Research Article Cite This: ACS Appl. Mater. Interfaces XXXX, XXX, XXX−XXX
www.acsami.org
Rheological Investigation on the Microstructure of Fuel Cell Catalyst Inks Sunilkumar Khandavalli,† Jae Hyung Park,‡ Nancy N. Kariuki,‡ Deborah J. Myers,‡ Jonathan J. Stickel,§ Katherine Hurst,† K. C. Neyerlin,† Michael Ulsh,† and Scott A. Mauger*,† †
ACS Appl. Mater. Interfaces Downloaded from pubs.acs.org by UNIV OF WINNIPEG on 12/10/18. For personal use only.
Chemistry and Nanoscience Department, National Renewable Energy Laboratory, 15013 Denver West Parkway, Golden, Colorado 80401, United States ‡ Chemical Sciences and Engineering Division, Argonne National Laboratory, Argonne, Illinois 60439, United States § National Bioenergy Center, National Renewable Energy Laboratory, Golden, Colorado 80401, United States S Supporting Information *
ABSTRACT: We present a rheological investigation of fuel cell catalyst inks. The effects of ink parameters, which include carbon black-support structure, Pt presence on carbon support (Pt−carbon), and ionomer (Nafion) concentration, on the ink microstructure of catalyst inks were studied using rheometry in combination with ultrasmall-angle X-ray scattering (USAXS) and dynamic light scattering (DLS). Dispersions of a high-surface-area carbon (HSC), or Ketjen black type, demonstrated a higher viscosity than Vulcan XC-72 carbon due to both a higher internal porosity and a more agglomerated structure that increased the effective particle volume fraction of the inks. The presence of Pt catalyst on both the carbon supports reduced the viscosity through electrostatic stabilization. For carbon-only dispersions (without Pt), the addition of ionomer up to a critical concentration decreased the viscosity due to electrosteric stabilization of carbon agglomerates. However, with Pt−carbon dispersions, the addition of ionomer showed contrasting behavior between Vulcan and HSC supports. In the Pt− Vulcan dispersions, the effect of ionomer addition on the rheology was qualitatively similar to Vulcan dispersions without Pt. The Pt−HSC dispersions showed an increased viscosity with ionomer addition and a strong shear-thinning nature, indicating that Nafion likely flocculated the Pt−HSC aggregates. These results were verified using DLS and USAXS. Further, the observations of the effect of ionomer:carbon ratio and a comparison between carbons of different surface areas provided insights on the microstructure of the catalyst ink corresponding to the optimized I/C ratio for fuel cell performance reported in the literature. KEYWORDS: proton-exchange membrane fuel cells, catalyst inks, rheology, platinum, carbon, ionomer, catalyst layer
■
INTRODUCTION Polymer electrolyte membrane fuel cells (PEMFCs) are a promising energy technology for many applications such as automobiles and portable or stationary power sources due to their clean emissions and high energy-conversion efficiency. A PEMFC produces electricity through electrochemical reactions of H2 and O2 to produce water. A critical component of a fuel cell is the catalyst layer (CL), where these electrochemical reactions occur. The microstructure of the CL controls the transport properties of electrons, protons, reactants, and products, which are key for fuel cell performance.1−3 One of the current challenges is the lack of a rational approach to efficiently design the CL with an optimal microstructure for fuel cell performance.2,3 The CL microstructure is primarily controlled by the nature of fabrication, which is commonly solution processing of an ink dispersion. The predominant catalyst ink formulation for PEMFCs consists of a platinum (Pt) nanoparticle catalyst, carbon black (CB, acting as a support surface for Pt), and a © XXXX American Chemical Society
perfluorosulphonic acid ionomer (which mainly serves as a proton-conducting phase in the CL), commonly Nafion. These are dispersed in water−alcohol solvent at specified ratios to form an ink. To fabricate a CL, the ink is coated (e.g., spraycoating, rod-coating, or hand-painting) onto a substrate (porous carbon-fiber diffusion media, polymer electrolyte membrane, or decal) and dried to form the CL.3 The resulting microstructure of the CL is a complex three-dimensional (3D) distribution of Pt−CB−ionomer components and pores across multiple length scales. An optimal microstructure of CL is desired for the best performance.2 Studies on high-performance CL microstructure have suggested that an optimal microstructure is characterized by the following:2−6 (i) a maximum Pt−CB−ionomer interaction or three-phase interface for efficient platinum utilization, (ii) a contiguous ionomer Received: August 30, 2018 Accepted: November 30, 2018
A
DOI: 10.1021/acsami.8b15039 ACS Appl. Mater. Interfaces XXXX, XXX, XXX−XXX
Research Article
ACS Applied Materials & Interfaces
(USAXS).18,21 These studies provided a structural picture of catalyst ink components, where a strong interaction of ionomer with Pt catalyst particles was observed. Yang et al.21 have elucidated that electrostatic interactions play a role in the interaction between Pt and ionomer by examining the effect of carbon surface charge modification on interaction with ionomer. Molecular dynamics simulations examining the interactions between ionomer and Pt−CB on the microstructure of inks13,14,19 have shown that the ionomer interacts differently between with CB than Pt−CB due to preferential hydrophobic and hydrophilic interactions from the segments of ionomer chains. As a result, the ionomer configuration and its distribution on CB surface were found to differ between CB and Pt−CB.13,14,19 There are a few experimental studies examining the effect of ionomer addition to carbon (but without Pt) on dispersion stability through a combination of sedimentation, electrophoretic mobility, and rheology experiments.22−25 The ionomer was observed to stabilize the carbon aggregates via electrosteric stabilization. Catalyst ink formulations can vary in carbon-support propertiessurface area, pore volume, and graphitization as noted above. Depending on the carbon-support type, the pore-size distribution and the Pt distribution on the support (internal vs external) were found to vary.26−28 Therefore, the carbon-support type was found to influence the optimal I/C from in situ electrochemical tests of fuel cell performance.28 However, most studies on catalyst ink microstructure have focused on single carbon-support type.15,16,18 We are unaware of any studies in the literature discussing the effect of carbonsupport type on the catalyst ink microstructure. Additionally, many of these studies used bare carbons (without Pt) and thus may not be fully representative of a real catalyst ink as the influence of the Pt on ink microstructure was not considered. To efficiently optimize catalyst ink formulation, understanding the interaction between Pt−CB and ionomer and the impact of bulk microstructure is critical. Experimental studies on complete catalyst ink systems consisting of Pt−carbon and ionomer15,16,18 are relatively few, and most of these studies are limited to an optimized ink formulation (i.e., a fixed I/C). Systematic investigations examining the effect of I/C on the Pt−CB ink microstructure and the underlying interactions are still lacking. Rheology is a promising method to quantitatively probe the bulk microstructure of inks. The rheological (or flow) behavior of complex fluids is very sensitive to the microstructure.29 Furthermore, rheological properties play a critical role in processing behavior during the fabrication of CL.30 Depending on the coating method and processing conditions, the rheological properties of the ink strongly influence both the coating characteristics (i.e., uniformity, thickness, and ink penetration into a porous substrate)23,31−33 and the macroscopic structure of CL, such as percolated network of both carbon and ionomer, and the pore structure of CL.23,34,35 Although there have been several rheological investigations on understanding the microstructure of carbon black dispersions36−40 and ionomer solutions41 and some on dispersions of ionomer and carbon mixture,23,25 we are unaware of any published rheological studies on catalyst inks or dispersions consisting of both Pt−CB particles and ionomer. Understanding the rheology of catalyst inks is important toward establishing structure−property−processing relationships. In this paper, we present a rheological study of PEMFC catalyst inks. We investigate the interactions between Pt−CB
distribution across the catalyst layer structure for efficient proton conductivity, (iii) a maximum carbon−carbon 3D connectivity for electron conductivity, and (iv) a certain porosity of the electrode for efficient mass transport of reactants and products. The microstructure of the catalyst inks, which is governed by the underlying interactions between Pt−CB and ionomer solvent, and how it evolves in the fabrication process of CL, that is, during coating5 and drying,7 plays a critical role in formation of an optimal CL. However, the current approach for optimizing ink formulation, particularly the ionomer−Pt− CB ratio (often termed as optimal I/C), based on carbonsupport type, is predominantly empirical. The nature of interaction between Pt−CB and ionomer, the resulting impact on Pt−CB particle agglomeration and the distribution of the ink components, and the effects of processing method and processing conditions on physical evolution of the bulk structure are less clear. Understanding of the structure− property−processing relationships of the catalyst inks is critical toward efficient formulation and processing of inks to design an optimal CL microstructure.2,3 The complexity of the catalyst ink system presents a challenge to understand the structure−property−processing relationships. The carbon particles, being fractal aggregates, are highly irregular in structure.8 The primary particles are spherical with size ∼10−100 nm, but are fused together in primary aggregates of a few hundred nm diameter. These primary aggregates can form agglomerates of larger size on the order of microns.9,10 The porosity of the carbon structure at multiple length scales adds additional complexity to the ink system. These include micropores (50 nm) that are associated with the fractal aggregates/agglomerated structure.8 The degree of porosity and the pore-size distribution can vary depending on the carbon-support type, of which there are many.8 Further, the Pt catalyst nanoparticles, which are distributed on the carbon structure, modify the local surface chemistry/charge on the carbon surface and can result in heterogeneous interparticle interactions.11,12 The presence of Pt and its distribution based on the carbon support used could strongly influence the interparticle/particle−ionomer interactions.13,14 Further complicating the ink system is the chemical structure of the ionomer, which consists of both a perfluorinated backbone and ionic side-chain segments. This presents competing hydrophobic and hydrophilic interactions with carbon and Pt in the ink.13−16 Additionally, the choice of solvent strongly influences the agglomerated structure of the ink.14,17−19 Thus, the overall microstructure of the catalyst ink and its processing behavior are governed by a complex interplay of Pt−CB−ionomersolvent interactions. There have been several investigations of catalyst ink microstructures through simulations13,14,19 and experiments utilizing different techniques.7,15,16,18,20,21 Contrast variationsmall-angle neutron scattering studies on inks containing CB (without Pt) and ionomer in aqueous media have confirmed a strong interaction of ionomer with carbon aggregates and a consequent reduction in the carbon agglomerated structure.20 Direct visualization of the structure of catalyst ink systems consisting of Pt−CB and ionomer dispersed in organic solvent media has also been conducted using cryogenic transmission electron microscopy (cryo-TEM),15,16,18,21 sometimes in combination with ultrasmall-angle X-ray scattering B
DOI: 10.1021/acsami.8b15039 ACS Appl. Mater. Interfaces XXXX, XXX, XXX−XXX
Research Article
ACS Applied Materials & Interfaces
dispersions was made using a pH meter (ThermoFisher Scientific, Orion 4-Star). X-ray Scattering Characterization. The carbon aggregated structure was characterized using X-ray scattering at beamline 9-ID-C at the Advanced Photon Source (APS) at Argonne National Laboratory. The ball-milled inks were collected into a glass capillary tube (1 mm diameter) and sealed with an epoxy resin. The capillary tubes were completely filled with the catalyst inks to eliminate any air bubbles. The sample tubes were mounted in the beamline hutch and exposed to a monochromatic X-ray beam in the range of 16.8−21 keV. The scattered X-ray intensity was measured using the combined USAXS−SAXS by a Bonse−Hart camera for USAXS and a Pilatus 100 K detector for pinhole SAXS. The complete intensity data was collected in two scattering vectors, q, ranging from 10−4 to 6 × 10−2 Å−1 for the USAXS and from 3 × 10−2 to 1 Å−1 for the pinhole SAXS. The scattering vector is reciprocal to length and is given by q = 4π/λ sin(θ/2), where λ is X-ray beam wavelength. The scattering q range spans approximately between 1 nm and 1 μm length scales, which could cover a wide range of carbon structure size scalesprimary particles, primary aggregates, and agglomerates. The background scattering data from the capillary tube filled with the solvent solution (1-propanol−water) was recorded and subtracted from scattering data for each catalyst ink. The scattering data were analyzed in a modeling macro package Irena (APS, X-ray science division, beamline 9-ID-C) for data manipulations and analysis on Igor Pro(WaveMetrics, OR) platform. Rheology Measurement Protocol. Rheological measurements were performed using a stress-controlled rheometer (Bohlin Gemini HR Nano, Malvern Instruments). Measurements were performed using a stainless steel parallel-plate geometry (40 mm diameter) with a gap of 250 μm at 25 °C hironomus; prior to making measurements, the samples were preconditioned to erase any sample loading history on the microstructure by conducting a preshear at 40 s−1 for 1 min and then allowed to rest for 1 min. The steady-shear rheology measurements were performed by imposing a decreasing stress sweep in logarithmic steps ranging from 40 to 0.0001 Pa. All the data reported were ensured to reach steady state.
and ionomer by examining the variation in the bulk microstructure as a function of I/C. To fully explore these interactions in the inks, we first independently investigate the effects of Pt presence on carbon and ionomer concentrations on the microstructure of carbon dispersions. To understand the effect of carbon-support types on the microstructure of inks, we present a comparative study between two carbon systems. These include a moderate surface area carbon (Vulcan XC-72) and a high-surface-area carbon (HSC), which are commonly used supports for PEMFC catalysts. In combination with rheology measurements, we use USAXS measurements and dynamic light scattering (DLS) techniques to examine the aggregated structure of the particles and verify the rheological observations. From the understandings of ionomer−Pt−CB interactions and the microstructure as a function of I/C for the two carbon-support types, we provide insights into the optimal I/C for performance and the influence of carbon-support type.
■
EXPERIMENTAL METHODS
Materials and Sample Preparation. The carbon and Pt−carbon powders used were Vulcan XC-72 (The Fuel Cell Store), 46.7 wt % Pt on Vulcan (TKK, TEC10V50E), high surface carbon (TKK, E-type carbon), and 45.8 wt % Pt on HSC (TKK, TEC10E50E). The particle dispersions for rheometry were prepared by mixing desired amounts of all the components, carbon or Pt−CB (0.25−3 wt % C) and Nafion (D2020, EW1000, Ion Power) (0−1.4 mass ratio with respect to carbon), in a water−1-propanol solvent mixture. For all inks, the solvent mixture was 75 wt % 1-propanol (Alfa Aesar) and 25 wt % water (Milli-Q). For rheometry and USAXS measurements, all the components of the ink were mixed together by ball milling with highdensity Zirconia beads (Glen Mills) of 5 mm in diameter. Ball milling was performed for 12 h at 80 rpm to ensure the ink dispersions were homogeneously mixed. In sample preparation for DLS and ζ-potential measurements, which require dilute concentrations (0.05 wt % carbon) to avoid multiple scattering events, the desired amounts of all the ink components were added together and mixed by probe sonication for 10 s, followed by bath sonication for 30 min. Note that the Z-average diameter of the particles characterized by DLS has been shown to depend on the sample concentration.24 The DLS size was found to increase with increasing particle concentration. Newer DLS techniques, for example, NanoPlus 3 and Micromeritics, require much dilute concentration for accurate measurements. Therefore, the DLS size results here are only qualitative in understanding the rheological observations due to the difference in the concentration regime. N2 Adsorption Experiments. The N2 adsorption isotherms were measured using a Micromeritics ASAP 2020. All samples were degassed to 200 °C in vacuum to eliminate any surface water prior to measuring. The Brunauer−Emmett−Teller (BET) surface area was determined in general in the range of 0.01−0.15 P/Po, (where Po is the measured saturation pressure of N2). The total pore volume was determined by the single-point desorption total pore volume of pores at P/Po of 0.98. The mesopore surface area and pore volumes were determined based on the desorption data and using the Barrett− Joyner−Halenda model, and the micropore volumes and surface areas were calculated by subtracting the mesopore values from the total values. Dynamic Light Scattering (DLS) and ζ-Potential Measurements. DLS and ζ-potential measurements were performed using a Zetasizer Nano ZS (Malvern Instruments Ltd, Malvern, U.K.). For the measurements, a high-concentration ζ cell (ZEN1010) was used, which allows measurements on more turbid samples such as carbon solutions, as well as concentrated solutions.42 All the measurements were performed at 25 °C. In the calculation of ζ-potential, a Helmholtz−Smoluchowski equation43 was used, which assumes that the electric double-layer thickness is much smaller than the particle size, which is true here. At least five measurements were taken to ensure repeatability of the results. The pH measurement of the ink
■
RESULTS AND DISCUSSION The following section details the findings on the rheology of the catalyst inks and elucidation of the underlying interactions between the catalyst ink components. Subsequently, the insights relevant to the optimal microstructure of CL for fuel cell performance are discussed. Owing to the complex nature of fuel cell catalyst inks, we have systematically studied the influence of each component in the systemcarbon black, Pt catalyst, and Nafion ionomeron the rheology and microstructure of the inks. This type of linear analysis allows for the elucidation of each component’s contribution to the inks’ overall behavior, while permitting examination of the interplay between all relevant factors. Particle Dispersions with No Ionomer. Carbon Black without Pt. The steady-shear rheology measurements of Vulcan and HSC dispersions without Pt or ionomer are presented in Figure 1a,b for particle concentrations ranging from 0.05 to 3 wt %. The low-shear viscosity and the degree of shear thinning increase with increasing particle concentration in both cases, as expected. At low shear rates where a Newtonian plateau can be observed (all concentrations of Vulcan, but only for lower concentrations of HSC), Brownian motion dominates the flow behavior. As the shear rates are increased, the strength of the hydrodynamic forces increases, and the viscosity becomes shear-rate dependent. As the carbon blacks are weakly charged, they have a strong tendency to agglomerate via Brownian motion.24,44 The shear-thinning behavior of the carbon dispersions is attributed to the C
DOI: 10.1021/acsami.8b15039 ACS Appl. Mater. Interfaces XXXX, XXX, XXX−XXX
Research Article
ACS Applied Materials & Interfaces
breakdown of the agglomerates and their rearrangements induced by the flow.38,45,46 HSC appears to exhibit a yield stress at higher concentrations (>1.5 wt %), where the viscosity continually increases with decreasing shear rate, without plateauing. The carbon agglomerates likely have grown from isolated agglomerates to a 3D percolated network at higher concentrations.38,46 Further, a weak shear-thickening behavior is observed at medium shear rates in both HSC and Vulcan at higher concentrations (beyond 0.75% in Vulcan and 1.5 wt % in HSC). Such shear thickening has been reported in flocculated carbon dispersions46,47 and is attributed to the breakup of agglomerates through hydrodynamic forces and a resulting increase in effective volume fraction of the particles.46 We will now examine the influence of carbon porous structure on the rheology by presenting a comparison between Vulcan and HSC dispersions. HSC has a larger surface area and internal pore volume (associated with micro-, meso-, and macropores) compared to Vulcan, by 4× and 2×, respectively, shown in Table 1, as characterized from N2 adsorption experiments. A comparison of their low-shear viscosity, η0, against particle volume fraction, ϕCB (estimated assuming primary particle density of ρCB = 1.8 g/cm3), in Figure 1c shows a larger viscosity of HSC dispersions compared to Vulcan. This is consistent with previous literature findings that compare carbons of different porosities.37 Owing to the porous structure of carbons, the viscosity was observed to be higher than hard-sphere dispersions and increased with increasing internal porosity of carbon.36,48−50 For hard-sphere dispersions, the viscosity scaling is described by volume fraction,51 whereas for porous particles such as carbon, the viscosity is found to be better described by a scaling against effective volume fraction (ϕeff), which accounts for the internal porosity of particles that can entrain a solvent and exclude it from hydrodynamic interactions.37 This is expressed as ϕeff = ϕCB/ ϕCB,Ragg, where ϕCB,Ragg is the volume fraction of carbon in its primary aggregate.37,48,50 To validate our observations, we rescaled the viscosities of Vulcan and HSC in terms of the effective volume fraction, also shown in Figure 1c. The ϕCB,Ragg values of 0.2 and 0.06 for Vulcan and HSC, respectively, were obtained from the literature.37 The η0 values of both Vulcan and HSC collapse onto a master curve against ϕeff, in agreement with the literature findings.37,48,50 We should note that the estimated ϕCB,Ragg of Vulcan and HSC might be lower than the real volume fraction of carbon, as the estimation assumes that carbons exist as well-dispersed primary aggregates. However, the carbons could agglomerate, which would further increase the effective particle volume fraction. Therefore, we attribute the differences in the rheological behavior between the HSC and Vulcan particles to the variation in the primary aggregate structure and the degree of agglomeration. We verify the rheological observations by characterizing the agglomerate sizes of HSC and Vulcan dispersions using USAXS and DLS, which are complementary to each other. The USAXS results, plotted as scattered intensity (I(q)) versus scattering vector (q) for HSC and Vulcan dispersions, are presented in Figure 2. The curves exhibit several kneelike power-law regimes. This indicates a multilevel structure of the carbon: primary particle, primary aggregate, and their agglomerates. The primary aggregate sizes of Vulcan and HSC (similar to Ketjen black) were reported to be 177 and
Figure 1. Steady-shear viscosities of (a) Vulcan and (b) HSC dispersions, and (c) low-shear viscosity at γ̇ = 2 s−1 comparison between Vulcan and HSC for different particle volume fractions. For the data sets where γ̇ = 2 s−1 was difficult to reach, the η at the lowest γ̇ achieved is presented. In (c), the solid lines correspond to η scaled against volume fraction, ϕCB, estimated based on primary carbon particle density, ρCB = 1.8 g/cm3. The dashed lines correspond to η0 scaling against effective volume fraction, ϕeff = ϕCB/ϕCB,Ragg, where ϕCB,Ragg values are 0.2 and 0.06 for Vulcan and HSC, respectively.37 D
DOI: 10.1021/acsami.8b15039 ACS Appl. Mater. Interfaces XXXX, XXX, XXX−XXX
Research Article
ACS Applied Materials & Interfaces
Table 1. Characterization of Carbon and Pt−Carbon Samples from N2 Adsorption Measurements, Including BET Total Surface Area As,total, Macro- and Mesopore Surface Area As>2 nm, Micropore Surface Area (As2 nm), and Micropore Volume (V2 nm (m2/g)
As2 nm (cm3/g)
V
