Office of Research and Development Nanoscale X.ray Computed Tomography of Solid Oxide Fuel Cell Electrodes William K EptingU 1.3 Shawn LitsterU 3 Kirk GerdesU 1 Paul A Salvador 1U2 1 National Energy Technology LaboratoryU UDSD Department of Energy; 2 Dept of Materials Science and EngineeringU Carnegie Mellon UnivD; 3 Dept of Mechanical EngineeringU Carnegie Mellon UnivD Acknowledgment: "s part of the National ±nergy Technology µaboratory5s research portfolioC this work was conducted under the R±S contract ?±v%±222*222 Using a Spherical Dilating Kernel of R = W: Ni µarge pores NivYSZC pores YSZC no pores µSMvYSZC pores µSM µarge pores (2 μm V2 μm V2 μm 82 μm ~(22 μm Hathode current collector "ctive cath. "ctive anode ±lectrolyte "node support RP8 Hoarser kernels RPx Spherical kernel N8: kernel Hubic kernel Starting voxel Dilate YSZ µSM Void YSZ1 µSM Void YSZ1 µSM1 Void1 YSZ1 µSM Void1 Void µSM YSZ Void YSZ YSZ mbright3 Ielow resolution limit Iorder region Raw virtual x? slice IutC the real structure looks like this• Void mdark3 Void µSM YSZ Segmentation greyscales based on real volumes of µSM elsewhere Removal Method: Morphological Dilation ThreexPhase Artifact Artifact removal by sequential dilation µSM artifact coating all YSZ mthere should be YSZvpore interfaces3 µSM artifact even coating electrolyte layer mnot physical3 Dilation: Different kernels and sizes Studying Dilation Parameters for Artifact Removal Reasonable outcomes Unreasonable outcomes Untreated Is it physical? Effect on phase frac? Effect on pore contact? Effect on phase percolation? Kernel: cube Radius: 3 • All features look very “cube<ified” • LSM reduced to single isolated cubes Virtual /D slice Thinnest LSM skins gone Remaining LSM thick enough to be real Qualitatively, shapes faithful to orig. data LSM covering all YSZ < not physical Void YSZ LSM Virtual /D slice Long limbs present% smaller than nano<CT’s resolution limit Kernel: faces only Radius: 2 LSM YSZ Void Z coord. [μm] Phase fraction LSM Fraction > Pw, Not likely true. LSM likely getting boost in volume from artifact LSM Void Phase fraction LSM <3z, seems unreasonably low Cube% radius x LSM YSZ Void Sphere% radius / Z coord. [μm] zw, LSM seems high% but possible mHommercial composition unknown3 LSM YSZ Void Sphere% radius x Z coord. [μm] xw5xw5xw, mix also seems reasonable LSM YSZ Pore exposure LSM<pore exposure has taken a large hit Cube% radius x LSM YSZ Sphere% radius / Z coord. [μm] YSZ<pore exposure has come from ~w to a measurable amount% LSM<pore exposure still sizeable LSM YSZ Sphere% radius x Z coord. [μm] YSZ<pore and LSM<pore very even now% though TPB may be about the same as above LSM YSZ Z coord. [μm] Pore exposure YSZ has virtually w contact with pores1 Unprocessed Sphere% R=/ Sphere% R=x Cube% R=x Kernel: sphere Radius: 2 YSZ connectivity pretty reasonable to begin with. Void connectivity lowC but some dead pores expected. Void connectivity higher v closer to expected characteristics. "ll three phases have excellent connectivity. µSM connectivity has been severely diminished by overvdilation. Radiograph ±lectrolyte V2 μm V? reconstruction Virtual x? slices Raw Segmented Nanoscale Xxray CT of Solid Oxide Fuel Cells Hathode materials• µSM YSZ µanthanum strontium manganateC electron conductor = catalyst Yttriavstabilized zirconiaC oxide conductor %OV X-ray CT with 65 nm resolution LD Spatial Analysis of TPB Distribution Shaftvlike 0dead zone0 Heterogeneity Analysis Ongoing Work: Advanced Sample Prep for LargerO More Representative Samples m* μm3 V TPIs determined vertexvbyvvertex in V? Vetices connected to orthogonal neighbors Paths smoothed for accurate length measurement µocal TPI length stored on voxelvbyvvoxel basis for V? spatial analysis 82x82x( μm region selected for heterogeneity analysis mbelow3 TPB density "cloud" visualization Multiple cubic ROqs placed within larger volume +ow much does TIP density differ between each cubeC as the cubes become smallerz Typ . %qIvS±M size ±*2Z deviation from mean TPIC based on U(Z confidence interval on Tvdistribution µarger volumes mV2~ μm3 required to accurately characterize more heterogeneous industrial cells v nanovHT is capable N P V) N P 8* NP* qdeal sample 0%racture and hope0 method may favor statistically unusual samples• 0why did it break therez0 qn factC this sample is more homogeneous than many other samples previously attempted in %qIvS±MC which exhibited large superporesC or *~ μm boulders of solid material. Cut-and-polish method ?irection of polishing Hut Polish Micrometerv controlled fixture Trim with laser ±xtract pillar Mount vertically for nanovHT Abstract Understanding performance degradation in SO%Hs requires an understanding of electrode microsctructure Prior methods yield small fields of view below scale of eterogeneity present in some cells Nanoscale Xvray HT can yield V? microstructure of a larger volume Here, we present imaging, artifact removal, and analysis of the scale of heterogeneity in industrial SOFC electrodes Virtual x? slice from V? image 82 μm µSM YSZ Pore ±lectrolyte 2.x 2.* 2.) 2.: 2 ( 82 8( x2 x( YSZ µSM Pore Z coord [μm] Vol fraction Phase fraction analysis Hurrent collector Phase connectivity analysis Analysis of larger cathode volume 445 x 32 x 25 μmx µSM YSZ Void 2 2.x 2.* 2.) 2.: 8 µargest contiguous frac. z