Introduction

Metal aluminum borates represent a fascinatingclass of materials. Depending on the composition,they can be useful as catalytic, laser, or nonlinearoptical materials. In the early stages of research anddevelopment, it can often be difficult to grow thesingle crystals “necessary” for detailed structuralcharacterization. Because knowledge of the crystalstructures of such materials is necessary for thedevelopment of structure/property correlations, weare sometimes forced to obtain detailed structuralinformation using only powder diffraction data. Inaddition, real catalysts are always polycrystalline,“forcing” the use of powder diffraction techniques tostudy such materials. Recent advances in hardwareand software, especially the availability of synchro-tron radiation, make it possible to answer manystructural questions using only powder diffractiondata. In this paper, we provide examples of ab initiostructure solution, and the use of powder diffractiontechniques to obtain information about the non-structural properties of a material.

POTASSIUM ALUMINUM BORATE,K2Al2B2O7

Only one ternary phase, K3AlB8O15 [1] in theK2O-Al2O3-B2O3 system has been characterizedstructurally. During exploration of this ternarysystem, we discovered a black amorphous semi-conducting phase of the approximate composition1K20:1Al2O3: 2B2O3. From preparations havingcompositions near lK20:1Al2O3:lB2O3, a phasehaving the X-ray powder diffraction pattern ofK2Al2B2O7 [2] was produced.

The powder pattern of K2Al2B2O7 can be indexedon a very high quality trigonal/hexagonal unit cellhaving a=8.5597(4) and c=8.4597(4) D. Magic anglespinning NMR indicated that the Al coordination istetrahedral, and that the borons are trigonal. Theobserved density indicated that the cell contents wereK6Al6B6O21. Because K2Al2B2O7 is formed near thesemiconducting phase in the phase diagram, webelieved that knowledge of the crystal structure, andcomparison of the structure to those of nearbycrystalline phases, would give some insight into the

17 The Rigaku Journal

The Rigaku Journal

Vol. 16/ number 2/ 1999

CONTRIBUTED PAPERS

CRYSTAL STRUCTURES OF METAL ALUMINUM BORATES

JAMES A. KADUK, LARRY C. SATEK, AND STEPHEN T. McKENNA

BP Amoco p.l.c., PO. Box 3011 MC F-9, 150 W Warrenville Rd., Naperville IL 60566, U.S.A.

The crystal structures of K2Al2B2O7, and SrAl2B2O7 have been solved ab initio by applying'traditional' single crystal techniques to structure factors extracted from X-ray powder patterns. K2Al2B2O7

crystallizes in the trigonal space group P321, with a=8.55802(2), c 8.45576(3) D, V=536.328 D3, and Z=3.The structure consists of a 3-dimensional network composed of corner-sharing BO3 triangles and Al2O7

units. The BO3 groups lie approximately parallel to the ab plane, and the planes containing them are joinedby Al2O7 “pillars” parallel to c. The K cations reside in channels parallel to c. SrAl2B2O7 crystallizes in therhombohedral space group R32, with a=4.90363(9), c 23.9346(6) D, V=498.416 D3, and Z=3. In thisstructure also, Al2O7 act as “pillars' between planes containing BO3 units. The structure is, however,layered, as the Al2O7 join pairs of BO3 planes. An ABC stacking sequence of these double layers createstrigonal prismatic cavities, in which the Sr reside.

The average structure of Cu2Al6B4O17 has been confirmed to higher accuracy and precision by aresonant scattering study. This compound crystallizes in the tetragonal space group I4/m, witha=10.57945(1), c=5.67357(6) D, V=635.013 D3, and Z=2. A new model for the local structure is proposed.Cu2+ and Al3+ equally occupy a trigonal bipyramidal site. The Cu occur as “cis" pairs around a “4-ring” ofthese 5-coordinate sites. A square planar oxygen is displaced from the center of this “4-ring” toward thepair of Al and away from the pair of Cu cations. The Cu2+ and Al3+ apparently occupy different locations inthe coordination sphere at low temperatures, only moving to a common site as the calcination temperatureis raised. Both the crystallinity and average crystallite size of this material can be controlled by controllingthe calcination temperature.

structure of the amorphous phase and the mechanismsof conductivity.

Experimental

The sample was mixed with 4.87wt% NIST 640bSi internal standard, and ground in a McCronemicronising mill using ethanol as the milling liquid.The X-ray powder diffraction pattern was measuredon beamline X3Bl at the National Synchrotron LightSource at Brookhaven National Laboratory, using awavelength of 0.89978 D. The pattern was measuredfrom 6-74E 22 in 0.004E steps, counting for 4sec/step. The sample was rocked from T-1 to T+lEduring each data point. An T-scan of the (112) peak at17.206E 22 indicated significant granularity. Thedegree of powder averaging obtained by a 2E rockingwould not be expected to affect solution of thestructure, but may affect the agreement of theobserved and calculated patterns.

No systematic absences were observed in thepowder pattern. Together with the trigonal/ hexagonal unit cell (no evidence was observed for lower,especially orthorhombic, symmetry), this “limits” thepossible space groups to: P3, P3, P321, P3ml, P3ml,P312, P31m, P3lm, P6, P6, P61m, P622, P6mm, P62m,P6m2 and P6/mmm. In all of the Laue classes but6/mmm, there are inequivalent reflections whichoverlap exactly. This overlap makes the extraction ofaccurate structure factors from the powder patternmore difficult. Many reflections, which would beabsent for a rhombohedral cell, are absent or veryweak in the observed pattern. These weak reflectionssuggested the likely presence of pseudosymmetry, and that the structure solution would be difficult.

Consideration of the coordination requirements of the cations and the approximate number of atoms inthe unit cell suggested that the presence of a 6-foldrotation axis was unlikely. Attempts to solve thestructure from laboratory powder data in P62m andP6m2 suggested a hexagonal array of heavy atoms at000, 002, ab0, ab2, ba0, and ba2. Acalculated pattern with K atoms at these sites hadmany features in common with the observed pattern,and suggested that there was some truth to this model.All attempts to locate more atoms in 6/mmm spacegroups were unsuccessful.

The strategy followed, therefore, was to try tosolve and refine the structure in the lowest possiblesymmetry space group, P3, and to seek additionalsymmetry elements to define the true space group.

Initial powder data processing was carried out usingGSAS [3]. A 3-term cosine Fourier series background function, scale factors for each of the four phases, thelattice parameters, and profile terms were refined.The structure model used the six K atoms at the abovecoordinates. The LeBail extraction procedure incor-porated into GSAS was used to extract 426 individualstructure factors for 22 < 56E (d > 0.96 D).

These structure factors were used to create aSHELXTL Plus [4] data file. Attempts to solve thestructure using direct methods were unsuccessful. APatterson synthesis suggested the presence of a Katom at the origin, and a second K at a, b, 0.5148.The z coordinate of the first atom was fixed to definethe origin. Successive cycles of refinement anddifference Fourier synthesis yielded the positions ofsome other reasonable atoms, but not the completestructure. The second-best heavy atom positionsuggested by the Patterson was aa0 -- a generalposition in P3. Successive cycles of least squaresrefinement and difference Fourier maps yieldedanother K on a general position, and six Al positionson the 3-fold axes. Further refinement of this modelgradually yielded the positions of the O and B atoms.The B and Al were distinguished by the coordinationnumber and the M-O distances.

Final refinement of the structure was carried outusing GSAS. The atom coordinates were refinedsubject to soft constraints of 1.74(l) D on the Al-Obonds and 1.37(l) D on the B-O bonds. The weight ofthese constraints was gradually decreased through the course of refinement. The atoms were refined with acommon isotropic displacement coefficient for eachatom type. Included in the final refinements werescale factors for each phase, and the lattice parameters for K2Al2B2O7 and an Al2O3 impurity. The sizebroadening terms X and ptec were refined forK2Al2B2O7 The profiles for the other phases weredescribed using fixed X and Y terms determined early in the refinement. The background indicated thepresence of an amorphous component, and was de-scribed by a 6-term real space pair correlationfunction, using two fixed distances of 2.80 and 0.50D.

The final refinement in P3 of 56 variables using17029 observations yielded the residuals Rp=0.1467and wRp=0.1821. The final reduced P2 was 6.388.The Bragg R(F) was 0.1489. At this point, theasymmetric unit contained two independent K, six Alon 3-fold axes, two B on general positions, and nine O

Vol. 16 No. 2 1999 18

(six on general positions, and three on 3-fold axes).One of the boron atoms was decidedly nonplanar (thesum of the O-B-O angles was 330E). The Al-O-Alangles of the Al2O7 units were 180E; this large angle is uncommon (but not unprecedented), and larger thanthe usual 130-160E Al-O-Al angle. Although theframework topology was almost certainly correct,these structural features, and the difficulty ofconvergence, pointed to a space group error.

The structure was examined for the presence ofadditional symmetry elements by application of theMISSYM program of the NRCVAX software system[5]. A number of additional symmetry elements werefound to relate the K, Al, and B atoms, but when thedefault tolerances were applied to all atoms, noadditional symmetry was detected. When thetolerances were increased slightly, additional 2-foldaxes parallel to [010], [100], and [110] weresuggested. Space group P321 (#150) thereforeseemed likely. The 2-fold axes intersected at ab0. Acoordinate transformation was therefore applied to the P3 atom coordinates, and the unique atoms in P321retained. The new asymmetric unit contained two Katoms on 2-fold axes, three Al on 3-fold axes, one B ata general position, three O at general positions, one Oon a 3-fold axis, and one O at the origin, a position of32 symmetry. Other coordinate transformations leadto wrong combinations of atoms on special positions,which led to the wrong stoichiometry.

Refinement using common isotropic displace-ment coefficients by atom type proceeded smoothly.The Al-O-Al are still linear in this model. Theconstraints on Uiso for O4 and O5 were removed, totest the possibility of disorder off the 3-fold axes. TheUiso of O4 essentially did not change, but that of O5increased greatly. (The largest hole in the )F map was at this atom-the origin.) O5 was thus moved off theorigin, to a position x00 on a 2-fold axis, and refined.The resulting Uiso is comparable to those of the otheroxygens, and the A13-O5-Al3 angle is 151E-- a moretypical value. There is no evidence that O4 isdisplaced off the 3-fold axis, and thus that the Al1-O4-Al2 angle is not 180E. The B is planar to withinexperimental error. In general, this structure ischemically more satisfactory.

The final refinement in P321 of 39 variablesusing 17014 observations yielded the residuals wRp = 0.1814 and Rp=0.1463. The final reduced P2 was5.800. The Bragg R(F) was 0.1451. The agreement ofobserved and calculated patterns is slightly better,

with fewer variables, in P321, suggesting that it is thecorrect space group. Application of MISSYMsuggested the presence of no additional symmetryelements, even with very large tolerances. Theagreement between Fo and Fc is poorest for thehighest-angle reflections. The observed, calculated,and difference patterns are illustrated in Fig. 1. Thelargest differences occur at K2Al2B2O7 reflections,and probably indicate that a true powder average wasnot achieved. The largest peak in a difference Fouriermap was 2.0 electrons, and the largest hole was -1.6electrons. The slope of the final Wilson plot suggeststhat the standard uncertainties are underestimated by a factor of 2.1. The final structural parameters arereported in Table 1.

Description of the Structure

The crystal structure of K2Al2B2O7 consists of a3-dimensional network composed of corner sharingBO3 triangles and Al207 units. The K cations reside inchannels parallel to the c-axis formed by thisframework. The cation coordination spheres and theatom numbering scheme are illustrated in Fig. 2. Theboron coordination is trigonal and very nearly planar;the sum of the O-B-O angles is 359E. The B atom is+0.05D, and the oxygens are -0.01-0.02E from themean BO3 plane.

There are two independent Al2O7 units. Both ofthem occupy 3-fold axes, and consequently theaverage Al-O-Al angles are 180E. Although thisconfiguration is uncommon in high-quality struc-tures, it is not unprecedented. The Al-O bonddistances and angles fall within the normal ranges.The deviations from ideal tetrahedral geometry aresmall. The configurations of the two Al207 units aredifferent. A convenient measure is the O-Al-Al-O“torsion angle”. For the Al1-Al2 group, this angle is39E, indicating that the tetrahedra are nearly in thestaggered configuration (60E). For the A13-Al3Agroup, the angle is 20E, indicating that this unit iscloser to the eclipsed configuration.

The occurrence of two distinct Al207 units, bothwith imposed 3-fold symmetry, reflects a subtlefeature of space group P321. There are two differenttypes of 3-fold axes in this space group. One type,passing though the origin and occupied by the Al3-Al3A units, is intersected by the 2-fold axes in the abplane. The other type, of which there are 2/cell, passthrough ab0 and aa0, and do not intersect theperpendicular 2-fold axes. These axes are occupied by the Al1-Al2 units. A consequence of the symmetry is

19 The Rigaku Journal

that there are two Al1-Al2 units for every Al3-Al3Aunit. The central oxygens also have different sitesymmetries. The oxygen O4 of the Al1-Al2 unit hassite symmetry 3, but the O5 of the Al3-Al3A unit hasthe higher symmetry 32 at its average position.

There are two independent K ions, each residingon a site of 2-fold symmetry. Considering only K-Odistances < 3.2 D, Kl is 5-coordinate, and thecoordination trigonal bipyramidal. Two O2 and one

O5 make up the equatorial plane, and two O3 are theaxial ligands. Two additional O1 are 3.32 D from Kl.If O5 is located at its average position (the origin), theKl-O5 distance is 3.07 D. Threefold disorder of O5off the origin leads to shorter distances to twodifferent Kl (2.88 D). Because Kl is relativelycoordinatively unsaturated, such distortion seems tobe reasonable to meet the bonding requirements of Kl. It is most reasonable to consider K2 to be 8-

Vol. 16 No. 2 1999 20

Fig. 1. Observed, calculated, and difference diffraction patterns of K2Al2B2O7. The crosses represent the observed datapoints, and the solid line through them the calculated pattern. The difference pattern is plotted a t the same scale as the other patterns. The bottom row of tick marks represents the calculated reflection positions for K2Al2B2O7. Successively higherrows indicate the Si, Al2O3, and K2B4O5(OH)4A2H2O peak positions. The vertical scales have been multiplied by factors of 5and 20 in the two halves of the pattern.

coordinate (two O1, two O2, two O3, and two O4).The geometry is irregular. Two additional O1 are 3.23 D from K2.

A convenient way of assessing the strength andreasonableness of a K-O bond is to calculate the bondvalence [6]. Considering the nearest oxygens, Kl isonly 5-coordinate, and all the bonds are relativelyshort. The sum of these bond valences is 1.09 close tothe expected K valence of +1. The K2-0 bonds are ingeneral longer than those to Kl. The K2 valencederived from the 8 shortest distances is only 0.90, butincreases to 1.00 when the two additional 3.23 D K2-O1 distances are included.

The crystal structure is most easily visualizedwhen viewed down the c-axis (Fig. 3). The BO3

triangles lie approximately parallel to the ab plane,

and planes containing them are joined by Al2O7

“pillars” parallel to c. Each B is connected to two All-AI2 pillars pointing “down” and one A13-Al3A pillarpointing “up” (or vice versa). The K ions lie within 6-ring channels parallel to c. The K alternate in layers.The 7-coordinate Kl lie in the A13-Al3A layers,where the atom density is relatively lower. The 8-coordinate K2 lie in the Al1-Al2 layers, in which there are more atoms. The K do not lie in the centers of thechannels, but are displaced toward the walls.

The structure has the same topology as, butdifferent symmetry than, Rb2Be2Si2O7 [7], whichcrystallizes in the orthorhombic space group P2nn.Normally such an isostructural relationship would bedetected by similarities in the lattice constants or thepowder diffraction patterns. In this case, the 40%difference in the atomic volumes of K and Rb lead tocell volumes which differ by 40%. This difference islarge enough that the relation between these twostructures was not detected with the usual softwaretools. The structure similarity was found because both compounds have the relatively unusual A2B2C2X7

stoichiometry.

STRONTIUM ALUMINUM BORATE,SrAl2B2O7

The interesting catalytic chemistry [8] and optical properties [9] of metal aluminum borates prompted an exploration of the SrO-Al2O3-B2O3 phase diagram. Areported ternary phase in this system, SrAlBO4, has achain structure [10], with corner-sharing AlO4

tetrahedra bridged by BO3 triangles. The ultra-low-expansion composition SrAl2B2O7 has been reported[11] to crystallize in cubic and hexagonal forms, butits structure had not been reported.

21 The Rigaku Journal

Name X Y Z Ui*100

K1 0 -0.3589(5) 0 2.86(5)

K2 0.6889(5) 0 1/2 2.86(5)

Al1 1/3 -1/3 0.3145(9) 2.25(7)

Al2 1/3 -1/3 0.7243(8) 2.25(7)

Al3 0 0 0.1994(7) 2.25(7)

B1 0.3388(14) 0.0070(16) 0.7162(20) 2.33(26)

O1 0.2077(7) 0.0393(9) 0.7252(8) 2.10(8)

O2 0.5090(8) 0.1168(7) 0.7533(8) 2.10(8)

O3 0.2446(10) -0.1994(8) 0.7879(8) 2.10(8)

O4 1/3 -1/3 0.5192(11) 2.13(32)

O5 0.0501(27) 0 0 1.68(85)

Table 1. Crystal structure of K2Al2B2O7. Space group P321,a=8.55802(2), c=8.45576(3) D

Fig. 2. The coordination of the B, Al, and K atoms inK2Al2B2O7, with the aotm numbering scheme. K-Odistances < 3.2 Dare illustrated by thin solid lines, and those > 3.2 D by dashed lines. 50% probability ellipsoids. Onlyone of the three disordered O5 positions is shown.

Fig. 3. A polyhedral rendering of the K2Al2B2O7 structureviewed down the c-axis. The open triangles representthe BO3 units, and the shaded tetrahedra indicate theAlO4 subunits. The K are represented by open circles.

Experimental

A solution of 38.45g Sr(NO3)2 and 22.46g H3BO3

in 153 mL H2O was blended with 226.19 g of PHFalumina hydrosol (8.2% Al2O3). The resulting sol wasgelled using approximately 5 mL aqueous ammonia,and air-dried. The solid was pre-calcined by heating to 673 K at 2 K/min, to 773 K at 1 K/min, holding at 773K for 1 h, heating to 873K at 0.5K/min, holding at873K for 1 h, heating to 973 K at 0.5 K/min, andcooling to ambient conditions. The resulting whiteflakes were heated to 1023 K at 3 K/min, 1213 K at 1K/ min, held for 8 h at 1213 K, cooled to 1023 K at 1K/min, and then cooled to ambient conditions.

A portion of the white solid was mixed with NIST640b silicon internal standard in a Spex 8000mixer/mill. The powder diffraction pattern wasmeasured on a Scintag PAD V diffractometer from 5-120E 22 in 0.022 steps, counting for 18 sec/step. Thepattern could be indexed [12] on a high-quality (figure of merit=108) rhombohedral cell having a = 4.904 and c = 23.936 D. A search of the Crystal DataIdentification File [13] yielded a few hits, but nonehad the proper stoichiometry to serve as a modelstructure. The systematic absences were consistentwith space arouos R3, R-3, R32, R3m, and Rm. MAS

11B and 27Al NMR studies indicated that the Al atomswere tetrahedral and the B atoms trigonal planar, andthat tetrahedral boron and octahedral aluminumimpurities were present. These observations wereconsistent with the presence of a few weak unindexedpeaks in the powder pattern, and a weak amorphousbackground. The density and the initial stoichiometrydetermined the cell contents as Sr3Al6B6O21.

To minimize the effects of surface roughness,microabsorption, and incomplete interception of thebeam at low angles, only the 20-120E portion of thepattern was included in the solution and refinement.Integrated intensities were extracted using the lowestsymmetry, R3, and converted into structure factors.The 173 reflections were used as input to a Pattersonsynthesis [4], which indicated that the Sr atom resided at the origin. Successive cycles of least squares anddifference Fourier synthesis revealed the positions ofthe remaining atoms. Rietveld refinements in R3 were unstable. Application of MISSYM [5] indicated thepresence of 2-fold axes parallel to 100, 010, and 1 10-and thus that the true space group was R32.

All atoms were refined isotropically. Softconstraints of 1.74(l) and 1.38(l) D were applied to the Al-O and B-O bonds. These soft constraints contri-

Vol. 16 No. 2 1999 22

Fig. 4. Observed, calculated, and difference diffraction patterns of SrAl2B2O7. The crossesrepresent the observed data points, and the solid line through them the calculated pattern. Thedifference pattern is plotted at the same scale as the other patterns. The bottom row of tickmarks indicates the calculated reflection positions for SrAl2B2O7, and the top row marks thepeaks of the Si internal standard.

buted 3% to the final reduced P2. Included in therefinements were scale factors for SrAl2B2O7 and Si,and the lattice parameters of SrAl2B2O7. A March-Dollase preferred orientation ratio (unique axis [001])was refined. The profiles were described by pseudo-Voigt functions. For SrAl2B2O7, the Cauchy X, Y,ptec, stec, and asym coefficients were refined; for thesilicon internal standard, only Y and asym wererefined. The background was modeled by a 3-termcosine Fourier series.

The final refinement of 29 variables using 5005observations yielded the residuals wRp=0.1282,Rp=0.0966, P2=7.469, R(F)=0.056, and R(F2) =0.1058. The largest peak in a difference Fourier mapwas 0.98eD-3 , and the largest difference hole was -1.16 eD-3 . A normal probability plot indicated that the standard uncertainties were underestimated by afactor of 2.23. The largest errors in the differencepattern (Fig. 4) are in the description of the shapes ofthe low-angle peaks, and at the impurity peaks. Therefined structural parameters are reported in Table 2.

Description of the Structure

The B atom in this new rhombohedral structure istrigonal planar, and the tetrahedral Al atoms occur asAl2O7 pairs. The Al2O7 is nearly “eclipsed”. The Srcoordination is trigonal prismatic (Fig. 5). In contrastto the network structure of K2Al2B2O7, SrAl2B2O7 has a layered structure (Fig. 6). The BO3 planes areperpendicular to the c-axis, and the Al2O7 act as pillars between pairs of BO3 planes. Both the top and thebottom of the pillars share corners with threeBO3triangles, but the top and bottom BO3 triangles are not eclipsed. This corner-sharing arrangement resultsin triangular 6-ring windows in the top and bottomsurfaces of the layers. The ABC stacking sequence ofthese layers creates the trigonal prismatic cavities inwhich the Sr reside. There are no channels parallel to

c; the stacking sequence and the structure of the layers block any potential channels.

The central O1 atoms of the pillars are 4.9 D apartin the equatorial plane, suggesting that small speciesmight be occluded in the cavities between the pillars.Difference Fourier maps indicate no significantoccupation of these cavities, or significant displace-ment of O1 off the site of 32 symmetry. The observedpowder pattern does not correspond to either thehexagonal or cubic forms reported by MacDowell

23 The Rigaku Journal

Atom X Y Z Uiso,D2

Sr 0 0 0 0.0092(5)

Al 2/3 1/3 -0.0963(1) 0.0068(12)

B 2/3 1/3 0.0686(4) 0.010(4)

O1 2/3 1/3 -1/6 0.0259(27)

O2 0.3896(8) 0.4159(6) -0.0691(2) 0.0264(23)

Table 2. Crystal structure of SrAl2B2O7. Space group R32,a=4.90363(9), c=23.9346(6) D

Fig. 6. The coordination of the B, Al and Sr atoms inSrAl2B2O7, with the atom numbering scheme. 50%probability spheroids.

Fig. 6. A polyhedral rendering of the SrAl2B2O7structure. The BO3 units are indicated by triangles, andthe AlO4 subunits by shaded tetrahedra. The Sr arerepresented by shaded circles.

[11]. This rhombohedral structure apparentlycorresponds to a third polymorph.

A search for structurally-similar compounds hasyielded only K2Pb2Ge2O7 [14], which contains layersof the same topology, with Ge playing the role of Al,and Pb serving as the analog to B. The Pb arepyramidal, but recessed into the layers, rather than atthe top and bottom surfaces. The Ge2O7 units arestaggered, and the K occupy the interlayer regions.

COPPER ALUMINUM BORATE, Cu2Al6B4O17

The unusual copper aluminum borateCu2Al6B4O17 is useful as a dehydrogenation catalyst[8]. The average structure (I4/m, a= 10.586(l),c=5.688(2) D) has been known for some time [151,and has been redetermined recently using singlecrystal techniques [16]. Structure determination hasbeen hampered by the difficulty of preparing homo-geneous materials. Recent advances in sol-gelpreparative chemistry [8] have led to the synthesis ofuniformly-green material, permitting a more-detailedstructural study.

The crystal structure (Fig. 7) is made up of edge-sharing chains of octahedral Al atoms parallel to thetetragonal c-axis. The AlO6 chains are joined in the a-and b-directions by trigonal planar BO3 groups. Thereis a 5-coordinate site, 50% occupied each by Cu and

Al, which shares a face with the AlO6 octahedron.These trigonal bipyramidal sites share equatorialcorners at a square planar oxygen, O1.

Trigonal bipyramidal coordination is relativelyunusual for both Cu2+ and Al3+. We considered itunlikely that the Cu and Al ions would occupy exactly the same position within the coordination sphere, andthat a split site model might be appropriate. Attemptsto refine such a model using laboratory powder datadid not yield improved residuals compared to aunified-site model. To study this site in more detail,we carried out a resonant scattering experiment,exploiting the tunability of synchrotron radiation.

To a first approximation, X-ray scattering is anelastic process, and the atomic scattering factor isindependent of the wavelength of the incidentradiation. When the frequency approaches an atomicabsorption edge, inelastic processes becomesignificant, and lead to a dependence of the scatteringfactor on the wavelength of the incident radiation[17]. Both the real and imaginary parts of theanomalous scattering vary, and their variation can beexploited to alter the contrast between atoms.Simultaneous refinement of one structural modelusing several datasets collected at different wave-lengths near an absorption edge can yield information about the populations of different atoms occupying

Vol. 16 No. 2 1999 24

Fig. 7. A stereo view of the crystal structure of Cu2Al6B4O17. The view is approximately down thetetragonal c-axis. The AlO6 bonds are highlighted.

the same site [18]. In this study, we sought to vary thescattering contrast between Cu and Al, to permit amore detailed understanding of the 5-coordinate site.

Experimental

The material studied was prepared fromCu(C2H3O2)2, Al(C3H7O)3, and B(C3H7O)3 using sol-gel techniques and a slow programmed drying andcalcination. The product was ground in a McCronemicronising mill using ethanol as the milling liquid.

Three powder diffraction patterns were measured:

1. On a Scintag PAD V diffractometer in ourlaboratory, using Cu K, radiation (8=1.540629 and1.544451 D) from 10-140E22 in 0.02E steps, counting for 10 sec/step

2. On beamline X7A at NSLS, using a wavelength of 1.293661 D, from 9-70E22 in 0.02E steps, countingfor 5-8 sec/step

3. On beamline X3A2 at NSLS, using awavelength of 1.37957 D, from 30-75E 22 in 0.01Esteps. This wavelength is very close to the Cu K edge.

The variation in the scattering contrast producedsignificant differences in the relative intensities of thelines of the powder patterns.

All three powder patterns were included in arefinement of the crystal structure using GSAS [3].The single crystal structure model [16] was used asthe initial model. The Cu1/Al1 site was refinedanisotropically. All other atoms were refined withisotropic displacement coefficients. The Uiso of O2,

O3, and O4 were fixed at the average value of thesingle crystal Uiso. The occupancies of Cul, Al1, Al2,and Cu2 were refined subject to the constraint that theoverall Cu :Al ratio was 1:3.

The powder patterns indicated the presence of atrace of CuB2O4 (PDF entry 25-268). This phase wasincluded in the refinement using a fixed structuralmodel [19]. Phase fractions for both phases wererefined, subject to the constraint that the sum of thephase fractions was unity.

Included in the refinement were histogram scalefactors for each pattern, and the lattice parameters ofthe two phases. A March-Dollase preferred orienta-tion ratio was refined for Cu2Al6B4O17 (unique axis[001]). To minimize the effects of surface roughness,microabsorption, and incomplete interception of thebeam, the low-angle portion of each histogram wasexcluded from the refinement.

The peak profiles were described by pseudo-Voigt functions. For the synchrotron data sets, a pureCauchy profile was used, while for the in house data aGaussian component was required. Sample transpar-ency and displacement terms were included for thelaboratory pattern. All three profiles included anisotropic size broadening term. The background ofeach histogram was modeled adequately by a 3-termcosine Fourier,series.

An Fobs Fourier map (using either the multiplepowder data sets or the single crystal data) calculatedin the xy0 plane, which contains the Cul/Al1 site andO1 and O2, exhibited no evidence for a split Cul/Al1site (Fig. 8), but shows clear evidence for multiple O1sites displaced from the origin. Refinement of aunified Cul/All site yielded the lowest residuals, butordered and disordered O1 refinements yielded equi-valent refinements. Non-crystallographic evidencecan be used to distinguish between the models.

The final refinement of 53 variables using 13099observations (Fig. 9) yielded the residuals wRp =0.0409 and Rp=0.0221; the final reduced P2 was1.750. The largest peak in the final difference Fouriermap was 0.80 eD-3 (1.85 D from Cu1/Al1, and 2.03and 2.00 D from two different O4 in the channelsparallel to the c-axis). All of the largest differencepeaks are in the xy0 plane, and are near frameworkatoms. The largest difference hole was -0.67eD-3

(0.90D from the B). The structural parameters fromthe combined refinement are reported in Table 3.

25 The Rigaku Journal

Fig. 8. An Fobs Fourier map of Cu2Al6B4O17, calculatedthrough the xy0 plane, which contains Cu1, O1 and O2. The contours represent electron denisites of 32, 16, 8,4, 2, 1, -1 and -2 eD3. Negative contours are dashed.

Vol. 16 No. 2 1999 26

Fig. 9. The three observed,calculated and difference powderpatterns of Cu2Al6B4O17 used in theresonant scattering study. Thecrosses represent the observed datapoints, and the solid lines throughthem the calculated patterns. Thedifference patterns are plotted at thesame scales as the other patterns.The lower rows of tick marks indicatethe calculated reflection positions forCu2Al6B4O17 and the upper rows thepeaks of the CuB2O4 impurity.

Description of the Structure

The average crystal structure of Cu2Al6B4O17 hasbeen confirmed, at higher precision. There was noevidence in the powder patterns for a different cell orspace group. The octahedral sites reside on centers ofsymmetry, and are occupied completely by Al. TheAl2-O3 and Al2-O4 distances of 1.934(2) and1.943(2) D fall within the normal range of octahedralAl-O distances, but the Al2-O2 distance of 1.822(2) Dis relatively short. The average deviation from theideal octahedral angles is 4.3E. The B3-O3 distance of 1.388(6) D and two B-04 distances of 1.352(4) D fallwithin the expected range. The O-B-0 angles differinsignificantly from the expected 120E.

The trigonal bipryamidal Cul/Al1 site is halfoccupied each by Cu and Al. The axial distances totwo O2 are long and short (1.998(3) and 1.854(3) D).Two of the equatorial distances (to O4) are short(1.872(2) D) and one (to O1) is long (2.038(l) D). Thecentral Cul/All site is displaced 0.24 D from the center of the coordination polyhedron.

The atomic valences, calculated from the sums ofbond valences [61, of the Cu and Al are 2.63 and 2.44,far from the expected values of 2 and 3. The calculated valence of O1 is only 1.54, reflecting the relativelylong bonds. These differences are indications that therefined structure represents an average.

Analysis of 81 Cu 2+O5 coordination sphereslocated in the Inorganic Crystal Structure Database

[20] indicates that the typical CuO5 coordinationsphere contains four bonds in the range 1.90-2.05 D,and one longer bond, averaging 2.2-2.3 D. Theaverage Cul coordination sphere is therefore veryunusual, in that all five bonds are shorter than 2.04 D.The Cul-O2 bond of 1.85 D is among the shortest Cu-O bonds ever reported.

XAFS experiments [21] provide evidence for Cuclustering. Each Cu has at least one Cu in its secondcoordination sphere. This observation, and theappearance of the Fobs map, suggest a new model forthe local structure.

Consider the four 5-coordinate sites surroundingare individual O1. Stoichiometry mandates that thereare two Cu and two Al in the average “4-ring” aroundO1, and that there is only one oxygen atom at thecenter of the “4-ring”. If, according to the XAFSresults, the Cu ions occur in “cis” pairs, adisplacement of the central oxygen away from the two Cu in the xy plane would result in two long Cu-O1bonds and two short Al1-O1 bonds (Fiq. 10). Adisplacement of approximately 0.27 D along [10], to0.018, -0.018, 0, permits the bonding requirements ofall atoms to be better-satisfied, is consistent with theXAFS data, yields comparable residuals to theordered model for O1, and describes the same average structure. The combination of crystallographic andspectroscopic information has resulted in a newmodel for the local structure, a model consistent withal observations and with the catalytic properties of

27 The Rigaku Journal

Atom X Y Z Uiso,D2

Cu1/Al1* 0.2424(12) 0.19110(12) 0 0.0267

Al2/Cu2H 1/4 1/4 1/4 0.0050(5)

B 0.2504(6) 0.5054(7) 0 0.0079

O1I 0.018 -0.018 0 0.0081(19)

O2 0.2086(3) 0.1501(3) 0 0.0098

O3 0.2477(4) 0.3742(3) 0 0.0098

O4 0.4319(2) 0.2407(3) 0.2068(3) 0.0098

Table 3. Structural parameters of Cu2Al6B4O17

Space group I4/m, a=10.57945(1), c=5.67357(6) D at 27EC.

*Cu1/Al1 frac=0.487(2)/0.513(2), U11=0.0364(13), U22=0.0148(10), U33=0.029(11), U12=0.0001(9), U13=U23=0H Al2/Cu2 frac=0.988(2)/0.012(2)

IO1 frac = 1/4

this material. This model could be tested by carryingour diffraction anomalous fine structure (DAFS)experiments on reflections containing largecontributions from the Cul/Al1 site.

Evolution of the Structure as a Function ofTemperature

Important to determining the catalytic propertiesof a material are its crystallinity and averagecrystallite size. These attributes can often be

controlled by controlling the thermal history of thematerial. Accordingly, the properties of copperaluminum borate were studied as a function ofcalcination temperature. When a study from 600-900EC indicated that the most dramatic changes inproperties occurred between 700 and 850EC, thistemperature range was studied in more detail (Fig.11). Separate portions of an initial gel were calcinedat 710, 740, 780, 810, and 850EC. The powderpatterns were measured on a Scintag PAD Vdiffractometer; in addition, patterns of the 740 and810'C materials were measured on beamline X7A atNSLS, using a wavelength of 1.29366 D. Aliquotscalcined for 8 and 24 hours were essentially identical.

Trends can be seen in both the global andstructural parameters. The crystallinity can be derived from the refined scale factors. Crystallization iscomplete for calcinations at 780EC and higher (Fig.12). The average crystallite size increases withcalcination temperature, and is slightly anisotropic.The appropriate crystallite size, morphology, andcrystallinity can be obtained by control of thecalcination temperature.

We can gain a hint (but only a hint) at thestructural changes which occur during calcination.The structure of copper aluminum borate calcined at780, 810, and 850EC is essentially identical to thatreported above. The material calcined at 710EC is not

Vol. 16 No. 2 1999 28

Fig. 10. The proposed model for the local environmentof the Cu1/Al1 sites in Cu2Al6B4O17. The true position of O1 is displaced approximately 0.27 D from the averageposition. 50% probability ellipsoids.

Fig. 11. Laboratory powder patterns of Cu2Al6B4O17 calcined at various temperatures.

crystalline enough to obtain useful structure structural information from the powder pattern. The multipledata sets collected on the 740, 810, and 850ECmaterials permit refinement of some structural detail.

The most significant changes in the structureoccur at the Cul/Al1 site (Fig. 13). At all threetemperatures, the total electron density at this sitecorresponds to half occupancy by Al and Cu, but at the lower temperatures, the average cation location ismuch closer to one of the oxygens O2 than to theother. In the 740 and 810EC materials, the anisotropic“thermal” ellipsoid is elongated approximately in thedirection of the O2AAAO2 vector. At 740EC, theenvironment looks essentially like the superpositionof a tetrahedral AlO4 site and a trigonal bipyramidalCuO5 site. It has not proven to be possible to refinesplit site models of this site.

It is tempting to conclude that the averagestructure reported here represents a “metastable high-temperature disordered" form of a more-complexreality present at lower temperatures. While it ispossible to answer many structural questions using X-

ray powder diffraction data, it is not always possibleto gain all the information one desires. Moreknowledge about the structural changes in copperaluminum borate with temperature will probablyrequire in situ neutron diffraction studies.

Acknowledgments

We thank Ying-Mei Chen and K. M. Hanzlik fortechnical support, John Faber for assistance insynchrotron data collection, and G. J. Ray forcollecting and analyzing the 11B and 27Al NMRspectra.

D. E. Cox and J. A. Hriljac provided assistanceduring data collection on X7A, and we thank J. A.Hriljac for supplying the single crystal data on copperaluminum borate in advance of publication. P. W.

29 The Rigaku Journal

Fig. 12. Crystallinity and average crystallite size ofCu2Al6B4O17 as a function of calcination temperature.

Fig. 13. The Cu1/Al1 coordination spheres aftercalcinations at 740, 810, and 850EC. 50% probabilityellipsoids.

Stephens and his colleagues were most gracious andhelpful during the collection of synchrotron data onX3B2. This work represents work carried out in part at the National Synchrotron Light Source at Brookhaven National Laboratory, which is supported by the U.S.Department of Energy, Division of MaterialsSciences, and Division of Chemical Sciences. TheSUNY X3 beamline at NSLS is supported by theDivision of Basic Energy Sciences of the U. S.Department of Energy (DE FG02-86ER45231).

References[1] Y Tanaka, J. Fukunaga, M. Setoguchi, T. Higashi, and M.

lhara, J. Ceram. Soc. Japan, 90, 458-463 (1982); ICSDcollection code 201351.

[2] I. I. Kozhina, E. E. Kornilova, G. T. Petrovskii, and S. A.Stepanov, Vestn. Leningr. Univ., Fiz., Khim., 40-46(1983).

[3] A. C. Larson and R. B. Von Dreele, GSAS, The GeneralStructure Analysis System. Los Alamos NationalLaboratory (1994).

[4] G. M. Sheldrick, SHELXTL Plus, Version 3.4, NicoletInstrument Corporation (1988).

[5] A. C. Larson, F. L. Lee, Y. LePage, M. Webster, J. P.Charland, and E. J. Gabe, The NRCVAX CrystalStructure System, Chemistry Division, NRC, Canada,1988.

[6] 1. D. Brown and D. Altermatt, Acta Cryst., B41, 244 247(1985).

[7] R. Howie and A. West, Acta Cryst., B33, 381-385 (1977);ICSD collection code 828.

[8] L. C. Satek, J. A. Kaduk, and P. E. McMahon, in W. E.Pascoe (ed.), Catalysis of Organic Reactions, p. 235-243. New York, Marcel Dekker (1992).

[9] D. A. Keszler, Curr. Opin. Solid State Mater. Sci., 1, 204-211 (1996).

[10] T. Nagai and M. Ihara, J. Ceram. Assoc. Japan, 80,432-437(1972).

[11] J. F. Macdowell, J. Amer. Ceram. Soc., 73, 2287-2292(1990).

[12] J. W. Visser, J. Applied. Cryst., 2, 89-95 (1969).

[13] A. D. Mighel and V. L. Karen, J. Res. Natl. Inst. Stand.Technol., 101, 273-280 (1996).

[14] G. Bassi and J. Lajerowicz, Bull. Soc. Fr. Miner. Crist.,88, 342-344 (1965).

[15] L. Richter, “Synthese und Strukturuntersuchungen vonEisen- und Kupfer-Aluminum-Boraten”, TechnischeHochschule, Aachen (1977).

[16] L. C. Satek, J. A. Hriijac, R. D. Brown, J. A. Kaduk, andA. K. Cheetham, unpublished results.

[17] P. Coppens, Synchrotron Radiation Crystallography.London, Academic Press (1992).

[18] V. Petricek, Y. Gao, P. Lee, and P. Coppens, Phys.Rev., B42, 387 (1990).

[19] M. Martinez-Ripoli, S. Martinez-Carrera, and S.Garcia Blanco, Acta Cryst., B27, 677-681 (1971).

[20] Bergerhoff, R. Sievers, and R. Hundt, in F. H. Allen, G.Bergerhoff, and R. Sievers, eds., CrystallographicDatabases, Chester, International Union of Crystallo-graphy (1987).

[21] G. W. Zajac, J. Faber, and S. Pei, unpublished results.

Vol. 16 No. 2 1999 30