Gas-Induced Structural Damages in Forward-Bias Bipolar Membrane CO2 Electrolysis Studied by Fast X-ray Tomography

Forward-bias bipolar membrane (BPM) CO2 coelectrolysis (CO2ELY) aims at overcoming the issues of salt precipitation and CO2 crossover in anion exchange membrane CO2ELY. Increasing the stability of BPM-CO2ELY is crucial for widespread application of the technique. In this study, we employ time-resolved X-ray tomographic microscopy to elucidate the structural dynamics that occur within the electrochemical cell during operation under various conditions. Using advanced image processing methods, including custom 4D machine learning segmentation, we can visualize and quantify damages in the membrane and anode catalyst layer (CL). We compare our results to a CO2 transport model and hypothesize gaseous CO2 as the cause of the observed damages. At any operation condition, CO2 is formed at the junction in the center of the BPM by recombination of carbonate ions. CO2 migrates to the anode by diffusion and goes into the gas phase at the interface of the membrane and anode CL. After sufficient CO2 accumulation and pressure buildup after only tens of minutes, small irreversible holes break into the CL distributed over the entire active area. Additionally, at higher current densities, the CO2 accumulation leads to membrane delamination at the BPM junction. Despite the clear degradation processes, we do not observe an obvious direct effect on the electrochemical performance. The poor stability of BPM-CO2ELY remains an open question.


S3 Catalyst layer image filter
The metallic catalyst layers appear very bright in XTM due to their high X-ray absorption.In order to darken the corresponding image pixels for better readability and highlight the catalyst layer edges, the grayvalue G(x, y) of a pixel at position x, y is reduced to Ĝ(x, y) using equation S1.Ĝ(x, y) = G(x, y) The image filter is inspired by the Butterworth Lowpass filter, but applied directly on the real space instead of the Fourier space.Bright pixels are mostly affected, while grayvalues way below the critical grayvalue g c remain unchanged.n controls the roll-off and s is a value between 0 and 1 describing the strength of the filter.Values are manually chosen as g c = 14500, n = 20 and s = 0.265.

S7 Catalyst layer damage for a cell with Ti porous transport layer at anode
To test the applicability of our result to the typical co-electrolysis design, the experiment was repeated qualitatively with a Titanium porous transport layer (Ti PTL) at the anode using a tube X-ray cone beam CT scanner (Phoenix nanotm m, General Electric, Germany).The experimental setup is the same as for time-resolved XTM at the synchrotron except for the replacement of the carbon paper GDLs as anode PTL with two stacked 0.25mm thick Ti felts (2GDL10-0.25,Bekaert, Zwevegem, Belgium) and a thicker (0.47mm, FC-FKM 200, Freudenberg) ice cube gasket.The X-ray tube was set to 80kV acceleration voltage and 230µA.A 0.1mm Cu filter was placed into the beam to filter out low energy photons and reduce metal artifacts.1000 radiographic projections were recorded equally spaced over one full rotation of 360°.Each projection is the average of three frames at the same angle with an exposure time of 0.5s each.The acquisition of one frame was skipped while the stage is moving to the next angular position.The sample was placed at a distance of 12mm to the X-ray source while the flat panel detector had a distance of 400mm to the source capturing the entire width of the cell in the field of view and resulting in a voxel size of 3µm.Tomographic reconstruction was performed using the implementation of the Feldkamp-algorithm for cone beams in the commercial software datos|rec (Waygate Technologies, Bake Hughes, Houston, USA).One tomographic scan was taken after cell assembly and a second scan after operating the cell for 30min at 100mA/cm 2 while all other parameters were kept the same.The tubing and wiring were disconnected from the cell before taking the after-operation scan.Figure S10 shows exemplary tomographic slices that contain qualitatively the same degradation process as observed by the synchrotron XTM including cavities within the BPM and between BPM and anode CL as well as anode CL perforations.It is not possible to obtain time-resolved XTM due to the longer (>30min) acquisition for one scan.The image quality at the given labCT settings is furthermore lower compared to the synchrotron XTM with a larger voxel size, higher noise and lower contrast.Additionally, artifacts can be observed including black streaks and blurring of the catalyst layers ("catalyst shining") most likely due to the interaction of X-ray with heavier elements (Ir, Ag, Ti).It was consequentially not possible to employ the same quantitative image analysis as for the synchrotron XTM data.However, the qualitative analysis gives confidence in the general applicability of our results to CO 2 co-electrolysis.
S-9 We devise a model for the CO 2 transport to discuss the obtained operando imaging results.The BPM in forward bias prevents direct CO 2 crossover to the anode in the form of (bi)carbonate ions by reaction at the BPM-junction with protons to CO 2 and water: However, CO 2 and water can accumulate over time at the junction.We expect that the formation of gaseous CO 2 at sufficient concentration causes BPM delamination (figures 4-7) and damage to the anode CL (figures 4 and 8).An analytical model based on Fickian diffusion is used for first order approximation of CO 2 mass fluxes and concentrations at various locations in the MEA (figure S11) to complement the experimental observations of structural damages induced by gaseous CO 2 .CO 2 is produced at the BPM junction at a rate ṅint mol/cm 2 s where I mA/cm 2 is the current density, e [C] is the elementary charge, N A [1/mol] is the Avogadro number and f CO2 is the fraction of produced CO 2 per charge carrier.The formation of CO 2 depends on the different charge carriers in the AEM part of the bipolar membrane (reaction equation S2), i.e.HCO 3 results in 1 molecule of CO 2 at the junction, while CO 3 2-in 0.5 CO 2 and OH -in no CO 2 molecule per electron.The charge carrier distribution in the AEM part is complex and needs to be estimated based on other modeling work.We assume that all charge carriers are CO 3 2-in accordance to the current understanding of AEM [8,2,11,7], i.e. f CO2 = 0.5.For AEM, it is expected that with increasing current density, the CO 3 2-concentration is decreasing in favor of higher OH − concentration [8,11] and a recent study [10] shows an equivalent behavior of the BPM-AEL compared to a pure AEM [11].The accumulating CO 2 at the junction leads to a concentration gradient (c i − c j ) and diffusive fluxes j ij towards the cathode and anode, respectively, described by Fick's first law where the subscripts i, j denote the layer interfaces 1-5 (figure S11), D 0 m 2 /s is the diffusivity of CO 2 in the transport medium, ϵ the volume fraction of the transport medium, τ the tortuosity of the transport medium, c mol/m 3 the CO 2 concentration at the interface and ∆ [m] the layer thickness.Accumulating CO 2 reaches the solubility limit and small pockets of gaseous CO 2 form at the interfaces, i.e. at the BPM junction and the membrane-CL interfaces.We consider virtual, zero-thickness, gas volumes at the interface with gas pressure p i [P a].The CO 2 concentration c i of dissolved CO 2 at the interface is assumed to be in equilibrium with the gas phase.Dissolved CO 2 concentration and partial pressure are then related by the ideal gas law, which is approximately valid for CO 2 dissolved in water by a dimensionless Henry's number close to 1 [9].
While CO 2 at the anode is instantly carried away by the water stream resulting in zero CO 2 pressure p 5 , the partial CO 2 pressure at the cathode p 3 is known as where p amb is the ambient pressure and p vap is the water vapor pressure depending on the set temperature T and relative humidity of the inlet CO 2 stream.Considering the conservation of mass and inserting (S5) in (S4) allows us to calculate the CO 2 crossover to the anode j an , the back-diffusion to the cathode j cat and the CO 2 pressure p 1 at the BPM junction at steady state.0 = ṅint + j cat + j an (S7a) The CO 2 flux is also influenced by the pressure drop across the cathode and anode CLs, which are calculated considering the permeability of the porous catalyst filled with gas (cathode) and water (anode) respectively.Since the crossover flux to the anode is equal to the flux across CEL and ACL, we can calculate the pressure drop at the anode CL and analogous for the cathode CL.
The employed parameters are given in table S1.The model results for I = 100mA/cm 2 are given in table S2.

Figure S1 :Figure S2 :
Figure S1: Drawings of the polar plates displaying the internal routing of the gas/water streams to the integrated flow field.Symmetric except for the screw thread.a) cathode, b) anode

Figure S3 :
FigureS3: Current density (I, black) and voltage (E, red) over time (t) for all tested samples.Vertical gray lines denote tomographic scans.t=0 corresponds to the end of the conditioning step.

Figure S5 :
Figure S4: Figure 4 in main text without filter.

Figure S6: Sample 7 :
Figure S6: Sample 7: Visualization of segmented anode catalyst layer (top row) and membrane cavities (green, bottom) for the period of decreasing "surface cavity" volume (≈4-15min) in figure 6a of the main text.

Figure S7 :Figure
Figure S7: Charge (black) and damaged anode catalyst area (red) over time for all tested samples.

Figure S10 :
Figure S10: Tomographic slice of assembled cell with Ti fiber felt porous transport layer at anode.a) before operation, b) after operating at 100mA/cm 2 for 30min with annotation: 1) carbon paper GDL, 2) cathode CL, 3) BPM, 4) anode CL, 5) Ti fiber, 6) cavities between BPM and anode CL, 7) anode CL hole, 8) mobilized piece of anode CL. c,d) Additional location at different slice showing anode CL perforation and membrane delamination.e,f) Close-up of white frame in a) and b)

Table S2 :
Calculated CO 2 mass fluxes and pressures within the MEA for I = 100mA/cm 2 .