Novel diamond cells for neutron diffraction using multi-carat CVD anvils

Typically, for neutron diffraction, sample volumes of several mm³ are required for large volume devices using sintered diamond anvils [for example, the so-called Paris-Edinburgh (PE) presses] limiting the pressure range to just above 20 GPa.^1–3 This limitation is overcome with the availability of new, high flux spallation sources such as ORNL’s Spallation Neutron Source (SNS) because much smaller samples can be studied in diamond cells. The advantages of diamond cells in addition to the higher pressure range are manifold: pressures can be measured optically by the ruby method,⁴ the samples can be visually and spectroscopically studied and even laser-heated. Moreover, the significant weight reduction of a factor of over ten and the absence of hydraulic seals allow convenient low temperature experiments. The use of compact diamond cells was pioneered by Goncharenko.⁵ 

It has been demonstrated recently that megabar pressures (up to 100 GPa) can be reached in neutron diffraction experiments⁶ given the high neutron flux and focused beam at the Spallation Neutrons and Pressure Diffractometer (SNAP) of the SNS. For a material with good neutron scattering properties such as D₂O-ice, a sample volume of at least 0.02 mm³ was required to obtain diffraction patterns, which for good quality, however, still required several hours of exposure time.⁷ The diamond anvils were relatively small (sub carat) because natural diamonds become rapidly unaffordable with increasing size. Therefore, due to the high load on millimeter size culets, the anvil supports had to be made from polycrystalline diamond, which requires extremely high and costly machining precision. Furthermore, the geometry of the anvil support limited the opening aperture to ±15°, corresponding to a Q range of 2.1-19.9 Å⁻¹ (d range 3–0.3 Å).

In the following, we introduce novel, large, multi-carat anvils produced via Chemical Vapor Deposition (CVD) and describe new designs of large-aperture diamond cells with low-mass integrated membrane presses. We demonstrate that high quality diffraction patterns with significantly larger Q-range can be obtained in a fraction of time compared with that of previous neutron diamond cells.

Recent years have seen an impressive development in the commercial growth of large single crystal diamonds by CVD, a technique introduced in the early 1950s and now optimized for diamond growth. Thus, there exists a large amount of literature on growth and properties of CVD diamonds.⁸ During the last 15 years, this technology has led to the growth of large single crystals at a variety of industrial and research institutions.^9,10 For example, the Gemological Institute of America reported a gem-quality synthetic diamond weighing over 5 carat in 2016.

Whereas the cost of natural diamond increases with size nearly exponentially, for CVD diamond, this relationship is more linear. Figure 1 shows a variety of CVD anvils ranging up to over 10 carat (2 g) in weight with a volume of approximately 1 cm³. This allows for culet sizes of up to 3 mm and sample volumes of ∼1 mm³. The largest anvils in this figure cost only about four times as much as the smallest ones.

In the previous study,⁶ the largest anvils had a diameter of 4 mm and a height of less than 3 mm with culets ranging from 1.0 to 1.6 mm diameter. The forces required for maximum pressure were up to 11 tons, and this required costly, high-strength, polycrystalline diamond (PCD)-anvil supports. These PCD seats, due to their brittleness, failed frequently resulting in anvil damage in spite of them being supported by binding rings. This is understandable due to the high stress concentrations caused by even a minute elastic deformation of the seat. Thus, in spite of the successes in using this technique to reach record pressures, it is not well suited for routine high pressure neutron diffraction measurements.

The diamond anvils used for the present study were developed in cooperation with the Washington Diamonds Corporation (WDC, now WD Lab Grown Diamonds, publication in preparation). The CVD technique to grow diamonds produces clear, nearly flawless diamonds with low nitrogen content (type IIa). The optical quality depends on both nitrogen content and growth rate. Thus, Raman low fluorescence grade anvils can be readily produced and have been tested to multi megabar pressures.¹¹ For neutron diffraction, much larger anvils can be made at much higher growth rate at significantly reduced cost because high optical quality is not essential. For example, the price of the largest anvil shown in Fig. 1 of about 10 carat weight and dimensions of about 10 × 10 mm is about $7000. For the present report, anvils with dimensions 6 × 6 mm were used, the price of which is about half of the anvil/seat combination in the previous study.⁶ 

First tests with simple cylindrical anvil/seat geometry proved to be unsuccessful. Due to the high force requirements exceeding 10 tons, radially unsupported cylindrical anvils and flat seats resulted in substantial stress concentrations and splitting of the anvils at very low loads (∼2 tons). Supporting the anvils radially by press-fitting into binding rings only resulted in an insignificant improvement with the increased risk of anvil damage during this process, based on the strongly anisotropic mechanical strength of the diamond.

A drastic improvement was achieved by using the well-developed conical anvil/seat design.¹² This design has the advantage of simultaneously increasing the radial support with increasing axial load and it is now widely used for standard-size diamond anvils. The drawback of this design is the high precision requirement for grinding and matching seat and anvil, which for small anvils is not problematic. For large anvils, however, grinding cones precisely is very costly and high strength seat materials again introduce failure from stress concentrations due to the strong elastic deformation as described above. We solved both problems by commercial laser-cutting the cones and using steel as seat material. To date, the highest load of 12 tons generated high quality diffraction patterns at above 40 GPa with 2 mm diameter culets. Neither anvils nor seats failed during compression, however, upon unloading, sometimes ring cracks developed on the culets. Figure 2 shows a cross section of typical anvils used in this study. For the present study, we used two types of high strength steels as conical supports for the anvils: Vascomax C350 and Inconel 718 with yield strengths of 2.3 and 1.2 GPa, respectively. 718 was tested in view of experiments using hydrogen loaded in a gas loader because we found it to be significantly less subject to hydrogen embrittlement than VM C350.

The cells, shown in Fig. 3, are made from CuBe alloy 25 (C172). The choice of this material was mainly based on strength and its high thermal conductivity, speeding up cooling if required. For example, cooling of the cell shown in Fig. 3 to 4.7 K was achieved in 4.5 h. Opening angles of 70° × 120° allow a calculated Q-range from 1.6 to 22 Å⁻¹ (d range 3.9–0.29 Å) for the wavelengths at SNAP (0.5-3.65 Å). Experimentally Q as low as 1.25 Å⁻¹ has been accessed due to asymmetric anvil positions and the absence of Cd shielding on the lower-Q side of the cell.

The upper Q range was somewhat reduced due to Cd shielding, which was necessary to reduce effects of scattering from the hexagonal boron nitride collimator. The cell dimensions were chosen to fit a new 0.3 GPa gas loader built at ORNL (manuscript in preparation). For stability and simplicity in handling, the anvils have no lateral or tilt adjustments as the anvil/seat units are precision machined to sufficient accuracy. The cells are loaded in a two-step process: first the cell is loaded with a hydraulic press to 0.5t to seal the sample. That load is stored in the cell by simultaneously tightening the top cap screw. Then, the cell is attached to the membrane press (see Fig. 4). The membrane was pressurized with He to maximum pressures of 120 bars achieving forces up to 12 metric tons (about 120 kN). We have not yet tested the maximum achievable pressure of these cells; however, given the linear character of the pressure load curves (Fig. 5), we believe that they are capable of routine neutron diffraction to at least 50 GPa with 2 mm culets.

The diffraction geometry for diamond cells leading to the best signal-to-noise ratio and best resolution is such that the neutron beam enters the cell along the pressure axis through the diamond anvil. The beam size is adjusted to the sample diameter by a neutron-absorbing (hexagonal boron nitride or boron carbide) collimator (see Fig. 4). The diffracted beam exits the cell at 90° ± 35° through the gasket. We found from numerous tests that stainless steel gaskets (for example, 15-5 PH) yield the lowest absorption and scattering, while providing reasonably high strength. Initially, we made gaskets by indenting simple discs, the traditional method. The large dimensions and high forces required for this process, however, resulted in strong, irregular deformation of the material and frequent anvil damage. We therefore developed new gaskets that resemble mini girdle pressure vessels¹³ shown in Fig. 6. These gaskets were formed by either conventional machining or by pressing with tungsten carbide dies. Using these gaskets, we observed significantly improved sample stability, very little drifting of the gasket chamber and reduced anvil damage (e.g., ring cracks), especially upon unloading. For the present work using 2 mm culets, the sample diameter was 0.95 mm with 0.2 mm thickness.

Pressures can be measured online using either the ruby scale⁴ or the equation-of-state of the sample. Naturally, for such large samples there exist substantial radial pressure gradients and the values given here are average pressures. The gradients depend on the mechanical properties of the material and in some cases diminish with time due to relaxation. In the near future, inert gases added using the new gas loader will help us to significantly reduce these gradients.

In order to demonstrate the advantage of single-crystal diamond cells over conventional large-volume cells, we loaded silicon samples of typical quantities for these devices. The volume of illuminated material was ∼17 mm³ in a PE press outfitted with double-toroidal anvils and ∼0.2 mm³ in the DAC. The spectra shown in Fig. 7, remarkably, have similar signal-to-noise ratio. The signal-to-background is a factor of at least 4 better in the DAC compared with PE data. Moreover, despite the large sample volume ratio, the DAC required a much shorter exposure time (1 versus 4 h). Furthermore, masking allows for the removal of the single-crystal diamond peaks in the DAC case but not for the removal of the powder diamond peaks in the PE cell. Last but not least, not only the exposure times are significantly reduced but also the long loading and unloading times required by the PE presses (hours) are reduced to only a few minutes.

One of the first pressure runs in the new cells was done with a pure nickel sample, as it has one of the highest neutron coherent scattering cross sections. Ni powder was loaded into the DAC and then pressurized using the membrane in 10 bars steps. Data were collected for 1 h at each step. Pressures were obtained from a third order Birch-Murnaghan equation of state of Ni with B₀ = 189 GPa and B₀′ = 4.7.¹⁴ Figure 8 shows representative spectra at ambient pressure and 30 GPa for a sample volume of 0.14 mm³. The spectra are typically normalized by accumulated protons on the target and incident spectrum/detector efficiency as determined by a vanadium measurement and by a polynomial fit to their background.

A third example demonstrates the uniqueness of neutron diffraction to probe disordered materials: A key advantage of neutron diffraction lies in the fact that atomic form factors are flat, in contrast to X-ray scattering where they are strongly Q-dependent. This simplifies the Fourier-transformations routinely performed during total scattering experiments. This is demonstrated in another pressure run, where disordered glassy carbon (GC, Sigradur G) was loaded into a DAC with anvils of the same dimensions. The high pressure phase behavior of GC has attracted recent interest due to its high potential as a precursor material for high pressure synthesis or transitions. For example, a reversible sp²-sp³ transition at ∼40 GPa has been reported based on X-ray Raman spectroscopy.¹⁵ In contrast, another very recent study reports an irreversible sp²-sp³ transition from GC to hexagonal diamond at 100 GPa and 400 °C.¹⁶ This transition behavior and specifically the structural changes in GC during compression are however, poorly understood since the light element carbon is not a good X-ray scattering sample. Instead it is a prime example for the advantages of neutron diffraction.

The sample was pressurized from ambient to 15 and 45 GPa taking spectra at each pressure with exposure durations of ∼8 h. This is thus far the largest pressure range for studying a disordered material using neutron scattering. The spectra, as is typical for disordered materials, are corrected based on the spectra from an empty cell and one loaded with vanadium. The {002} peak (the first sharp diffraction peak) of GC at Q = ∼1.80 Å⁻¹ corresponds to weak, interlayer, highly compressible bonds. The position of this peak was used to estimate the pressure based on an experimental equation of state obtained from X-ray diffraction (Shiell, manuscript in preparation). Due to the difference in these two run conditions, the uncertainty in the present pressure estimates may be as high as ±3 GPa. The pressure shift of this peak together with the resulting pressure measurements are shown in Fig. 9.

We demonstrated that routine neutron diffraction experiments to at least 40 GPa can be carried out in a diamond cell using large single crystal CVD anvils. The sample sizes of order 0.15–0.2 mm³ produce high quality diffraction spectra in much shorter exposure times than in large volume devices with high signal to noise ratios. This is shown for three test cases: silicon, nickel, and glassy carbon. The new cells provide a considerably expanded Q-range compared with previous high pressure cells ranging from 1.25 Å⁻¹ to Q = 22 Å⁻¹ (d range 5–0.29 Å). The cell/press assembly is compact and low weight (about 4.4 kg) allowing easy handling and cooling. A further development of large CVD anvil technology is promising for future routine neutron diffraction in the megabar pressure regime.

We most gratefully acknowledge Tom B. Shiell and Jodie E. Bradby (Australian National University, Australia), Dougal G. McCulloch (RMIT University, Australia), and David K. McKenzie (University of Sydney, Australia) for their collaboration on the work concerning glassy carbon. We also gratefully acknowledge Stas V. Sinogeikin (HPCAT/Geophysical Laboratory) for the membranes used. We greatly appreciate the continuous support of Yarden Tsach of VFG technologies in developing large CVD diamond anvils. Work by R.B. was supported by the Energy Frontier Research in Extreme Environments (EFree) Center, an Energy Frontier Research Center funded by the U.S. Department of Energy (DOE), Office of Science, Basic Energy Sciences (BES) under Award No. DE-SC0001057 and by the Laboratory Directed Research and Development Program of Oak Ridge National Laboratory, managed by UT-Battelle, LLC, for the U.S. Department of Energy. This research was partially conduced at the SNAP beamline of the Spallation Neutron Source, a DOE Office of Science User Facility operated by the Oak Ridge National Laboratory.