In-plane anisotropic electronics based on low-symmetry 2D materials: progress and prospects

Low-symmetry layered materials such as black phosphorus (BP) have been revived recently due to their high intrinsic mobility and in-plane anisotropic properties, which can be used in anisotropic electronic and optoelectronic devices. Since the anisotropic properties have a close relationship with their anisotropic structural characters, especially for materials with low-symmetry, exploring new low-symmetry layered materials and investigating their anisotropic properties have inspired numerous research efforts. In this paper, we review the recent experimental progresses on low-symmetry layered materials and their corresponding anisotropic electrical transport, magneto-transport, optoelectronic, thermoelectric, ferroelectric, and piezoelectric properties. The boom of new low-symmetry layered materials with high anisotropy could open up considerable possibilities for next-generation anisotropic multifunctional electronic devices.

It is known that reducing the symmetry of materials is generally associated with exceptional anisotropy in electronic energy band structure and can be regarded as a process of lowering the dimensionality of the carrier transport. Therefore, the electrical, optical, thermal, and phonon properties of these anisotropic materials are diverse along the different in-plane crystal directions. Since these unique intrinsic angle-dependent properties of low-symmetry 2D materials cannot be easily realized in highly symmetric 2D materials, the emergence of in-plane anisotropic properties can provide another new degree of freedom to tune the previous unexplored properties and supply a tremendous opportunity to the design of new devices, such as polarization sensitive photodetectors, 31,32 polarization sensors, 33 articial synaptic devices, 34 digital inverters, 35 and anisotropic memorizers 36 that are highly desired in integrated logic circuits. Thus, BP and other low-symmetry layered materials (Table 1) have attracted enormous research interest towards potential applications and become a hot topic in the community of nanoscience and nanotechnology.
Moreover, strong in-plane anisotropic transport properties of low-symmetry 2D materials are typically a result of the different energy band structure along the different in-plane directions of the layered crystal lattice, leading to drastically different carrier effective mass along the different crystal directions. Therefore, the study on anisotropic magnetotransport properties of low-symmetry layered 2D materials could offer a powerful and useful tool to investigate energy band structures and new physical phenomena of low-symmetry layered 2D materials, such as anisotropic weak localization, anisotropic superconducting, and anisotropic non-linear magneto-resistance, which provide more a comprehensive understanding of their physical properties and insights into potential applications. Table 1 Low-symmetry 2D layered materials classified by the crystal structure and space group and their basic parameters In addition, due to the anisotropy of transport properties offered by low-symmetry layered 2D materials, their optoelectronic, thermoelectric, piezoelectric, and ferroelectric properties should also be dependent on the crystalline directions. There is no doubt that the corresponding performance along a certain crystalline direction is better than the others. Therefore, investigating the anisotropic electronic properties along different crystalline orientations in low-symmetry 2D materials can optimize the performance of eld effect transistors, 35 photodetectors, 36 thermoelectric devices, 15 piezoelectric devices, 37 ferroelectric devices, 38 and so on. Some anisotropic semimetals exhibit large non-saturating magnetoresistance (MR) along a specular orientation and can be used in magnetic devices, e.g., magnetic sensors and magnetic memories. [37][38][39][40] Therefore, the study on anisotropic electronic properties in lowsymmetry 2D materials is of considerable interest and importance.
Herein, we summarize the recent advances in low-symmetry layered materials and their anisotropic electrical properties. We rstly classify these low-symmetry layered materials by the periodic table of elements and crystal structures. Secondly, we introduce the synthetics methods and their relative merits of these materials, followed by the common methods for characterizing the anisotropy including polarization-dependent absorption spectroscopy (PDAS), azimuth-dependent reectance difference microscopy (ADRDM), angle-resolved polarization Raman spectroscopy (ARPRS), and angle-resolved DC conductance. Then, the anisotropic electronic properties, e.g., optoelectronic, magneto-transport, thermoelectric, piezoelectric, and ferroelectric properties ( Fig. 1) with the applications using them are introduced and discussed. In the end, we conclude the challenges encountered and the future prospects of low symmetry layered materials.

Crystal structure and electronic band structure
Since materials' anisotropic properties and functionalities are strongly related to their crystal structures and compositions, it is crucial to study the low-symmetry 2D layered crystal structures rst. Aer early investigations on the characterization of structures and properties of bulk samples, the family of lowsymmetry 2D layered materials have recently attracted tremendous attentions due to the novel anisotropic properties. Here, we will categorize the low-symmetry 2D layered materials through the conductivity and periodic table of elements as shown in Fig. 2.

Elementary 2D material
Among the 2D layered anisotropic materials, black phosphorus (BP) has a wide thickness-tunable direct bandgap ($0.3 eV of monolayer to 2 eV of bulk) and high intrinsic mobility, with a puckered orthorhombic structure of a Cmca space group symmetry (see Fig. 3), which makes it a promising core material for next-generation (opto)electronic devices. 31,[62][63][64][65][66][67][68] In the atomic layer, each phosphorus atom in BP is connected to three adjacent phosphorus atoms, leading to two distinguishing dened directions: armchair and zigzag directions along the x and y axis, respectively. The highly anisotropic crystal lattice gives rise to its anisotropic in-plane electrical, optical, and phonon properties. Tellurene is another elementary in-plane anisotropic semiconductor, which is comprised of non-covalently bound parallel Te chains. Tellurene crystallizes in a structure composed of Te atomic chains in a triangular helix that are stacked together via van der Waals forces in a hexagonal array. In this structure, Te atoms form covalent bonds only to the two nearest neighboring Te atoms in the helical chain as shown in Fig. 3. The band gap of tellurene is also thickness-tunable varying from nearly direct 0.31 eV (bulk) to indirect 1.17 eV (2 L). Moreover, compared with BP, 2D tellurene also exhibits an extremely high hole mobility ($10 5 cm 2 V À1 s À1 ) but has a better environmental stability. Tellurene, therefore, is expected to rival black phosphorus in many applications. 15,60,[69][70][71][72][73][74] 2.2 Binary IV-VI chalcogenides Similar to BP, the anisotropic layered IV-VI metal monochalcogenides (MX, M ¼ Ge, Sn; X ¼ S, Se, etc.) also possess puckered orthorhombic (distorted NaCl-type) crystal structure and exhibit high Grüneisen parameters, which give rise to ultralow thermal conductivities and exceptionally high thermoelectric gures of merit. 12,75 In addition, their lowsymmetry crystal structures can lead to highly anisotropic behaviors manifested in, such as, the in-plane anisotropic carrier's mobility, [76][77][78] photoresponse, 42,79-82 and Raman intensity. 78,83,84 Conventionally, the zigzag accordion-like projection is dened as x-axis and y-axis denoting the armchair direction. Theoretical calculations have predicted the valley-dependent transport excited by linearly polarized light, 85 reversible in-plane anisotropy switching by strain or electric eld, 86 and anisotropic spin-transport properties. 87 Lin et al. demonstrated valley-dependent absorption excited by linearly polarized light. 88 Beyond that, in the bulk forms, SnSe and SnS exhibit both multi-valley features at valence bands and very low thermal conductivities, which result in high thermoelectric anisotropic properties. 12,75 Germanium disulde (GeS 2 ) and germanium diselenide (GeSe 2 ) are other typical layered materials among binary IV-VI chalcogenides. Monoclinic b-GeSe 2 is the most stable phase among all the GeSe 2 phases with relatively lower lattice symmetry and exhibits in-plane anisotropic behaviors. 56,89,90 Fig. 3 shows the crystal structure of GeS 2 and b-GeSe 2 . Unlike BP with strong interlayer coupling, the interlayer interactions of GeS 2 and b-GeSe 2 are relatively weak. 91 Owing to the high stability under ambient environment and large direct bandgap, GeS 2 and b-GeSe 2 are promising candidates for short-wave photodetection.
interactions between the layers are weak. 98 People have also investigated the in-plane anisotropic optical, electrical, and optoelectrical properties of them due to the highly anisotropic dispersions of the band structures. Both theoretical calculations and experiments have revealed that 2D SiP has a widely tunable direct band gap (1.69-2.59 eV), high carrier mobility (2.034 Â 10 3 cm 2 V À1 s À1 ) similar to BP and fast photoresponse. 44,99,100 Moreover, GeAs and GeAs 2 have been proved to be promising in thermoelectric materials by theoretical calculations and experiments. 44,99,100

Transition metal dichalcogenides (TMDCs)
TMDCs have attracted increasing research interest due to their attractive physicochemical properties. [101][102][103][104] Low level of inplane crystal symmetry can also occur in TMDCs, such as 1T 0molybdenum disulde (MoS 2 ), Td-molybdenum ditelluride (MoTe 2 ), Td-tungsten ditelluride (WTe 2 ), rhenium disulde (ReS 2 ), and rhenium diselenide (ReSe 2 ). 32,46,48,105,106 Stable metallic 1T 0 -MoS 2 (distorted octahedral MoS 2 ) can be obtained and crystallizes in the orthorhombic crystal structure (Pmn2 1 ). Unlike the trigonal prismatic (2H) or octahedral (1T) structure of MoS 2 , each Mo atom in 1T 0 -MoS 2 is linked with six sulfur atoms and connects with two adjacent Mo atoms. 107 Based on the distinct phased-induced anisotropy in 1T 0 -MoS 2 , people have investigated its anisotropic electrical transport properties and electrocatalytic performance. 46 As for Td-MoTe 2 and Td-WTe 2 , the Td phase shares the same in-plane crystal structure with the 1T 0 phase but stacks vertically in a different way as depicted in Fig. 3. Td phase MoTe 2 can be regarded as the distortion of MoTe 2 along the a-axis. From Fig. 3, we can see that each Mo(W) atom bonds to two adjacent Te atoms, leading to the formation of Mo(W) chains along the a-axis, perpendicular to the in-plane b-axis and the interlayer c-axis. Besides the in-plane anisotropic properties, Td-MoTe 2 and Td-WTe 2 are also good candidates of type-II Weyl semimetals, which present a large amount of novel physical properties to be undiscovered.
Unlike MoS 2 with hexagonal structures, group VI TMDCs with rhenium atoms (ReX 2 , X ¼ S, Se) have distorted CdCl 2 layer structure (denoted 1T 0 phase, see Fig. 3) leading to triclinic symmetry and large in-plane anisotropy. 108,109 In contrast to the 1T phase, the 1T 0 phase displays covalent bonding between the nearest Re atoms. The covalent bonded Re atoms form diamond-like pattern leading to quasi one-dimensional Re chains.

Transition metal trichalcogenides (TMTC)
Group IV transition metal trichalcogenides MX 3 are composed of transition metals M belonging to either group IVB (Ti, Zr, Hf) or group VB (Nb, Ta) and chalcogen atoms, X, from group VIA (S, Se, Te). 110,111 The MX 3 crystal structures (see Fig. 3) can be described as the stacking of individual chain units with the same orientation. Parallel neighbor chains are formed by sequential triangular prisms, where M and X atoms are respectively placed at the corners. These parallel chains in the same quasi-layer are bonded one to another with weak van der Waals interaction. Therefore, each layer of MX 3 consists of unique quasi-1D chain-like structure and contributes to its anisotropic properties. In particular, titanium trisulde (TiS 3 ) that crystallizes in the monoclinic crystal structure (P2 1 /m) with two formula units per unit cell has a direct bandgap of 1.13 eV. Apart from the in-plane electrical anisotropy, TiS 3 also exhibits ultrahigh efficiency of visible photoresponse, which makes it a suitable material for polarized photodetectors. 112 In addition, NbS 3 (triclinic structure), NbSe 3 (monoclinic structure), and TaS 3 (orthorhombic structure) also present the formation of charge density waves (CDW) and superconductivity at low temperature. 113,114 2.6 Group III-VI compounds Layered III-VI semiconductors, such as GaSe and InSe, are of wide interest due to their strong exciton peaks at room temperature absorption edge, large non-linear effect, and high intrinsic carrier mobility. [115][116][117] They open up more possibilities for applications in non-linear optics and electronics. In contrast to GaSe, gallium telluride (GaTe) crystallizes in the monoclinic system with space group (C2/m) and one-third of the Ga-Ga bonds lies in the plane of the layer, as shown in Fig. 3. These bonds are perpendicular to the b-axis and lead to in-plane anisotropic physical properties.
Bulk TlSe crystallizes in a tetragonal structure with the space group of I4/mcm. Two thallium ions, monovalent Tl + and trivalent Tl 3+ , exist in the crystal-line structure. The trivalent Tl 3+ ions form chains of tetrahedral bonds disposed along the tetragonal axis, while the monovalent Tl + ions are located between the chains and are held together by weak coupling interaction. 61

Ternary transition metal chalcogenides
Nowadays, many 2D ternary transition metal chalcogenides (i.e., Ta 2 NiS 5 , TaIrTe 4 ) have been successfully fabricated and are good candidates for excitonic insulator and type II Weyl semimetals. 49,118,119 In particular, the crystal structure of bulk Ta 2 NiS 5 is shown in Fig. 3. It crystallizes in the orthorhombic structure with the space group Cmcm. The octahedral coordinated Ta chain and the tetrahedral coordinated Ni chain form onedimensional structures along the a-axis and stack along the caxis in the order of Ta-Ni-Ta. The NiS 4 and TaS 6 units are formed by coordination with the nearest-neighbor S atoms arranged along the c axis with NiS 4 units, which are separated by two TaS 6 units. Therefore, the different arrangement of chains in the layer gives rise to the one-dimensional characteristic. 120 2.8 Group V 2 -VI 3 compounds V 2 -VI 3 compounds, such as Bi 2 Te 3 and Sb 2 Te 3 , have gained great interest and extensive research due to their striking thermoelectric properties and possibility to be topological insulators candidates. 121,122 Some other V 2 -VI 3 compounds such as Sb 2 S 3 , Sb 2 Se 3 , and Bi 2 S 3 are composed of one dimensional covalently linked ribbons stacking along the c-axis by weak van der Waals interactions. Take Sb 2 Se 3 for example; bulk Sb 2 Se 3 was recently studied as a light sensitizer in photovoltaic devices due to its narrow direct band gap of about 1.1-1.3 eV, which crystallizes in an orthorhombic structure with the space group Pnma. It consists of staggered, parallel layers of 1D (Sb 4 Se 6 ) n ribbons that are composed of strong Sb-Se bonds along the h001i direction. For the h100i and h010i directions, the ribbons are stacked owing to their van der Waals interactions. 43,123

Others
MoO 2 crystallizes in the monoclinic phase with the space group of P2 1 /c and its crystal structure is distorted to the rutile-type. 57 This is because O atoms are closely packed into octahedrons and Mo atoms occupy half the space of the octahedral void, which results in the edge-sharing MoO 6 octahedrons connected with each other and forms the distorted rutile structure. Although MoO 2 has a typical wide band gap, the Mo-Mo metallic bonds give rise to metallic transport properties. 124,125 The binary alkaline metal phosphide family MP 15 (M ¼ Li, Na, K) crystallizes in the triclinic phase with the space group of P 1. 59 It is demonstrated that the anisotropic carrier mobility ratio of single-layer MP 15 is extraordinarily large ($10 1 to 10 2 ) between the x-and y-directions. 126 MP 15 is composed of parallel units with two antiparallel rows of P tubes in one [P15] unit. In one [P15]-cell, one P atom has two adjacent P atoms and the other 14 P atoms have three adjacent P atoms, which causes a pentagonal arrangement cross-sectionally. This tubular phosphorus structure makes KP15 highly anisotropic.
As seen from Fig. 3, the low-symmetry layered 2D materials in the same crystal structure and space group exhibit similar physical properties, which is highly desirable and important for the analysis of anisotropic properties in low-symmetry layered 2D materials.

Fabrication methods
Mono-and few-layer low symmetry 2D materials could be produced by using either "top-down" or "bottom-up" approaches. Top-down approaches include mechanical or ultrasound-assisted liquid phase exfoliation from the single crystal bulk. Bottom-up approaches, whereby the low symmetry materials are grown layer by layer, involve physical vapor deposition (PVD), chemical vapor deposition (CVD), molecular beam epitaxy (MBE), as well as solution synthesis.

Bottom down
3.1.1 Mechanical exfoliation. Since Novoselov et al. successfully fabricated the rst graphene ake using Scotchtape in 2004, 1 the mechanical exfoliation method has become commonly used to get few-layer single crystalline akes of 2D materials due to the weak van der Waals interlayer interaction. In general, mechanical exfoliation is used to prepare monolayer or few-layer 2D materials by applying an adhesive tape to cleave bulk crystals repeatedly. Because of the as-cleaved clean surfaces and high crystallinity, the few-layer samples produced by mechanical exfoliation exhibit extraordinary physical properties. However, the exfoliated 2D materials still have some problems to be solved. Take BP for example; the pure exfoliated few-layer BP (see Fig. 4(a) and (b)) is relatively small sized with about 5 mm 2 size. 18 As shown in Fig. 4(c) and (d), although with the help of Ar + plasma during the exfoliated process, 127 the size of the monolayer BP can only reach 15 mm 2 , which is still far from our goals of large-scale fabrication, and well controlled morphology and edges of 2D materials.
3.1.2 Liquid phase exfoliation (LPE). One of the methods of LPE is chemically exfoliating the nanosheets of layered materials from the bulk powders with a solvent-free method by lithium intercalation. The powders are submerged in a lithiumcontaining solution such as n-butyllithium for days and lithium ions can intercalate in-between the layer space of the bulk material. Then, the nanosheets are separated when the intercalated ions are exposed to water. The schematic process is shown in Fig. 4(e). 128 However, the disadvantage of ionic intercalation is that the nanosheets might be damaged during the process. The TEM image and thickness histograms of the GeS nanosheets produced by LPE are depicted in Fig. 4(f) and (g). 129 Nowadays, ultrasonic assisted liquid phase exfoliation (UALPE) 130,131 is being utilized to provide scalable production of 2D materials. The schematic illustration of UALPE is clearly shown in Fig. 4(h). 132 In this method, the cavitation bubbles and shear force produced from the propagation of sonication waves could break the relatively weak van der Waals force between the layers without breaking the strong covalent intra-layer bindings. Therefore, this method can produce minimum defects on the as-exfoliated nanosheets due to the non-chemical and noncovalent interaction between the material and the liquid. The TEM and HRTEM images of the as-exfoliated GeSe are shown in Fig. 4(i) and (j), which show the high degree of crystallinity in the unbroken GeSe samples. 132 3.2 Bottom up 3.2.1 Physical vapor deposition (PVD) and chemical vapor deposition (CVD). 2D layered materials are foreseen to be the next-generation multifunctional materials, such as high-speed electronics and exible optoelectronics, which compels researchers to fabricate 2D layered materials at the wafer scale. Because the bottom down approach can only produce the sheets at a micrometer scale, exploring the bottom up method, which has great potential to get a sizable 2D sample, is necessary. Up to now, many kinds of 2D layered materials have been fabricated through bottom up methods, including physical vapor deposition (PVD), chemical vapor deposition (CVD), molecular beam epitaxy (MBE), and atomic layer deposition (ALD). [133][134][135][136] In contrast to bottom down techniques, the bottom up PVD or CVD methods can not only fabricate the 2D layered materials at a large scale and with controllable thickness but also maintain the extraordinary quality, which is desirable for both fundamental research and device applications. For instance, Tian et al. have developed a PVD method, whose schematic instrument is shown in Fig. 5(a), to obtain rhombic SnS nanoplates with different thickness (6-20 nm). 137 The AFM image of a 2D SnS nanoplate is shown in Fig. 5 (110) and (021) planes that were measured to be 0.58 and 0.61 nm, respectively. 90 3.2.2 Solution synthesis. As for PVD or CVD techniques, the crucial conditions for the nucleation and growth of 2D layered materials are at high temperature, controllable growth atmosphere, and appropriate epitaxial substrates, which limit the facile growth of 2D layered materials. A promising alternative to gas-phase deposition is solution-based synthetic strategies owing to its low demanded growth temperature and substratefree growth process. Therefore, one can simply disperse the as-fabricated freestanding 2D layered materials and make straightforward exible devices, assemblies, and thin lms through means such as inkjet printing, spray coating, or roll-toroll processing. [139][140][141] In addition, the sizes and thicknesses of 2D layered materials can be effectively modulated by controlling the ratio of the precursors as well. Consequently, bottom-up solution-phase syntheses of 2D layered materials lend themselves promising commercial methods. For example, researchers have successfully synthesized and separated GeS, GeSe, tellurene, and colloidal SnS nanosheets from solution. 72,142,143 Their corresponding characterizations are clearly shown in Fig. 6. The high degree of crystallinity and large-scale production demonstrate that solution-based synthetic strategy is one of the promising and desirable methods for manufacturing applications and devices in the future.

Characterization
The low-symmetry crystal structures and anisotropic band structures of highly asymmetric 2D layered materials enable their strong optical anisotropy. In order to rapidly and directly detect and characterize the optical anisotropy of the lowsymmetry 2D layered materials without destroying the samples, the azimuth-dependent reectance difference microscopy (ADRDM), angle-resolved polarization Raman spectroscopy (ARPRS), and polarization-dependent absorption spectroscopy (PDAS) are effective detection techniques. 96,112

Polarization-dependent absorption spectroscopy (PDAS)
The detection principle of PDAS is to directly measure the difference of light absorption, which makes it a reliable method for the identication of crystalline orientation. 32,[144][145][146] The scheme of the PDAS measurements is displayed in Fig. 7(a). Firstly, the few-layer 2D materials are exfoliated and transferred on a quartz substrate. Then, the incident light beam is focused onto the ake and the inverted microscope is used to collect the transmitted light. Simultaneously, a spectrometer equipped with a CCD camera can analyze the intensity of transmitted light. If the anisotropic reection can be neglected, the absorbance (A) is equal to ln(I 0 /I), where I 0 and I are the light intensities transmitted through the quartz substrate nearby the ake location and through the ake, respectively. For example, Li et al. carried out the PRAS measurements of the multilayered GeS ake by rotating the direction of the probe light's polarization from 0 to 180 . 145 The anisotropic absorption of GeS is clearly seen in Fig. 7(b) and the polar plot of absorption as a function of degree of polarization is shown in Fig. 7(c), thus presenting the linear dichroic characteristics of GeS. Since the polarization-dependent absorption spectroscopy only considers the electron-photon interaction, it is a reliable way to identify the crystalline orientations. Angle-resolved polarization Raman spectroscopy is another choice besides PDAS. However, it involves both electron-photon and electron-phonon interactions, which makes direct detection of crystalline orientation complicated.

Azimuth-dependent reectance difference microscopy (ADRDM)
The detection principle of ADRDM is to directly measure the difference in the normalized reectance (DR) between two arbitrary orthogonal directions in the surface plane (a and b) when the sample is illuminated by polarized light, which can be dened as: 147 where R a and R b are the reectance rate along a-and b-directions. The dimensionless value N(q) alters as the incident direction of linearly polarized light changes, which can be described as: where R a and R b are the reectance rate along the a-and bdirections of low-symmetry crystals, and q and q 0 denote the azimuthal angles of the incident light and a direction of the sample, respectively. By plotting the N(q) as a function of the azimuthal angle q, the crystalline orientation of the lowsymmetry crystals can be easily identied by according to the extreme values of the N(q). From the equation, we can get that the maximum and minimum RD signals correspond to the high and low reectance axes of the sample, respectively. In particular, ADRDM can collect N(q) at all the pixels in the eld and directly visualize the optical anisotropic contrast, which is especially useful for tiny sized 2D ake obtained from mechanical or liquid phase exfoliation. The scheme of ADRDM is shown in Fig. 7(d). 148 Take BP for example; a typical optical image (OM) of exfoliated BP on the Si/SiO 2 substrate is shown in Fig. 7(e). The DR/R values of BP in zone b as a function of the azimuthal angle q of the incident light is displayed in Fig. 7(f), which shows a cosine function dependent. With the ADRDM result, the BP ake has two extreme DR/R directions of 115 and 205 , respectively. All the RDM images at different angles are depicted in a color scale in Fig. 7(g). However, even though the ADRDM measurement is a reliable technique for detecting the crystal orientations, the interference effect between the interfaces must be taken into account when the BP sheet is placed on a multilayer substrate (e.g., SiO 2 /Si) because the interference effect will disturb the reection signals and cause a reversed result.

Angle-resolved polarization Raman spectroscopy (ARPRS)
Based on group theory, from Raman tensors and density functional theory (DFT) calculations, the intensity of Raman signals can be quantitatively expressed as: 149 where e i and e s are the unit polarization vectors of the incident and scattered light, and R is the Raman tensor for a certain vibrational mode. For incident light, e i ¼ (cos q, sin q, 0), where q is the angle between the incident light polarization and one crystalline orientation of the material. The schematic illustration of the angle-resolved polarized Raman spectroscopy is shown in Fig. 8(a). For the scattered light in the parallelpolarized conguration, e s ¼ (cos q, sin q, 0), while in the perpendicular-polarized conguration, e s ¼ (Àsin q, 0, cos q). Take TaIrTe 4 for example; the Raman tensors of A 1 , A 2 , B 1 , and B 2 modes can be expressed as: 49 where a, b, c, d, e, and f are the tensor elements determined by the cross section of Raman scattering. 49 Furthermore, the angledependent Raman scattering intensities of different modes can be expressed as: I k A 1 fa 2 cos 4 q þ c 2 sin 4 q þ 2ac cos 2 q sin 2 q cos 2d (5) From the equations above and the measured Raman intensities of A 1 and A 2 modes shown in Fig. 7(c) and 8(b), we can see that the intensity of A 1 mode varied in periods of 180 and 90 in parallel-polarized conguration, whereas the A 2 , B 1 , and B 2 modes have 90 variation periods. Therefore, we can deduce the crystalline orientations of the low-symmetry materials by investigating the maximum and minimum intensities of A 1 mode with a 180 variation period. The relative magnitude of matrix elements in A 1 , a > c or a < c determines whether the main axis is along the a-axis or c-axis. However, ARPRS alone cannot conrm the relative magnitude of a and c. In addition, because Raman scattering involves both electron-photon and electron-phonon interactions, the anisotropy of Raman scattering could be diverse at different detection conditions, such as the variable of laser wavelength and the thickness of sample. 96 Therefore, combining ARPRS with other techniques such as high resolution TEM (HRTEM), PDAS, ADRDM, or angleresolved DC conductance is an alternative method to conrm the crystalline orientations.

Angle-resolved DC conductance
Owing to the highly asymmetric crystal structure, the band dispersions along two perpendicular directions (e.g., G-X and G-Y) and electron-phonon scattering may be strongly anisotropic. Therefore, the effective mass of the carriers along different crystalline orientations may differ a lot. According to the deformation potential theory, the anisotropy of effective mass gives rise to the anisotropy of the carrier's mobility m and electrical conductivity s. Consequently, by using the angleresolved DC conductance measurement, one can independently determine the crystalline orientations for the lowsymmetry layered materials. For instance, the electrical anisotropy of ZrTe 5 was determined through the angle-resolved DC conductance measurement. 51 In order to eliminate the geometric factors that might affect the current ow, the measured region should be circular. 12 electrodes were patterned uniformly and spaced at an angle of 30 along the directions, as shown in the inset of Fig. 9(a). Fig. 9(a) schematically illustrates the structure of the device. DC conductance measurements across each pair of diagonal electrodes at zero back gate bias were performed and the results are shown in Fig. 9(b). The angle dependent DC conductance ts well with the measured data using the equation: where G x is the conductance along h100i direction and G y is the conductance along the h001i direction. The DC conductance along the a-axis is 1.5 times larger than that along the c-axis. Therefore, we can identify the a-or c-axis by measuring the angle-resolved DC conductance. Moreover, the researchers also measured the carrier concentration and Hall mobility along two directions at low temperatures. The carrier concentrations remain constant along the two crystalline orientations, while the hole mobility along the a-axis is around 2 times larger than the c-axis, as shown in Fig. 9(c).
In the same way, Guo et al. also investigated the angleresolved transport in multi-layered GaAs using the device shown in Fig. 9(d). 53 An obvious anisotropic characteristics can be found by the angle dependent eld-effect mobility, as shown in Fig. 9(d). The ratio of anisotropic mobility can reach as high as 4.8, which is comparable with black phosphorus and SnSe. 78,150 Besides, from Fig. 9(e), we can see that the angleresolved plot of Raman intensity at 272 cm À1 is very close to that of mobility, which means that the direction of maximum mobility (or conductance) is perpendicular to the b-direction of GeAs. As shown in Fig. 9(f), similar results can also be found in other low-symmetry layered materials such as Sb 2 Se 3 , whose ratio between maximum and minimum current is 16, which is the record of the in-plane anisotropic current (or conductance) ratio reported at room temperature. 43,54 Recently, Wang et al. discovered that the angle dependent conductance can be effectively modulated by gate bias in few- Review layered semiconducting GaTe. 36 The optical image of the device is shown in Fig. 9(g). By measuring the anisotropic DC conductance at V g ¼ À80 V, as shown in Fig. 9(h), one can identify that the maximum I ds ow is in 0 , which is parallel to the y-axis, as marked in Fig. 9(g). It is striking that the ratio of anisotropic conductance (I max /I min ) is gate-tunable and can reach as high as 10 3 at V g ¼ À30 V, as shown in Fig. 9(i). The gate-tunable anisotropic conductance is probably due to the different ratio of transmission channels in x-and y-directions at diverse gate bias. By calculating the transmission coefficient, the researchers found that at low gate voltage (À9.1 V), there is almost no x-direction transmission channel in the scattering region between the source energy level and drain energy, while a sizable y-direction transmission is observed, resulting in a large anisotropic ratio at low gate voltages. In contrast, at high gate voltage (À82 V), the transmission is comparable in both xand y-directions, thus greatly suppressing the anisotropic ratio in GaTe.
Beyond the results described above, researchers have also investigated the anisotropic carrier transport properties of other low-symmetry 2D materials by the angle-resolved DC conductance method as well. The predicted and measured anisotropic effective mass, mobility, and conductance of low-symmetry 2D materials are summarized and depicted in Table 1. We can see that the studies of anisotropic carrier transport properties of certain low-symmetry 2D materials are still missing. There is no doubt that one can fabricate anisotropic devices with higher performance if low-symmetry 2D materials with large anisotropy ratio of carrier transport were studied more deeply.
Based on the detection principles of different measurements mentioned above, we can see that PDAS and ADRDM are reliable ways to quickly and directly identify crystalline orientations without damaging the materials. However, if the 2D materials are extremely thin, the signals of PDAS are too weak to detect and resolve. In addition, the interference effect may cause a reversed result of ADRDM when the 2D materials are placed on a multilayer substrate (e.g., SiO 2 /Si). ARPRS is another choice besides PDAS and ADRDM. However, it involves both electronphoton and electron-phonon interactions, which make the direct detection of crystalline orientation complicated and difficult. Meanwhile, the anisotropy of Raman scattering is strongly dependent on the laser wavelength and the thickness of the sample. Therefore, ARPRS might be restricted to analysis when it is compared with other techniques. In the end, even though angle-resolved DC conductance measurement can effectively identify the crystal directions, the procedure of fabricating the electrodes is complicated and time consuming.

Multifunctionality
5.1 Anisotropic magneto-transport properties 5.1.1 Anisotropic magneto-resistance (MR). Investigating the magneto-transport properties of materials could provide a more comprehensive understanding of their physical properties and insights into potential applications. 151 Here, we review some recent reports on the anisotropic magnetotransport properties of low-symmetry layered 2D materials in order to explore the rich physics in them.
As the rst predicted candidate for a type-II Weyl semimetal, Td-WTe 2 has become an attractive topic owing to its exotic physical properties, such as huge non-saturated magnetoresistance (MR), chiral anomaly, and ultrahigh carrier mobilities. 13,[152][153][154] The non-saturable large MR and chiral anomaly of WTe 2 are strongly related to its Td crystal structure. Recently, Li et al. have proved that WTe 2 was indeed a type-II Weyl semimetal with topological Fermi arcs by observing the anisotropic chiral anomaly through magneto-transport measurements in one WTe 2 nanoribbon. 154 When the electric eld is applied along the k y (b-)-direction and the magnetic eld is applied along the z-(or c-) direction in the b-axis ribbon of WTe 2 , a closed Weyl orbit is formed (Fig. 10(a)) and corresponds to a trajectory in the xz-plane in real space. The temperature-dependent resistivity curves of the a-axis and b-axis ribbons shown in Fig. 10(b), which demonstrate the anisotropic transport properties in WTe 2 . A higher residual resistivity along the a-axis than that along the b-axis indicates that the average carrier mobility is smaller along the a-axis than along the b-axis (s ¼ (n + p)em), which is consistent with the transport anisotropy observed previously in bulk WTe 2 . 13,155 In order to conrm the existence of a Weyl orbit (Fermi arcs), the authors measured the MR of both the a-axis and b-axis ribbon at 2 K, as shown in Fig. 10(c), where quantum oscillations can be observed in the b-axis ribbon, while those cannot be seen in the a-axis ribbon. Therefore, it is demonstrated that the quantum oscillations came from the Weyl orbit (Fermi arcs) instead of the trivial surface states. The disappearance of quantum oscillations in the a-axis ribbon can be attributed to the strong mobility anisotropy m a < m b , which is supported by the data in Fig. 10(b). In addition, a negative MR induced by this chiral anomaly should be observed when a magnetic eld is applied parallel to the tilted direction of the Weyl cones. Inversely, the positive MR emerges when the unsaturated electric eld and the magnetic eld are mutually vertical, and the current is parallel to the Wchain. 156,157 The anisotropic MR curves measured at 2 K with Bka, Bkb, and Bkc are shown in Fig. 10(d). Although the magnetic elds are perpendicular to the current in both cases, Bka and Bkc, the positive MR ratio when Bka is almost two orders of magnitude smaller than that when Bkc is consistent with a previous observation in bulk WTe 2 . 33 Meanwhile, negative MR can be observed when Bkb. When magnetic eld tilts slightly from the E-eld direction, the absolute value of negative MR in the b-axis ribbon decreases quickly, while no negative MR can be observed in the a-axis ribbon. All these experimental data demonstrate that the Weyl points and Fermi arcs are found along the y-direction (b-axis) and are indeed induced by the chiral anomaly. This strongly anisotropic MR behavior is mainly ascribed to the strong anisotropy in the carrier mobility. 13,155 Therefore, the measured anisotropic magneto-transport properties can indeed give evidence to the band structure of some low-symmetry 2D materials.

Anisotropic nonlinear magneto-resistance (NLMR).
Apart from the linear magneto-transport in Td-WTe 2 , recently, He et al. have also investigated the spin-dependent non-linear magneto-transport in Td-WTe 2 to explore its spin-polarized bands and their interplay with Fermi surface. 158 The crystal and calculated band structure of distorted Td-WTe 2 are shown in Fig. 11(a) and (b). It is known that linear resistance is currentindependent, while non-linear resistance is current-dependent. In the non-linear magneto-transport measurements, a lowfrequency ac current was applied in the device and the second-harmonic longitudinal voltage was measured by lock-in techniques. Since there are strong local distortion of W ions in the Td-phase, as shown in Fig. 11(a), the non-linear magnetoresistance (NLMR) may be strongly anisotropic. Therefore, as displayed in the Fig. 11(c), four pairs of electrodes were patterned and spaced at an angle of 45 degree along the directions of a-, b-, ab-le-, and ab-right-axis, which were initially identied by polarized Raman spectrum. The values of linear resistivity along the b-axis are about three times larger than those along the a-axis at different temperatures, as measured and shown in Fig. 11(d). But for normalized NLMR under unit current (voltage) and magnetic eld, the sign inverses with temperature when the current is along the b-axis, while it does not inverse with temperature when the current is along the aaxis, which is shown in Fig. 11(e) and (f). Such large anisotropic NLMR has not been reported in other materials before. [159][160][161] By using the DFT calculations and tight-binding model, the authors found that the sign of the non-linear current was decided by the Fermi surface convexity and the strongly  anisotropic NLMR was attributed to the low symmetry of the Fermi surface.

Anisotropic weak localization (WL) effect. Through investigation of anisotropic magneto-transport properties, one
can not only conrm the anisotropic chiral anomaly in Td-WTe 2 but also investigate the anisotropic electronic structure in SnSe. Wang et al. have recently studied the highly anisotropic electronic structure of SnSe by combining angle-resolved photoemission spectroscopy with angular dependent magnetotransport measurements. 162 The authors have synthesized several batches of SnSe single crystals using different growth methods, including self-ux (SF) and Bridgeman (BR). On account of the different conditions during growth, the amount of Se vacancy in SnSe crystals is varied and gives rise to different doping and concentrations of carriers. By analyzing the measured SdH oscillations and the Hall effect, the obtained concentration of SF1 sample is about two times larger than that of the SF3 sample. Both samples show metallic transport from r-T curves and weak localization (WL) at low temperatures (below 50 K). The charge transport in the samples are dominated by the multivalley Fermi surfaces of the pudding-mould shaped VB1, which can result in exotic quantum phenomena in p-SnSe. Therefore, the anisotropic MR of both SF1 and SF3 samples are measured for comparison. As shown in Fig. 12(a), when the magnetic eld is perpendicular to the b-c-plane (Bkaaxis) and the current is along the armchair direction (Ikc-axis), the WL effect induced negative MRs is pronounced and does not show any sign of saturation up to 14 T.
However, such exotic behaviors of the Ikc-axis are in striking contrast to the MR behaviors when the current is parallel to the zigzag direction (Ikb-axis). The WL effect induced negative MRs are only dominant at low elds below 2 T before prevalence of positive MR, as shown in Fig. 12(b). However, the MR characteristics in the SF3 sample is less anisotropic than the SF1 sample, while the WL effect is more robust and dominant in the SF3 sample. From Fig. 12(c) and (d), we can see that the SF3 sample shows signicantly larger low-eld negative MR, which can reach as large as $ À3% at 2 T and 1.5 K, than the SF1 sample in the same conguration. But the magnitude of low-eld negative MR does not differ much in comparison with the SF1 sample. As shown in the insets of Fig. 12(a) and (c), because the hole doping in the SF1 sample is about two times higher than that of the SF3 sample, the Fermi energy level is shied downwards by about 5 meV, which reduces the separations between the two pudding-mould valleys. Thus, the momentum mismatch Dp is compensated by the dipole eld acceleration of hole carriers and the intervalley scattering is expected to be stronger in SF1 sample when Ikc. Generally, for non-relativistic fermions, the enhanced intervalley scattering gives rise to the suppression of WL because of the interruption of backscattering loops. Therefore, the WL effect induced negative MR is weakened in the SF1 sample and dependent on doping. Also, the in-plane anisotropic WL phenomena may be attributed to strong intervalley scattering along the ferroelectric dipole eld direction (c-axis). Moreover, the anisotropic and non-saturating MRs can also be observed in the BR1 sample, as shown in Fig. 12(e) and (f). have studied the anisotropic spin-orbital coupling (SOC) and demonstrated that the in-plane upper critical eld in the superconducting few-layer Td-MoTe 2 exceeded the Pauli limit in the whole in-plane directions. 163 Through the atom-resolved STEM image of few-layer MoTe 2 at a large scale, as displayed in Fig. 13(a), the Td phase of the CVD-grown MoTe 2 could be unambiguously conrmed. The MR of the 3 nm-thick MoTe 2 at 0.3 K (T ¼ 0.07T c ) with various in-plane tilted angle 4 is shown in Fig. 13(b) 4 is dened as the degree between x-axis of MoTe 2 and the magnetic eld. As we can see from Fig. 13(b), the superconducting transition moves from the lower magnetic eld to the higher magnetic eld when 4 rotates from 90 degree to 0 degree, which clearly shows the anisotropy of superconducting. Besides, the in-plane upper critical eld (H k c2 ) in this few-layer MoTe 2 also has an angular dependence at different temperatures, as shown in Fig. 13(c). The in-plane inversion asymmetry can induce strong SOC splitting and lead to effective Zeeman magnetic eld with opposite out-of-plane direction at K and ÀK valleys of the Brillouin zone. Consequently, the spins of electrons in Cooper pairs are orientated by the effective Zeeman magnetic eld and become insensitive to the external in-plane magnetic eld. 164,165 Therefore, the inplane upper critical eld of few-layer MoTe 2 can exceed the Pauli limit H p in the in-plane directions. In order to further conrm this interpretation, the band structure of the bilayer Td-MoTe 2 as well as the anisotropic spin texture calculated by the rst principle are presented in Fig. 13(d) and (e). The in-plane spin-orbit coupling (SOC) is highly anisotropic at the G pockets and the out-of-plane spin polarization dominates for the other two pockets. Moreover, the temperature phase diagrams for the superconducting state under different directions of the in-plane magnetic eld are plotted by the mean eld calculations for the pairing order parameter dependence on the in-plane magnetic eld along 4 ¼ 90 and 4 ¼ 0 directions, as shown in Fig. 13(e) and (f). The obvious difference of H c2 along 4 ¼ 90 and 4 ¼ 0 directions signies the in-plane anisotropic SOC at temperatures below T C and agrees well with the trend of the experimental data in Fig. 13(b).
From the reported results above, the anisotropic magnetotransport measurements have been proven to be powerful and useful in the study of band structures and new physical phenomena of low-symmetry layered 2D materials.

Anisotropic optoelectronic properties
The optoelectronic properties of 2D layered materials are strongly related to the band gap and light absorption coefficient, which are depicted in Table 1. Similar to the conventional semiconductors, low-symmetry 2D materials, e.g., black phosphorus, Td-MoTe 2 , tellurene, and ternary TaIrTe 4 , can also realize a wide response range across the electromagnetic spectrum because of their small bandgaps. The bandgap values of low-symmetry 2D materials and their corresponding detection range are summarized in Fig. 14.
Likewise, anisotropic optoelectronic properties can be introduced by reducing the lattice symmetry of layered materials. To explore the optoelectronic applications deeply, the detection of polarized light is exceptionally useful in several elds, including optical communication, remote sensing, and optical data storage. 166,167 Since highly asymmetric arrangement of atoms can lead to anisotropic band dispersions, further leading to the anisotropic electronic and optical properties, and thus optoelectronic properties, the materials with anisotropic optoelectronic properties are promising candidates for polarization-sensitive photodetectors. 31,106,168,169 In order to investigate the anisotropic optoelectronic properties of low-symmetry materials, two-terminal phototransistors were fabricated, as schematically shown in Fig. 15(a). 61 In the measurement, the polarized incident light was modulated by the l/2 plate and changed at a step of one certain degree. Then, the photocurrents at different polarized degrees can be obtained and plotted. Fig. 15(b) shows the typical polarization-sensitive photoresponse of the 2D TlSe ake with two-fold symmetry axes. 61 To characterize the magnitude of the linear dichromic photoresponse, a dichroic ratio g ¼ I max /I min can be introduced. The larger the value of the dichroic ratio that is measured, the more sensitivity to the polarized incident light the material exhibits. Furthermore, it is of vital importance to gure out the origin of  Fig. 15(c). 45 The authors also measured the polarizationdependent reectance contrast of the sample and compared it with the trend of polarization-dependent photocurrent, as shown in Fig. 15(d). Both the photocurrent and reectance contrast display similar polarization-dependent behavior, which manifests that the origin of polarization-dependent photocurrent is the sample's intrinsic linear dichroism. Javey et al. have investigated the polarization-dependent photoresponse of 2D Te nanoakes. 73 Surprisingly, the behavior of photoresponse as a function of polarization under the illumination of 1.5 and 3 mm laser is different, as shown in Fig. 15(e). Since Te has a direct band gap at 0.71 eV due to a strong absorption when the polarized light is along the direction of Te-chain and an indirect band gap at 0.31 eV owing to a weak absorption when the polarized light is perpendicular to the direction of Te-chain, the photoresponse of 3 mm (indirect band gap) is more anisotropic than that of 1.5 mm (direct band gap). Scanning photocurrent microscopy (SPCM) has been widely utilized for understanding the mechanism for the generation of photocurrent. In order to exclude the anisotropic collection of the photo-induced carriers, Yuan et al. have fabricated a ring-shaped electrode on the BP ake. By measuring the mapping of polarization-dependent photocurrent, they demonstrated that the mechanism of the photocurrent in BP is photothermoelectric effect and the dichroic ratio of BP was about 3.5 at 1200 nm. 31 Typical mappings of polarization-dependent photocurrent in a GeAs ake by SPCM measurement are displayed in Fig. 15(f). 170 As shown in Fig. 15(f), the photocurrent is predominantly generated near the contact between the electrodes and the sample, and has opposite sign at the two electrodes, from which we can deduce that the Schottky barriers are formed at the interface between the electrodes and the sample, and the mechanism of photocurrent is photovoltaic and photothermoelectric effect. Moreover, from Fig. 15(f), it is clearly seen that the maximum photoresponse direction under 520 nm light is along about 0 , while it differs by about 80 under 830 nm light. This interesting phenomenon may be related to the strongest absorption polarization reversing from b-axis to a-axis at 623 nm. Other low-symmetry layered 2D materials exhibit polarization-sensitive photoresponse as well. We have summarized the reported dichroic ratio of the low-symmetry layered 2D materials in Table 2.

Anisotropic thermal conductivity and thermoelectric properties
Thermoelectric (TE) devices can convert heat ow into electrical energy by utilizing the Seebeck effect and the efficiency of thermoelectric conversion is described by the gure of merit, where S, T, s, k e , and k l are Seebeck coefficient, absolute temperature, electrical conductivity, electronic thermal conductivity, and lattice thermal conductivity, respectively. From the denition of ZT, we can deduce that larger S and s with lower thermal conductivity k ¼ k e + k l should be simultaneously needed if we want to produce a TE device with outstanding performance. Nowadays, researchers have paid great attention to the investigation of k of 2D layered materials as well as their anisotropic properties. k can be measured using micro-Raman method, micro-bridge method, time domain thermo-reectance (TDTR), and time-resolved magneto-optical Kerr effect (TR-MOKE). In micro-Raman measurements, the suspending samples are transferred onto the micro-fabricated trenches or holes. The laser heats up the samples and creates a temperature gradient in the samples. Meanwhile, by measuring the Raman peak shi with temperature, we can obtain the in-plane thermal conductivity with the help of laser absorption and geometry. 171,172 The micro-bridge method was originally used to measure the thermal conductivity of onedimensional (1D) nanotubes or nanowires. 173 Recently, this method has been developed to detect the thermal conductivity of 2D materials. 174 The samples are transferred on the two suspended micro-fabricated silicon dioxide membrane islands several microns apart. One of them is the heating membrane and the other one is the sensing membrane. There are two metal resistors under the two suspended islands and a direct current is applied to the metal resistors. Consequently, the current gives rise to a temperature gradient in the sample owing to Joule heating effect. The temperature can be extracted from the resistance change of the sample and thus, we can calculate the thermal conductivity of the sample. The principle of TDTR method is to measure the thermo-reectance response as a function of delay time between the arrival of the pump and probe pulses on the sample surface. The modulation of the pump beam at rf frequencies is used for lock-in detection of the thermoreectance signal and to generate useful heat accumulation effects. The in-phase signal from the lock-in outputs V in is approximately proportional to the temperature difference induced from pump pulse and the out-of-phase signal V out is approximately proportional to the imaginary part of the temperature oscillations of the pump beam. We calculate the ratio between V in and V out voltages, and t them to a heat transfer model, from which the unknown thermal conductivities can be obtained. The experimental setup of TR-MOKE method is quite similar to TDTR. TR-MOKE method is used to detect the temperature dependence of Recently, many groups have investigated the thermoelectric behaviors of BP. Since the puckered crystal structure of BP results in a lower lattice anharmonicity and larger group velocity along the zigzag direction than the armchair direction, the thermal conductivity (k) along the armchair direction is several times smaller than that along the zigzag direction, as summarized by Kang et al. in Fig. 16(a). [175][176][177][178] The difference in the measured in-plane anisotropic thermal conductivity may be related to the different measuring methods and the easily degenerated surface of BP. 175,179,180 As seen in Fig. 16(a), the 3D anisotropic thermal conductivities also have thickness dependence in the specic region, which indicates the efficiency of surface or boundary scattering. 175 More results have proved that when the vibrations or the propagation directions of phonons are out-of-plane, the scattering is strongly enhanced, which results in the lowest thermal conductivity. But when vibrations and propagation directions of the phonons are in-plane (zigzag or armchair axis), the phonon relaxation time is almost the same. Thus, the phonon relaxation time only contributes to anisotropy in the through-plane thermal conductivity but not the in-plane thermal conductivity. 176,180 Kang et al. have also developed a method to reversibly modify the thermal conductivity of BP by Li ion intercalation. The thermal conductivities of pristine BP and Li 3 P are found to be highly anisotropic, as shown in Fig. 16(a), which shows that Li ion intercalation covers a remarkably large thermal conductivity tuning range. 178 Recently, Zhao et al. have reported that the ZT values of SnSe crystal are extremely high owing to its ultralow lattice thermal conductivity for the distinctive anharmonic structure of SnSe. 12,14 Similar to BP, SnSe also has in-plane anisotropic ZT values (2.6 and 2.3 at 950 K along the b and c axes, respectively) and thermal conductivities along different axes as shown in Fig. 16(b). 14,178 Strikingly, when SnSe is hole doped with Na, the values of thermal conductivities along three directions are enhanced due to the multiple valence band maxima that lie close together in energy by liing the Fermi level deep into the band structure. Chen et al. have explored the in-plane anisotropic thermal conductivity of Td-WTe 2 akes with different thickness using micro-Raman spectroscopy method. 181 The extracted thermal conductivity of the WTe 2 samples with different thickness are shown in Fig. 16(c). Especially for the 11.2 nm thick fewlayered WTe 2 , the thermal conductivity along the zigzag direction, k zigzag ¼ 0.743 W m À1 K À1 , is 16.3% larger than that along the armchair direction, k armchair ¼ 0.639 W m À1 K À1 , thus showing a strong anisotropy in the thermal conductivity. But as the thickness of WTe 2 increases, the anisotropy of the in-plane thermal conductivity decreases due to the rise of mean free path along the armchair and less phonon-boundary scattering, as shown in Fig. 16(d).
As a typical low-symmetry 2D material, ReS 2 also has anisotropic thermal conductivity, which has been studied using the TDTR method by Jang et al. 182 They found that the thermal conductivity along the Re-chains was larger than that along the   Fig. 16(e) plots a summary of 2D materials whose in-plane and through-plane thermal conductivities have been experimentally measured in the bulk limit at room temperature. From Fig. 16(e), we can see that the thermal conductivity of ReS 2 has a remarkably high anisotropy (130 AE 40 and 90 AE 30) for the two in-plane directions.
It is well known that heavy elements are preferred for thermo-electrical devices with high performance due to the enhanced phonon scattering and lower thermal conductivity. 188 Therefore, it has been predicted and experimentally demonstrated that Te is a good candidate as a thermoelectric material due to its high electrical conductivity and low thermal conductivity. 69,70,74 Peide D. Ye et al. recently fabricated a stateof-the-art thermoelectric device based on few-layered 2D Te akes. 15 The Seebeck coefficient of few-layered Te can be found to be 413 mV K À1 . Then, the thermal conductivity along the 1D chain direction of a similar suspended 2D Te ake is measured by micro-Raman spectroscopy and can be obtained to be about 1.50 W m À1 K À1 . Hence, the calculated value of ZT for fewlayered Te is about 0.63 at T ¼ 300 K.
Many groups have theoretically predicted that BP has excellent thermoelectric properties. [189][190][191] For example, Zhang et al. calculated and concluded the peak ZT of 1.1 and 0.6 with high electron and hole doping at 800 K. 190 However, few reports have been aimed at investigating the TE properties of BP in experiment. Yu Saito et al. have currently measured and successfully tuned the Seebeck coefficient of multilayered BP by gate voltage. 192 The maximum of S can reach as high as 510 mV K À1 at 210 K when BP is in the hole-depleted region, which is much higher than the reported bulk single crystal value of 340 mV K À1 at 300 K. 189 Zhao et al. have previously proved that single crystals of p-type SnSe exhibited an extremely high ZT of $2.6 at 923 K along crystallographic b-direction. 12 Lately, n-type Br-doped SnSe single crystals have exhibited a striking ZT of 2. 8. 193 Other low-symmetric 2D materials with low thermal conductivities and highly anisotropic transport properties also show potential promising thermoelectric applications. Tasuku Sakuma et al. recently measured the thermoelectric power S ¼ À530 mV K À1 and calculated ZT ¼ 0.0023 for quasi-onedimensional TiS 3 microribbon. 194

Ferroelectric and piezoelectric properties
Realizing ferroelectricity and piezoelectricity in 2D layered materials is intriguing for fundamental science and potential applications (e.g., non-volatile memories, generators, and sensors). Ferroelectricity is a characteristic of certain materials that have a spontaneous electric polarization that can be reversed by an external electric eld. Up to now, ferroelectricity has been successfully detected in monolayer SnTe, few-layered a-In 2 Se 3 , and CuInP 2 S 6 akes by different methods of measurement. [195][196][197][198][199] The ferroelectric behavior usually originates from the breaking of the structural centrosymmetry in the polarization direction. However, there are few reports on achieving ferroelectricity in low-symmetry layered materials, which is extremely impressionable to strain tuning with novel anisotropic properties. Recently, Fei et al. revealed that the unique ionic-potential anharmonicity can induce spontaneous in-plane ferroelectricity in monolayer group-IV monochalcogenides MX (M ¼ Ge, Sn; X ¼ S, Se). They deduced that the ferroelectricity in these materials was robust and their Curie temperatures are all above 300 K. The spontaneous electric polarization is in the order of 10 À10 C m À1 . 200 The monolayer b-GeSe with puckered lattice structure was also predicted to be a 2D ferroelectric material by Guan et al. The in-plane spontaneous electric polarization is about 0.16 nC m À1 , which is comparable to that of monolayer SnTe. The intrinsic Curie temperature T c is calculated to be 212 K by using Monte Carlo simulations. 201 Fei et al. have also found that two-or three-layer metallic Td-WTe 2 exhibits spontaneous out-of-plane electric polarization that can be switched by gate in experiment. 202 The authors estimated that the polarization intensity was about 2 Â 10 À4 C m À2 , which was about three orders of magnitude lower than that of classic ferroelectric BaTiO 3 . 203 Moreover, the researchers also found that the ferroelectric switching characteristics can be effectively tuned by the gate bias. The above observations are practical for ferroelectric applications and may be relevant to novel physical phenomena such as extreme and anisotropic magnetoresistance, 37,155 a polar axis, and Weyl points. 204,205 Since the ion-displacement of compounds can induce the dipole moment, most of the reported 2D ferroelectric materials are compounds. In comparison, elemental materials are predicted to have no ferroelectricity because there is no electronegativity difference in them. However, Wang et al. recently predicted that 2D few-layered tellurium is a stable ferroelectric material at temperature up to 600 K and the in-plane electric polarization is about 0.16 nC m À1 . 206 The origin of polarization is the in-plane ion-displacement due to interlayer interactions between the lone-pairs. Piezoelectric effect is the electric charge accumulated in the material in response to applied mechanical stress and has been used in several devices such as piezoelectric-gated diodes, eld effect transistors, and nano-sensors. 207 Electric polarization is caused by broken symmetry and exists in most of the noncentrosymmetric materials such as ZnO and Pb(Zr x Ti 1Àx )O 3 (PZT). 208 Recently, two dimensional piezoelectric materials have attracted tremendous interest because of their good ability to endure enormous strain. It should be noted that there are some materials in which the inversion symmetry can be preserved in the 3D forms but broken in the 2D ones. 209,210 For example, bulk MoS 2 is not piezoelectric but Wang et al. have proved that the monolayer MoS 2 ake can produce a piezoelectric voltage of 15 mV and a current of 20 pA with 0.53% strain. 102 For the lowsymmetry 2D materials, Fei et al. have predicted that there exists giant anisotropic piezoelectric effects in monolayer group-IV monochalcogenides. By virtue of their unique puckered C 2v symmetry and electronic structure, the piezoelectric coefficients of the monolayer group-IV monochalcogenides are surprisingly one to two orders of magnitude larger than other 2D piezoelectric materials such as MoS 2 , hexagonal BN (h-BN), and GaSe. 209,[211][212][213] However, as far as we know, no experimental results have been reported for measuring the piezoelectric voltage or current in monolayer group-IV monochalcogenides.
The attempt of exploring anisotropic piezoelectric polarization in low-symmetry 2D materials may offer new possibilities to tailor in-plane anisotropic piezoelectric response in nanotechnology and new accesses for harvesting of energy, which can be further used for self-running nano-devices without using additional energy.

Applications
By taking advantage of multifunctionality of low-symmetry 2D layered materials, applications with anisotropic properties can be manufactured. Owing to the ambipolar functionality and high-mobility of BP, Zhu et al. have demonstrated high performance exible amplitude-modulated (AM) demodulator, ambipolar digital inverter, and frequency doubler. 214 The BP AM demodulator is a single-transistor circuit, which can convert RF signal to audio signal. The optical image and schematic of exible BP AM demodulator are shown in Fig. 17(a) and (b). The FFT output signal spectrum for 100 mV input peak-to-peak carrier amplitude at 100% modulation index is shown in Fig. 17(c) and the authors further demonstrated that the BP AM demodulator worked well. Besides, BP ambipolar digital inverters and a frequency doubler are successfully manufactured based on the ambipolar transport characteristics and high drain current modulation. As shown in Fig. 17(e), a BP push-pull amplier was fabricated. Two identical bottom gate transistors share the same drain as the output and the bottom gate as the input. The amplied inverted signal with an output/ input voltage gain of $1.68 can be observed in Fig. 17(e). The frequency doubler is also widely used in analog circuits due to the low energy consumption. Compared to graphene, BP can offer lower power and higher power efficiency because of its lower DC power dissipation and off current. As shown in Fig. 17(f), the BP transistor was biased to realize symmetric transfer characteristic near the minimum conduction point and the output sinusoidal frequency was doubled with a conversion gain of 0.72. Similarly, Liu et al. have also fabricated a digital inverter based on the integration of two separated ReS 2 FETs along two orientations, as shown in Fig. 17(d). 35 The gain of the inverter is dened as |dV out /dV in | and can reach as high as 4.4 when V DD ¼ 3 V, which is comparable to the MoS 2 inverters. 215,216 Because of the gate-tunable anisotropic resistance in fewlayered GaTe, Wang et al. have manufactured a prototype anisotropic memristor based on GaTe akes with few-layered graphene as the oating gate. 36 The schematic view of the device is shown in Fig. 17(g). Since the anisotropic transport characteristics vary largely in GaTe, the hysteresis memory curves along the x-and y-directions differ a lot, which show a clear window of memory. When the programming gate sweeps from 0 V to À20 V, the device is in the 'off' state along both the xand y-directions. When the erasing gate sweeps from 0 V to 20 V, the device stays in two different 'on' states along the x-and ydirections. Therefore, the anisotropic memristor is realized by erasing and programming pulses measured in the y-direction, as shown in Fig. 17(h) and has great potential in directionsensitive data storage.
Optoelectronic devices rely on light-matter interactions and can convert light into electrical signal or vice versa. 218,219 Optoelectronic devices including detectors, lasers LEDs, solar cells, and optical switches are widely used in low-loss optical ber communications, power generation, and military measure systems. For low dimensional and exible photodetectors, 2D layered materials should exhibit high responsivity, large detectivity, and fast response time. 111,[220][221][222] Unlike graphene, many other 2D layered materials have a band gap and large absorption coefficient, which are benecent for high performance photodetectors. In addition, polarization-sensitive photodetectors based on low-symmetry layered materials are also desirable in optical communication, remote sensing, and optical data storage. 166,167 Apart from the photodetectors based on lowsymmetry materials mentioned above, similar polarizationsensitive photodetectors have also been investigated, including wide-band-gap ultraviolet photodetectors (e.g., GeS 2 ), visible-light photodetectors (e.g., ReS 2 , ReSe 2 , SnSe, SnS, GeSe, GeS 2 , GeSe 2 , GeP, SiP, TiS 3 , and Sb 2 Se 3 ), and narrow-band-gap infrared photodetector (e.g., Td-MoTe 2 , Td-WTe 2 , and TaIrTe 4 ). For comparison, more performance details of a series of 2D-based polarization-sensitive photodetectors are listed in Table 2. Take the high-performance polarization-sensitive photodetectors based on BP for example. 217 As shown in Fig. 17(i) and (j), the photo-response time is ultrafast, which is measured to be about 12.4 ms. The specic detectivity (D*) can be optimized by adjusting the thickness of BP to maximize the absorption and minimize the dark current. The maximum value of D* can reach as high as 6 Â 10 10 jones at room temperature, Fig. 18 Summarization of the values of photo-responsivity and dichroic ratios for low-symmetry 2D materials. The data are extracted from Table  2.
which is about one order of magnitude higher than commercial mid-wave infrared detectors operating at room temperature. As another critical index, the dichroic ratio of photocurrent can be obtained in Fig. 17(k). One can nd that the polarization ratio (dichroic ratio) between the two crystal orientations of BP at mid-infrared wavelengths is larger than 100, which is larger than all the other low-symmetry 2D materials. This value is limited by experimental instruments and approaches the extinction ratio of the polarizer used in this study. Since both photo-responsivity and dichroic ratio are the most important indices for the polarization-sensitive photodetectors, here, we have summarized these two values of some low-symmetry 2D materials. As we can see from Fig. 18, both photo-responsivity and dichroic ratios should be high for extraordinary polarization-sensitive photodetectors, which provides guidance for next-generation promising optoelectronic devices with inplane anisotropy.

Conclusions and outlook
Here, we have summarized the recent achievements in lowsymmetry 2D layered materials and their anisotropic properties, including anisotropic electronic, optoelectronic, magnetic transport, thermoelectric, piezoelectric, and ferroelectric properties, resulting from their anisotropic structures and band structures. On account of the intriguing anisotropic electronic properties, the applications have been fabricated and developed, such as in-plane anisotropic FETs, 63,243 anisotropic oating gate memristors, 36 digital inverters, 35,244 memristors, and polarization-sensitive photodetectors. 31,32,45,53,56,228 However, there are still many problems to be resolved to attain a comprehensive understanding of the properties of lowsymmetry 2D materials and to realize their full potential in multifunctional elds. The potential opportunities and challenges are listed as follows: (1) more work is needed to achieve low-symmetry 2D materials at a large scale. Although a few lowsymmetry 2D materials (e.g., SnS and GeSe) have already been manufactured by CVD and PVD methods, still a lot of fewlayered low-symmetry 2D materials have only been made by exfoliation, which limits the development of fabrication devices with anisotropic properties. (2) Since the anisotropic ratio of anisotropic 2D materials is still very low (for most of them, it is less than three), exploring new materials and techniques to enhance the in-plane anisotropy of 2D materials is very essential and promising for future anisotropic devices. (3) Searching new methods to effectively modulate and enhance the in-plane anisotropic ratio. Recently, Wang et al. have discovered that the in-plane anisotropic ratio of resistance in few-layered GaTe can be greatly enlarged by tuning the gate voltage. 36 But for other low-symmetry 2D materials, whether the gate voltage can also modulate the in-plane anisotropic properties is still unknown. (4) The predicted thermoelectric, piezoelectric, and ferroelectric properties and their anisotropy in some of the low-symmetry 2D materials are still needed to be conrmed and explored by experiments. (5) Although many researchers have deeply investigated the properties of heterostructures based on lowsymmetry 2D materials, the isotropic/anisotropic and anisotropic/anisotropic 2D stacked heterostructures require more in-depth study to elucidate the unique properties and upgrade the device performance. (6) Since Wu et al. have predicted that moiré bands of twisted transition metal dichalcogenide homo-bilayers can be topologically non-trivial, 245 new physical properties such as quantum spin Hall effect and superconductivity may be observed in twisted bilayer of some particular low-symmetry 2D materials. Overall, the recent ndings concerning anisotropic electronics indicate a broad promise in multifunctional applications.

Conflicts of interest
The authors declare no competing nancial interests.