# Fileset

[s41526-024-00371-x.pdf](https://mdr.nims.go.jp/filesets/ce0c4edf-6883-42e9-8222-4cc579bc09a5/download)

## Creator

[Stephen K. Wilke](https://orcid.org/0000-0003-4674-7049), Abdulrahman Al-Rubkhi, Chihiro Koyama, [Takehiko Ishikawa](https://orcid.org/0000-0003-0769-3869), Hirohisa Oda, Brian Topper, [Elizabeth M. Tsekrekas](https://orcid.org/0000-0001-9417-8827), [Doris Möncke](https://orcid.org/0000-0002-4197-5520), [Oliver L. G. Alderman](https://orcid.org/0000-0002-2342-811X), Vrishank Menon, Jared Rafferty, Emma Clark, Alan L. Kastengren, [Chris J. Benmore](https://orcid.org/0000-0001-7007-7749), [Jan Ilavsky](https://orcid.org/0000-0003-1982-8900), [Jörg Neuefeind](https://orcid.org/0000-0002-0563-1544), [Shinji Kohara](https://orcid.org/0000-0001-9596-2680), [Michael SanSoucie](https://orcid.org/0000-0002-3575-2275), Brandon Phillips, [Richard Weber](https://orcid.org/0000-0002-2145-1279)

## Rights

[Creative Commons BY Attribution 4.0 International](https://creativecommons.org/licenses/by/4.0/)

## Other metadata

[Microgravity effects on nonequilibrium melt processing of neodymium titanate: thermophysical properties, atomic structure, glass formation and crystallization](https://mdr.nims.go.jp/datasets/4eaebfc7-3d34-4059-99a0-042163846d28)

## Fulltext

Microgravity effects on nonequilibrium melt processing of neodymium titanate: thermophysical properties, atomic structure, glass formation and crystallizationnpj | microgravity ArticlePublished in cooperation with the Biodesign Institute at Arizona State University, with the support of NASAhttps://doi.org/10.1038/s41526-024-00371-xMicrogravity effects on nonequilibriummelt processing of neodymium titanate:thermophysical properties, atomicstructure, glass formation andcrystallizationCheck for updatesStephen K. Wilke 1,2 , Abdulrahman Al-Rubkhi1, Chihiro Koyama3, Takehiko Ishikawa 3,Hirohisa Oda3, Brian Topper4, Elizabeth M. Tsekrekas 5, Doris Möncke 5, Oliver L. G. Alderman 6,Vrishank Menon1, Jared Rafferty1, Emma Clark1, Alan L. Kastengren2, Chris J. Benmore 2,Jan Ilavsky 2, Jörg Neuefeind 7, Shinji Kohara 8, Michael SanSoucie 9, Brandon Phillips9 &Richard Weber 1,2The relationships between materials processing and structure can vary between terrestrial andreduced gravity environments. As one case study, we compare the nonequilibriummelt processing ofa rare-earth titanate, nominally 83TiO2-17Nd2O3, and the structure of its glassy and crystallineproducts. Density and thermal expansion for the liquid, supercooled liquid, and glass are measuredover 300–1850 °C using the Electrostatic Levitation Furnace (ELF) in microgravity, and two replicatedensitymeasurementswere reproducible towithin 0.4%.Cooling rates in ELF are 40–110 °C s−1 lowerthan those in a terrestrial aerodynamic levitator due to the absence of forced convection. X-ray/neutron total scattering and Raman spectroscopy indicate that glasses processed on Earth and inmicrogravity exhibit similar atomic structures, with only subtle differences that are consistent withcompositional variations of ~2mol. %Nd2O3. The glass atomic network contains amixture of corner-and edge-sharing Ti-O polyhedra, and the fraction of edge-sharing arrangements decreases withincreasing Nd2O3 content. X-ray tomography and electron microscopy of crystalline products revealsubstantial differences in microstructure, grain size, and crystalline phases, which arise fromdifferences in the melt processes.Manufacturing processes suitable to reduced gravity and microgravityenvironments must be developed to advance the low Earth orbit (LEO)economy, utilize resources from locations beyond Earth, and actualize self-sustainable extraterrestrial habitats with closed-loop materials recycling1.Melt processing is one common manufacturing step with particularlychallenging differences between terrestrial andmicrogravity conditions. Formicrogravity, Earth-based melt processes must be adapted to the lack ofbuoyancy-driven convection, density-induced sedimentation, and gravita-tional forces to direct fluid flow.Additive manufacturing techniques like selective laser melting aresome examples of melt processing that may be suitable for reduced gravityenvironments1–4. In such processing techniques, the product’s shape,1Materials Development, Inc., Evanston, IL 60202, USA. 2X-ray Science Division, Advanced Photon Source, Argonne National Laboratory, Lemont, IL 60439, USA.3JapanAerospaceExplorationAgency, Tsukuba, Japan. 4Center forHigh TechnologyMaterials, University of NewMexico, Albuquerque, NM87106,USA. 5InamoriSchool of Engineering at the New York State College of Ceramics, Alfred University, Alfred, NY 14802, USA. 6ISIS Neutron & Muon Source, Rutherford AppletonLaboratory, , ChiltonDidcot, Oxon, OX11 0QX, UK. 7Neutron Science Division, Spallation Neutron Source, Oak Ridge National Laboratory, Oak Ridge, TN 37831,USA. 8National Institute for Materials Science, Tsukuba, Japan. 9NASAMarshall Space Flight Center, Huntsville, AL 35812, USA. e-mail: swilke@matsdev.comnpj Microgravity |           (2024) 10:26 11234567890():,;1234567890():,;http://crossmark.crossref.org/dialog/?doi=10.1038/s41526-024-00371-x&domain=pdfhttp://crossmark.crossref.org/dialog/?doi=10.1038/s41526-024-00371-x&domain=pdfhttp://crossmark.crossref.org/dialog/?doi=10.1038/s41526-024-00371-x&domain=pdfhttp://orcid.org/0000-0003-4674-7049http://orcid.org/0000-0003-4674-7049http://orcid.org/0000-0003-4674-7049http://orcid.org/0000-0003-4674-7049http://orcid.org/0000-0003-4674-7049http://orcid.org/0000-0003-0769-3869http://orcid.org/0000-0003-0769-3869http://orcid.org/0000-0003-0769-3869http://orcid.org/0000-0003-0769-3869http://orcid.org/0000-0003-0769-3869http://orcid.org/0000-0001-9417-8827http://orcid.org/0000-0001-9417-8827http://orcid.org/0000-0001-9417-8827http://orcid.org/0000-0001-9417-8827http://orcid.org/0000-0001-9417-8827http://orcid.org/0000-0002-4197-5520http://orcid.org/0000-0002-4197-5520http://orcid.org/0000-0002-4197-5520http://orcid.org/0000-0002-4197-5520http://orcid.org/0000-0002-4197-5520http://orcid.org/0000-0002-2342-811Xhttp://orcid.org/0000-0002-2342-811Xhttp://orcid.org/0000-0002-2342-811Xhttp://orcid.org/0000-0002-2342-811Xhttp://orcid.org/0000-0002-2342-811Xhttp://orcid.org/0000-0001-7007-7749http://orcid.org/0000-0001-7007-7749http://orcid.org/0000-0001-7007-7749http://orcid.org/0000-0001-7007-7749http://orcid.org/0000-0001-7007-7749http://orcid.org/0000-0003-1982-8900http://orcid.org/0000-0003-1982-8900http://orcid.org/0000-0003-1982-8900http://orcid.org/0000-0003-1982-8900http://orcid.org/0000-0003-1982-8900http://orcid.org/0000-0002-0563-1544http://orcid.org/0000-0002-0563-1544http://orcid.org/0000-0002-0563-1544http://orcid.org/0000-0002-0563-1544http://orcid.org/0000-0002-0563-1544http://orcid.org/0000-0001-9596-2680http://orcid.org/0000-0001-9596-2680http://orcid.org/0000-0001-9596-2680http://orcid.org/0000-0001-9596-2680http://orcid.org/0000-0001-9596-2680http://orcid.org/0000-0002-3575-2275http://orcid.org/0000-0002-3575-2275http://orcid.org/0000-0002-3575-2275http://orcid.org/0000-0002-3575-2275http://orcid.org/0000-0002-3575-2275http://orcid.org/0000-0002-2145-1279http://orcid.org/0000-0002-2145-1279http://orcid.org/0000-0002-2145-1279http://orcid.org/0000-0002-2145-1279http://orcid.org/0000-0002-2145-1279mailto:swilke@matsdev.comchemical phase, and material microstructure are all influenced by theinterplay of energy deposition and the melt’s thermophysical properties,such as viscosity and surface tension5–7. These processing-structure-property relationships illustrate that accurate measurements of liquidthermophysical properties are crucial to enabling space-based manu-facturing. Additionally, melt processing often concludes with solidificationfrom nonequilibrium states, such as supercooled liquids. Because fluid flowand heat transfer differ between Earth and microgravity environments,different states of nonequilibrium will likely be encountered that providenew challenges and new opportunities for materials discovery and manu-facturing technology. Such opportunities may manifest through leveragingof curated large datasets for low-level machine learning and algorithmdevelopment, also known as Digital Twin interests. This would drive per-formance and design improvements that would have both terrestrial andspace-based benefits in the form of newer materials, processes, andefficiencies.Historically, microgravity research on melt processing and liquidthermophysical properties has focused on metallic materials, which havebeen studied in theTEMPUS8,9 campaign and theElectromagnetic LevitatorFacility10 onboard the International Space Station (ISS). Until recently, lessattention has been given to ceramic and glass materials. One notableexception is ZBLAN glass (ZrF4-BaF2-LaF3-AlF3-NaF). ZBLAN is a can-didate material for improved, low-loss optical fiber communicationsbecause it has a theoretical optical attenuation one to two orders of mag-nitude lower than that of silica11.However, ZBLAN is difficult to process dueto its sensitivity towater and small working temperature range, i.e. the rangebetween its glass transition and crystallization point. Early microgravityexperiments on ZBLAN suggested that crystallizationmay be suppressed inreduced gravity12, thereby widening the working temperature range andhypothetically making it possible to manufacture higher-quality ZBLANfiber. The increase in crystallization temperature was speculated to arisefrom a lack of buoyancy-driven convection13, though a full mechanisticexplanation remains lacking. Nonetheless, multiple commercial efforts arecurrently pursuing space-based processing of ZBLAN fibers14,15.Microgravity research on oxide ceramics, glasses, and their manu-facturing remains in its infancy. In 2016, the Japanese Aerospace Explora-tion Agency commissioned the Electrostatic Levitation Furnace (ELF)16onboard the International Space Station, which is well suited for high-temperaturemelt processing of oxides. The ELF uses three orthogonal pairsof electrostatic transducers to position and levitate spheroidal samples~2mm in diameter. During levitation, samples can be heated with four980 nm lasers, achieving melts up to ~3070 °C17. Silhouette imaging andoptical pyrometry enable volume measurements of molten samples as afunction of temperature, from which the sample density and thermalexpansion coefficient can be calculated based on its preflight and/or post-flight mass18. Liquid viscosity and surface tension can also be measuredusing a droplet oscillation technique19. By applying a sinusoidal variation tothe transducers’ voltages along one axis, oscillations in the liquid can beinduced. If the excitation frequency matches the natural resonance of thedroplet, analytical expressions can relate the resonance frequency andcharacteristic time for oscillation damping (upon stopping the excitation) tothe liquid’s surface tension and viscosity, respectively20. These relationshipsdepend on assumptions of fluid quiescence, which is why microgravityprovides a unique advantage formeasuring these thermophysical propertiesat high temperatures. The ELF can operate at up to 2 bar of gas pressure andcan utilize oxygen-containing gases, which are desirable for oxide studies.Using the ELF, we are reporting a comprehensive comparison of meltprocessing on Earth versus in microgravity for one particular oxide, aneodymium titanate (NT) with nominal composition 83TiO2–17Nd2O3.The NT composition is of interest technologically as an optical material: itsglass has a high refractive index (n > 2.1)21,22 and a wide transmission win-dow spanning the visible and infrared range out to ~5 μm21,23. NT is also offundamental scientific interest as a fragile liquid24 and poor glass-former25,with an atomic network of octahedrally coordinated Ti-O polyhedra26atypical of oxide glasses. Here, we compare the cooling rates, glassformation, and recalescence ofmolten and supercooledNTprocessed in theELF and in a terrestrial aerodynamic levitator. Density is reported as afunction of temperature for the equilibrium liquid (ca. 1470–1850 °C),supercooled liquid (ca. 780–1470 °C), and glass (300–780 °C). The glassatomic structure is assessed with X-ray and neutron total scattering andRaman spectroscopy.Microstructures in both glassy and crystalline samplesare probed with X-raymicrotomography, X-ray small-angle scattering, andelectronmicroscopy. This suite of characterizations aims to benchmark themicrogravity melt processing technique against established terrestrialmethods, while identifying the differences and advantages offered by space-based melt processing.ResultsMelt processing and glass formationSamples of 83TiO2–17Nd2O3 nominal composition (83mol. % TiO2;17mol. % Nd2O3) were prepared as spheroids ~2mm in diameter(20–23mg), and these samples were heated, melted, and quenched usingtwo different containerless processing techniques. In microgravity, sampleswere levitated in the ELF16 using electrostatic forces and heated with four980 nm lasers positioned around the sample like corners of a tetrahedron,with the sample in the center. On Earth, samples were processed in anaerodynamic levitator using a gas stream passing through a converging-diverging nozzle, and the top of the sample was heated with a 10.6 μmCO2laser27. In both techniques, the sampleswere held isothermally near 1850 °C,fully molten (Tm = 1467 °C28), for at least 30 s before turning off the heatinglasers. The samples then cooled freely and, depending on the cooling rate,either vitrified into glass or crystallized from a supercooled state. The con-tainerless processing techniques provide two major benefits for melt pro-cessing research: (i) avoiding chemical contamination from sample-cruciblereactions at high temperatures, and (ii) removing any source of hetero-geneous nucleation, so that supercooling and vitrification are easier atmodest cooling rates (i.e., <103 °C s−1) than in the presence of a container.Electrostatic levitation has been applied to liquids with a wide range ofsurface tensions, from propylene carbonate29 (43mNm−1) to tungsten30(~2500mNm−1).Exemplary cooling curves for samples inmicrogravity (“MG1”) andonEarth (“TG2”) are compared in Fig. 1a. Both measurements used opticalpyrometers with λ ~ 1.5 μm, but the range of the terrestrial pyrometer waslimited to a maximum of 1650 °C. (Terrestrial sample temperature above1650 °C was monitored with a pyrometer of different wavelength). Tem-perature measurement uncertainty is ca. ±30 °C for ELF and ±27 °C in theaerodynamic levitator (see Supplementary Discussion). The terrestrial andmicrogravity samples cooled from ~1850 to 300 °C in ~8.9 and20.6 seconds, respectively. Figure 1b shows the corresponding cooling ratesas a function of temperature. Comparing the two processing environmentsprovides insight into the relative contributions of different heat transfermodes. At high temperatures, radiative heat transfer (~T4) is dominant,while at the lowest temperatures, the samples cool due to convection and/orconduction. In microgravity, buoyancy-driven convection is minimal, sothe cooling rate of 20–60 °C s−1 observed below 800 °C is likely due to heatconduction through the gas surrounding the sample. For any given tem-perature, the terrestrial cooling rate is 40–110 °C s−1 larger than in micro-gravity. This difference in cooling rate represents the contribution fromforced convective cooling by the levitation gas around the sample.Both the terrestrial and microgravity cooling rates in Fig. 1b exhibitslope fluctuations at several temperatures. The most reproducible fluctua-tion is a transient increase in cooling rate ca. 570–740 °C, which was alwaysobserved in repeated terrestrial measurements and occurs just below theglass transition at Tg = 786 °C21. This fluctuation is most likely an artifactcaused by a combination of the pyrometer measurement and NT opticalproperties. NT glasses exhibit several absorption bands in the visible andnear-infrared range due to Nd3+ optical activity21; the sample transparencyand fluorescence near the pyrometer wavelength may change with tem-perature. Two auxiliary tests in the aerodynamic levitatorwere conducted totest this hypothesis. First, NT cooling rates were measured with threehttps://doi.org/10.1038/s41526-024-00371-x Articlenpj Microgravity |           (2024) 10:26 2different pyrometer wavelengths—λ = 0.9, 1.5, and 5.0 μm (SupplementaryFig. 1a, b)—and the magnitude of the cooling rate fluctuation varied sig-nificantly, appearingmost strongly for λ = 0.9. Second, cooling curves for anoptically inactive lanthanum titanate sample (83TiO2-17La2O3) did notexhibit this cooling rate fluctuation (Supplementary Fig. 1c, d). Theseobservations support the explanation that NT’s apparent fluctuation ca.570–740 °C is a measurement artifact. In Fig. 1b, the cooling rate alsoexhibitsmore subtle slopefluctuations at other temperatures, and thesemaybe due to slow sample rotation observed in both levitation techniques. Thesample surface is known to have spatial temperature variations (e.g., ~30 °Cfor liquids at 1500 °C in the aerodynamic levitator31), so sample rotationcould result in apparent temperature fluctuations where the pyrometer isfocused.In addition to the glass samples obtained by the cooling trajectories inFig. 1, crystalline samples were also produced by melt quenching. Crystal-lization typically occurred from the supercooled liquid at temperatures of900–1150 °C, evidenced by a visual recalescence and a momentary increasein the sample temperature of tens to hundreds of °C. One sample processedin microgravity was later found to contain trace amounts of crystalsthroughout the material (i.e., a glass-ceramic), despite its cooling curve notexhibiting recalescence.DensityThe densities of two NT samples melt quenched and vitrified in micro-gravity are shown in Fig. 2. These density values were calculated using thesamplemass aftermelt processing (i.e., after the sample return to Earth) andthe volume deduced from silhouette backlight imaging of the levitatedsamples during cooling18. Samples’ post-processing masses were 0.3–1.0%lower than their initial masses, likely due to volatilization. X-ray tomo-graphyof the recovered samples revealed internal porosity of up to0.25%, sothe sample volumes were corrected for this porosity before calculating thedensity shown in Fig. 2. This correction is based on porosity observed in theglassy samples, so it is possible that the internal pore sizes were different inthe molten state when volume measurements were collected.The density vs. temperature curves in Fig. 2 agree within 0.4% for thereplicate samples MG1 and MG2. Liquid density increases from 4.54 to4.92 g cm−3 during cooling from 1860 to 1030 °C, and the trend with tem-perature is described well by a linear fit:ρ Tð Þ ¼ 5:37 1ð Þ � 4:40ð5Þ× 10�4 ×T ;MG1; 1030 <T < 1860 ð1Þρ Tð Þ ¼ 5:38 1ð Þ � 4:43ð6Þ× 10�4 ×T ;MG2; 1030 <T < 1740 ð2ÞIn these equations, ρ is in units of g cm−3, T is in units of °C, and thevalues inparentheses indicate a 95%confidence interval. Themeanvaluesofthe linear equations for MG1 and MG2 are shown in Fig. 2 with a blackdashed line. The slope value of−dρ⁄dT ~ 4.4 × 10−4g cm−3 °C−1 for themeltis in reasonable agreementwith previously reported values for other titanatemelts32,33: BaTiO3 with 3.4 × 10−4, and BaTi2O5 with 4 × 10−4.For the glass, density increases from 5.15 to 5.22 during cooling from670 to 300 °C, with a trendwell described by a linearfit. However, this resultshould be accepted with some caution, since glass transparency at thepyrometer wavelength introduces temperature uncertainties that are diffi-cult to quantify.ρ Tð Þ ¼ 5:27 1ð Þ � 1:86 5ð Þ× 10�4 ×T ;MG1; 300 <T < 670 ð3Þρ Tð Þ ¼ 5:28 1ð Þ � 1:76ð4Þ× 10�4 ×T ;MG2; 300 <T < 670 ð4ÞIt is emphasized that the melt and supercooled liquid measurementsare the most valuable result that the ELF instrument enables; glass densityand its temperaturedependence canbemore easily andaccuratelymeasuredby other ground-based techniques.Extrapolation of the glass density to room temperature yields a value of5.28 g cm−3, which is larger than expected from prior measurements22(5.13 g cm−3, see Supplementary Discussion) and the compositionallysimilar crystalline phase (5.18 g cm−3 for Nd4Ti9O2434). However, this dis-crepancy of 2.9% is close to the relative uncertainty of 2.5% previouslyreported for ELF19.More puzzling is the observation ca. 720–930 °C of a regionwith largerthermal expansion than the glass or the liquid. The steeper slope in Fig. 2 forthis region generallywould suggest someheat release by the sample, possiblyby crystallization, but synchrotron X-ray scattering showed no sign ofcrystallinity. The process of melt quenching and vitrification is expected toshow a density-temperature plot with two linear regions, one each for glassand liquid, so this third anomalous region encompassing Tg representseither a measurement artifact or a new, unexpected behavior. Severalhypotheses for measurement artifacts were explored (see SupplementaryDiscussion). One explanation involves the sample becoming partiallytransparent at the pyrometer wavelength during cooling. If the transition topartial transparency began near 930 °C, then the pyrometer would startseeing some of the radiation emitted by the sample’s hotter interior. Thiswould result in a higher reading than the surface temperature, until thesample interior had also cooled enough to become partially transparent.This could explain the steeper slope inFig. 2 between720 and930 °C, yet it isnot a fully convincing argument. If instead the data are accurate and thisanomalous region is not due to a measurement artifact, the unexpecteddensity-temperature relationship could be caused by some kind of liquid-liquid phase transition or other thermodynamic anomaly in this fragileliquid approaching Tg35.Fig. 1 | Cooling of melt quenched neodymium titanate (NT) in air. a Sampletemperature vs. time after heating lasers were turned off, for melt processing inmicrogravity (MG1) and terrestrially (TG2). All temperatures are corrected forsample emissivity and window reflections. Both samples were cooled from ca.1850 °C. b Cooling rate as a function of temperature.https://doi.org/10.1038/s41526-024-00371-x Articlenpj Microgravity |           (2024) 10:26 3Density measurements are challenging for fragile liquids and theirglasses because melt processing typically requires containerless conditionswith small sample volumes. Glass density is typically measured using pyc-nometry of samples 2–3mm in diameter, as was done by Arai et al.22 onlanthanum titanate glasses, though this technique can have significantmeasurement uncertainty for small sample sizes26. Melt densities have beenmeasured using backlight imaging techniques in aerodynamic levitation,but partial occlusion of the sample and asphericity introduce an uncertaintyof ~5%33. Improved precision with backlight imaging is possible usingcombined aerodynamic and acoustic levitation techniques36, which provideoptical access to the entire sample.Given this context, electrostatic levitationand microgravity processing are useful tools for improving the precision ofdensity measurements on fragile liquids.Glass atomic structureThe atomic structures in terrestrial and microgravity glass samples wereprobed using high-energy X-ray diffraction, neutron diffraction, andRaman spectroscopy. Although gravity is not expected to directly influenceatomic arrangements, these characterizations remain important to bench-mark the similarities and differences between terrestrial and microgravityprocessing.We recently reported a thorough structural analysis of terrestrial NTglasses26, based on X-ray/neutron diffraction and structural modeling, soonly brief comments are provided here on the interpretation of the scat-tering data, focusing specifically on differences between microgravity andterrestrial samples. The glass atomic network has been shown to compriseTiO5 and TiO6 polyhedra connected primarily via corner-sharingarrangements, with ~23% edge-sharing, while the Nd3+ ions act similarlyto network modifiers26.X-ray total structure factors, S(Q), and differential pair distributionfunctions (PDFs), D(r), are shown in Fig. 3a, b for two glasses processed inmicrogravity (MG1 and MG2) and two glasses processed on Earth (TG1and TG2). In the structure factors, the principal peak near Q = 2.13 Å−1represents the packing of Ti-O polyhedra in the glass network. In the PDFs,the low-r region matches the anticipated slope of−4πρ, where ρ = 0.08163atoms Å−3 is the atomic number density (equivalent to 5.10 g cm−3 for NTwith 18.2mol. % Nd2O3). The first PDF peak corresponds to the Ti-O paircorrelation, with a mean bond distance of rTiO = 1.92 Å and mean coordi-nation number of nTiO = 5.2 based on peak fitting. The second PDF peak isthe Nd-O correlation, with rNdO = 2.45 Å and nNdO = 8.4. These values aresomewhat different than those from the published structural model26:rTiO = 1.984(11) Å and nTiO = 5.72(6); rNdO = 2.598(22) Å andnNdO = 7.70(26). These discrepancies are larger than the typical uncertaintyfrom peak fitting, e.g. ±0.5 for coordination numbers, and likely arise fromslight compositional differences as discussed below.Also, peakfitting here isbased on a single diffraction measurement, compared to the publishedstructural model that was refined based on six independent diffractionmeasurements, which resolves many of the issues related to overlappingfeatures of the different atomic partial pair correlations.All glasses exhibit nearly identical structures. From the peak fitting, allterrestrial and microgravity glasses in Fig. 3b have Ti-O and Nd-O meanbond distances within 0.01 Å and coordination numbers within 0.1 of oneanother. (These variations are smaller than the typical uncertainties for peakfitting.) There are only subtle differences between samples, most easilydiscerned in the third peak of the PDFs. This third peak corresponds to thecation-cationpair correlations,withoverlapping contributions forTi-Ti, Ti-Nd, and Nd-Nd. In Fig. 3b, PDFs are shown for terrestrial glasses TG1 andTG2, containing 20.5(2) or 17.1(2) mol. % Nd2O3, respectively. TG1 hasgreater asymmetry in the 3rd PDF peak, as compared to the nearly nominalcomposition inTG2, with stronger intensity on the higher-r shoulderwherethe Nd-Nd correlation is anticipated. This difference in the PDFs isexpected, since compositional changes alter the weighting factors thatdetermine how much each atomic partial pair contributes to the overallPDF37. Compositional enrichment by Nd would increase the weightingfactor for Nd-containing atomic partial pair correlations. For the micro-gravity glassesMG1 andMG2, their 3rd PDFpeaks aremore similar toTG1than TG2. For this reason, we hypothesize that the microgravity glasses areFig. 2 | Density during melt quenching. Measurements of two replicate samples(MG1 andMG2) that formed glass during melt cooling in ELF. All temperatures arecorrected for sample emissivity and window reflections, and density has been cor-rected for sample mass loss and internal porosity, as discussed in the main text.Linear fits to the glass and liquid thermal expansion are shown with dashedblack lines.Fig. 3 | X-ray diffraction of NT glasses. a X-ray total structure factors andb differential pair distribution functions for two replicate glasses melt quenched inmicrogravity (MG1 and MG2) and two glasses melt quenched on Earth (TG1 with20.5 mol. %Nd2O3; TG2with 17.1 mol. %Nd2O3). The gray dotted line indicates thelow-r slope inD(r) given by−4πρ, where ρ is the atomic number density. Curves areoffset vertically for visual clarity.https://doi.org/10.1038/s41526-024-00371-x Articlenpj Microgravity |           (2024) 10:26 4near ~19mol. % Nd2O3. This discrepancy vs. the nominal composition(17mol. % Nd2O3) can be explained by (i) the typical compositional var-iation of ~1mol. % Nd2O3 observed between replicate beads, and (ii) the~1% mass loss of samples processed in microgravity. (Compositionalmeasurements were not performed on the microgravity glasses due toconcerns about the required surface grinding and polishing for EDS.)X-ray measurements were collected at 20 locations across each glasssample, and the structure factors were found to be consistent, indicatinghomogeneity throughout each sample.Neutron diffraction is a technique for assessing atomic structure that iscomplementary to X-ray diffraction. Because X-rays scatter more stronglyfrom heavier elements, the X-ray structure factor for NT is predominantlyweighted by the atomic pair correlations that contain Nd or Ti. Neutronscattering is much more sensitive to O than X-rays are, so the neutronstructure factor contains more information on the O local environments26.For this reason, neutron diffraction was also conducted as a probe forstructural differences between the glasses processed in microgravity vs.terrestrially.Neutron diffractionmeasurements are shown in Fig. 4 for glasses meltprocessed in either microgravity (MG1 and MG2) or on Earth (TG3 andTG4). The scattering intensities, I(Q), are similar in Fig. 4a for all samples.To discern subtle differences, the scattering intensities in Fig. 4a were eachdivided by I(Q) for one of the terrestrial glasses (TG3), yielding the I(Q)ratios in Fig. 4b. The most notable differences between samples are theirslopes in Fig. 4b, which are likely due to slight variations in samples’ posi-tions relative to the neutron beam. Another possible explanation is differ-ences in inelastic scattering, perhaps arising from hydrogen content in thesamples. Optical measurements of rare-earth titanate glass have shown thatthey generally have low hydroxide impurity38, so it is likely that anyhydrogen differences in neutron scattering arise fromwater. Otherwise, theI(Q) ratios are nearly featureless aside from some faint fluctuationsnear Q = 4Å−1.The reduced and normalized Raman spectra for microgravity andterrestrial glasses are shown in Fig. 5a. Each Raman spectrum displaysexpected features of highly coordinated titanate polyhedra, as discussed bySu et al.39, andbears a strong resemblance to a spectrumrecently reported for83TiO2-17La2O338. These features include a highly polarized envelopespanning roughly 550–900 cm−1, belonging to stretching modes of titanatepolyhedra with coordination numbers larger than 4, and a depolarizedenvelope between 300–550 cm−1 containing bending modes.For the two terrestrial glasses (TG1 and TG2 with 20.5(2) and 17.1(2)mol. % Nd2O3, respectively), the spectra were adequately reproduced usingsix peaks. The peak fitting results are plotted in Supplementary Fig. 2, andthe band assignments are summarized in Supplementary Table 1 anddescribed as follows. First, there is a weak and relatively broad band at245 cm−1, which overlapswith the edge of the instrument range and, as suchFig. 4 | Neutron diffraction of NT glasses. a Scattering intensity after backgroundsubtraction and normalization to a vanadium standard; b the ratio of scatteringintensity for each sample to that of one terrestrial glass. Data is shown for tworeplicate glasses melt quenched in microgravity (MG1 and MG2) and two glassesmelt quenched on Earth (TG3 and TG4, both containing 17.6(2) mol. % Nd2O3).Curves are offset vertically for visual clarity.Fig. 5 | Raman spectra of NT glasses. a Unpolarized spectra for two glasses meltquenched in microgravity (MG1 and MG2), and two glasses melt quenched ter-restrially (TG1 with 20.5 mol. % Nd2O3; TG2 with 17.1 mol. % Nd2O3). Annotatedbands are discussed in the text. b Fitted component bands corresponding tostretching of Ti-O polyhedra that comprise the glass structural network, comparingcompositional effects for TG1 and TG2.https://doi.org/10.1038/s41526-024-00371-x Articlenpj Microgravity |           (2024) 10:26 5should not be over interpreted. Vibrational activity in this region is oftencharacteristic of network deformation modes. Rare-earth cations can alsodisplayRE-O stretching activity in the region between200–350 cm−1, whichcould contribute to the 245 cm−1 component. This RE-O activity likelycontributes strongly to a band at 337 cm−1, which increases in intensity withincreasing Nd2O3 content. For comparison, far infrared analysis of lithiumrare-earth orthoborate glasses containing 20mol. % RE2O3 displayed peakvibrational activity between 320–380 cm−1, depending on the RE3+fieldstrength40. Raman measurements on the same glasses40 as well as binaryrare-earth borates displayed similar features41. Considering that the337 cm−1 band displays a ~23% intensity increase as the Nd2O3 contentincreases just a few percent, it is likely that bending modes of some titanatearrangements are also contributing here. The band at 436 cm−1 is solely dueto bending of Ti-O-Ti linkages, based on its depolarized nature (Supple-mentary Fig. 3) and higher frequency than is expected from RE-relatedactivity.The high-frequency envelope is best described by three componentbands. The prominent 738 cm−1 is consistent with the stretching activity ofcorner-sharing TiO6moieties39. The broadest peak, centered at 596 cm−1, ismost likely due to edge-sharing TiO6 units and is just slightly below the600–650 cm−1 range where the corresponding feature is found in anataseand rutile39,42. Finally, the peak at ~828 cm−1 is most likely due to TiO5polyhedra. Activity in the 830–930 cm−1 range is generally unclear in TiO2-bearing glasses43–45: it has often been attributed to TiO5 species, as in fres-noite, but recently Santos et al. argued the assignment of an 825 cm−1 bandas belonging toTiO4 specieswithnonbridgingoxygen43,whichareknown tobe at higher frequencies when all bridging42,46,47. Since these NT glasses areexpected to have ~26% of Ti in TiO5 and only a small population of TiO4(~1%)26, the band at ~828 cm−1 here is assigned to TiO5. During the fittingprocedure, a band was positioned between 950–1050 cm−1 to simulate thesymmetric stretch of possible TiO4, but the fitting optimization returnedzero intensity for this component, consistent with the previously publishedstructural model26.The compositional variation between terrestrial glasses TG1 and TG2is subtle but discernible. Mainly, as the Nd2O3 content increases(TG2 < TG1), the intensities of the 337 cm−1 and 738 cm−1 bands increase,and the intensity of the 596 cm−1 band decreases. A comparison of thesefitted component bands for corner- and edge-sharing Ti-O polyhedra isshown in Fig. 5b. These differences may be due to an increase in corner-sharingTiO6 as the addedNd3+ cations depolymerize theTi-Onetwork anddestroy edge-sharing arrangements. Since the corner-sharing motifs arealready the majority species, such a scenario would lead to a less diversestructural landscape, and this may be the mechanism behind the decreasedFWHM values fitted for the high-frequency stretching bands (Supple-mentary Table 1).Raman spectra for the twoglassesprocessed inmicrogravity (MG1andMG2) are similar to each other, and their spectra are more similar to TG1with 20.5(2) mol. % Nd2O3 than to TG2 with 17.1(2) mol. % Nd2O3. Thisobservation is consistent with the compositional variation suggested by theX-ray diffraction data. The microgravity glasses and TG1 show some dif-ferences in the low-frequency range (i.e., 337 and 436 cm−1 bands).Microstructure of glasses and crystalsGlass samples prepared in microgravity and terrestrially all appearedhomogeneous in optical microscopy, X-ray tomography, and scanningelectron microscopy (SEM). The three-dimensional reconstructions fromtomography exhibited no density contrast, except for a few small internalvoids. Figure 6 provides exemplary cross-sections from tomography, andvideo animations showing the three-dimensional reconstructions are inSupplementary Movies 1–5. The two microgravity glasses (Fig. 6a) eachcontained 2–3 pores ca. 0.24mm in diameter or smaller, yielding totalporosities of 0.10–0.25%. These pores were spherical, suggestive of gasbubbles present in the melt. The terrestrial glasses (Fig. 6d) similarlyexhibited a few, small internal pores. SEM imaging of terrestrial glass cross-sections revealed no features except for surface scratches from polishing(Fig. 7a).Crystalline samples from melt quenching exhibited much moremicrostructural diversity between microgravity and terrestrial conditions.Crystalline terrestrial samples appearedhomogeneous inX-ray tomography(Fig. 6e), except for small pores and cracks. The lack of density contrastsuggests that the sample is single-phase or, if multi-phase, that its crystallinedomains are smaller than ~8 μm (i.e., ~6 voxels wide, given the voxel size of(1.37 μm)3). The crystalline microgravity sample exhibited clear densitycontrast in tomography (Fig. 6b), with lamellar crystalline grains 4–10 μmFig. 6 | Cross-sections of NT samples from X-ray microtomography. a Glass, b crystal, and c glass-ceramic formed during melt processing in microgravity. d Glass ande crystal formed during terrestrial melt processing. Scale bar is 500 μm. The red arrow indicates the likely crystal nucleation point in b.https://doi.org/10.1038/s41526-024-00371-x Articlenpj Microgravity |           (2024) 10:26 6thick and as wide as ~800 μm. Voids are present in the sample both inspherical form and as small gaps or cracks between crystalline domains.SEM (Fig. 7b) and energy dispersive spectroscopy (EDS, Fig. 8) revealdomains of three different compositions: 30.4(7), 17.3(2), and 2.5(1.3) mol.% Nd2O3. Of the previously identified crystalline phases for TiO2-Nd2O348,these compositions correspond most closely to Nd2Ti2O7 (33.3 mol. %Nd2O3), Nd4Ti9O24 (18.2 mol. % Nd2O3), and TiO2.The glass-ceramic prepared in microgravity also appeared homo-genous in X-ray tomography (Fig. 6c) and SEM (Fig. 7c). Its crystallinephase is a minor component of the sample, based on the very weak Braggpeaks observed in its X-ray diffraction. Indexing the Bragg peaks does notprovide a clear match to any of the expected phases (Supplementary Fig. 4).The closest match is Nd2Ti2O7, though some peak positions are shifted ormissing, which may be due to a highly distorted crystal lattice and/or thismeasurement’s lack of orientational averaging. The glass-ceramic Braggpeaks may also match Nd2Ti3O9 with a considerable contraction of thelattice. Small-angle X-ray scattering (SAXS, Fig. 9) shows a linear Porodregime in the range ofQ = 0.6–1.0 Å−1, which corresponds to the interfacesbetween crystalline and glassy domains. ThePorod slope is close to the valueof 2 expected for flat disc-shaped particles, which is consistent with thelamellar morphology of Nd2Ti2O7 crystal grains observed in the fullycrystalline product (Fig. 6b). The SAXS exhibits a broadpeak approximatelycentered at Q = 0.5 Å−1, corresponding qualitatively to a feature 2π/Q = 1.25 nm in size. The origin of this broad peak is unclear; it is too smallfor the crystallite size, given the Nd2Ti2O7 monoclinic unit cell49 witha = 7.67 Å, b = 13.00 Å, and c = 5.46Å.DiscussionThe discussion focuses on how melt processing in microgravity affects thesolidification behavior, both for vitrification and crystallization. We beginwith vitrification.Fragile liquids generally have lowglass-forming ability, as is the case forrare-earth titanates likeNT25. Glasses ofNThave been prepared by only twomethods: roller quenching and containerless processing. In roller quench-ing, exceptionally high cooling rates (~106 °C s−1) are the key to avoidingcrystallization, and only thin flakes (tens of μm thick) of product can beobtained. In containerless processes such as levitation, the lack of surfacecontact eliminates heterogeneous nucleation of crystals in the sample, soglasses can be formed at much lower cooling rates (102 to 103 °C s−1) andwith larger sample dimensions (spheres ~3mm indiameter). These benefitsmake levitation a useful tool for glass research on fragile liquids, providinglarge enough samples to investigate application-specific properties andguide the development of new glasses.For aerodynamic levitation, however, the sample size that can bevitrified is limited by fluid properties and/or heat transfer. Stable levitationrequires that the liquid sample maintains a nearly spherical shape, somaterials with low surface tension are challenging to process because liquidbeads deform or fragment in response to the forces of gravity and gas flowaround the sample. For fragile liquids with sufficient surface tension, themaximum sample size for glass formation is limited by the critical coolingrate for vitrification. As already discussed in Fig. 1b, the cooling rate(−dT⁄dt) is predominantly determined by radiative heat transfer at hightemperatures (e.g., above 900 °C) and by convection at lower temperatures.As sample size increases, the smaller ratio of surface area to volume results ina smaller radiative cooling rate. Once the sample is sufficiently large, it canno longer cool faster than the critical cooling rate, so it will crystallize. In theterrestrial aerodynamic levitator, NT melts up to 53mg consistentlyFig. 8 | Energy dispersive spectroscopy maps of NT crystalline sample formedduring melt processing in microgravity. The electron micrograph shows threecontrasting regions of increasing brightness, corresponding most closely topreviously identified phases TiO2, Nd4Ti9O24, and Nd2Ti2O7. The (fourth) darkestregions are voids in the sample. Scale bar is 10 μm.Fig. 9 | Small-angle X-ray scattering of NT glass-ceramic formed during meltprocessing in microgravity. The linear regime atQ = 0.6-1.0 Å-1 corresponds to theinterfaces between crystalline and glassy domains.Fig. 7 | Scanning electron microscope images of NT samples. a Terrestrial glass; b microgravity crystal; c microgravity glass-ceramic. Scale bar is 3 μm.https://doi.org/10.1038/s41526-024-00371-x Articlenpj Microgravity |           (2024) 10:26 7vitrified, whilemelts 66mg or larger consistently crystallized during cooling(Supplementary Fig. 1e, f). Crystallization occurred reproducibly ca.1040 °C, which is consistent with the sample that crystallized in micro-gravity (discussed further below). Based on comparison of samples withdifferent masses, the critical cooling rate in the aerodynamic levitator is~170 °C s−1 at 1040 °C.The cooling rate in microgravity was consistently 40–110 °C s−1 lowerthan on Earth for NT samples of the same size (20–23mg, Fig. 1b). Thisdifference is mainly due to the forced convective cooling associated withaerodynamic levitation, which is absent in electrostatic levitation inmicrogravity. Because aerodynamic levitation is not well suited for themicrogravity environment, these differences in cooling rate may alsorepresent an inherent difference in melt processing for Earth vs. reducedgravity conditions. If the critical cooling rate is the same in both environ-ments, larger samples may be more difficult to vitrify in microgravity.The cooling rates for a microgravity sample (21mg) and the largestterrestrial glass (53mg) provide an interesting comparison. For this pair, near1040 °C is a crossover point (Supplementary Fig. 1f), above which themicrogravity cooling rate is faster and below which the microgravity coolingrate is slower, compared to the terrestrial 53mg sample. Yet, themicrogravitysample does not crystallize before reaching the glass transition atTg = 786 °C,suggesting that the most crucial temperature range for investigating criticalcooling rates is near or above the 1040 °C crossover temperature.The atomic structures of glasses were nearly identical for the Earth andmicrogravity processing conditions, except for subtle differences that couldbe explained by compositional variations of ~2mol. % Nd2O3. This com-parison provides validation, at least for rare-earth titanates, that the sameglass can be manufactured in space as on Earth, aside from differences inthermal history.Next, we turn the discussion toward observations of crystallization inmicrogravity. One NT sample crystallized in microgravity and formed ahighly unusual microstructure compared to the terrestrial crystalline sam-ples (cf. Fig. 6b, e). The microgravity sample exhibits lamellar crystal grainsof Nd2Ti2O7 (the brightest regions in Figs. 6b and 7b), with some individuallamellae spanning most of the entire sample. Many of these lamellae ori-ginate from a common internal point (Fig. 6b, red arrow; SupplementaryMovie 6), suggesting that nucleation first occurred at this location.The unusual crystal nucleation of this sample is likely due to a combi-nation of unstable levitation and localized laser beam heating of the super-cooled melt. When this sample in ELF was being held isothermally ca.1750 °C, it began todrift in andoutof the central levitationpositionwhere thefour heating lasers were co-focused. (Levitation instability often occurs whenthe surface charge on the sample diminishes or changes polarity50.) Thesample exited the lasers’ focal spot for ~1.5 s, then partially reentered thelevitation position, and assumed a stable position such that the lasers’ focuswas off-center on the sample. Then, sudden crystallization occurred, evi-denced by the sample surface becoming faceted and loss of stable surfacecharge. From these observations and the clear nucleationpoint observedwithtomography, it seems likely that the small laser focal spot (0.5mm diameter)resulted in very localized heating, which led to crystal nucleation in the hotspot. This scenario is very different than the terrestrial aerodynamic levitator,in which a single, less focused laser provides sample heating.The microgravity crystalline sample contained a mixture of threecrystal phases, which can be explained using the equilibrium phasediagram48 (Supplementary Fig. 5), though not with certainty since the localtemperature during nucleation is not known because the sample positionwas not aligned with the pyrometer. First, we assume the liquid is likely~19mol. % Nd2O3, based on X-ray diffraction and Raman spectroscopy.From tomography, we infer that Nd2Ti2O7 was the first phase to nucleatefrom this liquid, so the sample temperature was either 1500–1589 °C(between the solidus and liquidus) or below 1199 °C. (Between 1199 and1500 °C, Nd2Ti3O9 is expected instead of Nd2Ti2O7.) The higher tem-perature range is most likely, given that the pyrometer corrected tempera-ture was 1410 °C at the time of nucleation, and the local temperature at thelasers’ focal spot would be higher. As the Nd2Ti2O7 crystals grew, the meltwould then become depleted in Nd2O3 until approaching the eutecticcomposition at 15.6mol. %Nd2O3. Further cooling would then result in thecrystallization of Nd4Ti9O24 and TiO2, consistent with the final mixture ofphases observed in SEM/EDS.MethodsSample preparation and melt processingSamples of nominal composition 83TiO2-17Nd2O3 (NT) were preparedfrom powders of TiO2 (99.98%, rutile, Aldrich) and Nd2O3 (99.999%,Cerac). Powders were dried in Pt crucibles at 600 °C for at least 3 h, thenweighed, and mechanically mixed. Portions of the powder mixture werethenheated andbrieflymelted in a copper hearth using a 10.6 μmCO2 laser.The resulting spheroids were then processed in an aerodynamic levitator27using air as the gas stream, and theywere remeltedwith laser beam heating.The melts were heated to 2000 °C for approximately 30 seconds and thenfreely cooled by turning off the laser. Samples for both the terrestrial andmicrogravity experimentswere prepared following this procedure, althoughfrom different batches of powder mixtures due to an insufficient amount ofthe original mixture. This contributed to the compositional variationsdescribed in the Results section.Cooling curves in the aerodynamic levitator were collected for samplesof varying mass, 20–66mg each, using pyrometers with λ = 0.9, 1.5, or5.0 μm (Chino Corporation, spot size 0.6mm diameter). Multiple coolingtraces for each sample and condition were recorded at 100Hz to ensurereproducibility. The apparent temperaturewas corrected assuminga sampleemissivity of 0.86.For microgravity experiments, samples were loaded to a sample car-tridge for the ELF16 and launched to the International Space Station. Eachsample was levitated and then laser beam heated. Levitation stability wor-sened upon heating as the sample charge decreased and switched polarity,particularly in the 1000–1400 °C range. If the sample motion could becontrolled, upon melting the levitation would again become stable oncethermionic emission created a stable surface charge. For the NT composi-tion, 5 out of 6heating attempts succeeded inmelting the samples. Themeltswere heated to ~1850 °C and held isothermally for at least several minutes.Then, the lasers were turned off, and the sample cooled freely. Pyrometrydata (IMPAC IGA140, λ = 1.55 μm, spot size 0.3–0.5mm diameter) werecollected at 100 Hz, and the apparent temperature was corrected assumingan effective emissivity of 0.63. This effective emissivity accounts for themultiplicative effects of twowindowspositionedbetween thepyrometer andthe sample, and a sample emissivity of 0.86. The temperature uncertainty isca. ±30 °C at 1800 °C and smaller at lower temperatures. A full discussion ofthe temperature correction estimations and temperature uncertainties isprovided in the Supplementary Discussion.Sample cooling rates were calculated from the pyrometry data. Foreach time point, a range of data ±0.5 s was defined and then fitted using asecond-order polynomial. The instantaneous cooling rate was then calcu-lated as the slope of the fitted polynomial, evaluated at the time point. Thisprocedure was tested using different fitting ranges (±0.1, 0.5, or 1.0 s), and±0.5 s was found to best fit the cooling data. However, subtle features likesmall recalescence events were not always resolved with this method, socooling traces were manually assessed for recalescence.Density measurements and analysisSamples’ densities were determinedusing theirmasses aftermelt processingand volume measurements made in ELF, as described previously17,51. As abrief summary, duringmelt processing a camera recorded backlit silhouettevideo of the sample (30 Hz, interlaced). Individual frames from these videoswere extracted and deinterlaced, and an edge detection algorithm was usedto define the perimeter of the sample. The sample volume was then calcu-lated from the edge fitting using a Legendre polynomial series18. Volumecalibration (mm3pixel−3)was performedby analyzing videoof stainless steelcalibration spheres that were levitated in ELF at the beginning and end ofeach sample holders’ experiments. Lastly, the sample volumes were cor-rected for the internal porosity as determined from X-ray tomography.https://doi.org/10.1038/s41526-024-00371-x Articlenpj Microgravity |           (2024) 10:26 8Pyrometrydata, collectedat 100Hz,were spline interpolated to a60 Hzbasis to align with the deinterlaced video frames. Sample volumes at a giventemperature were compared for dynamic (i.e., during sample cooling) andisothermal measurements, which were found to be consistent.Atomic structure measurements of glassesHigh-energy X-ray diffraction measurements were performed at Section6-ID-D of the Advanced Photon Source, Argonne National Laboratory(Argonne, IL, USA). The diffracted intensity of 99.9 keV X-rays fromwhole samples was measured in transmission geometry using a two-dimensional area detector (Varex 4343CT) at a distance of ~329mmfrom the sample. Samples were mounted on polyimide tape, and thesample was positioned so that the incoming X-ray beam (0.5 mm wideand 0.5mm tall) passed through near the edge of the glass bead, so thatX-ray transmission was 84–91%. The diffracted intensity was azimuthallyintegrated using Fit2D52, and the data were reduced and normalized inGudrunX53 to obtain the X-ray total structure factors, S(Q), followingprocedures described previously26. Free atom X-ray form factors wereused according to Waasmaier and Kirfel54, and the “top hat” convolutionin GudrunX55 was not applied. A sine Fourier transform was applied tothe structure factors to obtain the differential PDFs, D(r):D rð Þ ¼ 2πZ Qmax0Q S Qð Þ � 1ð ÞMðQÞ sin Qrð ÞdQ ð5Þwhere ρ = 0.08163 atoms Å−3 is the atomic number density, M(Q) is theLorch modification function56, and Qmax = 21.8 Å−3.To obtain mean bond distances and coordination numbers, NXFit57was used to fit the first two PDF peaks using Gaussian distributions con-volved with their associated peak-shape functions. For peak fitting, the totalPDF was used, T(r) =D(r)+ 4πρr. In NXFit, three Gaussian distributionswere initially defined, one each for Ti-O, Nd-O, and O-O. The initialcoordination numbers and O-O bond distances were taken from the pub-lished structural model26, and bond distances were set for Ti-O and Nd-Obased on the X-ray PDF maximum peak positions. The peak fitting para-meters were then refined over the range r = 1.335–2.60 Å using limits of±0.04 Å for bond distances and ±0.8 for coordination numbers.Neutron diffraction measurements were collected at the NOMADbeamline58 of the Spallation Neutron Source, Oak Ridge National Labora-tory (Oak Ridge, TN, USA). Individual glass beads were loaded into 3mmdiameter thin-walled silica capillaries and measured. Because the scatteringintensity from small samples is quite weak, measurements were also col-lectedoneach emptycapillary so that the background subtractionwouldnotleave silica signatures due to differences in the capillary wall thicknesses. Atthe sample position, the neutron beam was approximately Gaussian inshape with a FWHM of 6mm. After background subtraction, the samples’scattering intensities were normalized against that of a vanadium standard.Raman spectroscopy was carried out on a WITec Alpha300 RamanSpectrometer, equippedwithan1800groovesmm−1 diffraction gratingwitha ×50 objective lens. All measurements were made at room temperaturewith a resolution of 10 cm−1 or better using an excitation wavelength of488 nm. Each spectrum is an average of 200 scans in the range of85–1400 cm−1 accumulated in one-second intervals. Spectra were correctedfor the Bose-Einstein thermal population factor and area normalized59,60.Peak fitting of the terrestrial glasses’ Raman spectra was carried out usingcommercial graphing software (OriginLab). Lineshapes of componentswere assumed to be pseudo-Voigt functions. The Lorentzian and Gaussiancontributions were fixed at 60% and 40%, respectively61–63. The spectrum ofTG2 glass (17.1mol.%Nd2O3)was fit first, and the center frequencies of thebands were subsequently fixed while the bandwidths and areas were opti-mized. This methodology produced good fits for glasses with differentNd2O3 content, including TG1, though the constant center frequencyconstraint needed to be relaxed for a small shoulder band at ~828 cm−1. Thisfeature, as discussed in the Results section, has a strong overlap with a high-intensity feature at ~740 cm−1. Good fitting results were obtained byallowing the position of this band to vary within the range 828 ± 5 cm−1.Microstructure measurementsX-ray microtomography measurements were performed at Sector 7-BM-Bof the Advanced Photon Source. Incoming X-rays were attenuated with Cuand Ge filters to yield a polychromatic X-ray beam with a mean energy ca.80 keV. Each sample was positioned in a parallel beam configuration, andX-ray transmission was detected with a 25 μm thick LuAG scintillatorpositioned in front of a ×5 objective lens and FLIR Oryx camera. The setupprovided a pixel size of 1.37 μm in the recorded projections. For eachtomography scan, 2001 projections were collected at equally spaced rota-tional increments between 0 to 180° inclusive with a camera exposure timeof 20ms, in addition to dark and background images. Tomographicreconstructions were obtained with TomoPy64 and TomoCuPy65, using thegridrec algorithm66 and accounting for beamhardening. Image analysis wasperformed in ImageJ, first for segmentation using an automatic threshold.Then, porosity was measured by recording the number of “solid” pixelsbefore and after filling holes in the sample cross-section images.For SEM and energy dispersive spectroscopy (SEM/EDS, HitachiSU8030), samples were mounted in Crystalbond 509 (Buehler), mechani-cally grinded, andpolished to afinal stepwith 1 μmdiamond suspension.Atleast ten EDS measurements were collected across four sites on eachpolished cross-section to assess compositional uniformity.Small-angle X-ray scattering of the microgravity glass-ceramic samplewas measured at Sector 9-ID-C of the Advanced Photon Source67. Thesample was mounted with Scotch Magic Tape (3M), and scattering of21 keV X-rays was measured in a transmission geometry. Data werereduced with Nika software68.Reporting summaryFurther information on research design is available in the Nature ResearchReporting Summary linked to this article.Data availabilityAll data are available upon reasonable request.Code availabilityAll pertinent steps for analysis are detailed in the Methods section; noauthor-generated code is necessary. Code-related questions can be directedto the corresponding author.Received: 20 November 2023; Accepted: 19 February 2024;References1. National Academies of Sciences, Engineering, andMedicine. Thrivingin space: ensuring the future of biological and physical sciencesresearch, a decadal survey for 2023–2032 (2023).2. Sacco, E. &Moon, S. K. Additivemanufacturing for space: status andpromises. Int. J. Adv. Manuf. Technol. 105, 4123–4146 (2019).3. Reitz, B. et al. Additive manufacturing under lunar gravity andmicrogravity.Microgravity Sci. Technol. 33, 25 (2021).4. Fateri, M. & Gebhardt, A. Process parameters development ofselective laser melting of lunar regolith for on-site manufacturingapplications. Int. J. Appl. Ceram. Technol. 12, 46–52 (2015).5. Zhao, C. et al. Real-time monitoring of laser powder bed fusionprocess using high-speed X-ray imaging and diffraction. Sci. Rep. 7,3602 (2017).6. Levine, L. et al. Outcomes and conclusions from the 2018 AM-benchmeasurements, challenge problems, modeling submissions, andconference. Integr. Mater. Manuf. Innov. 9, 1–15 (2020).7. Stoudt, M. R. et al. Location-specific microstructure characterizationwithin IN625 additive manufacturing benchmark test artifacts. Integr.Mater. Manuf. Innov. 9, 54–69 (2020).https://doi.org/10.1038/s41526-024-00371-x Articlenpj Microgravity |           (2024) 10:26 98. Team Tempus. Containerless processing in space: recent result. inmaterials and fluids under low gravity. Lecture notes in physics, vol.464 (eds. Ratke, L., Walter, H. & Feuerbacher, B.) (Springer, 1996).9. Egry, I., Diefenbach, A.,Dreier,W.&Piller, J.Containerlessprocessingin space—thermophysical property measurements usingelectromagnetic levitation. Int. J. Thermophys. 22, 569–578 (2001).10. Soellner, W. & Aicher,W. The Electromagnetic Levitator Facility (EML)onBoard the ISS. In:Metallurgy in space (eds. Fecht,H.-J. &Mohr,M.)25–41 (The Minerals, Metals & Materials Society, 2022).11. Cozmuta, I., Cozic, S., Poulain, M., Poulain, S. & Martini, J. R. L.Breaking the silica ceiling: ZBLAN-based opportunities for photonicsapplications. Proc. SPIE 11276, 112760R (2020).12. Tucker, D. S., Ethridge, E. C., Smith, G. A. & Workman, G. Effects ofgravity on ZBLAN glass crystallization. Ann. N.Y. Acad. Sci. 1027,129–137 (2004).13. Torres, A., Ganley, J. &Maji, A. Understanding the role of gravity in thecrystallization suppression of ZBLAN glass. J. Mater. Sci. 49,7770–7781 (2014).14. Kasap, H. Exotic glass fibers from space: the race to manufactureZBLAN. Upward, Magazine of the ISS National Lab (2018).15. Tucker, D. S. & SanSoucie, M. Production of ZBLAN optical fiber inmicrogravity. In: Optical fiber sensors conference 2020 SpecialEdition (eds. CranchWang, A., Digonnet, M., and Dragic, P. G.) T2B.1(Optica Publishing Group, 2020).16. Tamaru, H. et al. Status of the electrostatic levitation furnace (ELF) inthe ISS-KIBO.Microgravity Sci. Technol. 30, 643–651 (2018).17. Koyama, C. et al. Densities of liquid lanthanoid sesquioxidesmeasured with the electrostatic levitation furnace in the ISS. J. Am.Ceram. Soc. 104, 2913–2918 (2020).18. Chung, S. K., Thiessen, D. B. & Rhim, W.-K. A noncontactmeasurement technique for the density and thermal expansioncoefficient of solid and liquid materials. Rev. Sci. Instrum. 67,3175–3181 (1996).19. Ishikawa, T., Koyama, C., Oda, H. & Saruwatari, H. Status of theelectrostatic levitation furnace in the ISS - surface tension andviscosity measurements. Int. J. Microgravity Sci. Appl. 39,390101 (2022).20. Rhim, W.-K., Ohsaka, K., Paradis, P.-F. & Spjut, R. E. Noncontacttechnique for measuring surface tension and viscosity of moltenmaterials using high temperature electrostatic levitation. Rev. Sci.Instrum. 70, 2796–2801 (1999).21. Kozuka, H., Ota, R. & Soga, N. Preparation and properties of binaryoxideglasses containing rare earth oxides. J. Soc.Mater. Sci. Jpn.35,73–79 (1986).22. Arai, Y., Itoh, K., Kohara, S. & Yu, J. Refractive index calculation usingthe structural properties of La4Ti9O24 glass. J. Appl. Phys. 103,094905 (2008).23. Masuno, A. et al. Glass-forming region and high refractive index ofTiO2-based glasses prepared by containerless processing. Phys.Status Solidi C. 9, 2424–2427 (2012).24. Angell, C. A. Formation of glasses from liquids and biopolymers.Science 267, 1924–1935 (1995).25. Alderman, O. L. G., Benmore, C. J., Tamalonis, A. & Weber, R. Rare-earth titanatemelt structureandglass formation. Int. J. Appl.Glas.Sci.10, 463–478 (2019).26. Wilke, S. K., Alderman, O. L. G., Benmore, C. J., Neuefeind, J. &Weber, R. Octahedral oxide glass network in ambient pressureneodymium titanate. Sci. Rep. 12, 8258 (2022).27. Weber, J. K. R. The containerless synthesis of glass. Int. J. Appl. Glas.Sci. 1, 248–256 (2010).28. Teterin, G. A., Zinchenko, V. F., Zagorodnyuk, A. V. & Minaev, I. M.Estimation of thermal stability of lanthanide dititanates in high-temperature region. Ukr. Khimicheskij Zh . 54, 252–255 (1988).29. Abe, Y. et al. Nonlinear dynamics of levitated droplet. Int. J.Microgravity Sci. Appl. 30, 42–49 (2013).30. Paradis, P.-F., Ishikawa, T., Fujii, R. & Yoda, S. Thermophysicalproperties of molten tungsten measured with an electrostaticlevitator. Heat. Transf. Res. 35, 152–164 (2006).31. McCormack, S. J., Tamalonis, A., Weber, R. J. K. & Kriven, W. M.Temperature gradients for thermophysical and thermochemicalproperty measurements to 3000 °C for an aerodynamically levitatedspheroid. Rev. Sci. Instrum. 90, 15109 (2019).32. Paradis, P. F., Yu, J., Ishikawa, T., Aoyama, T. & Yoda, S. Contactlessdensitymeasurement of high-temperature BiFeO3 and BaTiO3.Appl.Phys. 79, 1965–1969 (2004).33. Alderman, O. L. G. et al. Continuous structural transition in glass-forming molten titanate BaTi2O5. J. Phys. Chem. C. 120,26974–26985 (2016).34. Hübner, N. & Gruehn, R. Nd4Ti9O24: präparation und struktur. Z.Anorg. Allg. Chem. 616, 86–94 (1992).35. Hennet, L. et al. Development of structural order during supercoolingof a fragile oxide melt. J. Chem. Phys. 126, 074906 (2007).36. Ushakov, S. V. et al. Measurements of density of liquid oxides with anaero-acoustic levitator.Materials 14, 822 (2021).37. Benmore, C. J. A review of high-energy X-ray diffraction from glassesand liquids. ISRN Mater. Sci. 2012, 1–19 (2012).38. Topper, B. et al. Mid-infrared luminescence properties of erbium anddysprosium doped lanthanum titanate glasses. Opt. Mater. Express13, 2857–2868 (2023).39. Su, Y., Balmer, M. L. & Bunker, B. C. Raman spectroscopic studies ofsilicotitanates. J. Phys. Chem. B 104, 8160–8169 (2000).40. Topper, B., Tagiara, N. S., Herrmann, A., Kamitsos, E. I. & Möncke, D.Yttrium and rare-earth modified lithium orthoborates: glass formationand vibrational activity. J. Non-Cryst. Solids 575, 121152 (2022).41. Sasaki, S. et al. Structural origin of additional infrared transparencyand enhanced glass-forming ability in rare-earth-rich borate glasseswithout B–O networks. Inorg. Chem. 59, 13942–13951 (2020).42. Möncke, D., Lind, F., Topper, B. & Kamitsos, E. I. Anomalousdeformation behavior in ULE glass upon microindentation: avibrational spectroscopic investigation in the induced structuralchanges of a Ti-silicate glass. J. Phys. Chem. C. 125,4183–4195 (2021).43. Santos, A. G., Moulton, B. J. A. & Cabral, A. A. Discoveries about thestructure of alkaline earth-bearing borosilicate glasses doped withTiO2 revealed by Raman spectroscopy. J. Non-Cryst. Solids 578,121349 (2022).44. Henderson, G. S. & Fleet, M. E. The structure of Ti silicate glassesby micro-Raman spectroscopy. Canad. Mineral. 33,399–408 (1995).45. Osipov, A. A., Liška,M., Osipova, L.M., Chromčiková,M. &Hruška, B.Thermodynamic modeling and Raman spectroscopy study of Na2O-TiO2-SiO2 glasses. Vib. Spectrosc. 111, 103160 (2020).46. Richter, S. et al. Ultrashort pulse induced modifications in ULE - fromnanograting formation to laser darkening. Opt. Mater. Express 5,1834–1850 (2015).47. Efthimiopoulos, I. et al. Femtosecond laser-induced transformationsin ultra-low expansion glass: microstructure and local densityvariations by vibrational spectroscopy. J. Appl. Phys. 123,233105 (2018).48. Gong,W. & Zhang, R. Phase relationship in the TiO2-Nd2O3 pseudo-binary system. J. Alloy. Compd. 548, 216–221 (2013).49. Ishizawa, N., Ninomiya, K., Sakakura, T. & Wang, J. Redeterminationof Nd2Ti2O7: a non-centrosymmetric structure with perovskite-typeslabs. Acta Crystallogr. E69, i19 (2013).50. Ishikawa, T. et al. Status of the electrostatic levitation furnace in theISS - evaluation of sample position control. Int. J. Microgravity Sci.Appl. 35, 350205 (2018).51. Nawer, J. et al. Uncertainty analysis and performance evaluation ofthermophysical property measurement of liquid Au in microgravity.npj Microgravity 9, 38 (2023).https://doi.org/10.1038/s41526-024-00371-x Articlenpj Microgravity |           (2024) 10:26 1052. Hammersley, A. P. FIT2D: a multi-purpose data reduction, analysisand visualization program. J. Appl. Crystallogr. 49, 646–652 (2016).53. Soper, A. K. & Barney, E. R. Extracting the pair distribution functionfrom white-beam X-ray total scattering data. J. Appl. Crystallogr. 44,714–726 (2011).54. Waasmaier, D. & Kirfel, A. New analytical scattering-factor functionsfor free atoms and ions. Acta Crystallogr. A51, 416–431 (1995).55. Soper, A. K. Inelasticity corrections for time-of-flight and fixedwavelength neutron diffraction experiments.Mol. Phys. 107,1667–1684 (2009).56. Lorch, E. Neutron diffraction by germania, silica and radiation-damaged silica glasses. J. Phys. C. Solid State Phys. 2,229–237 (1969).57. Pickup, D., Moss, R. & Newport, R. NXFit: A program forsimultaneously fitting X-ray and neutron diffraction pair-distributionfunctions to provide optimized structural parameters. J. Appl.Crystallogr. 47, 1790–1796 (2014).58. Neuefeind, J., Feygenson, M., Carruth, J., Hoffmann, R. & Chipley, K.K. The nanoscale ordered materials diffractometer NOMAD at thespallation neutron source SNS. Nucl. Instrum. Methods Phys. Res.Sect. B 287, 68–75 (2012).59. Galeener, F. L. & Sen, P. N. Theory for the first-order vibrationalspectra of disordered solids. Phys. Rev. B 17, 1928–1933 (1978).60. Tagiara, N. S. et al. Synthesis, thermal and structural properties ofpure TeO2 glass and zinc-tellurite glasses. J. Non-Cryst. Solids 457,116–125 (2017).61. Palles, D. et al. Electro-thermal poling in bioactive sodium‑calciumphosphate-silicate glass: anodic near-surface network connectivitychanges and second harmonic generation. J. Non-Cryst. Solids X 17,100164 (2023).62. O’Shaughnessy, C., Henderson, G. S., Nesbitt, H.W., Bancroft, G. M.& Neuville, D. R. The influence of modifier cations on the Ramanstretching modes of Qn species in alkali silicate glasses. J. Am.Ceram. Soc. 103, 3991–4001 (2020).63. Bancroft, G.M. et al. Lorentzian dominated lineshapes and linewidthsfor Raman symmetric stretch peaks (800–1200 cm−1) in Qn (n = 1–3)species of alkali silicate glasses/melts. J. Non-Cryst. Solids 484,72–83 (2018).64. Gürsoy,D.,DeCarlo, F., Xiao, X.&Jacobsen,C. TomoPy: a frameworkfor the analysis of synchrotron tomographic data. J. SynchrotronRadiat. 21, 1188–1193 (2014).65. Nikitin, V. TomocuPy–efficient GPU-based tomographicreconstruction with asynchronous data processing. J. SynchrotronRadiat. 30, 179–191 (2023).66. Marone, F. & Stampanoni, M. Regridding reconstruction algorithm forreal-time tomographic imaging. J. Synchrotron Radiat. 19,1029–1037 (2012).67. Ilavsky, J. et al. Development of combined microstructure andstructure characterization facility for in situ and operando studies atthe advancedphoton source. J. Appl. Crystallogr. 51, 867–882 (2018).68. Ilavsky, J. Nika: software for two-dimensional data reduction. J. Appl.Crystallogr. 45, 324–328 (2012).AcknowledgementsThis work was supported by the National Aeronautics and SpaceAdministration (NASA) through grant 80NSSC19K1288 and the U.S.Department of Energy (DOE) through grant DE-SC0018601. ELF measure-ments were supported by JSPS KAKENHI through grants 20H05882 and20H05878. X-ray diffraction, tomography, and small-angle scattering mea-surements were made at Sectors 6-ID-D, 7-BM-B, and 9-ID-C of theAdvanced Photon Source, a U.S. DOE Office of Science User Facility,operated by Argonne National Laboratory under Contract No. DE-AC02-06CH11357. Neutron diffraction measurements were made at the NOMADbeamline of the Spallation Neutron Source, a DOE Office of Science UserFacility operated by Oak Ridge National Laboratory. Ramanmeasurementsused instrumentation supported by the National Science Foundation underGrant No. DMR-1626164. SEM/EDSmeasurementsweremade at the EPICfacility of Northwestern University’s NUANCE Center, which has receivedsupport from the SHyNE Resource (NSF ECCS-2025633), the IIN, andNorthwestern’s MRSEC program (NSF DMR-2308691). The authors wouldlike to thank Dr. Douglas Matson and Jannatun Nawer for guidance inmicrogravity experiment planning, and Dr. Robert Hyers and Dr. RichardBradshaw for guidance in the calibration of the density measurements.Author contributionsS.K.W. led the measurements and data analysis for cooling rates, X-ray dif-fraction, neutron diffraction, X-ray tomography, and electronmicroscopy. A.A.analyzed density and tomography data and assisted with tomography mea-surements. C.K. and T.I. conductedmicrogravity experimentswith E.L.F., withsupport from H.O. B.T. and E.T. performed and analyzed Raman measure-ments, with supervision byD.M.O.L.G.A. contributed to planningmicrogravityexperiments and sample preparation. V.M. performed density analysis, withassistance from J.R. E.C. prepared samples. User facilitymeasurementswerefacilitated by A.L.K. for X-ray tomography, C.J.B. for X-ray diffraction, J.I. forsmall-angle X-ray scattering, and J.N. for neutron diffraction. S.K. contributedto E.L.F. experiments. M.S. and B.P. provided programmatic support. R.W.conceived the project goals, supervised activities, contributed to X-ray dif-fraction and tomographymeasurements, to data interpretation, and procuredfunding.S.K.W.prepared themanuscriptdraftwithcontributions fromB.T.andE.T., and revisions were contributed by all.Competing interestsThe authors declare no competing interests.Additional informationSupplementary information The online version containssupplementary material available athttps://doi.org/10.1038/s41526-024-00371-x.Correspondence and requests for materials should be addressed toStephen K. Wilke.Reprints and permissions information is available athttp://www.nature.com/reprintsPublisher’s note Springer Nature remains neutral with regard tojurisdictional claims in published maps and institutional affiliations.Open Access This article is licensed under a Creative CommonsAttribution 4.0 International License, which permits use, sharing,adaptation, distribution and reproduction in anymedium or format, as longas you give appropriate credit to the original author(s) and the source,provide a link to the Creative Commons licence, and indicate if changeswere made. The images or other third party material in this article areincluded in the article’s Creative Commons licence, unless indicatedotherwise in a credit line to the material. If material is not included in thearticle’sCreativeCommons licence and your intended use is not permittedby statutory regulation or exceeds the permitted use, you will need toobtain permission directly from the copyright holder. To view a copy of thislicence, visit http://creativecommons.org/licenses/by/4.0/.© The Author(s) 2024https://doi.org/10.1038/s41526-024-00371-x Articlenpj Microgravity |           (2024) 10:26 11https://doi.org/10.1038/s41526-024-00371-xhttp://www.nature.com/reprintshttp://creativecommons.org/licenses/by/4.0/ Microgravity effects on nonequilibrium melt processing of neodymium titanate: thermophysical properties, atomic structure, glass formation and crystallization Results Melt processing and glass formation Density Glass atomic structure Microstructure of glasses and crystals Discussion Methods Sample preparation and melt processing Density measurements and analysis Atomic structure measurements of glasses Microstructure measurements Reporting summary Data availability Code availability References Acknowledgements Author contributions Competing interests Additional information