Ind. Eng. Chem. Res.
2010, 49, 6287–6290 6287
The Determination of Sodium Silicate Composition Using ATR FT-IR
James S. Falcone, Jr.,*,† Jonathan L. Bass,‡ Maria Angelella,†,§ Emily R. Schenk,†,| and
Karen A. Brensinger†,⊥
Department of Chemistry, West Chester UniVersity of PennsylVania, 229 Schmucker Science South,
West Chester, PennsylVania 19382
A procedure that measures both the silica concentration and the ratio of silica to alkali content of the
commercially important sodium silicate solutions using an ATR FT-IR spectrophotomer has been developed.
The accuracy and precision have been found to be equal to or better than the commonly used measures via
titrimetry and density determination.
Introduction viscosity, or refractive index. These relationships can be
expressed in charts, which can be used to interpolate values of
Commercial sodium silicates are water-soluble glasses gener-
the ratio (and %SiO2 by calculation) from measures of the
ally made from a very pure form of quartz sand and natural
physical property and %Na2O by titration. The most common
sodium carbonate in an open hearth regenerative furnace via
charts (commonly called AGV charts) rely on the use of density
the reaction:
measurements, using high precision gravity hydrometers, and
percent alkali by titration. Historically, these charts were drawn
Na2CO3(s) + rSiO2(s) f Na2O · rSiO2(s) + CO2(g) using density results generated internally or available in the
literature; see, for example, the discussion and data in Vail5 for
The composition is generally reported as the weight percent solutions ranging in ratio between 2 and 3.9 and %SiO2 between
alkali (%Na2O) and the ratio of SiO2 to Na2O on a weight basis. 0 and 30%. In the last 20 years, these charts have been replaced
This weight ratio is often confused with the mole ratio because by computer models,6 which streamline the calculation of
they are close in value with the mole ratio ) 1.032*weight ratio. solution composition, but the measurements still require sig-
Knowledge of the ratio of a sodium silicate is very important nificant time and are not easily automated.
in understanding the chemistry of these materials. Molten Recently, while studying the relationship between various
compositions of sodium silicate of any ratio can be prepared; instrumental methods of characterizing the nature of sodium
however, only those with values between 1.6 and 3.8 are usually silicate solutions, over 60 different silicate solutions were
available commercially in solution form. Manufacturers sell prepared and analyzed. Our studies identified key IR bands near
either concentrated aqueous solutions, glass lumps that can be 766, 874, 1006, 1051, and 1110 cm-1 and suggested that the
dissolved on site in suitable pressure dissolvers, or hydrous band near 1000 cm-1 is highly sensitive to the chemical
powders, which can be readily dissolved at atmospheric pressure. environment.7 Other applications of FT-IR to silicate solutions,8-15
The commercial aqueous solutions of the sodium silicates the increase in use of ATR FT-IR for in-process monitoring of
are colloidal in nature and contain complex silicate species, Na+ chemical processes over the last 10 years,16,17 and regularities
ions, and OH- ions.1 They are analyzed2 routinely using a in the FT-IR spectra of different solutions18 suggested that this
density measurement with a hydrometer and classical titrimetric technique might be useful in improving the ability to more
methods appropriate for a weak base3 to determine the alkali rapidly measure silicate solution composition. We observed that
content. Generally, standardized 0.2 M HCl is used as the titrant certain trends seen earlier by Bass and Turner were more regular
to a methyl orange end point, and the result is reported as than anticipated. The observed linearity suggested that a method
%Na2O with a relative precision 0.1%. The silica content can could be developed to quantitatively measure the concentration
be directly measured titrimetrically or gravimetrically. The and ratio of sodium silicate solutions without resorting to a
former procedure requires great skill because the end point is titrimetric method, possibly using an automated combination
difficult to determine and HCl(aq) and NaF(s) are used and the of density and IR measurements.
latter procedure is rather tedious and time-consuming as
described by Vail.4 Thus, on a routine basis, it has been Experimental Methods
determined that one can achieve sufficient accuracy utilizing
relationships between %SiO2, %Na2O, and various readily The sodium silicate solutions used in this study were
measurable physical properties of the solutions such as density, prepared with varying ratios of SiO2 to Na2O (w/w) and
concentrations of silica expressed as SiO2 molality (mol
* To whom correspondence should be addressed. E-mail: jfalcone@ kg-1). The ratios varied between 1 and 3.5, and concentra-
wcupa.edu.
† tions ranged from 1.6 to 8.8 mol kg-1 (8.8-30%) SiO2. All
West Chester University of Pennsylvania.
‡
Bass Chemical Consulting, 39 Lee Road, Audubon, Pennsylvania samples were aged for a minimum of 3 days after sample
19403. preparation to ensure that a steady state was reached. Steady
§
Currently at Department of Chemistry and Biochemistry, University state was identified by the change in the FT-IR spectrum over
of California, San Diego, Urey Hall 3118, 7835 Trade St., Ste 100, time. Samples were prepared from different sources of silica,
San Diego, CA 92121. including commercial sodium silicates solutions, silica sols
|
Currently at Department of Chemistry and Biochemistry, Florida
International University, 11200 SW 8th Street, Miami, FL 33199. (LUDOX HS30 30% silica sol and NYACOL NexSil 5000HP
⊥
Currently at U.S. Army Forensic Toxicology Drug Testing Lab, high purity 15% 4-6 nm silica sol), ACS reagent grade
2490 Wilson Street, Fort George G. Meade, MD 20755-5235. NaOH, and deionized water. All samples once prepared were
10.1021/ie1002747 2010 American Chemical Society
Published on Web 06/22/2010
6288 Ind. Eng. Chem. Res., Vol. 49, No. 14, 2010
Figure 3. The variation in the location of the peak near 1000 cm-1 with
ratio at constant silica concentration 2.0 m (10.2% SiO2).
Figure 1. Examples of FT-IR spectra qualitatively illustrating changes
with SiO2 molality at a constant ratio of 2.58: - 8.22 mol kg-1 (29.3%)
SiO2; · · · 6.16 mol kg-1 (24.5%) SiO2; - - - 4.9 mol kg-1 (20.9%)
SiO2; - · - 4.11 mol kg-1 (18.4%) SiO2; - · · - 3.09 mol kg-1 (14.8%)
SiO2.
Figure 4. The variation in the values of βr with ratio.
find peak tool. A typical graphical result is presented in the
Supporting Information along with the silicate concentrations
Figure 2. The variation in the location of the peak near 1000 cm-1 with and OMINC data for all solutions used in this study.
SiO2 concentration: ], 3.45 ratio sodium silicate; ×, 2.94 ratio sodium
silicate; 2, 2.517 ratio sodium silicate; and 9, 2.08 ratio sodium silicate. The samples used in the statistical test of the reliability of
the model were analyzed in the classical manner. The % alkali
or Na2O was determined using standardized 0.2 M HCl to titrate
kept sealed in plastic capped vials. One drop samples were a 2-3 g sample of the silicate solution diluted with about 20
run on a Nicolet Protégé 460 Fourier transform infrared mL of DI water to the methyl orange end point. Densities were
spectrometer (FT-IR) equipped with a Smiths Detection determined using VWR precision gravity hydrometers following
DuraSamplIR II ATR platform with the single reflection industry standard methods.19
Duradisk diamond window accessory. OMNIC FT-IR soft-
ware was used for data acquisition and analysis. All samples
were run at room temperature. The acquiring of data included Results
the analysis and subtraction of a deionized water spectrum In Figure 1 is shown the expansion of the 1250-750 cm-1
as background, 32 scans per sample in the wavenumber range range for different silica concentrations at a constant ratio of
of 4000-720 cm-1, while measuring the absorbance of the SiO2 to Na2O. Peak intensities and total area increase and the
sample. The resolution factor for the sampling parameters peak location for the major peak shifts to lower energy as the
established was 4 cm-1; all other possible experiment setup concentration of silica increases.
changes were left at the OMNIC stipulated default settings.
The linearity of the shift in the location of the maximum FT-
The diamond stage was cleaned between runs with dilute
aqueous sodium hydroxide and deionized water. The spectra IR band, ω(m,r), near 1000 cm-1 with SiO2 concentration at
were not corrected for the variation in penetration depth due constant ratio is shown more clearly in Figure 2. It became
to wavelength and sample index of refraction. The spectra apparent that equations for this trend of the type shown in eq 1
were analyzed without corrections. The area under the could be written.
absorption spectra in the characteristic region starting at 719
cm-1 and ending where the absorbance returns to 0.0 ω(m, r) ) Rr + βrm (1)
(approximately 1250 cm-1) was measured using the OMNIC
area tool. The location of the absorbance near 1000 cm-1 As shown in Figure 3, it was observed that the plot of location
(the maximum absorbance) was determined using the OMNIC of the maximum band varied linearly with the silicate ratio at
Ind. Eng. Chem. Res., Vol. 49, No. 14, 2010 6289
for {δω/δm}r or βr, slopes of plots of ω(m,r) versus m at
constant ratio, versus ratio.
Using the results from standard solutions prepared in a wide
ratio range at a constant silica molality of 2.0, we were able to
accurately predict the location of the maximum peak for a series
of solutions made from commercial silicates at much higher
concentrations using eq 4.
ω(m, r) ) ω(2.0, r) + (m - 2.0) { δm
δω
}r (4)
Krumrine and Falcone20 had earlier developed models requir-
ing five parameters that allowed the accurate calculation of
silicate ratio and concentration from density, F, and concentra-
tion measurements in a range similar to the one in this study.
In one model, the density expressed as the apparent molal
Figure 5. The absorption area in the range 1250-719 cm-1 at different volume,21 ΦV, was fit linearly to molality as generally expected
silica concentrations. for a colloidal solution, but the value of ΦVo varied linearly with
ratio, and the value of the second coefficient varied as a second-
Table 1. Results of the Multiple Regression Analysis on All Data order polynomial with ratio as shown in eq 5.
above a Ratio Value of 1.9
coefficient in eq 6 value std error ΦV(m, r) ) ΦVo(1 + ar) + (b + cr + dr2)m (5)
b0 1008.669 9.2
b1 -0.038116 1.2 where ΦV is the apparent molal volume, which is a function of
b2 -0.042292 0.025 F, r is the ratio, m is the molality of silica, mol kg-1, and ΦVo,
b3 4.318224 6.8 a, b, c, d are fitted coefficients.
b4 0.357498 0.14
b5 -1.013738 0.19 Thus, it seemed possible that a multiple linear equation of
the type shown in eq 6 might adequately fit our FT-IR data
Table 2. Reproducibility of FT-IR Results and Model Predictions over the range of concentration and ratio.
for SiO2 Concentration and SiO2:Na2O Ratio for 10 Trials on a
Known Sample with 3.43 Ratio and 6.00 mol kg-1 (24.6%) SiO2
ω(m, r) ) b0 + b1r2 + b2Ar2 + b3r + b4Ar + b5A (6)
-1
results ω, cm area molality, model ratio, model
1 1009.0 48.95 5.99 3.42 We found that we could substitute A for m because we observed
2 1009.0 48.88 5.98 3.42 that the area, A, data collected for all of the solutions studied
3 1008.9 49.15 6.02 3.41 above a ratio value of 1.9 correlated fairly closely with molality
4 1009.1 48.45 5.93 3.41 independent of ratio as shown in Figure 5 and that eq 6 provided
5 1009.1 48.97 5.99 3.43
6 1009.2 48.63 5.95 3.43 good estimates of ω(m,r) for all of the spectra in our portfolio
7 1009.1 48.77 5.97 3.43 from a ratio equal to 1.9 and above.
8 1009.2 48.9 5.98 3.44 The bn coefficients for eq 6 were derived using multiple
9 1009.3 48.66 5.95 3.45 regression analysis on the results for the 55 different silicate
10 1009.0 48.97 5.99 3.42 solutions; the standard error was 0.612, and the value of R2 for
average 1009.1 48.8 5.98 3.43
std deviation 0.1 0.2 0.025 0.013 prediction was 0.995. The full statistical output for this fit can
relative error, % 0.33 0.081 be found in the Supporting Information. The values for these
precision, % 0.42 0.37 coefficients are shown in Table 1.
The ratio may be determined when eq 6 is rewritten in a
several different concentrations, suggesting that representations quadratic form in r. Thus, one may directly deduce silicate ratio
like eq 2 for ω(m,r) were also reasonable. and silica molality from measures of the area and location of
the maximum peak in a simple FT-IR run. The correct solution
ω(m, r) ) Rm + βmr (2) for the value of ratio for our results was:
Finally, as shown in Figure 4, it was observed that we could -b + √b2 - 4ac
write a second-order polynomial expression of the type shown r)
2a
in eq 3:
where a ) [b2A + b1] ) -0.0386116 - 0.042292A; b )
{ }δω
δm
r ) εr2 + δr + φ (3) [b4A + b3] ) 4.318224 + 0.357498A; and c ) b0 + b5A -
ω(m,r) ) 1008.669 - 1.013738A - ω(m,r).
Table 3. Results of Measurements on Commercial Silicates from Around the Worlda
m calculated, % Na2O, ω, measured, m model,
sample mol kg-1 ratio calc by titration cm-1 area, measured mol kg-1 mrelative error % ratio model ratio relative error %
1 6.00 3.43 7.17 1009.1(1) 48.8(2) 5.98(3) 0.0 3.43(1) 0.0
2 7.69 3.42 8.46 1005.6(2) 60.8(6) 7.42(8) -3 3.41(4) -0.3
3 7.77 3.36 8.67 1004.9(3) 61.1(6) 7.47(7) -4 3.34(2) -0.6
4 6.39 3.52 7.31 1009.1(2) 51.4(6) 6.27(7) -2 3.52(4) 0.0
a
The values in parentheses are standard deviations for these results, that is, 5.98(3) ) 5.98 ( 0.03 sd.
6290 Ind. Eng. Chem. Res., Vol. 49, No. 14, 2010
Molality was calculated from the value of ω(m,r) and the Supporting Information Available: The raw data and model
calculated ratio using eq 7. results for each solution used in this study, along with a sample
spectrum showing the location of the maximum peak. Details
ω + 0.96245r2 - 996.056 - 11.2132r of the regression results for eq 6 using ω(m,r), measured ratio,
m) (7) and peak area in the first fit, and ω(m,r), measured molality,
-7.1680 + 2.3835r - 0.2860r2
and the model predicted ratio for the second fit. This material
is available free of charge via the Internet at http://pubs.acs.org.
Equation 7 was obtained from a multiple regression of the eq
6 using the actual molality and the model predicted values of r Literature Cited
for each solution. The full statistical output for this fit can be
found in the Supporting Information. When these algorithms (1) Falcone, J. S., Jr. Silicon Compounds: Anthropogenic Silicas and
were applied to our data, it was observed that the standard Silicates. Kirk-Othmer: Encyclopedia of Chemical Technology, 5th ed. -
Online Version; Wiley: New York, 2005.
deviations of predicted ratios and concentrations were 2.6% and (2) Bass, J. L. Modern Instrumental Methods for Analysis of Soluble
5.0%, respectively, over all ratios and concentrations measured Silicates. In Soluble Silicates; Falcone, J. S., Jr., Ed.; ACS Symposium
above a ratio value of 1.9. No attempt was made to determine Series; American Chemical Society: Washington, DC, 1982; pp 17-30.
instrument to instrument variations; however, it is expected that (3) Christian, G. D. Analytical Chemistry, 5th ed.; Wiley & Sons, Inc.:
New York, 1994; pp 223-236.
this procedure will have to be calibrated for each instrument (4) Vail, J. G. Soluble Silicates; Reinhold: New York, 1952; Vol. 1, pp
and application. 33-43, 78-81.
As a final test of the usefulness of the method to measure (5) Vail, J. G. Soluble Silicates in Industry; ACS Monograph No. 46;
ratios of solution, we randomly measured four well-characterized Reinhold: New York, 1928; pp 128-132.
commercial silicates, from three different continents, 10 times (6) Krumrine, P. SILICA - an unreleased proprietary software package
developed for The PQ Corp. to automate interpolation of a variety of sodium,
over 2 days. Table 2 illustrates typical data and the relative error potassium, and lithium silicate properties given any two of the following:
and precision of the model results as well as the sampling % alkali, % silica, % solids, ratio, or density (initially developed and
capabilities of a FT-IR with respect to sodium silicate solutions. functionally expanded during the early 1980s).
The known values of concentration and ratio were determined (7) Falcone, J. S., Jr.; Bass, J. L.; Krumrine, P. H.; Brensinger, K.;
Schenk, E. R. J. Phys. Chem. A 2010, 114, 2438.
from the measured value of %Na2O (alkali), carefully deter- (8) Dalby, K. N.; King, P. L. Am. Mineral. 2006, 91, 1783.
mined densities, and the known relationships between density, (9) Osswald, J.; Fehr, K. T. J. Mater. Sci. 2006, 41, 1335.
%Na2O, and ratio. The molality of each silicate was calculated (10) Halasz, I.; Agarwal, M.; Li, R.; Miller, N. Catal. Lett. 2007, 117,
once the %SiO2 was determined from these two values. 34.
Table 3 is a summary of the results for the model predicted (11) Medvedev, E. F.; Komarevskaya, A. Glass Ceram. 2007, 64, 7.
(12) Dutta, P. K.; Shieh, D. C. Zeolites 1985, 5, 135.
ratio and molality, and their percent variations from the (13) Couty, R.; Fernandez, L. J. Chim. Phys. 1998, 95, 384.
measured values are shown in the last four columns. (14) Halasz, I.; Agarwal, M.; Li, R.; Miller, N. Catal. Today 2007, 126,
196.
(15) Fukuhara, M.; Kohuta, A. CryoLett. 2005, 26, 251.
Conclusion (16) Liotta, V.; Sabesan, V. Org. Process Res. DeV. 2004, 8, 488–494.
(17) Andanson, J.; Kazarian, S. G. Macromol. Symp. 2008, 265, 195-
These results show that an ATR FT-IR can be used to 204.
determine the ratio of a commercial 3.4 ratio silicate solution (18) Bass, J. L.; Turner, G. J. Phys. Chem. 1997, 101, 10638.
with a relative error of better than 1%. The silica concentration (19) The PQ Corp. will, on request, provide procedures for the analyses
procedures used in the manufacture of sodium silicates.
can also be measured to better than 4% of the value measured
(20) Krumrine, P.; Falcone, J. S., Jr. Unpublished results of work
titrimetically. These results are comparable to those obtained performed at The PQ Corp. during the 1980s. At that time, it was
using the standard alkali and gravity determination. demonstrated that multiple linear regression models, based on roughly 500
measurements found both within PQ Corp. and the open literature, were
capable of replacing the manual AGV charts used at the time with similar
Acknowledgment accuracy. Both the charts and the computer model calculated the ratio of
the silicate from the measured values of density and % alkali. The algorithms
This work was made possible by the financial assistance are currently in use throughout the world as a process control tool as part
provided by Rhodia Silcea. Their purchase of FT-IR accessories of the SILICA program.
as well as other supplemental lab equipment needed was (21) Harned, H. S.; Owen, H. The Physical Chemistry of Electrolyte
Solutions, 2nd ed.; Reinhold: New York, 1950; p 250.
essential for this research. We are especially thankful to Kirk
Thompson and Joel Racinoux of Rhodia for their helpful ReceiVed for reView February 3, 2010
suggestions and support during this study. We acknowledge ReVised manuscript receiVed May 17, 2010
Accepted June 10, 2010
West Chester University of Pennsylvania for providing the
FT-IR and the laboratory space for this work. IE1002747
