Thermal Stability of Titanium Contacts to MoS₂

Abstract

Thermal annealing of Ti contacts is commonly implemented in the fabrication of MoS₂ devices; however, its effects on interface chemistry have not been previously reported in the literature. In this work, the thermal stability of titanium contacts deposited on geological bulk single crystals of MoS₂ in ultrahigh vacuum (UHV) is investigated with X-ray photoelectron spectroscopy and scanning transmission electron microscopy (STEM). In the as-deposited condition, the reaction of Ti with MoS₂ is observed resulting in a diffuse interface between the two materials that comprises metallic molybdenum and titanium sulfide compounds. Annealing Ti/MoS₂ sequentially at 100, 300, and 600 °C for 30 min in UHV results in a gradual increase in the reaction products as measured by XPS. Accordingly, STEM reveals the formation of a new ordered phase and a Mo-rich layer at the interface following heating. Due to the high degree of reactivity, the Ti/MoS₂ interface is not thermally stable even at a transistor operating temperature of 100 °C, while post-deposition annealing further enhances the interfacial reactions. These findings have important consequences for electrical transport properties, highlighting the importance of interface chemistry in the metal contact design and fabrication.

Graphical Abstract

INTRODUCTION

MoS₂ has gained tremendous attention for device applications, offering a high degree of electrostatic control for channel length scaling and the ability to tune electronic and optical properties by simply varying the number of layers.^1–3 It is widely known that the metal/MoS₂ contact interface is critical in determining device performance, especially at short channel lengths.^4–7 Annealing devices after contact deposition is a commonly implemented fabrication step in order to improve performance.^1,8–11 Baugher et al.¹¹ claimed that Schottky behavior was eliminated following vacuum annealing of devices with Ti/Au contacts. A decreased contact resistance for Ag contacts to MoS₂ was observed by Abraham and Mohney,⁸ who speculated that annealing causes the diffusion of Ag into MoS₂, which results in doping under the contact. In some cases, the effects of annealing are reported to be dependent on the contact metal. For example, Radisavljevic et al.¹ reported a factor of 10 decrease in resistance after annealing Au contacts to MoS₂, while Ti/Au contacts were reported to behave better without annealing. Similarly, English et al.¹⁰ demonstrated very low resistance for annealed Au contacts while Ti contacts behaved poorly.

Changes in device behavior upon annealing are undoubtedly linked to changes in the chemical composition and structure of the metal/MoS₂ interface. Interface chemistry is typically overlooked in discussions of device fabrication and electrical properties. Previous studies of the effects of annealing on interface chemistry of UHV-deposited metals on MoS₂ have been performed with Cr, Mn, and Fe by Lince et al.^12–15 These studies demonstrate that thermal annealing has a drastic effect on interface chemistry, particularly for metals that exhibit reactivity with MoS₂. Both Cr and Mn react with MoS₂ in their as-deposited condition at room temperature. Heating the Cr/MoS₂ and Mn/MoS₂ interfaces has been found to drive the reaction to completion. In contrast, heating the nonreactive Fe/MoS₂ interface was found to decouple the short-range Fe–S chemical interactions at the interface.¹⁵

In this work, we examine the effects of thermal annealing on the interface chemistry of Ti contacts to MoS₂. Ti contacts are commonly used in MoS₂ devices and often annealed at temperatures ranging from 200 to 400 °C.^10,16,17 McGovern et al.¹⁸ and McDonnell et al.¹⁹ have demonstrated that a reaction occurs at room temperature when Ti is deposited on MoS₂ in UHV, resulting in the formation of metallic Mo and Ti_xS_y compounds. Wu et al.²⁰ arrived at a similar conclusion with density functional theory and scanning transmission electron microscopy (STEM) with electron energy loss spectroscopy (EELS). Nevertheless, the literature currently lacks reports on the effects of annealing on the interface chemistry of Ti/MoS₂, which is essential for proper interpretation of device data.

RESULTS AND DISCUSSION

MoS₂ geological crystals were used for this work as described in the Experimental Section. Before investigating the effects of thermal annealing, we closely examine the chemistry of the interface in its as-deposited condition. Fitted XPS data of pristine and as-deposited Ti/MoS₂ is shown in Figure 1. Following the deposition of a thin layer of Ti in UHV, the Mo⁰ chemical state is observed in the Mo 3d spectrum. This state appears at 227.83 eV, whereas the Mo–S chemical state is at 229.50 eV. New chemical states in the S 2p spectrum correspond to Ti–S reaction products. This is consistent with previous reports by McGovern et al.¹⁸ and McDonnell et al.¹⁹ Calculations by McGovern et al.,¹⁸ Lince et al.,²¹ and Gonbeau et al.²⁴ conclude that the reaction to form titanium monosulfide (TiS) from Ti and MoS₂ is thermodynamically favorable at 298 K having a Gibbs free energy change of −24.1 kcal/mol. Reference spectra are only available for TiS₂ and TiS₃ compounds^22–25 but not for the other sulfides that exist in the Ti–S system.²⁶ TiS₂ has its 2p_3/2 component at 160.8 eV.^23,24 TiS₃ is characterized by two S 2p chemical states, one corresponding to the S²⁻ atom at 161.2 eV and one corresponding to the S₂²⁻ disulfide cluster at 162.5 eV.²³ Our S 2p spectra do not consistently match with either indicating that a different stoichiometry occurs. Our spectra indicate that new chemical states exist with 2p_3/2 components at 162.49 and 163.36 eV. The S–Mo chemical state is at 162.34 eV. Both the Mo–S and S–Mo states in Mo 3d and S 2p, respectively, exhibit a shift of 0.34 eV to higher binding energy. This is indicative of a shift in the position of the Fermi level, which indicates that the Ti overlayer and its reaction products result in n-type doping of the substrate. Ti and Mo metals have been shown to be n-type dopants in MoS₂.^27,28

Figure 1.

Mo 3d and S 2p core levels before and after Ti deposition.

Atomic-resolution annular dark field STEM (ADF-STEM), X-ray energy dispersive spectroscopy (EDS), and EELS spectrum-imaging (SI) data acquired from a cross section of the as-deposited Ti/MoS₂ are shown in Figure 2. The ADF-STEM image reveals the presence of a disordered Mo- and S-rich layer in contact with MoS₂ (Figure 2a). This layer exhibits a diffuse interface with a Ti-rich disordered layer at the surface. The EDS measurements along with EELS line-scans provide unambiguous evidence of Ti diffusion into MoS₂ as well as S and Mo out-diffusion into the Ti layer (Figure 2a–d). Similar STEM/EELS analysis of the Ti/MoS₂ interface by Wu et al.²⁰ has shown the presence of Ti atoms in the topmost five layers, while the chemical composition of the metal contact was assumed to be pure Ti. This assumption would appear to be inconsistent with our results, which show a substantial presence of Mo and S in the Ti layer (see Figure 2c,d). We note that Wu et al.²⁰ acknowledged that the density of Ti increased with the distance away from the Ti–MoS₂ interface, and therefore, these results may in fact be consistent. Cross-sectional HAADF-STEM analysis shows an ~7 nm thick Mo/S-rich amorphous layer due to the damage of MoS₂ during Ti deposition.

Figure 2.

(a) Cross-sectional ADF-STEM image of as deposited Ti/MoS₂ showing disordered Ti-rich and Mo/S-rich layers, (b) 90° rotated ADF-STEM image showing the dotted white line where an EDS and EELS line-scan was simultaneously acquired, and (c) the corresponding EDS line profiles of the Ti-K, Mo-K (enlarged five times), and the overlapped Mo-L and S-K edges, and (d) EELS line profiles of S-L and Ti-L edges. False colors are added in the ADF-STEM images to aid the eye.

The XPS spectra acquired during the in situ sequential annealing are shown in Figure 3a with the intensity ratios shown in Figure 3b. A substantial increase in reaction products is observed after 100 °C exposure as highlighted in Figure 3c. While this change is minor in comparison with the other annealing temperatures, it may have important implications for reliability of MoS₂ transistors since 100 °C is within the range of transistor operating temperatures.⁷ At 300 °C, we observe a clear increase in the Mo⁰ intensity relative to that of the Mo–S state, accompanied by an overall broadening of the S 2p spectra as more Mo–S bonds are broken to form new Ti_xS_y states. Figure 3b highlights the increase in the total S/MoS₂ ratio, indicative of the increased difiusion of S to the surface. At 600 °C, the Mo–S chemical state is nearly completely consumed. We note that additional 100 °C anneals after annealing to higher temperatures exhibited no detectable change in interface chemistry. This suggests that the interface could be stabilized to avoid changes at operating temperatures. However, we highlight that earlier work has demonstrated that as-deposited Ti/Au contacts perform better before annealing.¹

Figure 3.

(a) XPS spectra acquired following 30 min anneals at each temperature. These were performed sequentially on the same sample. (b) Intensity ratios based on the data in panel (a), whereas panel (c) highlights the changes that occur at 100 °C.

The Ti 2p spectra exhibit broadening due to the superposition of the different Ti–S states and show a decrease in intensity with increasing temperature. The reason for the apparent increase in the Ti/MoS₂ intensity ratio in Figure 3b can be understood as the result of the MoS₂ signal decreasing more significantly than the decrease of the Ti signal. The decrease in Ti intensity can be explained as a result of Ti diffusion into the MoS₂ substrate and its potential agglomeration. Durbin et al.¹³ reported that, in the case of Cr/MoS₂, annealing to 650 °C drives the reaction between Cr and MoS₂ while simultaneously causing coalescence of the Cr-containing portion of the film. Additionally, the core-level shift exhibited by the sample increases by another +0.36 eV after the final heating step at 600 °C, amounting to a total shift of 0.70 eV to higher binding energy compared to the pristine MoS₂ surface.

A cross-sectional ADF-STEM image of a sample annealed at 400 °C is shown in Figure 4a. The disordered Mo/S-rich layer on the top of MoS₂ in the as-deposited sample was partially recrystallized. The fast Fourier transform (FFT) images in Figure 4b,c, obtained from locations indicated in Figure 4a, suggest that this recrystallized layer has a structure similar to that of MoS₂. In addition, an approximately 2 nm thick layer with an enhanced contrast was found on the top of the recrystallized layer. The EELS line-scan in Figure 4d,e shows a fast decrease in S and Ti in this thin layer, while the EDS mapping in Figure 4f–i confirms that it is Mo-rich.

Figure 4.

(a) Cross-sectional ADF-STEM image of Ti/MoS₂ after 30 min anneal at 400 °C showing a Mo-rich layer and a partially recrystallized layer grown out from the disordered Mo/S-rich layer, (b,c) FFT images of the white dotted-line framed regions in panel (a), (d) 90° rotated ADF-STEM image showing the dotted white line where an EELS line-scan was acquired, and (e) the corresponding EELS line profiles of the S-L and Ti-L edges, the black dash-dot lines showing the fast dip of S and Ti in the Mo-rich layer. (f) ADF-STEM image showing the white dotted-line framed region where EDS mapping was acquired and (g) the corresponding Mo-K, (h) Ti-K, and (i) the overlapped Mo-L and S-K edge maps. False colors are added to aid the eye.

CONCLUSIONS

The results shown here indicate that the Ti/MoS₂ interface is highly reactive: a disordered, compositionally graded layer, composed of Mo₀ and Ti_xS_y species, forms at the surface of the MoS₂ crystal following the deposition of Ti. Thermal annealing in the 100–400 °C temperature range results in further chemical and structural changes at the Ti/MoS₂ interface and causes a degradation of the Ti metal contact. STEM provides evidence for the formation of new ordered phases upon heating to 400 °C that did not exist in the as-deposited sample. Our work strongly suggests that the effects of thermal annealing on the compositional and structural evolution of the Ti/MoS₂ interface must be taken into account when developing a post-deposition annealing process in the device fabrication steps and when interpreting changes in device characteristics after annealing.

EXPERIMENTAL SECTION

MoS₂ geological single crystals (SPI²⁹) were cleaned by mechanical exfoliation, mounted onto Mo plates with Ta foil, and then immediately loaded into a Scienta Omicron UHV³⁰ system described elsewhere with a base pressure of 2 × 10⁻⁹ mtorr.³¹ XPS data was acquired at a pass energy of 50 eV with a monochromated 1486.7 eV Al Kα X-ray source. XPS data was acquired on the pristine samples prior to metal deposition. A thin layer of metal was deposited by electron beam evaporation (Mantis Quad EV-C) in our UHV system, and XPS data was acquired in situ after deposition. Samples were heated in the UHV chamber to 100, 300, and 600 °C for 30 min at each temperature with ramp and cool times of approximately 30 min each. XPS measurements were performed following each heat treatment.

Two samples were prepared for STEM: the first one did not undergo heat treatment following deposition, and the second sample was heated directly to 400 °C for 30 min. An FEI Nova NanoLab 600 dual-beam scanning electron microscopy and focused ion beam (SEM/FIB) system was employed to prepare cross sections of the samples. Using electron beam deposition, 10 nm carbon and subsequently 100 nm Pt were deposited on the top of the sample to protect its surface followed by 2 μm Pt capping by ion-beam-induced deposition. To reduce Ga-ion damage in the final step of FIB preparation, the STEM samples were thinned with 2 kV Ga ions using a low beam current of 29 pA and a small incident angle of 3°. An FEI Titan 80–300 STEM/TEM equipped with a monochromator and a probe spherical-aberration corrector was employed to perform atomic-resolution STEM imaging, X-ray energy dispersive spectroscopy (EDS), and electron energy loss spectroscopy (EELS) spectrum-imaging (SI). Atomic-resolution annular dark field (ADF) STEM images were acquired with an operating voltage of 300 kV, probe convergence angle of 14 mrad, and collection angle of 34–195 mrad.