Effects of Annealing Temperature and Ambient on Metal/PtSe2 Contact Alloy Formation.

Forming gas annealing is a common process step used to improve the performance of devices based on transition-metal dichalcogenides (TMDs). Here, the impact of forming gas anneal is investigated for PtSe2-based devices. A range of annealing temperatures (150, 250, and 350 °C) were used both in inert (0/100% H2/N2) and forming gas (5/95% H2/N2) environments to separate the contribution of temperature and ambient. The samples are electrically characterized by circular transfer length method structures, from which contact resistance and sheet resistance are analyzed. Ti and Ni are used as metal contacts. Ti does not react with PtSe2 at any given annealing step. In contrast to this, Ni reacts with PtSe2, resulting in a contact alloy formation. The results are supported by a combination of X-ray photoelectron spectroscopy, Raman spectroscopy, energy-dispersive X-ray spectroscopy, and cross-sectional transmission electron microscopy. The work sheds light on the impact of forming gas annealing on TMD-metal interfaces, and on the TMD film itself, which could be of great interest to improve the contact resistance of TMD-based devices.

Introduction

Transition-metal dichalcogenides (TMDs) have been one of the most studied class of two-dimensional (2D) materials to date. Although promising, their integration with modern electronics is still limited due to the rather large Schottky barrier that usually characterizes the metal–TMD interface, which limits the device behavior, hiding their potential. In addition, most of the results are based on exfoliated flakes,[1,2] which causes a high variation from experiment to experiment, especially at low thickness. The lowest values of contact resistance are usually related to high-k induced doping[3,4] or interface engineering.[5−8] Although promising, another possible solution still mostly unexplored is the use of an opportune annealing step to modify the metal–TMD interface. Annealing is a common process step in the modern semiconductor industry. Laser annealing is used for Si and Ge to achieve highly doped and abrupt junctions.[9] Additionally, annealing in forming gas (FG) has been used to reduce fixed oxide charges and interface states at the Al2O3/InGaAs interface.[10] Similarly, for 2D-semiconductors, several publications have shown how different annealing steps can improve the performances of TMD-based devices. Vacuum annealing has been shown to reduce the concentration of surface contaminants.[11,12] Furthermore, annealing in an inert environment or FG can be beneficial in terms of reducing the density of interface states at the MoS2/oxide interface.[13] Annealing steps have also been shown to be valuable for the TMD–metal interface. FG annealing (FGA) on graphene/MoS2 contacts was shown to improve the contact resistance of the material.[14] Other annealing techniques, such as vacuum annealing,[15] microwave annealing,[16] or short-term pulsed annealing,[17] were also reported to be advantageous for the metal/MoS2 interface. Nonetheless, the values are still far from the requirement of the International Roadmap for Semiconductors and Systems,[18] and the reaction metal–TMD upon annealing as well as the optimal annealing temperature range and ambient is still unclear.

In this work, we systematically characterize large-area grown thin films of PtSe2 considering different postgrowth thermal treatments and their effects on the metal–TMD interface. PtSe2 has attracted increasing interest over the last few years. As other TMDs, a monolayer of PtSe2 is formed by a three-atom thick atomic cell, with the transition metal sandwiched between two chalcogen layers. Most TMDs show semiconductive properties, but interestingly, PtSe2 shows semimetallic characteristics in its bulk form, whereas experiments have shown the opening of a band-gap for thicknesses lower than 3 nm, with ∼1.2 eV for a monolayer.[19,20] Therefore, PtSe2 has gained attention due to its potential application in electronics,[21] optoelectronics,[22] and sensors.[23] PtSe2 can be uniformly grown over a large area by thermally assisted conversion (TAC), which does not require a high thermal budget (max temperature of 400 °C). This makes the growth compatible with back-end-of-line processing and integration in modern technologies, as it respects the typical thermal budgets of the CMOS industry (∼450 °C).[24−26] This is in sharp contrast with TMD films grown by chemical vapor deposition, chemical vapor transport, or molecular beam epitaxy.[27−30] The growth is not homogenous, and it is characterized mainly by triangular crystals that merge together after a sufficient growth time.[31−33] Above this, the thermal budget of these processes is usually high to obtain a reaction of the predeposited film with the chalcogen, limiting their integration in modern CMOS technology.

The PtSe2 was grown at 400 °C over a large area. Ti/Au and Ni/Au were used as metal contacts and circular transfer length method (c-TLM) structures were used to electrically characterize the metal/PtSe2 interface and the PtSe2 material itself. The samples were then annealed for 1 h in a FG ambient (5/95% H2/N2) at different annealing temperatures: 150, 250, and 350 °C. The same experiment was repeated in an inert environment (0/100% H2/N2). The electrical properties of the material were studied after each annealing step. The TMD–metal interface and alloy formation were studied and characterized by a combination of X-ray photoelectron spectroscopy (XPS), Raman spectroscopy, energy-dispersive X-ray (EDX) spectroscopy, and cross-sectional transmission electron microscopy (XTEM).

Results

Figure shows the process flow adopted to study the reaction of PtSe2 after each annealing step. First, PtSe2 (6–10 nm, with an initial Pt thickness of 1 nm) was grown on four different samples in a single growth run for high consistency using a TAC process.[34] Because of this, applying a similar FGA step allows the effect of the process on the material itself and the metal–TMD interface to be decoupled. c-TLM structures were patterned by standard photolithography, and Ti/Au or Ni/Au were deposited by metal evaporation. Then, samples were measured electrically. The presence of PtSe2 was confirmed by Raman analysis.

Figure 1

Experimental steps taken during this work.

Figure a,d shows the optical microscope images of the contacted PtSe2 with Ti and Ni contacts, respectively, before any annealing and no visible differences are present between the two samples. The samples then underwent different annealing steps. No obvious differences were found for the samples annealed in FG at 150 °C, as the temperature might have been too low to start any reactions. After annealing at 250 °C in forming or inert environment, the Ti/Au sample did not show any appreciable differences from optical inspection (Figure b,c). In contrast, the Ni/Au sample annealed at 250 °C in FG displayed what looks like an “alloy region” around the metal contacts (Figure e). This alloy region is less pronounced for the annealing at the same temperature in the inert environment (Figure f). It is noted that the reaction taking place between Ni and PtSe2 is accelerated by the presence of hydrogen during the annealing. The 350 °C anneal in FG (pictures not shown), as later confirmed by electrical and Raman characterizations, resulted in a degradation of the PtSe2. Considering that PtSe2 is grown at a temperature of 400 °C, a FG anneal at 350 °C may be too harsh an environment for this material, which highlights the importance of understanding the process window for optimal annealing temperature and its effects on the material properties.

Figure 2

Optical microscope pictures of the larger c-TLM of PtSe2 before annealing contacted with (a) Ti/Au and (d) Ni/Au. Same optical pictures repeated after annealing in FG and inert ambient at 250 °C for (b,c) Ti/Au and (e,f) Ni/Au, respectively. The darker gold ring in (e) clearly shows the alloy region.

Figure shows the electrical characterization carried out on each sample. Figure a shows representative current–voltage measurements on the Ni/Au samples considering a spacing of 43.5 μm. Although no appreciable differences are visible from optical inspection after a FG anneal at 150 °C, the current decreases with respect to the pre-annealed case. In contrast, considering the same FG environment but a temperature of 250 °C, the current increases, meaning that the reaction that took place positively changed the metal/PtSe2 interface. The difference is not related to a change in the properties of the channel as the annealing steps for the Ti/Au samples did not result in a change of the metal/PtSe2 or the channel properties (Figure c). Consider that the results at 250 °C in FG were corrected because of the alloy region because it changed the spacing of each structure (the actual channel length). The anneal at 250 °C in an inert environment, similarly to the 150 °C in FG, resulted in a lower current. These two anneal conditions seem to provide similar results. Although, one was done at a lower temperature, the presence of FG seems to have accelerated the reactions that took place. It is also important to consider that the 250 °C in the inert environment also resulted in an alloy region, although not as pronounced as the case in FG. In contrast, no alloy regions were found for the sample annealed at 150 °C in FG.

Figure 3

(a) Current–voltage measurements on the Ni/Au samples after each annealing step. (b) Total resistance vs spacing for the Ni/Au-contacted PtSe2. Inset: total resistance vs spacing for the Ti/Au-contacted PtSe2. (c) Extracted sheet resistance after each annealing step.

Figure b shows the total resistance versus spacing for the Ni/Au contacted PtSe2 at each annealing step. It is clear that each annealing is affecting the electrical behavior of the device. Nonetheless, this analysis is problematic for the Ni/Au samples. Because of the presence of the alloy region, the fitting process is less reliable because the spacing of the structures change with the annealing condition. Future work may include linear TLMs with smaller dimensions for a more accurate parameter extraction. The inset of Figure b shows the same analysis for the Ti/Au sample for each annealing condition. No appreciable differences are present, and the trend remains quite similar, meaning that the film itself, as well as the PtSe2/Ti interface, are not affected by any annealing steps.

Figure c shows the sheet resistance extracted from each measurement. The sheet resistance can be extracted from the Ti/Au measurements, considering the c-TLM theory.[35] However, for the Ni/Au samples, the total resistance was considered as the resistance of the PtSe2 owing to the low contact resistance. The results for the Ti/Au contacts confirm what is seen in Figure a–c and from XTEM images (Figure S1): the annealing temperature and conditions are not changing the metal/PtSe2 interface or the PtSe2 material itself. In contrast, a much higher variation is seen for the Ni/Au samples, which can be related to the reaction between Ni and PtSe2. Consider that, although obvious “alloy regions” are present from optical inspection it is possible that the Ni diffuses much further into the PtSe2, than what is possible to see with an optical microscope.

Raman spectroscopy was then performed to confirm the integrity of the PtSe2 and to probe the “alloy region” seen in the Ni/Au samples. The results are summarized in Table (all Raman results are reported in Figure S2). In particular, the Raman signal was collected on the bare PtSe2 and near the metal contact. For the Ti/Au sample, the signal of PtSe2 is always present, except for the 350 °C annealing in FG, which resulted in degradation of the PtSe2 film. Similarly, for the Ni/Au sample a clear PtSe2 signal is present up to 250 °C, in both forming and inert gas environment in the bare PtSe2 regions (far from the metal contact). For the samples that showed an “alloy region”, no PtSe2 signal is present close to the metal contact, suggesting metal interdiffusion and alloying.

Table 1

Summary of the Raman Analysis on the PtSe2 Samples after Each Annealing Condition

	Raman peaks of PtSe2
annealing condition	Ti/Au	Ni/Au
150 °C 5/95% H2/N2	present	present
250 °C 5/95% H2/N2	present	present, except in diffusion ring
350 °C 5/95% H2/N2	absent	absent
250 °C 100% N2	present	present
350 °C 100% N2	present	absent

Figure shows representative TEM images of the Ni/Au-contacted PtSe2 samples annealed at 250 °C in FG and in an inert environment. Figure a,b compare the PtSe2 structure inside and outside of the alloy region, respectively. It is noted that the layered structure is present outside the alloy region, whereas it is not inside it. Similarly, Figure c,d show the PtSe2 below the metal contact and outside of it, respectively. Again, the layered structure is not present below the metal contact but is preserved outside of it. These representative images confirm the hypothesis. The diffusion ring seen from optical images is a result of Ni interdiffusing and alloying with the PtSe2. This compound is not characterized by the same layered structure of PtSe2, but it has the form of a 3D alloy. Also, although the alloy region around the contact is not present in the sample annealed at 250 °C in an inert environment, a similar reaction has taken place below the metal contact. Therefore, it can be concluded that the presence of hydrogen accelerates the reaction, and the Ni can react more readily with PtSe2 away from the metal itself. It is also important to consider that defects as grain boundaries, steps or vacancies, are likely to lead to an increased activity compared to a defect-free film. These likely played a role in the process, considering both the polycrystalline nature of this film (Figure b) and the nanometer sized grain previously reported.[23]

Figure 4

Representative XTEM images of the PtSe2 contacted with Ni/Au. (a,b) show PtSe2 after annealing in FG at 250 °C inside and outside the diffusion ring, respectively. (c,d) show the PtSe2 after annealing in inert environment at 250 °C below the metal contact and outside the metal contact, respectively.

In order to explore the nature of the interdiffusional alloying observed, XPS was performed using near-identical samples. Figure shows XPS spectra of the Pt 4f and Se 3d core levels of PtSe2 before and after metal deposition, with no further annealing. No Au was used here as Au does not interface with the PtSe2. The Pt 4f spectra for bare PtSe2 (Figure a) shows the presence of two chemical states, both represented by a doublet, one related to PtSe2 (72.15 eV) and another due to surface oxidation (PtO at 71.1 eV). The Se 3d spectra of the same sample (Figure c) show PtSe2 and SeO features (53.55 and 54.25 eV, respectively), with the additional contribution of the Pt 5p3/2 (52.6 eV), which overlaps with the Se 3d.

Figure 5

XPS spectra of (a,b) Pt 4f and (c,d) Se 3d core levels for PtSe2 before and after Ni or Ti deposition.

Upon Ni deposition, a significant change in the peak shape of the Pt 4f is observed, with the emergence of a feature at low binding energy (BE). This peak is at lower BE than PtO, consistent with metallic Pt (Pt0). The Se 3d spectrum shows a subtler change with broadening to higher BE, attributed to the possible formation of a Ni/Se compound (NiSe). The Ni 2p spectra are difficult to interpret due to the presence of several satellite features. Although a change in line shape is observed in comparison to a reference Ni film, which could be due to Ni-oxidation or the reaction between Ni and a high electronegativity element such as Se. Additionally, the shift to higher BE may be due to charge transfer between the high work function Ni and PtSe2. The Ni 2p spectra and further discussion can be found in the Supporting Information. Taken together this suggests the reduction of PtSe2 by Ni, resulting in NiSe formation and Pt metal.

In the case of Ti, a much more subtle change in chemical state is observed. After Ti deposition, the PtO peak increases significantly and is accompanied by a higher BE feature consistent with PtO2. The Se 3d also shows a significant increase in oxidation with an increase in SeO and the appearance of a higher BE SeO2 peak. The Ti 2p spectra (not shown) show the complete conversion of Ti to TiO2, which is expected due to the oxygen-gettering nature of Ti. The presence of excess oxygen in the TiO2 film in turn oxidizes the underlying PtSe2.

Last, to confirm the presence of Ni in the reacted region, EDX was conducted on the TEM cross-section lamellas prepared with the Ni/Au-contacted devices before and after annealing at 250 °C in both FG and inert environment at 250 °C. Figure shows the compositional analysis performed via scanning TEM (STEM)-based EDX spectroscopy. Figure a,f shows the high-angle annular dark-field (HAADF) images of the region under study, which is at the edge of the metal contact. Figure b–d,g–j show the spectra for Au, Pt, and Se. No significant interdiffusion is observed for these elements after annealing, and the mapping results are fairly similar. Considering that Au and Pt have a similar atomic weight, what would seem to be diffusion of Pt in Au is actually just an artefact from the EDX. This is not an issue as the focus is on the behavior of Ni with PtSe2. Figure e,f shows clear Ni diffusion upon annealing. Before annealing, the Ni spectra is clearly confined between the PtSe2 and the Au region. Whereas after N2 annealing, significant diffusion of Ni into the Au and PtSe2 regions is observed. This is consistent with the diffusion seen in the optical images of c-TLM structures (Figure ).

Figure 6

Compositional analysis performed via STEM-based EDX on Ni/Au-contacted PtSe2 lamellas before and after annealing at 250 °C in N2. (a,f) shows the HAADF image of the region under study before and after annealing respectively. (b–e) and (g–j) refer to the Au, Pt, Se, and Ni signal, before and after annealing, respectively. The scale bar is 25 nm.

Discussion

The interaction between top contacts and TMDs is significantly different to contacts on conventional semiconductors due to the van der Waals nature of the TMD surface. Density functional theory simulations predict that because of the lack of surface covalent bonds for most deposited metals, a van der Waals gap exists between the TMD surface and the metal contact.[36] This van der Waals gap acts as a tunnel barrier for carrier injection into the TMD channel, in addition to the Schottky barrier between the metal and TMD,[37] increasing total contact resistance.

A study by Das et al.[38] showed that the Schottky barriers between metals and MoS2 are not consistent with the energy difference between the metal work function and the TMD Fermi level. They showed that even metals which would be expected to act as p-type contacts (i.e., Ni and Pt) instead exhibited electron injection consistent with n-type contacts. This was ascribed to Fermi level pinning in the MoS2 close to the conduction band minimum at the metal/MoS2 interface.[39]

Previous XPS studies on MoS2 and WSe2, considering a range of various metals,[40−43] have shown that similar reactions between the as-deposited metal and a TMD are possible. Nonetheless, an interface reaction does not necessarily result in a lower contact resistance because of the defective nature of exfoliated TMD flakes. However, a complete transformation of the TMD below the contact might mitigate the effect of contact resistance. Kappera et al.[7] demonstrated low contact resistance in a MoS2 FET varying the phase of MoS2 below the contact from semiconductive (2H) to metallic (1T). Similarly, here we showed a Ni/PtSe2 alloy at the contact and PtSe2 in the channel region. However, in this work, a contact resistance analysis was not possible. The PtSe2 under study had a thickness of ∼6–10 nm, and in this thickness range, the PtSe2 behaves as a semimetal, therefore it is not contact-resistance limited. Once the PtSe2 growth processes are tuned further to achieve few-layer (<3 L) films, a more detailed contact resistance study could be performed. Also, the annealing temperature and ambient need to be optimized in order to control the diffusion/reaction of Ni.

A behavior similar to Ni and PtSe2 has also been observed for Ag/MoS2.[44] In that work, the samples were annealed between 250 and 350 °C in a rapid thermal annealing furnace for 300 s and the contact resistance values decreased after the thermal treatment because of Ag diffusion and doping below the contact regions. Although Ag does not react with MoS2 at room temperature,[45] Souder and Brodie[46,47] showed that Ag diffused in bulk MoS2 after annealing at 400–600 °C for 5 min, estimating a concentration of approximately 1019 cm–3.

It is unclear if the Ni is acting as a dopant for PtSe2, however, the metal–TMD reaction could be of great interest to improve the contact resistance of TMD-based devices.

Conclusions

In this study, the Ti/PtSe2 and Ni/PtSe2 interfaces were explored, along with the impact of post-metallization anneals in FG and inert ambient. Electrical and chemical characterization show that Ti is unreactive after annealing, even at relatively high temperature, whereas Ni readily reacts with PtSe2 to form NiSe and reduced PtSe/Pt/metal. The reaction is enhanced with high temperatures or a hydrogen-rich environment, which can cause the Ni to diffuse laterally into the PtSe2. The metal/PtSe2 alloying is a possible solution for reducing contact resistance in few-layer devices based on PtSe2 and other TMD materials.

Experimental Section

Growth

All PtSe2 samples were grown over large area. Pt metal evaporated onto a SiO2/Si substrate. This was then converted in a furnace with a Se pressure of ∼1 mbar at 400 °C. The specifics of the growth are discussed in detail elsewhere.[48]

Metal Deposition

c-TLM metal contacts were patterned using standard photolithography followed by e-beam evaporation of the metal and a lift-off process. For these devices the targeted thicknesses were 10/90 nm for both Ti/Au and Ni/Au at a background pressure of ∼10–5 mbar. For the XPS samples, the targeted thicknesses were 3 nm for both Ti and Ni. No Au was used in the XPS samples as it does not interface with the PtSe2 film and the added metal thickness would exceed the photoelectron escape depth (of ∼5 nm). After an initial characterization, each sample was cleaved into three pieces to maintain consistency. FG anneals were performed at 150, 250, and 350 °C for 1 h using 5% H2/95% N2. The same annealing conditions were applied again in an inert environment (0% H2/100% N2).

Characterization

XPS characterization was performed using monochromated Al Kα X-rays from an Omicron XM1000 MkII X-ray source and Omicron EA125 hemispherical analyzer with ±0.05 eV resolution. A take-off angle of 45°, acceptance angle of 8°, and pass energy of 15 eV were employed during spectral acquisition. The BE scale was referenced to the adventitious carbon species in the C 1s core level (285.8 eV). Spectra were deconvolved using AAnalyzer,[49] a curve fitting software.

For structural analysis, cross-section samples were obtained by using the Dual Beam Helios NanoLab 600i system from FEI, using a Ga ion beam. Layers of the protective material were used consisting of electron beam deposited C, Pt, and ion beam deposited C. Lamellas were thinned and polished at 30 kV 100 pA and 5 kV 47 pA, respectively. XTEM imaging was carried out using a JEOL 2100 HRTEM, operated at 200 kV in the bright field mode using a Gatan double tilt holder. EDX mapping was carried out using a Thermo Fisher Scientific Titan Themis operated of 300 kV in the STEM mode using the Bruker SuperX silicon drift detector.