A1510 Journal of The Electrochemical Society, 160 (9) A1510-A1516 (2013)0013-4651/2013/160(9)/A1510/7/$31.00 © The Electrochemical Society

Carbon Nanofoam-Based Cathodes for Li–O2 Batteries:Correlation of Pore–Solid Architecture and ElectrochemicalPerformanceChristopher N. Chervin,z Michael J. Wattendorf, Jeffrey W. Long,∗,z Nathan W. Kucko,and Debra R. Rolison

Surface Chemistry Branch, U.S. Naval Research Laboratory, Washington, DC 20375, USA

Freestanding, binder-free carbon nanofoam papers afford the opportunity to gauge the influence of pore size on the dischargecapacity of Li–O2 cells. Four sets of carbon nanofoam papers were synthesized from resorcinol–formaldehyde sols, with pore sizedistributions in pyrolyzed forms ranging from mesopores (5–50 nm) to a size regime not represented in the literature for Li-O2cathodes—small macropores (50–200 nm). The first-cycle discharge capacity in cells containing 0.1 M LiClO4 in dipropyleneglycol dimethyl ether tracks the average pore size distribution in the carbon nanofoam cathode, rather than the specific surfacearea of the nanoscale carbon network or its total pore volume. The macroporous nanofoams yield cathode specific capacity of1000–1250 mA h g−1 at –0.1 mA cm−2 discharge rate, approximately twice that of the mesoporous nanofoams (∼580–670 mA h g−1),even though the macroporous foams have lower specific surface areas (270 and 375 vs. >400 m2 g−1). The specific capacity ofthe cathode decreases as the thickness of macroporous carbon nanofoam paper is increased from 180- to 530-μm, which indicatesthat the interior pore volume is underutilized, particularly with thicker nanofoams. For the four pore–solid nanofoam architecturesstudied, the specific capacity is limited by pore occlusion arising from solid Li2O2 product that is electrogenerated near the outerboundaries of the nanofoams.© 2013 The Electrochemical Society. [DOI: 10.1149/2.070309jes] All rights reserved.

Manuscript submitted April 11, 2013; revised manuscript received June 13, 2013. Published July 10, 2013. This was Paper 423presented at the Montreal, QC, Canada, Meeting of the Society, May 1–6, 2011.

Metal–air batteries, which function by balancing the oxidationof a metal anode (e.g., Zn, Li, Al, or Mg) with the reduction ofmolecular oxygen from air, combine the charge-storage capacity of abulk-metal negative electrode with a reactant-harvesting air cathode.By minimizing the mass/volume of one electrode, higher specific en-ergies, both theoretical and practical, are achievable as compared toother battery designs such as Li-ion batteries.1 Among the possiblemetal–air combinations, Li–air batteries have the highest theoreticalspecific energy (∼3,500 W h kg−1 assuming complete conversion ofthe anode and including the mass of oxygen),2 but only Zn-basedmetal–air batteries have demonstrated commercial success withconsumer-grade coin cells (∼300–440 W h kg−1; 0.050–6.5 A h) andmilitary-grade cells (∼300 W h kg−1; 60 A h).1,3 In practice, after con-sidering discharge inefficiency and including the weights of cathodematerials, electrolyte, and packaging, the actual specific energy of theLi–air battery at 800–900 W h kg−1 will be much lower than theoret-ical, but nonetheless could surpass state-of-the-art Li-ion and Zn–airtechnologies.2–5

Bringing nonaqueous rechargeable Li–air batteries closer to prac-tical fruition is inhibited by numerous factors including: (i) electrodeand electrolyte instability; (ii) charge–discharge inefficiency; and(iii) limited discharge capacity, which arises because solid dischargeproducts form within the porous air cathode and block ingress of themolecular reactant.2,4,6,7 These O2-choking solids may contain com-pounds beyond the anticipated reaction products of lithium peroxide(Li2O2) and lithium oxide (Li2O),7 including electrolyte and elec-trode decomposition species such as lithium carbonate (Li2CO3) andlithium alkyl carbonates.2,8–12 Once solids precipitate, diffusion of O2

into the electrode structure is hindered,13,14 in essence suffocatingthe air-breathing cathode and prematurely terminating the dischargeprocess.

The volume of accumulated discharge products and the result-ing discharge capacity of a Li–O2 cell depend on the size and dis-tribution of electrochemically wired voids in the air cathode (typi-cally a high-surface-area carbon), which comprise those surfaces thatare accessible to both O2 and to electrolyte. Researchers have corre-lated the discharge capacity of Li–O2 cells to the pore size distribu-tion of the carbon used to formulate conventional powder-compositecathodes, in which the carbon was predominantly microporous

∗Electrochemical Society Active Member.zE-mail: christopher.chervin@nrl.navy.mil; jeffrey.long@nrl.navy.mil

(<2 nm) or mesoporous (2–50 nm).15–21 Higher discharge capaci-ties were obtained for cathodes prepared using mesoporous carbons, aresult attributed to larger intraparticulate voids in which to accommo-date discharge products, but direct correlation of discharge capacityand pore size is complicated by the additional porosity that arises inpowder-composite electrodes from the ad-hoc packing of secondarycarbon particles/agglomerates and polymeric binder.22 Establishingcapacity–pore size correlations is further compromised because pore-size characterization is typically performed on the carbon powder, butnot on the final powder-composite structure in which the electrochem-ical processes occur.22 Some researchers anticipate that diminishingreturns with respect to capacity will be reached when pores approachtens of nanometers because the accumulation of electrically insulatingsolids (e.g., Li oxides and Li carbonates) will self-limit as electroactivecathode surfaces passivate.23–25

In this work, we leverage our ongoing efforts to optimize free-standing, three-dimensional (3D) porous carbon nanofoam papers26

as air-breathing cathodes for Zn–air batteries27–30 by exploring the re-lationship between the nature of the free volume of the nanofoamstructure—both the size of the pores therein and the size distri-bution of those pores—and the specific capacity of a Li–O2 cellconstructed with a nanofoam paper cathode. The well-characterizedmesopore-to-macropore size regime of the nanofoam-based air cath-odes (10s to 100s of nanometers) spans a range unrepresented in theliterature for Li–O2 cathodes, where the conductive component tog-gles between conventional microporous-to-mesoporous-carbons andnonconventional architectures such as nanotube/nanofiber assemblies(e.g., “buckypaper”) or freestanding metal foams, metal oxides, orcarbon, where voids are typically on the order of micrometers.13,31–34

Another advantage of nanofoam papers is that they require no binder,which is a necessary component in powder composites and onethat has been demonstrated to adversely affect capacity when poly-mer occludes small pores in the structure and within the carbonagglomerates.35

We synthesized a series of carbon nanofoam papers where the poresize was tuned to range from mesoporous (5–50 nm) to macroporous(50–200 nm), while cathode thickness was also varied for a macro-porous nanofoam (180, 360, or 530 μm). We test these cathodes inLi–O2 cells for single-discharge performance using an electrolyte of0.1 M LiClO4 in dipropylene glycol dimethyl ether (diproglyme),36

which is a low-vapor-pressure, nonflammable glyme that is more elec-trochemically stable than the organic carbonate-based electrolytes

) unless CC License in place (see abstract).  ecsdl.org/site/terms_use address. Redistribution subject to ECS terms of use (see 143.167.169.131Downloaded on 2014-09-19 to IP

Journal of The Electrochemical Society, 160 (9) A1510-A1516 (2013) A1511

that have been typically used for nonaqueous Li–O2 cells.10,37,38

For comparison, we also include prior results for cells containingcarbon nanofoam-based cathodes and discharged using carbonateelectrolytes.39 By combining microscopy, porosimetry, powder X-raydiffraction, and galvanostatic discharge, we demonstrate that incor-porating macropores on the order of 100–200 nm into O2-cathodes ofLi–O2 cells increases discharge capacity relative to nanofoam-basedelectrodes that contain only mesopores or small macropores (<100nm). The discharge capacity of even the macroporous nanofoams isultimately limited by pore occlusion as solids precipitate near theouter boundary of the carbon nanofoam electrode leaving the internalvoid volume of the cathode underutilized.

Experimental

Synthesis of carbon nanofoam-based O2 cathodes.— Carbonnanofoam papers were synthesized using our reported protocol,26 inwhich commercial carbon fiber papers (Lydall) are infiltrated witha resorcinol–formaldehyde (RF) sol followed by pressure cooking,drying, and pyrolysis. For this study, the molar ratio of resorcinol-to-formaldehyde was held at 1:2 and the combined weight fraction ofresorcinol and formaldehyde (wt% RF) and the resorcinol-to-catalystmolar ratio (R/C) in the sol were varied using either 40 or 50 wt% RFand either 500 or 1500 R/C to produce carbon nanofoams with differ-ent pore structures derived from the four formulations and designatedas: 40/1500; 40/500; 50/1500; and 50/500. The carbon fiber paperswere cut into 5 × 5 cm2 substrates and stacked to create two-plysheets that, after processing, produced carbon nanofoam papers with∼180-μm thickness (as measured by a digital micrometer). Addition-ally, four-ply and six-ply 40/1500 nanofoam papers were preparedto increase electrode thickness to 360 and 530 μm, respectively. Toachieve reproducible 40 wt% RF nanofoams, it was necessary to matchthe surface energies between the aqueous RF sol and the hydropho-bic carbon fibers by mildly oxidizing the fibers with glow-dischargeplasma etching under humidified conditions prior to infiltration.26

The RF sol was prepared by combining appropriate ratios of resor-cinol, 37 wt% formaldehyde, 18 M�cm water, and sodium carbon-ate (catalyst), stirring for 30 min, followed by standing unstirred for3 h at room temperature to increase viscosity of the infiltrating solvia oligomerization. Note: the pre-oligomerization step lowers theamount of unreacted formaldehyde, which has recently been reclas-sified by the Environmental Protection Agency as a carcinogen;40 allhandling of formaldehyde and formulation of the sol occurred in ahood while wearing personnel protection equipment.

The carbon fiber sheets were placed into glass containers andvacuum infiltrated with an excess of the oligomerized RF-sol. Theinfiltrated sheets were immediately sandwiched between two glassslides (previously calcined at 500◦C) with the addition of a few dropsof the RF-sol, sealed with binder clips, and wrapped tightly in ducttape. The duct-taped assembly was covered with an aluminum foilpouch, cured overnight in ambient conditions, and then placed in aconsumer-grade pressure cooker at 90◦C (steam setting) for 9.5 h fol-lowed by 80◦C (warm setting) for 6–10 h. After the samples werecooled to room temperature, the cured papers were unwrapped, rinsedin copious amounts of 18 M�cm water, and then soaked in acetone for3 h. The polymer foam papers were dried under ambient conditionsand pyrolyzed under argon at 1000◦C for 2 h with a 1◦C min−1 heatingand cooling ramp. Figure S1a shows an optical image of a commercialcarbon-fiber-paper scaffold before infiltration with the RF sol and a40/500 carbon nanofoam-filled paper after infiltration, gelation, andpyrolysis. The scanning electron micrograph image of the cross sec-tion of a 50/500 nanofoam paper demonstrates the high-quality fill ofnanofoam throughout the carbon paper (Figure S1b).39

Physical characterization of carbon nanofoam papers.— Thepore–solid architectures of carbon nanofoam papers prepared fromeach of the four RF–R/C formulations were characterized with scan-ning electron microscopy (SEM; Carl Zeiss Supra 55 electron micro-scope) and nitrogen physisorption (Micromeritics ASAP 2020 Surface

Area and Porosity Analyzer). The SEM samples were prepared by at-taching a small portion of nanofoam paper to an aluminum SEM stubusing conductive carbon tape and imaging was performed at 5 keV.For nitrogen porosimetry, samples were degassed at 120◦C under vac-uum for at least 24 h prior to analysis; the Brunauer–Emmett–Tellerspecific surface areas were determined from the linear portion of theadsorption isotherm and the pore size distributions and pore volumeswere calculated from the entire adsorption isotherm and fitted using adensity functional theory (DFT) model for a cylindrical geometry andHalsey curve thickness.

Fabrication and electrochemical characterization of Li–O2 cell.—Swagelok-type Li–O2 cells41 were prepared in an argon atmosphereglove box and contained a Li-foil anode (9-mm diameter; 99.9%,Aldrich), one disk of silica filter paper as separator (0.4-mm thick-ness, dried at 500◦C in air prior to transferring to glove box; Pallflextissuquartz, Pall Life Sciences), a carbon nanofoam O2 cathode disk(9-mm diameter; punched out of the nanofoam with a brass bore), alu-minum mesh as cathode current collector, and a stainless-steel washer.Each cell was fabricated by pressing the Li foil into a spring-loadedstainless-steel current collector disk, and then stacking atop the Li aseries of disks in the order of separator (pre-soaked in electrolyte), car-bon nanofoam (pre-soaked in electrolyte), and Al mesh. The stack wascompressed with a hollow, stainless-steel tube, which made contactto the aluminum current collector via a stainless-steel washer. Elec-trical contacts to the anode and cathode current-collecting tubes weremade with hose rings (see Figure S2 for a photograph and schematicof the Li–O2 cell). The electrolyte was 0.1M LiClO4 in dipropyleneglycol dimethyl ether (Proglyde DMM; Aldrich, used as received),1M LiClO4 in propylene carbonate (Aldrich, used as received), or 1MLiPF6 in 1:1:1 ethylene carbonate:diethyl carbonate:dimethyl carbon-ate (Novolyte; used as received).

After assembly, the O2 inlet valve on the Swagelok cell was closedand the cell was removed from the glove box and attached to a gasmanifold that was constantly purged with dry O2 (dried by flowingover a bed of Drierite desiccant). The valve was then opened and thecell flushed with O2 for 6 h before electrochemical measurementswere made with a Gamry Reference 600 potentiostat. The cells weredischarged at either –0.1 or –0.3 mA cm−2 (normalized to the areaof the cathode disk) until the cell voltage dropped to 2 V. Followingdischarge, selected cells were returned to the argon-atmosphere glovebox, disassembled, and the carbon nanofoam cathodes were soakedfor ∼12 h in several aliquots of diproglyme. The rinsed nanofoamcathodes were then dried under vacuum in the glove box antecham-ber over night before removal for further characterization with SEMand X-ray diffraction (XRD; Rigaku Smartlab). For SEM analysis,both the electrolyte- and O2-facing surfaces of the post-dischargednanofoams were analyzed. X-ray diffraction patterns were collectedfor post-discharged samples placed in a recessed sample holder suchthat the O2-facing surface of the cathode was approximately flushwith the diffraction plane. The diffraction patterns were acquired in0.02◦ intervals with 2-s dwell times using a Cu-Kα source operating at40 kV and 44 mA.

Results and Discussion

The carbon nanofoam paper is a device-ready, macroscopic archi-tecture built within a scaffolding of conductive carbon-fiber paper.26

The carbon nanofoam that fills the paper comprises two interpene-trating networks: a solid network of nanoscopic pyrolytic carbon anda pore network of through-connected nanoscale voids. Innate to thenanofoam architecture are physical attributes highly relevant to the de-sign of advanced air cathodes: (i) the 3D-interconnected pore networkthat spans the macroscale thickness of the nanofoam paper promotesO2 flux; (ii) the large free volume provides storage for solid dischargeproducts; (iii) the high surface area of the nanoscale carbon networkextends electrocatalytic interfaces throughout the electrode interior;(iv) the fiber paper/nanofoam structure has an electronic conductiv-ity typically >20 × that of the pyrolytic carbon nanofoam; (v) the

) unless CC License in place (see abstract).  ecsdl.org/site/terms_use address. Redistribution subject to ECS terms of use (see 143.167.169.131Downloaded on 2014-09-19 to IP

A1512 Journal of The Electrochemical Society, 160 (9) A1510-A1516 (2013)

Table I. Brunauer–Emmett–Teller (BET) surface areas, Barrett–Joyner–Halenda (BJH) pore volumes, and discharge capacities as a function ofwt% RF and R/C ratio.

Sample

Pore sizedistribution

(nm)

Specificsurface area(m2 g−1)a

Pore volume(cm3 g−1)b

Dischargecapacity

(mA h g−1electrode)c

Dischargecapacity

(mA h g−1nanofoam)d

Voids occupied byLi2O2 after

discharge (%)e

50 wt% RF (500 R/C) 2–60 420 0.94 670 1340 2650 wt% RF (1500 R/C) 2–80 440 0.91 575 1150 2340 wt% RF (500 R/C) 10–120 375 1.3 1250 2500 3640 wt% RF (1500 R/C) 20–200 270 2.2f 1025 2050 17

aBrunauer–Emmett–Teller (BET) specific surface area; replicate porosimetry analyses vary by ±1–2%; batch-to-batch variation is on the order of ±10%.bDetermined for pores between 1 and 300 nm from Barrett–Joyner–Halenda (BJH) using the adsorption isotherm.cDischarged in dry oxygen at –0.1 mA cm−2; capacities are normalized to the entire mass of the nanofoam paper cathode, with the carbon-fiber scaffoldcontributing 40–50% of the total mass of the nanofoam paper.dDischarged in dry oxygen at –0.1 mA cm−2; capacities are normalized to the mass of carbon nanofoam (pore-supplying active component), whichcontributes ∼50% of the total mass of the nanofoam paper.eVoid volume occupied by Li2O2 calculated from the measured pore volume of the cathode and assuming a two-electron discharge process with Li2O2(F.W. = 45.88 g mol−1, ρ = 2.31 g cm−3) as the only product.fValue derived from mercury porosimetry.26

binder-free electrode architecture eliminates complications of binderdecomposition during cell operation; (vi) the interbonded structure(the bonded carbon network of the foam and the bonded adhesionof the foam to the fibers of the paper) imparts mechanical stabil-ity to sustain internal pressures that arise as solids precipitate withinthe porous structure; and (vii) the nanofoam papers are scalable in xand y by adjusting the size of the fiber-paper scaffolds and in z bystacking multiple layers of the fiber papers before infiltration of theresorcinol–formaldehyde (RF) sol.26

Physical characterization of carbon nanofoams.— The nature ofporosity in carbon nanofoam paper is readily varied by adjusting theweight fraction of solids (wt% RF) and the resorcinol-to-catalyst ratio(R/C) in the aqueous precursor-sol.26,42,43 In this study we examinedfour RF–R/C formulations, which as shown by scanning electron mi-croscopy (SEM) produced pore sizes ranging from 5 to 200 nm (meso-pores to small macropores); see Figure 1. Changing the RF weightpercentage demarcates the pore structure between large mesopores(5–50 nm) at 50 wt% RF to small macropores (50–200 nm) at 40 wt%

RF. Further tuning of the pore structure is accomplished by varyingthe amount of catalyst (R/C molar ratio), which influences both thesize of the pores and the pore size distribution.

Nitrogen-sorption porosimetry provides more quantitative infor-mation regarding the pore structure of these various nanofoams. Themesoporous 50 wt% RF nanofoams exhibit higher specific surface ar-eas (440 and 420 m2 g−1) than their 40 wt% RF counterparts (270 and375 m2 g−1, Table I). Total pore volume of the 50 wt% RF nanofoamsis comparable (∼0.9 cm3 g−1), increases by ∼50% for the 40/500nanofoams (1.3 cm3 g−1), and doubles to 2.2 cm3 g−1 for the 40/1500nanofoam. The increase in pore volume for 40 wt% nanofoams isconsistent with the larger pores characteristic of these formulations(Figure 1). For the mesoporous 50 wt% RF nanofoams, reducing R/Cfrom 1500 to 500 decreases the average pore size from ∼40 nm to30 nm and narrows the pore size distribution from ∼40 nm to 20 nm(approximated as the full width at half maximum from the pore sizedistribution plot; Figure 2). Because macropores sized at >100 nmoccupy a significant fraction of the void volume in the 40 wt% RFnanofoams, the usefulness of N2-sorption to describe the total pore

Figure 1. Scanning electron micrographs of carbon nanofoam papers synthesized with weight fraction of solids (40 or 50 wt%) and resorcinol-to-catalyst ratio(500 or 1500) of (a) 50/500 (b) 50/1500 (c) 40/500 and (d) 40/1500.

) unless CC License in place (see abstract).  ecsdl.org/site/terms_use address. Redistribution subject to ECS terms of use (see 143.167.169.131Downloaded on 2014-09-19 to IP

Journal of The Electrochemical Society, 160 (9) A1510-A1516 (2013) A1513

Figure 2. Pore size distribution plots derived from N2-sorption porosimetryas a function of RF sol formulation: (— —) 40/500; (—�—) 40/1500; (—◦—)50/500; and (—∇—) 50/1500.

size distribution becomes limited, particularly for the 40/1500 for-mulation. In a previous study using Hg-sorption porosimetry, whichquantifies a greater range of pore sizes in the macropore regime, wedetermined that 40/1500 nanofoams have an average pore size of∼100 nm with a size distribution from 20 to 200 nm, in agreementwith the small macropores visible by SEM (see Figure 1d).26

Electrochemical characterization of carbon nanofoam cathodes inLi–O2 cells.— Within this set of four carbon nanofoam papers, threehave comparable specific surface area (50/500, 50/1500, and 40/500)at ∼400 m2 g−1, such that if specific surface area is the primary factorthat determines discharge capacity, the performance of Li–O2 cellsmade with these nanofoams should be comparable. If higher free vol-ume for discharge product accumulation is more critical, then onewould observe higher capacity with the 40 wt% nanofoams, partic-ularly for the 40/1500 formulation. To investigate these correlationsas well as the relevance of varying the pore size distribution, we as-sessed the single-discharge behavior of the four selected nanofoamformulations, tested as cathodes in Li–O2 cells. Discharge profilesfor galvanostatic tests at –0.1 mA cm−2 were relatively flat, exhibit-ing the ∼2.6-V plateau typical for Li–O2 cells with carbon-basedcathodes (Figure 3). The total capacity observed before reaching thenegative voltage limit depends on the particular pore–solid architec-ture. The 40 wt% RF nanofoams have approximately twice the spe-cific capacity of the 50 wt% RF nanofoams (1025–1250 mA h g−1 vs.575–670 mA h g−1; normalized to total cathode mass), demonstratingthat within the pore-size range of 5–200 nm, higher discharge capac-ity is obtained for nanofoams with a significant macropore contentcompared to related nanofoams that are predominantly mesoporous.

Specific discharge capacity of a Li–O2 cell is commonly normal-ized to the mass of active carbon in the cathode without including po-tentially significant mass contributions from binder, current collector,carbon support, or catalyst components that are required to fabricatea practical electrode.15,16,19,25,44,45 In the present case, we normalizespecific capacity to the total mass of the freestanding nanofoam papercathode, which includes the carbon-fiber paper scaffolding that doesnot contribute substantial void volume to accommodate dischargeproducts. If we normalize only to the mass of the pore-supplying ac-tive nanofoam, which represents 40–50% of total mass in the carbonnanofoam papers, our specific capacities increase by at least a fac-tor of two (e.g., the high-performing 40/500 cathode increases from1250 mA h g−1 to 2500 mA h g−1; see Table I). The 40 wt% nanofoamsprovide comparatively high specific discharge capacities relative toother carbon-based cathodes discharged at the same current density(–0.1 mA cm−2), where typical capacities as normalized only to theactive carbon range from 102–103 mA h g−1.13,15,46–48

Figure 3. Discharge of Swagelok-type Li–O2 cells at –0.1 mA cm−2 usingcarbon nanofoam papers prepared from different RF sol formulation: (———)40/500; (------) 40/1500; (— — —) 50/500; and (–••–••–) 50/1500. The cellswere discharged to 2 V in 0.1 M LiClO4 electrolyte in dipropylene glycoldimethyl ether. The discharge capacities are normalized to the entire mass ofthe nanofoam paper cathode.

We found a similar relationship between the pore size distributionof the nanofoam-based cathode and the discharge capacity of Li–O2

cells when 1:1:1 ethylene carbonate:diethyl carbonate:dimethyl car-bonate was used as the electrolyte solvent (see Figures S3 and S4).Organic carbonates have been the most commonly used electrolytesolvents for Li–O2 battery research, but the present consensus is thatthe discharge process is dominated by formation of electrochemicallyirreversible Li2CO3 and other carbonate decomposition solids ratherthan the desired reaction product, Li2O2.2,8,10 Various glymes are nowbeing explored as more stable alternatives to carbonate solvents, in-cluding higher molecular weight varieties, such as diproglyme, thatimpart low volatility to the electrolyte.36,49–51 Here, with diproglymeas the electrolyte solvent and using nanofoam-based air cathodes,Li2O2 is the predominant product on first discharge, with no evi-dence of significant byproducts as determined by XRD (see Figure 4).Whereas when the nanofoam cathodes are discharged in a carbonate-based electrolyte, Li2CO3 is the dominant discharge product in thediffraction pattern.

For the series of nanofoams used in this study, the size of porescorrelates to discharge capacity but the total void volume does not(see Table I and Figure 3). Specific capacity doubles for the 40/500nanofoam relative to either 50 wt% RF nanofoam, although its voidvolume increases only by ∼40%. The 40/1500 nanofoam has thelargest void volume, yet exhibits a modest capacity decrease (18%)relative to the 40/500 nanofoam. This observed disconnect betweena seemingly important parameter, void volume, and the dischargecapacity can be attributed to incomplete utilization of the pores andsurfaces in the interior of the cathode. To attain maximum dischargecapacity for a given cathode architecture, oxygen reduction must besustained until the solid discharge products fill the total void volumeof the porous electrode, ideally propagating uniformly out from theelectrified interface where O2 is reduced. In reality, solid productsthat form at the outer boundaries of the electrode hinder O2 transportpathways to the inner pores, thereby limiting utilization of the entireelectrode void volume,13,52 which explains the observed absence of amonotonic trend between pore volume and discharge capacity.

We estimate the percentage of void volume in the carbon nanofoampaper that is occupied by solid products at the end of dischargeby normalizing to the pore volume derived from either N2- orHg-porosimetry and assuming a two-electron discharge process withLi2O2 as the only product (see Table I). In the four nanofoamformulations discharged at –0.1 mA cm−2 in diproglyme, 17–36%of the void volume is occupied with Li2O2. The relatively large

) unless CC License in place (see abstract).  ecsdl.org/site/terms_use address. Redistribution subject to ECS terms of use (see 143.167.169.131Downloaded on 2014-09-19 to IP

A1514 Journal of The Electrochemical Society, 160 (9) A1510-A1516 (2013)

Figure 4. X-ray diffraction profiles of post-discharge air cathodes dischargedat –0.1 mA cm−2 in an electrolyte of either (a) 0.1 M LiClO4 in dipropyleneglycol dimethyl ether or (b) 1 M LiClO4 in propylene carbonate. The O2cathodes were two-ply 40/500 carbon nanofoam papers.

percentage of voids filled in the carbon nanofoam-based cathodes atthis moderate discharge rate indicates that Li2O2 accumulates withinthe interior of the nanofoams, and not just at the outer boundary. Incontrast, Albertus et al. found that for a Super P carbon-based cathode(slurry-coated onto a nickel foam current collector), only 2.4% of thecarbon void volume is occupied by solid products after discharge at–0.08 mA cm−2 in a propylene carbonate-based electrolyte; in thatcase, the authors attributed cell polarization to rapid passivation ofelectrified surfaces by discharge product.25

To determine if cathode thickness affects the volume of internalpores that are utilized in the carbon nanofoam paper, we compare thedischarge capacities of Li–O2 cells containing nanofoam cathodes ofvarying thickness but derived from the same RF(R/C) formulation.Uniform filling of the inner void volume as the electrode thicknessincreases should produce similar specific (i.e., mass-normalized) dis-charge capacity. Using 40/1500 carbon nanofoams (with nominallyidentical pore size distribution, void volume, and specific surfacearea), we prepared two-, four-, and six-ply nanofoams that are ∼180-,360-, and 530-μm thick. The respective discharge capacity at a forcingrate of –0.3 mA cm−2 for cells prepared with this series of nanofoamsis 890, 590, and 360 mA h g−1 (see Figure 5). Thus, each additional180-μm layer of carbon nanofoam paper results in ∼35–39% loss inspecific capacity.

The decrease in specific capacity with increasing electrode thick-ness was also reported for powder-composite electrodes13,16 and Xiaoet al. used area-specific capacity (mA h cm−2)16 as an alternative met-ric for comparing discharge capacity of cathodes prepared with vary-

Figure 5. Discharge of Swagelok-type Li–O2 cells at –0.3 mA cm−2 as afunction of carbon nanofoam thickness: (——) two-ply 180 μm; (– – –) four-ply 360 μm; and (— — —) six-ply 530 μm. The air cathodes were 40/1500carbon nanofoams and the cells were discharged to 2 V in 0.1 M LiClO4electrolyte in dipropylene glycol dimethyl ether. The discharge capacities arenormalized to the entire mass of the nanofoam paper cathode.

ing thickness, carbon loading, or by different methods. The corre-sponding area-normalized capacities for carbon nanofoams are 9.3,10.5, and 9.9 mA h cm−2 for the 180-, 360-, and 530-μm thicknanofoams, respectively. The invariance in area-normalized capaci-tance, despite the additional total pore volume available with increas-ing electrode thickness, indicates that much of the extra pore volumein thicker electrodes is inactive for discharge-product accumulation.Thus, as noted previously with powder-composite electrodes, there isa practical limit to cathode thickness for flooded Li–O2 cell operationthat will depend on such factors as discharge rate, O2 solubility in thechosen electrolyte, and pore–solid architecture.

Specific capacity also decreases with increasing electrode thick-ness for a cell using a 40/500 nanofoam, propylene carbonate elec-trolyte, and a lower discharge rate (–0.1 mA cm−2; see Figure S5).The loss in specific capacity is not as severe with increasing electrodethickness (25% capacity loss from two-ply to six-ply vs. the 60%capacity loss from above), which we attribute to greater participationby internal pores at the lower discharge rate.52 Although differentproducts are formed out of diproglyme- and propylene carbonate-based electrolytes, cell discharge terminates with incomplete utiliza-tion of the total cathode void volume in either electrolyte and by thesame mechanism, namely isolating the reactive surfaces of inner poresby precipitation of electrogenerated discharge products at the outerboundary of the electrode, where the concentration of oxygen is thehighest.

Cathode surface area is another possible factor affecting the dis-charge capacity of Li–O2 cells. In comparing the BET surface areas(see Table I) with the discharge profiles for the four electrode formula-tions, capacity increases with decreasing surface area, which is at firstglance, an unlikely correlation because the discharge process involvesthe electrochemical precipitation of solid products from species thatform at the electrically conductive surfaces.53 These results suggestthat when comparing mesoporous to macroporous carbon nanofoams,increasing pore size tracks increasing discharge capacity, whereas sur-face area is a less reliable indicator. The irrelevance of surface areaover pore size to predict discharge capacity has also been reported forpowder-composite cathodes using either mesoporous or microporouscarbons.15,16,18 In this study, we capitalize on the well-characterizedlarge pore sizes that are readily formed with the carbon nanofoampapers—an architectural regime that is unavailable in conventionalcarbon powder composite materials—to demonstrate that this trendextends into the small macropore range.

Scanning electron micrographs of post-discharge 40/500 and50/500 carbon nanofoams reveal that discharge products are formed

) unless CC License in place (see abstract).  ecsdl.org/site/terms_use address. Redistribution subject to ECS terms of use (see 143.167.169.131Downloaded on 2014-09-19 to IP

Journal of The Electrochemical Society, 160 (9) A1510-A1516 (2013) A1515

Figure 6. Scanning electron micrographs of post-discharge carbon nanofoam O2 cathodes as a function of RF formulation; (a) O2 side and (b) electrolyte sidefor a 40/500 nanofoam; (c) O2 side and (d) electrolyte side for a 50/500 nanofoam. The Li–O2 cells were discharged in dry O2 at –0.1 mA cm−2 to 2 V in 0.1 MLiClO4 electrolyte in dipropylene glycol dimethyl ether.

on both the O2- and the electrolyte-facing surfaces of the nanofoam pa-per to such an extent that the surface porosity is annihilated (compareFigure 6 to the corresponding as-prepared nanofoams of Figure 1aand 1c). These micrographs show only a small region of the total elec-trode surface but they are representative of the O2-facing boundary(which exhibits total pore blockage across the face of the electrode),whereas the electrolyte-facing side of the cathode retains some regionsof unblocked pores (see Figure S6). The elongated Li2O2 particlesshown here are distinctly different from the spherical Li2CO3 electro-generated out of carbonate-based solvents (Figure S7) with the particlesize for both products ranging from 10s to 100s of nanometers. Fill-ing exterior pores with discharge product, which preferentially formswhere O2 concentration is the highest, impedes further O2 transportinto the interior of the nanofoam cathode and limits the dischargecapacity that would otherwise be achieved based on the well-plumbednetwork of void volume characteristic of these nanofoams. This con-clusion, in agreement with prior results by Read et al.52 and Zhanget al.,13 is substantiated by the microscopy data, which show >90%of the pores at the outer boundary are covered by solid products, evenfor pores of 100–200 nm.

Conclusions

Using carbon nanofoam architectures as O2-breathing cathodes,we have demonstrated a correlation between pore–solid architectureand the specific capacity of cathodes in Li–O2 cells at relatively highcurrent density. In the limit of inhomogeneous product distribution—aproblem associated with most if not all Li–O2 cathode designs–macroporous nanofoams yield higher discharge capacity than compa-rable mesoporous nanofoams, which demonstrates the important rolethat macropores can play in establishing high specific capacity in Li–O2 cells. A design strategy that incorporates macropores—particularlyat the O2-facing boundary of the electrode—may be beneficial tocathodes for Li–O2 cells, as they are less susceptible to blockage bydischarge products than are mesopores. If an air-cathode architecturewith as well-wired an electron path and as well-plumbed a mass-transport path (for electrolyte and O2) as the carbon nanofoam paperscannot sustain uniform reaction fronts within the porous electrode—such that >100-nm pores are fully filled on the first discharge—even

well-catalyzed cathode walls bathed in a radical-resistant electrolytewill be in trouble. Gradients of pore size with larger pores at the O2

side of the electrode tapering to smaller pores near the electrolyte(O2-diffusion-limited region) may ultimately be desirable to achievean optimal combination of surface-to-volume ratio and more uniformdischarge product distribution, a hypothesis supported by conclusionsfrom a recent modeling study of mass transport and lithium peroxideformation in Li–O2 cathodes.54

Acknowledgments

This work was supported by the U.S. Office of Naval Research.M.J.W. (Princeton University) and N.W.K. (Alfred University) wereSTEP undergraduate science aides at the NRL.

References

1. D. Linden, Handbook of Batteries—Second Edition, p. 38.1–38.45, McGraw Hill,Texas (1995).

2. P. G. Bruce, S. A. Freunberger, L. J. Hardwick, and J.-M. Tarascon, Nature Mater.,11, 19 (2012).

3. Electric Fuel Battery website: http://www.efbpower.com/BA-8180.html (accessedApril 10, 2013); Duracell Technical Bulletin website: http://www.duracell.com/en-US/Global-Technical-Content-Library/Technical-Bulletins.jspx?icn=Prim/PrimNav/Technical-Bulletins&cc=Primary (accessed April 10, 2013).

4. G. Girishkumar, B. McCloskey, A. C. Luntz, and W. Wilcke, J. Phys. Chem. Lett., 1,2193 (2010).

5. J. Christensen, P. Albertus, R. S. Sanchez-Carrera, T. Lohmann, B. Kozinsky,R. Liedtke, J. Ahmed, and A. Kojic, J. Electrochem. Soc., 159, R1 (2012).

6. L. Trahey, C. S. Johnson, J. T. Vaughey, S. H. Kang, L. J. Hardwick,S. A. Freunberger, P. G. Bruce, and M. M. Thackeray, Electrochem. Solid-State Lett.,14, A64 (2011).

7. T. Ogasawara, A. Debart, M. Holzapfel, P. Novak, and P. G. Bruce, J. Am. Chem.Soc., 128, 1390 (2006).

8. G. M. Veith, N. J. Dudney, J. Howe, and J. Nanda, J. Phys. Chem. C, 115, 14325(2011).

9. Y. C. Lu, H. A. Gasteiger, E. Crumlin, R. McGuire, and Y. Shao-Horn, J. Electrochem.Soc., 157, A1016 (2010).

10. W. Xu, V. V. Viswanathan, D. Wang, S. A. Towne, J. Xiao, Z. Nie, D. Hu, andJ. G. Zhang, J. Power Sources, 196, 3894 (2011).

11. W. Xu, K. Xu, V. V. Viswanathan, S. A. Towne, J. S. Hardy, J. Xiao, Z. Nie, D. Hu,D. Wang, and J.-G. Zhang, J. Power Sources, 196, 9631 (2011).

) unless CC License in place (see abstract).  ecsdl.org/site/terms_use address. Redistribution subject to ECS terms of use (see 143.167.169.131Downloaded on 2014-09-19 to IP

A1516 Journal of The Electrochemical Society, 160 (9) A1510-A1516 (2013)

12. B. D. McCloskey, A. Speidel, R. Scheffler, D. C. Miller, V. Viswanathan,J. S. Hummelshoj, J. K. Norskov, and A. C. Luntz, J. Phys. Chem. Lett., 3, 997(2012).

13. G. Q. Zhang, J. P. Zheng, R. Liang, C. Zhang, B. Wang, M. Hendrickson, andE. J. Plichta, J. Electrochem. Soc., 157, A953 (2010).

14. J. Nanda, H. Bilheux, S. Voisin, G. M. Veith, R. Archibald, L. Walker, S. Allu,N. J. Dudney, and S. Pannala, J. Phys. Chem. C, 116, 8401 (2012).

15. X. H. Yang, P. He, and Y. Y. Xia, Electrochem. Commun., 11, 1127 (2009).16. J. Xiao, D. Wang, W. Xu, D. Wang, R. Williford, J. Liu, and J.-G. Zhang, J. Elec-

trochem. Soc., 157, A487 (2010).17. P. Kichambare, J. Kumar, S. Rodrigues, and B. Kumar, J. Power Sources, 196, 3310

(2011).18. C. Tran, X.-Q. Yang, and D. Qu, J. Power Sources, 195, 2057 (2010).19. M. Mirzaeian and P. J. Hall, Electrochim. Acta, 54, 7444 (2009).20. X. Ren, S. S. Zhang, D. T. Tran, and J. Read, J. Mater. Chem., 21, 10118 (2011).21. G. O. Shitta-Bey, M. Mirzaeian, and P. J. Hall, J. Electrochem. Soc., 159, A315

(2012).22. J. Hou, M. Yang, M. W. Ellis, R. B. Moore, and B. Yi, Phys. Chem. Chem. Phys., 14,

13487 (2012).23. J.-G. Zhang, D. Wang, W. Xu, J. Xiao, and R. E. Williford, J. Power Sources, 195,

4332 (2010).24. A. Kraytsberg and Y. Ein-Eli, Nano Energy, in press, doi: 10.1016/j.nanoen.

2012.11.016.25. P. Albertus, G. Girshkumar, B. McCloskey, R. S. Sanchez-Carrera, B. Kozinsky,

J. Christensen, and A. C. Luntz, J. Electrochem. Soc., 158, A343 (2011).26. J. C. Lytle, J. M. Wallace, M. B. Sassin, A. J. Barrow, J. W. Long, J. L. Dysart,

C. H. Renninger, M. P. Saunders, N. L. Brandell, and D. R. Rolison, Energy Environ.Sci., 4, 1913 (2011).

27. C. N. Chervin, J. W. Long, N. L. Brandell, J. M. Wallace, N. W. Kucko, andD. R. Rolison, J. Power Sources, 207, 191 (2012).

28. J. W. Long, C. N. Chervin, J. M. Wallace, N. L. Brandell, J. L. Dysart, andD. R. Rolison, SPIE: Micro- and Nanotechnology Sensors, Systems, and Applica-tions II, 7679, 767915 (2010).

29. D. R. Rolison, J. W. Long, and C. N. Chervin, U.S. Patent Application, 20120074908.30. J. W. Long, C. N. Chervin, N. W. Kucko, E. S. Nelson, and D. R. Rolison, Adv.

Energy Mater., 3, 584 (2013).31. Y. Cui, Z. Wen, and Y. Liu, Energy Environ. Sci., 4, 4727 (2011).32. Y. Cui, Z. Wen, X. Liang, Y. Lu, J. Jin, M. Wu, and X. Wu, Energy Environ. Sci., 5,

7893 (2012).

33. J. Wu, H. W. Park, A. Yu, D. Higgins, and Z. Chen, J. Phys. Chem. C, 116, 9427(2012).

34. R. R. Mitchell, B. M. Gallant, C. V. Thompson, and Y. Shao-Horn, Energy Environ.Sci., 4, 2952 (2011).

35. S. R. Younesi, S. Urbonaite, F. Bjorefors, and K. Edstrom, J. Power Sources, 196,9835 (2011).

36. C. N. Chervin, J. W. Long, M. J. Wattendorf, N. W. Kucko, and D. R. Rolison, Proc.45th Power Sources Conf., 567 (2012).

37. S. A. Freunberger, Y. Chen, Z. Peng, J. M. Griffin, L. J. Hardwick, F. Barde, P. Novak,and P. G. Bruce, J. Am. Chem. Soc., 133, 8040 (2011).

38. V. Bryantsev and M. Blanco, J. Phys. Chem. Lett., 2, 379 (2011).39. C. N. Chervin, J. W. Long, N. L. Brandell, and D. R. Rolison, ECS Trans., 35, 33

(2011).40. Toxicological Review of Formaldehyde Inhalation Assessment: In Support

of Summary Information on the Integrated Risk Information System [IRIS]<http://cfpub.epa.gov/ncea/iris_drafts/recordisplay.cfm?deid=223614>.

41. S. D. Beattie, D. M. Manolescu, and S. L. Blair, J. Electrochem. Soc., 156, A44(2009).

42. R. W. Pekala and D. W. Schaefer, Macromolecules, 26, 5487 (1993).43. V. Bock, A. Emmerling, and J. Fricke, J. Non-Cryst. Solids, 225,

69 (1998).44. H.-G. Jung, J. Hassoun, J.-B. Park, Y.-K. Sun, and B. Scrosati, Nature Chem., 4, 579

(2012).45. A. Debart, J. Bao, G. Armstrong, and P. G. Bruce, J. Power Sources, 174, 1177

(2007).46. O. Crowther, D. Keeny, D. M. Moureau, B. Meyer, M. Salomon, and M. Hendrickson,

J. Power Sources, 202, 347 (2012).47. X.-H. Yang and Y.-Y. Xia, J. Solid State Electrochem., 14, 109 (2010).48. J. Read, J. Electrochem. Soc., 149, A1190 (2002).49. S. A. Freunberger, Y. Chen, N. E. Drewett, L. J. Hardwick, F. Barde, and P. G. Bruce,

Angew. Chem. Int. Ed., 50, 8609 (2011).50. C. O. Laoire, S. Mukerjee, E. J. Plichta, M. A. Henrickson, and K. M. Abraham,

J. Electrochem. Soc., 158, A302 (2011).51. J. Read, J. Electrochem. Soc., 153, A96 (2005).52. J. Read, K. Mutolo, M. Ervin, W. Behl, J. Wolfenstine, A. Driedger, and D. Foster,

J. Electrochem. Soc., 150, A1351 (2003).53. J. S. Lee, S. T. Kim, R. Cao, N. S. Choi, M. Liu, K. T. Lee, and J. Cho, Adv. Energy

Mater., 1, 34 (2011).54. X. Li and A. Faghri, J. Electrochem. Soc., 159, A1747 (2012).

) unless CC License in place (see abstract).  ecsdl.org/site/terms_use address. Redistribution subject to ECS terms of use (see 143.167.169.131Downloaded on 2014-09-19 to IP