Manipulating ionic conductivity through chemical modifications in solid-state electrolytes prepared with binderless laser powder bed fusion processing

Additive manufacturing of solid-state batteries is advantageous for improving the power density by increasing the geometric complexity of battery components, such as electrodes and electrolytes. In the present study, bulk three-dimensional Li1+x Al x Ti2−x (PO4)3 (LATP) electrolyte samples were prepared using the laser powder bed fusion (L-PBF) additive manufacturing method. Li3PO4 (LPO) was added to LATP to compensate for lithium vaporization during processing. Chemical compositions included 0, 1, 3, and 5 wt. % LPO. Resulting ionic conductivity values ranged from 1.4 × 10−6–6.4 × 10−8 S cm−1, with the highest value for the sample with a chemical composition of 3 wt. % LPO. Microstructural features were carefully measured for each chemical composition and correlated with each other and with ionic conductivity. These features and their corresponding ranges include: porosity (ranging from 5% to 19%), crack density (0.09–0.15 mm mm−2), concentration of residual LPO (0%–16%), and concentration and Feret diameter of secondary phases, AlPO4 (11%–18%, 0.40–0.61 µm) and TiO2 (9%–11%, 0.50–0.78). Correlations between the microstructural features and ionic conductivity ranged from −0.88 to 0.99. The strongest negative correlation was between crack density and ionic conductivity (−0.88), confirming the important role that processing defects play in limiting the performance of bulk solid-state electrolytes. The strongest positive correlation was between the concentration of AlPO4 and ionic conductivity (0.99), which is attributed to AlPO4 acting as a sintering aid and the role it plays in reducing the crack density. Our results indicate that additions of LPO can be used to balance competing microstructural features to design bulk three-dimensional LATP samples with improved ionic conductivity. As such, refinement of the chemical composition offers a promising approach to improving the processability and performance of functional ceramics prepared using binderless, laser-based additive manufacturing for solid-state battery applications.


Introduction
Solid-state electrolyte batteries offer enhanced safety and improved energy density compared to conventional liquid-electrolyte batteries [1].Li 1.3 Al 0.3 Ti 1.7 (PO 4 ) 3 (LATP) is a widely explored solid-state electrolyte known for exhibiting high room-temperature conductivity [2][3][4][5].A promising fabrication method for preparing LATP samples utilizes a two-step glass-ceramic processing approach in which the LATP powder particles are: (1) melted in a furnace and splat-quenched on a metal plate to form a glass, and then (2) heat treated at high temperatures to develop the crystalline microstructure [5].While glass-ceramic processing produces LATP samples with high room-temperature conductivity, the simple geometries afforded by this approach limit further improvement of the performance.
Additive manufacturing has emerged as a promising tool for fabricating batteries with complex geometries [6].Several additive manufacturing techniques have been used to print battery components, including lithography-based methods, template-assisted electrodeposition, inkjet printing, direct ink writing, photopolymerization, and fused deposition [7].For example, Ao et al used direct ink writing to print electrode inks containing active materials and carbon nanotubes [8].They demonstrated that direct ink writing could be used to successfully prepare large-scale Li-ion battery devices with customized geometries, and without any loss in electrochemical performance.Additive manufacturing also shows promise as a method of fabricating all solid-state batteries [9].For example, Wen et al used photopolymerization to produce a NaClO 4 based polymer gel electrolyte for Na solid-state batteries [10].Full battery pouch cells based on this gel electrolyte were mechanically robust and retained 97% of their capacity after 1000 cycles.Martinez et al combined two different additive manufacturing techniques to print Na-ion batteries with NaMnO 2 as the active material [11].The authors demonstrated a viable half-cell battery using vat photopolymerization to prepare the electrolyte and direct ink writing to prepare the cathode.
Additive manufacturing enables the fabrication of geometries and designs with the potential to directly influence three-dimensional battery performance [12].For example, Long et al [13] reviewed the concept of improving battery performance by adopting a three-dimensional interdigitated battery architecture that reduces the distance over which ion transport occurs to enhance power density.Pearse et al used atomic layer deposition to prepare thin film three-dimensional batteries.They found that the three-dimensional battery architectures demonstrated improved performance over the standard planar design [14].Milroy et al printed highly porous cathodes containing carbon nanotubes for Li-S batteries [15].Their cathodes exhibited a high areal capacity of 7 mAh cm −2 and a coulombic efficiency of 95% for over 200 cycles.Three-dimensional printed batteries also allow for incorporation of novel features into battery design, such as active cooling through the addition of coolant flow channels [16].However, exploration of three-dimensional batteries has often been restricted to micro-battery applications [17] due to the limitations of fabrication techniques.Further development of fabrication techniques capable of producing battery components with complex geometries is necessary to further improve the performance of solid-state batteries.
The additive manufacturing technique, laser powder bed fusion (L-PBF), offers a novel approach for preparing three-dimensional battery components while reducing the number of processing steps required.During L-PBF, a high-energy laser beam is scanned over a bed of powder particles to selectively consolidate regions of the powder bed [18].After consolidation of each layer of material, fresh powder is added over the previously consolidated material and the process is repeated until the three-dimensional part is fully built.L-PBF has previously been explored for preparation of LATP solid-state electrolytes [19] and LiNi 0.80 Co 0.15 Al 0.05 O 2 (NCA) cathodes [20].The L-PBF LATP samples are dense with the desired phase state necessary for high ionic conductivity, however, the heterogeneous distribution of secondary phases throughout the columnar dendritic microstructure hinders the performance [19].Cao et al used laser based additive manufacturing to prepare a porous anode with Si islands embedded in a Cu matrix [21].Full cell batteries containing this anode retained 84% of their capacity after 100 cycles.Additionally, the anode remained undamaged after cycling due to the Cu matrix being able to accommodate volume expansion during lithiation.
During high temperature processing of LATP, material vaporization of volatile species, including Li 2 O, occurs leading to a change in stoichiometry that results in the formation of secondary phases, such as AlPO 4 , which lower the total conductivity [22].Therefore, we hypothesize that incorporating Li 3 PO 4 (LPO), which includes the elements that have a high propensity to vaporize (e.g.Li, P, O), into LATP will compensate for the expected change in stoichiometry during high temperature L-PBF processing.Further, previously prepared L-PBF LATP samples exhibited cracks that further limited ion transport [19].The addition of LPO to LATP is known to improve sample density and, as a result, increases the ionic conductivity [23].Hence, the influence of blending LATP with the sacrificial chemical species, LPO, during L-PBF, is investigated here, not only to directly control behavior but also to limit processing defect formation, such as cracks and pores.
The decomposition of LATP is known to be highly complex, with the exact decomposition products being sensitive to temperature [24].Given the highly non-equilibrium nature of L-PBF, it is not possible to propose an exact decomposition reaction for LATP and LATP-LPO.However, the decomposition products relevant to this work can be estimated based on our previous work and the results discussed below.We hypothesize that the decomposition products are: Li 2 O, LiTi 2 (PO 4 ) 3 , AlPO 3 , and TiO 2 .Thus, the relevant chemical reaction for LATP will be: We hypothesize that the addition of LPO will enhance the processability of LATP.Assuming equilibrium conditions, and products that LPO does not change the decomposition of LATP, the relevant chemical reaction of LATP-LPO is: (2) When x = 0.3 (the starting stoichiometry of powder we used for this study), the decomposition of LATP becomes: When x = 0.3, the decomposition of LATP-LPO becomes: From equation ( 4), it can be seen that increasing LPO addition will increase the formation of Li 2 O and TiO 2 , while suppressing the formation of LiTi 2 (PO 4 ) 3 and AlPO 4 .It is important to note that LATP is known to release P 4 O 10 or O 2 during decomposition [22].We have chosen to neglect these types of decomposition products in order to simplify the above equations.
As such, the present study investigates the influence of chemical composition on the processability, microstructure, and electrical conductivity of LATP samples produced using L-PBF.With the goal of improving performance, LATP powder was blended with LPO to compensate for the stoichiometric changes caused by materials vaporization during L-PBF.Three-dimensional LATP-LPO samples with the desired rhombohedral crystal structure were successfully fabricated using L-PBF processing.The grain size, phase composition, and electrical conductivity were characterized to unveil the relationship between microstructural features and performance.

Materials and methods
A flow chart summarizing the experimental procedure is provided in figure 1. Commercially available LATP powder (reported particle diameter < 53 µm) (NEI Corporation, Somerset, NJ, USA) and LPO powder (reported particle diameter < 1 µm) (Millipore Sigma, Burlington, MA, USA) were used in this study.Various concentrations of LPO powder (1 wt.%, 3 wt.%, 5 wt.%) were blended with LATP powder using low energy ball milling prior to L-PBF processing.Employing a 2:1 ball-to-powder ratio, the mixtures of LATP and LPO powders were blended for 19 h in atmosphere using a low energy ball milling system (Patterson Kelley Blend Master).The blended LATP and LPO powders are hereafter referred to as LATP-LPO powders.X-ray diffraction (XRD) (Rigaku SmartLab x-ray Diffractometer with D/teX Ultra) and the relative intensity ratio method [25] were used to confirm the LPO concentration in each LATP-LPO powder.The diffractometer had a Cu Kα x-ray source and was scanned from 2θ: 10 • -50 • using a scan rate of 4 • min −1 , a 2 mm incident slit, and step width of 0.02 • .
Bulk three-dimensional LATP-LPO samples were prepared from the LATP-LPO powders by using a custom powder-bed setup in an Optomec LENS ® 750 Workstation with an argon filled glove box chamber.The powder-bed setup was prepared by spreading the as-prepared powders onto an alumina substrate using a spatula.The thickness of each powder bed layer was 100 µm and was controlled using an aluminum shim.More details about the custom powder-bed setup can be found in our previous studies [19,20].
The LENS ® system includes a 1 kW continuous-wave IPG Photonics YLR fiber laser with a top-hat profile.The laser has a measured beam quality factor (M 2 ) of 24.7 and a wavelength of 1070 nm [26].The system contains a fused silica lens with a 150 mm focal length at the operating wavelength.The laser column position was adjusted to produce an approximate beam spot diameter of 0.60 mm at the substrate.During the experiments, the laser was operated with a laser power of 100 W and was rastered according to the computer aided design model to selectively consolidate regions of the LATP-LPO powder bed.The process parameters used to prepare bulk three-dimensional LATP-LPO samples are provided in table 1.These process parameters were selected because they were found to be the optimal processing parameters for LATP, based on our previous study [19].It is noted that the scanning velocity is relatively low, and the beam diameter is relatively large, compared to values typically used for L-PBF.The three-dimensional LATP-LPO  samples were prepared with a cube geometry having the dimensions of 10.2 mm × 10.2 mm × 1 mm, which could be achieved after depositing nine layers.The chosen geometry and dimensions were selected due to this size being found to be optimal for electrochemical characterization.The bulk, as-deposited LATP-LPO samples were rinsed with isopropanol to remove powder particles that adhered to the samples from the deposition powder bed.The crystal structure and phase composition of these bulk samples (five samples for each LPO concentration) were evaluated using XRD with the same parameters noted above.The samples were cross-sectioned using a slow speed diamond saw and polished (diamond slurry with particle sizes of 30 µm, 6 µm, 1 µm, and colloidal silica) for microstructural evaluation.Optical microscopy was employed to evaluate the processing defects, including cracks and pores.The software ImageJ [27] was used to threshold a region of interest within the optical micrographs using the 'particle analysis' tool to characterize the area of the sample occupied by processing defects.In this study, porosity is calculated as a percentage of the micrograph occupied by pores with respect to the total area of the micrograph.The cracks are quantified using crack density (mm mm −2 ), which is calculated by measuring the total length of the cracks in a micrograph divided by the total area of the micrograph.In quantifying the concentration of processing defects, at least ten micrographs from individual samples for each LPO concentration were employed using cross-sections that were parallel and perpendicular to the build direction.
The polished LATP-LPO samples were also sputter coated with 3 nm of Ir (EMS 150 T Sputter Coater) and evaluated using back scattered electrons during scanning electron microscopy (SEM) (FEI Magellan 400 XHR SEM).The primary dendrite arm spacing was determined, using ImageJ, from at least ten SEM micrographs of individual samples from cross-sections parallel to the build direction.Energy dispersive x-ray spectroscopy (EDS) (Oxford Silicon Drift Detector 80 mm 2 with Aztec software) was used to assess the spatial distribution of the chemical species within the bulk LATP-LPO samples.The different chemical species (i.e., secondary phases) identified with EDS were characterized using ImageJ by segmenting each phase with the 'Trainable Weka Segmentation' tool [28] and quantifying the Feret diameter (i.e., the longest diameter of a shape) of over 1500 particles per sample, as well as the concentration using the 'particle analysis' tool.
The total conductivity of the bulk, as-deposited LATP-LPO samples was characterized using electrochemical impedance spectroscopy (EIS) and chronoamperometry (BioLogic SP-200 Potentiostat with EC-Lab ® Software).LATP-LPO samples were thinned to a thickness of ∼0.5 mm using a 70 µm diamond polishing plate and polished using diamond polishing slurries with particle diameters of 30 µm, 6 µm, and 1 µm.The faces of the thinned LATP-LPO samples were coated with Ag paste to create an ion-blocking two-probe cell with an Ag/LATP-LPO/Ag configuration.The EIS measurements were performed using 10 mV AC perturbation (V rms ∼ 7.07 mV) across a frequency range of 100 mHz-500 kHz.The total conductivity was determined by fitting the raw data using an equivalent circuit with (R 1 + Q 1 )/R 2 /Q 2 , where R is the resistance, Q is a constant phase element, the '+' symbol denotes elements in series, and the '/' symbol denotes elements in parallel [29].The total conductivity values were calculated using σ = t/(R * A) where t is the sample thickness, A is the cross-sectional area of the sample, and R is the sum of R 1 and R 2 from the equivalent circuit fitting.The ionic and electronic conductivity values were calculated by determining the transference numbers from chronoamperometry experiments.During the chronoamperometry experiments, a set potential (0.2 V) was applied to the two-probe LATP-LPO cells, and the current was measured as a function of time for three minutes.The peak current at the beginning of the experiment (I o ) and the plateau current at the end of the experiment (I f ) were used to determine the transference numbers.The electronic and ionic transference numbers were evaluated using the equations: t e = I f /I o and t i = (I o −I f )/I o , respectively.

Results
Prior to L-PBF, the blended powders (figure 2(a)) are comprised of the rhombohedral crystal structure (with space group: R-3 c) common for LATP with minor diffraction peaks indicative of the orthorhombic crystal structure (Pmnb) of LPO.There is good agreement between the designed concentration of LPO in the blended powders (i.e. 1 wt.%, 3 wt.%, and 5 wt.%) and the concentration of LPO determined by the relative intensity ratio method (i.e.1.3 wt.%, 2.8 wt.%, and 4.4 wt.%).XRD results for the L-PBF processed LATP-LPO bulk samples reveal that the desired rhombohedral crystal structure with minor amounts of the LPO phase are present after deposition (figure 2(b)).As such, the XRD results confirm that preparation of bulk samples with the desired rhombohedral crystal structure is feasible using blended LATP and LPO powders during L-PBF.
Processing defects, such as cracks and pores, are observed in the normal (figure 3(a)) and in-plane (figure 3(b)) cross-sections of the LATP-LPO samples.In this study, the processing defects are quantified using % porosity and crack density.The concentration of processing defects (cracks and pores) in LATP-LPO samples (figure 3(c)) varies with LPO concentration, though a distinct trend is not observed.The similarity in the concentration of processing defects for the 0 wt.% LPO and 1 wt.% LPO samples suggests that the addition of a small concentration of LPO has limited influence on the propensity for crack and pore formation.In contrast, the average value for the concentration of pores in the 3 wt.% LPO samples decreases by 30% compared to the value for the 0 wt.% LPO samples, while the crack density decreases by 25%.Notably, adding an even higher concentration of LPO, such as 5 wt.%, increases the pore formation by ∼180% compared to the concentration in the 0 wt.% LPO samples, while the crack density increases by 52%.This variation in response suggests that an intermediate concentration of LPO assists in consolidation and reduces the number of cracks and pores that form.It is important to note that the dependance of LPO concentration on the values of porosity and crack density are very similar, indicating that the mechanism underpinning the formation of the various defects in these samples are similar.
As with our previous study on L-PBF of LATP [19], the bulk LATP-LPO samples prepared in this study exhibit a columnar dendritic microstructure, yet the LPO concentration influences the specific features in the microstructures of the LATP-LPO samples.Representative backscattered electron (BSE) SEM micrographs (figures 4(a)-(d)) reveal variations in the size and morphology of the columnar dendritic grains.The grain size in the present study is represented by the dendritic arm spacing (highlighted by the double-sided white arrow in figure 4(a)).Although the trend is difficult to interpret due to large standard deviation values, quantification of the dendrite arm spacing (figure 4(e)) from the BSE SEM micrographs suggest that samples with 0 wt.% LPO exhibit a finer dendritic microstructure with less variation in dendrite arm spacing values compared to samples with higher concentrations of LPO.The dendrite arm spacing increases from 2.1 µm at 0 wt.% LPO to ∼3.5 µm at 1 and 3 wt.% LPO.Finally, the dendrite arm spacing increases to 7.0 µm at 5 wt.% LPO.As the concentration of LPO increases, the dendrite arm spacing increases and grain growth becomes more irregular leading to more variation in the dendrite arm spacing.
Secondary phases reside within the interdendritic regions of the LATP dendrites.The representative BSE SEM micrograph in figure 5(a) showing a cross-section of an LATP-LPO sample with 3 wt.% LPO highlights the LATP dendrites (middle grayscale) and secondary phases, TiO 2 (lighter grayscale) and AlPO 4 (darker grayscale).The identification of the secondary phases, TiO 2 and AlPO 4 , in LATP-LPO samples is supported by the EDS results (figure 5(b)), as well as findings from the literature on LATP prepared by L-PBF [19].The secondary phases, including TiO 2 and AlPO 4 , are found within the interdendritic regions, leading to a heterogeneous distribution.Closer observation of the BSE micrographs reveals an additional secondary phase (highlighted in figure 5(a)) that resides within the interdendritic region but is less distinct than the AlPO 4 and TiO 2 secondary phases.We hypothesize that this phase is residual LPO (highlighted in figure 4(a)), which was not fully vaporized during L-PBF processing.Similar features are also clearly observed in the other micrographs, such as those presented in figure 3.
Weka analysis was used to quantify the secondary phase concentration (area %) observed in the BSE micrographs through the application of machine learning [28].During Weka analysis, the user initially trains the system by manually identifying a selection of the different phases present in the materials system.For this study, each phase was identified by different grayscales in the BSE micrographs.After training the Weka analysis system, the model was applied to the remaining BSE micrographs and over 1500 secondary phase particles were measured for each sample.The measurement of such a large number of particles allowed for a confident assessment of the size and concentration of the secondary phases.
Weka analysis reveals that the concentration of secondary phases within the LATP-LPO samples varies with the initial LPO concentration (see figure 6(a)).The most notable difference is found in the 3 wt.% LPO sample.The addition of 3 wt.% LPO leads to an increase in the concentration of AlPO 4 and a reduction in the concentration of residual LPO, whereas the concentration of TiO 2 does not significantly differ from that observed in the 0 wt.% LPO sample.When the concentration of LPO is increased to 5 wt.% the    particles are similar across all LPO concentrations.Notably, the size distributions differ for larger particle sizes of both secondary phases.For instance, the D50 value for both AlPO 4 and TiO 2 increases with increasing LPO concentration up to 3 wt.% LPO, followed by a decrease in D50 for both phases at 5 wt.% LPO (figure 6(d)).We also note that the D50 values for TiO 2 are consistently larger than those for AlPO 4 , regardless of LPO content.The D90 value for AlPO 4 particle size increases with increasing LPO concentration, whereas there is not a clear trend for TiO 2 particles.Conversely, the 5 wt.% LPO samples possess similar particle size distributions to the 0 wt.% samples, indicating that the AlPO 4 and TiO 2 particle sizes have decreased compared to the 1 wt.% and 3 wt.% LPO samples (figure 6(d)).The Nyquist plots for the bulk LATP-LPO samples (figure 7(a)) show that the electrical behavior changes when the initial concentration of LPO is changed.All four Nyquist plots resemble diagonal lines with a slope of ∼45 • , indicating highly diffusive transport.We attribute the diffusive behavior to our use of ion-blocking Ag electrodes.However, the electrical transport mechanisms do not change as a function of initial LPO concentration, since all the Nyquist plot curves exhibit a similar shape.

Discussion
When fabricating LATP, lithium containing chemical species are commonly added in excess prior to high temperature processing to compensate for the high volatility of lithium [30].In this study, LPO was added in excess to LATP to serve as a sacrificial phase that preferentially vaporizes during L-PBF processing to reduce the amount of secondary phase formation.Reducing the concentration of secondary phases is beneficial for the electrical behavior since the deleterious secondary phases, AlPO 4 and TiO 2 , exhibit lower ionic conductivity values than LATP and impede charge transport.Here, we discuss the effect of excess LPO on the microstructure, phase formation, and functional properties of bulk LATP prepared using L-PBF.
When preparing samples using L-PBF, the evaluation of processing defects is critical for the performance as cracks and pores can influence structural robustness, secondary phase formation, and functional behavior.The bulk L-PBF LATP samples exhibit processing defects, including cracks and pores.Process parameters and materials properties commonly influence the propensity for processing defects, such as cracks and pores, in L-PBF samples.For instance, lack-of-fusion pores often occur when the laser input energy is inadequate, leading to improper bonding between each deposited layer [31].Lack-of-fusion pores generally appear as a cluster of pores that align parallel to the substrate, as seen in the cross-section of the LATP-LPO sample with 0 wt.% LPO (figure 3(a)).Further, materials properties commonly determine whether cracks form.L-PBF of materials with high stiffness, commonly observed in ceramic materials, often result in crack formation due to thermal shock [32].Thermal shock occurs when the yield strength of a material is overcome by the stress that results from changes in temperature due to thermal expansion and contraction [32].The layer-by-layer consolidation of material during L-PBF processing leads to iterative heating and cooling cycles [33].The microcracks found in the LATP-LPO samples indicate that the stress from the iterative temperature changes during L-PBF was higher than the yield strength of LATP, which led to thermal shock and the formation of cracks.The evidence of thermal shock within the bulk LATP-LPO samples (figure 3(b)) resembles that of thermal barrier coatings that similarly undergo iterative heating and cooling cycles [34].
Further, the grain morphology of L-PBF LATP samples likely contributes to the concentration of processing defects.Conventionally prepared, sintered LATP samples exhibit equiaxed cubic grains, whereas L-PBF LATP samples exhibit coarse columnar grains, which is expected due to the solidification behavior during L-PBF [31].The coarse columnar grain morphology of L-PBF LATP samples likely contribute to the concentration of processing defects.For instance, the development of coarse columnar dendritic grains in LATP-LPO samples can lead to competitive grain growth and increased concentrations of processing defects, including cracks and pores.Though columnar dendritic grains commonly form in metals processed with L-PBF without leading to crack formation [35], the brittle nature of LATP [36] leads to crack formation during the iterative expansion and contraction of the material during the heating and cooling cycles of L-PBF processing.Additionally, the average dendrite arm spacing increases with the concentration of LPO in the sample, which indicates that LPO induces grain coarsening in the L-PBF LATP samples.This corroborates findings from Aono et al that demonstrate grain coarsening in LATP samples with added LPO [37].Larger grain sizes, when coupled with competitive grain growth, are expected to result in higher concentrations of processing defects.The correlation between grain size and the concentration of processing defects holds for all LPO concentrations except 3 wt.% LPO.This indicates that something different may occur when 3 wt.% LPO is added.
In addition to changes in grain structure, L-PBF influences secondary phase formation.While L-PBF processing of LATP-LPO samples does not significantly influence the type of secondary phases found in LATP samples, the distinct grain morphology previously discussed (i.e., columnar dendritic grain morphology) influences the distribution of secondary phases.In the current study L-PBF LATP-LPO samples exhibit similar secondary phases (AlPO 4 and TiO 2 confirmed with EDS in figure 5) as glass-ceramic processed LATP samples [5], however the distribution of the secondary phases is more heterogeneous when using L-PBF processing due to the formation of columnar dendritic grains.Specifically, the equiaxed cubic-shaped grains in conventional glass-ceramic processed LATP samples exhibit homogeneously distributed secondary phases (i.e., AlPO 4 and TiO 2 ) at the grain boundaries [38,39].Alternatively, the columnar dendritic grains of the L-PBF LATP-LPO samples produce a heterogeneous distribution of secondary phases throughout the microstructure.For instance, secondary phases commonly precipitate out of the primary phase during rapid solidification, which occurs in L-PBF, due to constitutional supercooling [40], and segregate to interdendritic regions.The SEM micrographs of the LATP-LPO samples show that the secondary phases, AlPO 4 , TiO 2 , and residual LPO, exist within the interdendritic regions suggesting that constitutional supercooling occurred during L-PBF processing (figures 5(a) and (b)).
Higher concentrations of processing defects in L-PBF LATP-LPO samples are expected to inhibit functional behavior, as shown by a previous study that demonstrates lower density LATP samples exhibited lower ionic conductivity values [23].The addition of LPO into LATP influences the processability of L-PBF LATP-LPO samples, evidenced by the changes in the concentration of processing defects (see figure 3).For instance, the concentration of processing defects remains fairly consistent for all concentrations of LPO (0 wt. %, 1 wt.%, 3 wt.%) until a noticeable increase is observed in the LATP sample with 5 wt.% LPO.It should also be noted that there is a reduction in the standard deviation for the concentration of processing defect values for the LATP sample with 3 wt.% LPO.The difference in melting temperature between LATP (∼1400 • C [5,41]) and LPO (∼850 • C [42,43]), as well as the vaporization temperatures of the more volatile elements (Li ∼ 1350 • C, P ∼ 280 • C [44]), likely contributes to the observed behavior.The lower melting temperature of LPO improves sample density through liquid-phase sintering in conventionally processed LATP [23].However, L-PBF is a more dynamic process than conventional sinter-based processing.During the initial consolidation step of L-PBF (figure 8(a)), LPO is expected to preferentially vaporize from the laser-matter interaction zone (denoted by the blue arrow in figure 8(a)) since the vaporization temperatures of Li (∼1350 • C [44]) and (P ∼ 280 • C [44]) are lower than the melting point of LATP (∼1400 • C [5,41]).During the rapid solidification, however, residual LPO that has not vaporized will solidify last (due to its lower melting temperature) and will remain in the grain boundaries with the AlPO 4 and TiO 2 secondary phases.The presence of residual LPO at the grain boundaries will wet the grains and reduce the amount of micro-scale porosity.We hypothesize that this grain wetting behavior occurs in the LATP sample containing 3 wt.% LPO, indicating that a moderate amount of excess LPO can improve the variation in processing defects throughout the sample, evidenced by the reduction in standard deviation for this sample.Following solidification, the subsequent passes of the laser beam will reheat previously consolidated layers of material [33].As shown in figure 8(b), the residual LPO may undergo melting during the reheating step due to the lower melting point, whereas the LATP primary phase may not.The partial melting of residual LPO within previously deposited layers of material is expected to reduce the propensity for crack formation, as the molten LPO phase will not restrict expansion and contraction of the (brittle) solidified LATP during reheating and cooling.As such, the presence of LPO likely contributes to maintaining a lower amount of porosity and crack formation during the L-PBF process.Notably, the concentration of LPO must be tailored since higher concentrations of LPO (i.e. 5 wt.%) were observed to increase the concentration of processing defects, which is expected to negatively influence the functional behavior.
Since higher ionic conductivity values for solid-state electrolytes are desirable for enhancing battery performance, a correlation study (figure 9) was performed to determine which microstructural features most significantly influence the ionic conductivity of the bulk L-PBF LATP-LPO samples.The features compared in this analysis are: concentration of processing defects (Defects), concentration of residual LPO (Residual LPO), concentration of TiO 2 (Conc.Ti), concentration of AlPO 4 (Conc.Al), ionic conductivity (Conductivity), and Feret diameters of AlPO 4 (Feret.Al) and TiO 2 (Feret.Ti), respectively.Although the set p-value is 0.30 (meaning that there is a 30% probability that the event occurred by chance), the correlation heat map provides insight into the relationship between the microstructural features and functional behavior.For instance, there is a high positive correlation between the concentration of AlPO 4 and the ionic conductivity.Additionally, we observe a strong negative correlation between the crack density and the ionic conductivity, which is consistent with expectations, given that cracks are known to decrease the ionic conductivity in bulk ceramic materials [45].
It has been shown that the presence of AlPO 4 secondary phase in LATP contributes to the heterogeneous distribution of lithium and exhibits higher impedance values, which lower the ionic conductivity when present in LATP samples [48].However, AlPO 4 is also known to act as a sintering aid, which increases the sample density and in turn increases the ionic conductivity [48].As such, the influence of AlPO 4 depends on a balance between reducing the concentration of processing defects, which improves the ionic conductivity, and reducing the concentration of higher impedance secondary phase particles, which reduces the ionic conductivity.We observe a positive correlation between the concentration of TiO 2 and the ionic conductivity, although it is less strong than that for AlPO 4 .In general, the formation of higher impedance secondary phases is known to negatively influence the ionic conductivity in LATP samples, therefore it is not surprising that the secondary phases influence the ionic conductivity in the L-PBF LATP-LPO samples studied here.We also find that there is a positive correlation between the relative concentrations of AlPO 4 and TiO 2 .This correlation suggests that the formation of AlPO 4 and TiO 2 coincide during L-PBF processing and that it is difficult to avoid formation of one type of secondary phase without the other.Although there is not a direct correlation between the concentration of residual LPO (listed as 'Residual LPO' in figure 9) and the ionic conductivity, we do see that the amount of residual LPO left in LATP-LPO samples following L-PBF processing influences the concentration of processing defects.Finally, the Feret diameter of TiO 2 has a negative correlation with the concentration of processing defects and the concentration of residual LPO.This means that as the concentration of processing defects increases and the concentration of residual LPO increases, due to increased excess LPO added prior to L-PBF, the Feret diameter of TiO 2 will decrease.This correlation map demonstrates the complex relationship between different microstructural features and the ionic conductivity of the bulk LATP-LPO samples when processed using L-PBF.Notably, though, with process refinement and tailoring of the LPO concentration, we were able to sufficiently balance competing microstructural features and processing effects, which ultimately led to improved performance.Specifically, we see a ∼250% increase in ionic conductivity values for the 3 wt.% LPO sample, compared with the other LATP-LPO samples.As such, further development of LATP-LPO samples prepared with L-PBF may exhibit further improvements in performance through careful refinement of process parameters and chemical composition.

Conclusions
This study demonstrates successful binderless additive manufacturing of bulk solid-state electrolyte LATP-LPO samples using L-PBF.The influence of chemical composition on the microstructure and performance of the solid-state electrolyte, LATP, was investigated.Specifically, LPO was blended with LATP powder prior to L-PBF, to serve as a sacrificial source of lithium during processing.We demonstrated that preparation of three-dimensional bulk samples with the desired rhombohedral crystal structure is achievable using blended LATP and LPO powders during L-PBF.While processing defects, such as cracks and pores, are present in all LATP-LPO samples, robust three-dimensional samples were producible.Microstructural analysis reveals that L-PBF processing led to the development of columnar dendritic LATP grains, as expected for this technique.Further, LPO serves as a grain modifier resulting in larger average grain diameters with increasing LPO concentration in the bulk LATP-LPO samples.
To explore the complex relationship between these microstructural features and the functional behavior of these LATP-LPO samples, a correlation map was created and discussed.We identified that the concentration of secondary phases, AlPO 4 and TiO 2 , influence the ionic conductivity of LATP-LPO samples prepared with L-PBF.We expect that these concentrations of secondary phases were influenced by the LPO concentration.As such, the manipulation of chemical composition can lead to a complex relationship between microstructural features and functional behavior.Further, this study shows the considerations required when designing for both processability and performance while enhancing the geometric complexity of functional ceramics for solid-state battery applications.Given our observed importance of chemical composition and materials selection on sample quality in L-PBF, future studies will focus on exploring materials with low isotropic thermal expansion values in order to improve thermal shock resistance and inhibit crack formation.Additionally, we plan to explore eutectic-forming materials that promote the simultaneous solidification of nano-scale microstructure features that promote a more homogeneous distribution of secondary phases.

Figure 2 .
Figure 2. X-ray diffraction patterns for: (a) the blended LATP (same indexing as LiTi2(PO4)3) and LPO (Li3PO4) powders, and (b) the as-deposited bulk LATP-LPO samples.The most prominent LPO diffraction peaks are highlighted with the asterisk data markers.

Figure 3 .
Figure 3. Representative optical micrographs of the cross-sections (a) normal to the build direction (BD), and (b) in the plane of the build direction show that LATP-LPO samples with 0 wt.% LPO display processing defects, including cracks and pores.(c) The concentration of processing defects, quantified using % porosity and crack density, are displayed as a function of LPO concentration.

Figure 4 .
Figure 4. Representative backscattered electron (BSE) micrographs of bulk LATP-LPO samples with (a) 0 wt.%, (b) 1 wt.%, (c) 3 wt.%, and (d) 5 wt.% LPO.(e) The dendrite arm spacing, highlighted by the double-sided white arrow in (a), is provided as a function of LPO concentration.The white arrow in the bottom left corner of (a) displays the build direction (BD) for all micrographs.

Figure 5 .
Figure 5. (a) Representative backscattered electron (BSE) micrograph and (b) energy dispersive x-ray spectroscopy (EDS) results for a bulk LATP-LPO sample with 3 wt.% LPO; AlPO4 (darker grayscale) and TiO2 (lighter grayscale) secondary phases within interdendritic regions are observed (highlighted by the white arrows and labeled in (a)).Residual LPO is also highlighted.The white arrow in the bottom left corner of (a) denotes the build direction (BD) for all micrographs.

Figure 6 .
Figure 6.(a) Weka analysis generated data for the concentration of secondary phases (% area) in bulk LATP-LPO samples as a function of LPO concentration (wt.%).Representation of cumulative distributions in the Feret diameters for the (b) AlPO4 and (c) TiO2 secondary phase particles.(d) Particle Feret diameter, D50, as a function of LPO concentration.

Figure 7 .
Figure 7. Room temperature (a) Nyquist plots and (b) ionic conductivity values for the bulk LATP-LPO samples with different concentrations of LPO.The dashed red line in (b) marks the conductivity of the 0 wt.% LPO sample, to which the conductivity of the other samples are compared.Note the logarithmic scale used for the y-axis in (b).

Figure 8 .
Figure 8. Schematic representation of vaporization and phase evolution dynamics during (a) deposition of layer 1 and (b) deposition of layer 2 during laser powder bed fusion processing.During initial consolidation of powder (a), LPO vaporizes from the melt pool as LATP solidifies via constitutional supercooling which leads to the development of secondary phases at interdendritic regions.Upon deposition of the next layer of material (layer 2) shown in (b), the previously consolidated layer (layer 1) undergoes reheating that melts residual LPO that resides between LATP dendrites.

Figure 9 .
Figure 9. Correlation heat map for bulk L-PBF LATP-LPO microstructural features.The features analyzed include: concentration of porosity (Porosity), crack density (Crack Density), concentration of residual LPO (Residual LPO), concentration of TiO2 (Conc.Ti), concentration of AlPO4 (Conc.Al), ionic conductivity (Conductivity), and the Feret diameters of AlPO4 (Feret.Al) and TiO2 (Feret.Ti, respectively).The size of each circle reflects the intensity of the correlation, while the positive/negative numbers and the blue/red colors indicate direct/inverse correlations, respectively.

Table 1 .
Laser powder bed fusion process parameters used to prepare LATP-LPO samples.

Table 2 .
Percentile values (D10, D50, and D90) for the Feret diameter cumulative distributions of the AlPO4 and TiO2 secondary phase particles within bulk LATP-LPO samples with different LPO concentrations.

table 2 .
The percentile values are also provided in table 2. The data presented in figures 6(b) and (c) and table 2 indicate that for smaller sizes (i.e.D10) the distribution in AlPO 4 and TiO 2

Table 3 .
Comparison of ionic conductivity values for LATP and LATP-LPO prepared through a variety of processing methods.