Bin Zhang(張斌) Hong Wu(吳宏) Kunling Peng(彭坤嶺) Xingchen Shen(沈星辰) Xiangnan Gong(公祥南)Sikang Zheng(鄭思康) Xu Lu(盧旭) Guoyu Wang(王國玉) and Xiaoyuan Zhou(周小元)

1Analytical and Testing Center of Chongqing University,Chongqing 401331,China

2Chongqing Key Laboratory of Soft Condensed Matter Physics and Smart Materials,College of Physics,Chongqing University,Chongqing 401331,China

3Chongqing Institute of Green and Intelligent Technology,Chinese Academy of Sciences,Chongqing 400714,China

4University of Chinese Academy of Sciences,Beijing 100044,China

Keywords: γ-InSe single crystals,structure identification,super deformability,thermoelectric properties

1. Introduction

In common sense, most inorganic semiconductors are usually brittle with poor mechanical properties.Seeking“soft”inorganic semiconductors with good mechanical behavior has drawn increasing interest, which is vital for developing flexible functional photoelectric devices. Very recently, an unusual mechanical behavior with super-plasticity/ductility on bulk-InSe single crystals was reported[1,2]which will bring new opportunities for two-dimensional (2D) field and flexible electronics.[1,2]Actually, indium selenide (InSe), a IIIV group semiconductor with layered structure, has been well known in various photoelectric fields due to its promising properties for potential applications in nanoelectronics,[3]nanophotonics,[4]thermoelectrics,[5]and flexible sensors.[6]In addition to the recent report on super-plasticity ofβ-InSe single crystals,[1,2]the mechanical properties of InSe have been intensely studied in the last decades.[7-12]The elastic modulus of bulkγ-InSe single crystal is given as 283 GPa by nano-indentation measurements.[12]While, using the atomic force microscope (AFM), Liet al.[8]measured the Young’s modulus to be 101.37±17.93 GPa in the multilayer 2Dγ-InSe (＞5 L) which is much higher than its bulk counterpart but softer than most existing 2D materials.

In ambient conditions,InSe crystals usually possess three polytype layered structures, i.e., two hexagonal phases ofβtype[13-16]andε-type[17-20]and one rhombohedral phase ofγ-type,[14,16,19,20]as shown in Fig. 1(a). All these layered structures are constituted by the same primitive layered blocks.Each unit layer consists of covalently bonded Se-In-In-Se intraplanes that are connected by weak Van der Waals interactions.The stacking sequences,that is,the key features of structural differences among different phases,are demonstrated as ABAB···in bothβandεphases,and ABCABC···inγphase.Basically,different InSe polytypes have similar constantaandb,while constantcusually corresponds to the number of primitive layers in a unit cell, i.e.,a=b=4.05 ?A,c=16.93 ?A(contains 2 unit layers)for bothβandεphases[17,21]anda=b=4.00 ?A,c=25.32 ?A (3 layers) forγphase.[22]To verify the actual structure of InSe polytypes,x-ray diffraction(XRD)alone is generally not accurate enough due to its strongly[00l] texture and very similar layered structure. In contrast,other techniques as Raman and infrared spectroscopy,[23-25]which are sensitive to the structure vibrational modes, are often combined with XRD to do the structural identification of InSe. Nowadays, the advanced electron transmission microscopy(TEM),especially the spherical-aberration corrected high angle annular dark field(Cs-corrected HAADF)imaging which gives out the atomic arrangements directly, has been widely applied for atomic structure analysis and a few reports on InSe were found.[18,26]Despite the commonβ-,γ-,andε-InSe polytypes,structural defects as plenty of stacking faults,twins, and local multiphases were also claimed to be present in InSe matrix[27,28]and three new layered polytypes asPm1,P63mc,andRmphases were theoretically predicted.[29]Moreover, it is worth noting that several nonlayered metastable phases also exist, such as zinc blende cubic structure in InSe nanomaterials[30,31]and monoclinic or NaCl-like cubic structure under high pressure and high temperature.[32]Generally,the knowledge of atomic structure is fundamental to the understanding of the physical properties, however different structural variations and abundant defects in InSe result in the difficulty in its structure determination. Therefore, accurate and comprehensive methods as well as special attention are highly desired in the structure determination of InSe-based materials.

Two-dimensional materials,owing to the intrinsically layered structure that is favorable for achieving a low lattice thermal conductivity, have been widely regarded as good candidates for thermoelectric materials. As typical 2D structures, indium selenides materials have been received growing attention in thermoelectrics owing to their intrinsic structural characteristics. For instance, thezTvalue of～1.48 at 705 K in In4Se3-δwas obtained, which is resulted from the low lattice thermal conductivity and the high Seebeck coefficient,[33]and thezTof～1.23 at 916 K was reported in Zn-doped In2Se3.[34]Meanwhile,a peakzTvalue of～0.23 in Sn doped InSe polycrystalline material,[35]and～0.14 in Si-doped InSe[36]were reported. In recent years, remarkable achievements have been achieved in single crystalline thermoelectric materials. Promising thermoelectric properties were reported in various 2D materials’single crystals, such as Nadoped SnSe,[37]SnS,[38]and SnS0.91Se0.09,[39]single crystals,and Br-doped SnSe single crystal.[40]However,intrinsic thermoelectric properties of InSe single crystals and the enhancement are still largely unknown.

In this work,γ-InSe bulk single-crystal ingots were grown via a modified Bridgeman method. Through comprehensive analysis of XRD, TEM, and Raman spectroscopy, the structure of InSe bulk crystals was systematically studied. The asgrown InSe ingot predominantly took theγphase with plenty of stacking faults and coexistence of local multiphases. Theγ-InSe bulk crystals also exhibited superior mechanical behaviors with super-plasticity as discovered in itsβ-type.[1]Finally,the thermoelectric performance of InSe bulk crystal was initially studied which showed an extraordinarily high thermal conductivity and it could be significantly reduced via bending/deformations.

2. Experiments and methods

2.1. Growth of InSe bulk crystal

Stoichiometric amounts of high purity elements In filament (99.999%) and Se powder (99.99%) of InSe were weighed and then loaded into cone-shaped silica tubes (φ=16 mm)under an argon atmosphere.The tubes were sealed under～5×10-4Pa and placed in another quartz tubes (larger diameter,φ=13 mm). The silica tubes were slowly heated up to 1273 K over 20 h, soaked at this temperature for 10 h,and subsequently cooled to 973 K as they were moving at a rate of 1.5 mm·h-1. Finally, InSe bulk crystal ingot with dimensions of 14 mm(diameter)×40 mm(length)was obtained(Fig.S1(a)).

2.2. Characterization

The powder was scraped from the as-grown InSe ingot and slightly grounded for XRD measurement using a PANalytical X’Pert diffractometer with CuKαradiation at 40 kV and 40 mA. Raman spectroscopy was carried out with HORIBA JobinYvon HR Evolution,1800 groove/mm holographic grating, and a thermoelectric cooled synapse charge coupled device (CCD) for high spectra resolution, and a solid-state Nd:YAG laser with the wavelength of 532 nm as the excitation source.Several InSe bulk crystal wafers/foils(Fig.S1(b))with a few millimeters in length/width and submillimeter in thickness were exfoliated from the InSe ingot for scanning electron microscopy (SEM) and mechanical analysis. The SEM image and energy dispersive x-ray spectroscopy(EDS)mapping were obtained via a Thermofisher Scientific Quattro S microscope. The InSe nanoflakes were prepared by the classical exfoliation method with a tape (SPV-224S), while the crosssectional samples were prepared by focused ion beam (FIB,Zesis Auriga Compact) for TEM studies. The selected area electron diffraction(SAED),high resolution TEM(HRTEM),HAADF images,and EDS mapping were obtained on a Thermoscientific Talos F200S microscope at 200 kV or a probecorrected Titan Themis microscope at 300 kV. HAADF simulations were performed by using KineHAADF program.[41]Finally, the carrier concentration and the thermoelectric performance were measured by Physical Property Measurement System(PPMS)DynaCool 9.

3. Results and discussion

3.1. Structure analysis

Figure 1(b) shows the XRD pattern of InSe powder. All the main peaks can be indexed to the (00l) reflections, indicating a strong[00l]texture,due to its inherent layered structure. Because of the similar layered structures among different phases,the precise phase composition cannot be convinced(00l)peaks in XRD.Some weak peaks in the XRD patterns are visible as well,but not precise enough for structure determination among InSe polytypes. This is because dense diffraction peaks appear in the simulated XRD patterns of all the InSe structures(the lower panel of Fig.1(b)),meaning some weak peaks may also ascribe to different structures. Thus,the actual structure of the InSe cannot be precisely clarified based on the present XRD pattern(Fig.1(b)),detailed structural analysis is necessary. The morphology of the cleavage surface and cross-sectional fracture(perpendicular to the cleavage surface)of the present InSe bulk was obtained by SEM and shown in Figs. 2(a) and 2(b). The dense parallel striations are presented in the fracture section. The SEM-EDS mappings were taken on the cleaved surface, as seen in Figs. 2(c) and 2(d).In addition, STEM-EDS mappings on the InSe nanoflakes were obtained as depicted in Figs. 2(e)-2(g). Both of them demonstrate the uniform element distribution of InSe at various scales. The chemical composition is identified as atomic ratio of In:Se=48.5:51.5(by SEM-EDS)and 49.7:50.3(by STEM-EDS), which is close to the nominal composition of InSe.

Fig. 1. (a) Structure models of InSe polytype layered structures. (b)XRD patterns of InSe powder accompanied by simulated XRD spectra.

Fig. 2. (a), (b) SEM images of in-plane (a) and cross section (b) surfaces of bulk InSe, respectively. (c), (d) SEM-EDS mappings corresponding to the dotted region in (a). (e)-(g) STEM-EDS mappings of InSe nanoflakes.

To reveal the actual structure of InSe at atomic scale,systematical TEM studies were performed. Firstly, the SAED pattern, HRTEM, and HAADF images of the exfoliated InSe nanosheet were obtained along the [001] direction, as shown in Figs.3(a)-3(c). When compared with the simulated SAED patterns (Fig. S2(b)), the experimental pattern (Fig. 3(a))seems to belong to a hexagonal phase(β-orε-InSe)or at least a mixed structure of hexagonal and rhombohedral phases (γ-InSe). Recently, a similar [001] SAED pattern was reported by Sunet al.,[17]and it was claimed to be aε-InSe single crystal. However, InSe bulk crystal in the present work is predominatelyγ-type instead ofε-type as confirmed by the following comprehensive analysis. Therefore, it seems that the[001]SAED pattern alone is generally not accurate enough for InSe structure determination.Regarding to the HRTEM image(Fig.3(b)),the regular lattices/dots can be roughly divided into three groups based on the image contrast,that is,one centered dimmest dot, three dimmer and three brighter dots(indicated by the circles)as surroundings. Thus,the HRTEM image suggests a 3-fold symmetry, which is suggested to be an indication of rhombohedral structure, i.e.,γ-InSe. Moreover, the[001] HAADF image (Fig. 3(c)) presents a regular hexagonal lattice with uniform contrast/intensity, which is well consistent with the simulated HAADF image ofγ-InSe but significantly different from that ofβ- orε-InSe (see Fig. S3),further suggesting theγ-InSe of the present sample. Subsequently, the sample was continuously tilted to another zone axis,and the corresponding SAED and HRTEM images were obtained(Figs.3(d)and 3(e)). Correspondingly,the same tilting operations (with the same tilting directions and angles)were carried on these polytype structure models. The corresponding atomic projections and diffraction patterns were presented in Figs.S2(a)and S2(b). Expectedly,the SAED pattern(Fig.3(d))can be well indexed to the[ˉ4ˉ21]zone axis ofγ-InSe,but different from the patterns ofβ-andε-InSe as revealed in Fig.S2(a). Similar analysis(serially tilted SAED patterns and HRTEM images) and results were obtained on several InSe nanoflakes. Therefore,on the basis of TEM analysis,it is concluded that the predominate structure of the present InSe isγphase, but the coexistence ofβandεphases still cannot be completely excluded. According to the literature,it seems that InSe single crystals grown by Bridgeman method prefer to formγphase,[21,42-47]and other polytypes ofβ-InSe[1]andε-InSe[17,18]single crystals were occasionally reported.

Fig. 3. TEM characterization of InSe crystals. (a)-(c) SAED pattern, the filtered HRTEM and HAADF images of the exfoliated nanosheets, along the[001]direction,respectively. (d),(e)SAED pattern and HRTEM image under the[1]zone axis through sample tilting,respectively. (f)-(h)SAED pattern,BF-TEM image,and HAADF image of the cross-sectional sample,respectively. Atomic models of γ-InSe are attached on the HAADF images.

As a matter of fact, cross-sectional samples (e.g., along the[010]direction,the same for the[100]direction)are more straightforward for structural determination through the direct observation of stacking sequence. Most of the pervious SAED patterns and HRTEM images were taken along the[001]zone axis,[1,17,21,23,42-45,48,49]which is a top view of the unit layer but instead of a side view of the cross section. Recently,based on the [100] HAADF images, Haoet al.,[18]clearly showed atomically resolved picture of ABAB···stacking sequence ofε-InSe phase on bulk single crystal, while Wuet al.[26]revealed ABCABC···arrangement ofγ-InSe based on a InSe/hBN/graphite heterostructure of thin films. In the present work, cross-sectional samples were prepared by FIB and further analyzed by HAADF imaging for stacking sequence investigation. The SAED pattern(Fig.3(f))can be well indexed to the[010]zone axis ofγ-InSe but completely different from that ofβ- andε-InSe, as compared to the simulated patterns in Fig. S2(c). The appearance of surrounding satellite spots in the SAED pattern further suggests the existence of stacking faults. A similar SAED pattern with satellite spots was also noted in theε-InSe single crystal[18]with stacking faults.Rigoultet al. claimed that the amount of stacking faults in theγ-InSe single crystals could be as large as～6%based on the XRD refinement.[27]Here,the appearance of stacking faults is further confirmed by a bright field(BF)TEM image with high density of “straight line” features, as seen in Fig. 3(g). As expected, the atomic-scale HAADF image(Fig.3(h))reveals that the matrix is indeedγ-InSe with ABCABC···stacking sequence as guided by the white lines. But localεphase with ABAB···(guided by the bule lines)and abnormal AA···(by the red lines,even might be a new phase)stacking sequences also coexist. Moreover, localβ-InSe, twins, and other defective structures were also observed (Fig. S4). Generally, the present bulkγ-InSe with coexistence of dense stacking faults and multiphases/impurifies was vividly identified by the comprehensive TEM analysis. When revisiting the [001] SAED pattern, we found that the experimental pattern (Fig. 3(a))exhibits several extinction diffraction spots as compared to the theoretical one (Fig. S2(b)) ofγ-InSe. These extinction diffraction spots could be ascribed to the above structural modifications, i.e., stacking faults and local multiphases. Interestingly and notably, when the sample tilted from the [010]to [120] zones, a feature of perfect “single crystal” was presented (Fig. S5), and all the features of stacking faults and local multiphases“disappeared”. This is because the polytype InSe possesses the same atomic arrangement projections along the[120]zone as shown in Fig.S2(d).

Raman spectroscopy was further performed on the exfoliated InSe flakes as displayed in Fig. S6. Generally, the Raman peaks in Fig.S6(b)at 40 cm-1,115 cm-1177 cm-1,and 226 cm-1are common in bothβ-,γ-,andε-InSe phases.[17,18]However, the Raman peak around 199 cm-1is thought to originate from the non-centrosymmetric structures asεorγphase.[17-20]Consequently, the～199 cm-1peak here could be resulted fromγ-InSe. Additionally, Raman spectrum in Fig. S6(b) shows significant thickness-dependence. Specifically, as the thickness decreases, the～199 cm-1peak gradually separates into two shoulder peaks. It is found that such separated peaks were also observed in theε-InSe.[21]Therefore, this peak splitting might be ascribed to the increased proportion of stacking faults or local impurities(as confirmed above) in the nano-thickγ-InSe flakes. Finally, based on the above analysis,we conclude that the predominateγ-phase with plenty of stacking faults and coexistence of local multiphases is present in bulk InSe single crystals grown by the modified Bridgeman method.

3.2. Mechanical property analysis

To investigate the mechanical behavior of the presentγ-InSe single crystals,several bulk slices with sub-mm thickness were prepared. The morphology evolution of InSe slices during bending deformation was studied by SEM, as shown in Fig.4(a). In the initial state of InSe slice/foil(marked as initial 0°state), typical “parallel striations” features parallel to the cleavage surface were found(Figs.4(a4)and 4(a5)).

Fig.4.The morphology evolution of γ-InSe slices/foils during bending deformation observed by SEM:(a)the original state(～0°),(b)bending of ～90°,(c)bending of ～180° (folded),(d)return to ～0° (unfolded). The red circles are markers for indicating the root of the bending InSe in(a2)-(d2). (e)The schematic for the bending deformation process.

Subsequently, InSe bulk slice was gently bent to～90°(Fig. 4(b)) and～180°(Fig. 4(c)). Interestingly, no embrittlement or cracking was observed at the bending root,and the initial“parallel striations”were gradually bent simultaneously.Finally, after a recovery treatment the sample was nearly returned to the～0°state,with some cracks and wrinkles being left inside the InSe slice (Fig. 4(d)). Schematic illustrations of the deformation process are shown in Fig. 4(e). Additionally, several similar cases related to the bending performance of InSe slices were shown in Figs. S7 and S8, which further confirm the good plasticity of theγ-InSe. Detailed morphology observations of normal and bending of InSe were depicted in Fig.S8.

Desides the bending deformation,γ-InSe bulk was further deformed by rolling. As shown in Figs.5(a)-5(d),original InSe bulk was enlarged gradually during the continuously rolling process under the compressive and shear stress. Areas of this InSe bulk at different stages(Fig.5)were roughly estimated based on the projection of the sample by using the particle analysis tool in Digital Micrograph software.The projected area of the sample increased from the initial 62.3 mm2(Fig.5(a))to the final 164.6 mm2(Fig.5(d)), with an expansion of about～264%. Both the above bending and rolling experiments and results show thatγ-InSe single crystal has good plasticity and ductility,which are similar to the recently reportedβ-type.[1,2]In addition to the super deformability, it is interesting to note that InSe ingot can be written smoothly like graphite/pencil and also be erased,as shown in Figs.5(e)and 5(f). This indicates that InSe is similar to graphite which is also a printable,writable,and erasable material.

Fig.5. (a)-(d)The ductility of γ-InSe under rolling. The red dotted line(in(a))indicates the projection area for evaluation. (e),(f)The writing experiment of InSe ingot(e)compared with HB pencil(f),respectively.

Fig.6. Wrapping the InSe slices on the surface of the copper wires.

For comparison, similar bending experiments were conducted on bulk graphite, i.e., the commercial highly oriented pyrolytic graphite(HOPG)with comparable thickness of a few hundred micrometers. Unlike the highly plasticity ofγ-InSe,HOPG bulk is brittle as breaking or fracture occurred when under the similar bending performance (even much less than 90°),as seen in Fig.S9. In addition,γ-InSe and graphite slices were crimped on the surface of copper wires. As shown in Fig. 6, the InSe slices can uniformly be wrapped on the surface of the copper wires. Beside the internal cracks or wrinkles, InSe slices were gradually bent along the surface of the copper wire without obvious broken. At the same time, the maximum tensile strain of InSe slices on the outer surface was roughly estimated according the bending beam model. Meanwhile,the strain of theγ-InSe slices under different conditions was analyzed,and bending strain over 12.5%was achieved,as shown in Fig.6(d). Contrarily,graphite flakes could not form a uniform coverage on the copper wire(Fig.S9)with the most area of graphite not being bent,while some brittle fracture occurred. These above findings indicate that bulkγ-InSe has superior mechanical properties than bulk HOPG.Such good mechanical properties ofγ-InSe single crystal are believed to be of great value for its practice applications especially in flexible functional devices.

Fig. 7. The electrical and thermal transport measurement of InSe. (a)Electrical conductivity and Seebeck coefficient of InSe,(b)thermal conductivity of InSe before and after bending twice.

In the following, the TE performance of pristineγ-InSe crystals was initially studied. Figure 7(a) demonstrates the temperature-dependent Seebeck coefficient (S) and the electrical conductivity (σ) for InSe single crystal. The negative Seebeck coefficient in the whole temperature (50-280 K) region is noted, indicating the n-type semiconductors behavior.TheSdecreases from-796 μV·K-1at 17 K to a minimum of-141 μV·K-1at 100 K, increasing to-386 μV·K-1before 280 K.As to the electricity,σincreases from 0.337 S·m-1at 17 K to 24 S·m-1at 80 K,and then decreases to 5.614 S·m-1at 280 K. The value ofσis comparatively small due to the low carrier concentration(Fig.S10),which indicates the high quality of the as-grown InSe single crystals. Currently,a very low thermoelectric performance of InSe single crystal withzTof～0.00027@280 K is obtained, mainly due to its poor electrical properties. The optimization of the electrical properties through alloying or doping to improve the thermoelectric properties of InSe-based single crystals needs further attempts. Figure 7(b)depicts the thermal conductivity as a function of temperature for InSe single crystal. Considering the low electrical conductivity, the total thermal conductivity is determined by the lattice part. Obviously,the thermal conductivity of InSe crystal sharply increases with the temperature arising from the enhanced specific heat capacity at low temperature(below 40 K),and then decreases following theT-1law at higher temperatures owing to the strengthened phonons scattering,which is in good agreement with the behavior of the thermal conductivity in crystalline materials. In contrast,thermal conductivity of SnSe is much lower in the range 50-280 K as shown in Fig.7(b),which may arise from the strong lattice anharmonicity in SnSe. As such, there is a large space to reduce the thermal conductivity for InSe system. One efficient way is to bind the sample to produce wrinkles by enhancing the phonons scattering as shown in Fig.7(b).

4. Conclusion

In summary, bulkγ-InSe single crystals are grown by modified Bridgeman method. The predominate structure ofγ-InSe with dense stacking faults and coexistence of local polytypes is identified by systematic TEM studies particularly on atomic scale analysis. The super deformable mechanical behaviors of the bulkγ-InSe crystals are discovered inγ-InSe. The thermoelectric performance of pristine bulkγ-InSe crystals is initially studied and the thermal conductivity is significantly reduced by applying plastic deformation.InSe-based materials present potentials in the application of flexible electronic devices. In addition to the super deformability,γ-InSe single crystals can also be written smoothly like graphite/pencil.These findings enrich the knowledge and clarify the underlying mechanisms of structural and mechanical property flexibility of InSe polytypes and shed light on developing mechanical plasticity and flexibility of van der Waals layered materials.