1. Structural investigation via optical microscopy and SEM

Single-layer graphene samples can be detected via optical microscopy when placed on silicon wafers covered with ∼300 nm thick SiO₂.^{Reference Novoselov, Geim, Morozov, Jiang, Zhang, Dubonos, Grigorieva and Firsov3} Consequently, optical microscopy has been used as the first step in the present study to detect and confirm the presence of CVD-grown graphene on SiO₂ substrates post-transfer. While optical microscopy readily confirms large-scale coverage (millimeters and above) of SiO₂ substrates by transferred graphene films, the spatial resolution is not sufficient to detect the presence of all structural defects such as tears and folds that frequently occur during transfer. As such, the next step in structural characterization of graphene films has involved imaging via SEM. Figure 3(a) shows the general appearance of CVD-grown graphene transferred onto SiO₂ substrates via a large-scale SEM image that independently confirms the successful transfer of extended graphene films onto SiO₂ substrates. Moreover, SEM images with higher resolution are utilized to investigate the presence of structural defects and contaminants in the transferred graphene films. For instance, Fig. 3(b) shows the morphology associated with a tear and additionally demonstrates the presence of contaminants in the form of small white dots on graphene samples that are occasionally encountered. While determining the exact chemical nature and origin of these contaminants is beyond the scope of the present study, energy dispersive x-ray spectroscopy indicates that oxygen is involved, supporting the idea that the white dots comprise copper oxide that originates from the foils used for CVD growth.^{Reference Celebi28}

FIG. 3.

(a) Large-scale SEM image confirming the transfer of extended graphene films onto SiO₂. (b) Higher-resolution SEM image showing a tear in the transferred film as well as the occasional presence of contaminants in the form of white dots.

2. Confirmation of single-layer character via Raman spectroscopy

Raman spectroscopy is frequently used to characterize the structure and chemistry of single- and multi-layer graphene.^{Reference Ferrari, Meyer, Scardaci, Casiraghi, Lazzeri, Mauri, Piscanec, Jiang, Novoselov, Roth and Geim23} Specifically, the number of layers of graphene samples can be precisely determined by detecting the relative intensities of the characteristic 2D and G peaks (I _2D/G), such that I _2D/G values of ≥2, ∼1, and <1, are expected for single-layer, bi-layer, and multi-layer graphene samples, respectively. Figure 4 presents a typical Raman spectrum obtained on a CVD-grown graphene sample transferred onto SiO₂. While the I _2D/G value of 2.80 is clearly indicative of the single-layer character, it should be mentioned that there is considerable variability in measured I _2D/G values between different graphene samples, and even for spectra obtained on the same graphene sample at different locations such that I _2D/G values between 2.50 and up to 5.50 have been recorded for the graphene films investigated in the present study. The observation of I _2D/G values significantly higher than 3.00 can be tentatively attributed to the presence of partially-suspended regions of transferred graphene films on SiO₂.^{Reference Berciaud, Ryu, Brus and Heinz29}

FIG. 4.

Representative Raman spectrum of a CVD-grown graphene sample transferred onto SiO₂.

3. Structural investigation via AFM

As already mentioned, AFM is particularly useful for graphene research in detecting the topographical height of graphene samples on SiO₂ and in observing structural defects and chemical contaminants with high resolution. In particular, thanks to the high spatial resolution of the method, AFM can be readily used to detect wrinkles on CVD-grown graphene samples that arise during the cool-down step of the CVD process due to the substantial difference in the thermal expansion coefficients of graphene and copper, leading to a build-up of in-plane stress in graphene and consequently, wrinkling.^{Reference Li, Cai, An, Kim, Nah, Yang, Piner, Velamakanni, Jung, Tutuc, Banerjee, Colombo and Ruoff15}

A representative topographical AFM image of a region of the SiO₂ substrate partially covered by graphene is provided in Fig. 5(a). It is observed that the graphene film features a number of structural defects including folds, and a substantial number of wrinkles. Moreover, the presence of contaminants (potentially, PMMA residues) on the graphene film is detected, which leads to occasional streaks in the AFM image due to the probe laterally manipulating the residues. Clearly, the high spatial resolution provided by the AFM method allows the investigation of graphene morphology and cleanliness in much higher detail when compared to optical microscopy and SEM. Figures 5(b) and 5(c) show topographical height profiles over the graphene–SiO₂ boundary and over a representative wrinkle, respectively. The height of the transferred graphene with respect to the SiO₂ substrate is measured as ∼1.45 nm, in line with certain values reported in the literature.^{Reference Chung, Shen, Cao, Jauregui, Wu, Yu, Newell and Chen30} Moreover, the height of a representative wrinkle with respect to the graphene surface is observed as ∼2.80 nm. A statistical analysis of wrinkle heights conducted via AFM on transferred graphene samples provides a mean value of 1.99 nm with a standard deviation (SD) of 0.87 nm. Finally, it should be indicated that no substantial change in the morphology and improvement in the cleanliness of transferred graphene samples is detected via AFM after both variants of post-transfer cleaning, confirming reports in the literature that currently used methods aiming at removing PMMA residues are not entirely successful.^{Reference Skakalova and Kaiser22}

FIG. 5.

(a) Topographical AFM image of CVD-grown graphene on SiO₂, demonstrating in high resolution the presence of defects such as a fold and wrinkles on the graphene surface, as well as contaminants. (b) Representative height profile of the boundary between graphene and SiO₂. (c) Representative height profile of a wrinkle on graphene. The height profiles in (b) and (c) have been recorded along the white lines designated with the letters “b” and “c” in (a), respectively.

1. Dependence of friction forces measured on CVD-grown graphene, wrinkles, and the SiO₂ substrate on normal load

Friction forces (F _f) recorded on CVD-grown graphene, increased friction forces encountered at wrinkles, as well as friction forces recorded on the SiO₂ substrate have been investigated as a function of normal load (F _n) in the present study. A representative friction force map and corresponding results of F _f versus F _n measurements acquired with a single AFM probe are provided in Fig. 6. The SiO₂ substrate exhibits the highest absolute F _f values, followed by wrinkles and CVD-grown graphene. The lubrication properties of CVD-grown graphene on the SiO₂ substrate become apparent, as a significant reduction in absolute F _f values. The increase in F _f values measured on wrinkles is attributed to the associated abrupt change in topography, in accordance with the so-called ratchet effect.^{Reference Sundararajan and Bhushan31} Please note that this is in contrast to previous observations of reduced friction on wrinkles of CVD-grown graphene patches on Cu foils which were attributed to multilayer character arising from a “fold-over” geometry at the wrinkle edges.^{Reference Egberts, Han, Liu, Johnson and Carpick12}

FIG. 6.

(a) Representative friction force map of CVD-grown graphene on SiO₂. (b) The dependence of friction force on normal load for graphene, wrinkles, and the SiO₂ substrate, together with linear fits. The error bars for measurements on graphene and SiO₂ have been determined by considering the standard deviation in friction forces measured on multiple regions of a given friction force map, while the error bars for the measurements on wrinkles have been determined by considering friction forces measured on 10 individual wrinkles. Data acquired with a single probe.

It is observed that the data in Fig. 6(b) can be fit reasonably well in a linear fashion. To eliminate the effect of potentially different degrees of adhesion exhibited by the SiO₂ substrate when compared to CVD-grown graphene on measured F _f values, tribological properties may be quantified by the slopes of linear fits applied to the data, to determine friction coefficients (μ). While the data presented in Fig. 6(b) already demonstrates that not only absolute F _f values, but also μ values are significantly lower on CVD-grown graphene when compared to SiO₂ (μ_Graphene: 0.006 ± 0.001, μ_Wrinkle: 0.018 ± 0.002, and ${\mu _{{\rm{Si}}{{\rm{O}}_2}}}$: 0.072 ± 0.003); a related, comprehensive analysis of lubrication properties of CVD-grown graphene is presented in Sec. III. B. 5.

2. Effect of using different probes on friction

It is well known that the specific nanoscale structure and chemistry of AFM probes affect measured friction force values significantly.^{Reference Szlufarska, Chandross and Carpick32} As such, several different probes have been used to perform the friction force measurements presented here. The results of F _f versus F _n experiments conducted on CVD-grown graphene with three different probes (probes 1–3) are presented in Fig. 7. It is observed that a significant degree of variability is present in measured F _f values from one probe to another, with Probe 2 generally exhibiting the highest friction. To eliminate the effect of using different probes on a rational evaluation of the lubrication properties of CVD-grown graphene, ratios of friction coefficients measured for CVD-grown graphene and SiO₂ substrates $\left( {{{\mu _{{\rm{Graphene}}} } \mathord{\left/ {\vphantom {{\mu _{{\rm{Graphene}}} } {\mu _{{\rm{SiO}}_2 } }}} \right. \kern-\nulldelimiterspace} {\mu _{{\rm{SiO}}_2 } }}} \right)$ have been determined for each probe and consequently averaged to obtain figures of merit for lubrication in Sec. III. B. 5.

FIG. 7.

The dependence of friction force on normal load on CVD-grown graphene for three different probes (Probe 1: blue, Probe 2: red, and Probe 3: gray).

It should be noted that certain degradations associated with tip apexes may occur during the measurements, sometimes resulting in appreciable enlargement of the apex radius and/or contamination by, e.g., post-transfer chemical residues on the sample surface. For instance, the significant increase in F _f observed for Probe 3 in Fig. 7 after 18 nN may be attributed to such a structural/chemical change. To evaluate whether significant degradation of probe apexes has occurred during the AFM measurements, SEM images of selected probe apexes have been acquired before and after each experiment. It has been observed that the apex radii have increased from 50 to 70 nm, from 25 to 45 nm, and from 55 to 65 nm for probes 1–3, respectively. A comparison describing a particularly severe degradation involving Probe 3 is depicted in Fig. 8. Such degradations of sliders are expected to also occur during potential practical implementations of CVD-grown graphene as a solid lubricant. The approach involving the calculation of ${{\mu _{{\rm{Graphene}}} } \mathord{\left/ {\vphantom {{\mu _{{\rm{Graphene}}} } {\mu _{{\rm{SiO}}_2 } }}} \right. \kern-\nulldelimiterspace} {\mu _{{\rm{SiO}}_2 } }}$ ratios discussed above allows a meaningful evaluation of lubrication performance under such conditions, as well.

FIG. 8.

SEM images of a specific probe apex (Probe 3) (a) before and (b) after AFM measurements, demonstrating significant degradation.

3. Effect of different post-transfer cleaning procedures on friction

Two post-transfer cleaning procedures have been optionally applied on CVD-grown graphene samples transferred onto SiO₂ substrates: (i) ultrasonic baths in acetone and isopropanol and (ii) Ar-gas annealing. Subsequently, the effect of post-transfer cleaning on the tribological behavior of CVD-grown graphene has been investigated (Fig. 9). It is observed that the Ar-gas-cleaned sample exhibits the highest absolute friction force values, followed by the ultrasonic-cleaned sample, with the as-transferred graphene sample exhibiting the lowest friction forces. It should be indicated that, from this comparison alone, it is not possible to determine the origin of the increase in absolute friction values as the cleaning procedure itself, and not structural/chemical changes in the probe apex that may occur during the experiments. To be able to make rational conclusions regarding this aspect, the previously mentioned approach involving the calculation of ${{\mu _{{\rm{Graphene}}} } \mathord{\left/ {\vphantom {{\mu _{{\rm{Graphene}}} } {\mu _{{\rm{SiO}}_2 } }}} \right. \kern-\nulldelimiterspace} {\mu _{{\rm{SiO}}_2 } }}$ ratios for multiple probes needs to be used.

FIG. 9.

The dependence of friction force on normal load on CVD-grown graphene for as-transferred (blue), ultrasonic-cleaned (red), and Ar-gas-cleaned (gray) graphene samples. Data acquired with a single probe.

4. Comparison with friction force measurements of mechanically-exfoliated graphene

While too small for most practical applications, the nanotribological properties of mechanically-exfoliated graphene have been nevertheless the subject of certain fundamental studies in the literature.^{Reference Lee, Li, Kalb, Liu, Berger, Carpick and Hone8} To determine whether there are significant differences in frictional behavior between single-layer, mechanically-exfoliated and single-layer, CVD-grown graphene, friction force maps have been recorded on both types of samples and comparative measurements of F _f versus F _n have been conducted (Fig. 10). The absence of wrinkles on the mechanically-exfoliated samples is noticeable. While absolute friction force values look very similar for the two graphene samples until a normal load of 18 nN (at which point the specific probe used in the measurements might have undergone a structural/chemical change, as discussed earlier), a thorough comparison of lubrication performance involving ${{\mu _{{\rm{Graphene}}} } \mathord{\left/ {\vphantom {{\mu _{{\rm{Graphene}}} } {\mu _{{\rm{SiO}}_2 } }}} \right. \kern-\nulldelimiterspace} {\mu _{{\rm{SiO}}_2 } }}$ ratios and multiple probes is provided in Sec. III. B. 5.

FIG. 10.

Representative friction force maps for (a) CVD-grown and (b) mechanically-exfoliated graphene samples. (c) The dependence of friction force on normal load for CVD-grown (blue) and mechanically-exfoliated (red) graphene. Data acquired with a single probe.

5. Lubrication performance of CVD-grown graphene on SiO₂ substrates

Rather than comparing absolute friction force values, it is useful to employ friction coefficients (extracted from linear fits to F _f versus F _n data) in evaluating graphene's tribological properties, to eliminate the influence of differences in adhesion behavior between graphene and SiO₂ substrates. Additionally, to eliminate the effect of (i) using AFM probes with different apex structure/chemistry, and (ii) potential changes in probe apex structure/chemistry during the experiments, it makes sense to determine ratios of friction coefficients between graphene and the substrate $\left( {{{\mu _{{\rm{Graphene}}} } \mathord{\left/ {\vphantom {{\mu _{{\rm{Graphene}}} } {\mu _{{\rm{SiO}}_2 } }}} \right. \kern-\nulldelimiterspace} {\mu _{{\rm{SiO}}_2 } }}} \right)$ for each measurement, and obtain average values from results delivered by the different tips. Mean $\left {{{\mu _{{\rm{Graphene}}} } \mathord{\left/ {\vphantom {{\mu _{{\rm{Graphene}}} } {\mu _{{\rm{SiO}}_2 } }}} \right. \kern-\nulldelimiterspace} {\mu _{{\rm{SiO}}_2 } }}} \right$ values obtained in this fashion represent true figures of merit associated with the lubricative properties of graphene samples.

Based on the discussion above, ${{\mu _{{\rm{Graphene}}} } \mathord{\left/ {\vphantom {{\mu _{{\rm{Graphene}}} } {\mu _{{\rm{SiO}}_2 } }}} \right. \kern-\nulldelimiterspace} {\mu _{{\rm{SiO}}_2 } }}$ ratios for CVD-grown graphene in three variants (as-transferred, ultrasonic-cleaned, and Ar-gas-cleaned) as well as mechanically-exfoliated graphene are provided in Table II, for (up to) five different probes, together with mean and standard deviation values. By studying the data, it is seen that graphene in all examined forms acts as a very good solid lubricant, reducing friction coefficients observed on SiO₂ substrates by ∼90%. Among CVD-grown graphene samples, the as-transferred variety exhibits the highest mean lubrication performance (thus, the lowest ${{\mu _{{\rm{Graphene}}} } \mathord{\left/ {\vphantom {{\mu _{{\rm{Graphene}}} } {\mu _{{\rm{SiO}}_2 } }}} \right. \kern-\nulldelimiterspace} {\mu _{{\rm{SiO}}_2 } }}$ ratio) and the mean lubrication performance drops slightly with ultrasonic cleaning, while Ar-gas-cleaned samples exhibit the lowest mean lubrication performance. It should be mentioned that a consideration of standard deviation values together with the mean values reported for as-transferred and ultrasonic-cleaned samples indicates that there is no statistically-significant difference in overall lubrication performance between the two varieties of graphene. On the other hand, Ar-gas-cleaned samples clearly exhibit a comparatively lower lubrication performance. A potential physical mechanism responsible for this effect could involve the thermal breakdown of PMMA residue trapped under the transferred graphene samples during annealing, leading to a wide distribution of trapped contaminants increasing average surface roughness and friction. Finally, it should be noted that while Raman spectroscopy has been previously used to successfully correlate the degree of fluorination of graphene samples to changes in frictional behavior,^{Reference Li, Liu, Kim, Shenoy, Sheehan, Robinson and Carpick33} no direct correlation between the “quality” of graphene samples (as inferred from I _2D/G values) and lubrication performance has been observed in our studies.

TABLE II.

Friction coefficient ratios associated with graphene and SiO₂$\left( {{{\mu _{{\rm{Graphene}}} } \mathord{\left/ {\vphantom {{\mu _{{\rm{Graphene}}} } {\mu _{{\rm{SiO}}_2 } }}} \right. \kern-\nulldelimiterspace} {\mu _{{\rm{SiO}}_2 } }}} \right)$.

Our results also allow a comparison of the lubrication performance of CVD-grown and mechanically-exfoliated graphene samples. While it is seen that as-transferred CVD-grown graphene samples on average exhibit slightly better lubrication performance than exfoliated ones, the existence of wrinkles on CVD-grown graphene and the associated increase in friction should be taken into account (please note that the data for CVD-grown graphene presented in Table II have been acquired on regions free from wrinkles). In that respect, the extent to which wrinkles increase friction values on CVD-grown graphene (as-transferred, ultrasonic-cleaned, and Ar-gas-cleaned) is examined in Table III. As expected, wrinkles lead to a significant (about 3-fold) enhancement of local friction values on CVD-grown graphene samples, representing a considerable detriment to overall lubrication performance.

TABLE III.

The ratios of friction forces measured on wrinkles to those on surrounding graphene $\left( {{{F_{{\rm{f,wrinkle}}} } \mathord{\left/ {\vphantom {{F_{{\rm{f,wrinkle}}} } {F_{{\rm{f,graphene}}} }}} \right. \kern-\nulldelimiterspace} {F_{{\rm{f,graphene}}} }}} \right)$. Data averaged over 10 wrinkles and a normal load range of 0–22 nN for each measurement.