Strengthening behavior of carbon/metal nanocomposites

Nanocomposites reinforced with nano-scale reinforcements exhibit excellent mechanical properties with low volume fraction of the reinforcement. For instance, only an addition of 0.7 vol.% few-layer graphene (FLG) into the pure titanium shows strength of ~1.5 GPa, obviously much superior to that of the monolithic titanium. The strengthening efficiency of composites is determined by several factors such as reinforcement geometrical/spatial characteristics and interfacial features between the matrix and the reinforcement. For the metal-matrix nanocomposites (MMNCs), since the nano-scale reinforcement has significantly high specific surface area, interfacial feature is more important and has to be clearly evaluated in understanding property of MMNCs. Although many researchers suggested the theoretical work using continuum mechanics in order to estimate the mechanical properties of the metallic composites, a clear determination has yet not to be proven by systematic experimental works. Here, we provide a new model to predict strength and stiffness of MMNCs based on quantitative analysis of efficiency parameters in which interface feature is strongly emphasized. To validate the model, we select multi-walled carbon nanotube (MWCNT) and FLG for reinforcement, and titanium (Ti) and aluminum (Al) for the matrix to modify bonding strength and specific surface area in the MMNCs.

Simulations with density-function theory (DFT) calculated the energetically favorable adsorption site in graphitic structure for certain metal elements. The equivalent bonding strength and the distance between graphene and metal were determined from the simulations as well 22,23 . As depicted in Fig. 1a, the carbon (C) atoms are held together by strong covalent bonds in the basal graphitic plane. The remaining p z -orbital of carbons allows them to bond with metals out of the plane. Non-transition metals (e.g. Al) form weak secondary bonds with graphene since they lack d-sub shell with very limited affinity with carbon. Transition metals (e.g. Ti), on the other hand, have unfilled d-orbitals where d-electrons may form ionic bonds with the dangling branches of carbon atoms in graphene. Calculations reveal that the overall bonding strength between the basal plane of Ti and a single plane of graphene was calculated to about five times higher than that between Al and carbon 21 .
The carbon and metal matrix composites for both Ti and Al were observed with high-resolution transmission electron microscopy (HRTEM) images and electron energy loss spectroscopy (EELS) analysis. Atomic-scale observation of the composite in powder-processed MMNCs can provide considerable information on the interface structure. The interface between FLG and Al matrix (Fig. 1b) appears different from that between FLG and Ti matrix (Fig. 1c). The HRTEM image of the FLG/Al composite clearly showed typical lattice fringes of a single plane of graphite with an inter-layer spacing of ~0.34 nm that confirms the presence of FLG (thickness: ~5 nm). This spacing is not observed in the FLG/Ti composite. Furthermore, the interface between Al matrix and FLG is obvious whereas the interface between Ti matrix and FLG appears in a stark contrast. Nonetheless, the Raman spectra confirm the presence of FLG for both composites ( Supplementary Fig. S4). Figure 1c shows HRTEM image of the FLG/Ti composite where carbon is detected using bright and dark field images ( Supplementary Fig. S5). The interference between FLG and Ti caused FLG to possess Moire fringes. It is hypothesized that Ti-C ionic bonds, partially formed between Ti and FLG, results in a heterogeneous arrangement of lattices near the FLG surfaces. The different bonding features between these two composites are clearly demonstrated in EELS analysis. Slight variations in EELS spectra along the detection points (from (i) to (iii) in the HRTEM images) indicate the presence of partly balanced incomplete metal-carbon bonds for both composites. Typically, C peaks at 285 eV near FLG while Al peaks at 1563 eV near Al matrix. Any types of Al-C ionic bonds (Al and C peaks at 73.4 and 282.2 eV, respectively) are not detected around the interface. The thermo-mechanical condition to produce the FLG/Al composite was not sufficient to form Al carbides along the interface 20 . Thus, clean interface was formed by means of micromechanical interlocking associated with the FLG and Al matrix. Transition metal, on the other hand, is strongly electrophilic and reactive to form Ti-C ionic bonds. Thereby, Ti provides the ionic-bonded C-Ti species at a relatively notable intensity of 458 eV in FLG as well as in Ti matrix.
High-resolution XPS analysis, shown in Fig. 2, compares the bonding characteristics of MWCNTs/ Ti, and FLG/Ti composites. The contributions of Ti (2p 1/2 ) and Ti (2p 3/2 ) spin-orbital splitting photoelectrons to Ti 4+ are identified at excitation energies of 463.5 and 457 eV. Moreover, carbon bonds with the π -π * shake-up, sp 2 , and sp 3   respectively 24,25 . Supplementary Table. S1 summarizes the binding energy and peak area of each peak in the XPS spectra. Furthermore, Supplementary Table. S2 calculates the portion of Ti-C bonds in the MWCNTs/Ti and FLG/Ti composites. Each composites are normalized with the reinforcement volume (i.e., 1 vol.%). FLG/Ti composites show twice as much volume fraction of Ti-C bonds than MWCNTs/ Ti composite. The chance of forming Ti-C ionic bonds on FLG may be higher than MWCNTs due to its superior specific surface area.
Mechanical properties of nano-C reinforced Ti and Al composites, as a function of the reinforcement volume fraction (V r ), are displayed in Fig. 3a,b. The original compressive stress-strain curves of the composites are shown in Supplementary Fig. S6. The elastic modulus (E c ) and yield stress (σ c ) of each composite increase with the volume fraction of reinforcements, although the slope of the increment starts to drastically decrease when the content of FLG and MWCNT is over 1 and 7 vol %, respectively. High volumes of MWCNT and FLG added within the metal matrix tend to agglomerate into clusters during the composite processing because of their large surface areas (as shown in Supplementary Fig.  S7). These carbon aggregates may restrict consolidation of the composite powder, generating pores or carbides in the final composite, thereby deteriorating the mechanical properties. On the other hand, the increment of E c and σ c are notably varied by the type of matrix and reinforcement used; FLG and Ti matrix show better strengthening efficiency over MWCNTs and Al matrix, respectively. Since the composites contain discontinuous reinforcements, primary deformation occurs in the matrix when the composites are loaded. The strained matrix transfers load to the reinforcements by means of shear stresses that generate along the matrix/reinforcement interface. The mechanism behind load transfer is governed by three important parameters: i) interfacial bonding between the reinforcements and the matrix (the bonding factor, k); ii) the aspect ratio and the surface-to-volume ratio of the reinforcement (the geometry factor, g); and iii) the average angle between the loading direction and the reinforcement axis (the alignment factor, s). We modified the rule of mixture with these three parameters, in order to emphasize their effects on the load transfer behavior, as expressed in Eq. (1) and (2); details in derivation of the formula are present in Supplementary Information: where E m and E r are the elastic modulus of matrix and the reinforcement, V m and V r are the volume fraction of the matrix and reinforcement, σ m and σ r are the yield strength of the matrix and the reinforcement, respectively. c and ′ c are empirical constants for E c and σ c , respectively. An efficiency factor f is given by ( Table 1): In Figure 3c,d, a differing interfacial bonding of Ti-C and Al-C is well contrasted with Ti and Al matrix with plots of normalized E c and σ c as a function of the surface area per unit volume (gV r ). The surface area per unit volume, equivalent quantity of reinforcements per arbitrary region (further details on the calculation each parameters are provided in Supplementary Information), therefore, relatively small amount of FLG in comparison with MWCNTs can be balanced using gV r . Accordingly, normalized  changes of metal-C bonding, in other words bonding character is not responsible for reinforcement geometry 26 . E c and σ c are further normalized by bonding factor using the Eqs (1) and (2), which expressed , respectively, and plotted as displayed in Fig. 3e,f, respectively. According to plots, we find values of 33.8 and 14.4 for the c (for E c ) and ′ c (for σ c ), respectively. Here, we show the numerically and experimentally demonstrated the contribution of factors to strengthening behavior of MMNCs reinforced by nano-C. Based on our results, when nano-C reinforcements are separately dispersed in metals, we can predict mechanical properties as a function of specific surface area when the bonding strength between the carbon and the metal is determined. Also, a combination of tight interfacial bonding and large specific surface area in the reinforcement/matrix interface would possess an efficient load transfer and enhanced properties of the composite. We explain the importance behind the dimensions and the interfacial properties of reinforcement that determine the performance of composites. The results provide a guideline for the design of MMNCs with diverse reinforcements using the reliable model to predict mechanical properties of MMNCs.

Methods
Composite powder preparation. Both Ti and Al-based composites reinforced by nano-C (MWCNTs and FLG) were produced by powder metallurgy. FLG is pre-milled using a planetary mill (Fritsch Co. Ltd., Pulverisette 5, Germany) before mechanical milling. FLG from graphite flakes (6− 8 nm thickness and 120− 150 m 2 /g typical specific surface area) were mechanically exfoliated by planetary ball milling with isopropyl alcohol ((CH3)2CHOH, IPA). A stainless steel bowl (500 mL) was charged with graphite flakes (2 g) and stainless steel balls (~5 mm diameter, 30 g) at a ball-to-powder weight ratio of 15:1, together with 50 ml of IPA. Planetary milling was performed at a rotation speed of 200 RPM for 1 h; it was paused for 75 min after every 15-min milling to maintain ambient processing temperature without any process control agent. Afterwards, the IPA was evaporated and dried at 150 o C for 3 h. Graphite flakes are supposed to be exfoliated by shear forces on contact between powder and balls during milling. Due to the weak van der Waals-like coupling between graphite layers, the graphene sheets in graphite can slide easily with respect to one another. For FLG/Ti composites, exfoliated graphite flakes and Ti powders were mixed without a process control agent using a planetary mill with a rotation speed of 100 RPM for 3 h; it was paused for 15 min after every 15-min milling to maintain processing temperature as the room temperature. For FLG/Al composites, firstly, exfoliated graphite flakes and Al powders were mixed with a process control agent of 1 wt.% stearic acid using a planetary mill with a rotation speed of 100 RPM for 3 h; it was paused for 15 min after every 15-min milling to maintain processing temperature as the room temperature. For both FLG/Ti, Al composites, to distribute FLG in Ti (Al) powders, the mixed powder was high energy ball-milled in an attrition mill at 500 RPM for 6 h under a purified argon atmosphere at a ball-to-powder weight ratio of 15:1, respectively. On the other hand, for MWCNTs/Ti, Al composites, to distribute MWCNTs in Ti (Al) powders, the mixed powder was high energy ball-milled in an attrition mill at 500 RPM for 6 h under a purified argon atmosphere at a ball-to-powder weight ratio of 15:1, respectively.
Composite Fabrication. Hot-pressing was carried out to consolidate a variety of ball-milled composite powders by varying the volume fraction of the reinforcements. Prior to pressing, the ball-milled powder was put in a stainless steel die with a diameter of 30 mm, which was surrounded by graphite foil. The punch and plate were sprayed with boron nitride, which was used as a lubricant to minimize the effect of friction. The powder was then pressed in the mold at 140 MPa at 450 °C for 1 h (Al-based composites) and at 570 °C for 1 h (Ti-based composites). After pressing, the graphite foil was easily peeled off. As the MWCNTs and FLG were deeply embedded in the metal powders, they were not expected to significantly interrupt the consolidation of powder during hot pressing, providing a fully dense composite compact.
Mechanical property testing. The compressive properties of the specimens were evaluated using an Instron-type machine under a constant crosshead speed condition of an initial strain rate of 1 × 10 −4 s −1 at room temperature. Rectangular specimens with 2:1 ratio of height-to-width were prepared for compression tests and evaluated. Two tungsten carbide plates were used to sandwich the compression specimens, and a sprayed film of boron nitride was used as a lubricant to minimize the effect of friction. Nanoindentation tests on the specimens were performed using a commercial nanohardness tester (Nanoindenter XP, MTS) equipped with a Berkovich indenter. In each test, the indenter was driven into the sample surface (loading half-cycle) at a rate of 10 nm/sec −1 and the peak load ranges from 20 mN to 200 mN.
Characterization techniques. The microstructure of the composites was observed using a high-resolution transmission electron microscope (HRTEM, Titan TM 80-300, FEI). Thin foil specimens from the sheets were carefully prepared by an ion-beam milling method (Gatan, Model 600, Oxford, UK). Electron energy loss spectroscopy (EELS) data were collected using an incident e-beam of 1200 eV, a 0.5 eV per step and resolution of 2 eV, measured from the full-width at half-maximum (FWHM) of backscattered electrons. For the X-ray photoelectron spectroscopy (XPS, K-alpha, Themo VG, UK) measurements, after exciting the films by the Al Kα line (1486.6 eV) and completed at resolution of 50 and 0.1 eV energy steps. The energy scale was measured in the Ag 3d 5/2 .