Raman spectroscopic characterizations of graphene on oxide substrates for remote epitaxy

Recently, graphene layers on oxide substrates, such as single-crystal SrTiO₃, i.e., graphene/SrTiO₃, are used as a platform for remote epitaxy in creating various oxide membranes, opening an unprecedented way for functional oxide device applications.^1,2 Characterizing the properties of graphene placed on various substrates or templates is essential for the advancement of this technology since the two-dimensional (2D) carbon atoms in a honeycomb lattice are strongly influenced by the substrate material.^3–7 For example, graphene layers on hexagonal boron nitrides can achieve a few orders of magnitude higher mobilities than those on silicon substrates.⁵ Graphene layers interfacing with high-k dielectric oxides, such as SrTiO₃, where high electric fields can be applied, can also exhibit improved field-effect transistor effects.^7–10 

However, characterizing graphene layers placed on SrTiO₃ substrates has been a formidable task. Raman spectroscopy is one of the standard non-destructive tools for inspecting graphene layers since Raman spectral features can tell us about the sample's defect densities and types,¹¹ thermal properties,¹² doping levels,¹³ and lattice strain.^14,15 For example, there are three well-known Raman scattering processes of graphene, the so-called G mode (i.e., the E_2g zone-center scattering), the D mode (i.e., the in-plane A_1g zone-edge intervalley scattering), and the 2D mode (i.e., an overtone of the D mode). Note that the G, D, and 2D modes are relevant to free-carrier doping, defects, and strain, respectively.¹⁶ Observation of these peaks in Raman spectra provides a comprehensive understanding of lab-prepared graphene samples. Nevertheless, this approach has had limitations for graphene/SrTiO₃ samples because strong multiphonon scattering of SrTiO₃ makes both the D mode and the G mode peaks untraceable in micro-Raman spectroscopic measurements³ with non-confocal optics, as shown in Fig. 1(a).

In this paper, we report that Raman spectroscopy with confocal optics combined with simple spectral subtraction can be used effectively for characterizing graphene layers placed on SrTiO₃ single crystals. By aligning the confocal plane of the laser beam, we can obtain the Raman spectra of graphene layers and SrTiO₃ substrates, respectively, with a substantially reduced spectral intensity overlap. This approach of confocal Raman spectroscopy overcomes the drawback of non-confocal micro-Raman spectroscopic measurements and reveals inelastic light scattering peaks of graphene layers, enabling quantitative spectral analysis. For example, Raman spectra of two different graphene samples that are synthesized using silicon carbide (SiC) and germanium (Ge), respectively, and transferred to SrTiO₃ substrates indicate that they have distinct defect densities and types. Temperature-dependent Raman spectra reveal that graphene layers on SrTiO₃ experience strain from the substrate that is different from either silicon (SiO₂/Si) substrates or copper (Cu) foils. The outcome of this confocal Raman spectroscopic approach on graphene/SrTiO₃ provides indispensable information not only in understanding the material but also in the device application for remote epitaxy of functional oxides.

Single-layered graphene was, respectively, grown on SiC and Ge via silicon sublimation of the Si-terminated face of SiC and chemical vapor deposition on hydrogen-terminated Ge and transferred to the surface of SrTiO₃ substrates using a dry transfer method as reported in Refs. 1, 2, and 17. We used a confocal micro-Raman spectrometer (JobinYvon LabRam HR800) with a 633-nm laser excitation having a focused beam spot size of ∼5 μm, which was passed through 600 grooves/mm grating with the energy resolution ∼5 cm⁻¹ to obtain inelastic light scattering spectra of our graphene/SrTiO₃ samples. For the comparison with our confocal Raman spectroscopic results, non-confocal Raman spectroscopic measurements were carried out at 532 nm excitation using a Thermo-Scientific DXR micro-Raman spectrometer.

Figure 1(b) illustrates how the Raman spectrum of the graphene layer on SrTiO₃ is obtained using confocal Raman spectroscopy. A well-defined confocal plane along the sample's surface normal direction with a vertical accuracy of ∼0.5 μm (Refs. 18 and 19) allows individual spectral measurements of the graphene layer and the SrTiO₃ substrate, respectively, as shown in the schematic diagram of Fig. 1(b). Note that both measurements are carried out on a single graphene/SrTiO₃ sample without a need for additional SrTiO₃ reference measurement. We can distinguish the spectral features of graphene (red curve) and SrTiO₃ (blue curve) even though there are some common peaks near 1200–1700 cm⁻¹ between the two. The graphene-focused spectrum (red curve) also exhibits multiphonon peaks of SrTiO₃ due to the confocal plane being thicker than the graphene layer, i.e., 0.5 μm > 0.35 nm. Nevertheless, by subtracting the SrTiO₃-focused spectrum (blue curve) from the graphene-focused spectrum (red curve), we can obtain the graphene-only spectrum (black curve), which is consistent with the previously reported Raman spectrum of graphene.^20–23 Note that the spectrum of Fig. 1(b) clearly shows weak features, such as the D and D′ modes, which are invisible in conventional micro-Raman spectroscopic measurements, as shown in Fig. 1(a).

We measured a few different graphene/SrTiO₃ samples to see if this confocal Raman spectroscopic approach is effective in examining their qualities since the Raman spectral features are correlated with the properties of defects in graphene. Figure 2 shows the graphene-only spectra of two different samples, i.e., graphene(SiC)/SrTiO₃ and graphene(Ge)/SrTiO₃. Note that both the D mode and the D′ mode (i.e., the in-plane A_1g zone-edge intravalley scattering) peaks of the graphene(Ge)/SrTiO₃ sample are significantly higher than those of graphene(SiC)/SrTiO₃, implying that the former has larger defect densities than the latter. The intensity ratios between the D and D′ modes (I_D/I_D′) are approximately 4.5 [for graphene(Ge)/SrTiO₃] and 2.7 [for graphene(SiC)/SrTiO₃], being smaller than 7 (see Table I), implying that both samples possess predominantly vacancy-type defects rather than sp³-type defects, as discussed in Refs. 11 and 24. Nevertheless, the intensity of the 2D mode normalized by the G mode, i.e., I_2D/I_G, from the graphene(SiC)/SrTiO₃ sample is approximately eight times larger than that of the graphene(Ge)/SrTiO₃ sample, indicating that the former is higher quality graphene than the latter.^11,24

Temperature-dependent confocal Raman spectra show that the graphene layer on SrTiO₃ substrates exhibits a distinct temperature coefficient compared to other graphene layers transferred to SiO₂/Si substrates and grown on Cu foils. Figure 3 shows temperature-dependent Raman spectra of a graphene(SiC)/SrTiO₃ sample from room temperature down to 10 K using a custom-built optical cryostat. All Raman modes of the graphene are shifted to higher energies at low temperatures, which is qualitatively consistent with the previous reports of Refs. 12, 25, and 26. Figure 4(a) shows the temperature dependence of the G mode energies, which is much higher compared to the theoretically estimated intrinsic temperature dependence.²⁷ The temperature dependence of the G mode is approximately −0.049 cm⁻¹ K⁻¹ and lies in between those of graphene layers placed on SiO₂/Si, i.e., −0.016 cm⁻¹ K⁻¹ (Ref. 12), and Cu, i.e., −0.101 cm⁻¹ K⁻¹ (Ref. 25). Since these substrate materials have different thermal expansion coefficients, i.e., 2.6 × 10⁻⁶ K⁻¹ (SiO₂/Si) < 9.0 × 10⁻⁶ K⁻¹ (SrTiO₃) < 16.5 × 10⁻⁶ K⁻¹ (Cu), the shift of the G mode energies is likely due to the strain between the graphene layer and the SrTiO₃ substrate.²⁸ Figure 4(b) shows the temperature dependence of the D mode and 2D mode energies. Note that both the G mode [in Fig. 4(a)] and the D mode [in Fig. 4(b)] are blue-shifted by 15 cm⁻¹ from room temperature to 10 K, whereas the 2D mode is shifted by 31 cm⁻¹. The double resonance process is due to the electronic transition and phonon energies that are affected by strain.^25,29 The biaxial tensile strain on the graphene layer by SrTiO₃ substrates is estimated to be approximately 0.23% (Refs. 30 and 31). However, the shift of Raman peak energies can also be affected by doping and electron correlations.^32,33 In particular, the large dielectric permittivity of SrTiO₃ at low temperatures may result in significant dielectric screening effects. We expect that future Raman spectroscopic studies combined with other experimental tools will shed light on these aspects.

Using confocal Raman spectroscopy and simple spectral subtraction methods, we obtained the information of defects and strain states of graphene layers placed on SrTiO₃ substrates, which had been difficult to characterize due to the strong spectral overlap between the graphene and SrTiO₃. We suggest that this approach will shed light on the investigation of not only this graphene/SrTiO₃ system but also various 2D materials where their substrate's Raman modes interfere with those of the 2D materials.

We thank Chengye Dong, Joshua Robinson, Ji-Yun Moon, Jae-Hyun Lee, and Bernhard Keimer for providing samples, experimental help, and valuable discussions and advice. The work at the University of Kentucky was supported by the National Science Foundation (Grant No. DMR-2104296). The team at MIT acknowledges support from the Air Force Research Laboratory (Award No. FA9550-19-S-0003). N.L.K. was partially supported by the Southern Company. A.S. thanks Hian for the careful correction of this manuscript.

The authors have no conflicts to disclose.

S. Shrestha: Data curation (equal); Formal analysis (lead); Investigation (equal); Methodology (equal); Writing – original draft (lead). C. S. Chang: Investigation (equal); Methodology (equal); Project administration (equal); Validation (equal); Writing – review & editing (equal). S. Lee: Investigation (equal); Methodology (equal); Validation (equal); Writing – review & editing (equal). N. L. Kothalawala: Methodology (equal); Writing – review & editing (equal). D. Y. Kim: Funding acquisition (equal); Methodology (equal); Resources (equal); Supervision (equal); Writing – review & editing (equal). M. Minola: Methodology (equal); Resources (equal); Writing – review & editing (equal). J. Kim: Funding acquisition (equal); Project administration (equal); Resources (equal); Supervision (equal); Validation (equal); Writing – review & editing (equal). A. Seo: Conceptualization (lead); Funding acquisition (lead); Investigation (equal); Methodology (equal); Project administration (equal); Resources (equal); Supervision (equal); Writing – original draft (equal).

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