Resist and Transfer Free Patterned CVD Graphene Growth on ALD Molybdenum Carbide Nano Layers
Eldad Grady, Chenhui Li, Oded Raz, W.M.M. Kessels, Ageeth A. Bol
RResist and Transfer Free Patterned CVDGraphene Growth on ALD
M oC x Nano Layers
Eldad Grady , Chenhui Li , Oded Raz , W.M.M. Kessels , andAgeeth A. Bol Department of Applied Physics, Eindhoven University of Technology, Den Dolech 2,P.O. Box 513, 5600 MB Eindhoven, The Netherlands Institute for Photonic Integration Eindhoven University of Technology Den Dolech2, 5612 AZ Eindhoven, the NetherlandsCorresponding author: [email protected]
Abstract
Multilayer graphene (MLG) films were grown by chemical vapour de-position (CVD) on molybdenum carbide (
MoC x ) substrates. We fabri-cated the catalytic MoC x films by plasma enhanced atomic layer deposi-tion (PEALD). The mechanism of graphene growth is studied and anal-ysed for amorphous and crystalline MoC x films. In addition, the uniqueadvantages of catalytic substrate PEALD are demonstrated in two ap-proaches to graphene device fabrication. First, we present a completebottom up, resist-free patterned graphene growth (GG) on pre-patterned MoC x PEALD performed at 50 ◦ C . Selective CVD GG eliminates theneed to pattern or transfer the graphene film to retain its pristine, asgrown, qualities. Furthermore, we fabricated MLG directly on PEALD MoC x on 100 nm suspended SiN membrane. We characterise the MLGqualities using Raman spectroscopy, and analyse the samples by opticalmicroscopy, scanning electron microscopy and X-ray diffraction measure-ments. The techniques of graphene device manufacturing demonstratedhere pave the path for large scale production of graphene applications. Next generation graphene-based applications would require a resist and trans-fer free graphene growth in order to realise the promise of graphene for flexibleelectronics[1, 2], and sub 10nm nodes’ interconnects [3]. Graphene based in-terconnects would significantly reduce the RC delay and thermal budget whichis currently the bottle neck in sub 10nm transistor nodes [4, 5] due to the ex-cellent room temperature carrier mobility and thermal conductivity. However,graphene based devices exhibit far lower carrier mobility than the theoreticalpredictions promise. Encapsulation of graphene with lattice matching dielectric1 a r X i v : . [ phy s i c s . a pp - ph ] D ec reprint Preprint Preprintmaterial such as hBN shows remarkable improvement, but performance is stillsubpar [6]. The main challenges of current commercial chemical vapour depo-sition (CVD) grown graphene are resist residue due to target transfer process[7, 8], and patterning of graphene, which degrade the graphene quality, resultingin the far lower carrier mobility than theoretical predictions. Known techniquesinclude photolithography, ion beam milling [9], shadow masking [10], area se-lective passivation layer [11] and stencil mask [12]. Despite these tremendousefforts, the goal of a scalable, low cost, high quality patterned graphene usingcommercially available tools, has not been achieved. Photolithography methodsleave resist residue, and ion beam etching causes broad lateral damage close topattern area due to scattering ions [13, 14]. Patterning graphene with oxygenplasma inevitably oxides graphene edges [15], and the stencil mask poses reso-lution limitation . Furthermore, transfer of graphene from one substrate ontoanother degrades the graphene quality due to induced film compression and ten-sion, and potential contamination between the graphene and target interfaces.A wholesome solution that avoids both the need to pattern the graphene orto transfer a pre patterned graphene onto a target, has yet to be introduced.While advancements have been made toward transfer free graphene grown onpatterned Mo catalytic layer [16], resist residue on the underlying growth sub-strate cannot be fully removed. As defects, non uniformity and contaminationwill imminently translate to graphene defects, it is vital to achieve a resist freecatalytic surface, in a uniform and conformal layer deposition. Atomic layerdeposition (ALD) is a cyclic, broad temperature window soft deposition, withprecise thickness control due to its self-limiting nature. It allows for unmatch-ing conformal deposition on high-aspect ratio (HAR) objects. Recently, wedemonstrated the plasma enhanced ALD (PEALD) of M oC x , with excellentcomposition control [17]. Consequently, we presented the growth of multilayergraphene (MLG) films on these substrates, and the correlation between the cat-alytic substrate physical and chemical properties to the grown MLG [18]. Inthis work, we describe the growth mechanism of graphene on M oC x , and com-pare between amorphous and highly crystalline catalytic substrates for graphenegrowth. Finally, we demonstrate a proof of concept for the merits of ALD basedgraphene growth for future interconnects (IC) in the form of low temperaturepatterned PEALD of M oC x , for selective CVD growth of MLG, so that noresist residue is present between the graphene and the catalytic material inter-face. Additionally, we show the advantage of the ALD soft deposition on 100nmthick suspended SiN membrane. Thereafter, MLG was grown on the suspendedSiN membranes to achieve suspended graphene based heterostructure withoutexposing the suspended MLG to wet chemicals or corrosive acids. M oC x thin films have been deposited by plasma enhanced atomic layer depo-sition (PEALD) at various temperatures and plasma conditions, as describedelsewhere [17]. 2reprint Preprint PreprintPEALD was performed on 100 mm Si (100) wafers coated with 450 nm ofthermally grown SiO . The depositions were performed in an Oxford instru-ments FlexAL2 ALD reactor, which is equipped with an inductively coupledremote RF plasma (ICP) source (13.56 MHz) with alumina dielectric tube. M oC x thin films have been deposited by PEALD at various temperatures andplasma conditions, with M oC x films varying from 15 µm to 30 µm in thickness.MLG was grown by low-pressure CVD (LPCVD) in a quartz tube (d=50mm,l=60cm) furnace with 3 heat zones set to 1050 ◦ C. The typical base pressurewhen evacuated is 10 − mbar. The furnace is set on cart wheels, to allow sam-ples to be rapid annealed, as furnace temperature stabilises within 3.5 minutesafter tube insertion. When moved away from the furnace, sample cooling downduration is typically 15 minutes. Carbon feedstock gas ( CH ) is fed along withArgon through a quartz inner tube of 5 mm in diameter to the sealed side ofthe outer tube. M oC x films have been saturated with carbon by annealing attemperatures between 500 ◦ C to 800 ◦ C with 100 sccm CH gas flow at 4 mbarpressure. Then, graphene films have been grown under similar conditions at1100 ◦ C for 10 minutes. The samples were then promptly extracted from thefurnace and allowed to cool down at ambient room temperature under Ar gasflow in the quartz tube. As shown in figure 3a photoresist (PR) ma-N 400 with4.1 µm thickness on 90 nm SiO on Si 2” wafers were used for low temperaturePEALD of M oC x film. After deposition, M oC x was patterned by lift-off pro-cess, and rinsed in isopropyl alcohol (IPA). 100 nm and 50 nm SiN membraneswere supplied by Philips Innovation Services (PInS) foundry. M oC x were alsodeposited at 300 ◦ C on 100 nm SiN membranes suspended on Si (5x5 mm sus-pended rectangular area). MLG were then grown the M oC x films as illustratedin figure 5. While deposition was successful for both SiN sample thicknesses,due to the brittle nature of the thinner membranes we present here resultsmeasured on the thicker 100nm based membranes. Raman spectroscopy wasperformed with Reinshaw InVia 514 nm laser. Film crystallinity and preferredcrystal orientation was studied by Gonio x-ray diffraction. Experiments wereconducted with PanAnlytical X’pert PROMRD diffractometer operated using CuKα ( λ = 1 . A ). This section is divided to three parts: first part deals with MLG growth mech-anism in
M oC x films and comparison between amorphous and crystalline cat-alytic substrate. The second part demonstrate a technique for resist and transferfree patterned graphene device. In the third part we fabricate MLG on a thinsuspended membrane using ALD and CVD. We show the limitation of Ramanmeasurements on suspended membranes low thermal conductivity.3reprint Preprint Preprint M oC x We studied the growth mechanism of graphene on
M oC x substrates, to optimisegrowth process for the various film compositions. The importance of saturat-ing the catalytic M oC x with free carbon is demonstrated and the effects offilm crystallinity on carbon precipitation to the surface during growth is shownby Raman and XRD measurements. The mechanism of graphene growth on M oC x films is explained in this section. When we subjected the M oC x filmto direct growth at 1100 ◦ C for 10 minutes. SEM images show ablation on thefilm surface, that has a characteristic graphene Raman signature [1a]. Outsidethese areas, no indication of graphene growth was measured. We added then acarburisation step in order to saturate the film with carbon and examined differ-ent temperatures around the crystalline phase change temperature ( ∼ ◦ C ).Figure 1c indicates that for crystalline M oC x film, ideal saturation takes placeabove the crystalline phase change, for MLG grown at the same growth time.The lack of sufficient carburisation affects the graphene growth significantly, ascan be seen in figure 1c. This effect is less dominant with amorphous, low massdensity M oC x films, which can be carburised at lower temperatures as well.After establishing an optimal carburisation temperature, we examined the idealgrowth time for various M oC x types. As seen in figure 2, M oC x with rich car-bon content, typically low mass density and crystallinity exhibit good qualityMLG growth after 10 minutes at 1100 ◦ C . M oC x films with higher mass densityand crystallinity display no graphene growth at this time duration, but ratherrequire a longer exposure time to CH at the growth temperature of 20 minutes.Longer exposure begin to deteriorate the graphene film. We study the physicalalteration for this crystalline film during the carburisation and graphene growthprocess, as seen in figure 2. XRD diffraction peak typical to cubic- M oC . are dominant for the deposited film. After 2 hours carburisation at 800 ◦ C , atransition Orthorhombic crystalline phase is noted, along with sharp graphite(101) plane diffraction peak. After 10 minutes graphene growth at 1100 ◦ C , theorthorhombic phase crystallinity increases, while no significant change in thegraphitic peak is noticed. Graphene film has been grown on patterned catalytic substrates. Patterning of
M oC x has been performed by a lift-off process, so that no exposed M oC x surfaceneeded to be coated with PR. When depositing M oC x at 150 ◦ C we found resistresidues on the SiO , due to hard baking of the photoresist at this temperature,as the thermal stability limits of the PR (110 ◦ C ). Moreover, after the CVDgrowth we found random patches of MLG coverage on the exposed M oC x , andno continuous MLG coverage. However, when the PEALD is performed at 50 ◦ C ,we could seamlessly remove the PR and no significant traces were found after lift-off. The CVD growth of graphene on these samples showed full MLG coveragewith excellent uniformity, albeit a relatively high D/G peak ratio. Moreover, ascan be seen in figure 4a, D and G peaks were detected throughout the exposed4reprint Preprint Preprint SiO areas, but no significant 2D peak. We have seen that a transfer free releaseof MLG on M oC x membranes is a direct result of wet etching the Mo basedcatalyst. This observation is valid for ultra thin layers with thicknesses below30 nm. Thicker layers ( > nm ) release the MLG film such that the graphenemembrane is afloat on the liquid surface. SiN membranes were suspended on Si substrate with 5x5 mm openings.
M oC x was then deposited by PEALD process. There after, MLG was grown by aCVD growth process. Thus, a suspended heterostructure of MLG/ M oC x /SiNwas fabricated on the supporting Si frame. MLG film grown on SiN membranewere characterised by Raman spectroscopy, as can be seen in 6. We measuredthe Raman signal at the edge of the suspended membrane where the SiN wassupported by the underlying Si, and at the centre of the membrane, where theMLG was on top of ∼ M oC x and 100 nm SiN. The Raman spectrum wasfitted and baseline corrected due to enhanced SiN background signal. AlthoughRaman spectroscopy is considered a non-destructive measurement, we discov-ered that suspended MLG/SiN membranes were highly sensitive to the Ramanlaser power. As figure 6c shows, when 10% power of the 20mW Laser was used,an increase in the D peak was measured, and the 2D peak was quenched in com-parison to 5% power. With 50% power of the Raman’s laser, we could punch ahole through the heterostructure suspended membrane as seen with an opticalmicroscopy image [see figure 6d ]. We have demonstrated in this work the growth mechanism and conditions ofMLG on PEALD
M oC x films. In order to achieve full graphene film coverage,and a uniform graphene growth, carburisation of the substrate is essential. Thecarburisation step allows for saturation of the catalytic film with free carbon.The free carbon then precipitate to the surface during annealing at 1100 ◦ C. Wefound that in order to saturate
M oC x film with higher mass density, carburisa-tion temperature has to exceed the crystalline phase change point. For M oC x film we found that point to be around ∼ M oC . to a polycrystalline or-thorhombic phase, which results in multiple strain points in the MLG film. Bycontrast, graphene growth on an amorphous carbon rich film is more facile andresults in a lower D/G peak ratios. After understanding the growth mechanism,5reprint Preprint Preprintwe demonstrated the advantages of graphene growth on PEALD catalytic films.The low temperature, soft and atomic precise deposition allows for pre pat-terning of catalytic substrates, such that no PR coating of the growth surfaceis needed. We demonstrated here an initial proof of concept, which showed afull film coverage with excellent uniformity. We speculate that decompositionof the CH molecules on the SiO surface with PR residue is the reason forthe carbide formation outside the patterned M oC x areas. The relatively highD/G peak could be addressed by growth process optimisation, and different ap-proaches to selective ALD which was out of the scope of this work. We cannotrule out potential surface impurities during the lift-off process, that could resultin additional amorphous carbon during the CVD growth process. A completebottom up selective area ALD process will avoid any potential resist contami-nation and should allow for a broader deposition window. We have also usedALD to deposit M oC x films on brittle SiN membranes of 50 and 100 nm. TheMLG growth demonstrated on 100 nm suspended membrane shows a viablefabrication route for sensor applications, and graphene based resonators for awide range of frequencies. We showed for the first time damage to a suspendedheterostructure caused directly by Raman measurements. The low thermal con-ductivity of the SiN is likely the cause for the local damage to the suspendedheterostructure, with no directly available heat sink - as oppose to the membraneedges. This however could prove useful, when punctured suspended grapheneheterostructures are required, for water filtration and desalination applicationsfor example. Building on the capabilities demonstrated here, the route for futureapplication based on patterned graphene or suspended graphene heterostructureis clearly marked. Further steps, such as Al O encapsulation, could be readilyperformed by ALD directly on the MLG without damage by a process of hy-drogenation and post ALD annealing [19]. Moreover, one can combine recentdevelopments in area selective ALD to realise a complete bottom up fabricationof the catalytic substrate resist free, with atomic-scale precision alignment. Acknowledgements
This research is supported by the Dutch Technology Foundation STW (projectnumber 140930), which is part of the Netherlands Organization for ScientificResearch (NWO), and partly funded by the Ministry of Economic Affairs as wellas ASML and ZEISS. E. Grady thanks Cristian Helvoirt, Janneke Zeegbregts,Jeroen van Gerwen and the lab technical staff for their support.
References [1] Veronica Strong, Sergey Dubin, Maher F El-Kady, Andrew Lech, YueWang, Bruce H Weiller, and Richard B Kaner. Patterning and electronictuning of laser scribed graphene for flexible all-carbon devices.
ACS nano ,6(2):1395–1403, 2012. 6reprint Preprint Preprint[2] Rak-Hwan Kim, Myung-Ho Bae, Dae Gon Kim, Huanyu Cheng, Bong HoonKim, Dae-Hyeong Kim, Ming Li, Jian Wu, Frank Du, Hoon-Sik Kim,et al. Stretchable, transparent graphene interconnects for arrays of mi-croscale inorganic light emitting diodes on rubber substrates.
Nano letters ,11(9):3881–3886, 2011.[3] Ning C Wang, Saurabh Sinha, Brian Cline, Chris D English, Greg Yeric,and Eric Pop. Replacing copper interconnects with graphene at a 7-nmnode. In , pages 1–3. IEEE, 2017.[4] Maria Politou, Xiangyu Wu, Inge Asselberghs, Antonino Contino, BartSoree, Iuliana Radu, Cedric Huyghebaert, Zsolt Tokei, Stefan De Gendt,and Marc Heyns. Evaluation of multilayer graphene for advanced intercon-nects.
Microelectronic Engineering , 167:1–5, 2017.[5] Yuji Awano. Graphene for vlsi: Fet and interconnect applications. In , pages 1–4. IEEE,2009.[6] Shaloo Rakheja, Vachan Kumar, and Azad Naeemi. Evaluation of thepotential performance of graphene nanoribbons as on-chip interconnects.
Proceedings of the IEEE , 101(7):1740–1765, 2013.[7] Junmo Kang, Dolly Shin, Sukang Bae, and Byung Hee Hong. Graphenetransfer: key for applications.
Nanoscale , 4(18):5527–5537, 2012.[8] Woosuk Choi, Muhammad Arslan Shehzad, Sanghoon Park, and YonghoSeo. Influence of removing pmma residues on surface of cvd graphene usinga contact-mode atomic force microscope.
RSC Advances , 7(12):6943–6949,2017.[9] Max C Lemme, David C Bell, James R Williams, Lewis A Stern, Brit-ton WH Baugher, Pablo Jarillo-Herrero, and Charles M Marcus. Etchingof graphene devices with a helium ion beam.
ACS nano , 3(9):2674–2676,2009.[10] Yong Seung Kim, Kisu Joo, Sahng-Kyoon Jerng, Jae Hong Lee, EuijoonYoon, and Seung-Hyun Chun. Direct growth of patterned graphene onsio 2 substrates without the use of catalysts or lithography.
Nanoscale ,6(17):10100–10105, 2014.[11] Mario Hofmann, Ya-Ping Hsieh, Allen L Hsu, and Jing Kong. Scalable, flex-ible and high resolution patterning of cvd graphene.
Nanoscale , 6(1):289–292, 2014.[12] Keong Yong, Ali Ashraf, Pilgyu Kang, and SungWoo Nam. Rapid stencilmask fabrication enabled one-step polymer-free graphene patterning anddirect transfer for flexible graphene devices.
Scientific reports , 6:24890,2016. 7reprint Preprint Preprint[13] L Gustavo Can¸cado, A Jorio, EH Martins Ferreira, F Stavale, Car-los Alberto Achete, Rodrigo Barbosa Capaz, Marcus Vinicius de OliveiraMoutinho, Antonio Lombardo, TS Kulmala, and Andrea Carlo Ferrari.Quantifying defects in graphene via raman spectroscopy at different exci-tation energies.
Nano letters , 11(8):3190–3196, 2011.[14] Nick FW Thissen, RHJ Vervuurt, JJL Mulders, JW Weber, WMM Kessels,and AA Bol. The effect of residual gas scattering on ga ion beam patterningof graphene.
Applied Physics Letters , 107(21):213101, 2015.[15] Isaac Childres, Luis A Jauregui, Jifa Tian, and Yong P Chen. Effect ofoxygen plasma etching on graphene studied using raman spectroscopy andelectronic transport measurements.
New Journal of Physics , 13(2):025008,2011.[16] S Vollebregt, B Alfano, F Ricciardella, AJM Giesbers, Y Grachova, HW vanZeijl, T Polichetti, and PM Sarro. A transfer-free wafer-scale cvd graphenefabrication process for mems/nems sensors. In , pages17–20. IEEE, 2016.[17] Eldad Grady, Marcel Verheijen, Tahsin Faraz, Saurabh Karwal, W. M. M.Kessels, and Ageeth A. Bol. Tailored molybdenum carbide properties andgraphitic nano layer formation by plasma and ion energy control duringplasma enhanced ald, 2019.[18] Eldad Grady, W. M. M. Kessels, and Ageeth A. Bol. Control of graphenelayer thickness grown on plasma enhanced atomic layer deposition of molyb-denum carbide, 2019.[19] Rene HJ Vervuurt, Bora Karasulu, Marcel A Verheijen, Wilhelmus (Er-win) MM Kessels, and Ageeth A Bol. Uniform atomic layer deposition ofal2o3 on graphene by reversible hydrogen plasma functionalization.
Chem-istry of Materials , 29(5):2090–2100, 2017.8reprint Preprint Preprint
Figures (a) SEM:
MoC x ablation(b) SEM: MLG grown on MoC x (c) Raman spectra (carb. temp) Figure 1: SEM images and Raman spectrum of
M oC x after CH annealing at1100 ◦ C . Top: (a) SEM image of films that were not carburised prior to growth.Bottom: (b) SEM image of MLG grown on carburised M oC x film. Smoothsurface no signs of inhomogeneity. (c) Raman spectrum after graphene growthon crystalline M oC x , as a function of carburisation temperature. Growth timeremained identical for all 3 experiments.10reprint Preprint PreprintFigure 2: Raman spectra and XRD measurements. Left: Raman measurementsof MLG grown on (a) amorphous M oC x film and (b) Single crystalline M oC x after 10 and 25 minutes growth at 1100 ◦ C . Left: (c) XRD measurements ofcrystalline M oC x for each process step from ALD, to carburisation, to graphenegrowth 11reprint Preprint Preprint (a) Fabrication flow graphics (b) Optical microscope Figure 3: Fabrication of patterned MLG. (a) Fabrication flow schematics: com-puter grphics and illustration of fabrication steps. (b) Top right: optical mi-croscopy of pre patterned
M oC x ALD at 50 ◦ C after deposition. (a) D/G peak ratios mapping (b) 2D/G peak ratios mapping Figure 4: Characterisation of patterned MLG: Raman mapping scan of40 µm x40 µm area with 1 µm step resolution of MLG grown on 20 µm thick M oC x ALD film . (a) left: D/G peak ratio shows a high D/G peak ratio, albeituniformal continuous coverage. (b) Right: 2D/G peak ratio show a uniformcontinuous MLG film. Variation in colour is inversely proportional to MLGuniformity 12reprint Preprint PreprintFigure 5: Illustration of suspended heterostructure and optical images.(a)+(b)+(c) Computer graphics of the fabrication schematics of suspendedgraphene heterostructure. (d) Optical images of SiN membranes before andafter PEALD of ∼
15 nm
M oC x film on top13reprint Preprint Preprint (a) MLG/SiN/Si substrate (b) 5% laser power(c) 10% laser power (d) Optical microscope: 50% laser power Figure 6: Raman spectrum of MLG CVD grown on ALD of
M oC xx