Parameter Space of Atomic Layer Deposition of Ultrathin Oxides on Graphene

Atomic layer deposition (ALD) of ultrathin aluminum oxide (AlOx) films was systematically studied on supported chemical vapor deposition (CVD) graphene. We show that by extending the precursor residence time, using either a multiple-pulse sequence or a soaking period, ultrathin continuous AlOx films can be achieved directly on graphene using standard H2O and trimethylaluminum (TMA) precursors even at a high deposition temperature of 200 °C, without the use of surfactants or other additional graphene surface modifications. To obtain conformal nucleation, a precursor residence time of >2s is needed, which is not prohibitively long but sufficient to account for the slow adsorption kinetics of the graphene surface. In contrast, a shorter residence time results in heterogeneous nucleation that is preferential to defect/selective sites on the graphene. These findings demonstrate that careful control of the ALD parameter space is imperative in governing the nucleation behavior of AlOx on CVD graphene. We consider our results to have model system character for rational two-dimensional (2D)/non-2D material process integration, relevant also to the interfacing and device integration of the many other emerging 2D materials.

here H 2 O vapor or O 3 , and B denotes the metal precursor, here TMA. The oxidant/precursor dose is calculated from the product of the delivery pressure (P dos ) and the residence time (t dos ), which in CM and PM are both governed by a single parameter ALD pulse time (t pul ). All samples are loaded while the chamber is at the preset deposition temperature (T dep ) and the process chamber is purged with N 2 for more than 10min (t purin ) before the ALD process is started. The purge time between oxidant/precursor pulses (t pur ) is varied depending on T dep . In PM, the samples are exposed to series of oxidant pulses prior to the ALD process, where the pretreatment time (t pretreat ) is determined by the total number of the pulses. In MM, each oxidant/precursor is delivered twice in a quick succession with a very short time interval (t intv ). In SM, the flow in the S-3 process chamber is stopped for several seconds (t hold ) to allow the samples to be soaked in the oxidant/precursor. Table S1. Details of common parameters for all ALD processes, i.e. continuous-flow mode  Figure S1.

Parameters Values
Deposition temperature ( using O 3 /TMA. Schematic of the process is shown in Figure S1a and S1b.

Parameters Values
Deposition  Figure S1c.

Parameters Values
Deposition temperature (T dep ) Figure S1d.

Parameters Values
Deposition

SI2. Aluminum oxide surface coverage calculation
The surface coverage of ALD AlO x (θ) is calculated based on the contrast observed in SEM images as shown in Figure S2, with bright regions indicate the AlO x covered graphene surface and dark regions indicate the absence of AlO x , i.e. bare graphene surface. Each SEM image (8bit grayscale) is first filtered with a set of Top-Hat and Bottom-Hat transformation using 64x64 morphological structuring element to enhance contrast and eliminate the uneven background illumination. 12 The filtered image is subsequently transformed into a binary image (black and S-6 white) by Otsu thresholding criterion. 3 The bright region area is calculated by estimating the total number of "white" pixels in the entire binary image using 2x2 area calculation rule. 4 The AlO x surface coverage (θ) is then calculated by normalizing the bright region area with the total area of the entire binary images, i.e. the total sum of both "white" and "black" pixels.    (Fig S3), AlO x surface coverage (θ) on graphene is found to be increasing with the increase of t pur . At t pur = 10s, AlO x predominantly nucleates as clusters at the ridges of G/Cu surface features. Such a highly heterogeneous nucleation is reflected by the extremely low θ of just ~21% (Fig S3a). While the nucleation is still largely heterogeneous, the nucleation density, particularly in the graphene troughs, improves significantly with the increase of t pur as reflected by the increase of θ to ~42% as t pur increases to 25s ( Fig S3b). Indeed, a nearly homogeneous nucleation with θ >98% can be observed when t pur = 60s (Fig S3c). Nevertheless, a significant decrease in AlO x nucleation density can be observed when t pur is significantly prolonged as reflected by the decrease of θ to ~61% as t pur increases to 300s ( Fig   S3d). Such a decrease in θ is expected as desorption, or even thermal decomposition, of H 2 O/TMA also takes place during the ALD process. Due to the relatively low desorption rate of H 2 O/TMA at T dep of 80°C, AlOx still nucleates on the graphene surface, albeit heterogeneously, even when t pur is set to be extremely long. In this study, t pur is carefully selected for each T dep such that it is sufficiently long to prevent the formation of prematurely hydrolyzed TMA species but not too long to result in undesired H 2 O/TMA desorption.

SI4. Aluminum oxide thickness measurement by AFM
The thickness of the ALD AlO x film is estimated by AFM measurement on HOPG surfaces.
HOPG surfaces were chosen as the representative samples due to their similarity to graphene samples in terms of wettability and chemical inertness. As HOPG surfaces are known to be much less wettable by H 2 O than G/Cu, 10,11 the measured values could serve as lower bound values of S-10 the actual thickness of the AlO x film on graphene. In addition, HOPG surfaces could be easily patterned without leaving behind significant residues. In contrast, graphene patterning often results in measurable residues that may skew the AFM measurement. As measured by AFM, 12 ALD cycles result in AlO x film thickness of 1.1(±0.2) nm and a RMS surface roughness of 0.52(±0.01) nm (Fig S4a). For 40 and 60 ALD cycles the AlO x film thickness is measured at 3.7(±0.3) nm and 5.9(±0.3) nm respectively with a RMS surface roughness of 0.54(±0.06) nm and 1.24(±0.47) nm respectively (Fig S4b and S4c). This is equivalent to a film growth rate of 0.088 nm/cycle (Fig S4d).
S-11 respectively. Figure S5b shows the leakage currents in these capacitors, where the leakage currents are lower than 1nA at 0.7 V and 2.2 V for AlO x film deposited in 20 and 50 ALD cycles respectively. These values match well with those of AlO x formation on graphene found in the literature. 12 Hence we are confident that the formation of AlO x in the present study is continuous and does show potential to act as efficient high-k dielectric in graphene electronics with EOT <1.3 nm. 13 Obtaining such an ultra-thin dielectric on graphene is very important for high frequency operation of FETs.   (Fig S7d). These ALD processes were selected because they yield conformal AlO x nucleation (θ>98%) with just 12 ALD cycles on all graphene samples (see the main article). The refractive index of the AlO x film (n) was calculated by fitting the obtained ellipsometry data (Ψ and Δ) to the available Al 2 O 3 thin film model for each wavelength at an incident angle of 60°. 6 As can be that the refractive index (n) of CM 80°C (Fig S7d) is consistently the lowest amongst all samples across the entire spectrum. The refractive index (n) of CM 200°C (Fig S7a) is very similar to that for SM 200°C (Fig S7b), and both are found to be consistently the highest amongst the samples across the entire spectrum. Note that a difference in refractive index (n) by 0.02 indicates a difference in AlO x density by ~0.12 g/cm 3 . 5 Thus, this finding suggests that the density of AlO x films deposited at a T dep of 200°C is higher, albeit only slightly, than that at a S-15 T dep of 80°C. However, it is important to note that this finding is obtained from ~10nm thick AlO x films on SiO 2 rather than directly from sub-2nm AlO x films on graphene. contributions, and (c) scatterplot of 2D peak linewidth (Г 2D ) against G peak linewidth (Г G ). All AlO x depositions are performed in 12 ALD cycles total. The doses for both H 2 O and TMA are maintained at ~0.14 Torr·s while that for O3 is set at ~28.65 Torr·s.

SI8. Effect of prolonged ozone pretreatment
The effect of prolonged ozone exposure during the ALD to the graphene quality is assessed by Raman spectroscopy analysis using photon excitation of 532nm (Fig S8).  (Fig S8a). Such an increase in I D /I G indicates that defects are being introduced to the graphene structure during O 3 pretreatment and a longer O 3 pretreatment time (t pretreat ) results in a higher defect density. Thus, the use of O 3 in the ALD process is not completely harmless for the graphene even at a low T dep of 80°C. To avoid introducing excessive defects to the graphene, it is therefore necessary to limit the t pretreat to just 2min as it is sufficient to introduce relatively homogenous AlOx with θ of ~96% (see the main article Fig 3d and 3f).
The peak frequency of the 2D band (ω 2D ) and G band (ω G ) are found to be shifted toward higher values with the increase of O 3 pretreatment time, where ω 2D ~2681cm -1 and ω G ~1591cm -1 are observed for PM-2m-O 3 and ω 2D ~2688cm -1 and ω G ~1595cm -1 are observed for PM-15m-O 3 S-17 ( Fig S8b). The shift in ω 2D and ω G modes toward higher wavenumbers indicates a significant increase in the graphene doping level from ~3x10 12 cm -2 to ~5x10 12 cm -2 when t pretreat is set to 2min and to ~6x10 12 cm -2 when t pretreat is set to 15min. While the mechanical strain level remains similar in magnitude between -0.1 --0.2% when t pretreat is set to 2min, it increases to -0.2 --0.3% when t pretreat is set to 15min. 8,9 In addition, the line width of the 2D band (Г 2D ) is found to be Torr, t dos : ~0.7s) for 12 cycles total.

S-18
As a control experiment, AlO x nucleation on bare Cu foils ( Fig S9a) and SiO 2 wafers ( Fig   S9b) Consequently, one should adjust the typically used ALD parameters if a homogeneous nucleation on graphene is to be achieved.

SI10. Effect of extended TMA residence time
The effect of prolonged TMA residence time during the ALD to the graphene quality is assessed by Raman spectroscopy analysis using photon excitation of 532nm (Fig S10) Fig. 5d).
The peak intensity ratio between D band and G band (I D /I G ) is found to be increasing with the increase of TMA residence time, where I D /I G of ~0.07 is observed for MM-10s. In contrast, SM-3.5s yields I D /I G of just ~0.04, which is similar to the as-transferred G/SiO 2 (Fig S10a). The peak frequency of the G band (ω G ) are found to be shifted toward higher values with the increase of TMA residence time, where ω G increases slightly from ~1584cm -1 for SM-3.5s to ~1585cm -1 for MM-10s ( Fig S10b). The line width of the 2D band (Г 2D ) and G band (Г G ) is found to be relatively constant with the increase of TMA residence time (Fig S10c). An increase in I D /I G indicates the formation of defects on graphene due to exposure to the highly reactive TMA when the residence time (t dos ) is set to be significantly long. In addition, a shift in ω G indicates a slight increase in the graphene doping level from <10 12 cm -2 to ~10 12 cm -2 while the mechanical strain level remains similar in magnitude between -0.05 --0.15%. 8,9 While the use of TMA with t dos ~3.5s is harmless for the graphene, the exposure to TMA for ~10s results in a more defective and doped graphene. Thus, it is necessary to find an optimal value of TMA residence time at which it is sufficiently long to obtain a conformal AlO x nucleation without inducing excessive damage or doping to the graphene. In this study, t dos of 2-3.5s is found to be optimal. It is important to note that the value of optimal t dos may be different from one ALD system to another.