An insight into what superconducts in polycrystalline boron-doped diamonds based on investigations of microstructure

Abstract

The discovery of superconductivity in polycrystalline boron-doped diamond (BDD) synthesized under high pressure and high temperatures [Ekimov, et al. (2004) Nature 428:542–545] has raised a number of questions on the origin of the superconducting state. It was suggested that the heavy boron doping of diamond eventually leads to superconductivity. To justify such statements more detailed information on the microstructure of the composite materials and on the exact boron content in the diamond grains is needed. For that we used high-resolution transmission electron microscopy and electron energy loss spectroscopy. For the studied superconducting BDD samples synthesized at high pressures and high temperatures the diamond grain sizes are ≈1–2 μm with a boron content between 0.2 (2) and 0.5 (1) at %. The grains are separated by 10- to 20-nm-thick layers and triangular-shaped pockets of predominantly (at least 95 at %) amorphous boron. These results render superconductivity caused by the heavy boron doping in diamond highly unlikely.

After the discovery of superconductivity in boron-doped diamond (BDD) (1), numerous theoretical and experimental studies (2–10) confirmed the phenomenon and went along with its explanation. Superconductivity in group IV semiconductors with diamond structure—such as silicon, germanium, and their alloys—was already predicted in the early 1960s to occur at very low temperatures (11). Except for diamond (1–3, 8, 10), there was a report on experimentally observed superconductivity in boron-doped silicon (9). For diamond, Ekimov et al. (1) suggested that the superconducting transition temperature (T_c) increases with heavy boron doping (T_c ≈ 4 K at 2.6 at % B). However, further investigations of superconducting BDD prepared by the high-pressure–high-temperature (HPHT) technique and by chemical vapor deposition (CVD) revealed strong inhomogeneities in these materials (1, 3, 7, 10, 12). In particular, B-rich phases [such as B₄C (1, 7, 10, 12), used as a reactant, and B₅₀C₂ (10, 12)] were found in HPHT BDD samples. In BDD prepared by CVD sp²-bonded carbon is unavoidable (3), and one cannot exclude that sp²-bonded amorphous or graphite-like carbon accumulates some amount of boron. The presence of extra phases and large discrepancies in the B content determined by various methods (1–3, 7–10, 13, 14) (such as secondary ion-mass spectroscopy, Hall-effect measurements, correlations with diamond lattice parameters, infrared spectroscopy, and microprobe analysis) and the absence of a clear correlation between T_c and the B concentration (10) raise the question of how much boron indeed is incorporated into the diamond structure of superconducting samples. It is remarkable that, although measurements of the temperature dependence of the resistance have been conducted on BDD single crystals with boron content ranging from ≈10¹⁹ to 2.7 × 10²¹ cm⁻³ (i.e., up to 1.53 at % B), no evidence of superconductivity was found down to 0.5 K (15, 16). This is at odds with the results for the polycrystalline materials and contrasts especially with one report (3), where superconducting onset temperatures as high as 7.4 K were observed for CVD BDD samples with even smaller boron concentrations [9.4 × 10²⁰ cm⁻³ (0.53 at %)] (3).

We investigated polycrystalline BDD samples that were synthesised at 20 (1) GPa and 2,300 (50) K (four samples) as well as at 9.0 (5) GPa and 2,500 (50) K (one sample) in a 5,000-tonne press by using a HPHT technique as described in refs. 10 and 12. A mixture of graphite (referred further to as ¹²C) or isotopically pure amorphous carbon ¹³C and B₄C in a ratio C:B = 13:1 (≈7 at % B) were used as starting materials. Synchrotron and in-house x-ray diffraction investigations revealed for all five samples studied that their main crystalline component (>99%) is diamond with a small amount of boron carbide B₅₀C₂. In some samples residuals of the starting material B₄C were found (10, 12). Microprobe analysis of the samples revealed a boron concentration of 2.6 at % (4.6 × 10²¹ cm⁻³) in the samples, whereas Hall-coefficient measurements gave a charge-carrier concentration of 1.4 × 10²¹ cm⁻³, that is, three times as low as the apparent B concentration in the material (supporting information (SI) Fig. S1). Both microprobe analysis and energy-dispersive x-ray (EDX) spectra did not show the presence of other elements than carbon and boron in BDD samples.

The superconducting transition temperatures were determined by means of electrical transport and specific heat measurements (Fig. S2 and Fig. S3), and were found to be in good agreement with earlier reports (1, 7, 10).

The boron content in our samples was characterized by measurements of diamond lattice parameters, Raman spectroscopy, Hall-effect measurements, scanning electron microscopy, and microprobe analysis (10, 12). All of the methods mentioned above give only sample-averaged results. In particular, no information on the spatial distribution of boron on the submicron level in the diamond samples was available. To investigate the microstructure and the exact boron location in our superconducting BDDs we used transmission electron microscopy (TEM). Electron transparent foils of BDD samples were prepared by means of focused ion beam (FIB) techniques (for details, see Materials and Methods). Fig. 1 shows two bright-field TEM images of our polycrystalline sample, which consists of micrometer-size carbon grains with predominantly diamond structure as revealed by x-ray and confirmed by electron-diffraction data. A number of grains from different samples were investigated by using high-resolution TEM (HRTEM) (SI Text, Fig. S4). All investigated grains are separated by layers of amorphous material along straight boundaries with thicknesses of 10–20 nm. The material constituting the layers along the grain boundaries also fills triangular-shaped pockets at the grain junctions (Fig. 2A). With side lengths reaching 0.5 μm some of the triangular pockets have sizes comparable to some diamond grains. The complete absence of any TEM diffraction contrast (uniform light-gray contrast during sample tilting) indicates the amorphous state of the filler material (Fig. 2A).

Fig. 1.

Bright-field TEM images of a polycrystalline BDD sample. (A) A micrometer-sized grain is separated from other grains by a clearly visible straight boron-rich boundary. (B) Twin boundaries and dislocations observed within a grain.

Fig. 2.

The results of TEM and EELS investigations. (A) Bright-field TEM image of a BDD sample. The amorphous material constituting the layers along the grain boundaries also fills triangular-shaped pockets at the grain junctions. (B) Electron energy-loss spectra (EELS) at the boron K and carbon K ionization edges from the pockets (Lower) and the diamond grains (Upper) reveal that the pocket material consists of ≈95 at % boron with a small amount (≈5–6 at %, varying in different pockets) of carbon. The amount of boron in the diamond grains does not exceed 0.5%. The π* peak due to sp²-bonded carbon is very common for the pure diamond investigated by TEM on carbon grid. A smaller beam size was selected to measure the intergranular pocket, thus resulting in lower intensity and a more “noisy” spectrum (Lower).

To probe the chemical nature of the inter- and intragranular material in our samples, we measured the electron energy-loss spectra (EELS) at boron K and carbon K ionization edges for numerous pockets and grains (Fig. 2B). The EELS spectra (Fig. 2B Lower) revealed that the pocket material contains ≈95% boron with a small amount of carbon (≈5–6%, varying for the different pockets). The appearance of some amorphous boron material under high-pressure, high-temperature syntheses in the B–C–N–O system was reported in refs. 17 and 18. The authors (17, 18) provided a detailed investigation of boron K edges from several α-rh B-bearing materials. They demonstrated that the B K edge energy-loss near-edge structure (ELNES) of α-rh B and the products with the α-rh B structure exhibit an edge rich in features (17). The ELNES for the B K edge from α-rh B and structurally related materials was divided into the π* (<195 eV) and σ* (195–210 eV) regions. The B K edge shapes from the α-rh B-bearing materials could be divided into three types (17): α-rh B, B₆O, and B₄C. However, the EELS spectra we obtained from the carbon-doped amorphous boron phase (with sharp features at ≈189 eV and 199 eV in the π* and σ* regions, respectively; Fig. 2B Lower) at the B K edge are different from that known for crystalline boron, boron carbides, nitrides, oxides, and boron carbooxinitrides (17–20). This suggests that the intergranular boron does not form B₁₂ icosahedra typical for the structure of pure boron and boron carbides. The observed EELS spectra are also different from the spectrum of amorphous boron carbide (21). An example for an elemental mapping using EELS is shown in Fig. 3. Practically all boron is concentrated in the grain boundaries and pockets.

Fig. 3.

Elemental B map obtained by using EELS and consisting of four images of an intergranular area. It shows four diamond grains, five grain boundaries, and two pockets. The vast majority of boron is concentrated in the grain boundaries and pockets. Boron is quite homogeneously dispersed in a very small amount within the diamond grains.

The accurate investigation of the interior of the diamond-structured grains (Fig. 3) shows that boron is quite homogeneously dispersed in a very small amount within the carbon grains. Boron mapping did not show any directional dependence in B concentration within the diamond grains. To quantify the amount of boron in the well crystallized diamond part of the grains, we measured a number of EELS spectra at various points (Fig. 2B Upper) Our results show consistently that the true amount of boron incorporated into the diamond structure is between 0.2 (2) and 0.5 (1) at % (see details of quantitative EELS analysis in Materials and Methods). That is considerably less than the 0.8 at % estimated from the number of charge carriers determined by the Hall-effect measurements, and almost one order of magnitude less than the overall amount of B (2.6 at %) as determined by electron microprobe analysis (10, 12). This result is consistent with our observation that boron is mainly localized in intergranular noncrystalline layers and pockets.

Our EELS data show as well that there are a few small (5–50 nm) isolated spots with very high B concentration within the diamond grains. These spots have platelet-like shapes (Fig. S5). HRTEM revealed their crystalline structure and the electron diffraction patterns showed reflections at 4.36 Å and 3.92 Å characteristic for B₅₀C₂, in agreement with synchrotron diffraction data obtained for this sample (12).

Our observations render it highly unlikely that the observed superconductivity in BDD synthesized at high pressures and high temperatures is actually related to the boron incorporated in the diamond. If so, then even <0.5 at % B would be sufficient to induce superconductivity in diamond with onset temperatures of ≈2.4 K and transition widths of <1.4 K. Our heat capacity measurements confirm the bulk nature of the superconductivity. However, the total amount of material converted to the superconducting state in our samples is of the order of 20% or less, consistent with other reports (7) (Fig. S3). If boron doping at a level of 0.5 at % would actually be enough to generate superconductivity in diamond, then, based on our observation of a homogeneous doping of all diamond grains, one would expect the majority of the sample to become superconducting (not only <20%). Moreover, as mentioned, presently available data for single-crystalline BDD contradict the results obtained for polycrystalline samples. Indeed, superconductivity has not been observed resistively in single-crystalline diamonds with B contents up to 1.53 at % (15, 16). However, the diamond grains in our superconducting HPHT BDD samples are completely isolated from each other by a boron-rich amorphous phase. Consequently, this phase controls the superconducting transition at least in the electrical transport data.

Besides the investigation of BDD produced by the HPHT technique we were able to study the microstructure of a superconducting (T_c = 5.2 K) boron-doped thin film deposited on a single-crystalline diamond (111) surface by use of the microwave plasma-assisted chemical vapor deposition (MPCVD) method (22, 23). The sample was kindly supplied by Dr. Y. Takano (National Institute for Materials Science, Tsukuba, Japan). Our HRTEM analysis surprisingly revealed, instead of the expected diamond morphology, a graphite-like structure (24) with a homogenous B distribution within the film according to EELS mapping (Fig. S6).^§§ Consequently, also for this sample, no evidence for superconductivity in BDD was found.

Based on the results of our research, we can suggest a few hypotheses for the explanation of the nature of superconductivity in the investigated samples: (i) Undistorted diamond becomes superconducting at B concentrations much lower than reported, but that strongly contradicts the results on single-crystal BDD samples (15, 16). This also does not agree with the experimental heat capacity data showing that only a small fraction of the sample becomes superconducting, although diamond forms its major part. (ii) The graphite-like regions with comparatively moderate B concentrations inside the grains become superconducting. However, these innermost parts of diamond grains are not connected to each other; this hypothesis cannot explain superconductivity observed in resistivity measurements. (iii) The intergranular boron-rich material becomes superconducting. In our opinion, the third hypothesis is most probable, because carbon-doped amorphous boron phase, which is filling all of the intergranular space in the polycrystalline samples, forms a continuous net through the whole sample and its amount agrees with the estimations of a superconducting part of the sample based on the heat capacity experimental data.

To prove that superconductivity in diamonds is caused by a certain amount of boron doping one would need to measure well characterized single-crystalline BDD. Otherwise, the presence of other complex phases in the available composite materials hampers any definite conclusion.

Materials and Methods

Details of the samples synthesis and preparation are described in refs. 10 and 12.

Transmission Electron Microscopy (TEM).

Electron transparent foils were prepared by focused ion beam (FIB) techniques. This allows the preparation of site-specific TEM foils with typical dimensions of 15–20 μm width, ≈10 μm length, and ≈0.150 μm thickness. Focused Ga ions have been used to sputter material off the sample. A special device, called a selected carbon mill (SCM) significantly enhanced the sputtering process of diamond. In this process OH-containing MgSO₄ was used, which was heated inside the SCM device. Water vapor was brought close to the ion beam by inserting a needle close to the location where the foil was cut. The H₂O decomposed by the ion beam and oxygen oxidized the diamond. Details of the FIB technique and the use of a SCM are described elsewhere (25).

TEM investigations were performed with a TECNAI F20 XTWIN transmission electron microscope operating at 200 kV with a field emission gun electron source. The TEM is equipped with a Gatan Tridiem filter, an EDAX Genesis x-ray analyzer with an ultrathin window and a Fishione high-angle annular dark-field detector (HAADF). The Tridiem filter was used for the acquisition of energy-filtered images applying a 20-eV window to the zero-loss peak. EELS spectra of the different K edges (B, C) were acquired in the diffraction mode with a dispersion of 0.1 eV per channel and an entrance aperture of 2 mm. The resolution of the filter was 0.9 eV at half-width of the full maximum of the zero-loss peak. The acquisition time was 1 s. Processing of the spectra (background subtraction, removal of multiscattering, and quantification) was performed by using the DigitalMicrograph software package.

Details of Quantitative EELS Analysis.

Removal of plural scattering by Fourier-ratio deconvolution by using the DM software package; t × lambda was in the range of 0.7 to 0.8 (thickness × electron mean free path); acceptance semiangle used was 1.8 mrad; EELS quantification occurred by using the DM software package (EELS Quantification); cross-sections were calculated by using the Hartree–Slater model; background model was Power Law; for background and signal windows, we used default values of the software.

From our long-standing experience with the FIB technique we can exclude that the observed noncrystalline layers and pockets are produced during the FIB milling. Irradiation damage as a cause of the noncrystallinity can be eliminated as well, because after insertion of the sample into the TEM electron beam focusing was avoided until the tilt experiments had been carried out.

EDX spectra were obtained in the scanning transmission mode (STEM) by using the TIA software package of the TEM. Significant mass loss during analysis was avoided by scanning the beam in a preselected window (20 × 20 nm or larger). The spot size was ≈0.1 nm, and the acquisition time was 60 s at an average count rate of 60–80 cps. This resulted in a counting error of about 4–5% at a 3σ level.