Tucker J.C.,Carnegie Mellon University |
Chan L.H.,EDAX |
Rohrer G.S.,Carnegie Mellon University |
Groeber M.A.,Air Force Research Lab |
Rollett A.D.,Carnegie Mellon University
Scripta Materialia | Year: 2012
A three-dimensional (3-D) dataset of Ni-based superalloy Inconel 100 is used as a validation case for using stereology to estimate 3-D grain sizes from 2-D data. 2-D sections of the IN100 dataset are extracted, from which 3-D size distributions are estimated through the use of the Saltykov method and compared to the true 3-D statistics. The Saltykov method corrected the upper tail disparity between the 2-D and 3-D grain size distributions, but the lower tail of the distribution was not improved. © 2012 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.
Fullwood D.,Brigham Young University |
Vaudin M.,U.S. National Institute of Standards and Technology |
Daniels C.,Brigham Young University |
Ruggles T.,Brigham Young University |
Materials Characterization | Year: 2015
Abstract Cross correlation techniques applied to EBSD patterns have led to what has been termed "high-resolution EBSD" (HR-EBSD). The technique yields higher accuracy orientation and strain data which is obtained by comparing a given EBSD pattern with either a real or a simulated reference pattern. Real reference patterns are often taken from a "central" position in a given grain, where it is hoped that the material is "strain-free", and they enable the determination of relative changes in orientation and strain from that present in the lattice at the reference position. Simulated patterns, on the other hand, enable comparison of the sample lattice with that of a perfect lattice, resulting in a measure of absolute strain and orientation. However, the simulated pattern method has several drawbacks, including the need to accurately specify microscope geometry, the lower fidelity/detail of simulated patterns compared with real patterns, and the potential for microscope-specific bias in the measured patterns (such as due to optical distortion). These drawbacks have led to much debate about the utility of the cross-correlation technique using simulated patterns. This paper is the first to assess the accuracy of the simulated pattern method relative to the real pattern approach in a setting where accuracy can be reasonably determined, thus providing a fair assessment of the potential of the simulated pattern technique. Based upon recent developments towards a standard material for assessing strain mapping techniques, this paper assesses the overall accuracy of the simulated pattern technique. Mismatch strains are calculated using both the real and simulated pattern techniques for a SiGe film deposited on a Si substrate. While the simulated pattern technique is not as accurate or precise as the real pattern technique for providing relative strains, it provides an estimate of absolute strain that is not available via the real pattern approach. © 2015 Elsevier Inc.
« ARPA-E to issue funding opportunity for advanced technologies for seaweed cultivation for fuels and chemicals | Main | Porsche introducing new Panamera 4 E-hybrid PHEV at Paris Show » Boskalis, a leading global dredging and marine expert, and GoodFuels Marine, a leading provider of sustainable marine biofuels to the global commercial shipping fleet, have successfully performed live tests on a sustainable wood-based drop-in biofuel called UPM BioVerno. (Earlier post.) The tests were conducted on a vessel working on the Dutch Marker Wadden nature restoration project in the middle of the Markermeer lake. UPM BioVerno is produced from wood-based tall oil. Crude tall is a natural extract of wood, mainly from conifers. The renewable raw material comes from sustainably managed forests. Crude tall oil is gained as a result of the separation process of fibrous material from wood; it is a residue of pulp manufacturing. The production process was developed in the UPM Biorefinery Research and Development Center in Lappeenranta, Finland. The fuel supplied by Finland-based UPM Biofuels is the first biofuel derived from wood residue used in a marine fleet. Boskalis vessel EDAX, a 1696 deadweight tonne (DWT) cutter suction dredger, has successfully used the fuel in bio/fossil blends going up to 50% as it worked on the first phase of the Marker Wadden project in the first half of 2016. This resulted in a CO saving of 600Mt over the operating period. The €33-million project includes the construction of an island with underwater landscaping to restore the Markermeer’s delicate ecosystem. The testing of this fuel marks yet another milestone for the marine biofuels consortium that was announced in October last year by GoodFuels Marine, Boskalis and Wärtsilä, the global supplier of engines and power systems to the marine industry. The consortium was launched with the mission to spearhead a two-year pilot programme to accelerate the development of truly sustainable, scalable and affordable marine biofuels. All fuels being live-tested on board of Boskalis vessels—including UPM BioVerno—were first extensively ground tested at the Wärtsilä lab in Vaasa, Finland. Sustainable marine biofuels offer ship operators a way to reduce a vessel’s CO emissions by 80-90%. They eliminate SO emissions, cut NO emissions by 10% and reduce particulate matter (PM) expelled in a ship’s exhaust plume by 50%. Current forecasts predict that marine biofuels could make up 5 - 10% of the marine fuel mix by 2030, significantly contributing towards the reduction of the shipping industry’s carbon footprint.
The starting material for experiments to determine the melting-phase relations of carbonated MORB (ATCM1) replicates basalts from the IODP 1256D from the Eastern Pacific Rise20 (the reported composition of IODP 1256D basalts is the average of all analyses presented in table T17 of ref. 20) with an added 2.5 wt% CO (Extended Data Table 1). This material was formed by mixing high-purity SiO , TiO , Al O , FeO, MnO, MgO, Ca (PO ) and CaCO , which were fired overnight at temperatures of 400–1,000 °C, of appropriate weights in an agate mortar under ethanol. This mixture was decarbonated and fused into a crystal-free glass in a one-atmosphere tube furnace by incrementally increasing the temperature from 400 to 1,500 °C before drop quenching into water. Subsequently weighed amounts of CaCO , Na CO and K CO were ground into the glass, introducing the alkali and CO components. After creation, the starting material was stored at 120 °C to avoid absorption of atmospheric water. Starting material ATCM2 replicates the near-solidus melt composition measured in melting experiments at 20.7 GPa and 1,400/1,480 °C. This was created by grinding natural magnesite and synthetic siderite with high-purity CaCO , Na CO , K CO , SiO , TiO , Al O and Ca (PO ) . Synthetic siderite was created in a cold-seal pressure vessel experiment run at 2 kbar and 375 °C for 7 days. A double Au capsule design containing iron (II) oxalate dehydrate in the inner and a 1:1 mixture of CaCO and SiO in the outer capsule produced a pale beige powder confirmed as siderite using Raman spectroscopy. The material for a sandwich experiment, to ensure near-solidus melt compositions were accurately determined at 20.7 GPa, was formed of a 3:1 mixture of ATCM1:ATCM2. The transition-zone peridotite mineral assemblage in reaction experiments was synthesized at 20.7 GPa and 1,600 °C for 8 h from a mixture of KR4003 natural peridotite31 with an added 2.5 wt% Fe metal. In reaction runs the recovered synthetic peridotite was loaded in a second capsule, surrounded by the ATCM2 near-solidus melt composition. Additional reaction-type experiments were performed on ground mixtures of peridotite and melt compositions. In these experiments PM1 pyrolite32 was used as the peridotite component and mixed with ATCM2 melt in 9:1, 7:3 and 1:1 weight ratios in Fe capsules. A single mixed experiment was performed in a Au capsule and used a starting mix of PM1:Fe:ATCM2 in 16:1:4 molar ratio. High-pressure experiments were performed using a combination of end-loaded piston cylinder (3 GPa) and Walker-type multi anvil (5–21 GPa) experiments at the University of Bristol. Piston cylinder experiments employed a NaCl-pyrex assembly with a straight graphite furnace and Al O inner parts. Temperature was measured using type D thermocouple wires contained in an alumina sleeve and positioned immediately adjacent to the Au Pd sample capsule that contained the powdered starting material. We assume that the temperature gradient across the entire capsule (<2 mm) was smaller than 20 °C (refs 33, 34). The hot piston-in technique was used with a friction correction of 3% applied to the theoretical oil pressure to achieve the desired run conditions35. Multi-anvil experiments were performed using Toshiba F-grade tungsten carbide cubes bearing 11, 8 or 4 mm truncated corners in combination with a pre-fabricated Cr-doped MgO octahedron of 18, 14 or 10 mm edge length, respectively. The relationship between oil-reservoir and sample pressure for each cell was calibrated at room and high temperature (1,200 °C) by detecting appropriate room temperature phase transitions of Bi, ZnTe and GaAs and bracketing transformations of SiO (quartz-coesite and coesite-stishovite), Mg SiO (α-β and β-γ) and CaGeO (garnet-perovskite). Calibrations are estimated to be accurate within ±1 GPa. In all experiments, desired run pressure was achieved using a slow, Eurotherm controlled, pressure ramp of ≤50 tonnes per hour. Experiments were heated after high pressure was reached with high temperatures generated using stepped graphite (18/11 cell) or straight LaCrO furnaces (14/8 and 10/4 cells) and monitored with type C thermocouple wires. Two 10/4 experiments, performed during a period of repeated LaCrO heater failures, used rolled 40-μm-thick Re furnaces. Temperature was quenched by turning off the furnace power before a slow decompression ramp (half the rate of experiment compression) to ambient conditions. Samples were contained in Au capsules unless temperatures exceeded its thermal stability, in which case Au Pd or Au Pd capsules were used. Run durations all exceeded 600 min and are reported in Extended Data Tables 2 and 3. Temperature uncertainties were believed to be less than ±20, 30 or 50 °C for 18/11, 14/8 and 10/4 cells respectively36, 37. Recovered samples were mounted longitudinally in epoxy, polished under oil and repeatedly re-impregnated with a low viscosity epoxy (Buelher EpoHeat) to preserve soft and water-soluble alkali carbonate components present in run products. Polished and carbon-coated run products were imaged in backscatter electron mode (BSE) using a Hitachi S-3500N scanning electron microscope (SEM) with an EDAX Genesis energy dispersive spectrometer to identify stable phases and observe product textures. Subsequently, wavelength dispersive spectroscopy (WDS) was performed using the Cameca SX100 Electron Microprobe or the Field Emission Gun Jeol JXA8530F Hyperprobe at the University of Bristol to achieve high-precision chemical analyses of run products. Analyses were performed using an accelerating voltage of 15 or 12 kV on the respective instruments, with a beam current of 10 nA. Calibrations were performed during each session using a range of natural mineral and metal standards and were verified by analysing secondary standards (as described previously6). Silicate phases were measured using a focused electron beam whereas carbonates and melts were analysed using an incident beam defocused up to a maximum size of 10 μm. Count times for Na and K were limited to 10 s on peak and 5 s on positive and negative background positions. Peak count times for other elements were 20–40 s. Additional analyses of the calcium perovskite phases grown during reaction experiments, measuring only SiO and MgO content, were made using the Jeol instrument at 5 kV and 10 nA to ensure reported MgO contents were not influenced by secondary fluorescence from surrounding material. The identity of experimental-produced minerals was determined using Raman spectroscopy as a fingerprint technique. Spectra were collected using a Thermo Scientific DXRxi Raman microscope equipped with an excitation laser of either 455 or 532 nm. Studies that investigate the alteration of oceanic crust have demonstrated that carbon incorporation does not simply occur by the addition of a single carbonate species to MORB9. It instead appears to occur by a complex amalgamation of hydrocarbon and graphite deposition related to hydrothermal fluxing above magma chambers at the mid-ocean ridge8 and underwater weathering9, 38, 39, 40 where seawater-derived CO reacts with leached crustal cations, often in veins. It is believed that the quantity of biotic organic carbon in the crustal assemblage is negligible compared with abiotic organic compounds and inorganic carbonates8. These processes result in a layered crustal assemblage that, in the uppermost few hundred metres can contain up to a maximum of 4 wt% CO in rare cases9, 39 but more commonly <2 wt% CO (refs 8, 9, 39). Beneath 500 m depth the carbon content drops to between 100 and 5,000 p.p.m. CO throughout the remainder of the 7-km-thick basaltic section8, and is mostly organic hydrocarbon species. The upper 300 m are regularly altered and can be generally thought to have compositions similar to the altered MORB rocks analysed previously41. Deeper portions of the MORB crust retain their pristine MORB compositions. It is therefore apparent that carbonated eclogite bulk compositions used in previous studies, where at least 4.4 wt% CO was added to an eclogite by addition of ~10 wt% carbonate minerals, may not be good analogues of naturally subducting crustal sections. The compositions of these starting materials from previous studies19, 42, 43, 44, 45, 46 can be found in Extended Data Table 1. We do not include the composition of the starting material used by refs 47 or 48 as these studies were conducted in simplified chemical systems so are not directly comparable with these natural system compositions. However, as some of the previous studies rightly identify and discuss, the composition of deeply subducted MORB is unlikely to be the same as that entering the subduction system. One process widely believed to alter the composition of downwelling MORB is sub-arc slab dehydration. Pressure (P)–temperature (T) paths of subducted slabs26 can be compared with experimental studies of hydrous, carbonated and H O-CO -bearing eclogite compositions12, 24, 42, 43, 49 and thermodynamic models11, 50 to conclude that slabs experience dehydration at sub-arc conditions (that is, 1–5 GPa) but will generally not reach high enough temperatures to undergo melting. Therefore, they will by and large retain their carbon components although some fraction may be lost by dissolution into aqueous fluids51, 52. It is believed that sub-arc dehydration is capable of removing SiO from the subducting assemblage, and previous carbonated MORB compositions were therefore designed to be considerably silica undersaturated (relative to fresh/altered MORB)19, 43, 44, 45. While studies53, 54, 55, 56 do indicate that SiO can become soluble in H O at high pressures, they infer that the solubility of silica in hydrous fluids only exceeds ~1 wt% at T > 900 °C at 1 GPa (higher T at higher P). In contrast, slab dehydration occurs on all prograde slab paths at T < 850 °C. Additionally, the composition of quenched hydrous fluids coexisting with MORB at 4 GPa and 800 °C (ref. 57) indicate that a maximum of ~12 wt% SiO can dissolve in the fluid. Given that there should be considerably less than 10 wt% H O (more likely << 5 wt% H O) in subducting assemblages, this suggests a maximum SiO loss in subducting MORB lithologies of ~0.6–1.2 wt%. The compositions used in previous studies have SiO depletions ranging from 3 wt% up to, more commonly, 6–10 wt% SiO relative to MORB. We further investigated the effect of oceanic crust alteration and sub-arc dehydration on the composition of subducted MORB rocks by compiling a data set of altered MORB41 and exhumed blueschist, greenschist and eclogite facies rocks from exhumed terrains worldwide to compare them with fresh MORB21, our starting material and previous starting materials. We then assess the relevance of our starting material based on the composition of natural MORB rocks, rather than using models of the subduction process that contain few observable constraints. Results of this comparison are plotted in Extended Data Fig. 1. This analysis confirms that relative to fresh MORB, altered MORB and exhumed crustal rocks are somewhat depleted in SiO , up to a maximum of 6 wt% SiO in the most extreme case, but more commonly 0–3 wt% SiO . Thus, many previous starting materials are too silica undersaturated to be good analogues of subducting MORB. Furthermore, this analysis reveals that altered and exhumed MORB are not enriched in CaO compared with fresh MORB, if anything they actually contain lower CaO on average. In contrast, all previous starting materials are enriched in CaO compared with fresh MORB. This is because most previous studies introduced the carbon component to their experiment by adding ~10 wt% calcite to an eclogite-base composition. We note that SLEC1 (ref. 43) was not created in this manner, but instead this composition falls far from the MORB field as the authors used an eclogite xenolith erupted by a Hawaiian volcano as a base material. By plotting the position of the maj–cpx join, defined by the composition of our experimental phases plotted in Extended Data Fig. 5, onto Extended Data Fig. 1a, we demonstrate that our bulk composition (ATCM1), ALL-MORB21, the vast majority of the fresh MORB field, altered41 and exhumed MORB samples fall on the CaO-poor side of this join, that is, on the Mg+Fe-rich side. Therefore, magnesite will be the stable carbonate phase in these compositions at high pressure (above dolomite breakdown). In contrast, all previous bulk compositions plot on the Ca-rich side of this join, or are very depleted in SiO , and therefore fall in a different phase field to the overwhelming majority of subducted MORB. This difference causes a considerable difference in the phase relations of our starting material relative to those used in previous studies. We acknowledge that no single bulk composition can be a perfect analogue for the entire range of subducting MORB compositions, however, ATCM1 is a good proxy for sections of the MORB crust between ~300 m and 7 km depth that have unaltered major element compositions and low CO contents. Additionally, ATCM1 remains a better analogue for the uppermost portions of the MORB crust than starting materials employed in previous studies because its CO content is within the range of natural rocks while it is also not oversaturated in CaO or over depleted in SiO . This is despite it falling towards the SiO -rich end of the compositional spectrum of subducting MORB rocks. Recent experiments have suggested that carbonate in eclogitic assemblages may be reduced to elemental carbon, either graphite or diamond, at depths shallower than 250 km (ref. 58). However, subducting slab geotherms are much colder than the experimental conditions investigated by this study, and additionally they are believed to contain considerable ferric iron that is further increased during de-serpentinization10. Indeed, several observations of carbonate inclusions in sub-lithospheric diamonds6, 7, 59 require that slab carbon remains oxidized and mobile until diamond formation, far deeper than 250 km. Given the numerous observations from natural diamond samples, the general uncertainty in the mantle’s fO structure and the lack of any conclusive experimental evidence that subducting carbon becomes reduced before reaching the transition zone we posit that nearly all subducting carbon is stable as carbonate throughout the upper mantle in subducting MORB assemblages. Extended Data Table 2 presents the run conditions, durations and phase proportions in all carbonated MORB melting experiments, which are also summarized in Extended Data Fig. 2. Phase and melt compositions are presented in the Supplementary Tables 1–4. Phase proportions are calculated by mass balance calculations that use the mean composition of each phase as well as the reported 1σ uncertainty in this mean as inputs. We note that the 1σ uncertainty for some oxides in garnet and clinopyroxene minerals occasionally exceeds 1 wt%, although it is normally much smaller than this. These large uncertainties are a function of the small crystal sizes present in some runs, and not a function of sluggish reaction kinetics. Phase proportion calculations were run in a Monte Carlo loop of 10,000 calculation cycles where a varying random error was added to each oxide in each mineral phase during each iteration. Overall the distribution of varying random errors for each oxide form a Gaussian distribution with standard deviation equal to the reported 1σ uncertainty of measurements. The reported proportions are the numerical mean of all calculation cycles and the r2 value reports the average squared sum of residuals. Low r2 values indicate that chemical equilibrium is likely to have been achieved and that mineral and melt compositions have been accurately determined. Representative BSE images of the polished experiments are shown in Extended Data Fig. 3. Garnets in experiments at all pressures contain abundant SiO inclusions. In subsolidus experiments the number of inclusions increases and the definition of mineral boundaries deteriorates, which makes accurate analysis of garnet compositions increasingly challenging. In supersolidus runs, garnet minerals adjacent, or near to, carbonatite melt pools have well defined edges and contain fewer inclusions. However, far from quenched melts the textures of garnets remain small and pervasively filled with inclusions, indicating the influence of melt fluxing on mineral growth. With increasing pressure, garnets become increasingly majoritic, with increasing quantities of octahedral silicon. Clinopyroxene was observed in all subsolidus experiments, as euhedral crystals that are often spatially associated with the carbon-bearing phase. Cpx abundance falls with increasing pressure and their compositions becoming increasingly dominated by sodic components (jadeite, aegerine and NaMg Si O ) at high pressure (Extended Data Fig. 5). Cpx only disappears from the stable phase assemblage in supersolidus experiments at 20.7 GPa. SiO is observed in all runs and are small, often elongated tabular-shaped crystals. An oxide, either TiO at low pressure or an Fe-Ti oxide above 13 GPa (as described previously24) are observed in all subsolidus runs. The carbon-bearing phase in subsolidus experiments changes with increasing pressure. At 3 GPa CO , marked by the presence of voids in the polished sample, is stable. This converts to dolomite at 7.9 GPa, consistent with the position of the reaction 2cs + dol = cpx + CO (ref. 22). Beyond ~9 GPa dolomite becomes unstable and breaks down into magnesite + aragonite23. Therefore, because the ATCM1 bulk composition lies on the Mg+Fe2+-rich side of the garnet–cpx join (Extended Data Figs 1a and 5), magnesite replaces dolomite as the carbon host in the experimental phase assemblage. This differs from experiments in previous studies, where aragonite was dominant because bulk compositions fall on the opposite side of the garnet–cpx join. It is clear from the ternary diagrams (Extended Data Fig. 5) that while the tie-line between garnet and cpx remains, magnesite and aragonite cannot coexist in a MORB bulk composition. Finally, at pressures above 15 GPa, Na-carbonate becomes stable in the subsolidus phase assemblage. This is chemographically explained by the rotation of the garnet–cpx tie-line with increasing pressure (EDF5). Its appearance can also be justified as a necessary host of sodium at increasing pressure, since aside from clinopyroxene there is no other Na-rich phase stable on the Mg+Fe side of the maj–cpx join. The appearance of silicate melt, containing dissolved CO (estimated by difference), defines the solidus at 3 GPa. This may initially appear to contradict the results of some previous studies, which find carbonatite melts are produced near the solidus of carbonated eclogite at pressures lower than 7 GPa (refs 43, 45, 46). However, this is easily explained by the differences in CO and SiO content used in these studies. The higher CO and lower SiO contents of previous studies stabilize carbonate melt to lower temperatures relative to silicate melts. Indeed, we note that our results are consistent with those described previously42, 44 (the two previous studies with the least depleted SiO ), which also observed that near-solidus melts below 5 GPa were basaltic to dacitic silicate melts containing dissolved CO . The results of one paper19 are not entirely self-consistent, in that at some pressures between 3.5 and 5.5 GPa the authors observed silicate melts before carbonate melts (4.5 and 5 GPa), whereas this relationship is sometimes reversed (5 GPa in AuPd capsules) or both melts were observed together (3.5 GPa). The observation of two immiscible melts in previous studies probably reflects the maximum CO solubility in silicate melts. Since our bulk composition has less CO , akin to natural rocks, we do not observe liquid immiscibility. In all experiments above 7 GPa, near-solidus melt compositions are carbonatititc and essentially silica-free. This result is notably different from those described previously19, which reported that near-solidus melts were a mixture of silicate, carbonated silicate and carbonatite melts. We believe this contrast is caused by the interpretation of experimental run textures. Whereas ref. 19 identified regions of fine-grained material consisting of mixtures of stable phases from elsewhere in the capsule as quenched melts, we have not followed the same interpretation of these features. Although we do recognize similar features in some run products, we have interpreted these features as a consequence of poor crystal growth in regions far from the influence of melt fluxing. In all supersolidus experiments, we observed regions of carbonatite material (typically <1 wt% SiO ) that is fully segregated from surrounding silicate minerals and possesses a typical carbonate-melt quench texture (Extended Data Fig. 3). Silicate minerals in close proximity to these melt pools are larger than those elsewhere in the same experiment, have well-defined crystal boundaries and contain few inclusions. Therefore, we attribute the variable texture and regions of fine-grained material present in experiments to the location of melt within experiments, which has a tendency to segregate to isolated regions of capsules under influence of temperature gradients. Although melt segregation occurs in all supersolidus experiments, the efficiency of segregation and size of melt pools considerably increases with rising temperature above the solidus. Extended Data Figure 4 shows the highly systematic evolution of the melt compositions reported from our study with increasing pressure, strongly supporting our interpretations. Carbonatite melts are calcic, Ca number > 0.5 (Ca number = Ca/[Ca+Mg+Fe]), despite subsolidus carbonates being dominated by magnesite (Extended Data Fig. 4). Melts have high concentrations of TiO (typically 1–3.5 wt%), P O (0.4–1.5 wt%) and K O (0.3–1.5 wt%) and a variable Mg number (0.33–0.7 defined as Mg/[Mg+Fe]). The alkali content of melts, strongly dominated by Na O due to the bulk composition, increases with pressure (from 1 to ~15 wt% Na O at 7.9 and 20.7 GPa respectively; Extended Data Fig. 4). This increasing Na O content is driven by the decreasing compatibility of Na O in the residual mantle phase assemblages as the abundance of stable clinopyroxene falls. At 20.7 GPa the melt composition, as evidenced both by constant phase proportions and consistent melt/majorite compositions, remains constant over a temperature interval of ~350 °C above the solidus. It is only when temperature reaches 1,530–1,600 °C (runs #16 and #31) that the silica content of the melt begins to increase (to 8.7 wt%) and CO content falls as melts start to become silica-carbonatites. One experiment (#33) aimed to verify that measured low-degree melt compositions are accurate, and are not affected by analytical problems related to the small size of melt pools, was conducted at 20.7 GPa. In this experiment the abundance of carbonate melt was increased by adding a mix replicating the low degree melt composition ATCM2 to ATCM1 in a mass ratio of 1:3. If the composition of low-degree melts has been accurately determined in ‘normal’ experiments then this addition will have a negligible effect on phase relations or the compositions of the garnet, SiO or melt; it would simply increase the melt abundance. The result of this experiment has a similar texture to all other experiments, where carbonatite melt segregates to one end of the capsule and is adjacent to large, well-formed majoritic garnets. The far end of the capsule has a much smaller crystal size, crystals have ragged edges, garnets are full of inclusions and SiO is present along grain-boundaries and triple junctions (Extended Data Fig. 3h). Mineral and melt compositions, although not exactly identical, are similar to those measured in ‘normal’ experiments (to achieve identical compositions an iterative approach would be required that was not deemed to be necessary) thus confirming that near-solidus melt compositions have been accurately determined. The presence of fine-grained material away from segregated melt also acts to further confirm our hypothesis regarding the vital importance of melt presence for growing large crystals during experiments. Comparing our starting material and results with those of previous studies using ternary and quaternary projections (Extended Data Fig. 5) reveals that it is not possible for both magnesite and aragonite to coexist alongside majorite and clinopyroxene owing to stable mineral phase fields (see earlier). Thus, in Mg-Fe-dominated compositions, such as our starting material, magnesite is the stable carbonate at high-pressure subsolidus conditions. Whereas in Ca-dominated compositions aragonite will be the stable carbonate beyond the pressure of dolomite dissociation. Natural subducting MORB compositions, which contain, at most, a similar quantity of CO to our bulk composition11, almost all lie on the Ca-poor side of the majorite–clinopyroxene join (Extended Data Figs 1 and 5). In this situation, as our experiments demonstrate, cpx remains an important Na-host in MORB assemblages to high pressures alongside [Na,K] Ca CO structured carbonate. Ca-rich compositions containing subsolidus CaCO experience different phase relations because aragonite can dissolve considerable Na O and so is the sole Na-host in these compositions. We conclude that because the majority of natural MORB rocks fall on the Mg+Fe side of the maj–cpx join, like our bulk composition, that the phase relations determined in this study are applicable to the case of natural subduction. Therefore, the melting point depression we observe along the carbonated MORB solidus at uppermost transition zone pressures is generally applicable to subducted oceanic crust. Without the influence of slab-derived melts, the anhydrous transition zone peridotite assemblage at 20.7 GPa and 1,600 °C (experiment G168 and G176) is dominated by Na-poor majorite and wadsleyite (Mg number = 0.90) (Extended Data Fig. 6, Extended Data Table 3 and Supplementary Table 5a). Upon reaction with the near-solidus alkaline carbonatite defined during melting experiments, ATCM2, a clearly defined reaction zone is observed between this ambient peridotite assemblage and the infiltrating melt (Extended Data Fig. 6). The products of this reaction are garnet containing a notable Na X2+Si O majorite component, Ca(Si,Ti)O perovskite, ringwoodite, ferropericlase and diamond. All of these phases were identified using Raman spectroscopy (Extended Data Fig. 7) and their compositions are presented in Supplementary Table 5a. Raman spectroscopy alone, which was performed before any sample polishing using diamond-based products, confirms the creation of diamond during these reactions. We have not observed diamond using SEM techniques and believe that it resides as sub-micrometre-sized inclusions in the various reaction-product minerals where it is seen by spectroscopic methods. The experiments performed on intimately mixed powders of melt and pyrolite also form the same phase assemblages (Extended Data Table 3) and mineral compositions from those runs are also presented in Supplementary Table 5b, c. We observed the reaction products as new crystals floating in the residual carbonatite melt and/or nucleated on the relics of the peridotite assemblage, thus creating zoned minerals. We have demonstrated that the composition of majorite minerals crystallizing during the reactions lie between those expected for peridotitic and eclogitic minerals at a similar pressure and possibly explain intermediate-composition diamond-hosted majorites (Fig. 2). We suggest that the full range of intermediate inclusion compositions might be created by the gradual shift in phase compositions, from those we observe towards more peridotitic minerals as the melt composition reacts with increasing quantities of mantle material. Additionally we have shown that the compositions of calcium perovskite (Extended Data Fig. 8) and ferropericlase (Fig. 3) formed during the reactions are consistent with diamond-hosted minerals of those species. Further experiments, across the solidus ledge and into the uppermost lower mantle pressure range are required to test whether melt–mantle interactions account for all diamond-hosted inclusions.
Nowell M.M.,EDAX |
Wright S.I.,EDAX |
Scarpulla M.A.,University of Utah |
Compaan A.D.,University of Toledo |
And 3 more authors.
Proceedings of the International Symposium on the Physical and Failure Analysis of Integrated Circuits, IPFA | Year: 2012
It is assumed that the performance of polycrystalline CdTe photovoltaic thin films is correlated to microstructure, especially aspects of the microstructure related to grain boundaries. However the role of grain boundaries on the electrical performance of these films is not well understood. In an effort to gain understanding into the correlation between the performance and grain boundaries in CdTe thin films, electron backscatter diffraction (EBSD) has been used to characterize the the grain size, grain boundary structure and texture of sputtered CdTe at varying deposition pressures before and after CdCl 2 treatment. Weak axisymmetric (or fiber) textures were observed in the as-deposited films. At lower deposition pressures (111) fiber textures were observed and (110) fiber textures were observed at higher lower deposition pressures. Samples treated with CdCl 2 exhibited significant grain growth with a high fraction of twin boundaries. A much stronger correlation of performance with grain size was found when the grain size was corrected to exclude the twin boundaries. This observation confirms that the twin boundaries do indeed have different electrical properties than random high-angle boundaries. © 2012 IEEE.