the peaks centered at 1.52, 1.59, and 2.14 eV to the donor−acceptor pair (DAP) or free-tobound radiative recombination, and the peak at 2.05 eV is due to bound exciton transition. Although the k value for 1.71 eV peak is about 1 within the error margin, this is likely not due to a free exciton, because the expected bandgap at 12.5 K is higher than the room-temperature bandgap of 2.12 eV. Hence, we attributed it also to DAP recombination. Electrical Properties and Device Characterization. The as-grown crystals form platelets with smooth surfaces so they can be used for device preparation without further mechanical polishing. The use of carbon contacts gives stable detector performance with no evidence of the chemical reaction between the electrodes and crystals. The intrinsic resistivities of Hg3S2I2, Hg3Se2I2, and Hg3Te2I2 fitted from the linear region of the I−V curve in the low bias range (Figure S4) were 2.0 × 1011, 1.2 × 1012, and 3.5 × 1012 Ω cm, respectively. These values had very good reproducibility, as the intrinsic resistivity of these compounds was always at least 1011 Ω cm regardless of various growth temperatures, starting materials, and transport agents. Also, intentional doping here is not necessary to compensate the intrinsic or extrinsic ionized defects inside the single crystals. In comparison, for CdTe and CZT, obtaining high

2.3 eV range (Figure 6a). Gaussian analysis of the peaks for a spectrum collected using 1.0 mW laser power is shown in

Figure 6. (a) PL spectra of a Hg3Se2I2 crystal at various laser intensities of the 405 nm laser at 15 K. (b) A plot of log(PL intensity) versus log(laser power) for Hg3Se2I2.

Figure 6a. Because PL intensity (I) is proportional to Lk, where L is the laser power and k is the exponent,62 we examined the

Figure 7. γ-Ray response under 57Co γ-ray source of Hg3S2I2 (a) and Hg3Se2I2 (b) and Hg3Te2I2 (c) devices with the corresponding μτ fitting (d−f) using the Hecht equation. The device thickness in (a−c) is 0.20, 0.20, and 0.30 mm, respectively. Note that the background spectrum under higher voltages was also measured as reference to distinguish the signals with background noise. The counting time was 200 s. For the μτ fitting in (d−f), the maximum channel numbers were extracted from each spectrum. The measurement uncertainty is about 10% (error bars). 7945

DOI: 10.1021/jacs.7b03174 J. Am. Chem. Soc. 2017, 139, 7939−7951

Article

Journal of the American Chemical Society

Figure 8. α-Particle response (a) of Hg3Se2I2 devices under various biases from 241Am α particle source with the μτ product fitting (e) according to the Hecht eq 1. (c) γ-Ray response under 241Am γ ray source. The counting time for each spectrum was 120 s. The histogram of rise time tr distribution (b) for the Hg3Se2I2 detector under various biases. The peaks from (b) were used to estimate the mobility of electrons in the Hg3Se2I2 detector (d) by linear fitting according to eq 2; the inset is the transient pulse from one radiation event recorded by the preamplifier. The error bar in (d) represents ±10% errors. For the μτ fitting, the channel numbers used in (e) were extracted from the centroid channel number of alpha peaks by Gaussian fitting in (a) and with ±10% uncertainty (error bars).

resistivity (109−1010 Ω cm) is not straightforward due to the precise in situ temperature and stoichiometric control required for detector-quality crystals.16 Photoresponse was assessed under 405 nm laser illumination using the ratio of photocurrent to dark current (σphoto/σdark) as well as the difference between photocurrent and dark current (Δσ).63 Typical I−V characteristics of Hg3Se2I2, Hg3Te2I2, and Hg3Te2I2 samples are shown in Figure S5a−c, respectively. In the cases of Hg3Se2I2 and Hg3Te2I2, the photoresponse was larger for electrons (negative bias) than for holes, whereas in Hg3S2I2, the holes and electrons had comparable photoresponse. A summary of average resistivity and photoresponse of these samples is listed in Table S3. Hg3Se2I2 exhibits the highest photo response (>230, negative bias) as compared to Hg3Te2I2 and Hg3S2I2 samples (10 10136 1012−1013 101280 >1011 >1012 >1012

me*

mh*

μe (cm2/V·s)

HVc (kg/mm2)

0.92m0 1.25m075a 0.31m076a 0.34m081 0.23m0 0.20m0 0.36m0

3.00m0 1.11m075a 2.06m076a 0.43m081 0.91m0 0.86m0 1.01m0