Production and Characterization of Bacterial Cellulose Separators for Nickel-Zinc Batteries

Abstract

The need for energy-storing technologies with lower environmental impact than Li-ion batteries but similar power metrics has revived research in Zn-based battery chemistries. The application of bio-based materials as a replacement for current components can additionally contribute to an improved sustainability of Zn battery systems. For that reason, bacterial cellulose (BC) was investigated as separator material in Ni-Zn batteries. Following the biotechnological production of BC, the biopolymer was purified, and differently shaped separators were generated while surveying the alterations of its crystalline structure via X-ray diffraction measurements during the whole manufacturing process. A decrease in crystallinity and a partial change of the BC crystal allomorph type I_α to II was determined upon soaking in electrolyte. Electrolyte uptake was found to be accompanied by dimensional shrinkage and swelling, which was associated with partial decrystallization and hydration of the amorphous content. The separator selectivity for hydroxide and zincate ions was higher for BC-based separators compared to commercial glass-fiber (GF) or polyolefin separators as estimated from the obtained diffusion coefficients. Electrochemical cycling showed good C-rate capability of cells based on BC and GF separators, whereas cell aging was pronounced in both cases due to Zn migration and anode passivation. Lower electrolyte retention was concluded as major reason for faster capacity fading due to zincate supersaturation within the BC separator. However, combining a dense BC separator with low zincate permeability with a porous one as electrolyte reservoir reduced ZnO accumulation within the separator and improved cycling stability, hence showing potentials for separator adjustment.

1. Introduction

On the way towards shifting energy sources from fossil fuels to renewable resources, inadequate energy storage is still a limiting factor. With rechargeable batteries as one of the available options, many cell chemistries for efficient, reliable, and long-term energy storage are in the focus of today’s research. In this field, Zn-based batteries are considered as promising candidates by cause of cheap and abundant raw materials, recyclability, ease of manufacturing, and safe handling due to aqueous and non-combustible electrolytes. Furthermore, Zn is environmentally insensitive, easily accessible from the earth crust, and is characterized by a high theoretical specific capacity (820 mAh ${\mathrm{g}}_{\mathrm{Zn}}^{-1}$) [1,2]. Among other long-investigated cell chemistries such as Ni-Zn, Ag-Zn, and MnO₂-Zn, Zn-air is regarded as ideal Zn-battery candidate in terms of sustainability and specific energy [2,3,4]. Zn-air batteries use oxygen from the atmosphere for the electrochemical reaction and are characterized by practical specific energies of up to 450 Wh kg⁻¹, which is still more than for currently available advanced Li-ion batteries but considerably less than the targeted 1000 Wh kg⁻¹ for Li-air [5,6]. Nevertheless, the latter goes along with distinct drawbacks in terms of environmental and cost aspects regarding the Li metal.

Despite the advantageous starting conditions, Zn-based batteries are facing major challenges for exploiting their theoretical potential. Most of these issues derive from the Zn anode, which is prone to passivation, self-corrosion, shape change, and self-discharge, and thus promotes capacity fading [3]. To overcome these challenges, many components of Zn batteries have been optimized, such as the anode structure [1,7,8] and composition [9,10] or the electrolyte in terms of composition and additives [3,11,12], pH range [13] or its fundamental chemistry [14]. Consistent with the progress of Zn-air batteries, the separating component between anode and cathode has become of great interest as well.

Separators are fundamental components in batteries separating the electrodes and thus preventing electrical short-circuits while allowing for ionic conductivity. Although being this important to the battery’s performance, separators are often less considered when it comes to optimization approaches. In Zn-based batteries in particular, the separator offers great opportunities for stabilizing the Zn electrode, for example by suppressing the expansion of dendrites [15,16]. Several separator modification methods were proposed with the aim to prevent dendrite growth, Zn migration, or zincate formation, thus optimizing the battery performance. Suppression of dendrite growth has been achieved by optimizing the mechanical stability of separators [15,17] or by separators that affect the zinc morphology [18,19]. Furthermore, Yuan et al. [20] described a negatively charged nanoporous membrane to alter the growth direction of the dendrites and, therefore, avoid separator piercing. Another challenge to address is the zincate crossover through separator membranes to the cathode. Herein, previous studies focused on adjusting the ionic selectivity of separators [21,22] or the trapping of zincate ions inside functionalized membrane layers [23]. Additional approaches to tackle shape change and zincate crossover involve the utilization of gel polymer or hydrogel electrolytes, which also replace the separator and provide favorable protective effects toward the Zn anode. By reducing the amount of free water compared to aqueous electrolytes, undesired reactions of Zn, including zincate formation, are minimized [24,25]. Biopolymers applied in this context may include xanthan [26], bacterial cellulose (BC) [27], or BC/polyvinyl alcohol composites [28].

In the present study, the performance of BC as separator in a Ni-Zn battery was investigated. Rechargeable Ni-Zn batteries offer specific energy and specific power comparable to present battery technologies such as Li-ion [2,4]. In addition, they benefit from comparatively low environmental impact and abundant material availability together with safe handling [4,29]. BC was chosen as separator material due to its exceptional properties such as high tensile strength, high degree of polymerization, thermal stability, purity, hydrophilicity, and crystallinity [30,31,32]. The three-dimensional network structure of BC combined with its hydrophilicity suggests a high electrolyte uptake and ionic conductivity at low environmental impact as the biopolymer can be produced from a variety of by-product streams from the food industry [32]. For these structural and environmental aspects, BC has been investigated as separator in a range of battery types such as Li-ion [31,33], Pb-acid [30], or Zn-air [17,27,28] before. In case of Zn batteries, cellulose-based materials such as cellophane have long been applied but lacked stability at lower alkalinities or in Ag-Zn batteries due to oxidative degradation by Ag ions [34,35]. Still, they are applied in several layers in Ni-Zn or Ag-Zn batteries to reduce the risk of a short circuit by degradation [11,36].

In view of the advantageous properties of BC, it was investigated to what extent material properties, such as the comparatively high crystallinity, affect the structural stability and performance in Ni-Zn batteries. Therefore, BC was biotechnologically produced with Komagataeibacter xylinus DSM 2325, purified, and processed into different kinds of separator types. The BC separator was then characterized prior to its battery application in terms of swelling behavior, hydroxide, and zincate permeability, as well as its crystalline structure via X-ray diffraction (XRD) measurements. Afterwards, the produced separators were electrochemically analyzed in comparison with a commercial glass-fiber (GF) reference separator and surveyed by XRD to evaluate alterations of the semi-crystalline cellulose-based material. Properties such as cycle life, coulombic efficiency, and C-rate capability were electrochemically investigated and discussed. Hence, this article gives a biopolymer-focused view on the production and application of BC in Zn batteries, which will contribute to a better understanding of suitable BC preparation processes, the behavior and ion permeability of BC separators in alkaline electrolytes, and potential improvements by the combination of tailored BC separators with complementary properties.

2. Materials and Methods

2.1. Manufacturing of BC Separator

2.1.1. BC Production

K. xylinus DSM 2325 was purchased from the German Collection of Microorganisms and Cell Cultures GmbH (DSMZ, Braunschweig, Germany) as a freeze-dried sample and was revived according to the provided DSMZ-protocol. As such, prepared glycerol stock cultures were used as inoculum for a preculture cultivated in Hestrin-Schramm medium with an initial pH value of 5.0 [37]. After 5 days of cultivation at 30 °C and a shaking frequency of 150 min⁻¹ with a shaking diameter of 50 mm using 250 mL shake flasks without baffles and 10% filling volume, the resulting cellulose agglomerates were dissected with sterile spatula and the leaking bacterial suspension was used to inoculate 1 L main cultures to an initial optical density of 0.05 at 600 nm. After 7 days of cultivation the developed cellulose was harvested, dissected, pelleted at 8500× g, and then washed with distilled water prior to incubation at 90 °C for 70 min in 0.1 N NaOH [38]. Subsequently, the BC was washed several times with distilled water to remove detached cells and NaOH. From there on, the resulting BC was kept hydrated and stored at 5 °C until further processing.

2.1.2. Separator Production

The purified BC was first processed at 4000 min⁻¹ with a dispersing instrument (Ultra Turrax T50, IKA, Staufen, Germany) and centrifuged followed by a second, more intense disintegration step with an increasing stirring rate up to 24,000 min⁻¹ (Ultra Turrax T25, IKA, Staufen, Germany). The subsequently pelleted BC was then diluted to a wet weight concentration of about 30 g L⁻¹ BC and was subjected to a high-pressure homogenizer (HPH, Lab Homogenizer PandaPLUS 2000, GEA Group, Düsseldorf, Germany) for five cycles at an initial pressure of 150 bar during the first cycle. Afterwards, the pressure was continuously increased for each cycle up to 1500 bar during the last one. The resulting BC suspension was finally pelleted and lyophilized. Three different BC separators were then prepared by two different methods: BC-5 and BC-10 by solution casting of a 5 and 10 g L⁻¹ BC suspension in a 1 and 2 mm thick mold containing additional 1 g L⁻¹ CMC (DT R HV 400, Mikro-Technik, Bürgstadt am Main, Germany) and 0.2 M propylene carbonate, which was dried at 55 °C following the procedure of Gwon et al. [31] and BC-10-L by lyophilizing a 10 g L⁻¹ BC suspension in order to obtain a thick and porous membrane.

2.2. Separator Characterization

2.2.1. Dimensional Characteristics and Changes upon Electrolyte Uptake

The thickness of dry and electrolyte-soaked separators was determined via a digital micrometer (MMO, Kloster-Lehnin Germany) in triplicate. Therefore, the separators were soaked in electrolyte solution (6 M KOH, 20 mM Zn(OAc)₂) for 2 h, which was followed by the removal of excess water with a paper towel. The area- and volume-based dimensional changes were then calculated from the measured separator diameter and thickness according to Equations (1) and (2) [21]. Gaussian error propagation was applied to consider for sample deviations during dry and wet thickness and diameter determinations.

$${\mathsf{\Delta }\mathrm{A} [\%]=\left[\right(\mathrm{A}}_{\mathrm{s}}{ - \mathrm{A}}_{\mathrm{d}}{)/\mathrm{A}}_{\mathrm{d}}] · 100$$

(1)

$${\mathsf{\Delta }\mathrm{V} [\%]=\left[\right(\mathrm{V}}_{\mathrm{s}}{ - \mathrm{V}}_{\mathrm{d}}{)/\mathrm{V}}_{\mathrm{d}}] · 100$$

(2)

Here, A_d and V_d represent the area and volume of the initially dry separators, respectively, whereas A_s and V_s indicate the measure for the swollen ones. The electrolyte uptake was investigated accordingly and measured gravimetrically. The relative weight difference ΔW resulting from the electrolyte uptake was calculated using Equation (3) with W_s and W_d representing the electrolyte-swollen and previously weighed dry separators, respectively [21].

$${\mathsf{\Delta }\mathrm{W} [\%]=\left[\right(\mathrm{W}}_{\mathrm{s}}{ - \mathrm{W}}_{\mathrm{d}}{)/\mathrm{W}}_{\mathrm{d}}] · 100$$

(3)

Electrolyte retention W_r is defined by Equation (4), where W_t represents the weight of the electrolyte-soaked separator after different incubation time spans at ambient air [39].

$${\mathrm{W}}_{\mathrm{r}}{ [\%]=(\mathrm{W}}_{\mathrm{t}}{/\mathrm{W}}_{\mathrm{s}}) · 100$$

(4)

2.2.2. Hydroxide and Zincate Diffusion

Diffusion coefficients for hydroxide (D_OH⁻) and tetrahydroxozincate ions (D_Zn(OH)4²⁻, abbreviated as zincate in this article) were determined in a diffusion cell consisting of two glass chambers filled with 50 mL feed or draw solution, respectively. The investigated separators were initially allowed to swell in the applied feed solution and were then clamped between the chambers, which were mixed with magnetic stirring bars at 300 min⁻¹ to avoid concentration polarization [40]. Hydroxide diffusion was recorded in the draw chamber via pH measurement (pH meter CG 843, Schott, Mainz, Germany with BlueLine 16 electrode, Xylem Analytics Germany Sales, Weilheim, Germany) every 15 s using 6 M KOH (>90%, Sigma Aldrich Chemie, Taufkirchen, Germany) as the feed and distilled water as the draw solution. For zincate diffusion, 8.5 M KOH with dissolved 0.5 M ZnO (99%, Carl Roth, Karlsruhe, Germany) was applied as feed and 8.5 M KOH as draw solution similar to Kolesnichenko et al. [22], which was sampled after specific time intervals (0, 2.5, 5, 10, 20, 40, and 80 min). Generated samples were then diluted 100-fold with distilled water to reduce their alkalinity for the following Zn analysis via inductively coupled plasma-optical emission spectroscopy (ICP-OES 715ES, Varian, Palo Alto, CA, USA). Quantification was carried out at 206.2 nm using Ar as control and Zn standards for calibration. Diffusion experiments for D_OH⁻ and D_Zn(OH)4²⁻ were conducted in duplicates for each separator and the diffusion coefficients were calculated from Equation (5) [22].

$$\ln \left[{\mathrm{C}}_{\mathrm{A}}/\left({\mathrm{C}}_{\mathrm{A}}{ - \mathrm{C}}_{\mathrm{B},\mathrm{t}}\right)\right]=\left[\left({\mathrm{D}}_{\mathrm{x}} · \mathrm{A}\right)/\left({\mathrm{V}}_{\mathrm{B}} · \mathrm{L}\right)\right] · \mathrm{t}$$

(5)

Therein, C_A corresponds to the initial concentration of hydroxide or zincate ions in the feed chamber, respectively. C_B,t indicates the concentration of the measured analyte after specific time spans t and the volume of the draw chamber V_B. The slope derived from Equation (5) can then be narrowed to the respective diffusion coefficients D_x by applying the diameter of the diffusion cell connection pathway as the effective diffusion area A and the previously determined separator thickness as length L. The given standard deviation of the diffusion coefficients thereby considers the Gaussian error propagation from the separator thickness determination and resulting slopes from the diffusion experiments.

2.2.3. XRD Analysis

The structural properties of BC were investigated along the process chain using XRD analysis. In detail, purified and homogenized BC was analyzed in terms of cellulose crystal type and crystallinity. Additionally, different separators (BC-5, BC-10, BC-10-L, BC-10/BC-10-L) were analyzed after their application within Ni-Zn batteries. All separators that had been in contact with the electrolyte were thoroughly soaked and cleaned in distilled water several times to remove residues from the electrolyte, which would otherwise interfere with the cellulose structures. Prior to XRD measurements, all samples were freeze-dried and stored air-tight thereafter to avoid extensive humidification, which is known to alter the cellulose diffractograms [41]. An Empyrean series 2 diffractometer (MalvernPanalytical, Almelo, The Netherlands) with Cu K_α radiation (λ = 1.5419 nm, K_α2: K_α1 = 0.5, Cu LFF HR as X-ray source) and a PIXcel3D detector was applied to record diffractograms in the range 10–40° 2θ and a step size of 0.053°. The samples were provided on a Si sample holder and measured in reflection mode (Bragg–Brentano geometry). BC crystallinity was determined by peak deconvolution and assignment of an 8th-order real form Fourier series model for the amorphous content as proposed by Yao et al. [42]. The obtained diffraction data were pre-treated in terms of background subtraction of the empty Si sample holder, polarization correction, and data smoothing using a FFT filtering tool in PeakFit software (Version 4.12). Crystalline peaks were assigned as Voigt functions (Voigt Amp) based on the structure determined by Nishiyama et al. [43] and adapted by French [44]. The peaks were then allowed to be automatically scaled by PeakFit software to achieve high coefficients of determination [44]. However, final manual adjustments were necessary to ensure reasonable crystallographic validity and avoid erroneous peak broadening. Crystallinity was then calculated based on the proportion of the crystalline to the total assigned peak area solely considering cellulosic contributions. The I_α phase content for highly crystalline type I celluloses was calculated as described by Wada et al. [45].

2.3. Ni-Zn Battery

2.3.1. Cathode Production

Cathodes for Ni-Zn batteries were generated by electrochemical deposition of nickel(II) hydroxide (Ni(OH)₂) from nickel(II) nitrate (Ni(NO₃)₂, ≥99%, Carl Roth, Karlsruhe, Germany) solution on Ni foil (thickness 0.05 mm, 99.98%, Goodfellow, Hamburg, Germany). Therefore, the foil was initially roughened with emery paper and successively immersed in acetone, 0.1 M NaOH, and 0.24 M HNO₃ solution for 10 min in an ultrasonic bath [46]. It was then rinsed with distilled water and dried at 40 °C for approximately 3 h. Electrochemical deposition was conducted in a three-electrode cell consisting of the pre-treated Ni foil fixed with a clamp as the working electrode (WE), a flat graphite counter electrode (CE, 5.7 cm × 4.7 cm × 0.6 cm), and a saturated Ag/AgCl reference electrode (RE, Xylem Analytics Germany Sales, Weilheim, Germany) in 0.08 M Ni(NO₃)₂. Electrodeposition was executed using an MPG-2 battery cycler (BioLogic, Seyssinet-Pariset, France) in constant current mode at −2 mA cm⁻² as described before [47,48,49], which resulted in an average measured voltage of about −1 V vs. Ag/AgCl (sat.) during the process. Here, a solely one-sided Ni(OH)₂ deposition was ensured by covering the backside of the Ni foil with isolating vinyl tape. The required deposition time and thus electrical charge Q was set by the targeted mass loading of 0.59 mg cm⁻² Ni(OH)₂, which was initially estimated by Faraday’s law (Equation (6))

$$\mathrm{Q}=(\mathrm{m} · \mathrm{F} · \mathrm{n})/\mathrm{M}$$

(6)

where m and M represent the desired mass and molecular weight of Ni(OH)₂, respectively. F is the Faraday constant and n = 1.6 the quantity of required electrons per deposited Ni atom [46,48]. As the efficiency of the electrochemical process is known to be easily overestimated by Faraday’s law [48,50], the mass loading was consistently checked gravimetrically and showed a lower than targeted but consistent mass loading of 0.39 mg cm⁻² Ni(OH)₂. The as-prepared Ni foils were finally punched to obtain cathodes with a diameter of 1.8 cm and 1 mg of cathodic active material.

2.3.2. Battery Assembly and Electrochemical Cycling

Battery cycling and C-rate investigations were conducted in an ECC-Std electrochemical test cell (EL-Cell, Hamburg, Germany) consisting of a Cu current collector (0.025 mm, 99.999%, Goodfellow, Hamburg, Germany), Zn foil (0.05 mm, 99.95%, Goodfellow, Hamburg, Germany) as the anode, and Ni foil with electrochemically deposited Ni(OH)₂ as the cathode. A glass-fiber (GF) separator (18 mm × 1.55 mm, EL-Cell, Hamburg, Germany) served as reference separator, and 500 µL of 6 M KOH with 20 mM zinc acetate dihydrate (Zn(OAc)₂ × 2 H₂O, reagent grade, Sigma Aldrich Chemie, Taufkirchen, Germany) was applied as electrolyte [51,52]. Deviating from the GF separator, BC separators were soaked in electrolyte prior to cell assembly to account for the visible and previously described initial shrinkage of cellulosic separators upon contact with the alkaline electrolyte [27]. Upon cell assembly, a 30 min resting phase was followed to allow for temperature equilibration at 25 °C and sufficient electrode wetting. Galvanostatic cycling was then conducted at 0.115 mA cm⁻² and thus C-rate of 1C, referring to the initially cell-limiting cathode capacity of 0.292 mAh (292.2 mAh ${\mathrm{g}}_{\mathrm{NiO}\left(\mathrm{OH}\right)}^{-1}$). Charging was conducted up to 1.9 and 1.85 V for GF and BC separators, respectively, with a cut-off voltage of 1.5 V. The first three charge and discharge cycles at 1C were considered as cell formation cycles initially showing flat voltage plateaus at the top of charge, which have been attributed to initial cathodic Ni(OH)₂ activation [53]. The calculated state of health was defined as the relative remaining capacity in relation to the initial discharge capacity after the cell formation phase. Charge/discharge rate tests were carried out at 0.5, 1, 2, 4, and 6C with three charge and discharge cycles for each condition.

3. Results and Discussion

3.1. BC Comminution and the Effect on BC Crystallinity

The separator production started with the comminution of the purified BC material by shredding with a dispersing instrument followed by several homogenization steps using an HPH. The progress was followed microscopically to prove the continuous comminution of the BC agglomerates. Figure 1a still shows fibrous BC after applying a disperser to initially shred the macroscopic biopolymer structure. This is strongly reduced in Figure 1b after the second homogenization step with an HPH, although thicker BC fiber bundles are still visible. After the fourth application of the HPH, no further alterations of the BC structure were microscopically detectable (Figure 1c,d). The reduction of the BC fiber size and increasing uniformity with increasing HPH cycles is thereby expected and in agreement with previous investigations applying different methods to homogenize BC [54,55]. The alteration of the BC appearance from loose fibers to denser entangled BC bundles is also confirmed and represents a result of the shear forces that lead to fragmentation and defibrillation of the BC fibers [54].

In addition to the microscopic investigation, XRD analysis followed by peak deconvolution was conducted to investigate effects of the homogenization process on the structural characteristics of BC. The detected main reflections of purified BC with their respective Miller indices at 14.7° (100), 17.0° (010), and 22.9° (110) in Figure 2a were assigned to the I_α structure, which is characteristic for celluloses with bacterial origin [43,44,45]. However, the phase content of the I_β allomorph is indicated by slight signal shifts from the I_α model structure. Therefrom, the phase content of the respective allomorphs I_α and I_β can be estimated from the diffraction patterns. This results in a I_α share of 82% for the purified BC, which agrees with previous investigations where the I_α content ranged from 69 to 86% for BC from K. xylinus strains using different approaches of allomorph structure determination [56,57,58,59]. Besides the I_α share, the polymer crystallinity represents an important structural property and can be especially high for BC [60]. The applied peak deconvolution method for its determination resulted in a crystallinity of 70% for the produced and purified BC, which is similar to 63–74.1% BC crystallinity determined in previous studies using an equivalent determination approach [58,60,61].

The BC processing, however, altered the structural features to some extent (Figure 2b): the I_α share is reduced to 74% and the crystallinity decreased to 59%. Both can be related to the initial mechanical shear forces occurring in the disperser and to the compression/decompression processes during the HPH cycles [54,55]. A reduction in crystallinity after several HPH cycles has been reported before and was attributed to the debonding of surface crystallites [55]. The partial conversion of the metastable I_α into the stable I_β phase could be a result of short thermal effects due to the cavitation occurring in HPH processes [62]. A susceptibility towards high temperatures and thus conversion of the I_α allomorph has been pointed out before [43,63]. Overall, the investigations in parallel to the BC comminution process prove microscopically a successful homogenization of the BC, which was accompanied by alterations of the crystalline structure that should be kept low as higher crystallinities are associated with higher mechanical strength [34].

3.2. Dimensional Properties and Interactions with the Electrolyte

The separators produced from the homogenized BC via solution casting and freeze-drying showed differing appearances (Figure 3). BC-5 displays a slightly irregular, sheer surface and thus differs highly from the dense and homogenous BC-10 membrane, which indicates a better applicability of the latter one due to a better mechanical rigidity. In contrast to the casted separators, BC-10-L is a highly porous and volumetric membrane that could thus stand out due to a higher electrolyte uptake capability.

The separator characterization was initiated with the determination of their dimensional properties (

Table 1. Dimensional properties of produced bacterial cellulose (BC) separators in comparison to commercial references.

Separator	Thickness, Dry [mm]	Thickness, Wet [mm]	Area Shrinkage ΔA [%]	Swelling Degree ΔV [%]
Glass-fiber	1.55 [71]	-	-	-
Polyolefin	0.025	-	-	-
BC-5	0.023 ± 0.005	0.062 ± 0.013	39.3 ± 1.7	63.5 ± 59.0
BC-10	0.045 ± 0.005	0.109 ± 0.013	39.3 ± 5.6	45.3 ± 27.0
BC-10-L	1.54 ± 0.15	0.64 ± 0.19	58.1 ± 3.4	−82.4 ± 9.5

). A greater thickness of the BC-10 membrane in comparison to BC-5 is reasonable since more BC was used during its preparation. The lyophilization of the BC-10-L sample solution resulted in a highly porous membrane with a dry thickness similar to the GF separator, whereas the casted membranes are equal to the polyolefin separator. Electrolyte soaking caused a 2.7- and 2.4-fold increase in thickness of BC-5 and BC-10, respectively, whereas the area shrank by 39.3%, which has been reported for BC in alkaline environment before [27,64]. These structural changes are attributed to several and partially counteracting processes occurring in parallel: The increasing thickness mainly derives from electrostatic repulsion due to an excess of carboxylate anions from the CMC incorporated in the polymer matrix [65,66] and the swelling of amorphous content of the BC [34,67,68]. The equally longitudinal and lateral shrinkage of the membranes is probably caused by the partial decrystallization of the microfibril structure due to the alkaline electrolyte. The initially crystalline and straight microfibrils rearrange into randomly oriented cellulose chains that contract and thus cause a shrinkage of the cellulosic membrane, which at the same time causes a small lateral expansion [69]. The latter is, however, expected to be less pronounced and thus compensated in case of BC-5 and BC-10 by the shrinkage. The comparably higher area shrinkage for the BC-10-L might be due to the lack of CMC in this membrane type: The electrostatic repulsions occurring in BC-5 and BC-10 seem to partially compensate the contraction of the decrystallized fibers which explains the higher area shrinkage and reduced swelling degree of BC-10-L. However, the swelling effects due to amorphous parts within the BC microfibrils highlight the impact of crystallinity and the potential for its targeted adjustment to tailor BC separator properties such as the swelling degree, which is, for example, much lower for hydrophobic polyolefin separators, such as Celgard 3501, with only 3% [27,34,70].

The electrolyte-soaked separators were additionally characterized in terms of the electrolyte uptake and its retention at ambient conditions. BC-5 showed the highest relative electrolyte uptake in relation to its dry weight (Figure 4a), which is the result of a higher surface to volume ratio than for BC-10 or handling difficulties during the removal of surface-associated electrolyte residues due to its mechanical sensitivity. However, the swelling degree (Table 1) also indicated a potentially high electrolyte uptake in relation to its dry weight for BC-5, which corroborates the result. Volume increases of similar extent due to polymer swelling in alkaline environment was pointed out for BC and other separator materials such as cellophane, polyvinyl alcohol (PVA), and Nafion before [27,34,72]. Interestingly, BC-10 had a much lower electrolyte uptake than the BC-10-L, which is attributed to the highly porous structure of the latter. This might additionally indicate the advantage of a greater electrolyte reservoir within the battery cell when using BC-10-L. The commercial GF separator equals the BC-10 in this context although its porous structure is more like BC-10-L (Figure 4a). This is most likely due to the differing uptake mechanisms: the electrolyte almost solely penetrates the available pores in GF whereas BC-10 additionally swells and thus reaches a similar electrolyte uptake despite its smaller wet thickness.

The electrolyte retention of the investigated separators shows differences between GF and BC separators (Figure 4b). Both casted separators BC-5 and BC-10 could retain <75% of their electrolyte content within the first hour, which was more for BC-10-L (86%) or the GF (94%) separators. After 19 h, BC-10 and GF retained less than 70% of their electrolyte content, whereas the course of electrolyte loss differs. After a fast initial drop of the electrolyte content in BC-10, it remained almost constant afterwards, which was shown for cellulosic hydrogels before [65]. The different evaporation course could indicate in this case an initial evaporation of surface-associated water followed by a much slower evaporation of electrolyte stored within the separator pores. Due to highly differing surface to volume ratios of BC-10 and the GF separator, this difference might be mostly attributed to geometric effects. This theory is also supported by the initially slower electrolyte loss for the voluminous BC-10-L. For a PVA separator, an electrolyte retention of about 65% was reported after 12 h, showing an almost linear loss in electrolytes [39]. However, the combination with polyacrylic acid (PAA) and other additives lead to retention values > 90% after 12 h, which highlights the potential of polymer combinations. BC-5 and BC-10-L retained less than 60% of their weight after 12 h, which, however, is not expected to occur in this extent within a sealed Ni-Zn battery cell but must be considered for applications in half-open Zn-air batteries [73,74].

3.3. Seperator Permeability for Hydroxide and Zincate Ions

For a successful application of separators within a Ni-Zn cell, the electrolyte and separator need to allow for a high hydroxide ion transport but must hinder the crossover of zincate ions. Hence, the permeability of both ions for the commercial and the produced BC separators BC-10 and BC-10-L was investigated in an H-cell (

Table 2. Diffusion coefficients and selectivity for hydroxide and zincate ions for the investigated separators.

Separator	DOH− · 10−9 [m2 s−1]	DZn(OH)42− · 10−9 [m2 s−1]	Selectivity [-]
Glass-fiber	63.2 ± 5.0	13.2 ± 3.4	4.8 ± 1.3
Polyolefin	0.17 ± 0.02	0.018 ± 0.003	9.5 ± 2.0
BC-10	1.6 ± 0.2	0.036 ± 0.012	45.1 ± 16.5
BC-10-L	125.1 ± 37.2	3.0 ± 1.0	41.2 ± 17.9

). BC-5 was not characterized here due to its lacking mechanical rigidity. The flexible but stable BC-10 and BC-10-L allowed for distinctly better handling in dry and wet state similar to previous reports where high tensile strengths of BC membranes have been described [17,31]. However, this favorable property was shown to lessen for cellulose membranes during the incubation in alkaline electrolytes unless the polymer crystallinity is high as it is the case for BC [34].

From the investigated membranes, BC-10-L showed the highest permeability for hydroxide, followed by GF. This difference is mainly attributed to differing structural properties: The thickness of BC-10-L is halved compared to GF because of the shrinkage upon electrolyte soaking. This could cause a reduced hydroxide diffusion for GF due to a higher tortuosity and a longer diffusion pathway. BC-10 and the polyolefin separator had remarkably lower permeability for hydroxide, which is attributed to smaller pores and thus a decreased diffusion (Figure 5). BC-10 allowed for improved hydroxide diffusion in comparison to the polyolefin separator, since the discrete and almost unswollen pore structure of the latter might be smaller and less hydrophilic than cellulosic polymer, which additionally leads to inferior hydroxide diffusion processes. This thereby agrees with the findings that a cellophane membrane had a higher hydroxide diffusivity and higher ionic conductivity than a polyolefin-based Celgard 3501 separator [22]. However, in previous investigations a 15.5-fold greater hydroxide diffusion coefficient was determined for Celgard 3501 than for the herein investigated commercial polyolefin separator [22]. This discrepancy is thought to be due to differences in the respective polyolefin composition or to methodical differences that have been reported to greatly affect the total values of the zincate diffusion coefficients [75].

The structural effects of the investigated separators are also reflected by the zincate diffusion coefficients. Once again, the highly porous separators GF and BC-10-L facilitated higher zincate permeability, whereas in BC-10 and polyolefin membranes, the steric hinderances decreased the ion transport capability. The determined values for zincate ion permeability for the polyolefin separator with 0.018 ± 0.003 m s⁻² are within the range of 0.0092–0.0387 · 10⁻⁹ m s⁻² [21,22,70,72] reported for the frequently investigated Celgard 3501, which validates the general methodical approach and reference separator choice, although deviations are expected due to varying testing conditions or differing material properties. Interestingly, both BC separators not only provide high hydroxide (BC-10-L) or low zincate permeability (BC-10). The selectivity, representing the ratio between the hydroxide and zincate diffusion coefficient, provides evidence for a better applicability of the BC-based separators than the commercial ones as an improved differentiation between hydroxide and zincate ions is possible and could favor longer battery cycle life and enhanced coulombic efficiency. The selectivity obtained for the polyolefin separator with 9.5 ± 2.0 is thereby in proximity to 1.2 ± 0.2 for the commercially available Celgard 3501, which was determined by Kolesnichenko et al. [22], and thus corroborates the tendency to lower ion differentiation of polyolefin-based separators.

3.4. Ni-Zn Battery Cycling

Following the physical characterizations, the produced separators were tested during electrochemical cycling in Ni-Zn batteries. At first, three BC-5, one BC-10, and one BC-10-L membrane were applied as the separator. Three BC-5 were used in a serial configuration to compensate for the lower mechanical rigidity of this separator type and thus ensure the separation of the anode and cathode. The results of the initial cycling experiments in Figure 6 confirm the low applicability of the BC-5 membranes, since only three charge/discharge cycles were feasible. After those, a sudden drop of the cell potential to less than 0.3 V occurred during the charging procedure, which could not be overcome afterwards. This suggests a potential contacting of the anode and cathode, passivation of an electrode, or extensive water loss. The latter one would be unlikely after these few cycles unless the separator was squeezed during battery assembly or the wetting of both electrodes consumed a great share of the electrolyte volume. The lack of water could then relate the low coulombic efficiency with high internal resistances (Figure 6b). Additionally, parasitic side reactions, such as oxygen evolution reaction (OER) or hydrogen evolution reaction (HER), could have occurred, which can severely affect battery performance and lifetime [76]. The HER would also initiate the formation of ZnO dendrites that were found to reduce the cycle life of MnO₂-Zn batteries when applying thin BC-separators [17]. For that reason, a short circuit due to the low mechanical rigidity of BC-5 membranes or ZnO dendrites must have caused a contacting between the electrodes and led to cell failure.

The BC-10 membrane showed much better mechanical stability and suitability, which is probably due to a denser biopolymer fiber network and enabled a longer battery lifetime with 28 cycles. After this number of cycles, water consumption via HER or Zn corrosion could have critically contributed to the dry out of the separator [77,78]. In Zn-air batteries, the overall electrolyte loss was quantified to be 34.6% after 24 h battery cycling and was assigned to HER, Zn corrosion, and evaporation. The effects of continuous water consumption were, however, not electrochemically detectable until shortly before the cell failure, which coincides here with the sudden capacity decrease during the last two discharge cycles (Figure 6a) [78]. However, soft short circuits due to the build-up of precipitated ZnO within the separator because of zincate supersaturation in the electrolyte (~40 µL) [79], Zn migration to the cathode due to the Zn²⁺-rich electrolyte [11,12], or passivation of the Zn anode [80] could also account for the early cell failure.

The lyophilized BC-10-L separator facilitated a coulombic efficiency of about 90% after 20 cycles and, overall, 100 charge and discharge cycles were achieved, which is attributed to its higher electrolyte content (~130 µL) and hydroxide permeability, albeit a continuous capacity fade was found leading to a final capacity retention of 22%. The latter might be due to anode passivation or its zincate permeability (Table 2), which can lead to a migration and incorporation of Zn into the cathode and thus a decrease in capacity [11,12]. Similarly, the coulombic efficiency of both BC-10 and BC-10-L exhibited an ongoing increase after 10 cycles, which indicates alterations of the electrochemical processes that could be related to each other. An increase of the coulombic efficiency can imply a lower cell resistance or less side reactions leading to less energetic losses. Another explanation approach could result from the cellulosic material itself: as discussed for the results presented in Table 1, the alkaline electrolyte possibly interacted with the amorphous content of the cellulose, which might be ongoing during cycling and caused structural changes that could enlarge pores in the separator membrane, which in turn would enhance ion diffusion.

The conclusions drawn from these cycling experiments suggest possible improvements for the further application of the produced separators. Since BC-10 could be susceptible to dehydration due to a lower electrolyte uptake (Figure 4a) and BC-10-L to zincate migration to the cathode (Table 2), their advantageous properties of low zincate permeability and high electrolyte content or hydroxide diffusion, respectively, were combined by applying the separators simultaneously. Therefore, BC-10 was placed adjacent to the Zn anode, followed by BC-10-L and the cathode. The results of the electrochemical cycling are shown in Figure 7 with a commercial GF separator as reference. The initial specific discharge capacities of the Ni-Zn batteries differed between both separators with the GF cell having 1.3-fold higher capacity than the BC-10/BC-10-L one (Figure 7a). This could arise from the lower total electrolyte uptake of the BC separators (~170 µL) compared to GF (~440 µL), as the contained Zn(OAc)₂ can act as an additional small anode capacity reservoir next to the anode [11]. This could partially explain the higher specific discharge capacity of BC-10/BC-10-L compared to the stand-alone application of BC-10 or BC-10-L (Figure 6). However, the overall specific capacities were within the range of other Ni-Zn batteries of similar composition [52,81,82]. Another explanation lies in the charging voltage of 1.85 V compared to 1.9 V for the GF separator, which was selected as the cell voltage of BC-10/BC-10-L cells could only reach 1.9 V after prolonged charging steps that can be accompanied by Zn redistribution and side reactions that should be avoided.

In regard of the cell cycling course, the capacities of both cell setups remained almost constant for about 10 charge/discharge cycles (Figure 7a), which is different from the BC-10-L (Figure 6) where the capacity depleted from the very beginning. The state of health shows a slower decrease for GF with 74 and 58% after 50 and 100 cycles, respectively, whereas BC-10/BC-10-L could retain 53 and 27%. Thus, both battery configurations underwent a continuous capacity loss but with differing separator-based side effects as indicated by the continuously rising coulombic efficiency for BC-10/BC-10-L with increasing cycle number and decreasing one for GF from the 20th cycle on. The most common reasons for capacity fading in alkaline electrolytes when using Zn anodes are the diffusion of zincate from the vicinity of the electrode and the passivation of the anode surface by layers of precipitated ZnO, which can be exacerbated by HER, which is likely to have occurred here, too [3]. The steady loss of capacity is thereby comparable to previous studies where this phenomenon was mainly ascribed to insulating ZnO precipitation due to its supersaturation within the drying separator [17,76,79].

The different extent of capacity loss is a result of several overlapping effects: Due to the overall lower total electrolyte uptake of BC-10/BC-10-L in comparison to GF, the supersaturation of zincate and thus precipitation of ZnO occurred earlier at the anode/separator boundary layer. This must have been intensified by the greater zincate shielding capability of BC-10/BC-10-L, which led to an accumulation of zincate in the vicinity of the anode and thus precipitation of ZnO. This is corroborated by the whitish deposit that was found on the anode surface after cell disassembly and was more pronounced for batteries with the BC separator (Figure S1). A 10% lower electrolyte retention determined for stand-alone separators might have additionally intensified this process (Figure 4b). Based on these results, an additional or more electrolyte retaining BC-10-L separator layer could have lessened capacity fading and allowed it to benefit from the zincate shielding capability of the BC-10 separator. The application of Zn anodes with a high surface area, coating, and heavy metal additives could additionally counteract the cell aging processes [3].

The initial coulombic efficiency was similar for both Ni-Zn cell set-ups, whereas it increased sharply for the GF up to about 93% with high similarity to Hu et al. [81] and linearly for BC-10/BC-10-L during the whole cycling experiment, which could indicate differing aging mechanisms or alterations of the separator structure (Figure 7b). Possibly, the lower hydroxide permeability of BC-10 compared to GF resulted in an initially lower charge transfer during the charging process, which in turn caused overpotentials and thus energetic losses due to OER. The improving coulombic efficiency is most likely due to favorable structural changes of the BC separators. The alteration of the semi-crystalline material discussed in Section 3.2 could continue in the alkaline electrolyte and thus improve the hydrophilicity or the ionic conduction, which will be evaluated in Section 3.6. Another explanation could be derived from Figure 4b, where a better electrolyte retention was found for GF than for BC-10 or BC-10-L. The loss or continuous consumption of water via HER could increase the KOH concentration in BC-10/BC-10-L, which in turn could improve the conductivity to a small extent, although this potential enhancement would be limited since a conductivity decrease can already be expected around 6.5 M KOH [83,84]. The decrease in efficiency for GF from the 20th cycle on could be attributed to parasitic side reactions, such as the OER or HER, during the charging and discharging procedures.

The exemplary discharge voltage profiles (Figure 7c) after specific cycle numbers revealed a discharge plateau at about 1.73 V, which is equal to previous observations [36,52,82]. However, the capacity fading with increasing cycle life is indicated by a steepening voltage profile, which is more pronounced for BC-10/BC-10-L than GF, which correlates to the slower aging of the cell with a GF separator in this configuration.

3.5. C-Rate Dependence

The rate capability of the GF and BC-10/BC-10-L separator is displayed in Figure 8a. The initial cycles resemble Figure 7 in terms of capacity retention and coulombic efficiency but change upon a reduced C-rate. The cell capacity for GF is increased whereas it is slightly reduced for BC-10/BC-10-L. The former is expected based on the results of previous rate capability investigations in Ni-Zn cells [35,52,81], and a decreasing coulombic efficiency at decreasing C-rates has also been shown before [81].

With increasing current densities, the state of health for each Ni-Zn cell is reduced in addition to the general capacity fading, which was discussed in Section 3.4. The losses for both separators due to higher C-rates were very similar and only the values obtained for the BC separator at 6C seems to stand out. Since the difference in state of health between both cell types was about 7% at this point anyway (Figure 7a), this deviation can be ascribed to the faster cell aging process. A sixfold increase of the C-rate from 1 to 6C caused a capacity reduction of 21.6% (Figure 8b), although the state of health would have already been reduced by 10% after these cycles at 1C. Hence, the capacity loss due to higher current densities is to be expected around 10–15%, which has been reported before for Ni-Zn batteries and proves a good rate capability for the BC-10/BC-10-L separator [35,52]. Once the initial current densities at 1C were reached again, the course of the cell capacity for GF and BC-10/BC-10-L was in high agreement with Figure 7a,b, which shows that short-term cycling at higher C-rates did not further impair the Ni-Zn battery.

3.6. XRD Analysis of BC Separators

The comminution process at the beginning of the separator production has been shown to affect the cellulose structure (Figure 2). Since it was hypothesized that the contribution of the amorphous content of the cellulose would play a role in the separator swelling, possible alterations of the BC structure upon contact with the electrolyte were followed via XRD measurements. Therefore, the crystalline structure of BC-10 (Figure 9a) and BC-10-L (Figure 9b) was investigated after the separator production, swelling in the electrolyte, and after the electrochemical cycling experiments.

Following their production, the XRD patterns of both BC membranes were still in high accordance with the I_α triclinic unit cell based on the diffraction signals, although the diffraction pattern of BC-10-L is slightly shifted towards higher 2θ values, which is attributed to its three-dimensional structure that can complicate X-ray beam focusing. Still, the pattern is reasonable and agrees with BC-10. The crystallinities of 62 and 61% for BC-10 and BC-10-L, respectively, were somewhat higher than the one determined for purified BC with 59%, which is within the methods accuracy range and does not indicate severe structural alterations. Changes of the BC-10 diffraction pattern due to the addition of CMC were not detectable and would have exhibited a broad diffraction signal around 20.9° due to the semi-crystalline nature of the material [85].

Soaking of both separators in the alkaline electrolyte solution followed by thorough washing with distilled water altered their crystalline structure. The diffraction signals with Miller indices (110) and (010) were barely detectable for BC-10 and are completely absent in case of BC-10-L. This difference is probably due to the different wetting behavior, as the BC-10-L membrane had a much greater contact area towards the electrolyte because of its highly porous structure. The (110) reflection was broadened (BC-10) or extended by two overlapping diffraction signals (BC-10-L) from the cellulose type II at 19.95° (110) and 22.1° (020), indicating a partial mercerization of the cellulose structure [59,64,67]. Thereby, the crystallinity is also reduced for both samples to 30%, which is similar to previous investigations where a decrease in crystallinity from 66 to 40% was reported after a strong alkaline treatment [59]. It additionally proves a partial degradation of the cellulose structure towards a greater amorphous content, which has been pointed out before and was found to depend on the concentration of the alkaline solution [67,86]. This underlines that the differences in shrinkage and swelling behavior between BC-10 and BC-10-L (Table 1) must have been caused by more than amorphous swelling, since otherwise the dimensional alterations should have been more similar. For that reason, additional effects such as electrostatic repulsion due to the CMC content in BC-10 become more evident.

Following the cycling in the Ni-Zn batteries, the structural changes towards cellulose II became more pronounced for both separators. The expected rise of the cellulose II reflection at 12.1° (1–10) is hardly visible in the diffraction patterns [44]. Interestingly, BC-10-L showed three distinct signals at 32.2°, 34.6°, and 36.8° after the battery cycling, which cannot be attributed to one of the cellulose allomorphs. Figure 10a shows that these reflections correspond to ZnO (ICSD collection code: 166353), which was trapped within the porous separator. The precipitation of ZnO and clogging of separator pores during the cycling of Zn batteries has been reported before and represents a severe cause of cell aging [11,79].

Signs of ZnO were also detected in BC-10-L when BC-10 and BC-10-L separators were applied simultaneously, although the diffraction signals were much less pronounced (Figure 10b). This indicates that the simultaneous application of BC-10 and BC-10-L enhanced shielding of the zincate ions at the anode, which was assumed before due to a lower diffusion coefficient of BC-10 for zincate ions (Table 2) and can thus be attributed to a size-exclusion-based blocking effect [22,36,75]. It additionally suggests that one major reason for the determined capacity loss found in BC separators could be the formation of zincate and the following precipitation of ZnO at the anode/separator boundary layer as laid out in Section 3.4. Due to the greater dimension and higher zincate permeability of the GF separator, more zincate could be incorporated into the separator instead of its precipitation at the anode surface and was thus lost for further cycle numbers, which was to some extent compensated by the excess of Zn from the anode. In this electrode set-up and without further additives the desirable property of BC-10/BC-10-L of less active material loss cannot be fully utilized as water consumption reactions reduced the solvation of ZnO. However, it is of great promise in batteries with more advanced anode and electrolyte compositions and can thus allow for better active material utilization, reduced dendrite formation, and the switch to a bio-based separator material.

4. Conclusions

This investigation aimed at applying BC as separator in Ni-Zn batteries to further reduce the environmental impact of rechargeable Zn-based batteries as an important beyond lithium-ion technology. Therefore, the biotechnologically produced biopolymer was purified and different approaches, such as solvent-casting and lyophilization, were applied for separator manufacturing. The resulting BC separators were characterized in terms of their dimensional properties and interaction with the alkaline electrolyte that was found to cause alterations of the cellulose structure and by that dimensional shrinkage and swelling of the separators. The BC crystal allomorph I_α partially transformed into the thermodynamically more stable type II, which was accompanied by a reduction in polymer crystallinity from 59 to 30%. The permeability of dense BC-10 and porous BC-10-L separators for zincate and hydroxide ions was characterized and compared to commercial ones. For the porous BC-10-L separator, the highest hydroxide and second highest zincate ion diffusion coefficient were determined based on the investigated samples together with a high electrolyte uptake. In contrast, the thinner and denser BC-10 membrane showed high zincate shielding capability at the cost of lower hydroxide permeability. Electrochemical cycling, ion diffusion, and XRD analysis of stand-alone BC separators revealed zincate migration, anode passivation, and water loss as possible causes of cell failure or aging. The combination of the zincate shielding BC-10 separator and the BC-10-L separator with higher electrolyte storage capability led to slightly slower cell aging and less ZnO to be found in the BC separator pores. The capacity retention after 50 and 100 cycles was 53 and 27%, respectively, which might be improved by higher polymer crystallinity induced by improved biotechnological production processes, denser BC membranes for enhanced water retention, or by combinations of BC with other hydrophilic biopolymers that can favor electrolyte retention. C-rate tests showed comparable dependencies of the discharge rate from the current densities thus proving that BC separators show high promise for potential use as tailored separators with zincate blocking or electrolyte retaining properties in various Zn-based battery systems.