1 Introduction

Solid lubricants are widely used in many applications involving harsh environments such as aerospace, automotive, and nuclear energy [1, 2].] They are critical to the improvement of energy efficiency of moving mechanical systems and the reduction of carbon footprint [3, 4]. Graphite and molybdenum disulfide (MoS2) are among the most commonly used solid lubricants [5, 6]. They are usually applied as thin films by physical vapor deposition (PVD) and/or chemical vapor deposition (CVD) [7]. Such “bulk” thin films typically provide friction coefficients of around 0.1 in pure sliding owing to their easy shear properties. Being thin films of finite thickness, wear life is limited, and replenishment is a challenge [8].

In recent years, atomically thin 2D materials, especially graphene, graphene oxide (GO), and 2D MoS2, have been extensively studied to overcome certain limitations of their bulk counterparts because those nanosheets can be easily applied and replenished [9,10,11,12]. However, the main drawbacks of atomically thin 2D materials as solid lubricants include easy aggregation, limited stability, and inconsistent performance at the macroscale [13, 14].

Several previous studies have found that during the friction process of graphene or other carbon-based solid lubricants (amorphous carbon and diamond-like carbon), graphene nanoscrolls can be formed. Nanoscrolls are thin layers of material that roll-up upon themselves to form a scroll-like structure. Such nanoscrolls are critical to reducing contact area and coefficient of friction [15, 16]. For instance, Berman et al. showed that superlubricity, defined as coefficient of friction (COF) less than 0.01, can be realized at engineering scale in a nitrogen environment when graphene was used in combination with nanodiamond particles and diamond-like carbon [15]. The graphene sheets wrapped around nanodiamonds to form nanoscrolls at the sliding interface. Graphene nanoscrolls (NS) consisting of graphene shell and amorphous carbon core have also been observed in tribofilm formed by amorphous carbon coating [16], which is believed to be the main reason for achieving low friction. In another report, macroscale superlubricity at ambient conditions was achieved when graphene-coating microspheres were used as solid lubricants between the graphene-coated plate and ball [17].

Despite the importance and high potential of graphene nanoscrolls in achieving macroscopic superlubricity, it is very challenging to generate them in a controlled or scalable way. In the above-mentioned studies, graphene nanoscrolls are formed among the wear products during the friction process, which is not efficient or well controlled. There are a few recent reports on bottom-up approach for the synthesis of graphene nanoscrolls [18,19,20,21], but such nanoscrolls have quite different structures and sizes, and they have yet to be investigated for solid lubrication applications.

Moreover, it is important and beneficial to integrate inorganic nanoparticles (NPs) within the graphene to further enhance their solid lubrication performance [22, 23]. The metal or metal oxide nanoparticles provide robust film formation ability or act as nano-scale ball bearings in addition to the easy shearing of graphene nanosheets, which usually leads to enhanced lubrication especially under harsh conditions [24,25,26]. The inorganic nanoparticles that have been combined with graphene for lubrication applications include Cu, Ag, Al2O3, Fe3O4, Mn3O4, etc. [11, 27]. It is relatively simple to incorporate 2D graphene nanosheets with inorganic NPs by direct mixing or bottom-up synthesis. But it is more challenging to incorporate graphene nanoscrolls with inorganic NPs due to the 1D morphology and size mismatch. To the best of our knowledge, such hybrid graphene nanoscrolls from bottom-up synthesis have never been explored for lubrication applications.

To address those issues and fill the knowledge gap, here we report a scalable method for the synthesis of hybrid graphene-titanium oxide (G–TiO2) nanoscrolls and study their potential as solid lubricants. Our method for creating the hybrid nanoscrolls is based on in situ sol–gel synthesis of TiO2 on graphene followed by spray-freeze-drying-induced shape transformation. Such G–TiO2 nanoscrolls show macroscopic coefficient of friction of around 0.1 at ambient conditions, which can be further improved to as low as 0.02 when graphene oxide is used as the primer layer. The high solid lubrication performance can be related to the unique sharkskin-like morphology formed by the alignment and assembly of those hybrid nanoscrolls during friction.

2 Materials and Methods

2.1 Materials

Graphite powder, potassium permanganate, sodium bicarbonate, tris–HCl buffer, ammonium hydroxide solution (28–30%), titanium (IV) n-butoxide, and ethanol were all purchased from Sigma-Aldrich and used as received. Sulfuric acid and hydrogen peroxide were purchased from VWR Chemicals and used as received. Hydrazine monohydrate (98%) was purchased from Fisher Scientific. Silicon wafers with a diameter of 2 inch were purchased from UniversityWafer.

2.2 Preparation of G–TiO 2 nanoscrolls

Graphene oxide (GO) nanosheets were prepared by a modified Hummer’s method as described in our previous report [28]. For the in situ sol–gel synthesis of TiO2 on GO nanosheets, the GO water suspension (0.1 mg/mL, 200 mL) was first solvent exchanged with ethanol to get GO ethanol suspension. Then, 0.4 mL of ammonium hydroxide was added to the solution and stirred for 30 min at 25 °C. Titanium (IV) n-butoxide (3 mL or 1.5 mL) was then added dropwise to the solution and kept for stirring at 25 °C for 24 h. After the sol–gel reaction, the product was washed by centrifugation and replaced the supernatant with DI water, and the washing was repeated three times.

The GO-TiO2 nanosheets prepared from the previous step were dispersed in DI water and added to a glass flask. A mixture of 40-µl hydrazine monohydrate and 280-µl ammonium hydroxide was slowly added to the flask while stirring. The flask was then put in an oil bath heated to 95 °C and kept for 1 h. During this step, the reduction of GO to graphene as well as the amorphous to crystalline transition of TiO2 nanostructures took place, and the product is named G-TiO2 nanosheets.

The G-TiO2 nanosheets water suspension was then directly sprayed into a container filled with liquid nitrogen using a spray gun. The frozen sample was then freeze-dried at -58 °C to obtain the final product, which is G–TiO2 NS.

2.3 Tribology Characterization

A controlled amount of G–TiO2 NS was dispersed in ethanol to prepare a uniform suspension (0.5 mg/mL). The suspension (20 mL) was then drop casted on a 2-inch silicon wafer and after evaporation of ethanol, a uniform coating layer was formed on the wafer. Other control samples were prepared in the same way by drop casting on silicon wafer from their corresponding ethanol suspensions. For coatings with graphene or GO primer layer, graphene or GO ethanol suspension was used to drop cast and obtain the primer layer, followed by drop casting of the G–TiO2 NS.

The coefficient of friction (COF) was measured using CSEM High-Temperature Pin-on-disk tribometer. Stainless Steel (SS) balls (440 C) with a diameter of 6 mm were used. The substrates are silicon wafers with different coatings mentioned above. The normal load was varied in the range of 2 N to 5 N with a maximum Hertz contact pressures of 0.74 and 1 GPa, respectively. The linear speed of the ball was kept at 5 cm/s for a total sliding distance ranging from 50 to 100 m. The data collection frequency was 7 Hz. All the measurements were conducted in ambient conditions with temperature around 25 °C and humidity around 50%.

2.4 Other Characterization

Scanning electron microscopy (SEM) was conducted using JEOL-7401 FE-SEM at an accelerating voltage of 10 kV. Transmission electron microscopy (TEM) images were obtained using Tecnai G2 F20 at an acceleration voltage of 110 kV. Raman spectra were collected with a Renishaw inVia confocal Raman microscope with an excitation laser of wavelength 514 nm. Optical profilometer (Zygo NewView 7300) was used for studying the dimension and morphology of the wear tracks and balls. Olympus BX51 optical microscope was used for capturing optical images. X-ray photoelectron spectroscopy (XPS) was conducted with a PHI VersaProbe III surface analysis instrument (Physical Electronics) at a 45° take-off angle. Surveys were conducted at a pass energy of 117 eV and high-resolution spectra were obtained at a pass energy of 11.7 eV.

3 Results and Discussion

3.1 Synthesis and Characterization of the G–TiO 2 Nanoscrolls

The fabrication process of the G–TiO2 hybrid nanoscrolls is shown in Fig. 1. Briefly, in situ sol–gel synthesis of TiO2 was conducted on the GO surface in ethanol using titanium butoxide as the precursor. During this step, the hydrolysis and subsequent polycondensation reaction occur to form amorphous TiO2 network. Subsequently, the hybrid nanosheets were chemically reduced by a hydrazine/ammonia mixture. During this step, GO was reduced to reduced graphene oxide, which is referred to as graphene here for simplicity, although the structure is different from physically exfoliated graphene. Meanwhile, the hydrazine acted as a catalyst to further promote the dehydroxylation and polycondensation reaction of amorphous TiO2 on graphene surface [29]. In addition, the heating (95 °C) during the hydrazine reduction step promotes the transition of amorphous TiO2 to crystalline structures, as confirmed by spectroscopy data in the following discussion.

Fig. 1
figure 1

Schematic of the in situ sol–gel synthesis and shape transformation to generate the G-TiO2 NS

Full size image

The as-prepared G–TiO2 hybrid nanosheets were transformed into nanoscrolls through a scalable spray-freeze-drying method. This method has been used to prepare simple graphene nanoscrolls [18, 19]. In our case, the prepared G–TiO2 hybrid nanosheets were dispersed in DI water, and the dispersion was sprayed directly into liquid nitrogen to achieve very fast freezing. Then, the frozen sample was freeze-dried, and during the sublimation process of ice, the 2D nanosheets transformed into nanoscrolls. Such shape transformation process was induced by the competition between the elastic bending energy and free energy. When the scrolling occurs, there is a decrease in the surface free energy of the hydrophobic graphene nanosheets originating from the van der Waals interaction of the overlap** domains; at the same time, there is an increase in the elastic energy caused by rolling and distorting of the graphene. When the former overweighs the latter, spontaneous formation and stabilization of G–TiO2 NS takes place, as is the case for our system.

The morphology of the fabricated G–TiO2 NS as well as graphene NS (as a control sample) was studied with SEM and TEM. A dense network of the formed graphene NS can be seen in SEM image (Fig. 2a), the average length is on the order of one to several μm, and the average diameter is on the order of 10 nm. TEM image of the graphene NS (Fig. 2b) further confirms the tubular shape of the NS with multilayered graphene shell. When TiO2 is integrated by in situ sol–gel synthesis, the formed G–TiO2 NS maintains the nanoscroll shape (Fig. 2c), while the average diameter is increased to about 50 nm, and the hybrid NS have higher degree of curling compared with pristine graphene NS. High-resolution TEM image (Fig. 2d) further shows that TiO2 nanostructures exist on both the surface and interior of the G-TiO2 NS. In addition, we studied another control sample: the physical mixture of graphene and TiO2 NPs by SEM (fig S1), which does not show such rolled-up morphology, and the two components are randomly distributed or aggregated in the mixture.

Fig. 2
figure 2

Morphological study of graphene and G-TiO2 nanoscrolls. a SEM of graphene nanoscrolls, b TEM of one representative graphene nanoscroll, c SEM of G-TiO2 nanoscrolls, and d TEM of one representative G-TiO2 nanoscroll

Full size image

Raman spectroscopy was used to characterize the structure of the G–TiO2 NS as well as its precursors and control samples (Fig. 3a). Graphene shows its characteristic G band at 1596 cm−1, D band at 1348 cm−1, 2D bands at 2677 cm−1, as well as the D + G band at 2865 cm−1 [30]. As a control sample, the TiO2 NPs prepared by hydrothermal method show Raman peaks at 145, 394, 515, and 637 cm−1, which indicates the TiO2 NPs have anatase crystalline structure. Raman spectrum of the physical mixture of graphene and TiO2 NPs (Figure S2) shows the characteristic graphene G, D, and 2D bands, as well as the anatase TiO2 peaks at 145, 394, and 630 cm−1.

Fig. 3
figure 3

Spectroscopy characterization of the G-TiO2 NS and control samples. a Raman spectra of the G–TiO2 NS and control samples: graphene and TiO2 NPs. b FTIR spectrum of the G–TiO2 NS. c XPS spectra of graphene and G-TiO2 NS. d High-resolution XPS scan of the C1s and Ti2p peaks for graphene and/or G–TiO2 NS

Full size image

Raman spectrum of the G–TiO2 NS shows the D and G bands of graphene, but interestingly, the peaks for TiO2 component are located at 150, 257, 421, and 610 cm−1, which indicate the TiO2 nanostructures in the G–TiO2 NS have primarily rutile crystalline phase [31], which are different from the control TiO2 NPs prepared by hydrothermal method. The crystalline structure of TiO2 in the G–TiO2 NS will also undergo interesting transformation during the friction process, as will be discussed later.

FTIR was used to characterize the chemical structure of the G–TiO2 NS (Fig. 3b). The peak at 742 cm−1 corresponds to Ti–O–Ti bond [32], and the peak at 863 cm−1 corresponds to Ti–O–C bond [33]. Such results indicate that there are chemical interactions and bond formation between graphene and TiO2 in the G–TiO2 NS. The peaks at around 1600 cm−1 and 3500 cm−1 are ascribed to bending vibration and stretching vibration of H–O–H and O–H, which comes from the absorption of moisture on the surface of G–TiO2 NS.

X-ray photoelectron spectroscopy (XPS) provides further information on the structure and composition of the samples (Fig. 3c). The survey XPS scan shows that graphene has a pronounced C1s peak as well as an O1s peak. The O1s peak mainly originates from the underlying SiO2 substrate as well as a small amount of residual C–O groups on graphene. The G–TiO2 NS, on the other hand, has not only C1s and O1s peaks but also a substantial Ti2p peak due to the in situ-synthesized TiO2 nanostructures.

High-resolution XPS scans of the C1s peaks and peak fittings are shown in Fig. 3d. The C1s peak of graphene can be deconvoluted into a strong sp2 carbon peak at 284.1 eV, sp3 carbon peak at 285.0 eV, as well as a small C–O peak at 287.1 eV. For G–TiO2 NS, the C1s peak not only has contribution from the above-mentioned three components, but also an additional peak at 289.3 eV, which can be ascribed to the Ti–O–C=O structure [8e-f). There are high density of partially overlap** “scales” in the wear track. The coverage, density, and thickness of those scales are substantially higher than that of the G-TiO2 NS wear track without GO primer layer. We conducted TEM study of the wear particles from G-TiO2 NS coating after tribology study (Fig. 8g-h). It can be seen that the G-TiO2 NS somewhat maintain the scroll-like morphology with TiO2 nanostructures dispersed in rolled-up graphene, although the nanoscrolls are less well defined and have more interconnected structure compared with those before tribology testing.

We also characterized the wear tracks with Raman spectroscopy (Figure S7). For graphene coating after tribology testing, there are only very weak peaks corresponding to the G and D band of graphene that can be observed. For TiO2 NPs coating after testing, there is no visible peak except for the underlying Si substrate, which indicates complete removal of the coating. For the physical mixture of graphene and TiO2 NPs, only the G and D bands of graphene at 1596 and 1348 cm−1 were observed, which shows that TiO2 NPs were mostly lost due to the weak physical interaction between graphene and TiO2 NPs. On the other hand, for the G-TiO2 NS wear tracks, besides the G, D, and 2D bands for graphene, interestingly, peaks for anatase TiO2 at 394, 515, and 637 cm−1 were also observed. This result shows that there is a rutile to anatase transformation of the TiO2 nanostructures in G-TiO2 NS during mechanical friction, which has not been observed before.

The possible mechanism for the formation of such hierarchical sharkskin-like morphology is shown in Fig. 8i. The as-deposited G-TiO2 NS has a random orientation and distribution in the coating layer. During the tribology test, the application of normal force and the lateral friction force can effectively induce the rolling and alignment of the G-TiO2 NS. Many of the NS also undergoes deformation or become partially collapsed during the friction process. Multilayered structures with each layer consisting of closely packed NS can be formed in the wear track, which mimics the structure of sharkskin. Such sharkskin-like morphology decreases the contact area and reduces the shear stress during the friction process [39, 40], which contributes to the substantially enhanced solid lubrication performance of the G-TiO2 NS.

4 Conclusion

An efficient and scalable method for the controlled synthesis of hybrid graphene nanoscrolls was developed. This method integrates in situ sol–gel metal oxide synthesis and spray-freeze-drying-induced shape transformation of graphene. The generated G–TiO2 NS have well-defined size, shape, and tunable composition. The COF of such G–TiO2 NS measured under ambient conditions is around 0.10, which is substantially lower than its individual components: graphene or TiO2 NPs, also much lower than the physical mixture of graphene and TiO2 NPs. Moreover, with the application of GO as the primer layer for such G–TiO2 NS coating, the COF can be further reduced to 0.02. Friction and wear resistance of both the silicon substrate and the steel ball is significantly enhanced (by one to two orders of magnitude) with such G–TiO2 NS coating. Morphological study of the wear track shows that sharkskin-like hierarchical structures are generated during friction, which is ascribed to the assembly and layering of the G-TiO2 NS. The incommensurate contact area and rolling of G–TiO2 NS during this process are believed to be the major reasons for the achieved outstanding solid lubrication performance. The scalable generation of such hybrid nanoscrolls as high-performance solid lubricants will have impact on aerospace, automotive, and precision manufacturing.