Etching processes of transferred and non-transferred multi-layer graphene in the presence of extreme UV, H2O and H2

In this work we expose transferred multi-layer graphene (i.e. graphene separated from the substrate used for graphene growth and placed onto a carrier sample) to Extreme Ultraviolet light and water, in the presence of hydrogen. It is observed that graphene oxidizes by breaking down its aromatic structure into the enol form of 1,3-diketone. Furthermore, this oxidation is compared with studies on as-grown multi-layer graphene samples, which are shown to not oxidize. Our work shows that oxidation is most likely limited to a few layers of graphene at the interface between the graphene and the substrate onto which it is transferred. This is attributed to the presence of water, trapped during the transfer process, which is dissociated by secondary electrons from the


Introduction
Graphene, a 2D hexagonal lattice structure of carbon with covalently bound atoms, has been shown to possess a number of unique physical, thermal, and chemical properties [1][2][3][4][5][6][7][8]. Defect-free single layer graphene has been shown to be impermeable to gases [9,10], whereas defective multi-layer graphene has been shown to act as a membrane [11], and can be used to transfer protons through atomic defects [12]. Multi-layer graphene as a membrane has been proposed as a potential candidate for inexpensive water desalination [13], food packing [14], and anti-corrosion paint [15].
Optically, graphene is known to have highly desirable traits such as being transparent for both visible [16] and Extreme Ultraviolet (EUV) wavelengths [17]. These properties make graphene an attractive candidate for transparent conductive electrodes, such as flexible screens for smartphones and wearable devices, and EUV optics. High EUV transparency is very desirable, as most EUV optics require a layer to protect them from surface chemical processes, induced by high photon-energy radiation [18].
Single layer graphene, known for its low chemical reactivity when defect free, is highly susceptible to the formation of defects during growth, via chemical vapor deposition, and graphene transfer [19]. Defects may initiate oxidation and act as the nucleation point for further defect formation when graphene interacts with its environment [20,21]. Furthermore, the chemical reactivity of single layer graphene, being one atomic layer thick, is highly dependent on the substrate on which it is transferred [22]. Using single layer graphene as a protective layer in chemically reactive environments may be unfavorable due to these drawbacks. Multi-layer graphene has been suggested as an alternative, which would still maintain most of the properties of single layer graphene, while being, in practice, less reactive and more robust due to its larger thickness [23].
The robustness of protective layers is of critical importance in EUV optical applications. Optical components are required to last the full lifetime (years) of beamline or imaging optics [18]. Therefore, it is important to reach a detailed understanding of the physical, chemical and electronic stability of multi-layer graphene and its reaction pathways in an EUV environment [24,25].
Previous research has shown that single-layer graphene in an EUV environment starts to oxidize when dosed with less than a monolayer of water (0.75 ML) and exposed in a partial pressure of 10 −5 mbar of H 2 [25] . This oxidation was in the form of an enol form of 1,3-diketone which was characterized in-situ using Reflection Absorption Infrared Spectroscopy (RAIRS). It was concluded that diketone formation is due to the aromatic structure breakdown of graphene. The formation of the diketone was shown to saturate for prolonged exposure to water, EUV and H 2 under identical exposure conditions.
In this work, we further expand on 1,3-diketone formation and study its formation in multi-layer graphene under the same EUV exposure conditions using in-situ RAIRS characterization. We show that as-grown multi-layer graphene is resistant to oxidation when dosed with H 2 O and exposed to EUV and H 2 . However, transferred MLG is shown to oxidize. Oxidation is driven by the presence of water that is trapped at the interface between the graphene layer and the underlying substrate.

Experimental
The substrate for graphene growth consisted of 300 nm SiO 2 , grown on a Si(1 0 0) wafer by dry oxidation at 1100°C. Next, 70 nm of Mo was sputter deposited. Subsequently, multi-layer graphene was grown by chemical vapor deposition (CVD) at 1000°C using a feedstock of Ar/ H 2 /CH 4 at 8 mbar in a cold-wall reaction chamber (VG Scienta at the NanoLab facilities at the University of Twente). The resultant graphene layer was then transferred onto an amorphous Si surface on a 10 mm × 10 mm substrate, using H 2 O 2 as an etchant. The etchant dissolves the molybdenum film onto which the graphene is grown, resulting in a graphene sheet floating on top of the solution, which can be scooped with the substrate onto which the graphene has to be transferred [26]. This transfer process was performed without the use of any support polymer (e.g. PMMA) as is commonly used for the transfer of single layer graphene, enabling residue-free graphene transfer.
The substrates onto which MLG was transferred have a top layer of 22 nm of amorphous Si (aSi) on top of 9 nm of Mo deposited on a Si wafer (see Fig. 1). Both layers of Mo and aSi were sputter deposited, with their thicknesses calibrated by X-ray reflectivity measurements. Since the amorphous silicon is not protected from the environment after deposition, the top 1-2 nm of the surface is oxidized. The Mo layer was sandwiched between the aSi layer and the Si wafer to provide an infrared reflective sample, which is required to conduct in-situ RAIRS.
For EUV exposure experiments, the MLG-on-aSi sample was mounted on a 3-axis positioning manipulator in Ultra-High Vacuum conditions and maintained at a base pressure of 10 −10 mbar. A K-type thermocouple is fixed to the sample to measure the surface temperature, and a Eurotherm temperature controller is used to control the surface temperature by radiative heating. Surface chemistry experiments, such as water dosing and Ar + ion exposure, were conducted in the same experimental setup without breaking vacuum.
The chamber is equipped with RAIRS and Temperature Programmed Desorption (TPD) spectroscopy [27]. RAIRS spectra are measured at grazing incidence using a Fourier Transform Infrared (FTIR) spectrometer (Bruker Vertex 70 V), which is equipped with a liquid N 2 cooled detector. Each IR spectrum is summed over 256 scans with a resolution of 4 cm −1 . Background and sample scans were recorded at the lowest possible stable surface temperature (~85 K). The sample is placed~1 mm away from the aperture of a Hiden Analytical 3F/PIC quadruple mass spectrometer to obtain TPD spectra. Deionized water is dosed on the surface using a retractable quartz tube connected via a pinhole. Surface coverage is calibrated against reference TPD spectra obtained from a clean Ru(0001) surface [28].
The chamber is also attached to a Xe plasma discharge Extreme Ultraviolet source (Philips EUV Alpha Source 2) with a repetition rate of 500 Hz. EUV light is filtered from the source using a Mo/Si reflective mirror (55% reflectivity at 13.5 nm) and transmission through a Si/Mo/ Zr membrane (35% transmission at 13.5 nm) [25,[29][30][31]. This results in a total transmission of 19.25% at 13.5 nm with a FWHM of 0.2 nm. The beam has a Gaussian profile (FWHM = 3 mm) with an intensity of 35-55 mW/cm 2 , corresponding to a dose of 90-110 µJ/cm 2 per pulse.
The experimental procedure is, unless otherwise stated, as follows: a background IR spectrum at 85 K is recorded, the surface is dosed with 0.75 ML of water and exposed to EUV with different partial pressures of hydrogen (either 1 × 10 −5 mbar or 1 × 10 −4 mbar, unless otherwise stated). Further IR spectra are recorded at 10 min intervals until the end of the 60 min of EUV + H 2 exposure. Finally, TPD is performed by heating the surface to 450 K at a rate of 1 K/s. X-ray Photoelectron Spectroscopy (XPS) measurements are performed ex-situ using monochromatic Al-Kα radiation, employing a ThermoFisher Theta Probe with a beam spot size of about 1 mm in diameter. Scanning Electron Microscopy (also ex-situ) measurements were performed with a high resolution SEM (HR-SEM) using a voltage of 2 kV, with a point resolution of 1.2 nm.

As grown multi-layer graphene on Mo
Previous results, mentioned in the introduction and reproduced in Fig. 2, line (a), show the RAIRS data of a transferred single layer graphene when exposed to EUV, H 2 (10 −5 mbar) and H 2 O (0.75 ML) for 60 min. Apart from the water peaks (O-H stretch, O-H bend and libration), we see a peak at~2480 cm −1 , corresponding to the formation of the enol form of 1,3-diketone. This indicates oxidation of the graphene layer and degradation of its aromatic structure [25].
For this work, an as-grown~6 nm thick multi-layer graphene sample was first dosed with 0.75 ML of H 2 O and then exposed to EUV at H 2 partial pressures ranging from 1 × 10 −5 to 1 × 10 −3 mbar. Prior to exposure, a defect density (I(D)/I(G)) of 0.11 was measured using Raman spectroscopy, which corresponds to an inter-defect distance of > 24 nm [32]). As shown in Fig. 2, line (b), with a partial pressure of up to 10 −3 mbar of H 2 , there is no increase in the spectral region associated with keto-enol formation, even after 60 min of exposure. This indicates that as-grown multi-layer graphene is more resistant to oxidation compared to transferred single-layer graphene.
The absence of the expected formation of a keto-enol peak on asgrown multi-layer graphene might be explained by a lower defect density for the multi-layer graphene sample. The previously investigated single layer graphene samples (line (a) in Fig. 2) were reported to have an I(D)/I(G) ratio on the order of 0.7 and an inter-defect distance of 20 nm. To test this hypothesis, a graphene layer with a higher initial defect density was exposed under identical exposure conditions. In this experiment, an as-grown multilayer graphene stack, 6 nm thick, with a higher defect ratio of I(D)/I(G) = 0.31 was used. Again, RAIRS shows that there is no formation of a keto-enol peak when the water dosed surface (0.75 ML) is exposed to EUV at a relatively high partial pressure of H 2 at 10 −3 mbar (Fig. 2, line (c)). The RAIRS spectra indicate, therefore, that no oxidation is taking place in the MLG structure. This conclusion is supported by XPS measurements, which show no increase in the C-O group contribution to the C1s, nor any significant change in the O1s signal. Furthermore, no changes in C1s sp 2 to sp 3 ratio were observed. In contrast, for transferred single layer graphene, changes to both the sp 2 to sp 3 ratio and the C-O group contribution to C1s were observed under these exposure conditions [25].
To confirm that defect density plays little role in oxidation, the defect ratio was modified during exposure using Ar + sputtering. The graphene surface was exposed to Ar ions with an energy of 2.8 keV for a total of 35 min (813 µC/cm 2 ) which is enough to remove a monolayer of graphene, according to the Bohdansky-Yamamura [33] model. Similar to earlier results, there is no evidence of keto-enol formation in RAIRS for as-grown MLG dosed with water and exposed to EUV and H 2 (Fig. 2, line (d)). Furthermore, while ex-situ experiments, such as Raman spectroscopy, showed an increase in the I(D)/I(G) ratio from 0.31 to 0.71 after sputtering and exposure, this did not correspond to noticeable changes in the overall sp 2 to sp 3 ratio after exposure or change in the C-O peak, which would reflect an oxidized MLG. RAIRS spectra of the O-H stretch peak (3800-3200 cm −1 ), originating from water adsorbed on the sample, show variations in intensity between different samples and exposures. The reason for this is twofold: (i) the residual water pressure during exposure depends on the exposure history of the vacuum system and (ii) depending on initial surface condition of the graphene, the orientation of the adsorbed water may change, so that a different fraction of the O-H stretches is out-of-plane (the in-plane mode is dipole forbidden in reflection mode). However, it should be noted that prior to each EUV exposure, the sample was reproducibly dosed with 0.75 ML of water. Since water in direct contact with the surface is expected to have the highest potential contribution to oxidation of the graphene, possible variations in adsorption of water from residual gas are considered less important compared to initially dosed water.

Transferred multi-layer graphene onto amorphous Si
Exposures were also performed for multi-layer graphene which was transferred (I(D)/I(G) = 0.31 after transfer) to an amorphous Si layer. As described earlier, the surface was dosed with 0.75 ML of water, and exposed to EUV with different partial pressures of hydrogen. As shown in Fig. 3, a keto-enol peak [34] (2300-2700 cm −1 ) forms when the water dosed surface is exposed to EUV at a partial pressure of 1 × 10 −4 mbar of H 2 . Similar to the case for SLG, the keto-enol peak (2484 cm −1 ) is visible after 10 min of exposure to H 2 O, EUV and H 2 .
The increasing peak intensity for longer exposure times suggests that 1,3-diketone continues to form during exposure, likely at the edges of graphene grain boundaries, which are relatively more hydrophilic and reactive than mid-grain points on the graphene layer [35]. Later in this section we discuss differences between transferred and non-transferred graphene samples that likely explain the difference in reactivity.
The formation of this peak is seen in more detail in Fig. 4, which displays RAIRS spectra after sequential exposures of the same sample (on the same spot) at different H 2 partial pressures. Curve I shows that the keto-enol peak becomes visible when the H 2 partial pressure is 1 × 10 −5 mbar. Firstly, the peak continues to grow over 60 min with peak intensity at 2467 cm −1 . Secondly, after increasing the partial pressure of H 2 to 10 −4 mbar (curve II), the keto-enol peak is observed to grow faster, and the peak intensity shifts to 2490 cm −1 . (Note that a new background spectrum was taken before each exposure step. The RAIRS signal thus only corresponds to molecular groups formed during each exposure run.) Thirdly, in curve III, where the exposure is repeated at 10 −5 mbar H 2 , we see that the growth rate of the keto-enol peak is slower compared to the previous exposure at 10 −5 mbar H 2 (curve I). Finally, as the experiment conditions from curve II are repeated again, curve IV shows that the rise in peak intensity is lower compared to curve II.
A comparison of the growth of the keto-enol peak intensity for these different experiments is displayed in Fig. 5, where the peak intensity has been plotted against EUV exposure time. From Fig. 5 it is observed that the growth of the keto-enol peak for curve I (EUV + H 2 at 10 −5 mbar) starts at 0.002%/min, while the rate of keto-enol growth increases to 0.014%/min when the hydrogen partial pressure is increased to 10 −4 mbar (curve II). Finally, repeating/continuing the exposure experiments results in curve III and IV, with growth rates of 0.001%/ min and 0.006%/min respectively, indicating that the growth rate of keto-enol formation is indeed decreasing with exposure time.
XPS data for a photoelectron take off angle of 34°(with respect to the surface normal) has been compared for transferred unexposed graphene and graphene after EUV and H 2 exposure in Fig. 6. Table 1 quantifies the spectral components seen in the XPS data. The C1s sp 2 and sp 3 peak percentages before and after exposure fall within margins of error of each other, indicating that the graphene layer has remained unchanged after EUV exposure up to the information depth of XPS, which is about 5 nm for carbon [36]. Also, a Si 2p percentage around 1 at.% is detected before and after exposure. The silicon signal may be due to the Si substrate, which is on the edge of detection by the XPS, because the graphene layer is almost as thick (~5-8 nm) as the information depth of the XPS. Furthermore, due to manufacturing defects, the graphene layer has holes that account for 2-3% of the total surface area. The presence of holes results in a photoelectron signal from the Si substrate. This is further discussed in the section below.
In our previous work on single layer graphene exposed to H 2 O, EUV and H 2 , changes in the sp 2 and sp 3 peaks were easily observable by eye [25]. In the case of multilayer graphene, possible chemical changes in the topmost monolayers are expected to be less visible due to the background signal from the graphene layers below. We also confirmed that no detectable changes in the C1s peak were observed for a more grazing angle of 72°with respect to the sample normal. If diketone formation occurred at the upper-most surface, the absolute increase in sp 3 peak area of the C1s peak should be similar to our previous EUV exposure experiments on single-layer graphene, where 12 at.% of sp 3 C Fig. 3. Keto-enol formation (2300-2700 cm −1 ) seen in transferred multi-layer graphene dosed with water (0.75 ML) and exposed to EUV with a partial pressure of 10 -4 mbar of H 2 . The other peaks shown correspond to different O-H modes of adsorbed water: stretch (3200-3800 cm −1 ), bend (1400-1600 cm −1 ) and libration (600-100 cm −1 ) [28]. Fig. 4. Formation of keto-enol peak in a transferred MLG stack on aSi seen around~2400-2500 cm −1 , with EUV and H 2 O at different partial pressures of H 2 , after 60 min of exposure. Also, O-H stretches of adsorbed water (3200-3400 cm −1 ) are seen. The experiments are done sequentially from I to IV, with a new background taken before every experiment.
was observed [25]. Analysis of the peak fitting indicates that the same increase in sp 3 peak area as observed for damaged SLG would be detectable even if deeper graphene layers are not damaged. Moreover, it should be noted that no increase in the concentration of oxygen (based on the O 1s peak) was observed, which is an additional indication that no oxidation of the graphene is detectable within the information depth of XPS. Based on the absence of oxidized graphene in the XPS signal, we conclude that the keto-enol oxidation observed in RAIRS is most likely originating from the interface between the graphene and the Si, which cannot be characterized by XPS.
SEM images for the transferred multi-layer graphene before and after exposure are shown in Fig. 7. As seen in the pictures, multiple holes are visible in the graphene layer. These holes are mostly around~100 nm in diameter, with largest holes being~300 nm in diameter. The smaller hole sizes (< 50 nm) are quantified in Fig. 8. Furthermore, the holes are through the depth of the graphene layer, exposing the Si substrate to the environment above the graphene layer, which was confirmed after looking at a cross section of a graphene layer in the SEM, as seen in Fig. 9.
Comparing a set of 10 images (field-of-view of 3 × 2.5 µm 2 ) taken before and after exposure to EUV and H 2 , we find that the number of holes less than 50 nm in diameter increases after exposure, whereas larger holes (> 50 nm) remain largely unaffected by EUV exposure. This has been quantified in Fig. 8, where we see that the average hole density for MLG before exposure is~2.41 holes/µm 2 (σ = 0.25), which increases to~3.90 holes/µm 2 (σ = 0.39) after exposure. Furthermore, we see a slight increase in hole percentage to total graphene area (as pictured in Fig. 7) from 2.21% before exposure to 2.68% after exposure. This increase in holes is most likely caused by oxidation of MLG and the breakdown of the graphene layer [37].
The cross-section of transferred multi-layer graphene in Fig. 9 shows that the graphene film retains the rough grain structure of the Mo film it was grown on. As a result, the film rests on the aSi surface with an irregular waviness of approximately 32 nm peak-valley. This waviness could result in water being physisorbed between the MLG and aSi surface during the transfer process, with further water being introduced (via holes) during dosing and exposures. As a result, large amounts of water can remain trapped between the graphene layer and the substrate, which may dissociate into reactive species such as OHand H + during EUV exposure [37]. This could support the fact that EUV photons react with water molecules, resulting in a trapped energy high enough to break C = C bonds and form small holes in the graphene layer.
Finally, the difference in EUV-induced secondary electron yield between graphene and the aSi substrate can contribute to the different reactivity of as-grown and transferred multilayer graphene. Carbon has a reasonably low secondary electron yield δ at 92 eV (13.5 nm) (δ = 0.37) as compared to the underlying substrate such as Si (δ = 0.74) or Mo (δ = 0.87) [38,39]. This means that water trapped between graphene and the a-Si substrate has a higher probability to be dissociated, resulting in production of O radical species, compared to water adsorbed onto the graphene surface.

Summary and conclusions
Previously it has been shown that single layer graphene sheets, as-grown or transferred, are likely to oxidize and form defects when introduced to an EUV environment in the presence of water and hydrogen [24,25,37]. This oxidation could be observed with RAIRS through the formation of keto-enol groups. In this work, we show that as grown multi-layer graphene is resistant to oxidation under the same exposure conditions up to 10 -3 mbar of partial pressure of H 2 . On the contrary, transferred MLG does show evidence of oxidation in RAIRS similar to SLG. Since the employed transfer process is polymer free, we can exclude that keto-enol formation occurs due to polymer residues such as PMMA. Two factors are identified as likely cause for the difference between as-grown and transferred MLG: firstly, MLG conforms to the grain boundaries of the Mo growth substrate, so EUV induced reactions with water can only take place on the topmost layers of the graphene. Secondly, the low EUV-induced secondary electron yield of carbon (graphene) reduces the generation of reactive O species from water adsorbed on MLG. The MLG thickness is larger than the secondary electron mean free path [36], thus, secondary electrons generated in the Mo substrate cannot reach adsorbed water on the graphene surface. This in in contrast to the case for single-layer graphene on a substrate with a relatively high SE yield, where the thickness of the graphene is smaller than the electron mean free path.
XPS data shows that graphene oxidation does not occur at the top surface of the graphene layer, leading to the conclusion that oxidation must occur at the interface between the graphene and the aSi. SEM pictures indicate that the transferred graphene is not conformal with the surface, which could result in trapped water during transfer and/or exposure, potentially leading to oxidation of the graphene when irradiated by EUV photons. Water trapped at the graphene-substrate interface is exposed to the relatively higher flux of secondary electrons from the substrate and is, thereby, more likely to initiate oxidation. SEM pictures of exposed transferred MLG show formation of additional small holes (< 50 nm in diameter), similar to previous reports on EUV exposed SLG [37].

Declaration of Competing Interest
We declare that we have no conflict of interest.

Acknowledgements
This research is supported by the Dutch Technology Foundation STW (project number 140930), which is part of the Netherlands   Comparison of the density of holes less than 50 nm in diameter seen in transferred MLG (in field-of-view = 3 × 2.5 µm 2 ) in a set of 10 images before and after exposure to EUV and H 2 . Fig. 9. Cross-section of exposed transferred MLG on amorphous Si, displaying the morphology of graphene on the transferred surface. The distance between the graphene layer and aSi surface is also indicated.