Imaging microscopic electronic contrasts at the interface of single-layer WS$_2$ with oxide and boron nitride substrates

TL;DR

This study employs photoemission electron microscopy to analyze the microscopic electronic contrasts at the interfaces of single-layer WS$_2$ with various substrates, revealing insights into band alignment crucial for device engineering.

Contribution

It introduces a microscopy-based approach to directly image and analyze the electronic structures at WS$_2$ interfaces with different substrates, advancing understanding of substrate effects on 2D materials.

Findings

Identified microscopic regions with clean interfaces using work function and X-ray imaging.

Measured valence band offsets to determine band alignments.

Discussed implications for vertical band structure engineering.

Abstract

The electronic properties of devices based on two-dimensional materials are significantly influenced by interactions with substrate and electrode materials. Here, we use photoemission electron microscopy to investigate the real- and momentum-space electronic structures of electrically contacted single-layer WS$_2$ stacked on hBN, SiO$_2$ and TiO$_2$ substrates. Using work function and X-ray absorption imaging we single-out clean microscopic regions of each interface type and collect the valence band dispersion. We infer the alignments of the electronic band gaps and electron affinities from the measured valence band offsets of WS$_2$ and the three substrate materials using a simple electron affinity rule and discuss the implications for vertical band structure engineering using mixed three- and two-dimensional materials.

Tables

Table 1. Table 1: Summary of dielectric constants ϵ italic-ϵ \epsilon , quasiparticle band gaps E g subscript 𝐸 𝑔 E_{g} , electron affinities χ 𝜒 \chi and band offsets with respect to K ¯ ¯ K \bar{\mathrm{K}}  of SL WS 2 for SiO 2 , TiO 2 and hBN.

Substrate	$ϵ$	$E_g$ (eV)	$\chi$ (eV)	$E-E_{\overline{K}}^{{\mathrm{WS}}_2}$ (eV)
SiO2	3.9	9.7 Kresse et al. (2012)	0.9 Williams (1965)	4.0
TiO2	113	3.4 Landmann et al. (2012)	4.8 Scanlon et al. (2013)	1.0
hBN	4.0	6.0 Arnaud et al. (2006)	2.0 Sup Choi et al. (2013)	1.2

Full text

Semiconducting transition metal dichalcogenides (TMDs) at the single-layer (SL) limit offer entirely new possibilities for fabricating field-effect transistors with atomically thin gating materials and sophisticated contact electrode geometries leading to nanoscale engineered unipolar and ambipolar charge carrier transport Radisavljevic et al. (2011); Wang et al. (2012); Jariwala et al. (2013, 2014); Gong et al. (2014); Allain et al. (2015). These properties are determined by the electronic band alignments at the vertically stacked interfaces of the active device components, which can be tailored using junctions of TMDs in combination with other TMDs Schlaf et al. (1999), TMDs and oxides McDonnell et al. (2014); Ulstrup et al. (2016) as well as mixed two-dimensional (2D) and three-dimensional (3D) materials Jariwala et al. (2016). Understanding how key band alignment parameters such as the valence band (VB) offsets, quasiparticle band gap energies $E_{g}$, and electron affinities $\chi$, depend on the interface type and quality as well as environmental screening remains an important issue for band structure engineering utilizing 2D materials Guo and Robertson (2016).

The interplay of these parameters on the electronic properties of SL TMD devices is ideally investigated using spectromicroscopic probes of the electronic structure Klein (2012). Photoemission electron microscopy (PEEM) is a powerful method in this regard because it offers fast switching between real space and $k$-space imaging modes with work function, core level absorption and VB contrasts Fujikawa et al. (2009); Ulstrup et al. (2016); Koch et al. (2018). The use of $k$-resolved PEEM for performing microscale angle-resolved photoemission spectroscopy (microARPES) has been an essential tool for observing band structures of SL and few-layer MoS2 Jin et al. (2013, 2015) and WSe2 Yeh et al. (2015) exfoliated on SiO2 substrates.

Here, we use the SPECS PEEM P90 microscope installed at the Microscopic And Electronic STRucture Observatory (MAESTRO) at the Advanced Light Source to investigate the electronic properties of vertical stacks based on SL WS2 transferred on oxide and hexagonal boron nitride (hBN) substrates. The thickness of WS2 is checked before and after transfer using photoluminescence and Raman spectroscopy as shown in our earlier works McCreary et al. (2016); Ulstrup et al. (2016). The influence of the dielectric environment on the electronic properties of SL WS2 is studied using insulating 300 nm SiO2 on Si (SiO2/Si) with relative permittivity $\epsilon_{\mathrm{SiO}_{2}}=3.9$ and 0.5 wt % Nb-doped rutile TiO2(100) (Shinkosha Co., Ltd) with $\epsilon_{\mathrm{TiO}_{2}}=113$ as supporting substrate. We assemble WS2/hBN heterostructures ($\epsilon_{\mathrm{hBN}}\approx 4$) on both oxides utilizing a similar transfer technique as previously reported Ulstrup et al. (2016); Katoch et al. (2018) and as described further in the Supplementary Material. On SiO2 we deposit an Au electrode that is contacted to both SL WS2 and hBN on the side (see optical microscope image in Fig. 1(a)) which is essential to avoid charging during photoemission measurements. The Nb doping of TiO2 is sufficient to prevent charging. By shorting the WS2 flake on hBN to the TiO2 we avoid using a metal electrode in this system.

The rationale of using SiO2, hBN and TiO2 as substrates for SL WS2 is three-fold: (i) These materials are commonly used in devices where they are known to exhibit strong variations in interfacial quality with other 2D materials Chen et al. (2008); Dean et al. (2010); Ulstrup et al. (2016), (ii) the dielectric properties vary strongly across the interfaces, potentially affecting the electronic bandstructure of the adjacent SL WS2 Rösner et al. (2016); Raja et al. (2017), and (iii) the quasiparticle band gaps and electron affinities are very different and thus give rise to substantially different band alignments. Here, we address these key points by first presenting PEEM measurements of electronic contrasts to identify the three types of interfaces and investigate their quality from a photoemission perspective. We then discuss $k$-resolved electronic structure measurements and use these to infer the band alignments of the systems.

The photoemission intensity variations during in situ annealing of the SiO2 supported sample to 380 *∘*C are studied in PEEM using a Hg excitation source as shown in Figs. 1(b)-(c). The average contrast levels for Au, SL WS2 and hBN areas are similar before annealing (panel (b)) making it difficult to distinguish the materials. During annealing the intensity of the Au electrode increases (panel (c)). This behavior indicates a lowering of the Au work function giving rise to higher secondary electron emission and therefore higher intensity. The reduction of secondary electron emission from WS2 on hBN during annealing indicates an increase in work function, possibly due to a change in doping caused by the desorption of water. The intensity levels from patches of WS2 on SiO2 and on hBN adjust slightly after cooling down. Most importantly, we observe that there is no sign of Au diffusion on the surface at these annealing conditions in ultra-high vacuum (UHV) at 380 *∘*C.

The same piece of transferred SL WS2 covers the SiO2 substrate in the part marked by a dashed red box in Fig. 1(a). We can therefore compare the contrast levels on both hBN and SiO2 as shown in Figs. 1(d)-(f). On WS2/hBN in panel (d) the intensity exhibits only minor fluctuations with respect to the average, as demonstrated by the line profile in Fig. 1(f). Much stronger contrasts are observed on WS2/SiO2 in panel (e), which are quantified in panel (f) as intensity fluctuations within a scale of 2 $\mu$m and a slow intensity increase over the full 30 $\mu$m range of the profile. These features are indicative of both long range and short range potential energy variations on the SiO2, which are likely caused by remaining charge impurities that inevitably form in such WS2/SiO2 interfaces Ghatak et al. (2011). Removing such strong potential energy fluctuations is essential for electronic structure measurements as this greatly reduces energy broadening of the measured bands. This may be achieved using the conductive TiO2 interface seen in the optical microscope image in Fig. 2(a) and the Hg PEEM image in Fig. 2(b) obtained after annealing to 380 *∘*C. Parts of a transferred WS2 triangle straddle both the TiO2 and the hBN flake. Representative line profiles from these two regions are compared with the SiO2 sample in Fig. 1(f) and exhibit much less fluctuations as expected for the conductive and thus more strongly screening TiO2 interface Ulstrup et al. (2016).

X-ray PEEM (XPEEM) is applied for X-ray absorption spectroscopy (XAS) and imaging of the absorption peaks of the boron K-edge and titanium L-edge in Figs. 2(c)-(f). The image in Fig. 2(c) was obtained using secondary electron contrast of the boron $\pi^{\ast}$ resonance such that bare hBN areas exhibit a high intensity Koch et al. (2018). This reveals cracks and tears in the transferred WS2 as well as dark sub-micron spots (see white arrows for a few examples in panel (c)) which are trapped bubbles that form in transferred van der Waals heterostructures Khestanova et al. (2016). The spatially resolved XAS spectra in Fig. 2(d) are obtained by integrating the intensity within the blue and magenta boxes on bare and WS2 covered hBN shown in panel (c). The expected $\pi^{\ast}$ resonance is observed in addition to a shoulder which appears after SL WS2 transfer Koch et al. (2018). Using secondary electron contrast from the t2g resonance on the TiO2 L-edge we are able to distinguish bare and WS2 covered TiO2 in Fig. 2(e). The area-selective XAS spectra over the entire edge shown in Fig. 2(f) resemble typical pristine TiO2 spectra, indicating the cleanliness of the interface Ulstrup et al. (2016).

Having established the characteristic real space electronic contrasts, we collect distinct microARPES spectra with $k$-resolved PEEM from clean areas of the three vertical interfaces WS2/SiO2 (Fig. 3(a)), WS2/TiO2 (Fig. 3(b)) and WS2/hBN (Fig. 3(c)). The WS2/hBN dispersion in Fig. 3(c) is measured on the SiO2 supported sample, but we get similar spectra from WS2/hBN on TiO2 Katoch et al. (2018). The data were obtained along the $\bar{\mathrm{K}}$-$\bar{\mathrm{\Gamma}}$ high symmetry direction of the SL WS2 Brillouin zone (BZ), permitting us to identify the global valence band maximum (VBM) at $\bar{\mathrm{K}}$ and the local maximum at $\bar{\mathrm{\Gamma}}$ as expected for SL WS2 Klein et al. (2001). Note that the energy scale is referenced to the energy of the VBM at $\bar{\mathrm{K}}$. Energy distribution curve (EDC) fits to Voigt line shapes on a linear background at $\bar{\mathrm{\Gamma}}$ provide an offset of 0.20(4) eV from the peak position to the VBM at $\bar{\mathrm{K}}$ for all interfaces as seen in Fig. 3(d). The full width at half maximum (FWHM) values for the fitted Voigt peaks demonstrate sharpest SL WS2 bands on hBN with a FWHM value of 0.39(1) eV (see arrows in Fig. 3(d)). Extensive broadening is observed across the oxides with the FWHM value more than doubled on SiO2.

Measurements along $\bar{\mathrm{M}}$-$\bar{\mathrm{K}}$ further reveal the spin-orbit split VBs at $\bar{\mathrm{K}}$ as seen in Figs. 4(a)-(c). EDC fits lead to a value for the spin-orbit splitting of 0.42(6) eV for WS2/hBN as demonstrated in Fig. 4(d), which is in agreement with other studies Ulstrup et al. (2016); Katoch et al. (2018). The linewidth broadening masks the spin-orbit splitting to such an extent that the EDC fits for SiO2 and TiO2 in Fig. 3(g) had to be performed with the peak separations constrained to the values obtained on hBN. The broad VB states of WS2/SiO2 are consistent with similar measurements on MoS2/SiO2 Jin et al. (2013, 2015); Yuan et al. (2016), which may be explained by charge impurities rigidly shifting and broadening the bands as hinted by the work function contrast in Fig. 1(e). Such effects are also present in TiO2, although less dramatic Ulstrup et al. (2016). The surface roughness in the oxides is expected to be substantially higher than in hBN Lui et al. (2009), which causes additional momentum broadening.

We determine the VBM offsets for the substrates (marked by dashed horizontal lines in Figs. 3(a)-(c)) as described in the Supplementary Material and apply the electron affinity rule as the simplest method of constructing the band alignment diagrams of our mixed 2D-3D heterojunctions with respect to the vacuum level $E_{vac}$ in Fig. 5 Klein (2012). In all cases we assume the measured quasiparticle band gap of SL WS2 on SiO2 given by $E^{\mathrm{WS_{2}}}_{g}=2.4$ eV Chernikov et al. (2015). Substrate values for $E_{g}$ and $\chi$ are given in Fig. 5 and Table 1. On both SiO2 and hBN a straddling band gap configuration appears due to the wide gaps of the substrates (see panels (a) and (c)). On TiO2 the conduction band offsets are very close and may form a staggered band gap (panel (b)), which could lead to substantial electron (hole) transfer to TiO2 (SL WS2). This may explain our previous observation of less electron doping of SL WS2 on TiO2 compared to other oxides Ulstrup et al. (2016).

This simple construction suggests that $\chi_{\mathrm{WS_{2}}}$ is substrate dependent and generally larger than a recently determined theoretical value of 3.75 eV Guo and Robertson (2016). Caution should be exercised when considering the values here because of the variation in literature values of $\chi$ for the substrates. This issue is most pronounced in the case of $\chi_{hBN}$ where we used an often cited value of 2.0 eV Sup Choi et al. (2013) in Fig. 5(c). However, a value of 1.1 eV can also be found Fiori et al. (2012) and even a negative $\chi_{hBN}$ has been suggested Powers et al. (1995). Note also that on TiO2 the Nb doping as well as annealing- and beam-induced oxygen vacancies may modify the band offsets from their intrinsic values Onda et al. (2004), which could lead to an overestimation of $\chi_{\mathrm{WS_{2}}}$. Additionally, the simple electron affinity rule may break down due to a substrate dependent quasiparticle band gap of SL WS2 or possibly due to unusually strong interfacial dipoles that vary between substrates Schlaf et al. (1999).

In conclusion, we have fabricated SL WS2/hBN heterostructures supported on SiO2 and TiO2 substrates implementing device architectures in photoemission spectromicroscopy experiments that we believe will be compatible with charge transport measurements in gated conditions and with current passing through the materials Kaminski et al. (2016). The electronic transport properties of these mixed 2D-3D junctions will be defined by the vertical band alignments which we here inferred using an electron affinity rule incorporating the measured VB offsets.

supplementary material

See online Supplementary Material for further details on PEEM measurements, for the sample fabrication procedure and for the determination of the SiO2, TiO2 and hBN VB offsets.

acknowledgement

S. U. acknowledges financial support from VILLUM FONDEN (Grant. No. 15375). R. J. K. is supported by a fellowship within the Postdoc-Program of the German Academic Exchange Service (DAAD). D. S. acknowledges financial support from the Netherlands Organisation for Scientific Research under the Rubicon Program (Grant 680-50-1305). The Advanced Light Source is supported by the Director, Office of Science, Office of Basic Energy Sciences, of the U.S. Department of Energy under Contract No. DE-AC02-05CH11231. This work was supported by IBS-R009-D1. The work at NRL was supported by core programs and the Nanoscience Institute.