Effect of physical vapor deposition on contacts to 2D MoS₂

The two-dimensional (2D) transition metal dichalcogenide (TMD) molybdenum disulfide (MoS₂) is a desirable channel material for field-effect transistors (FETs) due to its structural and semiconducting properties for devices approaching the sub-nanometer scale.^1,2 Molybdenum disulfide has strong covalent bonds within its layers and weak van der Waals bonds between its layers, which can be cleanly broken to produce a single layer of MoS₂. Monolayer MoS₂ provides an atomically flat and potentially dangling-bond-free van der Waals surface and a bandgap of 1.8 eV for high-performance FETs.^3,4 However, high contact resistance (RC)at the metal/semiconductor interface impedes FET performance.^3,5 In addition, interfacial reactions between contact metals and 1L MoS₂ could lead to high contact resistance in MoS₂-based FETs.⁶ 

Preparing atomically sharp and clean metal/semiconductor interfaces can help overcome the Fermi-level pinning effect in 2D semiconductor devices,^7–9 allowing engineers to tailor the Schottky barrier height and reduce contact resistance. Sputtering allows for the deposition of alloys and compounds with precise control over composition, making it easier to achieve the desired stoichiometry from a single source.^10,11 The ability to deposit alloys and compounds is particularly valuable when forming contacts to 2D materials, making it possible to tailor the work function^12,13 or topological properties¹⁴ or superconductivity¹⁵ of a contact metallization. However, metallization processes can introduce crystal lattice disorder near the metal/2D material interface and result in Fermi-level pinning.^8,16,17 Despite the higher energy of sputtered atoms (typically 2–7 vs ~0.1 eV for evaporated atoms),¹⁸ the approach has been of interest in the literature. Radio frequency (RF) sputtering of Sb₂Te₃¹⁹ has been used to prepare van der Waals (vdW) contacts to MoS₂, offering high thermal stability and low contact resistance. The improved performance is attributed to Fermi-level unpinning, a small band offset between Sb₂Te₃ and MoS₂, and a low density of states in Sb₂Te₃, leading to better results in MoS₂ MOSFETs with Sb₂Te₃/W contacts compared to Sb/W contacts.

Thermal and electron-beam evaporation are often used for contact deposition on MoS₂, but they come with additional concerns such as the need for precise control over the deposition rate and chamber pressure. These factors can impact the performance of contacts, potentially leading to a variation in device characteristics and reliability. For example, Wang et al. showed that when the deposition rate of Au by electron-beam evaporation is high, the performance of FETs was degraded, possibly due to radiative heating.²⁰ Other recent research on 2D memristors reveals that high Au electron-beam evaporation rates can cause local damage to MoS₂, with Au atoms penetrating the layers, impacting memristor performance. Their study found reliability improvements with low Au deposition rates and thicker MoS₂, as higher rates negatively affected memristor performance by inducing localized disorder at the contact inferface.²¹ To minimize radiation damage from x-ray exposure during e-beam evaporation, Wang et al.²² recommend intermittent deposition at a relatively high rate (as reported earlier by Wang et al.²⁰) to reduce x-ray exposure, with pauses to stabilize pressure between steps. Another study reports that Au electron-beam evaporated in ultrahigh vacuum (∼10^–9 Torr) offers three times lower contact resistance compared to ordinary conditions (∼10⁻⁶ Torr),²³ while others claim that a moderate vacuum of 5 × 10⁻⁶ Torr still permits the formation of a high-quality²⁴ quasi-van der Waals interface. Yet other studies have examined the effect of deposition pressure, revealing the potential for oxide formation at the contact interface (for Ni²⁵ or Ti²⁶) if the base pressure is not sufficiently low, although Bi contacts²⁵ showed no significant difference in contact resistance or interfacial chemistry between ultrahigh vacuum and high vacuum conditions. High reactivity of metal atoms involved in the deposition process can also greatly alter or even destroy a monolayer MoS₂ film.^27,28 It is, therefore, important to investigate metal deposition to understand better and control electronic transport in nanoscale MoS₂ devices.

In this work, we investigate the deposition of different metals on MoS₂ using two physical vapor deposition techniques for each and study the impact of metallization on the vibrational and electronic properties of MoS₂. We find that sputter deposition causes more damage to monolayer MoS₂ than electron-beam evaporation, as evidenced by greater suppression of first-order Raman modes, and we correlate this damage with the cohesive energies of the metals. Furthermore, we compare the performance of field-effect transistors fabricated with e-beam evaporated and sputtered Sn/Au contacts. We selected Sn/Au contacts because low-melting-point metals such as Bi,²⁹ Sn,³⁰ and In³⁰ have been reported to offer low contact resistance on 1L MoS₂ in recent studies without damage to 1L MoS₂.^29,30 Among these metals, Sn was chosen to explore differences in contact resistance, with Au added as a capping layer for ease of probing.

Monolayer MoS₂, synthesized by metalorganic chemical vapor deposition (MOCVD) on 2 in. sapphire substrates as previously described,^31,32 were obtained from the 2D Crystal Consortium (2DCC) at Pennsylvania State University and cleaved into smaller pieces for sample fabrication. Metal thin films were deposited directly onto the 1L MoS₂ by electron-beam evaporation or DC magnetron sputtering to determine the effects of each physical vapor deposition technique on MoS₂. We compared electron-beam evaporated Bi, Cu, In, Pd, Pt, and Sn films to their sputtered counterparts. These metals have been used previously in reported contacts to MoS₂.^4,29,33–35 All electron-beam evaporation occurred in a chamber with base pressure <5 × 10⁻⁷ Torr, and the chamber was gettered prior to electron beam evaporating the metals of interest by depositing 30 nm of Ti with the sample shutter closed to remove background reactive species (such as H₂O vapor) in the deposition system. More details on the conditions for the electron-beam evaporation of each metal are found in Table S1.1 in the supplementary material 1). Details about the conditions for the sputter deposition of each metal are also provided in Table S1.2 in supplementary material 1. The metal films were nominally 100 nm thick except for electron beam evaporated Pt (70 nm), which was intentionally made thinner to mitigate the heat generated during its evaporation. Samples remained in the vacuum chamber or were vacuum sealed or stored in a nitrogen dry box until it was time to collect Raman spectra to reduce possible effects from the ambient environment.

Raman spectroscopy is a fast and non-destructive technique previously used to study doping, strain, and damage in 1L MoS₂³⁶ and 1L MoS₂ top-gated FETs.³⁷ Studies that used Raman spectroscopy to observe the effects of metal contacts on TMDs were limited to studying ultra-thin (<2 nm) metal films due to the restriction of maintaining metal transparency.⁶ We need an alternative approach to understand the buried metal/MoS₂ interface in a FET structure, since ultra-thin contact metals could result in discontinuous films.²⁷ We have already established a backside-illumination Raman spectroscopy technique and demonstrated its use in probing various metal/1LWS₂ interfaces to observe interfacial reactions.²⁸ In this study, we applied the backside-illumination method to subsequent experiments in this paper to study the effects of electron beam evaporated and sputtered metal thin films on 1L MoS₂.

We collected numerous Raman spectra from various areas of the metal/1LMoS₂/sapphire samples using a LabRAM HR Evolution (Horiba Scientific Co.) spectrometer with ultra-low-frequency (ULF) 532 nm laser excitation at a power of 1.7 mW (5% of 34 mW @ 100%). We collected each scan twice using a grating with 1800 lines/mm and a near infrared 50× objective lens over a spectrum range from −150 to 800 cm⁻¹ for 30 s. The 532 nm laser was focused through the transparent sapphire substrate (backside illumination) using a backscattering geometry (Fig. 1). The backscattering geometry coupled with the backside illumination allows us to probe a buried metal/TMD interface and non-destructively detect an underlying TMD monolayer, if still present, after metallization.²⁸ We did not observe any differences in the Raman spectra for 1L MoS₂ between the data collected with the topside illumination backscattering geometry and data collected via the backside-illumination backscattering geometry, as described in more detail in Fig. S2.1 in supplementary material 2. The Raman spectra were plotted using the LabSpec 6 Imaging and Spectroscopy Software (Horiba Scientific Co.) and normalized to the sapphire peak. Spectra were fitted with the Gaussian–Lorentzian fit to determine the peak positions and full width at half maxima (FWHM). Further information is provided in supplementary material 3, which provides more information about the process and shows the fit for each metal. However, we report FWHM only for spectra with low signal-to-noise ratios. The peak positions and corresponding FWHM were averaged over all acquired spectra for each sample with a standard deviation of ≤0.7 cm⁻¹ for the peak position and ≤1.1 cm⁻¹ for the FWHM.

To fabricate the devices, the MoS₂ monolayer was transferred onto Al₂O₃/Pt/TiN/p + Si via poly (methyl methacrylate) (PMMA)-assisted wet-transfer.³⁸ To define the channel regions of the MoS₂ FETs in this study, substrates with transferred MoS₂ were spin-coated with PMMA A6 at 4000 RPM for 45 s and then baked at 180 °C for 180 s. The resist was exposed using electron beam (e-beam) lithography and developed with a 1:1 mixture of 4-methyl-2-pentanone (MIBK) and isopropyl alcohol (IPA) for 80 and 40 s, respectively. The exposed monolayer MoS₂ was etched using an SF₆ and O₂ plasma at 5 °C for 25 s. Following this, the samples were rinsed in acetone and IPA to remove the e-beam resist. For defining the source and drain contacts, a bilayer resists of A3 PMMA and A6 PMMA was spin-coated onto the samples. E-beam lithography was used again to define these contacts, with development performed using the same MIBK and IPA mixture. Metal contacts were deposited using e-beam evaporation or sputter deposition (30/20 nm Sn/Au). Finally, the samples underwent a lift-off process to remove excess resist and metal by immersing in acetone for 1 h, followed by IPA for an additional 30 min. Measurements were performed using a semi-automated Formfactor 12 000 probe station under atmospheric conditions with a Keysight B1500A parameter analyzer.

A reference Raman spectrum was collected for bare 1L MoS₂ using the backside-illumination method as the baseline for analysis, and it provides the same peak positions as for topside illumination (confirmation is provided in supplementary material 2 where deconvoluted spectra are compared). Peak positions for bare and metallized 1L MoS₂ are listed in Table I. In Fig. 2, this same spectrum for bare MoS₂ is shown as the top figure in each part along with spectra collected after electron-beam evaporation and sputter deposition of each metal, with the metals Cu, Pd, Pt, Bi, Sn, and In shown in Figs. 2(a)–2(f), respectively. Peak deconvolution is shown for each metallized sample in the figures in supplementary material 3. We detected the first-order in-plane and out-of-plane Raman modes for 1L MoS₂, $E ′ ( Γ ) α$ at $386.1 c m − 1$ and $A ′ 1 ( Γ ) α$ at 405.3 cm⁻¹, respectively.^36,39 The broad shoulder on the low-frequency side of the $E ′ ( Γ ) α$ peak is a convolution of the $E ′ ( M ) L O 2$ mode referenced at 377.4 cm⁻¹ for 1L MoS₂³⁶ and a sapphire peak at 379 cm⁻¹. We assigned the $2 LA( M)$ acoustic phonon and the $E ′ ′ ( M ) T O 1− ZA( M)$ multiphonon modes to the broad peak from ~455to 467 cm⁻¹.⁴⁰ Peaks originating from the sapphire substrate are located at 379, 417, 430, and 448 cm⁻¹.⁴¹ The metalized 1L MoS₂ Raman spectra are presented in Fig. 2 for Cu, Pd, Pt, Bi, Sn, and In, respectively.

The first-order $E ′ ( Γ ) α$ and $A ′ 1 ( Γ ) α$ 1L MoS₂ Raman modes remained for all electron-beam evaporated metals [Figs. 2(a)–2(f)], although they are weak after the deposition of Pd and very weak after the deposition of Pt. The $E ′ ( Γ ) α$ mode for MoS₂ beneath Cu, Pd, Pt, Bi, Sn, and In is located at 384.1, 384.0, 384.6, 383.8, 384.4, and 384.6 cm⁻¹, respectively. All evaporated metals exhibited a red shift of approximately 2 cm⁻¹ in the $E ′ ( Γ ) α$ mode. The $A ′ 1 ( Γ ) α$ modes for evaporated Cu, Pd, Pt, Bi, Sn, and In samples are located at 404.5, 405.6, 405.6, 405.1, 406.0, and 405.5 cm⁻¹, respectively [Figs. 2(a)–2(f)]. Evaporated Pd, Pt, Sn, and In samples exhibited a blue shift of <1 cm⁻¹ [Figs. 2(b), 2(c), 2(e) and 2(f)], but Cu and Bi samples exhibited a red shift of <1 cm⁻¹ in the $A ′ 1 ( Γ ) α$ mode [Figs. 2(a) and 2(d)]. The shifts of the $A ′ 1 ( Γ ) α$ modes are within the spectral resolution of the Raman data.

An additional peak, which we identified as the $E ′ ( M ) L O 2$ mode, remains after depositing Cu with a peak position of 373.6 cm⁻¹ after evaporation and 375.1 cm⁻¹ after sputtering, although much broader, resulting in red shift of ~4 and ~2 cm⁻¹ compared with bare 1L MoS₂ [Fig. 2(a)]. This peak does not show up for the other metals deposited [Figs. 2(b)–2(f)]. We do not know why we observe the $E ′ ( M ) L O 2$ mode in some cases but not others. Although Cu₂S, Cu₉S₅, Cu₉S₈, or CuS could contribute peaks near 470 +/− or 261 +/−5 cm⁻¹, there are no peaks expected from copper sulfides from 373 to 375 cm⁻¹.⁴² Using polarized light could have helped us further confirm the identification, but we did not take this step.

The first-order $E ′ ( Γ ) α$ and $A ′ 1 ( Γ ) α$ 1L MoS₂ Raman modes remained after sputtering Cu, Pd, Bi, and Sn [Figs. 2(a), 2(b), 2(d), and 2(e)], but disappeared after sputtering Pt and In [Figs. 2(c) and 2(f)]. The $E ′ ( Γ ) α$ mode for Cu, Pd, Bi, and Sn is located at 384.3, 384.5, 384.5, and 384.0 cm⁻¹, respectively, and exhibited a red shift of approximately 2 cm⁻¹. The $A ′ 1 ( Γ ) α$ mode for Cu, Pd, Bi, and Sn is located at 404.6, 404.7, 406.2, and 404.7, respectively [Figs. 2(a), 2(b), 2(d) and 2(e)]. These shifts in the $A ′ 1 ( Γ ) α$ mode for the sputtered metals were <1 cm⁻¹ and within the spectral resolution of the Raman data.

According to the literature, red shifts observed in the $E ′ ( Γ ) α$ modes could be due to strain or structural effects,⁴³ and shifts observed in the $A ′ 1 ( Γ ) α$ modes can be due to doping or electronic effects.³⁶ The $A ′ 1 ( Γ ) α$ mode has been shown to be sensitive to doping,⁴⁴ shifting to lower frequencies (red shift) when the MoS₂ undergoes n-type doping⁴⁵ and higher frequencies (blue shift) when it undergoes p-type doping.⁴⁶ Shifts in the $A ′ 1 ( Γ ) α$ mode can be associated with electron doping due to the strong electron–phonon coupling observed for the out-of-plane vibrational mode.³⁶ Structural disorder could lead to the appearance of new defect-induced peaks such as the LA(M) mode referenced to be around 226 cm^−1,47 but we did not find this peak.

The disappearance of first-order Raman modes of TMDs has been observed in cases where the monolayer TMD was consumed in a reaction.²⁸  Table I shows that the first-order 1L MoS₂ Raman mode after sputtering is missing for some metals, as discussed in more detail below, leading us to conclude that sputtering may induce significant damage to the 1L MoS₂.

We observed a decrease in the intensity of the first-order 1L MoS₂ Raman modes after metal deposition. The decrease in the intensity for the $E ′ ( Γ ) α$ and $A ′ 1 ( Γ ) α$ modes was severe for sputtered Pd, Pt, Sn, and In films [Figs. 2(b), 2(c), 2(e) and 2(f)], resulting in our inability to detect the first-order MoS₂ Raman modes for sputtered In [Fig. 2(f)], nor were we able to detect the $E ′ ( Γ ) α$ mode for sputtered Pd, Pt, and Sn films [Figs. 2(b), 2(c) and 2(e)]. We detected only a weak remnant of the $A ′ 1 ( Γ ) α$ mode in the spectra from the sputtered Pd, Pt, and Sn samples [Figs. 2(b), 2(c), and 2(e)], which shows up as a very weak bump around the location of the referenced $A ′ 1 ( Γ ) α$ mode. Although the peak intensity decreased for sputtered Cu and Bi, we were able to detect the first-order MoS₂ Raman modes [Figs. 2(a) and 2(d)].

The peak intensities of first-order Raman modes for various electron-beam evaporated metals are summarized in Table II as high intensity for Sn and In, medium intensity for Cu and Bi, low intensity for Pd, and very low intensity for Pt. We were still able to detect the first-order 1L MoS₂ Raman modes for all evaporated films, suggesting that evaporation is less destructive to the 1L MoS₂ than sputtering. Decreasing intensity has been observed in first-order Raman modes after evaporating very thin layers of Ag, Al, Au, and Ni onto monolayer MoS₂⁴⁸ and Au on multilayer MoS₂.⁴⁹ 

Peak broadening has been associated with a decrease in the MoS₂ grain size due to phonon confinement at grain boundaries. Peak broadening can be determined by measuring the FWHM of the first-order 1L MoS₂ Raman modes and comparing it to metallized 1L MoS₂. The FWHM of the $E ′ ( Γ ) α$ and $A ′ 1 ( Γ ) α$ modes for the bare 1L MoS₂ was measured to be 4.4 and 5.6 cm⁻¹, respectively. We measured the FWHM of the $E ′ ( Γ ) α$ and $A ′ 1 ( Γ ) α$ modes for the evaporated samples (Table II). (We do not provide FWHM from the sputtered samples because of lower intensity or disappearance of first-order Raman modes.) The $E ′ ( Γ ) α$ peak narrowed by approximately 0.4 cm⁻¹ after evaporating some metals (Cu, Pd, and In) remained unchanged for Pt, and increased by 1.1 and 0.3 cm⁻¹ after depositing Bi and Sn, respectively. The $A ′ 1 ( Γ ) α$ peak narrowed by 0.6–0.9 cm⁻¹ for Cu, Pd, Pt, and Bi, but it increased by ~0.4 cm⁻¹ for Sn and In. These changes to the FWHM are too small to conclude that the deposition of the evaporated metal films influenced the peak broadness in the underlying MoS₂ monolayer. It should be noted that the samples with evaporated metals, except for Sn and In, exhibited very low signal-to-noise ratios, making it hard to determine the FWHM with confidence. Although broadening of first-order Raman modes for 1L MoS₂ has been reported to be associated with an increase in disorder through theoretical⁵⁰ and experimental⁶ studies, we do not observe significant peak broadening that could be a signature of structural damage after evaporating the metal films. Peak broadening in the first-order Raman modes of MoS₂ induced by structural damage in the TMD reported in the literature is larger (change in FWHM > 4 cm⁻¹)^44,47 than what is measured for our samples.

We expect atoms arriving at the 1L MoS₂ surface with high energy to cause damage, but we are unable to directly measure the energies of the sputtered atoms. However, Spethmann et al. determined that for a given angle of incidence of an Ar ion beam, the maximum force a sputtered atom exerts on a surface upon impact increases with decreasing sputter yield.⁵¹ In other words, metals that are easily sputtered require less energy to be released from the target surface and also arrive on the substrate surface with less energy, assuming limited collisions during transport. The sputter yield also tends to increase with the cohesive energy of the target material.⁵² We, therefore, compare our Raman spectra of 1L MoS₂ beneath sputtered metals to the cohesive energy of the metals (Fig. 3). Indeed, we observe a correlation between the expected force of sputtered metals of interest (Bi, In, Sn, Cu, Pd, and Pt) and their cohesive energies (taken from).⁵³ The metals Bi, Sn, Cu with low cohesive energies (2.18, 3.14, and 2.49 eV, respectively) are the same ones with the first-order 1L MoS₂ Raman modes still present after sputtering, with the exception of In (2.52 eV). Interestingly, this correlation also applies to the adsorption energies of the metals on MoS₂, where metals with lower adsorption energies (Bi, Sn, and Cu with adsorption energies of 0.09, 0.63, and =0.77 eV) retained, respectively, the first-order 1L MoS₂ Raman modes, with the exception of In, which has a slightly higher adsorption energy. Hence, it is possible that the disappearance of the 1L MoS₂ peaks beneath In might also be influenced by a slightly higher adsorption energy.

To understand how the electrical properties of the MOCVD-grown MoS₂ films change with PVD technique, a study of Sn/Au contacts for 1L MoS₂ FETs has been conducted. Figure 4(a) shows the FET structure with the MoS₂ channel contacted by 20/30 nm Sn/Au. To create transfer length method (TLM) test structures, we fabricated FETs with 100, 300, 500, and 800 nm channels; a channel width of 4 μm; and a contact length of 1.5 μm on each mesa. Atomic layer deposition (ALD) grown 50 nm Al₂O₃ (ɛ_ox ≈ 9) serves as the back-gate dielectric on Pt/TiN/p⁺⁺-Si substrates. Figures 4(b) and 4(c) show the transfer characteristics, i.e., source-to-drain current $( I D)$ vs back-gate bias $( V B G)$ for different drain voltages $( V D)$ for a channel length $( L C H)$ of 100 nm for electron beam evaporated and sputtered contacts, respectively. All the devices exhibited an n-type channel, as expected.⁵⁴  Figures 4(b) and 4(c) show the ratio of the maximum to minimum current (I_max/I_min) for the best set of devices for different drain voltages $( V D)$⁠. Here, I_max is the maximum current obtained from the transfer characteristics for V_D = 1 V, and I_min is the average noise floor. I_max/I_min was found to be 3 × 10⁷ with evaporated Sn/Au contacts and 4 × 10⁶ with sputtered Sn/Au contact, respectively. The best device had driven currents of $I ON=58.5μ A /μ m$ and $I ON=9.3μ A /μ m$ with electron-beam evaporated and sputtered Sn/Au contacts, respectively, as shown in Figs. 4(d) and 4(e), while the mean values were 52.25 and 7.76 μA/μm, respectively.

Here, $C O X$ is the capacitance of the back-gate oxide (~1.6 × 10⁻³ F m⁻²). The threshold voltage (V_TH) was extracted at 10 nA/μm iso-current and showed a slight variation between sputtered and evaporated contacts. Specifically, the mean V_TH for sputtered contacts was −4.7 ± 0.8 V, while for electron-beam evaporated contacts, it was −5.3 ± 0.7 V.

Figure 5(a) depicts the distribution of contact resistances of 25 TLM structures with channel lengths of 100, 300, 500, and 800 nm. The best TLM set gave contact resistances of 8.75 and 72 kΩ μm, while the mean Rc values were 11.25 and 135 kΩ μm for evaporated and sputtered Sn/Au contacts for n_S = 9 × 10¹² cm⁻², respectively. Hence, sputtered contacts consistently offer higher resistance than evaporated contacts. Interestingly, the slopes of the plots for the total resistance vs channel width were slightly higher for the sputtered contacts than for the evaporated contacts [Fig. 4(f)], which could reflect an effect on the MoS₂ adjacent to the contacts due to sputtering. For the three-dimensional semiconductor GaN, Molina and Mohney showed that sputtering led to less ideal as-deposited Schottky barrier diodes when rhenium was sputtered compared to when it was electron-beam evaporated.⁵⁶ Another study found that process-induced defects in GaN can be introduced well below the metal/semiconductor interface, as detected by deep level transient spectroscopy. Their concentrations depend strongly on metal deposition conditions.⁵⁷ We, therefore, suspect that sputter deposition could also have introduced defects in the 2D semiconductor in the adjacent channel of the FETs.

Here, g_m is the transconductance, W is the channel width, I_D is the drain current, and V_D is the applied drain bias. The subthreshold slope (SS) was calculated as SS = [dlog(I_D)/dV_(BG)]⁻¹.

The mobility of 2D MoS₂ devices with electron-beam evaporated Sn/Au contact demonstrated superior performance, with a maximum mobility of 2.14 cm²/V s, a minimum mobility of 1.68 cm²/V s, and a mean mobility of 2.1 cm²/V s. In contrast, devices with sputtered contacts exhibited significantly lower mobility values, with a maximum of 0.55 cm²/V s, a minimum of 0.24 cm²/V s, and an average mobility of 0.5 cm²/V s [Fig. 5(b)]. The subthreshold slope (SS) was extracted over two orders of magnitude change in I_D for all MoS₂ FETs with mean values of 880 and 1500 mV/dec for evaporated and sputtered Sn/Au contacts, respectively [Fig. 5(c)].

This study analyzed the effects of electron-beam evaporated and DC magnetron sputtered metal contacts (Cu, Pd, Bi, Sn, Pt, and In) on 1L MoS₂ using non-destructive backside-illumination Raman spectroscopy. Our results indicated that of the two PVD techniques, the 1L MoS₂ retained its first-order Raman modes after evaporating metal films, although the intensities for Pt were quite low. We were able to observe damage induced in 1L MoS₂ by sputtered metals by the disappearance of the $E ′ ( Γ ) α$ and $A ′ 1 ( Γ ) α$ modes or the significant decrease in their peak intensity, and we found that higher cohesive energies generally correlate with more damage to the semiconductor. Field-effect transistors prepared with evaporated Sn/Au contacts exhibit superior performance metrics compared to FETs with sputtered Sn/Au contacts, including lower contact resistance, enhanced mobility, and improved subthreshold slope compared to their sputtered counterparts. The higher energy of arriving species makes sputter deposition more damaging to monolayer MoS₂ than evaporation of metals, leading to FETs with poorer performance. This work demonstrates the utility of backside Raman spectra, which correlate in an obvious way with device performance, and the approach is recommended for the study of other transition metal dichalcogenide devices.

The supplementary material provides supporting information on (1) evaporation and sputtering conditions, (2) comparison of backside and topside illumination, and (3) the Gaussian–Lorentzian fit to determine the peak positions and full width at half maxima (FWHMs), including the fit for each metal.

The authors acknowledge the primary financial support of the National Science Foundation (NSF) through ECCS 2227346. Epilayers were provided by the Penn State 2D Crystal Consortium—Materials Innovation Platform (2DCC-MIP) under NSF DMR 2039351.

The authors have no conflicts to disclose.

M. Saifur Rahman: Data curation (equal); Formal analysis (equal); Investigation (equal); Methodology (equal); Writing – original draft (equal). Ama D. Agyapong: Data curation (equal); Formal analysis (equal); Investigation (equal); Methodology (equal); Writing – original draft (equal). Suzanne E. Mohney: Conceptualization (equal); Investigation (equal); Writing – review & editing (equal).

The data that support the findings of this study are available from the corresponding author upon reasonable request.