Growth of diamond in liquid metal at 1 atm pressure.
Liquid metal containing Ga, Ni, Fe, and Si was used for the growth of diamonds. We used a home-built cold wall vacuum system that can rapidly heat and cool the metal (Fig. 1a). A graphite crucible connected to two water-cooled copper electrodes is joule heated by electrical current at a given voltage (Fig. S1). A pyrometer is connected into a feedback loop and used to measure and control the temperature of the graphite crucible; its ~ 1-mm diameter laser spot was focused at the center of the outside part of the side of the graphite crucible (Fig. 1b, upper). A small piece of silicon wafer was placed on the bottom of a 20 mm (length) x 10mm (width) x 15 mm (depth) cavity in the crucible and liquid gallium was then added, followed by adding nickel and iron ‘ingots’ (Fig. 1b (lower) shows this mixture in the cavity prior to heating). A typical growth run was done with the pyrometer reading 1175°C; prior to heating, methane (CH4) and hydrogen (H2) were added (after the chamber had been pumped down to the base pressure of its vacuum pump) to reach 760 Torr. Ni, Fe, and Si dissolve entirely into the liquid gallium forming a molten (liquid) metal. The temperature at different regions inside the liquid metal was measured by alumina-sheathed thermocouples, Fig. S2, Table S1-2. We refer to exposure to the methane/hydrogen mixture while heating for a given length of time as a ‘growth run’. After a growth run the electrical current was turned off, and the crucible (temperature at the bottom center) cools in about 167 seconds to 25°C (Fig. S3).
Diamonds typically grew in the central region of the bottom surface of this liquid metal alloy, at its interface with either the bottom of the cavity of the graphite crucible, or with thin pieces of HOPG (Grade ZYB, SPI Supplies) or EDM-3 Poco Graphite (0.0190-inch-thick) that we placed at the bottom of the crucible cavity to explore the potential role of different interfaces. A parametric study was undertaken in which these parameters were varied: growth temperatures, the concentrations of Ni, Fe, and Si in liquid Ga (Fig. S4-Fig. S9). We found diamonds were grown in a pyrometer reading temperature range from 1165 ºC to 1190 ºC, and that these diamonds grew most abundantly for a 77.75/11.0/11.0/0.25 mix (atomic percentages) of Ga/Ni/Fe/Si at 1175 ºC during exposure to a gas mixture of methane and hydrogen at 760 Torr that we refer to as an ‘optimized growth condition’ for growing diamonds and was used for further studies of the growth. When 13CH4 was substituted for normal methane, regions with 13C-pure diamonds could be found.
A typical growth result by using the optimized growth condition is shown in Fig. 1c. Diamonds were found on the bottom surface of the solidified Ga-Fe-Ni-Si alloy piece and had “rainbow” colors to the eye (Fig. 1c), and these regions (that were later unequivocally identified as diamond by Raman spectroscopy, transmission electron microscopy (TEM), X-ray diffraction (XRD), by observation of the negatively charged silicon-vacancy (SiV−) center in diamond, and, for X-ray photoelectron spectroscopy (XPS), by observation of only sp3-bonded carbon). As mentioned, diamonds nucleate and grow in the center region but not in other regions (some diamond crystals appear to be embedded in the solidified liquid metal surface); we suggest this is because the temperature of this center region (~ 1025 ºC) is the lowest in the liquid metal in the cavity (Fig. S2, Table S1). The diamond size and areal density were highest in the central part of this ‘diamond region’, and both are reduced as one progresses towards the outermost part of this region (Fig. S10). And, diamonds were only found in the bottom surface of the solidified liquid metal piece (we did not find diamonds on, or embedded in, the 4 sides of the metal piece or its top surface).
Both the size and the density of the diamond crystals increased with time of exposure (the ‘growth time’) to the methane/hydrogen mixture (Fig. 1d to Fig. 1g) up to 150 min growth time. For a growth time of 15 min, small polyhedral diamond crystals of a low areal density were observed on certain regions (Fig. 1d). For a growth time of 30 min, larger diamond crystals with higher areal density were observed and facets are clearly evident (Fig. 1e). For a growth time of 60 min, the diamond crystals were still larger and had merged together in some regions to form “islands” (Fig. 1f). Some islands with a size of a few microns are shown in Fig. S11. For a growth time of 150 min a nearly continuous diamond film was formed (Fig. 1g). We note that there were a few gaps in the diamond film region (Fig. S12). Scanning electron microscopy (SEM) images of the as-grown diamond with growth times of 15 min, 30 min, and 60 min were acquired at 50.0° tilt and are shown in Fig. S13. When observed in an optical microscope the continuous film shows various colors including red, yellow, and green (Fig. 1h). We note, for growth times longer than 150 min, the thickness and morphology of diamond film did not change (Fig. S14 and Fig. S15).
Characterization of diamond and the solidified liquid metal interface region
The as-grown diamond film can be delaminated and is readily transferrable to other substrate surfaces by dissolving the metal alloy piece using HCl(aq) solution (see Methods). The optical images of the as-transferred diamond film on a Quantifoil holey amorphous carbon film coated Cu TEM grid show it is transparent (Fig. 1i). Atomic force microscopy (AFM) of the as-transferred diamond film on the Cu TEM grid is shown in Fig. 1j and the facets of the diamonds are clearly seen. The plan-view TEM images of the as-transferred diamond film show diamond particles with different sizes and orientations and that there are, in some regions, gaps between the diamond particles (Fig. S16). There was no graphitic structure found in these gaps (Fig. S17). C1s XPS spectrum of an as-transferred film on a 300 nm SiO2/Si wafer (Inset image in Fig. 1k and Fig. S18) showing a single and symmetric peak at 285.1 eV correlating to sp3-bonded carbon suggests the film is diamond (Fig. S19)34. Synchrotron two-dimensional XRD (2D-XRD) pattern in grazing incidence mode of the same transferred diamond film suggests that it has a cubic structure (Fig. 1k), with (111), (220) and (311) diffraction Debye rings observed, suggesting that the as-transferred diamond film is polycrystalline.
We used 13CH4 (99 at% 13C) instead of normal methane (98.9 at% 12C) for some growth runs. We label the as-grown samples as 13C-D150-GC for a 150 min growth run, for the configuration that the entire bottom surface of the crucible cavity was contacted with liquid metal (Fig. 2a). Raman spectrum (all of which were obtained with a 266 nm excitation source) acquired on 13C-D150-GC (Fig. 2b) showed diamond peaks at 1283 cm− 1 and 1332 cm− 1, and graphite G band peaks at 1521 cm− 1 and 1580 cm− 1, which are the Raman bands of essentially pure 13C-labeled diamond (13D) and normal diamond (12D)35, and the G bands of essentially pure 13C-labeled graphite (13G) and normal graphite (12G)36, respectively. Such Raman spectrum can be rationalized by both the methane and the graphite crucible contributing carbon to the growth of diamond and graphite. With the observed much larger intensity of Raman 12D and 12G peaks as compared to 13D and 13G peaks, the graphite crucible seems to contribute larger amounts of carbon for growth of both diamond and graphite, than methane. Observing Raman 13D and 12D peaks (and the 13G and 12G peaks) with no peaks between them was unexpected. We speculate that the diffusion pathways of the carbon species coming from methane and coming from the graphite crucible are perhaps different.
The graphite crucible is composed of isotropic graphite. We asked: could different types of carbon (carbon material) ‘interfaced’ with the liquid metal, influence diamond growth at this interface? We thus covered regions of the bottom of the cavity with flat (square or rectangular) pieces of different types of graphite than the crucible graphite, among them HOPG pieces and EDM-3 Poco Graphite plates, and here we present results obtained with EDM-3 Poco Graphite (about HOPG, Fig. S20). Placing one piece of EDM-3 Poco Graphite plate, or a stack of ten of them, on the bottom surface of the cavity prior to adding the Ga, Fe, Ni, and Si led to as-grown diamonds using 13CH4 that we name 13C-D150-EDM and 13C-D150-SEDM (“S” for stacked), see Fig. 2a. Raman spectrum of 13C-D150-EDM showed 13D and 12D peaks (and 13G and 12G peaks) but the intensities of the 12D and 12G peaks were significantly lower compared to that of the 13D and 13G peaks, showing the growth was mainly from 13CH4 (that is, from methane). The spectrum of 13C-D150-SEDM showed only 13D and 13G, thus all newly grown carbon was from 13CH4. It is thus reasonable to assume that 12D and 12G Raman peaks observed in 13C-D150-EDM were coming from carbon from the graphite crucible rather than the inserted EDM-3 Poco Graphite plate: that is, the EDM-3 Poco Graphite plate does not contribute carbon to the growth of diamond and graphite. The morphologies of 13C-D150-GC, 13C-D150-EDM, and 13C-D150-SEDM were quite similar (Fig. S21). SEM images (Fig. S22) showing that the surface of the graphite crucible is rougher and has larger pores present than EDM-3 Poco Graphite plate possibly rationalizes why some carbon in the crucible contribute to the growth of diamond and graphite.
The intensity ratio of the Raman peaks for diamond and graphite (ID/IG) was higher for 13C-D150-SEDM than for 13C-D150-GC or 13C-D150-EDM. Raman maps of the 13D intensity (Fig. S23) and the ID/IG ratio (Fig. 2c) of 13C-D150-SEDM show that diamond and graphite formed continuously throughout the mapped region. (The black spots in Fig. 2c are due to cosmic rays incident on the detector during mapping (Fig. 2d)). Essentially similar ID/IG peak ratio was also observed for the as-grown diamonds grown with the same configuration with normal methane (Fig. S24).
A typical photoluminescence (PL) spectrum of 13C-D150-SEDM excited by a 532 nm laser is shown in Fig. 2e. The strong peak at 738.5 nm (1.679 eV) with a full width at half-maximum of 6.4 nm can be assigned to the zero-phonon line (ZPL) of the SiV− color center37. A photoluminescence map of ZPL intensity over a 50 µm × 50 µm region showed this ZPL everywhere but with somewhat varying intensity (Fig. 2f).
Cross-sectional TEM analysis was used to study the atomic scale structure and elemental composition of as-grown diamonds, as-grown graphite, and different regions of the solidified liquid metal piece that were interfaced with these solid carbons. The cross-section specimens were prepared by focused ion beam (FIB) milling of the metal pieces from growth times of 150 min (D150) and 30 min (D30) (see Methods). A typical large-area cross-section TEM image of D150 showed a diamond film at the solidified liquid metal surface (Fig. 3a). (We note the growth of diamond was usually accompanied with the growth of graphite in our experimental conditions, and the surface of solidified liquid metal was not always flat. More details are shown in Fig. S25 and Fig. S26.) High-resolution TEM (HR-TEM) images of the diamond-metal interface show that the metal region has two different structures: one (labeled as M1) with a thickness of about 30 nm, while the other (labeled as M2) is interfacing with M1 (Fig. 3b-c). The fast Fourier transform (FFT) patterns of M1 (upper inset) and M2 (lower inset), show that M1 is amorphous but M2 is crystalline (Fig. 3c). M2 shows a crystal lattice with clearly identifiable fringe spacings and without any structural defects or disorder. (While our ‘quenching’ cooled the liquid metal due to turning off the electric current that had been driving the Joule heating, as described above, it did not “freeze in” the amorphous liquid metal state—crystallization of the interior of the liquid metal did occur in the few tens of seconds time scale of cooling—and without the formation of separated phases.)
The near-surface solidified liquid metal region that is beneath the diamond film was studied by TEM Energy-dispersive X-ray spectroscopy (TEM-EDS) line profiling, by scanning along the surface of M1 to the bulk of M2 (Fig. 3d), and we found a significant amount of carbon (as atoms or very small clusters) is present in the M1 region. The carbon concentration decreases significantly from ~ 26.5 at% at the top surface to ~ 5.0 at% at a depth of ~ 40 nm, that is matched to the thickness of M1, and (seemingly) plateaus at ~ 3.0–5.0 at% for the M2 region. However, the M2 region might be carbon-free as the detection limit of this method of identifying carbon concentration is ~ 2.0–4.0 at%. Thus, a high percentage of carbon is “dissolved” in the M1 region of the (solidified) liquid metal prior to very rapid solidification but not to any, or any appreciable degree, in the bulk of the liquid metal; the presence of such a high concentration of carbon in the M1 region is perhaps the reason for this region being amorphous.
Electron energy loss spectroscopy (EELS) spectra acquired from 20 randomly selected regions over the diamond region show only one major σ* peak around 292 eV (Fig. 3e)38, suggesting the whole region is a uniform diamond film. Atomic resolution TEM (AR-TEM) image obtained from the diamond region (marked by the yellow cross in Fig. 3b) observed from the [110] zone axis (that is, the beam direction (BD) is [110]) show the region is running parallel to the (110) diamond planes, with the C − C ‘dumbbell’ units observed as bright spots in Fig. 3f39,40.
We further imaged the interface between diamond and the (solidified) liquid metal surface at the early stage of growth (D30). Isolated diamond crystals (for example, the one labeled as D1 in Fig. 3g) were found at the interface (Fig. 3g), and this agrees well with the SEM images (Fig. 1e). Amorphous M1 and crystalline M2 regions were again found underneath the diamond crystals (Fig. 3g-h, Fig. S27; we note M1 in some regions was delaminated from M2). TEM-EDS maps of C, Ga, Ni, Fe and Si in the region where D1 is located were obtained (Fig. S28). TEM-EDS analysis shows the M1 region of D30 is carbon-rich (Fig. S29).
D1 was found to be directly contacting the surface of the M1 region (Fig. 3h). The D1 crystal lattice fringes are not aligned with the same angle to the M1 surface (i.e., there is not a constant orientation between the D1 lattice fringes and the M1 surface; Fig. 3i).
How does diamond grow?
Here, we speculate about how nucleation and growth might progress, based on our experimental results. The formation of a thin carbon rich M1 region suggests that the growth of diamonds was facilitated by surface catalysis and diffusion of carbon atoms in a thin subsurface region of the liquid metal. That is, carbon is present only in this subsurface region of the liquid metal and this carbon leads to the nucleation and growth of diamonds. During the growth at high temperature, Ni, Fe and Si are all dissolved in the Ga solvent and perhaps all 4 elements (or some more than others) contribute to catalytic generation of C atom precursors that grow diamond. The bottom surface of this liquid metal (at the interface between the metal and the graphite crucible) dissolves carbon atoms or C-H radicals from methane and/or the graphite crucible surface, and the concentration of the precursor carbon species likely follows the temperature gradient at the interface (Fig. S30). We speculate that the relevant carbon species diffused from the high temperature regions (the sides of the graphite crucible) to the center regions (of the bottom surface of the molten metal that is at a relatively lower temperature (Fig. S2, Table S1)), to form diamond crystals. Our control experiments with only Ni dissolved in Ga yielded a very thick graphite film on the entire surface of the liquid metal (Fig. S7 and Fig. S9). The formation of graphite was suppressed with Fe dissolved in the liquid metal alloy (Fig. S8 and Fig. S9).
We emphasize that Si plays a critical role in our growth of diamond. Si was also relevant in our prior studies of homoepitaxial growth of diamond on single crystal diamond41. 0.25 at% Si used for our typical growth yields the diamonds of largest crystal sizes at 150 min growth. 0.50 at% Si (and 150 min growth) leads to a much higher density growth of diamonds of smaller crystal sizes (Fig. S5 and Fig. S6). These two results along with our observation of the SiV− color center, suggest Si plays a role in the nucleation of diamond; a higher nucleation density would rationalize the higher density growth of smaller diamonds at 0.50 at%. We note that the absence of Si (per TEM-EDS) in the M1 region correlates with cessation of diamond growth (Fig. S31). For example, Fig. S25 and Fig. S31c-d shows C present in M1 and a graphite layer beneath the diamond film layer—but no Si. We found by AFM measurements essentially the same thickness diamond film for 300 min growth runs vs 150 min growth runs (Fig. S15). (That is: after ~ 150 minutes there is no more Si present, and longer growth runs do not yield more diamond.)