Temperature-Dependent Phase Variations in Van Der Waals CdPS₃ Revealed by Raman Spectroscopy

Abstract

In addition to graphene, the transition metal dichalcogenides, black phosphorus and multiple other layered materials have undergone immense investigations. Among them, metal thiophosphates (MPS_x) have emerged as a promising material for various applications. While several layered metal thiophosphates with general-formula MPS_x have been scrutinized extensively, van der Waals (vdW) CdPS₃ has been overlooked in the literature. Here we report on the extensive Raman scattering of layered CdPS₃, showing structural phase transition at a low temperature. The emergence of multiple new peaks at low frequency and a significant shift in peak position with temperature implied a probable change in crystal symmetry from trigonal D_3d to triclinic C_i below the phase transition temperature, T_K~180 K. In addition, we also showed a p-type performance of CdPS₃ FET fabricated using Au electrodes. This work adds CdPS₃ to the list of potential layered materials for energy application.

1. Introduction

The discovery of intrinsic van der Waals two-dimensional (2D) material has attracted immense interest from the material science and nanomaterials community [1,2,3,4,5]. Extensive investigations have been carried out on the electrical, optical, mechanical and chemical properties of layered materials, which have witnessed applications in multidisciplinary fields [2,6,7,8]. In addition to numerous optoelectronic devices [9,10], like transistors, diode, LEDs, photovoltaics and solar cells, layered materials have also been applied to catalytic activity for future roles in energy alternatives, (bio)sensing systems and advanced electronic devices [11,12,13,14]. With the advent of new synthesis techniques and emergence of various materials in the vdW 2D family, attention has recently been turned to other 2D materials besides graphene [15,16,17], transition metal dichalcogenides (TMDs) [18,19,20] and black phosphorus [12,21]. The search for enhanced electrical, optical and magnetic properties has led to the manifestation and synthesis of metal phosphorus chalcogenides with a general formula of MPX_Y [22].

The family of MPX_Y has a layered structure possessing the van der Waals gap. M represents the transition metal (TM) atom (e.g., Mn, Fe. Co, Ni and Cd), X is a chalcogenide atom (either S or Se) and Y is the number of chalcogen atoms [23]. Several studies have reported on the alluring electronic, anisotropic, magnetic, dielectric, structural and optical properties of these materials, which can be easily exfoliated down to monolayer thickness [22,24,25,26,27]. This is because they are held together by weak van der Waals forces, similar to other layered compounds, and can be thinned down to fewer sheets [28]. Their structures are also similar where the metal and phosphorus atoms within each layer are sandwiched in between the chalcogenide atoms [23]. MPX_Y compounds have vital intrinsic properties that are useful in multifarious processes. Their layered structures have manifested anisotropic properties, giving them the ability to act as host lattices for intercalation compounds [29]. MPX_Y compounds also tend to possess unusual intercalation–reduction behavior, in addition to incipient ionic conductivity, influencing their usage in Li-ion batteries [30,31], gas storage [32] and various other photo-electrochemical reactions [33]. They also tend to have a wide range of band gaps that suggest possible optoelectronic applications in a broad wavelength range.

According to computational results and experimental verification, MPX_Y compounds are the most popular functional materials because of their intermediate bandgaps from 1.3 to 3.5 eV, implying their enhanced light absorption compared to transition metal dichalcogenides (TMDs) [34,35]. Most interestingly, magnetic moments have been found to be localized in the transition metal ions that form a honeycomb network structure, giving rise to antiferromagnetism (AFM) beyond a certain transition or Neel temperature, T_K [36]. For instance, MnPS₃ shows AFM beyond 78 K, while FePS₃ and NiPS₃ show AFM beyond 116 K and 155 K, respectively [37,38,39]. However, these were only manifested when the transition atom was magnetic [40]. Furthermore, it has been demonstrated that the magnetic properties can also be tuned via alloying with nonmagnetic transition metals [41]. Raman scattering has been promoted as the pivotal and significant tool to study such phase transitions in these materials, where clear signatures of phase transition emerges beyond T_K in form of new phonon peaks or a large positional shift in the existing phonon modes [38,39,42,43].

While a majority of the members in the MPX_Y family have been shown to have AFM and have been studied extensively, some essential compounds, especially the nonmagnetic ones, have been mostly overlooked in the literature. CdPS₃, for example, is a diamagnetic compound due to its core electrons, but it does not manifest AFM due to the Cd ion being nonmagnetic [44]. It generally belongs to the C2/m space group with monoclinic symmetry, which results from a small distortion in the rhombohedral structure (trigonal space group R3(C²_3i)) but can take other phases or symmetry at room temperature when synthesized in various forms. Earlier investigation by Lifshitz et al. [45], using electron paramagnetic resonance (EPR) and X-ray diffraction on Mn-doped CdPS₃ crystals, revealed a phase transition from monoclinic to orthorhombic, around 260 K. With the decreasing temperature, they found that the crystal symmetry became higher, and there was a loss of unique axis corresponding to the monoclinic phase.

Some researchers have explored the phase change in the CdPS₃ compound from C2/m to R3 through the introduction of external pressure and analyzing pressure-dependent Raman spectra [46]. Others have investigated the interlayer coupling in a CdPS₃ nanosheet-based membrane through conductivity and transport measurements [47]. Computational and theoretical efforts have also been instigated with an aim to view the molecular structure and bond vibrations [48]; however, it was not substantiated through intensive experimental analysis. Regardless, there is no comprehensive reporting on the nonlinear optical properties of layered CdPS₃, despite its being an important compound for future applications in energy storage devices and the development of biosensing system [44]. It is essential to completely understand the vibrational properties, phase segregation and lattice dynamics incurred by the thermal effects on this material, as they can have vital implications for future device fabrication and advanced manufacturing using CdPS₃. Low-frequency Raman spectroscopy provides an intriguing non-evasive technique to probe such properties and comprehend the phase structure of this material. The aim of this study mostly involves providing a detailed analytical investigation on the low- and high-frequency vibrational dynamics of CdPS₃ and phase segregation with temperature. In this work, we used high-resolution Raman spectroscopy to characterize bulk CdPS₃ to uphold its precise phase transition from trigonal to monoclinic. The thermal dependence of the Raman spectrum clearly indicates a phase transition temperature of T_K~178–180 K. We analyzed the peak position and vibrational intensity of several Raman modes that correlate well with symmetric change to monoclinic phase. Furthermore, we also fabricated a field effect transistor (FET) by using thin layers of CdPS₃, which showed a p-type response and good hole mobility, allowing its application in major electronic devices.

2. Results and Discussions

CdPS₃ belongs to a family of quasi two-dimensional materials. Figure 1a shows the schematic representation of the crystal structure, and Figure 1b,c show the top view and side of the crystal. It is isostructural, with the transition metal atom, Cd, forming a honeycomb lattice in planes that are weakly bound by van der Waals forces [P₂S₆]⁴⁻. An anion sublattice appears within each layer, and the honeycomb arrangement of the transition metal ion is distributed around the [P₂S₆]⁴⁻ bipyramids such that TM atom is surrounded by six sulfur atoms. These sulfur atoms are further connected to two phosphorus atoms above and below the TM plane. Single crystals of CdPS₃ were sourced commercially (2D semiconductors) and exfoliated mechanically (see Section 4). The samples were then transferred to Si/SiO₂ substrate for further optical characterization. Figure 1d shows a typical CdPS₃ bulk sample on a SiO₂ substrate, exfoliated mechanically and transferred using the vdW transfer technique (see Methods section). At first, we confirmed the sample stoichiometry by using scanning electron microscope (SEM) and energy-dispersive X-ray spectroscopy (EDX). Figure 1e shows the SEM image of a section of the bulk sample in Figure 1d, showing a relatively smooth surface. The corresponding energy-dispersive spectrometry (EDS) spectrum is shown in Figure 1f–h, highlighting the uniform elemental composition of Cd, P and S in the layered material. The Cd-L line and K line of P and S can be easily excited using 5 kV excitation energy during SEM and EDS measurements. Figure 1i shows the EDS spectrum, indicating the peak energies of each of the constituent elements. Hence, within the limitation of the technique, the samples examined in this work does not exhibit detectable unwanted impurities.

Raman measurements were consequently carried out on bulk CdPS₃ samples, using a commercial micro-Raman system. Figure 2a,b show the temperature dependence of Raman spectra from 55 to 95 cm⁻¹ and from 100 to 600 cm⁻¹, respectively, from RT down to 77 K. As witnessed, multiple peaks can be identified both at RT and 77 K. In the low-frequency regime, only the P₂ peak is recorded around 79.9 cm⁻¹ at RT. With a decreasing temperature, more new phonon modes start to appear, marked by P₀, P₁, P_2a, P_2b and P₃. This signifies that the material is undergoing substantial changes as the temperature is lowered. In the high-frequency regime, peaks P₄, P₅, P₆, P₇, P₈ and P₉ can be detected from RT, but the visual examination manifests a clear change in the peak position and vibrational intensity. Such a change in mode position and intensity can also be observed for low-frequency peaks. To clearly emancipate the transition temperature, T_K, we carried out Lorentzian fitting of the experimental curves to disintegrate and analyze each peak separately. Figure 2c shows some representative Raman spectra with fitting curves, with particular focus around T_K. Peak 2 splits into P_2a and P_2b at around ~178–180 K, manifested meticulously by Lorentzian fitting curves, implying a clear phase transition. Figure 2d shows the full width at half maximum (FWHM) of peak P_2b, which decreases progressively with the temperature, but around T_K, it shows a sudden augmentation before decreasing again. Figure 2e shows the normalized intensity of peaks P₀, P₁ and P₃ in the low-frequency regime, which suddenly increases in strength around 178–180 K. This clearly signifies that the material is undergoing a phase transition around that temperature. This is in slight contradiction with the previously reported value of 228 K for CdPS₃ single crystals deduced from X-ray diffraction [49]. Figure 3 shows the summary of peak position and intensity of other high-frequency Raman modes, thus further reaffirming the phase transition of CdPS₃. The peak intensity of P₄, P₆, P₇ and P₈ increases gradually with the decrease in temperature; around T_K, it shows sudden nonuniformity. Similarly, the peak position of the higher Raman modes shows a sudden blue shift around T_K, and then the modes become insensitive with temperature.

Such an observation in Raman spectra can be correlated with the phase transition of CdPS₃ from trigonal at a high temperature to monoclinic at a low temperature. The phonon mode P_2a appears only at the low temperature and vanishes above ~180 K. This soft mode is strongly related to the structural phase transition. Liftshitz et al. [45] observed a hysteresis loop in CdPS₃ single crystals at a higher temperature and suggested the co-existence of multiple phases before the first order phase transition occurs. However, we observed that the P_2a mode disappears completely above ~180 K; in principle, first-order phase transition should not nullify the frequency of the mode at the transition temperature [50]. Hence, this indicates that the phase transition is more of a displacive type. Generally, it is plausible to observe more Raman peaks in the low symmetric phase than in the high symmetric phase. The appearance of new mode P_2a and the increase in intensity of other low-frequency Raman modes, like P₀, P₁ and P₃, suggest that the crystal symmetry decrements with the reduction in temperature through the phase transition.

The high-temperature phase of CdPS₃ consists of degenerate phonon modes, as denoted by the two- or three-dimensional representation, because the peak P₂ splits into two peaks at a low temperature [50]. Since CdPS₃ has a layered structure, peak P₂ can be assigned to the doubly degenerate Raman active mode in the high-temperature phase. Splitting into two peaks can be denoted as one-dimensional representations in the low-temperature phase. Hence, the symmetry of the crystal should be reduced at a low temperature. Lifshitz et al. [45] claimed, in an earlier investigation, that the high-temperature phase would be monoclinic; however, there are no degenerate modes that can be denoted by the two- or three-dimensional representations, thus contradicting their conclusion.

Alternately, our observation can be explained well considering the fact that the high-symmetry phase belongs to trigonal symmetry, as claimed by Covino et al. in an earlier report [51]. All MPX_y crystals have a layered structure due to the weakly interacting van der Waals interaction, and the symmetry of a single layer belongs to the trigonal symmetry D_3d in almost all members of MPX_y family. Also, two formula units are present in the primitive unit cells. Assuming that the crystal symmetry of CdPS₃ belongs to trigonal symmetry, D_3d, the phonon modes at Γ can be decomposed into the following representations [50,52]:

Γ = 3A_1g + 2A_2g + 5E_g + A_1u + 4A_2u + 5E_u

A_1g and E_g modes are Raman active among these, and their Raman tensors can be denoted as follows:

$$A_{1g}:\left(\begin{array}{ccc}\alpha & 0& 0\\ 0& \alpha & 0\\ 0& 0& \beta \end{array}\right)E_g :\left(\begin{array}{ccc}\Upsilon & 0& 0\\ 0& -\Upsilon & \delta \\ 0& \delta & 0\end{array}\right), \left(\begin{array}{ccc}0& -\Upsilon & -\delta \\ -\Upsilon & 0& 0\\ -\delta & 0& 0\end{array}\right)$$

One A_2u mode and E_u mode can be considered to be acoustic modes, while A_2u and four E_u modes are infrared active. Hence, they cannot be detected in the Raman spectra extracted using a green laser. Therefore, at RT, the high-frequency Raman active modes can be assigned to the vibration of P₂S₆ units with D_3d symmetry [44]. Modes P₆ and P₈ at ~248 cm⁻¹ and ~378 cm⁻¹, respectively, can be assigned to A_1g; meanwhile, modes P₄, P₅, P₇ and P₉ at 126.5 cm⁻¹, 224.8 cm⁻¹, 273.42 cm⁻¹ and 563.30 cm⁻¹, respectively, can assigned to E_g [44]. Such phonon modes originating from the symmetry of P₂S₆ units are also common in compounds of the MPS₃ family [53]. Infrared spectroscopic analyses of CdPS₃ single crystals on the basal layer planes were carried out by Mathey et al. [54]; they observed three strong absorption bands at 193 cm⁻¹, 252 cm⁻¹ and 564 cm⁻¹, and they also noted another band at around 110 cm⁻¹ with dual peaks that was regarded as a superposition of a strong and weak peak. These were assigned to E_u modes that are only detectable under infrared excitation. Furthermore, Mathey et al. observed some Raman modes at 310 and 449 cm⁻¹ that correlated with A_u modes. Hence, these signatures were absent in our Raman study. Therefore, based on the current analysis, it is reasonable to expect that the high-temperature phase of CdPS₃ is D_3d and belongs to trigonal symmetry (R3). Peak P₂ can be assigned to the E_g mode in the trigonal symmetry, which was split into two peaks in the low-temperature phase. If we assume that the Landau theory is applicable to the present phase transition of CdPS₃, then it can be regarded as a weakly first-order variation. Hence, it is plausible to assume that the crystal point group of the low-temperature phase is one of the subgroups of the high-temperature phase. Considering the appearance of new Raman modes, enhancement of vibrational intensity and significant blue shift of the phonon modes, it can be suggested that crystal symmetry changes from the trigonal symmetry, D_3d, to triclinic symmetry, C_i. This transition occurs at around T_K = ~178–180 K. Tomoyuki Sekine et al. [50] observed some similar transition on CdPS₃ single crystals and mentioned the possibility of other monoclinic-symmetry C_2h, C₂ and C_s at a low temperature. However, we also observed peaks P₀, P₁ and P₃ at a low temperature, which has not been reported before. It is reasonable to assume that the low-temperature phase forms a complex phase structure consisting of one of the monoclinic symmetries (C2/m). This can make some Raman modes active at 61.37 cm⁻¹, 70 cm⁻¹ and 93.3 cm⁻¹. As deduced before, high-frequency phonon modes originate from the internal vibrations of P₂S₆ octahedrons, while the low-frequency Raman modes are mostly due to the vibrations of the transition metal ion, Cd. Hence, the structural phase transition occurs as a result of the displacement of the Cd ion. This can be related to the observation of new peaks (P₀, P₁ and P₃) and the increase in their intensity below the transition temperature.

Finally, we fabricated the field effect transistor (FET) from the vdW-layered CdPS₃. The schematic structure of back-gated FET is shown in Figure 4a. SiO₂ was used as the dielectric, and Au contacts were used as the source and drain terminals. Highly doped Si provides a good source for back-gate voltage. To fabricate the FET structure, conventional optical lithography was deployed to pattern the Au electrodes on the Si/SiO₂ substrate. The process involved UV exposure after spin-coating with photoresist metal evaporation, using e-beam evaporation and, finally, lift-off to obtain the patterns. The 2D material was then transferred mechanically using the vdW transfer technique, such that the source and drain electrodes were in contact with the sample. An optical image of a typical FET is shown in Figure 4b. The channel length is 2 µm, and a large-sized bulk sample is present in between the channels. The output characteristics of the device, source-drain current (I_ds) as a function of source-drain voltage (V_ds) and transfer characteristic, drain current as function of back gate (V_g) is shown in Figure 4c,d. All electrical properties of CdPS₃ back gated FET were measured at room temperature in air. The current increases with increasing negative back gate voltage, suggesting p-type behaviors [55]. P-type semiconductors are indispensable in many complex digital electronics, such as inverters, logic gates, memory devices and integrated circuits (ICs) [56]. The nonlinear behavior may arise from the Schottky barrier contact with Au metal [57]. The field effect mobility can be extracted from the I_ds-V_g curve, using the following expression:

$$\mu =\left[\frac{d{I}_{ds}}{d{V}_{ds}}\right]\times \left[\frac{L}{W{C}_i{V}_{ds}}\right]$$

where µ is the mobility, W (2 µm) is the channel width, L is the channel length and C_i (1.418 × 10⁻⁴ F.m⁻¹) is the capacitance between the channel and back gate per unit area ($C_i=\frac{{\epsilon }_0{\epsilon }_r}{d}$; ε₀ = 8.85 × 10⁻¹² F.m⁻¹; ε_r = 3.9; d = 275 nm). The mobility value can be determined to be 0.1–0.5 cm²/Vs. The low mobility compared to several other devices in the literature could be due to the poor contact with metal electrode, as reported earlier in case of WSe₂ and MoS₂ [55]. It may also be possible to improve the performance by using high-K dielectric materials in top-gated devices [58,59,60].

Two-dimensional p-type semiconductors provide extensive advantages and benefits for designing CMOS inverters with unique and desirable properties. Such CMOS designs with p-type FET allow for a simple electrode structure (V_in) and circuit planning that result in an elevated gain and lower operating voltage (V_dd). P-type 2D semiconductors can also provide a solid support for developing new junction inverters with lower power consumption, which is vital for future consumer electronics. With the development of the p-type 2D semiconductor, CMOS logic circuits based on p-type FET present multifarious advantages in terms of computational densities and reconfigurable capabilities of single devices [56]. They also offer key advantages due to their ability to operate with positive voltage in regard to the source terminal, which simplifies the integration process of these components in circuits where a positive power supply is essential. This also makes p-type FET well suited in power electronics, like load switching and battery management. In addition, the negligible ON-state resistance of P-channel FETs suppresses power losses and contributes to enhanced energy efficiency, making them a favorable candidate for battery-operated devices and energy-conscious systems.

The development of digital electronics has manifested specific requirements in terms of channel length, switching characteristic and contact resistivity, eventually integrating them into modern Si-dominant electronics. Recent studies have also indicated that p-type 2D semiconductors exhibit great application potential in logic device design, although their performance is yet to reach the theoretical bottleneck. Fortunately, the p-type 2D semiconductors with atomic thickness and surfaces with dangling bonds are promising channel candidates, and CdPS₃ provides a unique leverage in the family of transition metal thiophosphates. Although mobility is not at the ideal mark, multiple device improvements, like reducing the Schottky barrier and improving the metal contacts, can assist in reaching the desired outcome. In addition, due to ease of fabrication of 2D CdPS₃, it can be used more swiftly in future electronics based on p-type 2D semiconductors to achieve complexity, integration and functional diversification.

3. Conclusions

In this work, we carried out an extensive structural analysis of vdW layered CdPS₃, using temperature-dependent Raman scattering. In addition to the high-frequency regime, Raman signatures at a low wavenumber provided convincing information on the structural features and thermally induced phase variation. The emergence of new modes and the large blue shift of the peak position suggested a structural phase transition below a certain temperature. This was highly influenced by the appearance of new peaks in low-frequency Raman signatures that was clearly induced by thermal variation. In addition, the vibrational intensity of these Raman modes also incremented below the phase transition temperature. The elemental composition was verified using EDS prior to the Raman analysis. Furthermore, we analyzed the CdPS₃ FET to show a p-type transport behavior that is crucial for the development of future power electronics and energy applications. This work adds CdPS₃ to potential candidates for multifarious applications in demanding sectors of energy and electronics.

4. Methods

4.1. Material Fabrication

Two-dimensional flakes of CdPS₃ were prepared by the well-established mechanical exfoliation method. Firstly, the SiO₂/Si wafers were cut into 10 × 10 mm² size chips; this stage was followed by thorough cleaning of the wafers in an ultrasonic bath of acetone, isopropyl alcohol and deionized water, in sequence. Bulk crystals of CdPS₃ (purchased commercially from 2D semiconductors) were exfoliated with Scotch tape from the same batch of crystals. Tapes containing CdPS₃ were directly brought into contact with the substrate. During the transfer process, the substrate was baked or mildly annealed on a hot plate at 60−65 °C for 2 min. This helps to remove any trapped air molecules between the sample and the substrate interface and also facilitates the obtainment of large area samples, as used by researchers previously. This method is also effective for obtaining intact and ultraclean surfaces, in addition to achieving contamination-free samples. The flakes were then identified by an ultrahigh-resolution optical microscope from Nikon. All results obtained are from the same batch of crystals.

For device fabrication, the 100 nm thickness gold electrodes were directly patterned in a plain SiO₂/Si substrate via conventional photolithography, metal deposition and lift-off process. Exfoliated 2D materials were then dry-transferred onto the electrodes, using the vdW transfer technique. A micro actuator was used to accurately position the samples.

4.2. Optical Characterization

Raman measurements were conducted using a Horiba LabRAM system equipped with a confocal microscope, a charge-coupled device Si detector and a 532 nm diode-pumped solid-state laser as the excitation source. An objective lens was used to focus the laser light on the surface of the sample. Estimated diameter of the illuminated spot on the samples is ∼2 μm. The spectral response of the entire system was determined with a calibrated halo-gen–tungsten light source. For temperature-dependent measurements, the sample was placed on a microscope compatible. Linkam chamber connected to a temperature con-troller; liquid nitrogen was used as the coolant.

4.3. Electron Microscopy Characterization

Scanning Electron Microscopy was carried out in a Thermo Fisher Scientific Verios 460 L and an Oxford Instruments X-MaxN 80, incorporating a field emission gun and a monochromator suitable for ultrahigh-resolution imaging and equipped with an Everhart-Thornley in-lens detector. Energy-dispersive spectrometry was carried out using the same instrument equipped with an Oxford electron-dispersive X-ray (EDX) spectrometer with an 80 mm² silicon drift detector and X-Max detector.