lable at ScienceDirect

Journal of Power Sources 321 (2016) 135e142

Contents lists avai

Journal of Power Sources

journal homepage: www.elsevier .com/locate/ jpowsour

Operando identification of the point of [Mn2]O4 spinel formationduring g-MnO2 discharge within batteries

Joshua W. Gallaway a, *, Benjamin J. Hertzberg b, Zhong Zhong c, Mark Croft d,Damon E. Turney a, Gautam G. Yadav a, Daniel A. Steingart b, Can K. Erdonmez e,Sanjoy Banerjee a

a The CUNY Energy Institute, Department of Chemical Engineering, The City College of New York, 160 Convent Ave, New York, NY 10031, USAb Department of Mechanical and Aerospace Engineering, Andlinger Center for Energy and the Environment, Princeton University, Princeton, NJ 08544, USAc Photon Sciences Directorate, Brookhaven National Laboratory, Upton, NY 11973, USAd Department of Physics and Astronomy, Rutgers University, 136 Frelinghuysen Road, Piscataway, NJ 08854, USAe Energy Storage Group, Brookhaven National Laboratory, Upton, NY 11973, USA

h i g h l i g h t s

* Corresponding author.E-mail address: jgallaway@che.ccny.cuny.edu (J.W

http://dx.doi.org/10.1016/j.jpowsour.2016.05.0020378-7753/© 2016 Elsevier B.V. All rights reserved.

g r a p h i c a l a b s t r a c t

� Discharge of g-MnO2 was trackedwithin sealed batteries as a functionof position.

� Spinel formation immediately fol-lowed phase transition to a-MnOOH.

� Well-formed a-MnOOH occurred af-ter insertion of 0.79 Hþ per Mn atom.

� Insertion of 0.79 Hþ correlated to104% of the 2 � 1 ramsdellite tunnelcapacity.

a r t i c l e i n f o

Article history:Received 15 April 2016Accepted 1 May 2016Available online 7 May 2016

Keywords:Alkaline batteryManganese dioxideZincSpinelProton insertionOperando diffraction

a b s t r a c t

The rechargeability of g-MnO2 cathodes in alkaline batteries is limited by the formation of the [Mn2]O4

spinels ZnMn2O4 (hetaerolite) and Mn3O4 (hausmannite). However, the time and formation mechanismsof these spinels are not well understood. Here we directly observe g-MnO2 discharge at a range of re-action extents distributed across a thick porous electrode. Coupled with a battery model, this reveals thatspinel formation occurs at a precise and predictable point in the reaction, regardless of reaction rate.Observation is accomplished by energy dispersive X-ray diffraction (EDXRD) using photons of high en-ergy and high flux, which penetrate the cell and provide diffraction data as a function of location andtime. After insertion of 0.79 protons per g-MnO2 the a-MnOOH phase forms rapidly. a-MnOOH is theprecursor to spinel, which closely follows. ZnMn2O4 and Mn3O4 form at the same discharge depth, by thesame mechanism. The results show the final discharge product, Mn3O4 or Mn(OH)2, is not an intrinsicproperty of g-MnO2. While several studies have identified Mn(OH)2 as the final g-MnO2 dischargeproduct, we observe direct conversion to Mn3O4 with no Mn(OH)2.

© 2016 Elsevier B.V. All rights reserved.

. Gallaway).

1. Introduction

The oxide g-MnO2, which is a staple of global electrochemicalenergy storage, shows a high complexity in its structure anddischarge mechanism. The mechanism in alkaline electrolyte is

J.W. Gallaway et al. / Journal of Power Sources 321 (2016) 135e142136

often treated in shorthand as a first electron reaction followed by asecond electron reaction [1,2].

MnO2 þ H2O þ e� / MnOOH þ OH� (1)

MnOOH þ H2O þ e� / Mn(OH)2 þ OH� (2)

This understanding is true only in limited circumstances. Whiledischarge always begins via the proton insertion in Reaction (1),upon reaction beyond a poorly defined extent various other man-ganese oxides are observed, including the spinels Mn3O4 (haus-mannite) and ZnMn2O4 (hetaerolite), the latter of which isproduced in batteries with zinc anodes. These spinels have lowelectrochemical activity, and will not recharge to MnO2. Despitetheir importance in keeping g-MnO2 batteries from being widelyused rechargeably, the reaction mechanisms leading to these spi-nels are largely unknown, as are their precise points of formationduring discharge.

Details from the literature show a strong influence of experi-mental factors, for example the mass percent of g-MnO2, whichmust be diluted with a conductive material like carbon. Variousdilutions show a range of discharge results, and the studies mostcited are difficult to compare directly. In one, constant currentdischarge at 87.5% mass loading led first to Mn3O4 followed byMn(OH)2, while in a second, potential scanning at 60% loading ledto Mn(OH)2, which was recharged through g-Mn2O3, b-MnOOH,and g-MnOOH, and gave Mn3O4 only on the second cycle [3,4]. Inthe ensuing years there has been no resolution of this mechanism.A later in situ study at 80% loading showed simultaneous formationof Mn3O4 and Mn(OH)2 at 0.8 extent of Reaction (1) [5]. Anotherstudy at 3% loading showed Mn(OH)2 with no Mn3O4, suggestingthat Mn3O4 in earlier reports was a consequence of sample prep-aration [6]. Addition of a zinc anode has complicated in-terpretations further. It has been reported that ZnMn2O4 wasobserved in situ during the first electron reaction, and thatZnMn2O4 and Mn3O4 form at different stages, governed by the cellpotential [6,7]. These variations can also be attributed to datainterpretation, as the manganese oxides are difficult to distinguishexperimentally. Also in play is the nature of porous electrodes,which almost always display nonuniform reaction rates, especiallyat high mass loading or high rate. Reaction extents calculatedassuming a bulk average will differ from the true values and thuscall mechanisms into question.

In this study we target the pathway to and point of spinel for-mation under the most relevant experimental conditions: thosefound within commercial g-MnO2 batteries. Knowledge of theprecisemolecular pathway leading to spinel formationmay be usedto engineer a method to block or reverse this pathway. For practicaluse, g-MnO2 electrodes operate at high mass loading and signifi-cant discharge rate. The technique of using photons of high energyand high flux for energy dispersive X-ray diffraction (EDXRD) hasbeen used to track material evolution within electrochemical sys-tems that more resemble actual devices thanmodel systems, whichmay exclude factors that have a significant impact [8e11]. Thus weuse commercial alkaline batteries as the test bed. These havequalities not found in many model systems: ~97% active materialloading, thick electrodes, and relatively starved electrolyte. Wehave previously demonstrated that high energy EDXRD has thesensitivity required to spatially distinguish ZnMn2O4 and Mn3O4,which have a d-spacing difference of only 0.04 A, within the sealedsteel casing of a battery, using a moderate data collection time of20 s per location [12]. To account for the distribution of currentwithin the thick electrodes, we use a proven macrohomogeneousbattery model to calculate local discharge rates based on the bulkdischarge rate. The practical interest in this mechanism is due to

the emerging need for inexpensive and safe electricity storage onthe scale of the power grid [13]. Because g-MnO2 is inexpensive,non-toxic, and non-flammable, it is uniquely suited for massive-scale stationary batteries, fulfilling requirements challenging forother battery chemistries [14].

2. Experimental

Because the photons used were highly penetrating it waspossible to use LR6 commercial alkaline batteries as a test bed, andthese were Duracell MN1500 AAwith a rated capacity of 2.85 Ah at0.8 V. The reported specific and volumetric energy densities were143 Wh kg�1 and 428 Wh L�1. In operando battery discharge wasperformed using an eight channel MTI battery cycler.

EDXRD experiments were conducted at Brookhaven NationalLaboratory (BNL) at beamline X17B1 of the National SynchrotronLight Source (NSLS). An incident X-ray beam penetrated the battery,and the diffracted beam was detected at a fixed angle of 2q ¼ 3�.The incident beamwas white beam radiation with an energy rangeof ~20e200 keV. The diffracted beam intensity was measured usinga high resolution germanium detector with a digital signal pro-cessor and a 8192-channel multichannel analyzer. The channelnumber to inverse d-spacing calibration was made using LaB6 andCeO2 standards. Diffraction data was smoothed using a Savitz-kyeGolay filter.

Collimation slits were used to set the sizes of the incident anddiffracted beams, and the intersection of the two beam cross-sectional areas defined a gauge volume (GV) in space [10,15,16].By positioning the GV inside the battery, diffraction data wascollected at a well-defined spatial location as illustrated in Fig. 1.Settings were di ¼ 130 mm, ds ¼ 100 mm, and width in the x3 di-rection of 2 mm. At these conditions the GV was 130 mm along theradius of the battery. For the 100 mA battery, a larger di setting of300 mm was used to better resolve the (110) reflection of rams-dellite. Batteries were “mapped” by positioning the GV at the top ofthe can and moving the battery, using an x-y-z stage, to scan the GVradially inward to the battery center. Data was collected at eachpoint for 20 s. Previously it was found that batterymaps collected atdifferent axial or azimuthal locations could be expected to be ingeneral agreement with one another [11]. Axially, the data wascollected at the battery half-height. The beam caused no damage tothe battery materials on the time scale of the experiments.

3. Results

3.1. Battery discharge rates

Batteries were discharged at 400, 143, 100, and 50 mA, as shownin Fig. 2. Upon reaching 0.02 V the discharge continued at constantpotential until the experiment was halted. The gradual drop in cellpotential was chiefly due to the g-MnO2 cathode material, whosepotential drops from its value of E0 ¼ 0.26 V vs. Hg|HgO (1.64 V vs.Zn) as it is discharged. The batteries were continually mapped byEDXRD to track the crystal structure of the active material as afunction of position inside the battery. EDXRD map times areoverlaid on the potential curves. Other than the 400 mA case, thedischarge curves displayed characteristic dips in potential. To makethemmore visible, the first derivatives of the potential curves dV/dtare show, and in these the dips appear as points of inflection.

The initial MnO2 discharge reaction is the proton insertion inEquation (3).

MnO2 þ xH2O þ xe� / MnO2�x(OH)x þ xOH� (3)

The variable x indicates the number of electron equivalents that

Fig. 1. (A) Experimental setup showing the EDXRD signal from a GV localized withinthe battery. (B) Schematic of white beam X-ray diffraction from a gauge volume (GV)inside the battery. The GV size was defined by the collimation slits. The blue colorshows the diffracted beam path to the detector. Beam and GV sizes are enlarged manytimes for visibility. (For interpretation of the references to colour in this figure legend,the reader is referred to the web version of this article.)

J.W. Gallaway et al. / Journal of Power Sources 321 (2016) 135e142 137

have reacted per Mn atom. As the reduction of Mn4þ to Mn3þ isaccompanied by a proton insertion into the MnO2 tunnel structure,x is also the number of protons inserted. At the endpoint of thisreaction x ¼ 1 and a-MnOOH is formed, which has the mineralname groutite. These commercial cells had an extremely high massloading of MnO2 in the cathode of 96.8% of total solids [17]. Thetheoretical capacity of the first election reaction in Equation (3) was3.19 Ah. Table 1 lists times, values of xb, and relevant EDXRD mapsof observed potential dips in the discharge curves.

3.2. Battery modeling

Batteries discharged at these rates were expected to have

Fig. 2. Discharge curves of 3 AA alkaline batteries. Cell potential is in the top panel,with its first derivative (dotted). Dips in the potential curves are indicated. Current isshown in the bottom panel, with electron equivalents xb for the bulk-averaged cath-odes (dashed). Numbered EDXRD maps are shown on the potential curves. Width ofthe rectangles represents the actual data collection time.

variable current distributions and concentrations as a function ofradius within the cathode, from r ¼ 4.2e6.9 mm. The alkalinebattery model of Chen and Cheh was adapted to calculate localvalues of x at a given radius, xr [18]. This model was based on themacrohomogeneous theory of porous electrodes, and accounted fordiffusion and convection in the electrolyte, changes in porosity andelectrolyte composition, ionic migration in concentrated solution,and proton transport in the MnO2 particles [19,20]. Details of themodel and its solution are discussed in the supporting information.Current distributions in these batteries evolved in diverse ways,and these are plotted in Fig. 3 as volumetric electrochemical reac-tion rate or transfer current. The initial disparity between the frontand back of the cathode varied from only 13% at 50 mA to 140% at400 mA.

3.3. Structure of the g-MnO2 starting material

The g-MnO2 active material has a complex and defect-filledstructure that is known to be an intergrowth of two MnO2 poly-morphs. The first is ramsdellite, an orthorhombic material (spacegroup Pbnm) with a 2� 1 tunnel structure. The second is pyrolusiteor b-MnO2, a tetragonal material (rutile structure, space group P42/mnm) with a 1 � 1 tunnel structure. The MnO6 octahedra of thesetwo pure materials are shown in Fig. 4. g-MnO2 can be described byassuming it has a ramsdellite structure affected by two kinds ofdefects: (i) a stacking disorder in which pyrolusite layers areinterspersed within the ramsdellite layers, and (ii) a microtwinningof the ramsdellite lattice [21]. These defects can be quantified withtwo parameters: defect type (i) by Pr, which is the fraction of py-rolusite layers in the structure; and defect type (ii) by Tw, which is astacking transition probability in the ramsdellite part of the mate-rial [22]. A Tw of 0% is perfect ramsdellite and 100% is maximallytwinned. Highly active g-MnO2 is highly twinned and incorporatesboth ramsdellite and pyrolusite into its structure. Bailey and Donneprovide examples of a wide range of g-MnO2 sample properties[23].

The g-MnO2 used was extensively twinned, Tw > 80%, typical forelectrolytic manganese dioxide (EMD). This meant that the ortho-rhombic ramsdellite structure was not evident from the diffractionpattern. Instead the MnO2 X-ray reflections were characteristic of apseudo-hexagonal unit cell, and the material was thus indexed ashexagonal ε-MnO2 (space group P63/mc) [24]. However as ex-pected, the (110) reflection of an orthorhombic Pbnm space group,indicative of the presence of ramsdellite, was also clearly apparent.From the average of several EDXRD patterns this reflection waslocated at a value of 0.2488 Å�1, corresponding to a Cu Ka 2q of22.12�. The method of Chabre and Pannetier was used to calculatePr, which was 0.39 [22]. This was the fraction of pyrolusite layers inthe starting material.

3.4. Operando EDXRD

Cathode reflections during discharge of the 400 mA batteryevolved as a function of position in the battery, as shown in Fig. 5.Initially the reflections associated with MnO2 shifted to a lowerinverse d-spacing as proton insertion expanded the unit cell. Theproton insertion in Equation (3) begins as a single-phase reaction,with protons inserted into the 2 � 1 tunnels of ramsdellite [1,25].The first evidence of a-MnOOH was due to a diffuse peak corre-sponding to the (111) reflection and a more pronounced (400)reflection. This (400) reflection rapidly split from the hexagonal(100)h reflection of ε-MnO2 as it shifted leftward. Immediatelyfollowing this split, the spinel structure of ZnMn2O4 formed directlyat the separator. After further reaction, which was at constantvoltage in this battery, the other reflections of a-MnOOH became

Table 1Data at dips in cell potential during the battery discharges.

Rate (mA) Dip in cell potential Map of formation at the separator Model calculation

Time (h) EDXRD map xb a-MnOOH (400) ZnMn2O4 xr at r ¼ 4.2 mm xr at r ¼ 5.0 mm

400 e e e (6)a 6 e e

143 16.5 16 0.69 16 17 0.79 0.75100 24.0 25 0.71 25 26 0.79 0.7350 47.3 48 0.72 48 49 0.80 0.76

a Data was not collected at map 5 in the 400 mA cell due to an X-ray beam injection.

J.W. Gallaway et al. / Journal of Power Sources 321 (2016) 135e142138

distinct. The a-MnOOH (210) reflection transitioned smoothly from(100)h of ε-MnO2, and a-MnOOH (211) transitioned from ε-MnO2(101)h. An intensifying signal from ZnMn2O4 at the separator wasaccompanied by loss of the a-MnOOH signal, marked by note (c). Bythe time of map 32, after the experiment was halted and the cellrested, only ZnMn2O4 was apparent at the separator, with no traceof a-MnOOH. Finally, the spinel structure of Mn3O4 formedthroughout the cathode from r ¼ 4.57e6.55 mm. Mn3O4 wasaccompanied by a related structure labeled spinel 2. Traces ofMn3O4 and spinel 2 such as those marked by (a) and (b) intermit-tently appeared near the detection limit beginning at map 12, withtheir full patterns forming at map 16.

Localized reflections adjacent to the separator in the 100mA cellare shown in Fig. 6. a-MnOOH first became apparent as the diffuse(111) signal at map 23, and this was followed by splitting off of themore intense (400) reflection at map 25. ZnMn2O4 then formedsuddenly, indicated by the marker (a), with the three sharpest re-flections at map 26 and the remainder of the pattern by the next

Fig. 3. Battery model calculations for cathodes in the cells discharged in Fig. 2. The final timLocal values for the transfer current j in mA cm�3 and xr are shown. (A) 50 mA; (B) 100 m

map. This result was general, and it was the case in every batterytested that spinel appeared directly following the a-MnOOH (400)reflection. It was also the case that dips in the cell potential coin-cided with formation of the a-MnOOH (400) reflection in thecathode adjacent to the separator. Table 1 lists the times of theseoccurrences to make the connection clear. Of course with thehighest discharge rate of 400 mA no potential dip was seen. Insteadthe potential fell precipitously at map 6, and this was caused byheavy formation of ZnO at the separator on the anode side.We havepreviously detailed the growth of such ZnO blocking layers [11].

Evolution of the 143 mA battery during the end of discharge isshown in Fig. 7. During the time between maps 18 and 19 the cellpotential fell, tripping the constant voltage stage after map 19. Atr¼ 4.17mm the appearance of a-MnOOH (400) occurred at map 16,followed by ZnMn2O4 at map 17. Behind this layer, a similar pro-gression was observed in which a-MnOOH (400) was insteadclosely followed by Mn3O4. The maps of first formation for thesematerials are listed for each radius in Table 2. In some cases there

e corresponds to observed formation of the a-MnOOH (400) reflection at the separator.A; (C) 143 mA; (D) 400 mA.

Fig. 4. Structures of (a) ramsdellite (R-MnO2, 2 � 1 tunnels) and (b) pyrolucite (b-MnO2, 1 � 1 tunnels) in the direction along the tunnels. Bottom panels show the correspondingfully proton-inserted structures (c) a-MnOOH and (d) g-MnOOH. The dotted line shows the incompatibility between groutite, which is inserted, and pyrolusite, which is not.

J.W. Gallaway et al. / Journal of Power Sources 321 (2016) 135e142 139

was ambiguity, if only one reflection appeared and was near thenoise level of the EDXRD signal, and for these a map range is listed.By map 19 Mn3O4 had formed throughout the cathode. As withZnMn2O4, Mn3O4 closely followed the well defined signature of a-MnOOH. We conclude a-MnOOH is the precursor to any spinelstructure.

Fig. 5. Cathode reflections during discharge of the 400 mA battery. Map numbers corresponr ¼ 4.17 mm. That nearest the current collector was at r ¼ 6.55 mm. This shows the full ra

The final calculated times in Fig. 3 correspond to the suddenappearance of the a-MnOOH (400) reflection at the separator, andthis occurred at xr ¼ 0.79e0.80. These values are in Table 1. Fromthe model calculation for the 400 mA cell in Fig. 3D, a-MnOOHformation at the separator would have occurred just before map 6.However, since maps were collected at the same rate as the other

d to the times shown in Fig. 2. The section of the cathode nearest the separator was atnge of discharge products.

Fig. 6. Cathode reflections during discharge of the 100 mA battery from EDXRD maps 19e29 near the separator at r ¼ 4.15 mm. Details from near note (a) are shown to the right,where the progression to ZnMn2O4 formation is clearly apparent. A small ZnO signal was due to the GV overlapping slightly into the separator and anode.

J.W. Gallaway et al. / Journal of Power Sources 321 (2016) 135e142140

cells despite the much more rapid discharge, this moment was notresolved from that of spinel formation. For the 143 mA cell, cathodeutilization values were also calculated beyond this point. While themodel did not account for ZnMn2O4 formation, this provided good

Fig. 7. Cathode reflections during discharge of the 143 mA battery near the end ofdischarge. The section of the cathode nearest the separator was r ¼ 4.17 mm. For-mation of Mn3O4 begins behind the ZnMn2O4 layer and then progress to larger radii.The peaks marked (b) were ZnO reflections from the separator/anode.

approximations for xr values during the Mn3O4 formation in Fig. 7.These are listed in Table 2 and reveal that a-MnOOH formation inthe bulk of the cathode occurred at xr ¼ 0.78e0.79, similar to nearthe separator. Mn3O4 formation closely followed, analogous toZnMn2O4. Thus, the formation pathway of these two spinels isfundamentally the same, following the same chain of events.

4. Discussion

Direct observation showed that [Mn2]O4 spinels formedimmediately following appearance of the a-MnOOH (400) reflec-tion. This transition to the a-MnOOH structure occurred in a tightrange of x ¼ 0.78e0.81 at all locations in all batteries, regardless oflocal reaction rate. We thus define a single value of state of chargeat the phase transition of xt ¼ 0.79, the most common value. The g-MnO2 was calculated to have a fraction of 0.39 pyrolusite layers inthe crystal lattice, and state of charge to fully insert protons into all2 � 1 ramsdellite tunnels was thus

x2 ¼ ð2� 2PrÞð2� PrÞ ¼ 0:76 (4)

This meant the a-MnOOH phase transition occurred universallyat 104% capacity required to fill the ramsdellite fraction of the g-MnO2. Ramsdellite and pyrolusite discharge to different poly-morphs of MnOOH [26,27]. a-MnOOH has a 2 � 1 tunnel structure,corresponding to the fully proton-inserted structure of ramsdellite.g-MnOOH (manganite) has a 1 � 1 tunnel structure, correspondingto that of pyrolusite. This relationship between filled and unfilledstructures is illustrated in Fig. 4. At the phase transition somenumber of the inserted protons had occupied the tunnels of bothramsdellite and pyrolusite, xRt and xrt respectively, andxRt þ xrt ¼ xt. The discharge reaction up to the point of a-MnOOHformation can be written:

½ð1� x2Þr$MnO2�½x2R$MnO2� þ xtH2Oþ xte�/ð1� x2Þ r$MnO2�xrt=ð1�x2ÞðOHÞxrt=ð1�x2Þþ

x2 a$MnOOHxRt=x2 þ xtOH�(5)

Pyrolusite domains, which have a rutile structure, are labeled r-MnO2; ramsdellite domains are labeled R-MnO2. Here the bracketsindicate joined domains in the g-MnO2 crystal lattice. Reaction to xtcaused these domains to be sheared apart due to the latticemismatch between a-MnOOH and pyrolusite. At this point a-MnOOH became the dominant diffraction signal.

Table 2Mn3O4 formation in the 143 mA cell.

Radius (mm) Map of first formation Model calculation xr Spinel (103) reflection d�1 (�1), map 19

a-MnOOH (400) Spinel Map 17 Map 18 Map 19

4.17 16 17 0.87 0.91 0.94 0.368b

4.57 16e17 17e18 0.85 0.88 0.92 0.3644.96 17 18 0.81a 0.86 0.90 0.3635.36 17e18 18 0.78a 0.83 0.88 0.3625.76 18 18e19 0.76 0.81a 0.86 0.3626.16 18 19 0.74 0.79a 0.85 0.3626.55 18e19 19 0.73 0.78a 0.84 0.361

a First appearance of the a-MnOOH (400) reflection occurred after local state of charge xr passed ~0.79.b This spinel was ZnMn2O4.

J.W. Gallaway et al. / Journal of Power Sources 321 (2016) 135e142 141

Ripert and co-workers have reported that this lattice shearingoccurs after full filling of the ramsdellite domains [28,29]. Usingoperando neutron diffraction on g-MnO2 with Pr ¼ 0.3 andx2 ¼ 0.82, they observed Mn(OH)2 formation at xb ¼ 0.8 and notedthis matched x2, suggesting the ramsdellite is filled while the rutiledomains remain inactive until shearing. However, this cannot bethe case as a-MnOOH reflections shifted leftward after their for-mation due to continued proton insertion, illustrated by note (b) inFig. 6. Thus the ramsdellite remained incompletely filled and non-stoichiometric immediately upon shearing, with its formula givenby a$MnOOHxRt=x2. This is supported by 2HMAS NMR results of Paikand Grey, who found that the signal from deuterons in 1 � 1 tun-nels was strong during initial discharge [30]. This means xrt s 0.Because a multitude of materials formed after xt and the diffractionpatterns were not strictly quantitative, it was challenging tocalculate xrt with certainty. It was possible to define the bounds0.03 < xrt < 0.24, corresponding to a-MnOOH0.72 or more filled.However, as no g-MnOOH signal was detectedwe assume a value ofxrt below 0.12 was more likely, consistent with rutile layers beingsheared into isolated surface sites before their half-reduction. Thiswould mean ramsdellite was more filled than a-MnOOH0.88 uponshearing.

After a-MnOOH formation and shearing of the lattice, [Mn2]O4spinel building occurs based on whichever complex ion is availableto fill the tetrahedral sublattice.

2 a­MnOOH þ ZnðOHÞ2�4 /ZnMn2O4þ2 OH�þ2 H2O (6)

2 a­MnOOH þ MnðOHÞ2�4 /Mn3O4þ2 OH�þ2 H2O (7)

Results above, such as note (c) of Fig. 5, demonstrate that spinelformation corresponds to loss of the a-MnOOH signal. The source ofMn(OH)42� ion may be dissolution of MnIII followed by dispropor-tionation, or via an electrochemical reaction. It has been variouslysuggested that spinel formation is either by a reaction through thesolution or essentially solid state [31e34]. In batteries dischargedand rested, ZnMn2O4 and Mn3O4 were distinguished by their (103)reflections, which when co-located appears as peak splitting [12].Viewed in operando the (103) reflection showed significant shift-ing, which was a function of radius or distance from the separator.This is marked in Fig. 7 by note (a) and the corresponding values ofd�1 are listed in Table 2. Since the peak was not split, this shows asingle-phase mingling of ZnMn2O4 and Mn3O4, with more Znbringing the d�1 value higher. The unknown spinel 2 d-spacing waseven smaller than that of ZnMn2O4, and is evidence of partially-formed spinel, for example the [Mn2]O4 spinel framework withthe tetrahedral site not yet filled or temporarily filled by a smalleratom. Both ZnMn2O4 and Mn3O4 have distorted tetragonal spinelstructures (space group I41/amd), and spinel 2 appeared analogousto the Mn3O4 (112) and (200) reflections. While difficult to identify

based on the two visible reflections, spinel 2, as well as the lack ofpeak splitting, suggest spinel building is slow because the operandoresults show intermediate species.

The reaction mechanism observed in the current work was atvariance with other studies that show Mn(OH)2 as a dischargeproduct, or as a precursor toMn3O4 formation. In other workwith alower mass loading we have observed Mn(OH)2 is easily distin-guished in EDXRD patterns by reflections at d�1 ¼ 0.348 Å�1 and0.407 Å�1, but was never observed here. Tarascon and co-workersobserved Mn(OH)2 as the final discharge product in their experi-ments, but the mass loading of g-MnO2 was 3%, with an extremeexcess of carbon [6]. They also observed that ZnMn2O4 does notform from MnOOH chemically produced via reduction by hydra-zine. From this we conclude that the a-MnOOH reactants inEquations (6) and (7) must be the electrochemically-producedmaterial, rather than ones chemically synthesized. The resultspresented here show the final discharge product, Mn3O4 orMn(OH)2, is not an intrinsic property of g-MnO2. However, theconversion to a-MnOOH is fundamental, and occurs at a precise andobservable point in the discharge reaction.

5. Conclusions

In a discharging g-MnO2 electrode, formation of ZnMn2O4 andMn3O4 follow the same mechanism, which occurs immediatelyfollowing the appearance of a-MnOOH as a distinct phase. Time offormation of the a-MnOOH phase is identified by the splitting off ofits (400) reflection from the hexagonal (100)h reflection of ε-MnO2.After this split, the ε-MnO2 (100)h reflection smoothly transitions tothe a-MnOOH (210) peak upon further discharge. If the electrode isrecharged, the ZnMn2O4 and Mn3O4 materials remain persistent.Because these spinels are electrochemically inactive, their presencein the electrode causes a profound drop in capacity. This means theevent that ultimately leads to loss of reversibility for a g-MnO2cathode is the occurrence of well-formed a-MnOOH, after whichspinel formation is imminent. Furthermore, this splitting off of thea-MnOOH (400) was observed to always occur at a local value of thestate of charge of xr z 0.79. This was most likely filling of theramsdellite tunnels in the g-MnO2 corresponding to a-MnOOH0.88or greater. The first occurrence of a-MnOOH (400) next to thebattery separator corresponded to dips in battery potentialobservable in Zn-MnO2 discharge curves.

Acknowledgements

The authors would like to thank Hui Zhong for helpful assistanceat the beamline. This work was supported by the LaboratoryDirected Research and Development Program of Brookhaven Na-tional Laboratory (LDRD-BNL) Under Contract No. DE-AC02-98CH10866with the U.S. Department of Energy. Use of the National

J.W. Gallaway et al. / Journal of Power Sources 321 (2016) 135e142142

Synchrotron Light Source, Brookhaven National Laboratory, wassupported by the U.S. Department of Energy, Office of Science, Of-fice of Basic Energy Sciences, under Contract No. DE-AC02-98CH10886.

Appendix A. Supplementary data

Supplementary data related to this article can be found at http://dx.doi.org/10.1016/j.jpowsour.2016.05.002.

References

[1] A. Kozawa, R.A. Powers, J. Electrochem Soc. 113 (1966) 870e878.[2] A. Kozawa, R.A. Powers, J. Electrochem Soc. 115 (1968) 122e126.[3] D. Boden, C.J. Venuto, D. Wisler, R.B. Wylie, J. Electrochem Soc. 114 (1967)

415e417.[4] J. Mcbreen, Electrochim. Acta 20 (1975) 221e225.[5] C. Mondoloni, M. Laborde, J. Rioux, E. Andoni, C. Levyclement, J. Electrochem

Soc. 139 (1992) 954e959.[6] R. Patrice, B. Gerand, J.B. Leriche, L. Seguin, E. Wang, R. Moses, K. Brandt,

J.M. Tarascon, J. Electrochem Soc. 148 (2001) A448eA455.[7] Y. Paik, W. Bowden, T. Richards, R. Sirotina, C.P. Grey, J. Electrochem Soc. 151

(2004) A998eA1011.[8] K. Kirshenbaum, D.C. Bock, C.Y. Lee, Z. Zhong, K.J. Takeuchi, A.C. Marschilok,

E.S. Takeuchi, Science 347 (2015) 149e154.[9] K.C. Kirshenbaum, D.C. Bock, Z. Zhong, A.C. Marschilok, K.J. Takeuchi,

E.S. Takeuchi, J. Mater. Chem. A 3 (2015) 18027e18035.[10] J. Rijssenbeek, Y. Gao, Z. Zhong, M. Croft, N. Jisrawi, A. Ignatov, T. Tsakalakos,

J. Power Sources 196 (2011) 2332e2339.[11] J.W. Gallaway, C.K. Erdonmez, Z. Zhong, M. Croft, L.A. Sviridov,

T.Z. Sholklapper, D.E. Turney, S. Banerjee, D.A. Steingart, J. Mater. Chem. A 2(2014) 2757e2764.

[12] J.W. Gallaway, M. Menard, B. Hertzberg, Z. Zhong, M. Croft, L.A. Sviridov,D.E. Turney, S. Banerjee, D.A. Steingart, C.K. Erdonmez, J. Electrochem Soc. 162(2015) A162eA168.

[13] J. Eyer, G. Corey, Sandia report, SAND2010-0815, 2010.[14] N.D. Ingale, J.W. Gallaway, M. Nyce, A. Couzis, S. Banerjee, J. Power Sources

276 (2015) 7e18.[15] M. Croft, V. Shukla, E.K. Akdogan, N. Jisrawi, Z. Zhong, R. Sadangi, A. Ignatov,

L. Balarinni, K. Horvath, T. Tsakalakos, J. Appl. Phys. 105 (2009).[16] M. Croft, I. Zakharchenko, Z. Zhong, Y. Gurlak, J. Hastings, J. Hu, R. Holtz,

M. DaSilva, T. Tsakalakos, J. Appl. Phys. 92 (2002) 578e586.[17] W. Luo, A.B. Shelekhin, M. Sylvestre, in: US 6,858,349 B1, The Gillette Com-

pany, 2005.[18] J.S. Chen, H.Y. Cheh, J. Electrochem Soc. 140 (1993) 1205e1213.[19] J. Newman, W. Tiedemann, Aiche J. 21 (1975) 25e41.[20] R. Pollard, J. Newman, J. Electrochem Soc. 128 (1981) 503e507.[21] D.E. Simon, R.W. Morton, J.J. Gislason, Adv. X-ray Anal. 47 (2004) 267e280.[22] Y. Chabre, J. Pannetier, Prog. Solid State Chem. 23 (1995) 1e130.[23] M.R. Bailey, S.W. Donne, Electrochim. Acta 56 (2011) 5037e5045.[24] P.M. de Wolff, J.W. Visser, R. Giovanoli, R. Brutsch, Chimia 32 (1978) 257e259.[25] A. Kozawa, J.F. Yeager, J. Electrochem Soc. 112 (1965) 959e963.[26] T. Kohler, T. Armbruster, E. Libowitzky, J. Solid State Chem. 133 (1997)

486e500.[27] L.A.H. MacLean, F.L. Tye, J. Solid State Chem. 123 (1996) 150e160.[28] Pannetier Ripert, Poinsignon Chabre, Mat. Res. Soc. Symp. Proc. 210 (1991)

359e365.[29] M. Ripert, C. Poinsignon, Y. Chabre, J. Pannetier, Phase Transit 32 (1991)

205e209.[30] Y. Paik, J.P. Osegovic, F. Wang, W. Bowden, C.P. Grey, J. Am. Chem. Soc. 123

(2001) 9367e9377.[31] L.T. Yu, J. Electrochem Soc. 144 (1997) 802e809.[32] D. Im, A. Manthiram, B. Coffey, J. Electrochem Soc. 150 (2003) A1651eA1659.[33] M.M. Thackeray, M.H. Rossouw, R.J. Gummow, D.C. Liles, K. Pearce, A. Dekock,

W.I.F. David, S. Hull, Electrochim. Acta 38 (1993) 1259e1267.[34] W.I.F. David, M.M. Thackeray, P.G. Bruce, J.B. Goodenough, Mater Res. Bull. 19

(1984) 99e106.