Data for Nanoscale reservoirs store solar wind-derived water on the lunar surface

<p>The presence of water has been identified across the lunar surface by remote sensing observations. The primary mechanism proposed to drive the production of widespread water is the implantation of solar wind hydrogen ions into the lunar regolith via space weathering, a process that alters the optical, chemical, and microstructural properties of regolith grains on the surfaces of airless planetary bodies. During solar wind irradiation, energetic H<sup>+</sup> ions (with a mean energy of ~1 keV) bombard the lunar regolith and penetrate the surfaces of individual grains. This irradiation produces structural and chemical changes in the upper ~100 nm of grain surfaces, including the sputtering of atoms from grain rims, the production of metallic iron nanoparticles, the accumulation of implanted solar wind species, and the progressive disruption of the original crystalline microstructure demonstrated through the formation of defects. Defect-rich, space weathered rims are predicted to be the primary sites where implanted hydrogen may be trapped and can react with structural oxygen in the host material to form OH/H<sub>2</sub>O. To understand this relationship, we can employ high-resolution analytical techniques that not only detect water, but also provide spatial context to connect these observations with the microstructural and chemical environment within space weathered grains. Here, we present the results of a novel coordinated study using transmission electron microscopy (TEM), electron energy loss spectroscopy (EELS), and atom probe tomography (APT) to directly measure the presence of solar wind-derived water and link to its microstructural context in space weathered lunar grains.</p>

<p><strong>TEM and EELS analyses </strong></p>

<p>We performed preliminary began preliminary characterization of the electron transparent sections in the 200 keV Tecnai T20 and 200 keV Talos 200X transmission electron microscopes at Purdue University to investigate the microstructural and chemical characteristics of the grain. We used bright-field (BF), high-resolution TEM (HRTEM) phase contrast imaging, and high-angle annular dark field (HAADF) scanning TEM (STEM) imaging of these FIB sections to identify evidence of space weathering. We targeted grains that contained radiation-damaged rims of uniform thickness (~100 nm) extending across the grain surfaces in the FIB section. We also looked for evidence of other space weathering features such as nanoparticles and vesicles within the rim and solar flare tracks retained below the rim in the crystalline interior. We performed EDS mapping of these sections to compare the composition of the rims to the crystalline interiors to determine whether the rims were formed via solar wind irradiation or via melt/vapor deposition. Elemental abundances from EDS analyses were quantified in the Velox software using the Cliff-Lorimer method.</p>

<p>Electron energy loss spectroscopy measurements were performed on the Thermo Fisher Scientific 300 keV Themis Z monochromated and aberration-corrected S/TEM equipped with a Gatan Quantum 965 EELS detector at the Birck Nanotechnology Center at Purdue University and at the Center for Electron Microscopy and Analysis (CEMAS) at the Ohio State University. Instrument alignments for EELS analyses included a monochromator excitation of 0.8, a camera length of 29.5 mm, a convergence angle of 15.8 mrad, an energy resolution of ~0.18 eV, and dispersions of 0.01 eV/ch and 0.025 eV/ch. Measurements were collected in dual-EELS mode to collect data on the zero-loss peak (ZLP) and the low-loss energy region. An energy range of ~2-25 eV was defined within the low-loss region to remove signal from the ZLP and to include the plasmon feature of the material, a feature that arises due to the distribution of valence and conduction electrons in a material, as well as other spectral features of interest that are superimposed on the plasmon. Measurements were performed as linescans and spectrum images on areas within the rim, including vesicles, and in the grain interior. The pixel time of each linescan varied between 0.05 and 0.5 s. The pixel size was varied from 1 nm to 16 nm based on the size of features of interest and the resulting scan time. When acquiring spectrum images, we performed sub-pixel scanning in a 16x16 array. The spectral data was binned at 65 and 130.</p>

<p> </p>

<p><strong>EELS data processing</strong></p>

<p>EELS spectra and spectrum images were processed and analyzed using the Gatan Digital Micrograph software and the HyperSpy open-source Python library. Dual-EELS spectra were first corrected for energy shifts that occurred during acquisition using the “Align ZLP” tool in Digital Micrograph. After alignment of the spectra with the zero-loss peak, random spikes of high intensity were identified and removed using the HyperSpy “spikes_diagnosis” tool and “spikes_removal_tool,” respectively. We then applied a background subtraction to the model spectra to remove contributions of intensity from the ZLP. The decay in intensity of the ZLP can be modeled using a power law in the form of AE<sup>-r</sup>. A background fitting window was applied to each spectrum and the residual spectra were extracted.</p>

<p>We then performed a series of multivariate statistical analyses (MSA) such as principal component analysis (PCA) on the spectrum images to identify spectral features of interest in the spectrum images. We first performed a single variable deconvolution (SVD) to the spectrum images and applied a weighted scaling function that would “de-noise” the spectra of Poissonian noise (Bosman et al., 2006) using the HyperSpy “deconvolution” tool. Spectra were then reconstructed using a variable number of principal components that comprised the real variability within the data. The resultant spectra had reduced noise and scan artefacts relative to the acquired spectra and also highlighted spectral features of interest and their variability and spatial distribution.</p>

<p>From here, EELS spectrum images and linescans were evaluated for the presence of spectral features of interest using PCA and non-negative matrix factorization (NMF). NMF components were chosen to highlight the spectral features of interest and their spatial distributions because NMF provides a more physically interpretable signal.  We applied these analyses to subsets of this spectral range to better identify the presence and variability of features such as the energy gap of water (EG) which has a peak at ~8.6 eV, the hydrogen H-edge (H-K) which has variable peak shapes and reported energies and between ~13-14 eV, and the helium K-edge (He-K) with variable peak shapes and energies at ~22 eV.  </p>