Journal of Non-Crystalline Solids 261 (2000) 282±286

www.elsevier.com/locate/jnoncrysol

Letter to the Editor

The ring structure of boron trioxide glass

C. Joo, U. Werner-Zwanziger, J.W. Zwanziger *
Department of Chemistry, Chemistry Building, Indiana University, Bloomington, IN 47405-4001, USA
Received 13 July 1999; received in revised form 2 August 1999

Abstract
A new experiment, sensitive to the dihedral angle distribution of neighboring structural units, was applied to boron
trioxide glass. The experiment consisted of correlating the NMR frequencies of neighboring boron nuclei. The fre-
quencies are identical if the boron oxide polyhedra containing the two borons are co-planar, as found in ring systems,
but di€er for non-co-planar geometries. Through the rate of growth of signal intensity arising from the non-co-planar
case, as compared to that obtained in a boron trioxide crystal of known structure, the fraction of boron contained in
rings in the glass was determined to be 0:70  0:08. Ó 2000 Elsevier Science B.V. All rights reserved.

Boron trioxide glass has composition B2 O3 and
consists of planar BO3 triangles (Fig. 1(a)) [1].
These polyhedra share oxygen atoms, resulting in
a three-dimensional network. The extent of inter-
mediate range order in this material, and in par-
ticular the extent of planar boroxol ring
formation, is the subject of this contribution.
Boron trioxide glass attracts interest because it
has a simple composition and is an excellent glass-
former. Its structure has been modeled using a
variety of approaches, and it is considered a
bench-mark among network glasses. That its
structure remains controversial can be attributed
to the experimental diculty of probing disor-
dered materials on the 0.3±1.0 nm length scale.
While indeed pair-wise inter-atomic correlations
on this length scale can be extracted from di€rac-
tion data, modeling such data to determine dihe-
*
Corresponding author. Tel.: +1-812 855 3994; fax: +1-812
855 8300.
E-mail address: jzwanzig@indiana.edu (J.W. Zwanziger).
0022-3093/00/$ - see front matter Ó 2000 Elsevier Science B.V. All rights reserved.
PII: S 0 0 2 2 - 3 0 9 3 ( 9 9 ) 0 0 6 0 9 - 2
dral angles between coupled polyhedra has proven
very dicult.
The speci®c issue in question here is the extent
to which six-membered boroxol rings (Fig. 1(b))
are found in boron trioxide glass. A variety of
experimental probes has been interpreted as
showing the existence of a signi®cant fraction of
rings, including Raman spectroscopy [2,3], which
shows an intense sharp feature assignable to the
ring breathing mode; inelastic neutron scattering
[4], which detects the same feature seen in Raman
spectroscopy; neutron di€raction, which shows
inter-atomic correlations consistent with boroxol
rings [5]; and 17 O NMR [6,7] and 11 B NMR [8],
which show spectroscopically distinct oxygen and
boron sites. This body of evidence has been in-
terpreted as showing that the fraction of boron in
boroxol rings in the glass is in the range 0.65±0.8.
It is in our view equally important that, despite
numerous attempts, models which match the low
glass density and which include a large fraction of
boroxol rings have not been built [9±13]. The di-
verse methods used consistently yield model
 C. Joo et al. / Journal of Non-Crystalline Solids 261 (2000) 282±286
Fig. 1. (a) Planar BO3 triangle as found in boron trioxide glass
and crystal. (b) Planar boroxol ring.
structures consistent with di€raction data and with
the correct density, but which have few if any rings
(at most, the fraction of rings expected purely by
chance are found, which typically contain about
0.15 of the boron).
Because of the above sharp divergence of in-
terpretations, and the fundamental importance of
this glass to understanding intermediate range
order in inorganic glass-formers generally, we un-
dertook a measurement sensitive to the dihedral
angle distribution and hence the ring fraction
using a two-dimensional NMR experiment. In our
approach, the powder-patterns of neighboring
boron atoms were correlated through a spectral
spin-di€usion process. Anisotropy correlations
have been used in numerous ways to determine
local geometry around spin-1/2 nuclei such as
carbon [14], but for quadrupolar nuclei this ap-
proach has been developed only very recently [15].
In the experiment [15], a powdered sample ro-
tates in the NMR probe at about 5 kHz; the re-
sidual anisotropic interactions yield an
inhomogeneously broadened line shape. The ex-
periment was executed by applying a resonant
radio-frequency pulse to the boron nuclei under
MAS conditions, and allowing the magnetization
to evolve for a time t1 . This served to label each
spin by its anisotropic frequency shift. A second
pulse was then applied, to store the evolved mag-
netization parallel to the external magnetic ®eld,
and the rotation axis of the sample was ¯ipped o€
the magic angle (OMAS). For a sample rotating
about an axis at an angle H with respect to the
external ®eld, the inter-nuclear magnetic dipole
couplings are scaled by P2 …cos H†. Thus under
OMAS conditions dipole coupling can be rein-
troduced, allowing magnetization to transfer be-
tween nearby sites. The sample rotated under
OMAS conditions for a mixing time sm , and thus
283
an e€ective mixing time seff ˆ P2 …cos H†sm , and
then the rotor axis was ¯ipped back to the magic
angle. A third pulse converted the stored magne-
tization back to the transverse plane, where it was
detected during a time t2 . The signal was collected
in the time-domain as S…t1 ; t2 †, and a double-
Fourier transform yielded the spectrum I…m1 ; m2 †.
The sequence of pulses was phase-cycled to retain
only the signal that had evolved in the transverse
plane during both t1 and t2 , and was accumulated
in hyper-complex mode to obtain pure absorption
phase two-dimensional spectra.
The above experiment was applied for various
mixing times to both glassy and crystalline B2 O3 .
The glass was made by melting boric acid in a
platinum crucible, and holding it for 1 h at
1000°C. The melt was quenched on a brass plate,
and stored under dry conditions until use. The
glass was crushed to a ®ne powder to facilitate
sample spinning. Crystalline B2 O3 was prepared
following a literature procedure, by seeding a melt
with borophosphate followed by a very slow
cooling and annealing cycle [16]. Phase purity was
veri®ed by powder X-ray di€raction and compar-
ison to the known crystal structure [17].
Representative two-dimensional spectra at var-
ious mixing times are shown in Fig. 2. As ex-
plained below, at short mixing times only a signal
on the diagonal is seen, because magnetization
that begins on a given nucleus during t1 is observed
on the same nucleus during t2 . However, if during
the mixing time the magnetization transfers via the
magnetic dipole coupling from one nucleus to a
neighbor, and the neighboring nucleus has a dif-
ferent resonance frequency due to a di€ering an-
isotropic shift, the signal will appear o€-diagonal
in the two-dimensional spectrum. Fig. 2 shows
examples of both e€ects, namely a signi®cant di-
agonal component due to non-transferred mag-
netization and transferred magnetization that
happens to have the same shift, and o€-diagonal
signal due to spin di€usion to neighbors with dif-
ferent frequency shifts.
To interpret this experiment, note ®rst that the
dominant local interaction to which the boron
nucleus is subject is the coupling between the
quadrupole moment and local electric ®eld gradi-
ents. Due to the local three-fold symmetry of the
284 C. Joo et al. / Journal of Non-Crystalline Solids 261 (2000) 282±286

Fig. 2. Two-dimensional spectra of boron-11 in boron trioxide glass, obtained by correlating the anisotropic frequencies of the boron
nuclei with those of neighboring sites, for increasing e€ective magnetization transfer times. The diagonal intensity re¯ects both non-
exchanged magnetization and magnetization that transferred during the mixing time to sites with equivalent frequencies; the o€-di-
agonal intensity re¯ects magnetization transfer to sites with di€ering anisotropic frequencies.

boron sites (Fig. 1(a)), the asymmetry of the
quadrupole interaction is essentially zero and the
principal axis is located along the three-fold sym-
metry axis. The interaction can thus be represented
as an ellipsoid centered on the boron atom with its
long axis orthogonal to the BO3 plane (Fig. 3). The
NMR frequency of a particular boron depends on
the orientation of the quadrupole interaction with
respect to the rotor-®xed frame; even after partial
averaging due to MAS, some anisotropy remains
and the transition frequency is given by
m ˆ miso ‡ maniso …bPR †. Here miso is the isotropic fre-
Fig. 3. The electric quadrupole interaction tensor at each boron
can be represented by an ellipse, orienting it with respect to the
molecular frame. The Euler angle b gives the relative orienta-
tion between neighboring tensors; only one such angle is nec-
essary because the interaction is approximately cylindrically
symmetric.
quency shift, and the anisotropic component maniso
depends on the Euler angle bPR that ®xes the in-
teraction in a rotor-®xed frame of reference. Only
one such Euler angle is needed within the ap-
proximation of a symmetric quadrupole interac-
tion.
For two coupled boron atoms we must consider
the relation between their respective principal axis
frames, which for the present case of symmetric
interactions can be described with one additional
Euler angle b, relating the two quadrupole tensors
(Fig. 3). If the neighboring boron atoms are in
mutually co-planar polyhedra, then b ˆ 0° and
they will exhibit identical frequency shifts. Then,
even after magnetization transfer, only a signal on
the diagonal in the two-dimensional spectrum will
result. However, for non-zero dihedral angles the
coupled boron atoms will have di€erent frequency
shifts, and hence yield o€-diagonal intensity. At
short transfer times seff , all signal appears on the
diagonal, because there has not been sucient time
for transfer. At long seff signi®cant o€-diagonal
intensity is observed in samples with non-zero b
between adjacent sites. We con®rmed this predic-
tion by performing the experiment in crystalline
K3 B3 O6 , which consists entirely of well-separated,
planar oxide-capped boroxol rings; the signal re-
mained diagonal even at long mixing times.
The rate of growth of the o€-diagonal signal in
this experiment is shown in Fig. 4, for both glass
 C. Joo et al. / Journal of Non-Crystalline Solids 261 (2000) 282±286
Fig. 4. O€-diagonal spectral intensity, normalized to the total
intensity, as a function of the e€ective mixing time
seff ˆ P2 …cos H†sm , for glassy B2 O3 (squares) and crystalline
B2 O3 (circles).
and crystalline B2 O3 as a function of
seff ˆ P2 …cos H†sm . We extracted this result from
the spectra (Fig. 2) by projecting along the diag-
onal to obtain a one-dimensional representation,
using the short time data to model the diagonal
portion of the spectrum, removing the diagonal
contribution from the projection, and plotting the
o€-diagonal integrated intensity normalized by the
total intensity for each value of seff . The data show
several qualitative features: ®rst, the initial growth
of o€-diagonal intensity in the glass is somewhat
faster than in the crystal, and secondly, both show
a break in slope at about the same e€ective mixing
time.
We interpret the growth of o€-diagonal inten-
sity in the glass by comparing it to the crystal. In
the crystal, each boron is at the center of a planar
BO3 group (Fig. 1(a)), and these polyhedra are
linked such that each boron has three nearest
neighbor polyhedra: two nearly co-planar with it
(dihedral angles b < 5°) and one neighbor with
b ˆ 78°, although the crystal contains no rings
[17]. While the details of magnetization transfer in
systems like this are very complex, it is clear that at
short times the transfer should be proportional to
the number of available partners [14]. It is im-
portant to note here that, while the bulk density of
the glass is much lower than the crystal, the local
inter-atomic distances, as assessed from neutron
di€raction in the glass [5] and the X-ray structure
285
of the crystal [17], are very similar. Thus the two
samples are directly comparable in terms of dipole
coupling strengths, for short magnetization trans-
fer times.
From the slopes of the intensity data at short
times in Fig. 4, we ®nd that the rate of increase of
o€-diagonal intensity in the glass is greater than in
the crystal by a factor of 1:60  0:16. This number
can be used to estimate the fraction of boron in the
glass contained in boroxol rings. We assume a
maximally disordered ring model, such that, in the
glass, neighboring boron polyhedra will be mutu-
ally co-planar only if constrained to be so by virtue
of being bound within the same boroxol ring, and
that the dihedral angles between other neighboring
pairs of boron polyhedra will be in general non-
zero. Note that, in this model, all non-aligned
neighbor interactions give rise to observable o€-
diagonal intensity in the spectrum, and the faster
growth rate of signal in the glass relative to the
crystal indicates that the glass contains a larger
fraction of non-aligned interactions. In a sample
with N boron atoms, there are 3N =2 near-neighbor
interactions (3 per boron, and the factor of 2
avoids double-counting). Let f be the fraction of
boron in boroxol rings. Then the number of mu-
tually co-planar interactions is 2Nf =2 ˆ Nf (each
of the Nf ring boron participates in two, and di-
vision by two avoids double-counting). The other
3N =2 ÿ Nf contribute to the o€-diagonal signal.
In the crystal, 1/3 of the interactions are of the
non-co-planar type, or N =2. Thus the ratio of non-
co-planar interactions in glass to crystal is
N …3=2 ÿ f †=…N =2† ˆ 3 ÿ 2f . Setting this equal to
the ratio of the o€-diagonal intensity growth rate,
or 1.60, yields the fraction of boron in rings in the
glass as f ˆ 0:70. By propagating uncertainties in
the data, we put this ®gure at f ˆ 0:70  0:08.
This ®gure is in accord with the Raman, neu-
tron, and NMR studies mentioned in the intro-
ductory section. In those experiments, the ring
fraction was determined somewhat indirectly; the
present results provide a more direct measurement.
The modeling studies noted above indicate distri-
butions of dihedral angles that are essentially ¯at
[13], or linearly decreasing, from b ˆ 0° to b ˆ 90°
[11]. For these two distributions, 12% and 21%
respectively of the interactions have dihedral
286 C. Joo et al. / Journal of Non-Crystalline Solids 261 (2000) 282±286
angles less than 10°, the angle we use as a rough
divider between signal on and o€ the diagonal (our
line±shape simulations show that, for greater val-
ues of b, o€-diagonal intensity is signi®cant). The
ratio of slopes between glassy and crystalline B2 O3
for these distributions would be 2.7 and 2.4, re-
spectively, well outside the uncertainty in our ob-
served ratio of 1:60  0:16. Our observed ratio of
slopes is consistent with a distribution for which
some 50% of the nearest neighbor interactions
have dihedral angles less than 10°, while the re-
maining 50% have larger dihedral angles.
In summary, we have performed a two-dimen-
sional NMR experiment that is sensitive to the
dihedral angle distribution in boron trioxide glass.
By comparing the results of the glass to those
obtained from crystalline boron trioxide, we de-
termine the fraction of boron in six-membered
rings in the glass to be f ˆ 0:70  0:08. A signi®-
cant challenge in this ®eld remains to build struc-
tural models of the glass with the correct density
and that contain signi®cant ring fractions, or on
the other hand to build models with no rings that
can still explain not just the neutron di€raction
data but also the Raman spectroscopy, inelastic
neutron data, NMR spectral data, and the present
experiment.
Acknowledgements
We thank Dr Peter Hartmann and Professor
Christian Jager for helpful conversations, and
thank Professor Philip Bray for providing us with
his method for preparing crystalline boron triox-
ide. This work was supported by the NSF under
grant DMR-9870246.
References
[1] A.H. Silver, P.J. Bray, J. Chem. Phys. 29 (1958) 984.
[2] T.W. Bril, Philips Res. Rep. Suppl. 2 (1976) 1.
[3] C. Windisch Jr., W. Risen Jr., J. Non-Cryst. Solids 48
(1982) 307.
[4] A. Hannon, A. Wright, J. Blackman, R. Sinclair, J. Non-
Cryst. Solids 182 (1995) 78.
[5] P.A.V. Johnson, A.C. Wright, R.N. Sinclair, J. Non-Cryst.
Solids 50 (1980) 281.
[6] G.E. Jellison Jr., L.W. Panek, P.J. Bray, G.B. Rouse Jr., J.
Chem. Phys. 66 (1977) 802.
[7] R.E. Youngman, S.T. Haubrich, J.W. Zwanziger, M.T.
Janicke, B.F. Chmelka, Science 269 (1995) 1416.
[8] R.E. Youngman, J.W. Zwanziger, J. Non-Cryst. Solids 168
(1994) 293.
[9] M. Amini, S.K. Mitra, R.W. Hockney, J. Phys. C 14 (1981)
3689.
[10] S.J. Williams, S.R. Elliott, Proc. R. Soc. London A 380
(1982 ) 427.
[11] A.H. Verhoef, H.W. den Hartog, J. Non-Cryst. Solids 146
(1992) 267.
[12] M. Teter, in: A.C. Wright, S.A. Feller, A.C. Hannon
(Eds.), Borate Glasses, Crystals, and Melts, Soc. Glass
Tech., Sheeld, UK, 1997, p. 407.
[13] J. Swenson, L. B orjesson, Phys. Rev. B 55 (1997) 11138.
[14] M. Ernst, B.H. Meier, in: I. Ando, T. Asakura (Eds.), Solid
State NMR of Polymers, Studies in Physical and Theoret-
ical Chemistry, vol. 84, Elsevier, Amsterdam, 1998, p. 83.
[15] P. Hartmann, C. J ager, J.W. Zwanziger, Solid State NMR
13 (1999) 245.
[16] D. Kline, P.J. Bray, H.M. Kriz, J. Chem. Phys. 48 (1968)
5277.
[17] G.E. Gurr, P.W. Montgomery, C.D. Knutson, B.T.
Gorres, Acta Crystallogr. B 26 (1970) 906.