Introduction of silver into a silicate glass often leads to reduction and coloration. The electron requiredfor silver reduction is extracted from a nonbridging oxygen atom. The use of high-field-strength ionslimits the number of nonbridging oxygens and silver reduction.
IntroductionThe production of spatial patterns in variations inrefractive index has found increased use in opticalengineering in recent years. One technique for pro-ducing variations in the refractive index is the ex-change of ions with different polarizabilities. In thisprocess a univalent ion in a glass object is usuallyexchanged for another univalent ion from a salt bath.Frequently one alkali is exchanged for another.Modest changes in refractive index are achieved bysuch an exchange. The exchange of thallium for analkali produces large changes in refractive index.The poisonous nature of thallium has limited theextent to which this technique is used. The ex-change of silver for an alkali produces a change inrefractive index that is comparable with that pro-duced by the thallium exchange without the introduc-tion of toxicity problems. The potential advantagesof this exchange have not been realized fully becausethe introduction of more than minimal amounts ofsilver into a silicate glass by ion-exchange techniqueshas often led to extensive chemical reduction of thesilver. 3 The intense color that characterizes theformation of colloids when silver is reduced is unac-ceptable for most applications.
Polyvalent impurities such as arsenic or tin canprovide electrons leading to the reduction of a smallamount of silver and consequently can cause a smalldegree of coloration. It is unrealistic, however, tobelieve that small levels of impurities can be responsi-ble for the considerable coloration observed in alkali-silicate glasses on the introduction of modest amountsof silver. In this paper extraction of an electronfrom atoms that are intrinsic to the glass is suggestedas an alternative mechanism of silver reduction.
ModelIn the present model electrons are extracted from thevalence band. The holes are assumed to be trappedby the evolution of molecular oxygen rather than bynetwork relaxation. In other words the process isdiscussed in terms of the reversible chemical reactionbetween the metal, silica, and oxygen shown in Fig. 1.
Since large amounts of thallium or alkali-metaloxides can be incorporated in silicate glasses withoutchemical reduction, the equilibrium must lie far tothe right when any of them serves as the monovalentmetal. The observation of reduced silver indicatesthat the equilibrium lies considerably less far to theright when silver is one of the monovalent ions. Theparameter in the free-energy change for the reaction,which is most closely related to the nature of themetal, is the ionization potential. The values for themetals in question are shown in Table 1.
The highest ionization potential of the listed metalsthat are not reduced in silicates is that of thallium.The ionization potential of silver exceeds that by only1.5 eV. This suggests that, if the energy at the top ofthe valence band in some modified alkali-silicate glasscould be decreased by more than 1.5 eV, silver wouldnot be reduced in such a glass. Table 2 indicatesthat the band gap is much larger in silica than inalkali silicates.
Since metals with such low ionization potentialsseem unlikely to introduce states at the bottom of theconduction band, it is plausible that the oxygen atomsmodified by the presence of the monovalent ions (theso-called nonbridging oxygen atoms) introduce high-energy states at the top of the valence band. Thisinterpretation of the smaller band gap in the alkalisilicates is corroborated by the observation that whilethe ionization potentials of sodium and potassiumdiffer by 0.8 eV, the band gaps differ by only 0.2 eV.Furthermore, the larger band gap is observed for thesilicate of sodium, which has the larger ionizationpotential. If the width of the band gap were depen-
dent primarily on states introduced at the bottom ofthe conduction band by the monovalent metals, so-dium silicates should have a smaller band gap thanpotassium silicates.
The incorporation of high-field-strength ions suchas A1+ 3 removes nonbridging oxygen (NBO) atoms inthe manner shown in Fig. 2. It is plausible thereforethat alumina might remove the most energetic elec-tron states since they are associated with NBO atoms.This expectation is confirmed by Sigel's measure-ments4 shown in Fig. 3. They indicate that the bandgap in fused silica doped with equal parts of aluminaand alkali is 1.4 eV larger than the correspondingglass with no alumina.
ExperimentalTo test the idea that high-field-strength ions such asaluminum remove the energetic NBO atoms andprevent silver reduction, silver was introduced by anion-exchange technique into several glasses contain-ing varying ratios of alumina to alkali. All theglasses listed in Table 3 were immersed in moltensilver chloride at a temperature of 5000C for 10 days.Diffusion of silver throughout the thickness of thesample is expected to be complete, and the amount ofsilver introduced represents its equilibrium value.Since no concentration gradients are expected toresult from such a long immersion, the change inrefractive index as a function of the amount of silverintroduced is easily determined in transparent glasses.
Since one atom of aluminum is expected to removeone NBO atom from the glass, one can calculate fromthe composition the fraction of oxygen atoms that arebonded into the structure of an alkali-aluminosilicateglass as NBO. The expected fraction of NBO atoms
Table 2. Comparison of Band Gaps in Silica and Alkali Silicates
Glass
SiO 2 Na2 SiO3 K2SiO 3
Edge (eV) 10 6.0 5.8
is given by
NBO= -M2 0 - A1203
X SiO2 + 1.5 x A120 3 + 0.5 x M20
where all the concentrations are expressed in cationpercent. Table 3 shows qualitatively the anticorrela-tion between silver reduction and the fraction of NBOatoms in the glass.
The extensive reduction of silver in glass 1 wasindicated by a deep magenta color when the samplewas viewed in transmitted light and by a green colorwhen the sample was viewed in reflected light.Microscopic examination indicated the existence of aplethora of gaseous inclusions. Mass spectrometrydisclosed no evidence that the bubbles containedchlorine. Thus one can rule out the possibility thatchloride ions diffusing into the glass concurrentlywith the exchange of cations provide the electrons forthe reduction of silver. Consistent with the thesisthat the nonbridging oxygen atoms supply the elec-trons was the observation that the bubbles containedprimarily oxygen. The observation of small amountsof nitrogen in some of the bubbles has not beenexplained.
Glasses 2 and 3 contain fewer but nonethelessplentiful supplies of NBO atoms. After they hadbeen subjected to ion exchange, both samples wereblack. In glass 4 only slightly more than 1% of theoxygen atoms are of the nonbridging type. Ionexchange of this resulted in several small areas thatmanifested a red color that was much paler than thatseen in sample 1. Furthermore, no scattered greenlight was observed nor were any gas bubbles. Thelargest fraction of the glass displayed only a paleyellow color. Microprobe examination revealed that- 18% of the sodium originally contained in either
area of the glass was replaced by silver. In glass 5
Nonbridging Oxygenso o lI ' X 1 K ( 4 1
(1) O-Si-O-Si-O + K2 0 O-Si-o K+ o-si-OI I I Io o 0 0
Fig. 2. Structural units in alkali silicate glasses. This schematicrepresentation shows (1) the formation of NBO atoms by theaddition of K20 and (2) the formation of bridging oxygen atoms bythe addition of equal parts of K20 + A1203.
AEr is the energy of chemical reaction between an oxide ion and silica.
70
a 60
w 50
E0 40Cc
i 30C* 20
EL 10
7.7 ev 6.3 ev
180 200Wavelength (nm)
Fig. 3. Transmittance spectra of alkali silicate glasses. Theeffect of alumina on absorption edge in alkali-doped fused silica is(1) 0.2 mole % K2 0 + 0.2 mole % A1203 and (2) 0.2 mole % K2 0.
the density of NBO atoms is diminished still furtherand no coloration is caused by the ion exchange.
A glass with a silver content exceeding 23% byweight and with a refractive-index change exceeding0.05 was produced by performing the ion exchangefor five days at 650°C. An exchange employing thesame schedule but using a mixed salt bath composedof 70% by weight silver chloride and 30% silversulfate led to a glass in which 90% of the alkali wasreplaced. The refractive-index change exceeded 0.13.Both of these glasses were virtually colorless, butboth suffered considerable cracking as a result of thestress generated by the ion exchange.
Melting the silicate glasses in which the number ofNBO atoms is low requires high temperatures.Lower-melting glasses with few or no NBO atoms atthe temperature of ion exchange can be obtained bythe use of ions that are known to behave like alumi-num at low temperatures, in that they remove NBOatoms from alkali silicates but do not do so at hightemperatures. These ions include hafnium, tanta-lum, niobium, and zirconium, which, because of theirlarge positive charge, have a high field strength andremove NBO atoms from the system in spite of theirlarge size. Also included are the small ions such aszinc, beryllium, or magnesium that remove NBOatoms in the process of attaining tetrahedral coordina-tion.
The efficacy of magnesium, gallium, yttrium, lan-thanum, and niobium in glasses that are ion ex-changed in the low-temperature regime is docu-
Table 3. Influence of NBO Atoms on Color in Aluminosilicates
mented in Table 4. It is expected that, because oftheir low field strengths, the larger alkaline earthions do not remove NBO atoms. Rather they aresimilar to the alkalies in producing NBO atoms. Theyellow color observed in sample 2 is consistent withthis expectation.
Kinder5 confirmed the effectiveness of Ta2C 5 inproducing glasses that can be melted at relatively lowtemperatures and that contain no NBO atoms at thetemperatures maintained during ion exchange. Inglasses containing no alumina but containing equalnumbers of tantalum and alkali atoms, he obtainedan almost complete exchange of silver for alkaliwithout discoloration.
In alkali-borate glasses and in alkali-borosilicateglasses low levels of alkali do not introduce NBOatoms. Instead the boron enters the structure intetrahedral coordination much the way alumina does.Yun and Bray6 used nuclear magnetic resonance todetermine the density of NBO atoms in borosilicatesat room temperature. The density of NBO atoms inthese glasses is known to increase with tempera-ture. 1 2
Thus properly chosen borosilicate glasses are easilymelted and contain no NBO atoms at the tempera-tures used for ion exchange. Table 5 illustrates therelationship between the estimated fraction of NBOatoms and the color that is observed in borosilicateglasses. The first two glasses appear to the nakedeye to be completely colorless while an extremely paleyellow color can be detected in sample 3. A deeporange color is observed in sample 4. It is not knownwhether the shift in (0.5) in sample 2 is due topolyvalent ion impurities or to a slight error in theestimated value of NBO introduced by assuming thatthe temperature of ion exchange lies within the realmof the low-temperature limit. These results seem toconfirm the efficacy of boron in preventing the reduc-tion of silver introduced by ion exchange.
The ideal glass for producing a large refractive-index change by the ion exchange of a large amount of
Table 4. Influence of NBO Atoms on Color in Aluminosilicates
silver is probably a glass with sufficient boron oxide tosoften the glass so that it can be melted at lowtemperatures and so that stress relaxation at theion-exchange temperature is sufficient to preventcracking. In the ideal glass the boron concentrationshould not greatly exceed the minimum level requiredto achieve these objectives. Too great a concentra-tion of boron oxide may make the structure of theglass unduly sensitive to the temperature of the ionexchange and introduce variability into the results.
A glass comprising 35% SiO2, 27.5% A120 3, 7.5%B203, and 30% Na2O illustrates a nearly ideal compo-sition. This glass was exchanged for five days at650'C in a mixed silver chloride-silver sulfate bath.The index increased from 1.506 to 1.613. No crack-ing was observed. A slight yellow tint was observedafter the ion exchange. The wavelength at whichthe transmittance of a 1-mm-thick sample was mea-sured to be 50% was 280 nm before exchange and 368nm after exchange. No absorption peak at 400 nm,which would suggest the presence of colloidal silver,
was observed. In fact no indication of atomic silverwas disclosed by a spin resonance investigation. Itseems fairly certain therefore that no reduction ofsilver occurred when it was introduced to the glass bythe ion-exchange process.
SummaryIt has been noted in past studies that the introductionof appreciable concentrations of silver into an alkalisilicate glass by an ion-exchange process resulted incoloration. This has been ascribed to a reduction ofthe silver by polyvalent ions.3 The present studyindicates that reduction can also be accomplished bythe extraction of an electron from an NBO atom.Test glasses without NBO were prepared by incorpo-rating high-field-strength ions such as aluminum ortantalum. High concentrations of silver ions wereintroduced to such glasses without coloration. Achange in refractive index in excess of 0.1 was achievedwithout the formation of colloidal silver formationand without the formation of cracks.
References1. R. J. Araujo, "Statistical mechanical model of boron coordina-
tion," J. Non-Cryst. Solids 42, 209-230 (1980).2. R. J. Araujo, "Statistical mechanics of chemical disorder:
application to alkali borate glasses," J. Non-Cryst. Solids 58,201-208 (1983).
3. T. Findakly, "Glass waveguides by ion exchange: a review,"Opt. Eng. 24,244-250 (1985).
4. G. H. Sigel, "Vacuum ultraviolet absorption in alkali dopedfused silica and silicate glasses," J. Phys. Chem. Solids 32,2373-2383 (1971).
5. Y. H. Yun, and P. J. Bray, "Nuclear magnetic resonance studiesof the glasses in the system Na 2O-B203-SiO2," J. Non-Cryst.Solids 27,363-380 (1978).
6. D. Kinder, University of Rochester, Rochester, N.Y. 14627(personal communication).
