American Mineralogist, Volume 79, pages 80-90, 1994

In-situ study of the thermal behavior of cryptomelane by high-voltage andanalytical electron microscopy

Pl,ur,o Mlncos YlscoNcnlos, HaNs-Ruoor,r WnNrDepartment of Geology and Geophysics, University of California, Berkeley, California 94720, U.S.A.

Cnucx EcrrnnNational Center for Electron Microscopy, University of California, Lawrence Berkeley I-aboratory,

Berkeley, California 9 47 20, U.S. A.

AssrRAcr

Scanning electron microscope (SEM) and transmission electron microscope (TEM) stud-ies show that cryptomelane and Cu-rich cryptomelane crystals from weathering profiles inBrazilrangefrom l0to200nmindiameterandhaveaspectratiosof l: l0to l: l00betweenthe short and long dimensions. Acicular crystals identified in hand specimen and in thescanning electron microscope are often composed of bundles of fibers elongated along thec axis (tetragonal) or the b axis (monoclinic). In-situ heating with a high-voltage trans-mission electron microscope (HVEM) and an analytical electron microscope (AEM) in-dicates that upon heating in vacuum cryptomelane crystals begin to transform into a mixedhausmannite and manganosite phase at 648 'C. Mineral transformations are followed bythe loss of K from the cryptomelane tunnel sites. Cu-rich cryptomelane transforms intohausmannite between 330 and 515 'C, and manganosite begins to form at 620 "C. Cu-richcryptomelane also loses K and exsolves native copper upon heating, suggesting that Cuoccupies the same site (the A site) as K. Similar mineral transformations are observedwhen the same samples are heated in air, although the transformations occur at highertemperatures than those observed in vacuum.

The variation in the thermal stability of hollandite-group minerals suggests that the sizeofA-site occupants may control the mobility of cations along the tunnel direction. Relativegrain sizes and the degree of crystallinity of supergene cryptomelane may also influencethe thermal stability of hollandite-group minerals. The applicability of cryptomelane andother hollandite-group minerals to K-Ar and 40Ar/3eAr dating is discussed in light of theelectron microscope investigation results. Ar release from the cryptomelane samples duringlo{r/3eAr vacuum laser heating is also discussed in terms of the phase transformationsobserved by electron microscopy.

INrnonucttoN

Recent 4o{r/3e{r laser-heating dating has shown thatK-bearing manganese oxides (cryptomelane and hollan-dite) are potentially suitable minerals for dating weath-ering processes by the K-Ar and 4o{r/3e Ar methods (Vas-concelos et al., 1992). The ubiquitous presence of thesedatable oxides in weathering profiles and soils throughoutthe world may provide information on rates of weather-ing processes, soil formation, paleoclimates, and land-scape evolution. To establish the applicability of cryp-tomelane to K-Ar ar.d 40Ar/3eAr dating it is necessary todefine its composition, structure, and range of stability.It is imperative to determine the size range of crypto-melane crystals in order to assess the possibility of 3eArrecoil from the mineral structures during the neutron ir-radiation procedure required for 4oArl3eAr analysis. Inaddition, it is necessary to investigate ordering ofthe tun-nel site occupants to determine if cations, HrO, and the

0003-004x/94l0 I 02-0080$02.00

radiogenic aoAr occupying these sites are easily exchange-able. Finally, in order to interpret the results of 40 Ar/3e Arexperiments better we must understand the thermal be-havior of hollandite-group minerals. In particular, thetemperature and the mineral transformations associatedwith the Ar release steps need to be known to determineif cryptomelane retains Ar at normal Earth's surface tem-perature and to determine that thermal degassing of man-ganese oxides does not occur during neutron irradiation.

Cryptomelane and hollandite belong to the hollanditegroup and have the general formula (A),-rBrO,..xHrO.The A site can be occupied by large cations such as K+,Ba2*, Na*, Cu*, Pb2*, Rb2*, Sr2t, Cs2+, etc., whereasthe 16lB site includes n4ttr*, Mno*, ps:+, dl3+, Sio*, andMg'?+ (Burns and Burns, 1979). Cryptomelane is the Kand hollandite is the Ba end-member of a solid-solutionseries (Bystrdm and Bystr6m, 1950). Bystr<im and Bys-tritm (1950) showed that ,4 cations occupy a tunnel struc-ture formed by edge-sharing B-O octahedra. Electrostatic

80

VASCONCELOS ET AL.: HVEM AND AEM STUDIES OF CRYPTOMELANE 8 l

repulsion between the I cations requires that the A sitesbe only partially occupied. Bystrtim and Bystrdm (1950)have proposed that the tunnel cations occupy a specialposition along the tunnel, which should be 0, t/2,0 in themonoclinic hollandite structure and 0, 0, t/z in the tetrag-onal cryptomelane structure. Charge-balance require-ments indicate that lower valence cations must occupythe B sites to compensate for the presence of the A cations(Post et al., 1982; Post and Burnham, 1986; Vicat et al.,1986). Vacancies in the B sites have also been proposedas a charge-balance mechanism (Bystrdm and Bystr6m,1950). Among the common K-bearing manganese oxides,hollandite is the most promising datable group becauseof its large range of stability and its occurring as a hy-pogene phase in hydrothermal environments (Hewett andFleischer, 1960), as metamorphic minerals in high p-Zenvironments (Roy and Purkait, 1965; Miura et a1., 1987),and as supergene phases in soils and weathering profiles(Hewett and Fleischer, 1960). In addition, the 2 x 2 tun-nel structure of hollandite-group minerals is more likelyto retain K and Ar than the more open structures of otherK-bearing manganese oxides, such as romanechite (2 x 3tunnels), todorokite (3 x 3 tunnels), and birnessite (lay-ered structure) (Burns and Burns, 1979; Post and Veblen,1990).

Thermogravimetric analysis (TGA) and differential-thermal analysis (DTA) of manganese oxides have indi-cated that dehydration reactions occur at 200-400 .C,

and dehydroxylation occurs at higher temperatures (-650"C) (Kulp and Perfetti, 1950; Ljunggren, 1955). Heatingexperiments in open air have indicated that cryptomelanetransforms into bixbyite at 600 "C, into tetragonal haus-mannite at 825 "C, and finally into cubic hausmannite at1050 (Faulring et al., 1960) or at I160'C (Van Hook andKeith, 1958). DTA analyses by Kudo et al. (1990) showedthat tetragonal cryptomelane transforms into a monoclin-ic structure at 7 l0 'C in air and into hausmannite throughbixbyite at 850-950 "C.

To investigate the mineralogical and crystallographicrelations of supergene potassium manganese oxides, weanalyzed cryptomelane samples by powder XRD, elec-tron microprobe (EM)" SEM, and TEM. We also moni-tored phase transformations of two of these samples insitu, with a heating stage in the HVEM. The degassinghistory of two samples was further studied by laser-heat-ing aoAr/3eAr analysis, and the degassing steps are corre-lated with the mineralogical transformations observed inthc TEM.

Frnr,o srrnsWe investigated hollandite-group minerals from two

weathering profiles in Brazil. The first sample, pEG-01(Table 1), came from deep weathering profiles in the peg-matite fields in the state of Minas Gerais (Vasconcelos etal., 1992). This sample occurs as intimately intercon-nected botryoidal potassium manganese oxide nodulesforming a three-dimensional dendrite-like pattern. Dur-ing weathering of primary minerals, supergene solutions

TABLE 1. Microprobe analyses of manganese oxides

Monoclinic Tetragonal

F226-121 F40-142 F115-139.8 pEG-01

0.03 0.050.08 0.13o.22 0.00

56.43 57.570.14 0.000.00 0.051.40 0.000.17 0 .120.09 0.072.98 4.910.01 0.550.08 0.130.24 0.44

34.79 34,7533.62 35.7196.66 98.77

0.010.020.037.660.020.007.74

0.01o.070.007.540.000.017.63

0.08 0.000.03 0.020.03 0.020.57 0.900.00 0.030.01 0.010.72 0.99

0.06 0.108.60 8.72

7't7 .55 717.564.31

0.60 0.601 .26 1.310.48 0.46

Note.' elemental weight percentages and formulas based on 16 O atoms.S.G. : specific gravity in grams per cubic centimeters.

enriched in K+ and Mn2* fill open spaces provided bymiarolitic cavities in the pegmatites, promoting the pre-cipitation of high-purity cryptomelane inside these cavi-ties. The autocatalytic nature of manganese oxide precip-itation leads to the continuous adsorption and subsequentprecipitation of dissolved Mn2+ onto previously precipi-tated Mna+-bearing oxide surfaces, forming concentricgrowth bands (Fig. 1) (Crerar and Barnes, 1,974 Creruret al., 1972). Backscattered SEM investigation and mi-croprobe analysis showed that differences in the compo-sition of botryoidal growth bands in PEG-01 suggestsrhythmic variations in the composition of the weatheringsolutions (Fig. l).

Samples F226-121, F40-142, and Fl l5-139.8 (Tablel) came from laterites developed over the Igarap6 Bahiahydrothermal Cu-Au deposit in the Caraj6s Mountains,eastern Amazon (Vasconcelos et al., 1993). The samplesoccur as massive manganese oxides replacing weatheredschists. Randomly oriented manganese oxide crystals haveprecipitated in the pore spaces around silicate crystals;after dissolution of the silicate matrix. thev form a net-

Si4+Al3*Fe3*Mn4+Co,*

Cu*Ca2*Na+K+Ba2*S12+P5+o*""O""n

Total

S i4*Al3*Fe3*Mno*Co2+

Total

Cu*Ca2*Na+K +Ba,*Sr'?*

Total

P5*Total cations

Molecular WeightCalculated S.G.

RBRARB/RA

0.06 0.050.52 0.180.02 0.01

55.05 56.44o.25 0.200.00 0.002.80 1.510.11 0 .100.08 0.133.01 3.040.14 0.360.04 0.14o.21 0.20

35.29 34.6534 01 34.3797.58 97.01

B cations0.o2 0.010.15 0.050.00 0.007.56 7.670.03 0.030.00 0.007.76 7.76

A cations0.33 0.18o.02 0.020.03 0.040.58 0.580.01 0.020 00 0.010.97 0.85

0.05 0.058.78 8.66

724.92 720.434.35 4.320.60 0.601 .19 1 .210.50 0.50

VASCONCELOS ET AL.: HVEM AND AEM STUDIES OF CRYPTOMELANE

Fig. 1. (a) Backscattered SEM photomicrograph of borryoi-dal cryptomelane (PEG-01); (b) SEM photomicrograph of acic-ular cryptomelane crystals surrounding cavities created by dis-solution of silicate grains; (c) pseudomorphic (p) texture gradinginto massive (m) texture as cryptomelane fills cavities previouslyoccupied by silicates.

work of crosscutting needles that preserve the originaltexture ofthe host rock (Fig. l). Further precipitation ofmanganese oxides in the void spaces creates massivemanganese oxide blocks (Fig. 1). K, Ba, and Cu contentsin supergene manganese oxides and the absence ofthese

elements from the host schists suggest free mobility ofthese elements in the weathering solutions. Probablesources of these elements in the weathering solutions aremetavolcanic horizons interbedded with the metasedi-ments. The difference in compositions between the Cu-rich samples and the K-rich samples and their distinctdegassing behavior during laser-heating +oAr/3eAr analy-sis (Vasconcelos et al., 1993) made these samples ideallysuited for this study.

ExpnnrrvrnxTAl- METHODS

The four samples were analyzed using an eight-channelARL-SEMQ electron microprobe equipped with wave-length-dispersive spectrometers at the Department of Ge-ology and Geophysics at the University of California,Berkeley. The sample current was 30 nA on MgO, theaccelerating voltage l5 kV, and the beam diameter 3-5pm. PEG-01 was also arralyzed with a l5-prm beam di-ameter, but the results are indistinguishable from thoseobtained with a focused beam.

Powder X-ray diffraction analyses were performed witha Rigaku Geigerflex X-ray diffractometer with CuKa ra-diation over the range 3 < 20 < 90'with scan speeds of1. 0.5. and 0.25/rnin. Measurements of each sample wererepeated three to four times, and the average peak posi-

tions were calibrated against a Si standard under the sameconditions. Unit-cell parameters were calculated using amodified version of the least-squares refinement programLCLS (Burnham, 1962) (Table 2) and confirm all foursamples as cryptomelane. Calculations for a body-cen-tered tetragonal unit cell (space group 14/ m) and a mono-clinic unit cell (space grotry 12/ m) were carried out forcomparlson.

After petrographic observations and electron micro-probe and scanning electron microscope analysis, suitablethin foils of samples PEG-01 and F226-l2l were pre-pared by ion-beam thinning for study in the heating stageof the Kratos 1.5-MeV HVEM at the National Center forElectron Microscopy (NCEM), Lawrence Berkeley Lab-oratory. The samples were characterized before heatingby a series of selected-area diffraction (SAD) patterns andbright-field photomicrographs. The samples were heatedal a rate of l0 "C/min to a maximum Z of 848 'C andSAD patterns and bright-field photomicrographs were re-corded throughout this process. The temperature withinthe heating stage was monitored by a Chromel-Alumelthermocouple, and calibration procedures indicated thatthe temperature readings were within 5 'C from actualsample temperature.

Two foils of sample F226-l2l and one foil of PEG-01were examined in the JEOL 200-kV analytical electronmicroscope at the NCEM. The AEM has a Kevex EDXdetector and corresponding processing facilities, whichpermitted the variations in composition of the samplesbefore and after heating to be investigated. A berylliumsample holder and grid was employed to obviate spuriousX-rays from the holder that might interfere with the sam-ple analysis. The low energy ofthe 200-kV electrons and

VASCONCELOS ET AL.: HVEM AND AEM STUDIES OF CRYPTOMELANE 83

Trau2. Unit-cell parameters and structural informationof cryp-tomelane

Space group:Sample:

l4lmPEG.O1

l4lm l4lmF40-142 F226-121

a (A)b (A)c(A)o (")p f )? (')v(A")No. of reflectionsSpecific Aravity (calc.)

9.846(1 6)s.84606)2.860(5)

909090

277.24(90)1 84.30

9.814(24)9.814(24)2.872(9)

909090

276.65(1 36)1 2

4.34

9.834(31)9.8i]4(31)2.857(9)

909090

276.28(1 68)1 04.38

/Vote; standard deviations are in parentheses.

the highly condensed beam caused significant damage tosample F226-121, which underwent a phase transfor-mation in the beam. This feature allows mineral trans-formations induced in the sample by electron beam andheating damage to be investigated and compared withphase transformations observed during external heating.Sample PEG-01 was much more stable in the electronbeam, and, in spite of extended investigation, no damagewas observed.

Single grains of sample PEG-01 and a Cu-rich crypto-melane sample (Fl l5- 139.8) from the lgarupe Bahia pro-file were further analyzed by aoArl3eAr analysis (Vascon-celos et al., 1993). Sample preparation and irradiationprocedures, experimental conditions, and analytical re-sults were presented in Vasconcelos et al. (1993).

Rnsur,rsTable I shows mineral compositions and site occupan-

cies calculated from the average results of electron mi-croprobe analyses of each sample (592 points analyzedfor sample PEG-0 I , I 9 8 points for sample F226-l2l , 100points analyzed for F40-142, and 49 points for Fl 15-139.8). The low totals reported cannot be accounted forby unanalyzed elements. These samples were checked byEDX and WDS scans, and no other elements were pres-ent in significant amounts. Previous analyses of crypto-melane by other researchers have indicated that signifi-cant amounts of HrO (1-3.8 wto/o) can be present in thesesupergene oxides (Bystrdm and Bystriim, 1950; Mathie-son and Wadsley, 1950; Perseil and Pinet, 1976; Miuraet al., 1987; Ostwald, 1988). We conclude that H,O ispresent in the cryptomelane samples analyzed.

The cell occupancy calculations shown in Table I sug-gest that the B site is only partially occupied. In addition,charge-balance considerations (based on l6 O atoms) andthe measured O contents suggest that all Mn cations pres-ent in the B site are in the Mna+ oxidation state. Previousauthors (Bystriim and Bystrdm, 1950; Burns and Burns,1979; Miura et al., 1987) have suggested that the chargeof I cations may be compensated by the presence of va-cancies in the B site. In the general formula proposed byBystrom and Bystrdm (1950), A2_"B8-.Xr6, z varies from0. I to 0.5 for the minerals analyzed. Twelve hollandite

1 0

56 57

Wt% Mn

1':t

- a a- a a .

"'i;: ri-.'.'r3:

4 4 4 6 o ' * n o * u o s z 5 4

Fig.2. Elemental correlation from electron microprobe anal-yses of sample PEG-O1.

and 36 cryptomelane samples analyzed by Miura et al.(1987) also showed vacancies in the B site, with z rangingfrom 0.300 to 0.153. These vacancies could allow forcharge compensation of the A-site cations and would pre-clude the necessity for the presence ofa lower oxidationstate of Mn in the B site.

The Mn contents of the samples analyzed should alsodecrease with increasing Si, Al, and Fe competition forthe octahedral site. Figures 2-4 illustrate the inverse cor-relation between Mn and Fe + Si * Al contents in thecryptomelane samples analyzed, suggesting competitionbetween Mn and these elements for the B site. Similarly,competition for the tunnel sites may account for the in-verse relationship between K and Cu (Fig. 3) and K +Ba * Na * Ca and Cu (Fig. 4). Toledo-Groke et al.(1990), studying Cu-rich cryptomelane from the Saloboprofile in Carajis, also noticed evidence for competitionbetween Cu and K for the A sites. Although it is yet notcertain whether Cu occupies the tunnel sites or substitutesfor Mn in the B sites, the elemental correlations and thebehavior of Cu during heating of the Cu-rich cryptome-

0)l!+

o " "+as o u=

o 4

6aos 0 6

=0 4

PEG-01 ELECTRON MICROPROBE ANALYSES

a a

a.a

a'a

a o o

a ao

a a

l f t o

F226.121 ELECTRON MICROPROBE ANALYSES

t . l .a a

,O t t a

'.ta.

84 VASCONCELOS ET AL.: HVEM AND AEM STUDIES OF CRYPTOMELANE

1 2

F4O.142 ELECTRON MICROPROBE ANALYSES

t.ta 1.tf.a a

.-.'-. t"l$.r6!r

1 0ol!+

.tos

0 6

=

€)ttJ 3 0

.!ah

S 2 0

=

4 A

1 0

0 05 4 0 5 4 5 5 5 5 5 6 0

Wt% Mn5 7 0 5 7 5

Wt% Mn

G i 6

o+6.D+ - "(Ez+Y 4 0

s= 3 5

Y

I5

'.:'* r .r o a a

a aa

a aa a a a

aa

1 5

" * r3 r i " u t ' 34 36 38

Fig. 3. Elemental correlation from electron microprobe anal-yses of sample F226-121.

lane samples described below are suggestive of Cu occu-pation of the tunnel sites. Structural considerations in-dicate that Cu, if present in the A site, must occur as Cu+(r,: 0.96 A1 and not as the smaller Cu2* ion (r, : 0.72 L).

The cell-refinement calculations (Table 2) indicate thatthe cryptomelane samples analyzed are either tetragonalor monoclinic, with B nearly 90'. The powder patternresults refined assuming tetragonal symmetry Q4/m) (Ta-ble 2) yield lower residuals than when fitted assuming amonoclinic symmetry Q2/m). On the other hand, thepresence ofdoublets and triplets in the powder pattern ofsome samples is suggestive of a lower symmetry than 14/m. The resolution of the powder diffractometer is insuf-ficient to positively identify the space group. Another cri-terion for distinguishing tetragonal from monocliniccryptomelane is the ratio of the average ionic radius ofB-site to A-site occupants (Post et al., 1982; Vicat et al.,1986). When this ratio is >0.48, the structure should bemonoclinic. The ratio of average ionic radius of B cationsand I cations in the samples studied (Table 1) suggeststhat sample PEG-01 should be tetragonal and the Cu-rich cryptomelane should be monoclinic.

o l

,.1.1..i....a

a

a

. J'.lr ]" +"3 ""3 0

2 5t o * , * a r t u 1 8 2 0

Fig. 4. Elemental correlation from electron microprobe anal-yses of sample F40-142.

Bright-field images and SAD patterns for sample PEG-0l are shown in Figure 5. PEG-01 crystals are prismaticand elongated along [001] (Fie. 5). Prism diameters varyfrom 50 to 1000 nm, and their lengths range from 500 to10000 nm; these prisms display square and diamond-shaped cross sections (Fig. 5, arrows). Because ofthe fine-grained nature of the sample, only ring patterns wereobtainable by SAD. In order to obtain single-crystal in-formation, convergent beam electron diffraction (CBED)patterns were necessary (Fig. 5). The CBED patterns forPEG-01 crystals indicate that the mineral is possibly te-tragonal and elongated along the c axis. The CBED pat-tern in Figure 5 shows the a*c* plane of tetragonal cryp-tomelane or the a*b* plane of monoclinic cryptomelane(Fig. 5). It was possible to obtain lattice fringes (Fig. 5).The measured d value of 7.1 A corresponds to the (l l0)planes of cryptomelane (Fie. 5). In addition to the 7.1-Alattice fringes of cryptomelane, some l0-A httice fringeswere also present. Similar patterns were observed by Tur-ner and Buseck (1979), who interpreted the l0-A dis-tances as romanechite intergrown with the 7-A holan-dite. Chen et al. (1986) also observed similar lattice

tq,

tYq

VASCONCELOS ET AL.: HVEM AND AEM STUDIES OF CRYPTOMELANE

- - r oA

85

o o o o o

o o o o o

Fig. 5. (a) BF image of crosscutting network of cryptomelanecrystals in PEG-01. Elongated prismatic and rod-shaped crystalsare typical habits for this cryptomelane sample. Arrows showthe cross sectional shape ofa crystal oriented perpendicularly tothe plane of the photograph; (b) SAD ring pattern of cryptome-lane (see Table 4 for indexed d values for ring pattern); (c) mi-crobeam diffraction pattem of tetragonal cryptomelane crystalsalong the b axis; (d) indexing of SAD pattern shown in c; (e)TEM image of properly oriented cryptomelane crystals showingpredominantly well-ordered 7.1-A fringes [(110) planes of cryp-tomelanel and a few l0-A fringes [possibly (100) planes ofpsi-lomelane or (001) planes of todorokitel.

irregularities in synthetic cryptomelane crystals formedepitaxially afler birnessite.

Bright-field images and SAD ring patterns of sampleF226-l2l before and after phase transformations areshown in Figure 6. Diffuse SAD ring patterns for sampleF226-l2l (Fig. 6) indicate that single crystals of crypto-melane are much smaller than the diameter of the elec-tron beam (- I pm). These cryptomelane crystals areacicular and have diameters ranging from l0 to 120 nm;their lengths may reach 1000 nm.

Mineralogical transformations during heating andelectron beam damage

Indexing of SAD patterns taken for the two samplesduring the heating experiment (Fig. 7 and Table 3 forsample F226-l2l and Figure 8 and Table 3 for samplePEG-01) indicated that the two samples exhibit slightlydifferent behaviors during the HVEM heating experi-ments. The fibrous and prismatic cryptomelane crystalsin the K-rich sample (PEG-01) were surprisingly resistantto phase transformations, persisting in vacuum up to Z: 648"C (Fig. 8 and Table 3). The Cu-rich sample (F226-

Fig. 6. (a) BF image of crosscutting network of cryptomelaneneedles in sample F226-l2l before heat-induced phase transfor-mation; (b) SAD ring patterns of cryptomelane crystals; (c) BFimage of granoblastic texture formed by intergrown hausman-nite, manganosite, and dispersed native copper in the sampleafter heat-induced phase transformation; (d) SAD ring patternof newly formed phases after heat treatment.

l2l) (Fig. 7 and Table 3) underwent phase transforma-tions at lower temperatures, as indicated by the almostcomplete disappearance of the cryptomelane lines in theSAD pattern for the sample aI T : 620 "C.

Both samples underwent sudden phase transforma-tions during in-situ heating investigations, but not all

45'cut ""

T

8 6 VASCONCELOS ET AL.: HVEM AND AEM STUDIES OF CRYPTOMELANE

T = 1 4 0 . C T = 3 3 0 ' C T = 5 1 5 ' C

Fig. 7. (a-h) SAD ring patterns depicting mineral transformations in sample F226-l2l during in-situ HVEM heating. Mineralparageneses are summarized in Table 3. Temperature and some diagnostic reflections are indicated. C : cryptomelane; H :hausmannite; M : manganosite. Several diffraction patterns were taken at the same temperature (d-f and g-h). Note that at 620"C hausmannite disappears during prolonged heating and manganosite becomes the dominating phase.

crystals underwent transformation at the same time. Thedifference in behavior between the two samples and thedifferent behavior of cryptomelane crystals in a singlesample suggest that small differences in composition be-

tween cryptomelane crystals possibly accelerate or retardthe transition from low- to high-temperature phases. Inaddition, other factors, such as grain size, crystal mor-phology and orientation, surface properties, and kinetic

620 'C 7 1 5 ' C

Fig. 8. (a-h) SAD ring patterns illustrating mineral transformations in sample PEG-01 during in-situ HVEM heating. Mineralparageneses are shown in Table 3. Temperature and some diagnostic reflections are indicated. C : cryptomelane; H : hausmannite;M : manganosite. At 648 'C (c) hausmannite crystallizes and later transforms to manganosite (e). Notice coarsening at 848 'C ascontinuous rings (g) change to a spot pattern (h).

VASCONCELOS ET AL.: HVEM AND AEM STUDIES OF CRYPTOMELANE

Tlale 3. Mineral paragenesis during heating in HVEM

Sample F226-1 21140 .C 330 "C s15 .C 620 .C 620 "C 620 .C 715 .C 715 "C

8 7

\ ,.wic(+) c(+) B(-) H(+)

H(+) M( - )^

H(+) M(+) M(+) M(+)M H ( - ) H ( - ) H ( - )

Sample PEG-0160.c 490.c 648.C 648.C 648"C 781"C 848.C

c(+) c(+)H

H(+) H(+)C MM C

HM

M(+)H ( - )

Note; C : cryptomelane, B : bixbyite, H : hausmannite, M : man-ganosite; + : abundant, - : minor.

constraints, may also retard the phase transitions ob-served and must be evaluated.

During heating with a heating stage, the Cu-rich cryp-tomelane sample transformed to hausmannite and nativecopper. Hausmannite crystals are coarser grained thanthe starting cryptomelane needles (Fig. 6b). As the phasetransformations progressed, the crosscutting network ofacicular cryptomelane crystals (Fig. 6a) was progressivelyreplaced by a granoblastic texture (Fig. 6b), in whichcoarser grained idiomorphic crystals of hausmannite andmanganosite are joined at l20 angles and native coppergrains, <0.005 pm, occur dispersed throughout the sam-ple. Textural evolution of the samples during heating sug-gests that crystals oriented perpendicularly to the samplesurface recrystallize more readily than the crystals ori-ented parallel to it (Fig. 6, arrow). This observation isconsistent with recrystallization by breakdown of the tun-nel structure due to migration of I cations out of thetunnel sites. Electron beam damage produced phasetransformations similar to external heating (Fig. 9b).Hausmannite and native copper suddenly crystallizedduring cryptomelane breakdown (Fig. 9b). It is possiblethat the temperatures reached in the electron beam werehigh enough to promote the phase transformationsobserved.

EDX semiquantitative analysis of sample F226-I2lbefore and after phase transformation by electron beamdamage showed the loss of K and Cu from the manganeseoxide structures as it transformed from cryptomelane tohausmannite. The loss of these elements was accompa-nied by a relative enrichment in Al and Co, whereas Mnremained relatively unaffected. This is interpreted as ev-

---J

Fig. 9. (a) BF image of crosscutting network of cryptomelaneneedles in sample F226-l2l before phase transformation in-duced by electron beam; (b) BF image of granoblastic textureformed by intergrown hausmannite, manganosite, and dispersednative copper in the sample (arrow) after phase transformationinduced by electron beam; (c) BF image of sample F226-l2lshowing granoblastic texture and single-crystal microbeam dif-fraction patterns of the newly formed hausmannite (d) and man-ganosite (e) crystals.

8 8 VASCONCELOS ET AL.: HVEM AND AEM STUDIES OF CRYPTOMELANE

tloooo q n0)rcoo( , 6 0

os 4 0oE€ 2 0

tro

0

1 0 0

a) ,

K-CryptomelaneSample PEG-01Lab # 3363-06

a "/o4oAr* Released

I % t 'Ar Re leased

b)

Cu-CryptomelaneSample F115-139.8

Lab # 4315-01

O o/o noAr' Released

I o/o t'Ar Released

1 . 0 2 0 3 . 0

Laser Power (Watts)

Fig. 10. (a) Release of Ar isotopes during laser-heating o0Ar/3eAr analysis of sample PEG-01 ; (b) Release of Ar isotopes dur-ing laser-heating40Ar/]eAr analysis of sample Fl l5-139.8. Thecurve in a indicates that the K-rich cryptomelane sample releasesits radiogenic aoAr* and nucleogenic seAr at a narrow range oflaser power. The Cu-rich cryptomelane sample (b) also releasesits ooAr* and seAr isotopes at a narrow range of laser power, butless laser energy is required to degas the Cu-rich sample.

idence for the presence of Al and Co in the octahedralsites; the octahedral sites undergo structural rearrange-ment but remain otherwise undisturbed during the phasetransformation. The loss of K and Cu is attributed to thebreakdown ofthe tunnel structure. K and Cu are exsolvedfrom the tunnel site; Cu forms a distinct phase (nativecopper) (Fig. 9b), and K is lost by volatilization. Nativecopper has also been found as a residue together withmanganosite in samples fused for K-Ar analysis and insamples heated in a furnace to 900 'C, further strength-ening this hypothesis.

Hausmannite forms euhedral cube-shaped crystals (Fig.9b, 9c). Indexed SAD ring patterns (Figs. 7, 8, and 9) andmicrobeam single-crystal diffraction pattems (Fig. 9d)suggest a tetragonal crystal lattice for hausmannite. Thenoticeable presence ofa tweed structure (Fig. 9c) in haus-mannite crystals may be attributed to the distortion ofthe crystal structure during the phase transformation froma high-temperature cubic spinel structure into the lowertemperature tetragonal hausmannite structure (space

group 14,) (Frenzel, 1980). Although under normal pres-sure conditions the tetragonal-cubic phase transforma-tion in hausmannite occurs between 1050 and I160'C,it is possible that in vacuum this transformalion mayoccur below 848 "C. Another explanation for the modu-lated structure is an incipient chemical fluctuation. SADring patterns (Figs. 6, 8, 9) and microbeam diffractionpatterns (Fig. 9e) also show the presence of manganositecrystals after the heating of cryptomelane crystals invacuum.

Laser-heating q Ar /3e Ar results

The Ar release history of the K-rich and the Cu-richcryptomelane samples are shown in Figure 10. The de-gassing history of the two samples is remarkably similar;most of the radiogenic 40Art< and nucleogenic 3eAr frac-tions are released at a narrow range of laser power. De-spite the similarities in the Ar degassing patterns for thetwo samples, the K-rich sample requires a higher laserenergy output to release its Ar gas. This difference in de-gassing history may be accounted for by different cou-pling between each sample and the laser beam. Anotherpossible explanation is that the K-rich sample undergoesphase transformation at a higher temperature than theCu-rich sample. The laser-heating ooAr/3eAr experimentalso shows a close similarity between the o0Ar* and 3eAr

release curves (Fig. I 0), which suggests that these isotopeshave the same activation energy, i.e., they occupy thesame site in the crystal structure. This is indirect yet strongevidence that significant amounts of 3eAr are not dis-placed by recoil out ofthe tunnel sites.

DrscussroN

The investigation ofsupergene manganese oxides at theappropriate scale and with the proper methodology sug-gests that the oxides we studied are, on the whole, well-crystallized single minerals with only minor occurrencesof other intergrown phases. The high degree of crystallin-ity and the purity of these minerals, in spite of their ex-tremely fine-grained nature, suggest that they should besuitable for radiometric dating, provided that aoAr and Kdo not easily diffuse into and out of the mineral structuresand that 3eAr recoil loss does not occur.

The diffusion of K from cryptomelane structures wasstudied by electrodialysis (Sreenivas and Roy, 196l).These authors proposed that the large tunnel structure, Ication disordering, and partial occupancy ofthe channelsites led to low activation energies for the I cations, re-sulting in a high cation-exchange capacily for these min-erals. High cation exchange would also imply loss of Kand radiogenic aoAr*, leading to inconclusive dating re-sults. On the other hand, Vicat et al. (1986) interpretedthe presence of a superstructure along the c axis in syn-thetic barium titanium hollandite as evidence oforderingofthe tunnel A cations. This superstructure involves thetripling of the cell along the c axis, which Vicat et al.(1986) interpreted as one-dimensional ordering ofK cat-ions in the partially filled A sites, following the sequence

EoooO o ^

oG66o o o

o

s 4 0o.z6a z vE

IC)

VASCONCELOS ET AL.: HVEM AND AEM STUDIES OF CRYPTOMELANE 89

K-K-n-K-K-tr. Ordering of thel cations implies restrict-ed mobility of the cations along the tunnel direction, animportant feature for the K and Ar retentivity of hollan-dite-group minerals. Restricted mobility of I cations hasalso been proposed by Sinclair et al. (1980), who noticedthe presence ofsquare-shaped bottlenecks along the chan-nels. These bottlenecks restrict cation motion and explainthe high resistance to leaching presented by hollandite-type structures (Sinclair et al., 1980). Cation mobility alongthe tunnel structure must be a function of the cation ra-dius, and one should expect greater mobility and loweractivation energy for smaller cations occupying the A sites.Some authors (Sinclair et al., 1980; Post et al., 1982; Vi-cat et al., 1986) have shown that some of the A cationsare displaced from the special position in cryptomelane.Vicat et al. (1986) proposed that K is shifted 0.125 x caway from the 0072 position. Post et al. (1982) have shownthat the smaller cations (Na and Sr) form shorter A-Ocontacts and will be displaced from the (000) positionoccupied by K. Structural energy calculations also showthat smaller tunnel cations are displaced farther from thespecial positions than larger cations (Post and Burnham,1986). The model calculations also indicate that the en-ergy minimum calculated for smaller cations is broaderand shallower than for larger cations, suggesting that thesmaller cations have larger temperature factors (Post andBurnham, 1986). The in-situ mineral transformations ob-served above suggest that K-rich cryptomelane are morethermally stable than Cu-rich cryptomelane, consistentwith the prediction ofa larger activation energy for largercalions occupying the A site.

Ar diffusion from mineral structures is controlled bythe energy required to distort neighboring atoms and toallow passage of Ar (McDougall and Harrison, 1988). Thelarge atomic radius of Ar (R : 1.96 A; imposes structuralconstraints in the tunnel site that impede the migrationof Ar out of these sites unless the other A-site occupants(K, Ba, Na, Cu, etc.) are also dislocated. This process willpromote the collapse of the hollandite tunnel structureand will favor the formation of bixbyite, hausmannite,and manganosite. The heating-stage EM experimentssuggest that K loss and possibly Ar degassing from thefine-grained network of cryptomelane crystals occur bythe breakdown ofindividual needles or fibers. This pos-tulated Ar degassing mechanism differs from volume dif-fusion models postulated for Ar loss from other mineralstructures (McDougall and Harrison, 1988) and is similarto the mechanisms proposed for Ar release from biotiteand amphibole by Gaber et al. (1988).

The laser-heating experiments have shown that, in spiteof a similar behavior between the K-rich and the Cu-richcryptomelane samples, these samples release their tunnel-site Ar components at different laser power levels, whichmost probably indicates degassing at different tempera-tures. These results are consistent with the observationthat samples richer in small cations in the tunnel sites(Cu-rich samples) undergo phase transformations morereadily than samples rich in large cations in the A site.

CoNcr-usroNs

The range of crystal sizes observed by electron mlcros-copy indicates that the supergene manganese oxides stud-ied, even the well-crystallized ones, are very fine grained(crystal diameters are generally smaller than I pm). In-situ observation of mineral transformations during heat-ing of supergene manganese oxides suggests that thesephases have a broad range of thermal stability, even invacuum. The large range of thermal stability of crypto-melane in vacuum suggests that the mobility of the Ication is fairly restricted and requires a large activationenergy. The refractory nature of the cryptomelane crystalsduring in-situ heating-stage experiments also suggests thatcryptomelane should not be affected by the temperaturesreached inside the nuclear reactor during neutron irradi-at ion (Z - 180'C).

We postulate that the temperature required for themineral transformations observed in the TEM, crypto-melane + hausmannite -' manganosite, correspond tothe temperatures in which radiogenic and nucleogenic Arisotopes are released from the mineral structures. Thisindicates that Ar release by hollandite-group minerals isnot a volume diffusion phenomena. The identical degas-sing patterns for radiogenic o0Ar* and nucleogenic seAr

observed in the laser-heating 4oAr/3eAr experiments sug-gest that these isotopes occupy the same site in the crystallattice and that 3eAr does not recoil out ofthe tunnel sitesduring neutron irradiation in any significant quantities.The very fine-grained nature of the samples studied, onthe other hand, suggests that 3eAr recoil is theoreticallypossible and should be monitored in future experiments.

AcxNowr-rncMENTS

We acknowledge the support ofthe Brazilian Research Council (CNPq)

to P.V. (grant 2O0392/88-3/GL) and ofthe National Science Foundationto R.w. (gant EAR-91-04605). We also are thankful for the access to the

facilities of the NCEM and the help with the HVEM provided by Doug

Owen. We gratefully acknowledge John Donovan for helping with the

electron microprobe analysis, Tim Teage for sample preparation, and Ron

Wilson for helping with the SEM studies. This project has also benefited

from discussions with Tim Becker, Garniss Curtis, and Paul Renne. We

also thank the garimpeiro Orilio Vaz de Aguilar from Divino das lar-

anjeiras for help with the collection of manganese oxide samples frompegmatites, and the staffof CVRD mining company, in particular Augusto

Kishida, Jose Luzimar do Rego, and Paulo Seryio, for their help and field

support. The final manuscript also benefited from insightful comments

and suggestions by J Post and D. Veblen.

RnrpnnNcns cITED

Burnham, C.W. (1962) httice constant refinement. Carnegie Institute ofWashington Year Book, 61,132-135.

Burns, R.G, and Burns, v M. (1979) Manganese oxides. MineralogicalSociety of America Reviews in Mineralogy, 6, l-46.

Bystr6m, A., and Bystrdm, A.M. (1950) The crystal structure of hollan-dite, the related manganese oxide minerals, and a-MnOr. Acta Crys-tallogaphica, 3, 146-154.

Chen, C.C., Golden, D.C., and Dixon, J.B. (1986) Transformation ofsynthetic birnessite to cryptomelane: An electron microscope study.Clays and Clay Minerals, 34, 565-57 |

Crerar, D.A., and Barnes, H.L. (1974) Deposition of deep-sea manganesenodules. Geochimica et Cosmochimica Acta, 38, 279-300.

Crerar, D.A., Cormick, R.K., and Barnes, H.L. (1972) Organic controls

90 VASCONCELOS ET AL.: HVEM AND AEM STUDIES OF CRYPTOMELANE

on the sedimentary geochemistry of manganese Acta Mineralogica-Petrographica, 20, 217 -226.

Faulring, G.M., Zwicker, W.K., and Forgeng, W.D. (1960) Thermal trans-formations and properties of cryptomelane. American Mineralogist,45,946-959.

Frenzel, G. (1980) The manganese ore minerals. In I.M. Varentsov andG. Grasselly, Eds., Geology and geochemistry of manganese, p. 25-I 58. E. Schweizerbart'sche Verlagsbuchhandlung, Stuttgart, Germany.

Gaber, L.J., Foland, KA., and Corbat6, C.E. (1988) On the significanceofargon release from biotite and amphibole during noAr/rrAr vacuumheating. Geochimica et Cosmochimica Acta, 52,2457-2465.

Hewett, D.F., and Fleischer, M (l 960) Deposits of the manganese oxides.Economic Geology, 55, l-55.

Kudo, H , Miura, H., and Haiya,Y. (1990) Tetragonal-monoclinic trans-formation of cryptomelane at high temperature. Mineralogical Journal,l 5, 50-63.

Kulp, J.L., and Perfetti, J.N. (1950) Thermal study of some manganeseoxide minerals. Mineralogical Magazine, 29, 239-25 l.

Ljunggxen, P. (1955) Differential thermal analysis and x-ray examinationof Fe and Mn bog ores. Geologiska Fiireningens iStockholm Ftirhand-lingar,77 /2, 135-147 .

Mathieson, A.M., and Wadsley, A.D (1950) The crystal structure of cryp-tomelane. American Mineralogist, 35, 99-101.

McDougall, I., and Harrison, T.M. (1988) Geochronology and thermo-chronology by the aoArlr,Ar method: Oxford monographs on geologyand geophysics, 212 p. Oxford University Press, New York.

Miura, H., Banerjee, H., Hariya, Y., and Dasgupra, S. (1987) Hotlanditeand cryptomelane in the manganese oxide deposits ofthe Sausar Group,India. Mineralogical Journal, 13, 424-433.

Ostwald, J. (1988) Mineralogy ofthe Groote Eylandt manganese oxides:A review. Ore Geology Review, 4, 3-45.

Perseil, EA., and Pinet, M. (1976) Contribution i la connaissance desroman6chites et des cryptomelanes-coronadites-hollandites: Traits es-sentiels et paragendses. Contributions to Mineralogy and Petrology, 55,19r-204.

Post, J.E., and Burnham, C.W. (1986) Modeling tunnel-cation displace-ments in hollandites using structure-energy calculations. AmericanMineralogist, 7 I, I 178-l 185.

Post, J.E., and Veblen, D.R. (1990) Crystal structure determinations ofsynthetic sodium, magnesium, and potassium birnessite using TEMand the Rietveld method. American Mineralogist, 75,477-489.

Post, J.E., Von Dreele, R.B., and Buseck, P.R. (1982) Symmetry andcation displacements in hollandites: Stnrcture refinements of hollan-dite, cryptomelane and priderite. Acta Crystallographica, 838, 1056-1065 .

Roy, S., and Purkait, P.K. (1965) Stability relations ofmanganese oxideminerals in metamorphic orebodies corresponding to sillimanite gradein Gowari Wadhona Mine Area, Chindwara District, Madhya Pradesh,India Economic Geology, 60, 601-613.

Sinclair, W., Mclaughlin, G.M., and Ringwood, A.E. (1980) The struc-ture and chemistry of a barium titanate hollandite-type phase. ActaCrystallographica, B36, 29 13-29 18.

Sreenivas, B.L., and Roy, R. (1961) Observations on cation exchange insome manganese minerals by electrodialysis. Economic Geology, 56,r 98-203.

Toledo-Groke, M.C., Melfi, A.J, and Parisot, J.C. (1990) Criptomelanacuprifera formada durante o intemperismo das rochas da jazida decobre do Salobo 3a, Serra dos Carajiis. Revista do Instituto Geol6gico,t t , 3s -42 .

Turner, S, and Buseck, P.R. (1979) Manganese oxide structures and theirintergrowths. Science, 203, lO24-1O27.

Van Hook, H.J., and Keith, M.L. (1958) The system FerOo-MnrOo.American Mineralogist, 43, 69-83.

Vasconcelos, P M , Becker, T.A., Renne, P.R., and Brimhan, G.H. (1992)Age and duration of weathering by aoK-aoAr and 4oAr-3'qAr analysis ofK-Mn oxides. Science. 258.451.

-(1993) Direct dating of weathering phenomena by ro11-r04. und4oAr-reAr analysis of supergene K-Mn oxides. Geochimica et Cosmo-chimica Acta, in press.

Vicat, J., Fanchon, E., Strobel, P., and Qui, D.T. (1986) The structure ofK, ,.MnrO,u and cation ordering in hollandite-type structures. ActaCrystallographica, B'42, | 62- | 67 .

M,rluscnrrr RECETvED Fennuenv 4, 1993Me.Nuscnrrr AccEprED Sep,rpvsen 20, 1993