Perfluorodecyltrichlorosilane-based seed-layer for improved chemical vapour deposition of ultrathin hafnium dioxide films on graphene

We investigate the use of perfluorodecyltrichlorosilane-based self-assembled monolayer as seeding layer for chemical vapour deposition of HfO2 on large area CVD graphene. The deposition and evolution of the FDTS-based seed layer is investigated by X-ray photoelectron spectroscopy, Auger electron spectroscopy, and transmission electron microscopy. Crystalline quality of graphene transferred from Cu is monitored during formation of the seed layer as well as the HfO2 growth using Raman spectroscopy. We demonstrate that FDTS-based seed layer significantly improves nucleation of HfO2 layers so that graphene can be coated in a conformal way with HfO2 layers as thin as 10 nm. Proof-of-concept experiments on 200 mm wafers presented here validate applicability of the proposed approach to wafer scale graphene device fabrication.


Results and Discussion
The representative Raman spectra of a graphene layer transferred to 300 nm SiO 2 /Si(100) wafers before and after FDTS coating are shown in Fig. 1. A strong G and 2D bands and a small D band confirm a high quality of the as-transferred layer 20,21 . After coating with FDTS the D band remains barely visible. G and 2D bands are narrow (FWHM(G): 14 cm −1 , FWHM(2D): 36 cm −1 ) proving that the FDTS coating process does not negatively affect the crystalline quality of graphene sheet. The observed red shift of about 8 cm −1 after SAM formation may be associated with a process induced electrostatic doping of graphene 22,23 .
The optical microscope image of a transferred graphene layer after the FDTS deposition process is shown Fig. 2a. Caused by the growth and transfer process, a small number of multi-layer islands and few cracks and holes are visible in the microscopic pattern. However, the FDTS based self-assembled monolayer is not visible by optical microscope imaging. In order to demonstrate the successful coating of graphene surface with FDTS, combined secondary electron microscopy (SEM) and Auger electron spectroscopy (AES) measurements were performed, as illustrated in Fig. 2b,c, respectively. The bright region in Fig. 2b (blue rectangle) is associated with a hole in the graphene layer whereas the remaining area is covered with graphene. As expected, the corresponding AES spectrum recorded in region 3 (Fig. 2c) shows intense Si and O signals originating from the SiO 2 substrate and weaker C signal due to the absence of graphene. The additional scans performed on the remaining area (regions 1 and 2) show reduced Si and O signals and an intense C peak due to the closed graphene layer. A fluorine signal of similar intensity is detected in all investigated regions proving that the FDTS SAM was coated  homogeneously on the substrate surface. More detailed chemical analysis of the SiO 2 /graphene/FDTS interfaces was performed by X-ray photoelectron spectroscopy (XPS), as shown in Fig. 3.
Overview XPS spectra of pristine transferred graphene layers and of the processed samples were performed to monitor the chemical induced changes by FDTS coating, as illustrated in Fig. 3a. The chemical differences are clearly visible. In the spectra of the pristine graphene layer no F peak was monitored, whereby the O1s peak and the Si peaks are visible due to the SiO 2 substrate which can be detected through the graphene layer. After the FDTS coating process, a clear F1s appears and the O1s peak and Si 2s and 2p peaks are attenuated. Based on these XPS results, we can also conclude, that no additional elements are incorporated during the FDTS deposition into the graphene layer (Fig. 3a).
A detailed compositional analysis of the C1s peak is presented in Fig. 3b. The pristine graphene (Fig. 3b) shows the typically XPS spectrum of C1s with the presence of a C-C peak around 284.8 eV which is characteristic for sp 2 bonded carbon atoms 24,25 . Due to the hydrocarbon contamination by air exposure, an additional C-O peak is recorded and carbonyl and carboxyl groups caused by grain boundary defects are visible 24,25 . The spectrum of the C1s signal after FDTS coating is completely different. Now, the typical XPS spectrum of FDTS SAM is recorded. The main peak from graphene (284.8 eV) is significantly reduced and new peaks are detected at larger binding energies. These peaks are attributed to carbon-fluorine bonds: CF 2 functional backbone group and CF 3 tail group at energy values of 291 eV and 293 eV respectively [26][27][28][29] (see curve fitting in Fig. 3b). The oxidized species C-O are detected at a binding energy of around 286 eV. The intense CF 2 signal at 297 eV is caused by the fluorine-chains which were deposited on the graphene surface.
To determine the impact of temperature to the stability of the FDTS based seed layer, a study at different temperatures was performed in ultra-high vacuum. The XPS spectra were recorded in situ. The sample was firstly investigated at room temperature; afterwards the first annealing step was performed at 200 °C for 5 minutes. Since the HfO 2 deposition process after seed layer formation is performed at 400 °C, the second annealing step was set also to 400 °C for 5 minutes. As illustrated in Fig. 4, XPS measurements were directly performed in-situ in order to record the chemical changes of the coated FDTS layer.
The intensity of the aromatic C-C signal is reduced after the annealing at 400 °C, as illustrated in Fig. 4a. Also the CF 2 signal is reduced; while the intensity of the oxidized carbon species (C-O) rises with temperature. As to be seen in the spectra, the characteristic oxidized carbon species (COOR) signals are also observed at a binding energy of 288 eV 28,29 . These changes in the C1s spectrum are caused by the cracking of the long fluorine chains in FDTS at 400 °C. Due to this cracking, oxygen from SiO 2 /Si substrate is adsorbed, resulting in an increase of the signal of oxidized carbon species C-O and COOR and a peak shift of the C-O component 28,29 . In order to study the chemical changes more detailed, the F1s signal is also evaluated at the annealing temperature of 200 °C. The results of the F1s signal analysis are shown in Fig. 4b. It is obvious that the binding energy does not shift with increasing temperature, but the intensity of the signal decreases with raising temperature. As illustrated in Fig. 4b the peak area as well as the full width at half maxima decrease with raising temperature, meaning that the desorption of fluorine species starts already at 200 °C and is even more pronounced at 400 °C.
As the decomposed FDTS chains could interact with graphene, a Raman analysis was performed after the annealing procedure in order to investigate the quality of the graphene layer. As illustrated in Fig. 4c, no destruction of the graphene layer occurs when the FDTS layer is decomposed by temperature.
The following conclusions can be drawn from the discussions above: The FDTS seed layer covers graphene homogenously at room temperature and this process does not affect the crystalline quality of graphene. During thermal treatment, long fluorine chains partially decompose and some fluorine species desorb from the surface as it is evidenced by reduced F1s signal. After annealing the high crystalline quality of graphene remains unaffected as proved by Raman spectroscopy investigations.
The next section is dedicated to the growth of HfO 2 by CVD on pretreated graphene layers. As demonstrated by Lukosius et al. 30 , the direct growth of HfO 2 on transferred graphene is feasible without any initial pretreatment of the graphene surface with the restriction, that HfO 2 films have to be thicker than 20 nm to be uniform and closed. Due to nucleation sites created by the FDTS based seed layer, the minimum HfO 2 film thickness should be drastically reduced. In the first step, we chose a thickness of nominally 10 nm HfO 2 to test the influence of the FDTS nucleation capability. Two graphene samples were transferred to pre-structured 75 nm TiN/Si substrates, which will be used for ongoing electrical characterization. After the transfer process one sample was coated with FDTS, while the other sample was left pristine. Finally, 10 nm HfO 2 were deposited by CVD onto both samples.
In order to investigate the quality of the HfO 2 layer, transmission electron microscopy (TEM) was applied. The samples were firstly coated with an Al-layer to protect the sample and then a lamella was prepared. According to the high resolution TEM images, there is a huge difference between the two samples, as illustrated in Fig. 5. While the sample without FDTS coating (Fig. 5a,c) shows a non-uniform and very rough HfO 2 layer, the sample with FDTS coating (Fig. 5b,d) shows a closed, uniform and smooth HfO 2 layer. The FDTS coating clearly optimizes the growth of thin layers of HfO 2 on graphene. To proof the chemical composition of the deposited HfO 2 layer an EDX-analysis (Fig. 5e) and an Auger electron spectroscopy (Fig. 5f) were done. In the AES-spectrum no Al peak can be seen, because the measurement was done before coating the sample with Aluminium for TEM. The Auger analysis was measured before coating the sample with aluminium. Both measurements show a thin HfO 2 layer, grown on top of the sample.

Conclusions
The pre-treatment, done by a modified FDTS process can be used to improve the growth of high-k dielectric layers on graphene with an excellent quality. We have demonstrated, that the minimum thickness of closed and uniform grown of HfO 2 by CVD on pre-treated FDTS graphene layers can be scaled down to 10 nm. The results implicate that the pre-treatment with FDTS provide a new route to deposit thin conformal layers of HfO 2 dielectrics on graphene surfaces up to 200mm wafer size.

Methods
Sample preparation. Commercially available graphene was transferred from Cu growth substrates by detaching graphene from Cu foils with poly(methyl methacrylate) (PMMA) assisted wet etching in Ammonium persulfate (80 mg/ml in water) 31 . After etching the PMMA/graphene stack was rinsed in fresh distilled water for several times to remove etchant solution. The stack with a size of 1 × 1 cm 2 was then transferred to target substrate and immersed in acetone and finally rinsed in isopropyl alcohol (IPA) to remove the polymer. Target substrates in this study were 200 mm Si(100) wafers covered with either 300 nm thermal SiO 2 or 75 nm CVD TiN layers.  exposure to FDTS and DI water at a process pressure of 40 Torr and for a duration of 300 s. The flow rate of the two carrier gases are 20 sccm for FDTS and 100 sccm for DI water. As O 2 plasma readily introduces defects in the crystalline lattice of graphene, this step was omitted in the process flow.
HfO 2 growth was performed in an atomic vapor deposition tool with Hf(NMeEt) 4 precursor and oxygen as the reactive gas 9 . The substrate temperature during HfO 2 deposition was set to 400 °C for a duration of 3 min. The deposited layers are polycrystalline. More details about the depositions of HfO 2 by CVD have been reported elsewhere 32 .

Characterization.
A Renishaw InVia micro-Raman spectrometer equipped with a 514 nm wavelength laser, an 1800 lines/mm grating and 50x objective was used to acquire Raman spectra. XPS measurements were performed using PHI 5000 Versa Probe II (ULVAC-PHI) with monochromatic Al-Kα irradiation. The chamber pressure was ~3 × 10 −8 Torr and the spectra were measured with X-ray beam size of 100 × 100 μ m. A PHI 670 Auger Nanoprobe (Perkin Elmer) was used for Auger electron spectroscopy. The spectra were measured with a beam voltage of 10 keV. Structural investigations and EDX-measurements were done by transmission electron microscopy using a FEI Tecnai Osiris instrument operated at 200 kV.