Detection and identification of corrosion products of sodium aluminoborosilicate glasses by 23Na MQMAS and 1H→23Na CPMAS NMR

Article abstract of DOI:10.1021/jp992999m

²³Na multiple-quantum (MQ) MAS NMR is applicable for monitoring the chemical and structural changes resulting from atmospheric exposure of a series of alkali aluminoborosilicate glasses with compositions RNa₂O: 1B₂O₃:1SiO₂:0.25Al₂O₃ (where R = 0.5-2.5). Glasses with high alkali concentrations possess greater numbers of nonbridging oxygens within the bulk structure and presumably at the initial surface of a fresh sample, and for three samples with R≥1.5 sharp resonances are revealed in the isotropic dimension of an MQMAS NMR experiment conducted after prolonged atmospheric exposure. The MQMAS NMR experiments, combined with ¹H→²³Na cross-polarization magic-angle spinning (CPMAS) NMR measurements, indicate that these resonances arise from sodium cations no longer participating in the glass network. Two new phases are formed as corrosion products and have been identified as an anhydrous Na₂CO₃ phase and a NaBO₂·1H₂O phase through comparison with ²³Na MQMAS and ¹H→²³Na CPMAS NMR spectra of crystalline samples. Due to an inherent difficulty with direct quantification of populations based on MQMAS spectra, a simplified approach for quantification of the amount of the new carbonate phase is presented. Values are then calculated for relative amounts of corrosion product formation for different exposure times and bulk glass compositions.

Full text of DOI:10.1021/jp992999m

9580
J. Phys. Chem. B 2000, 104, 9580-9586
Detection and Identification of Corrosion Products of Sodium Aluminoborosilicate Glasses
23
1
23
by Na MQMAS and H f Na CPMAS NMR
J. M. Egan and K. T. Mueller*
Department of Chemistry, The PennsylVania State UniVersity, 152 DaVey Laboratory,
UniVersity Park, PennsylVania 16802
ReceiVed: August 24, 1999; In Final Form: July 3, 2000
23Na multiple-quantum (MQ) MAS NMR is applicable for monitoring the chemical and structural changes
resulting from atmospheric exposure of a series of alkali aluminoborosilicate glasses with compositions RNa2O:
1B2O3:1SiO2:0.25Al2O3 (where R ) 0.5-2.5). Glasses with high alkali concentrations possess greater numbers
of nonbridging oxygens within the bulk structure and presumably at the initial surface of a fresh sample, and
for three samples with R G 1.5 sharp resonances are revealed in the isotropic dimension of an MQMAS
NMR experiment conducted after prolonged atmospheric exposure. The MQMAS NMR experiments, combined
1
with H f 23Na cross-polarization magic-angle spinning (CPMAS) NMR measurements, indicate that these
resonances arise from sodium cations no longer participating in the glass network. Two new phases are formed
as corrosion products and have been identified as an anhydrous Na2CO3 phase and a NaBO2‚1H2O phase
1
through comparison with 23Na MQMAS and H f 23Na CPMAS NMR spectra of crystalline samples. Due
to an inherent difficulty with direct quantification of populations based on MQMAS spectra, a simplified
approach for quantification of the amount of the new carbonate phase is presented. Values are then calculated
for relatiVe amounts of corrosion product formation for different exposure times and bulk glass compositions.
Introduction
In sodium aluminoborosilicate glasses, the sodium cations
act as a network modifier and their chemistry is tied to the
Understanding the durability of a glass within a particular
compositional series under typical environmental conditions is
crucial for basic and advanced technological applications of the
material. Borosilicate glasses are considered for various ap-
plications including laboratory glassware, nuclear waste disposal
and containment, and optical glasses. In each of these applica-
tions both understanding and identifying corrosive processes
affecting glass networks or surfaces are important, especially
for prediction of failure after long-term interactions with the
environment. One significant challenge is the development of
quantitative methods, and the ease of implementation of those
methods, for determination and comparison of the severity of
corrosion.
The formation of sodium carbonate on the surface of sodium
silicate glasses has long been known and studied. In silicate
glass systems, corrosion resistance decreases as the percentage
of alkali metal oxide increases.1 Experimental studies have
investigated the formation of Na2CO3 on glass surfaces utilizing
atomic force microscopy (AFM),2-4 electron microscopy,5 and
secondary ion mass spectrometry (SIMS).6 Most importantly,
the formation of a surface phase can be observed, and X-ray
photoelectron spectroscopy (XPS) has been utilized to identify
the phase as Na2CO3. Carter et al.4 studied the surfaces of silicate
glass via AFM with a “stiffer” tip then normally used (nominal
spring constant of 0.064 N/m) and found features on the glass
surface that were over 300 nm in height and up to 2 µm in
width after just 20 days of exposure to air. They attributed these
features to alkali carbonates, a hypothesis consistent with their
Auger spectroscopic measurements. However, the presence of
alkali silicates could not be ruled out.
corrosion behavior and other physicochemical properties of these
advanced materials. Thus, sodium is of central interest for
corrosion studies. Solid-state magic-angle spinning (MAS) NMR
spectra of 23Na nuclei have been published in numerous studies
of glass systems.7,8 Unfortunately, 23Na MAS NMR spectra are
usually uninformative as to the specific identity of environments
surrounding the 23Na nuclei in amorphous systems, essentially
due to the broad line shapes encountered. Partly responsible
for the broad resonances is the distribution of bonding param-
eters (bond angles and internuclear distances) associated with
the amorphous state of a glassy material. Another difficulty
encountered in 23Na NMR is the inability of magic-angle
spinning to completely average away the effects of second-order
quadrupolar interactions, leading to residual linebroadening in
23Na MAS spectra. Improved solid-state NMR techniques are
needed for higher resolution 23Na spectra. For example, a recent
study using a technique called off-resonance quadrupole nutation
has reported the resolution of different sodium environments
in hydrous albite glass.9
Multiple-quantum (MQ) MAS10,11 is another recent experi-
mental development in solid-state NMR that eliminates the
quadrupolar broadenings in spectra of half-integer quadrupolar
nuclei. MQMAS has the advantage that commercially available
MAS probes can be used to acquire higher-resolution spectra.
MQMAS is essentially a two-dimensional experiment where
spin magnetization evolves along specific multiple- and single-
quantum spin coherence pathways during two separate time
periods. The two evolution periods allow for the averaging to
zero of residual anisotropic broadening while under MAS
conditions. The resulting two-dimensional correlation spectrum
consists of a dimension free from anisotropic line broadening,
with direct correlation to an anisotropic MAS line shape in a
* Author to whom correspondence should be addressed. Fax: (814) 863-
8403. E-mail: ktm2@psu.edu.
10.1021/jp992999m CCC: $19.00 © 2000 American Chemical Society
Published on Web 09/23/2000

Corrosion of Sodium Aluminoborosilicate Glasses
J. Phys. Chem. B, Vol. 104, No. 41, 2000 9581
second dimension. Spectral editing (utilizing the isotropic
dimension) coupled with the anisotropic information of the
second dimension results in superior site resolution, discrimina-
tion, and characterization. Detailed investigations of the theoreti-
cal formalism and optimal experimental conditions for MQMAS
are found in the literature12-14 and are not the focus of this study.
In studies of amorphous samples, MQMAS experiments have
been used to investigate 11B and 17O nuclei participating in oxide
glass systems, and the results have provided enhanced models
of the structures and linkages present.15-17 In the past, 23Na
MQMAS studies have been successfully performed on crystal-
line18-23 or highly ordered materials,24 and limited 23Na MQ-
MAS studies have been focused toward investigations of glass
systems.15,25 Precise structural analysis is hindered in the glass
experiments due to the lack of long-range order found in glasses,
so that discussions of these 23Na MQMAS studies are limited
to the broadness of resonances present in both dimensions of
the spectra.
for the highest-alkali content (R ) 2.5) glass sample after
extensive atmospheric exposure. FTIR spectra of selected
samples were acquired on a Perkin-Elmer 1600 series instru-
ment.
All solid-state NMR experiments were performed using a
Chemagnetics CMX-300 spectrometer operating at a magnetic
field strength of 7.0 T. The 23Na MAS and MQMAS experi-
ments were performed utilizing the low-frequency channel of
a 4 mm Chemagnetics HX double-resonance probe. Rotor
frequencies of 12.000 kHz were kept constant to within (3 Hz
with a Chemagnetics MAS speed controller. The 23Na resonance
frequency at this field strength is 78.660 MHz and all spectra
are externally referenced to the 23Na resonance from a 1 M NaCl
solution. A 0.5 µs pulse length was used to obtain 23Na MAS
spectra. Recycle delays were 1 s for both MAS and MQMAS
experiments.
A three-pulse sequence employing TPPI (time-proportional
phase incrementation)26 for two-dimensional data acquisition
was used for all MQMAS experiments. The three pulses include
a triple-quantum excitation pulse and two pulses for the z-filter
pulse scheme.14,27,28 The first two pulses were hard, nonselective
pulses of 3.5 µs and 1.4 µs duration, while the last pulse was
attenuated in order to produce a “soft” pulse, selective for the
central transition of the 23Na nuclei (5.50 µs pulse duration).
All pulse lengths were optimized to produce maximum signal
for the spectrum of the first t1 data point of a two-dimensional
experiment. The t1 increment used provided for a 50 kHz sweep
width in the first (high resolution) spectral dimension. A total
of 936 scans were acquired for each t1 point, and two-
dimensional data sets consisted of either 128 t1 points × 256 t2
points or 256 t1 points × 256 t2 points. All two-dimensional
data sets have been processed with a shearing transformation
by the reported methods12,14 prior to the complex Fourier
transformation. Deconvolutions of the MQMAS projections and
determinations of the centers of gravity of resonances were
performed using GRAMS/32 deconvolution software.29
Additional 1H f 23Na CPMAS experiments were performed
to identify sodium nuclei that exhibit dipolar connectivity to
the hydrogen nuclei present in hydroxyl groups or molecular
water. The 1H resonance frequency at 7.0 T was 297.372 MHz.
A sample of Na2CO3‚1H2O was utilized to establish the
optimum matching condition for the CP experiment. A 1 ms
contact time was found to produce maximum signal and a 5 s
recycle delay was used between acquisitions. Results of the 1H
This paper reports on the chemical and structural results of
the reactions of alkali aluminoborosilicate glasses with a typical
laboratory atmosphere. Changes in the local environments
surrounding sodium nuclei are monitored after different exposure
times and from samples with different bulk glass compositions.
Chemical changes are initially revealed by 23Na MQMAS
1
experiments, and subsequent investigations incorporate H f
23Na cross-polarization MAS (CPMAS) measurements. New
resonances that appear as a function of exposure time are
identified on the basis of comparisons to 23Na MQMAS and
1H f 23Na CPMAS spectra of reference carbonate, borate, and
silicate compounds. In this study, one advantage to the MQMAS
method is that the new phases formed by corrosion contain local
environments for sodium that are overall more ordered than the
sodium sites in the original glass samples, providing sharper
and more easily resolved resonances in MQMAS spectra. The
relative extent of chemical reactivity of each glass is quantified
through deconvolution of MQMAS spectra, providing analyses
that allow comparisons of corrosion across a family of samples
with the same reaction products.
Experimental Section
A family of glass samples with batched mole ratios; RNa2O:
1B2O3:1SiO2:0.25 Al2O3 (where R ) 0.5-2.5) were prepared.
The samples will be referred to by their sodium oxide content
(R value) in the text. The raw materials consisted of Na2CO3,
calcined alumina, and H3BO3 (all purchased from J. T. Baker
and used as received), and silica (Min-u-sil from U.S. Silica).
The raw materials were crushed and mixed, and then heated in
a Pt crucible to 1200 °C, where they were held for 3 h. The
melts were then quenched between metal plates and crushed to
powder for the NMR experiments. Once the samples were
crushed, they were stored in glass vials within a desiccator for
varying amounts of exposure times. The uniformly sized glass
vials each held 2-3 mL of sample in a total volume of 20 mL.
The atmosphere in the vials was the standard laboratory
atmosphere. All glass samples were quenched and crushed on
the same day and placed in sample vials within a desiccator
immediately upon finishing the crushing. Exposure times refer
to the time period between the crushing of the glass samples
and the start of the NMR experiments, so that time zero
corresponds to performing the NMR experiment immediately
after crushing the glass sample. Powder X-ray diffraction,
performed on a Philips XPert MPD System using Cu KR
radiation, proved that all original samples were amorphous.
Sharp peaks indicating crystalline material were only observed
f
23Na CPMAS experiments were pivotal in establishing the
identity of phases that were hydrated, as well as indirectly
supporting the presence of anhydrous phases (through a lack
1
of H f 23Na CP signal).
Results and Discussion
23Na MAS NMR spectra of the R ) 2.5 glass sample for
three different atmospheric exposure times are shown in Figure
1. The 23Na MAS NMR spectrum of the R ) 2.5 sample at
time zero (Figure 1a) demonstrates the lack of resolution
commonly observed for amorphous samples. As the powdered
sample is exposed to the atmosphere for further time, the center
of gravity of the broad resonance shifts downfield from -31.6
to -20.9 ppm. In addition to a spectral shift, a sharp contribution
to the NMR signal begins to appear at +1.0 ppm after two weeks
of atmospheric exposure. After months of exposure this ad-
ditional feature in the spectrum becomes more prevalent and is
a strong indication of changes occurring in the local environment
surrounding some (but obviously not all) of the sodium nuclei.
Due to the broad nature of the other portions of the 23Na MAS

9582 J. Phys. Chem. B, Vol. 104, No. 41, 2000
Egan and Mueller
Figure 2. A 23Na two-dimensional MQMAS spectrum of the R ) 2.5
glass sample exposed for eight months is shown. Contour lines are
drawn to indicate 5-95% of signal intensity and are spaced equally
apart.
Figure 1. 23Na MAS spectra of an R ) 2.5 glass sample are shown
for three different times of exposure to laboratory atmospheric
conditions, as indicated. All spectra were processed by Fourier
transformation of 128 transients collected with a 1 s recycle delay.
The arrow indicates the sharp singularity that appears as a result of
atmospheric exposure.
spectra, it is difficult to quantify or even further characterize
the changes occurring from 23Na MAS experiments alone. 23Na
MAS spectra of glass samples with different alkali oxide
compositions (not shown) that have been exposed for 8 months
exhibit broad lineshapes whose centers of gravity range from
-37.0 ppm (R ) 0.5 glass) to -20.9 ppm (R ) 2.5 glass). The
shift becomes more positive as the sodium oxide content
increases in the samples. After eight months, only the R ) 2.5
glass sample clearly shows a visible change in the 23Na MAS
spectrum at 7.0 T, with the presence of a “sharp” resonance at
+1.0 ppm.
MQMAS experiments were performed to aid in the identi-
fication of any new species that slowly appear over time in the
glass samples. Figure 2 displays a two-dimensional MQMAS
correlation spectrum of the R ) 2.5 sample after exposure to
laboratory atmosphere (as described above) for a period of 8
months. The horizontal axis corresponds to the anisotropic MAS
dimension and the vertical axis corresponds to the isotropic
dimension. Along the isotropic dimension three features can
be observed. Two of the features are sharp peaks identifying
resonances that have been narrowed significantly by use of the
MQMAS experiment, while the remaining feature is a broad
resonance that shows little change in line shape when compared
to the MAS data. We propose that the two sharp resonances
arise from sodium atoms that have been removed from the bulk
glass structure and have been incorporated into a more ordered
environment compared to the glass network. 23Na MQMAS
spectra of an unexposed R ) 2.5 sample display evidence of
neither new site. The positions of these two new resonances
are +2880 and +5000 Hz in the isotropic dimension. The broad
resonance represents sodium cations that remain in the structural
network of the bulk glass. This sodium resonance shows little
or no narrowing because of a distribution of local electromag-
netic environments not averaged away by the MQMAS experi-
ment.
Figure 3. Single-slice projections (taken at +1.0 ppm in the MAS
dimension) of the 23Na isotropic dimension from two-dimensional
MQMAS experiments performed on the R ) 2.5 sample are shown
for three different exposure times, as indicated. These single-slice
projections were chosen to display the presence (or absence) of the
two sharp peaks from more-ordered sodium environments.
A single slice of the isotropic dimension for the R ) 2.5
sample (after 8 months total exposure) is shown in Figure 3c.
From this projection the two sharp resonances can be clearly
observed with the broad component making up the remainder
of the spectrum. Also depicted in Figure 3 are single-slice
projections obtained at two previous exposure times for the R
) 2.5 sample. Figure 3a shows no indication of specific
narrowed resonances appearing in data from a freshly crushed
R ) 2.5 glass. Figure 3b shows the emergence of the two sharp
resonances after two weeks of exposure to the atmosphere.
Another characteristic of the R ) 2.5 glass after two weeks
exposure time is the approximately equal intensities of the two
narrower peaks. At much longer exposure times the peak at
+2880 Hz is of greater intensity.
Figure 4 displays full projections along the isotropic dimen-
sion for three glass samples with different alkali oxide composi-

Corrosion of Sodium Aluminoborosilicate Glasses
J. Phys. Chem. B, Vol. 104, No. 41, 2000 9583
Figure 5. (a) A projection from the 23Na MQMAS spectrum of an R
) 2.5 glass sample reveals the MAS spectrum of the species that is
directly correlated to the sharp resonance at +2880 Hz in the isotropic
dimension. (b) A similar MAS projection from the 23Na MQMAS
spectrum of pure, crystalline Na2CO3 is shown to compare with the
phase present in the glass samples. (c) The 23Na MAS spectrum of
pure, crystalline Na2CO3 is displayed, and its comparison to spectrum
(b) underscores the deficiencies of triple-quantum excitation during the
MQMAS experiments implemented here.
Figure 4. Changes in the 23Na MQMAS spectra are charted with the
emergence of one of the sharp resonances in the full isotropic
projections as a function of sodium content. No evidence is seen for
changes in the sodium environment in (a) the R ) 0.5 glass, while a
new resonance clearly appears in the (b) R ) 2.0 and (c) R ) 2.5 glass
samples.
tions and after exposure for an 8-month time period. Both the
R ) 0.5 (Figure 4a) and R ) 1.0 (not shown) samples lack any
evidence of singularities in the 23Na MQMAS spectra. These
two samples exhibit broad NMR line shapes in both dimensions,
consistent with no measurable change in the environment of
sodium nuclei. Isotropic projections of the R ) 1.5 glass sample
(not shown) begin to display evidence of a newly formed sodium
environment but the full projection does not indicate the
presence of any sharp resonances because the bulk glass remains
mostly intact and its signal dominates the spectrum. When the
glass has a sodium content with R ) 2.0, full projections indicate
the presence of at least one sharp resonance indicative of a
highly ordered sodium environment, as well as the remaining
broad component from sodium in the bulk glass. The presence
of any additional distinct resonances is not obvious in the full
projections because the broad component dominates most
regions of the spectra.
TABLE 1: Comparison of Measured NMR Parametersa for
Pure Na2CO3 and a Putative Na2CO3 Phase from a Glass
Sample (R ) 2.5) after Exposure to the Atmosphere
sample
crystalline Na2CO3
R ) 2.5 glass
V1′ av freq. (Hz)
V2 av freq. (Hz)
V0Q (Hz)
+3267
-520
-291
+4.5
+2983
-504
-271
+4.0
V
CS (ppm)
a See ref 12 for definitions.
interaction parameters.12 At the radio frequency field strengths
utilized in these experiments, only one of the two Na2CO3 sites
observed can be excited efficiently. Based on the spectra
presented in Figure 5, and the fact that this match is the best
obtained from any crystalline compound considered, the pre-
liminary assignment of the resonance at +2880 Hz in the
MQMAS spectrum is to an anhydrous sodium carbonate phase.
While the known chemistry for leaching and corrosion of
similar glasses indicates the probable formation of carbonate
species, determination of the nature and identity of the particular
species responsible for the sharp resonances observed in the
MQMAS experiment necessitates a comparison of the spectra
with results from numerous crystalline samples of potential
corrosion products or intermediates. The 23Na MAS spectrum
of crystalline Na2CO3 is displayed in Figure 5c and does not
agree well with a MAS (low-resolution dimension) projection
from the R ) 2.5 sample that is correlated to the sharp resonance
at +2880 Hz in the MQMAS spectrum (Figure 5a). A much
superior match is observed when a MAS slice is taken (Figure
5b) from the two-dimensional MQMAS spectrum of crystalline
Na2CO3 at the same frequency shift in the isotropic dimension
as the slice from the glass sample. The difference can be
explained by the efficiency of the radio frequency pulses in the
MQMAS experiments used to excite and convert triple-quantum
coherence for various orientations of crystallites or quadrupolar
The two-dimensional MQMAS experiment provides two
direct quantitative measures of the similarity of the spectra, again
providing strong evidence that this predominant reaction product
is anhydrous Na2CO3. By measuring the average frequencies
in both MQMAS spectral dimensions (here designated as V1′
and V2) we calculate the isotropic quadrupolar and isotropic
chemical shift parameters (VQ0 and VCS, respectively).12 Table 1
lists the average frequencies in both dimensions and the
calculated isotropic shift values for crystalline Na2CO3, and
values are compared with those from the sharp resonance for
the R ) 2.5 glass. The average frequencies are within 10% of
one another in both cases, and discrepancies in the numbers
may be explained by the difficulty of determining an accurate
center of gravity for the sharp resonance in the isotropic
dimension of the glass sample.
Additional analytic techniques also indicate that Na2CO3 is
present in the R ) 2.5 glass sample after exposure to atmosphere

9584 J. Phys. Chem. B, Vol. 104, No. 41, 2000
Egan and Mueller
Figure 6. The XRD pattern for the R ) 2.5 glass exposed for eight
months displays minor peaks assignable to crystalline Na2CO3. The
arrows indicate positions that have been computer-chosen as peaks via
an algorithm supplied with the diffractometer system.
for eight months. Powder X-ray diffraction studies of the glass
samples, both exposed and unexposed, contribute little under-
standing to the bulk glass structure because of the absence of
crystallinity. However, the XRD pattern for the R ) 2.5 glass
after exposure for eight months (see Figure 6) does indicate
some evidence of a small amount of crystalline material. The
peaks identified in Figure 6 were chosen automatically by the
XRD software package, and two of the computer-chosen peaks
correspond to peaks that exhibit strong intensity for crystalline
Na2CO3. While the 2.61 and 2.96 Å peaks are found for Na2CO3,
the absence of any other peak match indicates that the
concentration of the Na2CO3 phase is still at or beyond the
detection limits of the technique, or possibly that the phase is
not completely crystalline. The monohydrate form of Na2CO3
provides an XRD pattern with crystalline peaks that occur at
smaller angles than those observed for the glass sample,
imparting further support for an anhydrous assignment. XRD
patterns for the R ) 1.5 and 2.0 glasses (8 months exposure
time) do not reveal any detectable crystalline peaks, demonstrat-
ing that in this instance the NMR methods are more sensitive
than XRD.
Figure 7. A major difference between the FTIR spectra of the (a)
fresh and (b) corroded glass (with R ) 2.5) is the relative intensity of
the peak at 1430-1460 cm-1. Pure Na2CO3 exhibits a very intense
stretching vibration at these wavenumbers.
detection of sodium environments near to (or in the same phase
as) H2O, and thus will provide additional data for identifying
the particular species present in the corroded glass samples.
Two conclusions can be made based on the 1H f 23Na
CPMAS spectra of Figure 8. The first conclusion supports the
assignment of the first site at +2880 kHz to an anhydrous Na2-
CO3 phase. In the CPMAS spectrum from the R ) 2.5 glass
(Figure 8a), no resonance peak is observed around +1.0 ppm,
indicating a lack of proton or H2O interactions with the sodium
in the putative Na2CO3 phase. The second conclusion is the
assignment of the other corrosion product to a sodium metabo-
rate monohydrate compound. Comparison of the CPMAS
spectrum of the glass to the CPMAS spectra of NaBO2‚1H2O
and Na2SiO3‚5H2O (Figure 8b,c, respectively) illustrates the
superior match between the glass and borate resonances. The
frequency shift of maximum intensity occurs at identical
positions in both spectra 8a,b, and the line shapes of both
2-
The presence of the CO3 anion in a sample of corroded
glass could also be expected to produce modification of the
FTIR spectrum of the sample after exposure, revealing an
additional stretching vibration from the CO32- anion. The main
change in the FTIR spectra depicted in Figure 7 between
exposed and unexposed glass samples (R ) 2.5) is the increased
intensity in the region of 1430-1460 cm-1. This peak is
consistent with the presence of Na2CO3, based on the large
spectral intensity observed between 1400 and 1500 cm-1 for
crystalline Na2CO3. No other significant differences are seen
in the FTIR spectra for the two R ) 2.5 glass samples.
The assignment of the second resonance (at +5000 Hz in
the isotropic dimension) presented a more difficult challenge.
Two crystalline compounds possessed resonances with very
similar frequency shifts in the MQMAS isotropic dimensions
that coincide with the +5000 Hz shift observed for the glass
sample. Both potential candidates, Na2SiO3‚5H2O and NaBO2‚
1H2O, are hydrated compounds. Comparison of the MAS slice
of the glass sample at +5 kHz with the 23Na MAS spectra of
the two candidates is extremely difficult because of the
contribution of the broad sodium background present from the
glass that obscures the true line shape of interest. Because both
potential compounds share the characteristic of being hydrated,
1H f 23Na CPMAS NMR would in principle enable the
1
samples coincide closely with one another. The H f 23Na
CPMAS spectrum of Na2CO3‚1H2O and the MAS spectrum of
NaBO2‚1H2O are shown in Figure 8d,e for further comparison.
Estimation of the relative percentage of glass that is converted
into the major Na2CO3 phase in these samples is documented
in Table 2. The glasses with R < 1.5 are not included in this
table because no indication of any new sodium environments
is observed. Percentages were determined by deconvolution of
the full projections in the isotropic dimension into three peaks:
one peak corresponding to the Na2CO3 phase, one for the
NaBO2‚1H2O phase (especially prominent at short exposure
times), and a third peak, which is considerably broader,
representing the sodium remaining in the bulk glass network.
The percentages given in Table 2 are not absolute percentages
of Na2CO3 because the MQMAS experiment is not perfectly

Corrosion of Sodium Aluminoborosilicate Glasses
J. Phys. Chem. B, Vol. 104, No. 41, 2000 9585
observing the disappearance of resonances is useful for deter-
mining the types of sodium environments most susceptible to
aqueous attack.
It is commonly believed that two reaction schemes are
involved in the atmospheric attack of the aluminoborosilicate
glasses considered in this study. Ion-exchange reactions take
place between hydrogen and sodium cations involved in the
charge balance of the nonbridging oxygens of the silicate
network. This aqueous attack on the structural network does
not appear at significant (detectable) levels until the glass
samples contain alkali contents of greater than R ) 1.0. Sodium
oxide initially associates with the aluminate and borate networks,
facilitating conversion to tetrahedral units. Formation of a
hydrous layer containing Na+ and OH- ions results from the
atmospheric water vapor attack. Secondary reactions in the basic
surface layer begin to occur with corrosive gases such as CO2
and progress to form carbonate species:
CO2 + M+ + OH- a HCO3- + M+
HCO3- + 2M+ + OH- a CO32- + 2M+ + H2O (2)
2M+ + CO32- a M2CO3
(1)
Figure 8. (a) The 1H f 23Na CPMAS spectrum of the R ) 2.5 glass
(after eight months of exposure) is shown, and compared to (b) the 1H
f 23Na CPMAS spectrum of NaBO2‚1H2O. These resonances are very
similar in position and line shape, arguing strongly for the assignment
of this hydrated species. (c) The 1H f 23Na CPMAS spectrum of
Na2SiO3‚5H2O is shown for comparison, as is (d) the 1H f 23Na
CPMAS spectrum of Na2CO3‚1H2O. (e) The 23Na MAS spectrum of
NaBO2‚1H2O is also presented for completeness, and compares well
with spectrum (a). The dashed lines are drawn as guides for the eye.
(3)
The fact that the Na2CO3 phase is anhydrous indicates that H2O
is not retained by this phase and instead continues to contribute
to glass degradation.
Hydrolysis of the network is also occurring in the basic layer
by a nucleophilic attack on borosilicate linkages:
TABLE 2: Relative Percentages of Na2CO3 Phase in
Modified Glasses from Deconvolution of Full Projections of
the MQMAS Experiments
R content
(exposure time)
freq. of max.
intensity (Hz)
% of 23Na MQMAS signal
in putative Na2CO3 phasea
1.5 (8 months)
2.0 (8 months)
2.5 (2 weeks)
2.5 (8 months)
+2841
+2976
+2386
+2945
2.0
3.4
2.3
11.5
and the NMR experiments presented here indicate that sodium
metaborate monohydrate is the main product of these hydrolysis
reactions.
a Estimated error of 10% in reported values.
quantitative using our experimental conditions. The important
purpose of these ratios is to illustrate the change in the amount
of Na2CO3 phase between glasses measured under identical
experimental conditions.
Although both degradation processes are proceeding during
the first two weeks of exposure, as demonstrated by the presence
of both new resonances in Figure 3b, it is also observed that
ion-exchange is the more dominant mechanism after longer
exposure times, while the primary hydrolysis product (NaBO2‚
1H2O) appears to reach a maximum concentration at shorter
exposure times. This latter conclusion is reached from Figure
3 by comparing the relative intensity of the NaBO2‚1H2O
resonance peak to the intensity of the broad component present
for sodium remaining in the glass network (which remains
relatively constant).
Sodium resonances that disappear upon interaction with the
atmosphere can also be tracked by inspection of Figure 3.
Evident in these full projections is a loss of signal intensity for
23Na nuclei with total frequency shifts that appear at more
negative values than that of the sharp resonance at +2880 Hz.
Spectral intensity at these frequencies represent sodium environ-
ments that are initially present in structural units correlated with
nonbridging oxygens on silicon, and these environments are
expected to be destroyed by the corrosion reactions. Aside from
the loss in intensity of the broad component due to sodium
remaining in the glass network, there is a gain in 23Na signal
Many other NMR studies have been performed on a variety
of glass systems investigating the effects of H2O in the structure.
Most often 1H MAS NMR,30-33 or a related technique, has been
utilized to probe the hydrogen environments. An unfortunate
experimental shortcoming of 1H MAS NMR is a small chemical
shift range, further complicated by the broad resonances often
encountered. In many systems, resonances from molecular H2O
can dominate and obscure the remaining resonances in the
1
frequency spectrum. Also, H MAS spectroscopic techniques
are limited to investigating structures that are created by
corrosion. This disadvantage is common to many analytical tools
in that only the created structures are detected or observed.
Because corrosion is a dynamic process, it is vital for a complete
understanding to simultaneously detect the nature and number
of structures created and destroyed in the process. Probing the
23Na environments accomplishes a 2-fold purpose in one
experiment. Observing the appearance of resonances can first
provide information on the type of corrosion products formed
from sodium cations removed from the glass network. Second,

9586 J. Phys. Chem. B, Vol. 104, No. 41, 2000
Egan and Mueller
(2) Rudd, G. I.; Garofalini, S. H.; Hensley, D. A.; Mate, C. M. J. Am.
Ceram. Soc. 1993, 76, 2555.
(3) Schmitz, I.; Schreiner, M.; Friedbacher, G.; Grasserbauer, M. Anal.
Chem. 1997, 69, 1012.
(4) Carter, M. M. C.; McIntyre, N. S.; King, H. W.; Pratt, A. R. J.
Non-Cryst. Solids 1997, 220, 127.
intensity at large positive values of frequency shift for the glasses
after atmospheric exposure. The positive frequency shift may
be a signature of modified structural units in the glass that still
contain sodium cations in close proximity.
Conclusions
(5) Tichane, R. M. Glass Technol. 1966, 7, 26.
(6) Schreiner, M. Glasstech. Ber. 1988, 61, 223.
The MQMAS NMR technique applied to 23Na nuclei provides
resolution and identification of new sodium environments
formed from sodium aluminoborosilicate glasses exposed to a
common laboratory atmosphere. The appearance of narrowed
resonance lines in the isotropic MQMAS dimension leads to
the identification of two corrosion products, both containing
sodium in more-ordered phases than the glass samples. The
enhanced sensitivity of the MQMAS NMR technique in these
studies allows for detection of corrosion products before
noticeable changes can be observed in MAS or diffraction
experiments. Comparisons with MQMAS spectra from crystal-
line sodium-containing materials indicate that one reaction
product in the glasses, predominant at long exposure times, is
Na2CO3. 1H f 23Na CPMAS experiments also provide evidence
that the second phase, formed in greater quantities at shorter
times, is NaBO2‚1H2O. Both the MQMAS and CPMAS data
presented are inconsistent with the formation of detectable
quantities of other possible or probable reaction products,
including hydrated carbonate species. As the corrosion reaction
products are formed, the observed disappearance of sodium
environments associated with nonbridging oxygens confirms that
sodium nuclei in these environments are most susceptible to
attack and removal. Heteronuclear connectivity experiments
coupled with MQMAS may provide a more potent tool to fully
characterize these systems, and such investigations are currently
in progress.
(7) Bunker, B. C.; Tallant, D. R.; Kirkpatrick, R. J.; Turner, G. L. Phys.
Chem. Glasses 1990, 31, 30.
(8) Loshagin, A. V.; Sosnin, E. P. Glass Phys. Chem. 1994, 20, 14.
(9) Kohn, S. C.; Smith, M. E.; Dirken, P. J.; Eck, E. R. H. v.; Kentgens,
A. P. M.; Dupree, R. Geochim. Cosmochim. Acta 1998, 62, 79.
(10) Frydman, L.; Harwood: J. S. J. Am. Chem. Soc. 1995, 117, 5367.
(11) Fernandez, C.; Amoureux, J. P. Solid State NMR 1996, 5, 315.
(12) Medek, A.; Harwood, J. S.; Frydman, L. J. Am. Chem. Soc. 1995,
117, 12779.
(13) Amoureux, J.-P.; Fernandez, C.; Frydman, L. Chem. Phys. Lett
1996, 259, 347.
(14) Hanaya, M.; Harris, R. K. J. Phys. Chem. A 1997, 101, 6903.
(15) Hwang, S.-J.; Fernandez, C.; Amoureux, J. P.; Han, J.-W.; Cho,
J.; Martin, S. W.; Pruski, M. J. Am. Chem. Soc. 1998, 120, 7337.
(16) Stebbins, J. F.; Lee, S. K.; Oglesby, J. V. Am. Miner. 1999, 84,
983.
(17) Wang, S. H.; Stebbins, J. F. J. Am. Ceram. Soc. 1999, 82, 1519.
(18) Brown, S. P.; Wimperis, S. J. Magn. Reson. 1997, 128, 42.
(19) Vosegaard, T.; Larsen, F. H.; Jakobsen, H. J.; Ellis, P. D.; Nielson,
N. C. J. Am. Chem. Soc. 1997, 119, 9055.
(20) Wu, G.; Rovnyank, D.; Sun, B.; Griffin, R. G. Chem. Phys. Lett.
1995, 249, 210.
(21) Wu, G.; Rovnyak, D.; Griffin, R. G. J. Am. Chem. Soc. 1996, 118,
9326.
(22) Brown, S. P.; Wimperis, S. J. Magn. Reson. 1997, 124, 279.
(23) Fyfe, C. A.; Altenschildesche, H. M. z.; Skibsted, J. Inorg. Chem.
1999, 38, 84.
(24) Hanaya, M.; Harris, R. K. J. Mater. Chem. 1998, 8, 1073.
(25) Brow, R. K.; Alam, T. M.; Tallant, D. R.; Kirkpatrick, R. J. MRS
Bull. 1998, 23, 63.
Acknowledgment. Acknowledgment is made to the donors
of the Petroleum Research Fund, administered by the American
Chemical Society, for partial support of this research. This report
is also based upon work supported by the National Science
Foundation under Grant No. DMR-9458053 and Grant No.
DMR-9413674. The authors also acknowledge the support of
the Henry Dreyfus Teacher-Scholar Awards Program. Amy
Barnes, Dr. Carlo Pantano, and Dr. Mary Strzelecki are also
acknowledged for their helpful discussions.
(26) Ernst, R.; Bodenhausen, G.; Wokaun, A. Principles of NMR in One
and Two Dimensions; Oxford University Press: New York, 1987.
(27) Sorensen, O. W.; Rance, M.; Ernst, R. R. J. Magn. Reson. 1984,
56, 527.
(28) Amoureux, J.-P.; Fernandez, C.; Steuernagel, S. J. Magn. Reson.
A 1996, 123, 116.
(29) Galactic Industries Corporation, 395 Main Street, Salem, NH 03079.
(30) Kummerlen, J.; Merwin, L. H.; Sebald, A.; Keppler, H. J. Phys.
Chem. 1992, 96, 6405.
(31) Zeng, Q.; Nekvasil, H.; Grey, C. P. J. Phys. Chem. B 1999, 103,
7406.
(32) Kohn, S. C.; Dupree, R.; Smith, M. E. Nature 1989, 337, 539.
(33) Eckert, H.; Yesinowski, J. P.; Silver, L. A.; Stolper, E. M. J. Phys.
Chem. 1988, 92, 2055.
References and Notes
(1) Rana, M. A.; Douglas, R. W. Phys. Chem. of Glasses 1961, 2, 196.