0
Research Papers

# Neutron Imaging and Electrochemical Characterization of a Glucose Oxidase-Based Enzymatic Electrochemical CellOPEN ACCESS

[+] Author and Article Information
Ryan S. Longchamps

Department of Mechanical and
Aerospace Engineering,
University of Alabama in Huntsville,
301 Sparkman Drive,
Huntsville, AL 35899
e-mail: rsl0002@uah.edu

Zachary K. van Zandt

Department of Mechanical and
Aerospace Engineering,
University of Alabama in Huntsville,
301 Sparkman Drive,
Huntsville, AL 35899
e-mail: zkz0001@uah.edu

Hassina Z. Bilheux

Chemical and Engineering Materials Division,
Oak Ridge National Laboratory,
P.O. Box 2008,
Oak Ridge, TN 37831
e-mail: bilheuxhn@ornl.gov

Indu Dhiman

Chemical and Engineering Materials Division,
Oak Ridge National Laboratory,
P.O. Box 2008,
Oak Ridge, TN 37831
e-mail: dhimani@ornl.gov

Louis J. Santodonato

Instrument and Source Division,
Oak Ridge National Laboratory,
P.O. Box 2008,
Oak Ridge, TN 37831
e-mail: santodonatol@ornl.gov

Yevgenia Ulyanova

Hexcel Corporation,
3300 Mallard Fox Drive,
Decatur, AL 35601
e-mail: jenny.ulyanova@hexcel.com

Sameer Singhal

CFD Research Corporation,
701 McMillian Way NW,
Huntsville, AL 35806
e-mail: ss2@cfdrc.com

George J. Nelson

Department of Mechanical and
Aerospace Engineering,
University of Alabama in Huntsville,
301 Sparkman Drive,
Huntsville, AL 35899
e-mail: george.nelson@uah.edu

1Corresponding author.

Manuscript received May 31, 2017; final manuscript received September 12, 2017; published online November 7, 2017. Assoc. Editor: Partha P. Mukherjee. The United States Government retains, and by accepting the article for publication, the publisher acknowledges that the United States Government retains, a nonexclusive, paid-up, irrevocable, worldwide license to publish or reproduce the published form of this work, or allow others to do so, for United States Government purposes.

J. Electrochem. En. Conv. Stor. 15(1), 011007 (Nov 07, 2017) (10 pages) Paper No: JEECS-17-1059; doi: 10.1115/1.4038244 History: Received May 31, 2017; Revised September 12, 2017

## Abstract

Enzymatic electrochemical cells (EECs) are a candidate for providing “green” solutions to a plethora of low-power, long-lifetime applications. A prototype three-electrode biobattery configuration of an EEC has been designed and fabricated for neutron imaging and electrochemical testing to characterize cell performance. The working electrode (WE) was catalyzed by a polymer ink-based biocatalyst with carbon felt (CF) serving as the supporting material. Results of both ex situ and in operando neutron imaging are presented as methods for relating fuel distribution, the distribution of the enzymes, and cell electrochemical performance. Neutron radiography (NR) was also performed on fuel solutions of varied concentrations to calibrate fuel solution thickness and allow for transient mapping of the fuel distribution. The calibration data proved useful in mapping the thickness of fuel solution during transient radiography. When refueled after electrochemical testing and neutron imaging, the cell surpassed its original performance, indicating that exposure to the neutron beam had not detrimentally affected enzyme activity. In operando mapping of the fuel solution suggests that increased wetting of the catalyst region increases cell performance. The relation of this performance increase to active region wetting is further supported by fuel distributions observed via the ex situ tomography. While useful in mapping aggregate solution wetting, the calibration data did not support reliable mapping of detailed glucose concentration in the WE. The results presented further demonstrate potential for the application of neutron imaging for the study of EECs, particularly with respect to mapping the distribution of aqueous fuel solutions.

<>

## Introduction

For electrochemical energy conversion and storage devices, natural processes provide a source of inspiration for novel cell chemistries and architecture. Such biomimetic design has resulted in the development of the enzymatic fuel cell (EFC), which employs a cell architecture comparable to the polymer–electrolyte–membrane fuel cell (PEMFC), utilizing the catalytic capability of enzymes to support the anodic and cathodic activities in an electrochemical cell that consumes organic fuels (i.e., sugars and alcohols) in a fashion similar to the processes that occur in the human body [1]. Though the EFC was first conceived as a fuel cell arrangement, the technology can more generally be referred to as an enzymatic electrochemical cell (EEC), of which there are two configurations: fuel cell and battery. The fuel cell arrangement utilizes a constant flow of fuel and oxidant into the anode and cathode, respectively, with outflows of the byproducts of the oxidation and reduction reactions flowing out of the respective electrodes. In a biobattery configuration, the cell is provided an initial amount of fuel solution during the “charging” process and “discharges” until the cell has consumed all available fuel [2]. While most device-level research has focused on the EFC arrangement, the biobattery configuration shows promise for remote applications of the technology.

The catalytic activity of certain enzymes in the presence of sugar was first observed in 1911 by Potter [3], and the application of this trait was first employed in a fuel cell architecture in 1964 by Yahiro et al. [4]. Since then, many challenges have been discovered in working toward making this technology competitive with comparable electrochemical systems for use in a variety of applications, such as implantable biosensors or portable electronic device primary power. The first of the challenges the research community has faced is the immobilization of the enzymes in either the cathode or anode for constrained operation [5,6]. Furthermore, two methods for electron transfer from the anodic enzyme to the current collector have been employed: direct electron transfer and mediated electron transfer, both possessing advantageous traits [5,79]. An additional challenge in advancing EFC technology is the development of various chemistries in combination with different enzymes or enzyme cascades to increase cell performance and expand the range of applicable fuels [10]. Possibly, one of the most important factors limiting the production of an end-use EFC is the lack of understanding of the impact of various operating conditions [2]. The elements of the operating environment that strongly impact performance are the pH, temperature, and humidity/saturation level in certain regions of the cell [1,1113]. When enzymes are exposed to extreme conditions with respect to these environmental characteristics, a process called denaturation occurs, in which the complex, twisted geometry of the molecular complexes comprising the enzyme unravels and renders the enzymes inactive [14,15]. Thus, it is highly important to understand the effects of these environmental parameters on the enzymes and the EFC as a whole. Given the limited in situ investigation into the water content within EFCs as well as some of the reaction-based dynamics that occur within the cell during operation, techniques for investigation on environmental conditions on EFC performance are needed [8]. It has been demonstrated that neutron imaging can be effective in tracking the wetting conditions within a PEMFC [1619]. In 2008, Weber and Hickner presented a qualitative comparison between the water content profile across a PEMFC measured with neutron imaging and predicted via simulation [19]. Also using neutron imaging, Cho et al. investigated the water content within a prototype PEMFC undergoing relative humidity-controlled gas purge for preservation of cell components during a freeze/thaw cycle [16]. These neutron imaging techniques were applied to an EFC in a similar fashion in the work presented by Aaron et al. [13]. This work utilized control of feed gas humidity level and flow rate to observe variations in cathode and membrane wetting during operation. Ultimately, it demonstrated the ability of neutron radiography (NR) to qualitatively depict the wetting conditions within an EFC [13]. In 2015, Looney et al. also used NR to demonstrate the ability of the technique to track wetting conditions within a prototype pouch cell during operation. In addition, the ability to map the distribution of the anodic biocatalyst was observed through neutron tomography (NT) due to the distinct difference in attenuation properties of the carbon felt (CF) electrode and polymer network containing the enzymes [2]. The pouch cell design that was implemented rendered repeatable postprocessing of image data difficult due to the lack of consistent geometry between samples. Furthermore, the cell architecture did not provide sufficient sealing during in situ NR, leading to the varying transmission behavior primarily as a result of electrode drying rather than the presence of electrochemical activity. Thus, it was suggested that future prototype cell designs would benefit from a more rigid geometry with proper sealing for containing the fuel solution.

In addition to these studies of electrochemical cells using neutron imaging, a study of water saturation in a porous media was conducted by Kang et al. in 2013 using neutron imaging techniques [20]. In this work, the presence of water in moist flint sand was tracked via neutron radiography by correlating the transmitted intensity to the thickness of water the neutron beam using calibration cells containing only water. This method was validated by calculating the presence of water in the flint sand sample through the correlation and comparing the results to the known presence of water, controllable through the experimental apparatus [20].

Incorporating many of the techniques utilized by Kang et al., the present investigation employs neutron imaging as a technique for demonstrating the effects of the distribution of aqueous fuel solution on the electrochemical performance within a prototype, biobattery half-cell configuration of an EEC. With the implementation of a rigid, three-electrode prototype cell with a machined aluminum housing, the previously demonstrated ability to map the biocatalyst region via NT was harnessed to inform the analysis of the cell performance in conjunction with the cell wetting around the active region. The fuel distribution throughout the working electrode (WE) was determined using ex situ NT for two wetting cases with the electrochemical activity of the cell being assessed through both cyclic voltammetry (CV) and chronoamperometry (CA). Using the calibration method developed by Kang et al., the aqueous fuel solution content of the cell was also mapped during in operando NR performed during cell discharge. These results are compared to those obtained from the ex situ computed tomography (CT) scans. These results support the applicability of neutron imaging techniques for observing aqueous the fuel solution in EFCs and the effects of localized wetting on cell performance.

## Methodology

###### Cell Design and Preparation.

The prototype cell used for all experiments was designed based on the need for a rigid geometry and relative transparency to neutrons. Aluminum was selected as the construction material, and the primary structural components were machined in-house. An overall schematic of the cell is presented in Fig. 1(a). As indicated in the figure, the WE of the test cell was housed in the smaller cavity of the holder (6.4 mm × 15.8 mm × 4.8 mm). The 6.4 mm thickness of the cavity parallel to the neutron beam was chosen to minimize the attenuation path of the neutrons. This decision provides the secondary benefit of reducing the radioactive decay time of the fuel and electrode, allowing faster sample manipulation between scans. The counter electrode (CE) (12.7 mm × 38.1 mm × 4.8 mm) was designed to be significantly larger than the WE to minimize limitations of the performance of the cell by providing sufficient pathways for electronic transport. The two electrodes of the cell were separated by two pieces of filter paper with a reference electrode contained between the separators. Carbon yarn obtained from Alfa Aesar (Ward Hill, MA) coated in silver ink obtained from Ted Pella, Inc., Redding, CA, was used for the reference electrode (∼90 mV versus Ag/AgCl). A rectangular piece of Teflon was placed in between each of the filter paper separators and the electrode holders to act as a sealing gasket in an effort to minimize leakage of aqueous fuel and buffer solutions from the cell. This gasket also provided an electronically insulating barrier between the WE and CE holders. To expose each electrode to the filter paper, allowing for ionic transport within the prototype, a smaller rectangular portion of the Teflon was cut out of the center of each gasket approximately corresponding to the respective area of each electrode. The entire assembly was held together by glass-filled nylon screws that were placed above and below the electrode holder cavities so as to not interfere with the imaging of the regions of interest (ROI) within the cell. Finally, aluminum tape was adhered to the outer surfaces of the electrode holder cavities, providing connection points for electrochemical testing. The WE holder, CE holder, and the mounting bracket were machined from 6061 aluminum bar stock sourced from McMaster-Carr (Elmhurst, IL). The electrodes of the cell were comprised of one primary material, 99% purity carbon felt which was sourced from Alfa Aesar. In the WE, this served as a structure to which the enzymes could be attached. The thickness of the felt used throughout the cell was 3.18 mm. All carbon felt was plasma treated for 5 min to make the inherently hydrophobic CF hydrophilic [21]. In the case of the working electrode, two pieces were used—one loaded with enzymes and one without. Figure 1(b) shows the two cell components disassembled with the carbon felt inserted into the electrode wells.

The present work focuses on testing an EEC anode operating on an aqueous solution of glucose fuel containing 0.243 M sodium phosphate and a 0.1:1.0 M mixing ratio of hydroquinone to glucose. Oxidation of the glucose in the fuel solution follows the reaction R1, catalyzed by the glucose-oxidase (GOx) enzyme. Future studies may address the performance of EEC cathodes. Display Formula

(R1)$C6H12O6+H2O→C6H12O7+2H++2e−$

The enzymes were introduced and immobilized on the CF using the proprietary carbon nanotube, polymer ink mixed with GOx developed by CFDRC [2,22]. After combining the components of the enzymatic ink, a gentle vortexing process was carried out to achieve a homogeneous mixture. The enzyme loading was 36.8 mgmL−1 with a total of 0.85 mL of enzyme mixture being applied to the piece of carbon felt that would provide catalytic activity within the WE. After introducing the enzyme solution onto the CF, the electrodes were stored at 4 °C for a minimum of 12 h. Prior to operation, the anodes were removed from the subambient temperature environment and allowed to naturally come to equilibrium with the environmental test conditions (i.e., room temperature, atmospheric pressure).

###### Experimental Process.

All NR and NT was performed at the CG-1D beamline at the Oak Ridge National Laboratory (ORNL) High Flux Isotope Reactor. The wavelengths present in the neutron flux at the CG-1D facility range from 0.8 to 6.0 Å, peaking at approximately 2.6 Å. Furthermore, all images were obtained employing the LiF/ZnS scintillator in combination with a CCD camera. This configuration provides a 7.4 cm × 7.4 cm field of view, sufficient for the prototype cell with a maximum projected profile of 5.4 cm × 2.9 cm. Furthermore, the CCD camera provides a ∼100 μm spatial resolution with a 36 μm pixel size for the sample position applied during imaging [2]. It is noted that an exposure time of 50 s was applied for all images captured for this study.

After capturing the initial alignment images, a computed tomography scan of the dry test cell was performed prior to the introduction of fuel. This scan was performed over a total angular sweep of 183 deg with an angular step of 0.3 deg for a total of 610 images. At approximately 60 s per image (including exposure and file writing times), the total duration of the CT scan was approximately 10 h. Following the CT scan of the dry cell, ten open beam (OB) and dark field (DF) images were taken for the purpose of image normalization during postprocessing. Ten OB and DF images were taken each day for normalization of sample images taken on each respective day. OB images were captured with the CCD exposed to the neutron beam with no sample in the field of view, while DF images were captured with the shutter closed. Following the first set of OB and DF images, the cell was opened and the fuel solution was applied with a pipette until both electrodes appeared to be saturated. The fuel solution applied during all cell operation was 0.1 M glucose in 0.243 M sodium phosphate with a mixing ratio of 0.1:1.0 M hydroquinone to glucose. A second CT scan was then performed to map out the distribution of the fuel within the cell. The scan parameter was again 183 deg of total angular sweep with an angular step of 0.3 deg. This scan was followed by two sets of in operando NR scans which corresponded to the two types of chronoamperometry to be discussed in more detail later in this section. For the first set of radiographic scans, ten images were taken before the respective electrochemical tests were performed, and at least ten images were taken after the completion of each respective electrochemical test. During the entire duration of each scan, images were taken at an approximate capture rate of one image per minute.

After the completion of the first round of neutron imaging, it was observed that portions of both the WE and the CE were not fully wetted. In an attempt to better saturate the cell and observe comparable data, the cell was opened, additional fuel was applied, and then the cell was reassembled with the same CF electrodes that were used in the first wetting. A second round of neutron imaging tests were then performed for comparison to the initial wetting, beginning with the in operando NR. These images were captured in exactly the same manner as for the first wetting. A final CT scan was setup to run after the NR to allow for mapping of the fuel distribution for the second wetting with the scan parameters being the same as all previous CT scans.

Following the CT scan of the cell after the first wetting, an initial CV scan was performed with a scan rate of 100 mVs−1 and upper and lower voltage limits of 0.5 and −0.8 V versus Ag/AgCl, respectively. Then a “staircase” CA scan was performed. The term staircase is introduced to describe the nature of the potential that the cell is subjected to during the test. Specifically, the voltage is held at increments of 0.1 V versus open-circuit voltage (OCV) up to 0.9 V versus OCV, holding each potential for 10 min for a total of 90 min. The OCV of the cell tested in the neutron imaging experiments was approximately −0.4 V versus Ag/AgCl. Similar OCV values were obtained in tests of the prototype cell at the University of Alabama in Huntsville and in prior neutron imaging tests [2]. The staircase CA scan was performed to observe the nature of the current response from the cell when subjected to these potentials as well as to discharge the cell to a state of lower average glucose concentration. The performance of the cell after the staircase CA was assessed through a CV scan with a scan rate of 100 mVs−1 and upper and lower voltage limits of 0.5 V and −0.75 V, respectively. Then, the cell was further discharged by running a constant-potential chronoamperometry scan for a total of 60 min while the second set of in operando images were taken. The voltage for this scan was measured versus OCV and was selected based on the potential at which the oxidation peak occurred to observe the cell operating at a state of maximum catalytic effectiveness at the onset of the scan. The value selected based on the observations within the previous CV results was 0.6 V versus OCV. A final CV scan was performed after this “long discharge” with the same settings as the previous CV scan.

At this point, the cell was removed from the beamline and flushed with unused fuel as previously mentioned. Then, the same series of electrochemical tests were performed as were carried out for the first wetting case. The CV scans occurred at the same key points in the operation of the cell (e.g., before staircase CA and after staircase CA). The CV scan performed prior to the staircase CA applied the following: a scan rate of 100 mVs−1 and upper and lower voltage limits of 1.0 V and −1.0 V, respectively. These settings remained the same for the remaining two CV scans. The staircase CA was again run at intervals of 0.1 V versus OCV up to 0.9 V versus OCV. It is noted that the OCV for the second wetting was higher than that of the first wetting. The second long discharge was then performed with the potential set relative to the reference electrode (∼90 mV versus Ag/AgCl) as opposed to OCV. This was deemed easier than approximating the potential versus OCV that would be necessary to drive the cell at the potential corresponding to the peak oxidation current during the previous CV scan. Ultimately, the potential was set at 0.12 V versus Ag/AgCl. This CA scan ran for 90 min, the time required to observe a negligible current output from the cell.

The final neutron imaging scan performed was radiography of three solutions in cylindrical, aluminum calibration holders of known diameter. The solutions corresponded to the fuel solution with glucose concentrations of 0.1 M, 0.05 M, and 0.0 M. The diameters of the holders corresponding to the solutions in the order previously listed are 9.13 mm, 9.09 mm, and 9.01 mm. All holders were present in each of the twenty total images captured. This operation was motivated by the previous demonstration of water management in porous media via a calibration sample by Kang et al. These images were taken in an attempt to extend this work to tracking average concentration of a species within the solution. The selection of concentrations to image was based on the maximum and minimum glucose concentrations that could be achieved in the cell during operation. Moreover, the additional 0.05 M glucose solution was imaged to provide more resolution to the calibration.

## Results and Discussion

###### General Image Processing.

As previously mentioned, all sample images were normalized based on the OB and DF images taken on the corresponding day of imaging. This process allows for the removal of any detector defects and variations in beam intensity by referencing the maximum and minimum grayscale values attributed to each pixel. It also allows for comparisons to be made between sample images taken during separate scans. This process is mathematically defined by the below equation: Display Formula

(1)$(In)I,j=(Iraw−IDF¯IOB¯−IDF¯)I,j$

where $(In)i,j$ is the normalized intensity of the ith and jth pixel within the image under consideration, $Iraw$ is the pixel intensity prior to normalization, $IDF¯$ is the pixel intensity averaged over all ten DF images, and $IOB¯$ is the pixel intensity averaged over the ten OB images.

###### Computed Tomography.

All tomographic results were reconstructed using Octopus 8.6, a common tomogram reconstruction software developed at Ghent University. During this process, an automated tilt correction was performed based on slight misalignments in the prototype as mounted on the stage. Also, the default ring filter in Octopus was applied to remove potential artifacts resulting from imbalances in the detector. Upon completion, the results were deemed acceptable based on the absence of obvious artifacts. The filter was therefore not adjusted. By cropping the reconstructed stack to the anodic region of interest (Fig. 2; region 2) and applying a binary threshold using the image processing software, FIJI [23,24], the polymer-based ink could be distinguished from the CF. This distinct difference in attenuation behavior is illustrated in Figs. 1(c) and 1(d) where the biocatalyst and CF regions are indicated. Ultimately, this provided the data necessary to map the spatial distribution of the enzymes within the WE. A three-dimensional representation of the biocatalyst distribution (isosurface) is shown in Fig. 2(e) with the entire region analyzed corresponding to the volume of the WE well.

During the binarization process, the grayscale equivalent of several slices of the WE were compared to the thresholded equivalent to visually assess the appropriate limits of the grayscale range for thresholding. At any setting of these values, one slice may provide an overestimate of the actual active region while another slice may provide an underestimate. Thus, the lack of a quantitative means for validating the threshold limits necessitates this method, producing some inherent error in the final result.

The same reconstruction and binarization process was applied for the CT scan results for both applications of solution. The solution thickness when viewing the cell with a line of site parallel to the beam path (side view) is extracted by counting the number of voxels attributed to fuel after the thresholding process at each pixel in the ROI. These values represent the volume occupied by solution, enzymes, and the CF present within those regions. Since the biocatalyst distribution was known, its presence was first accounted for by subtracting the length calculated from the dry CT results from that calculated for each wet CT scan. The remaining lengths represent regions with solution and CF only. Thus, the solution thickness was obtained by multiplying by the porosity. This process generated the results shown in Fig. 3, where the two wetting distributions are compared via the applied scale. The analysis of the WE resulted in a total solution volume of 447 μL and 470 μL for the first and second wettings, respectively. With a total void space of 582 μL analyzed, this represents a 4% increase in the solution volume with respect to the total volume.

###### Calibration Image Analysis.

The first step in extracting the data from the calibration images was to generate a single image as a result of the average intensity of the 20 calibration images taken. Furthermore, a mean filter with a pixel radius of 4 was applied to reduce noise within the data. This is in accordance with the image processing performed in the work by Kang et al. [20]. Figure 4(a) displays this average, filtered image with the calibration cells from left to right corresponding to 0.05 M glucose, 0.1 M glucose, and 0.0 M glucose fuel solutions.

After aligning the calibration tubes graphically, a custom matlab script was developed to extract the intensity data from the ROIs shown in Fig. 4(a) with respect to the attenuation length through the solution as illustrated by the graphic on the right of Fig. 4(b). This was done by referencing the relationship between a lateral distance from the center line of the tube and the chord length as depicted in Fig. 4(b). The mathematical relationship between these parameters is presented in the following equation: Display Formula

(2)$C(i)=2r2−a(i)2$

where $C(i)$ is the chord length corresponding to the ith column away from the centerline, r is the radius of the tube, and $ai$ is the lateral distance away from the centerline corresponding to the ith column away from the centerline.

As shown by Kang et al., the intensity change due to the transmission of the neutrons through the solution can be isolated by subtracting the natural logarithm of the intensity of the cell imaged in the absence of the solution, Idry, from the natural logarithm of the intensity of the image captured with the solution present, Iwet [20]. This is accomplished with a single image for each of the calibration tubes by referencing the empty portion of each tube at the top of the image as indicated by the upper left ROI in Fig. 4(a) and an ROI corresponding to the solution within the tube, as indicated by the lower rectangle. The data for a dry ROI outside of any of the holders was also referenced as indicated by the rightmost ROI shown in Fig. 4(a) in an effort to remove the effect of any tube defects. After extracting the transmission behavior for each calibration solution, the average was taken at each chord length to generate the data that serve as the overall calibration data set for the aqueous fuel mixture. These data, which do not distinguish between glucose concentration, are presented in Fig. 5, along with the 95% confidence interval and the fit line generated through a least squares fitting technique. The model for the fit used, an adaptation of the Beer–Lambert Attenuation Law, was based on the attenuation model applied by Kang et al. [20] and is presented as the below equation: Display Formula

(3)$ln(Idry¯Iwet¯)(i)=Στ(i)+βτ(i)2$

where $τ(i)$ represents the attenuation length through the solution, $Σ$ is a macroscopic linear attenuation coefficient, and $β$ is a correction coefficient that allows for compensation for beam hardening (i.e., selective attenuation of longer wavelengths). Here, $Σ$ is used to represent the attenuation coefficient to maintain consistency with the nomenclature applied by Kang et al. $β$ is inherently negative; thus, it works by reducing the intensity ratio for increasing attenuation thicknesses [20]. It is again noted that beam hardening has a more significant effect for longer attenuation lengths [25]. The parameter values determined for the fit equation in Fig. 5 are $Σ=0.6mm−1$ and $β=−0.033mm−2$. Though the calibration curve strays from the average value slightly, it remains within the 95% confidence interval over the domain of chord lengths sampled.

###### In Operando Imaging.

Custom matlab code was also developed to analyze the WE ROI within each in operando radiograph to compare the transmission behavior with the calibration data, ultimately allowing for a secondary calculation of the solution distribution in this region. By determining the characteristic transmission parameter, $ln(Idry¯/Iwet¯)$, for each pixel in the WE ROI, the corresponding solution thickness calculated through the fit equation was determined across all images taken during the four in operando NR scans by applying the fit equation. Three select solution distributions are presented in Fig. 6(a) for the long discharge performed for the first wetting case. Qualitatively, these images match well with the results from the CT scan seen on the left of Fig. 3. To evaluate these results relative to the distributions determined from the CT scans, the absolute error was calculated with the CT results serving as the “true” value. Figure 6(b) presents maps of the percent error corresponding to the WE ROI for the same three images shown in Fig. 6(a). These images were selected since they represent the beginning, middle, and end of the long discharge, the period of time in which any transmission variations observed were expected to be the largest. All three images exhibit ∼10% agreement throughout most of the ROI. It is noted that a few small regions exhibit error of 100% or greater, a result that is attributed to the negligible presence of fuel solution in the affected regions. As attenuation lengths approach zero, the calibration used for mapping reaches its resolution limit, and an erroneous result is expected. The observed agreement between the two thickness mapping approaches in the first image speaks to the ability of the fit equation to reproduce the fuel distributions determined from the CT scans since the cell has not been operated. One source of this error may lie in the threshold levels selected for the CT scans, as this would change the fuel distribution calculated as the reference value in this case. Furthermore, the calibration data contain some uncertainty that is propagated to the fit equation, introducing another possible source of error. The images did not show any significant variation in average error for the duration of the long CA scan, as the maps remain essentially identical. The average error for each radiograph was also calculated, and the trend is plotted in Fig. 6(d). Since the error is negative, an average decrease in the value corresponds to an average decrease in solution volume in the WE ROI. Thus, with a decrease of less than 1% over the entire duration of the scan, the cell was deemed effective in containing the fuel during cell operation. This observation is largely attributed to fuel loss since the trend occurs both before and after the driving potential was applied for the CA scan, indicating that the source of the trend is not related to the electrochemical activity of the cell. The results were similar to the other three sets of CA and in operando NR scans. The average percentage error values are presented in Table 1 along with the range of the data, the difference in the maximum and minimum value in each data set. The average values demonstrate bimodal behavior, separated by the wetting case. This further supports the theory that the results are biased due to error induced by the thresholding process since the CT scan is the only component of the calculation that changes from the first to the second wetting.

Having observed relatively constant fuel content and distribution throughout the scans, the possibility of determining the average concentration through each pixel in the WE ROI was explored. The constant fuel presence is necessary since the results of the CT scans must be used to represent the fuel distribution at all points in time for the corresponding wetting case. For this method, the calibration curves for each concentration were plotted along with the 95% confidence interval as seen in Fig. 7.

These results provide a clear picture of the limitations of applying these data to the final sample for detailed glucose concentration mapping by showing the overlap in the uncertainty in the calibration data. The data presented only represent a portion of the attenuation lengths achievable with the prototype cell. The portion selected was that of the higher chord lengths where the largest spread of mean values is observed. For this range, the lower limit of the uncertainty corresponding to the 0.1 M fuel solution (upper band) overlaps the upper limit of the uncertainty corresponding to the 0.0 M (lower band). This overlap is significant, representing 55% of the total uncertainty band on the 0.1 M glucose solution data on average across the domain of chord lengths shown. Having observed the upper mean calibration values varying into the range of values for the lower calibration values, a more definite understanding of the source of the limitations in the application of this technique for studying the glucose concentration distribution is obtained. This result is most likely explained by the nature of the interactions of the neutrons with the sample. The neutrons interact with the atomic nuclei, and their attenuation is not sensitive to chemical bonding. This characteristic functionality leads to macroscopic attenuation behavior that is largely unchanged when a species reacts while maintaining a similar atomic content on average from the reactants to the products. If an EEC is operated as a biobattery, the molecular concentrations within the cell will change, while the number of each type of atom will remain unchanged since there is approximately no mass flux out of the cell (i.e., no leakage). Therefore, it is difficult to support further pursuit of this technique as a method for tracking a molecular species concentration within a solution. However, the method does prove useful in tracking the distribution of the glucose fuel solution. When combined with the mapping of active regions via ex situ tomography, this fuel distribution mapping can provide insight into how localized wetting of the active regions can impact cell performance.

###### Electrochemical Performance and Fuel Distribution.

The electrochemical testing performed at ORNL provides data that are useful in forming connections between the functionality of the cell and its relationship to fuel distributions observed through neutron imaging. All CV results for the testing performed at ORNL consisted of ten cycles. The data for all cycles were averaged to produce the results that will be presented in the proceeding figures. The main function of the CV scans was to assess the performance of the cell before and after the two discharge processes performed for each wetting. The averaged CV results just prior to the stepped CA, after the stepped CA, and after the long discharge for the first wetting are presented in Fig. 8.

The data for the first wetting provide similar enough results to that of the second wetting to facilitate a qualitative discussion applicable to both cases. The first observation made is that the peak oxidation current for the earliest scan is the highest out of the three, indicating that the enzymes were most active here in combination with a higher average glucose concentration within the fuel in the cell. As the cell is discharged through both the stepped CA and the long discharge, the peak oxidation current continues to decrease. Converse to the results of the first CV scan, these results indicate that some of the glucose has been consumed and the enzymes have become less active.

A second observation from this data is the leftward shift in the potential at which the peak oxidation current occurs. This shift occurs as the cell is discharged further from its original state. This behavior is explained by a slowing of the reaction in the presence of a decreasing concentration of glucose and increasing concentration of chemical inhibitors related to the products of the oxidation reaction. The oxidation peak occurs when the mass diffusion of the substrate to the active material surface is balanced by the maximum reaction rate at the active surface (e.g., maximum cation flux). Thus, if this occurs at a lower potential, it means the enzymes are not able to continue to produce a significant enough ionic concentration potential to overcome the diffusive resistance presented by the flux of the glucose fuel to the active surface. This means the rate of oxidation catalysis will decrease, yielding a subsequent decrease in the electron flux measured as a decrease in the current output from the cell.

In addition to comparing the performance of the cell before and after each discharge for a single wetting, it is also useful to juxtapose the results obtained from the first and second wetting. Figure 9 facilitates this comparison, including the CV results obtained after the stepped CA and the long discharge for both wettings. When comparing the scans taken at the same point in the discharge process for each wetting, an increase in peak oxidation current by a factor of approximately two was observed. This demonstrates the ability of the cell to perform the same or better after being exposed to the neutron beam for an extended period after simply introducing new fuel. It also demonstrates the strong sensitivity the output of the cell has in relation to the fuel loading and average glucose concentration. Moreover, this observation demonstrates that the enzymes, showing signs of being chemically inhibited during the first wetting discharge tests, were not denatured and were ultimately able to return to their initial equilibrium state.

For Fig. 9, it is noted that the peak oxidation potential for the CV scans performed before the long discharge decreased from the first wetting to the second, indicating a slower reaction due to decreased glucose concentration and the presence of inhibitors in the oxidation products. However, the concentration potentials may have been greater in a proportional manner, increasing the overall measured transport of electrons (current). In contrast, the peak oxidation potential for the second wetting after the long discharge occurs at a higher value than that of the first wetting. This may be a result of the less substantial effect of chemical inhibitors that the reaction inherently introduces. With a greater presence of unreacted glucose fuel near the active region during the second wetting, there is a larger sink for the products of the reaction, resulting in a lower average concentration of the gluconic acid in the fuel. This better saturation of the active region may provide a more stable local operating environment for the enzymes in terms of pH.

The observation of an approximately 200% increase in peak oxidation current from the first wetting to the second does not align with the observation of a 23 μL increase in solution out of the 600 μL volume analyzed, a 3.8% increase. To understand how this minimal increase in total solution volume could contribute to the increase in peak oxidation current by a factor of two, the distribution of the solution is compared between the two wetting cases. A qualitative assessment of the change in distribution can be made through the thickness maps shown in Fig. 3 that demonstrate the dry regions at the upper right and bottom right corners of the ROI becoming significantly more saturated after the second wetting. From this observation, it appears that the increase in wetting occurred predominantly in the regions around the biocatalyst ink. To confirm this theory, the increase in solution presence in decreasing WE volumes of interest (VOIs) is analyzed where the profile of the VOI is defined by the side view observed in a radiograph of the cell. The third dimension corresponds to the attenuation length, or thickness of the WE holder (6.4 mm). The change in the control volume is better understood by referencing Fig. 10(b), which shows the initial ROI laid over the radiograph of the prototype cell. The $x$-vector indicates the initial position of the left edge, which is subsequently shifted to the right to analyze VOIs with decreasing volume. In addition, a flowchart describing the analysis process is provided in (c) for clarity.

Due to the presence of the biocatalyst in most VOIs, the enzymes and the supporting ink occupy a significant volume that must be removed for consistent comparison of the wetting throughout the domain of ROIs. For both wetting cases, the distribution of the biocatalyst is assumed to be known based on the data produced from the first, dry cell CT scan. The initial VOI analyzed includes 100% of the total volume of enzymes present. As the left boundary shifts toward, and ultimately through the bulk of the active region, the percentage of volume occupied by the biocatalyst in the slice at a given value of $x*$ increases, where $x*=x/L$. This trend is observed in the enzyme presence plotted for each slice through the entire VOI in Fig. 10(a). This more specifically indicates that the active region is primarily located beyond 80% of the electrode thickness as defined by the $x$-vector and $L$.

Working toward a comparison of the distribution of solution from the first to second wetting, the presence of the carbon felt was also considered by employing the porosity. This facilitates the discussion of the void space within the electrode. This space, free of both enzymes and CF, is used to normalize the comparison of VOIs for the first and second wettings and is calculated as $Vvoid$ through the below equation: Display Formula

(4)$Vvoid=ε(VVOI−Venz)$

where $VVOI$ is the total volume of interest, $Venz$, is the volume of the biocatalyst present within a given VOI, and $ε$ is the porosity of the CF. Since the distribution of the biocatalyst region was assumed constant throughout the experimentation process, the resulting value for $Vvoid$ is applicable for both wetting cases.

To compare the fuel distribution between the two wetting cases, the difference between those cases was calculated in the form of a normalized solution volume change, $ΔV*$Display Formula

(5)$ΔV*=(Vsol)2−(Vsol)1Vvoid$

where $(Vsol)i$ is the solution volume corresponding to the ith wetting case present within the VOI. This value is calculated by Eq. (4) with $VVOI$ replaced with the solution volume calculated by voxel counting. The trend of $ΔV*$ with the position of the right boundary of the ROI is shown as the solid line in Fig. 10(a). The peak in $ΔV*$ is centered about the approximate center of mass of the biocatalyst region, indicating that the quantitative analysis supports the qualitative assessment of the change in wetting distribution as being localized around the biocatalyst region. This observation is again based on calculations of fuel distribution resulting from CT scans that require a large amount of experimental time. If an EEC was calibrated based on the presence or absence of fuel similar to the work of Kang et al., radiographic image data could be captured, from which the solution distribution could be calculated [20]. This would allow for a more thorough study of EECs with a wider set of experimental parameters since a single radiograph took 1 min to capture in comparison with the ∼10 h duration of a CT scan. In a fuel cell arrangement, this would be especially advantageous since its operation requires a continuous flux of reactants and products into and out of the cell, introducing a higher potential for transient cell wetting variations during operation. Such distinct wetting variations are not observed for the biobattery configuration employed in this study since the cell is well sealed upon assembly. As this work is moved forward, these methods may yield observations related to the distribution of fuel around the active region that could better inform future cell configurations and numerical modeling of EFCs.

## Conclusion

This study has demonstrated that neutron computed tomography can produce useful results for determining the distribution of aqueous fuel solutions and the biocatalyst throughout an EEC. Ultimately, this drives a better understanding of the genesis of variations in electrochemical performance as the increase in localized fuel saturation around the biocatalyst region provided a qualitative correlation to the ∼200% increase in the peak oxidation current. Moreover, biocatalyst mapping could become a tool to inform the process of biocatalyst deposition and penetration into these porous, three-dimensional electrode materials. Furthermore, the ability to track the approximate fuel distribution during in operando NR was observed in agreement with the CT scans ranging from 8% to 14% on average. However, the extrapolation of the common calibration techniques used for tracking the presence of a solution to tracking the concentration of glucose within that solution via neutron radiography has been shown to be an intractable problem with the current experimental architecture. The ability to draw insight from the ties between cell characteristics determined through NR and NT and the electrochemical performance of the cell shows promise for further, quantitative investigation into the effects of environmental conditions on performance of EECs, specifically EFCs, and the development of more effective EEC designs and improved implementation.

## Funding data

• The University of Alabama in Huntsville New Faculty Grant Program (Grant No. UAH-NFR-2015-116).

• A portion of this research at ORNL's High Flux Iso-tope Reactor was sponsored by the Scientific User Facilities Division, Office of Basic Energy Sciences, and U.S. Department of Energy.

## References

Zhu, Z. , Sun, F. , Zhang, X. , and Zhang, Y.-H. , 2012, “ Deep Oxidation of Glucose in Enzymatic Fuel Cells Through a Synthetic Enzymatic Pathway Containing a Cascade of Two Thermostable Dehydrogenases,” Biosens. Bioelectron., 36(1), pp. 110–115. [PubMed]
Looney, E. E. , Nelson, G. J. , van Zandt, Z. K. , Ulyanova, Y. , Singhal, S. , Santodonato, L. J. , and Bilheux, H. Z. , 2016, “ Ex Situ and In Situ Neutron Imaging of Enzymatic Electrochemical Cells,” Electrochim. Acta, 213, pp. 244–251.
Potter, M. C. , 1911, “ Electrical Effects Accompanying the Decomposition of Organic Compounds,” Proc. R. Soc. London B, 84(571), pp. 260–276.
Yahiro, A. T. , Lee, S. M. , and Kimble, D. O. , 1964, “ Bioelectrochemistry: I. Enzyme Utilizing Bio-Fuel Cell Studies,” Biochim. Biophys. Acta, 88(2), pp. 375–383. [PubMed]
Rasmussen, M. , Abdellaoui, S. , and Minteer, S. D. , 2016, “ Enzymatic Biofuel Cells: 30 Years of Critical Advancements,” Biosens. Bioelectron., 76, pp. 91–102. [PubMed]
Ghassemi, Z. , and Slaughter, G. , 2017, “ Biological Fuel Cells and Membranes,” Membranes, 7(1), p. 3.
Calabrese Barton, S. , Gallaway, J. , and Atanassov, P. , 2004, “ Enzymatic Biofuel Cells for Implantable and Microscale Devices,” Chem. Rev., 104(10), pp. 4867–4886. [PubMed]
Minteer, S. D. , Liaw, B. Y. , and Cooney, M. J. , 2007, “ Enzyme-Based Biofuel Cells,” Curr. Opin. Biotechnol., 18(3), pp. 228–234. [PubMed]
Zhao, C. , Gai, P. , Song, R. , Chen, Y. , Zhang, J. , and Zhu, J. J. , 2017, “ Nanostructured Material-Based Biofuel Cells: Recent Advances and Future Prospects,” Chem. Soc. Rev., 46(5), pp. 1545–1564. [PubMed]
Palmore, G. T. R. , Bertschy, H. , Bergens, S. H. , and Whitesides, G. M. , 1998, “ A Methanol/Dioxygen Biofuel Cell That Uses NAD+-Dependent Dehydrogenases as Catalysts: Application of an Electro-Enzymatic Method to Regenerate Nicotinamide Adenine Dinucleotide at Low Overpotentials,” J. Electroanal. Chem., 443(1), pp. 155–161.
Stolarczyk, K. , Kizling, M. , Majdecka, D. , Żelechowska, K. , Biernat, J. F. , Rogalski, J. , and Bilewicz, R. , 2013, “ Biobatteries and Biofuel Cells With Biphenylated Carbon Nanotubes,” J. Power Sources, 249, pp. 263–269.
Wu, X. , Zhao, F. , Varcoe, J. R. , Thumser, A. E. , Avignone-Rossa, C. , and Slade, R. C. T. , 2009, “ A One-Compartment Fructose/Air Biological Fuel Cell Based on Direct Electron Transfer,” Biosens. Bioelectron., 25(2), pp. 326–331. [PubMed]
Aaron, D. S. , Borole, A. P. , Hussey, D. S. , Jacobson, D. L. , Yiacoumi, S. , and Tsouris, C. , 2011, “ Quantifying the Water Content in the Cathode of Enzyme Fuel Cells Via Neutron Imaging,” J. Power Sources, 196(4), pp. 1769–1775.
Dill, K. A. , 1990, “ Dominant Forces in Protein Folding,” Biochemistry, 29(31), pp. 7133–7155. [PubMed]
Gouda, M. D. , Thakur, M. S. , and Karanth, N. G. , 2002, “ Reversible Denaturation Behavior of Immobilized Glucose Oxidase,” Appl. Biochem. Biotechnol., 102(1–6), pp. 471–480. [PubMed]
Cho, K. T. , Turhan, A. , Lee, J. H. , Brenizer, J. S. , Heller, A. K. , Shi, L. , and Mench, M. M. , 2009, “ Probing Water Transport in Polymer Electrolyte Fuel Cells With Neutron Radiography,” Nucl. Instrum. Methods Phys. Res. Sect. A, 605(1–2), pp. 119–122.
Manke, I. , Markötter, H. , Tötzke, C. , Kardjilov, N. , Grothausmann, R. , Dawson, M. , and Hartnig, C. , 2011, “ Investigation of Energy-Relevant Materials With Synchrotron X-Rays and Neutrons,” Adv. Eng. Mater., 13(8), pp. 712–729.
Strobl, M. , Manke, I. , Kardjilov, N. , Hilger, A. , Dawson, M. , and Banhart, J. , 2009, “ Advances in Neutron Radiography and Tomography,” J. Phys. D, 42(24), p. 243001.
Weber, A. Z. , and Hickner, M. A. , 2008, “ Modeling and High-Resolution-Imaging Studies of Water-Content Profiles in a Polymer-Electrolyte-Fuel-Cell Membrane-Electrode Assembly,” Electrochim. Acta, 53(26), pp. 7668–7674.
Kang, M. , Bilheux, H. Z. , Voisin, S. , Cheng, Z. L. , Perfect, E. , Horita, J. , and Warren, J. M. , 2013, “ Water Calibration Measurements for Neutron Radiography: Application to Water Content Quantification in Porous Media,” Nucl. Instrum. Methods Phys. Res. Sect. A, 708, pp. 24–31.
González-García, J. , Bonete, P. , Expósito, E. , Montiel, V. , Aldaz, A. , and Torregrosa-Maciá, R. , 1999, “ Characterization of a Carbon Felt Electrode: Structural and Physical Properties,” J. Mater. Chem., 9(2), pp. 419–426.
Ulyanova, Y. , Babanova, S. , Pinchon, E. , Matanovic, I. , Singhal, S. , and Atanassov, P. , 2014, “ Effect of Enzymatic Orientation Through the Use of Syringaldazine Molecules on Multiple Multi-Copper Oxidase Enzymes,” Phys. Chem. Chem. Phys., 16(26), pp. 13367–13375. [PubMed]
Schindelin, J. , Arganda-Carreras, I. , Frise, E. , Kaynig, V. , Longair, M. , and Pietzsch, T. , 2012, “ FIJI: An Open-Source Platform for Biological-Image Analysis,” Nat. Methods, 9, pp. 676–682. [PubMed]
Schneider, C. A. , Rasband, W. S. , and Eliceiri, K. W. , 2012, “ NIH Image to ImageJ: 25 Years of Image Analysis,” Nat. Methods, 9, pp. 671–675. [PubMed]
Zawisky, M. , Bastürk, M. , Rehacek, J. , and Hradil, Z. , 2004, “ Neutron Tomographic Investigations of Boron-Alloyed Steels,” J. Nucl. Mater., 327(2–3), pp. 188–193.
View article in PDF format.

## References

Zhu, Z. , Sun, F. , Zhang, X. , and Zhang, Y.-H. , 2012, “ Deep Oxidation of Glucose in Enzymatic Fuel Cells Through a Synthetic Enzymatic Pathway Containing a Cascade of Two Thermostable Dehydrogenases,” Biosens. Bioelectron., 36(1), pp. 110–115. [PubMed]
Looney, E. E. , Nelson, G. J. , van Zandt, Z. K. , Ulyanova, Y. , Singhal, S. , Santodonato, L. J. , and Bilheux, H. Z. , 2016, “ Ex Situ and In Situ Neutron Imaging of Enzymatic Electrochemical Cells,” Electrochim. Acta, 213, pp. 244–251.
Potter, M. C. , 1911, “ Electrical Effects Accompanying the Decomposition of Organic Compounds,” Proc. R. Soc. London B, 84(571), pp. 260–276.
Yahiro, A. T. , Lee, S. M. , and Kimble, D. O. , 1964, “ Bioelectrochemistry: I. Enzyme Utilizing Bio-Fuel Cell Studies,” Biochim. Biophys. Acta, 88(2), pp. 375–383. [PubMed]
Rasmussen, M. , Abdellaoui, S. , and Minteer, S. D. , 2016, “ Enzymatic Biofuel Cells: 30 Years of Critical Advancements,” Biosens. Bioelectron., 76, pp. 91–102. [PubMed]
Ghassemi, Z. , and Slaughter, G. , 2017, “ Biological Fuel Cells and Membranes,” Membranes, 7(1), p. 3.
Calabrese Barton, S. , Gallaway, J. , and Atanassov, P. , 2004, “ Enzymatic Biofuel Cells for Implantable and Microscale Devices,” Chem. Rev., 104(10), pp. 4867–4886. [PubMed]
Minteer, S. D. , Liaw, B. Y. , and Cooney, M. J. , 2007, “ Enzyme-Based Biofuel Cells,” Curr. Opin. Biotechnol., 18(3), pp. 228–234. [PubMed]
Zhao, C. , Gai, P. , Song, R. , Chen, Y. , Zhang, J. , and Zhu, J. J. , 2017, “ Nanostructured Material-Based Biofuel Cells: Recent Advances and Future Prospects,” Chem. Soc. Rev., 46(5), pp. 1545–1564. [PubMed]
Palmore, G. T. R. , Bertschy, H. , Bergens, S. H. , and Whitesides, G. M. , 1998, “ A Methanol/Dioxygen Biofuel Cell That Uses NAD+-Dependent Dehydrogenases as Catalysts: Application of an Electro-Enzymatic Method to Regenerate Nicotinamide Adenine Dinucleotide at Low Overpotentials,” J. Electroanal. Chem., 443(1), pp. 155–161.
Stolarczyk, K. , Kizling, M. , Majdecka, D. , Żelechowska, K. , Biernat, J. F. , Rogalski, J. , and Bilewicz, R. , 2013, “ Biobatteries and Biofuel Cells With Biphenylated Carbon Nanotubes,” J. Power Sources, 249, pp. 263–269.
Wu, X. , Zhao, F. , Varcoe, J. R. , Thumser, A. E. , Avignone-Rossa, C. , and Slade, R. C. T. , 2009, “ A One-Compartment Fructose/Air Biological Fuel Cell Based on Direct Electron Transfer,” Biosens. Bioelectron., 25(2), pp. 326–331. [PubMed]
Aaron, D. S. , Borole, A. P. , Hussey, D. S. , Jacobson, D. L. , Yiacoumi, S. , and Tsouris, C. , 2011, “ Quantifying the Water Content in the Cathode of Enzyme Fuel Cells Via Neutron Imaging,” J. Power Sources, 196(4), pp. 1769–1775.
Dill, K. A. , 1990, “ Dominant Forces in Protein Folding,” Biochemistry, 29(31), pp. 7133–7155. [PubMed]
Gouda, M. D. , Thakur, M. S. , and Karanth, N. G. , 2002, “ Reversible Denaturation Behavior of Immobilized Glucose Oxidase,” Appl. Biochem. Biotechnol., 102(1–6), pp. 471–480. [PubMed]
Cho, K. T. , Turhan, A. , Lee, J. H. , Brenizer, J. S. , Heller, A. K. , Shi, L. , and Mench, M. M. , 2009, “ Probing Water Transport in Polymer Electrolyte Fuel Cells With Neutron Radiography,” Nucl. Instrum. Methods Phys. Res. Sect. A, 605(1–2), pp. 119–122.
Manke, I. , Markötter, H. , Tötzke, C. , Kardjilov, N. , Grothausmann, R. , Dawson, M. , and Hartnig, C. , 2011, “ Investigation of Energy-Relevant Materials With Synchrotron X-Rays and Neutrons,” Adv. Eng. Mater., 13(8), pp. 712–729.
Strobl, M. , Manke, I. , Kardjilov, N. , Hilger, A. , Dawson, M. , and Banhart, J. , 2009, “ Advances in Neutron Radiography and Tomography,” J. Phys. D, 42(24), p. 243001.
Weber, A. Z. , and Hickner, M. A. , 2008, “ Modeling and High-Resolution-Imaging Studies of Water-Content Profiles in a Polymer-Electrolyte-Fuel-Cell Membrane-Electrode Assembly,” Electrochim. Acta, 53(26), pp. 7668–7674.
Kang, M. , Bilheux, H. Z. , Voisin, S. , Cheng, Z. L. , Perfect, E. , Horita, J. , and Warren, J. M. , 2013, “ Water Calibration Measurements for Neutron Radiography: Application to Water Content Quantification in Porous Media,” Nucl. Instrum. Methods Phys. Res. Sect. A, 708, pp. 24–31.
González-García, J. , Bonete, P. , Expósito, E. , Montiel, V. , Aldaz, A. , and Torregrosa-Maciá, R. , 1999, “ Characterization of a Carbon Felt Electrode: Structural and Physical Properties,” J. Mater. Chem., 9(2), pp. 419–426.
Ulyanova, Y. , Babanova, S. , Pinchon, E. , Matanovic, I. , Singhal, S. , and Atanassov, P. , 2014, “ Effect of Enzymatic Orientation Through the Use of Syringaldazine Molecules on Multiple Multi-Copper Oxidase Enzymes,” Phys. Chem. Chem. Phys., 16(26), pp. 13367–13375. [PubMed]
Schindelin, J. , Arganda-Carreras, I. , Frise, E. , Kaynig, V. , Longair, M. , and Pietzsch, T. , 2012, “ FIJI: An Open-Source Platform for Biological-Image Analysis,” Nat. Methods, 9, pp. 676–682. [PubMed]
Schneider, C. A. , Rasband, W. S. , and Eliceiri, K. W. , 2012, “ NIH Image to ImageJ: 25 Years of Image Analysis,” Nat. Methods, 9, pp. 671–675. [PubMed]
Zawisky, M. , Bastürk, M. , Rehacek, J. , and Hradil, Z. , 2004, “ Neutron Tomographic Investigations of Boron-Alloyed Steels,” J. Nucl. Mater., 327(2–3), pp. 188–193.

## Figures

Fig. 1

Prototype, three-electrode cell schematic (a) and images (b), illustrating the assembly of the final cell for electrochemical testing and neutron imaging. (c) A representative radiograph of the cell prior to the introduction of any fuel, with the expanded ROI shown in (d) demonstrating the observable difference in transmission behavior between the CF and the biocatalyst ink.

Fig. 2

Sample slice and isosurface from reconstructed tomograms of the dry cell. (a) Shows the full field of view of the tomogram with two distinct regions indicated: region 1 and region 2. (b) Presents the cropping of (a) to region 1 with the CF and biocatalyst ink indicated with distinct intensity differences. (c) Represents (a) cropped to region 2 with the image data presented in grayscale. (d) Represents the same region as (c); however, the data are presented in binary for after the threshold was applied. (e) Shows the isosurface of the biocatalyst region and its orientation with respect to the WE holder.

Fig. 3

Solution thickness maps for the first wetting on the left and the second wetting on the right. The solution attenuation thickness is correlated with the map via the values on the right in millimeters.

Fig. 4

(a) Calibration image resulting from the average of the 20 individual images taken. The calibration tubes from left to right correspond to 0.05 M glucose, 0.1 M glucose, and 0.0 M glucose fuel solutions with diameters of 9.09 mm, 9.13 mm, and 9.01 mm, respectively. (b) An illustration of the chord length, c as it is referenced to a lateral distance from the centerline of the calibration cell, a, and how this is oriented with respect to the neutron flux.

Fig. 5

Average calibration transmission data extracted from the three sample solutions of concentration 0.1 M, 0.05 M, and 0.0 M glucose in 0.243 M sodium phosphate. The data are plotted versus chord length along with the associated fit curve determined through a standard, least squares fitting technique. The 95% confidence interval for the mean values is plotted as the two black lines displaced evenly from the mean values.

Fig. 6

(a) Fuel solution thickness distribution in mm calculated for the three radiographs taken at 1, 10, and 60 min after the onset of the long CA scan for the first wetting case. (b) The percent difference between the solution thickness calculated via the calibration fit curve and the solution thickness calculated through the reconstructed CT scan data for the same elapsed times. The associated scale for each set of images is shown to the right. (c) Presents the trend of the average percentage deviation of the calibrated length values from the CT scan data. The electrochemical data associated with this image data are presented in (d).

Fig. 7

Average calibration data for all three solution concentrations imaged with the uncertainty bands presented as the shaded regions

Fig. 8

Averaged CV results for the first wetting case shown for the state of the cell just prior to the stepped CA scan, after the stepped CA scan, and after the long discharge. The approximate locations of peak oxidation current are indicated by the diamond markers.

Fig. 9

Average CV results for the scans take prior to and after the long discharge for both wettings. The data for the CV scan performed before and after the long discharge are shown in upper and lower series, respectively. The approximate locations of the oxidation peaks are identified by the diamond marker on each series.

Fig. 10

(a) shows the increase in the normalized solution volume as the volumetric ROI is decreased with a shift of the left edge moving toward the right edge (i.e., x*) and the volume of enzymes normalized on the total volume of a slice present at x*. (b) Shows a radiograph of the fueled sample cell with the WE ROI identified and magnified with the definition of the reference for the left edge of the ROI, x, and the overall length, L. (c) Presents a high-level flowchart describing the process carried out to achieve the data shown in (a) with the nomenclature, CT1 and CT2, corresponding to the CT scan data for the first and second wetting cases, respectively.

## Tables

Table 1 Average percent error in solution thickness estimates for the four in operando NR scans presented with the respective percent error range, the difference between the maximum and minimum of each data set

## Discussions

Some tools below are only available to our subscribers or users with an online account.

### Related Content

Customize your page view by dragging and repositioning the boxes below.

Related Journal Articles
Related Proceedings Articles
Related eBook Content
Topic Collections