J.C.S. Faraday I, 1980, 76, 9-18Titanium Dioxide-Electrolyte InterfacePart 2.-Surface Charge (Titration) StudiesBY DAVID E. YATES~ AND THOMAS W. HEALY*Colloid and Surface Chemistry Group, Department of Physical Chemistry,University of Melbourne, Parkville, Victoria 3052, AustraliaReceived 22nd March, 1977Surface charge-pH isotherms for TiOz in K+, Li+, Mg2+ and tetramethylammonium electrolytesolutions are reported for a TiOz colloid sample shown previously to be essentially holocrystallineand non-porous. The charge data are similar to other oxides with the exception of precipitatedsilica. The adsorption sequence obtained is MgZ+ % Li+ > K+ 21 (CH3)4N+.In the present paper we present the potentiometric titration surface charge pro-perties of the TiO,-electrolyte interface for various concentrations of 1 : 1 and 2 : 1electrolytes. The TiO, colloid sample was characterized in detail ; gas adsorptionand tritium exchange studies of the sample showing it to be non-porous are reportedin detail in the previous publication.Electrokinetic potential and coagulationstudies of the same sample have been reported in detail elsewhere.,The surface charge of the oxide-water interface can be determined by measuringthe uptake of the potential-determining ions, H+ and OH-, as a function of the pHof the suspension. The method is a modification of the potentiometric titrationtechnique used successfully by, for example, Lyklema et aL3~ to study the doublelayer at the Agl-solution interface.Potentiometric titrations were first applied to the study of oxide double layers byBolt with a-Fe,O, (haematite) sus-pensions.Their investigations confirmed the potential-determining role of Hf andindicated that the oxide double layers were significantly different from the doublelayers at the Hg and A@-solution interfaces.A different method of measuring the surface charge has been developed byAhmed;’ fresh dry samples of oxide are added to solutions freshly prepared atdifferent pH values. The surface charge is then calculated from the initial pH changesupon immersion. This technique has some advantage in that the oxide is, in principle,not given time to dissolve and so any effects due to soluble species are minimized.Detailed potentiometric titration studies on silica and earlier work by de Bruynand co-workers on a-Fe203 9* lo and later on Ti02 l1 and ZnO 12* l3 showed theuptake of the potential-determining ions occurs in two steps.The first adsorptionstep is rapid and complete within a few minutes, while the second is a slow processwhich may continue for weeks. The occurrence of the slow adsorption step meansthat cro against pH curves determined by potentiometric titration depend to a greateror lesser extent on the titration rate.The fast adsorption process has been attributed to the ionization of the surfacehydroxyl groups, as described by ionization equilibria, which by analogy to mostproton transfer reactions are expected to be practically instantaneous.t Present address : I.C.I. Australia Ltd, Central Research Laboratory, Newson Street, AscotVale, Victoria 3052, Australia.with SiO, sols and by Parks and de Bruyn10 TiO2-ELECTROLYTE INTERFACEIn general the rapid adsorption is the result of the surface charge development,while the slow adsorption reflects slow changes in the P.Z.C.Therefore, rapid acid-base titrations, which suppress the slow adsorption step, provide the most usefulinformation with respect to the oxide-solution interface. The reliability of the acid-base titration technique for determination of o0 values has been substantiated, at leastat low electrolyte concentrations, by direct determination of the concurrent counterion adsorption; results of Hingston et aZ.149 l5 show that the adsorbed counter-ioncharge is, within experimental error, equal and opposite to the surface charge forgoethite and gibbsite suspensions; similar results were obtained by Huang l6 foralumina and, in more detail, by Breeuwsma and Lyklema l7 and Li and de Bruyn l 8for haematite suspensions.calomel supporting suspensionelectrode electrolyte in supportingwith solution electrolytesalt bridgeEXPERIMENTALglasselectrodeMATERIALSThe pure rutile sample had been prepared by hydrolysis of re-distilled titanium tetra-chloride followed by drying and calcining at 150°C.It was cleaned by soxhlet extractionwith water and had a B.E.T. surface area of 19.8 m2 g-l. In the previous paper it wasfully described and shown to be non-porous.All distilled water was triply distilled, the second stage being from alkaline permanganate.All reagents used were A.R.quality unless otherwise stated. The KN03, NaN03, LiN03,Mg(N03)2 and N(CH3)&1 were further purified by recrystallizing from distilled water,after which the KN03 and NaN03, contained in aluminium dishes, were baked overnightat 100°C. The recrystallized LiN03 and N(CH3)&1 were made up as very concentratedsolutions and their concentrations determined by analysis of nitrate and chloride. Thetitrants, 0.10 mol dm-3 HN03 and 0.10 mol dm-3 KOH, were prepared from May andBaker or B.D.H. concentrated standard solutions. Nitrogen gas used to degas solutionsand to exclude C 0 2 was prepurified through concentrated NaOH, HN03 and two aqueousTi02 suspensionsD. E. YATES AND T.W. HEALY 11Following Bates 21 the pH of the suspension was determined relative to the pH of twostandard buffer solutions. The standard buffer solutions used were Merck (Titrisols)pH 4.00 citrate+HCl and pH 9.00 boric acid+ KClf NaOH where the pH is given by-log yk[H+] and y* is the mean ion activity coefficient and [H+] the proton concentration.The electrodes were calibrated in this way before and after each titration.Although at least slight irreversibility between rapid acid and base titrations appearsto be general for metal oxides, repeated titrations on the same suspension were found to bereproducible. Experiments also showed that adsorption isotherms determined on a seriesof identical suspensions were reproducible to better than k0.5 pC cmW2.Outside the pHrange 4-10 the reproducibility decreases due to the relatively large correction for the blank.The absolute accuracy of the titrations depends on the cancellation of the liquid junctionpotentials, which leads to an uncertainty in the pH measurements of w k0.05, and theaccuracy of the surface area determination discussed in the previous paper.lRESULTS AND DISCUSSIONSURFACE CHARGE AGAINST pH ISOTHERMSThe variation of surface charge density of Ti02 as a function of pH for four con-centrations of KN03 is shown in fig. 1. The point of intersection of the o0 againstciI 5\ Yb”2015105051 1 I 1 1 II 5 6 7 8 9 10PHFIG. 1.-Variation of the surface charge density of Ti02 as a function of pH in aqueous solutions ofKN03.A, 2.9 ; x , 0.1 ; 0, 0.01 ; A, 0.001 mol dm-3 KN03.pH curves for different electrolyte concentrations defines the P.Z.C. at pH = 5.8k0.1in KNO3. This value agrees well with the i.e.p. of 5.S5+0.1 for the same sample inKN03 determined by Wiese using microelectrophoresis and coagulation ratetechniques and is well within the range of values previously reported 22 for syntheti12 TiO2-ELECTROLYTE INTERFACEand natural Ti02.* The P.Z.C. also coincides with the inflection points of the lO-l,mol dm-3 curves. The general shape of the curves is typical of thereported ao-pH curves for synthetic oxides other than Si02. The ao-pH curveon 2.9 mol dm-3 KN03 appears to be approaching a maximum in a. at high pH.andPHFIG. 2.-Variation of the surface charge density of TiOz as a function of pH in aqueous solutions ofLiN03.A,A, 1.0; x, 10-1 ; 0, ; 0, mol dm-3.The results for LiN03 axe shown in fig. 2 where a number of different features areobserved. For LiN03 concentrations of 10-1 moldm-3 and lower, the P.Z.C. ispH 5.9k0.1, which is, within experimental error, the same as observed in KN03.However, in 1.0 mol dm-3 LiN03 the P.Z.C. shifts quite significantly towards lowervalues. Furthermore, the go against pH curves are much steeper and reach muchhigher values than were observed in KN03. The difference between the LiN03 andKN03 curves becomes larger with increasing concentration. This is similar to thebehaviour observed by Breeuwsma 1 7 9 2o for a-Fe203 in concentrated LiCl solutions.The titration of the 1.0 mol dm-3 LiN03 +Ti02 suspension also showed greaterhysteresis than the other titrations and when the suspension was equilibrated for* P.z.c.and i.e.p. definitions are as given by Lyklema and Healy (see Disc. Faraday Soc., 1971,52, 318)D. E. YATES AND T. W. HEALY 13several days at pH 4.8, the pH slowly drifted downscale, even though atmosphericCOz was rigorously excluded. Subsequent rapid acid-base titrations showed thatthe oo against pH curve was moved to lower pH values. This behaviour indicatesthat the slow adsorption step, discussed above, is not completely suppressed by therapid titration procedure and that slow adsorption of Li+ continues even at pHvalues below 5.0. All these trends indicate that Li+ is much more strongly adsorbedthan NO; and K+, i.e., Li+ is specifically adsorbed on TiOz.BCrubC and de Bruyn 23also observed high values of cro in the presence of 10-1 moldm-3 Li+, but did notreport any results for 1.0 mol dm-3 LiN03 or any shift in the P.Z.C.PHFIG. 3.-Variation of the surface charge density of TiOz as a function of pH in aqueous solutions of(CH&NCl. Scales are larger than those in fig. 4 ad 5. 77, 10-1 ; x , ; 0, rnol dm-3.and 10-1 mol dm-3 NaN03 again gave aP.Z.C. of pH 5.9 40.1 in good agreement with the P.Z.C. in KN03 and in 10-3-10-1 moldm-3 LiN03. The surface charge densities indicated that the magnitude of Na+adsorption was closer to the extent of Li+ adsorption than to K+ adsorption.The results for (CH3)4NCl are shown in fig. 3, where again the P.Z.C.ofpH = 6.0+0.1 agrees with the P.Z.C. determined in the other 1 : 1 electrolytes. Thecurves are qualitatively similar to the curves obtained in the other electrolytes butdiffer in that, for mol dm-3 (CH3)4NCl, go for the negative branch is higher thanin the presence of K+ or Lif and then does not increase as much with increasing con-centration. In the only other study of oxide surfaces in the presence of tetra-alkylammonium ions, Tadros and Lyklema 24 observed that oo for porous SiOz in(C2H.J4NC1 solutions was almost independent of concentration. This is a morepronounced version of the trend in the oo against (CH3)4NC1 concentration behaviourobserved here. In 0.1 mol dm-3 (CH3)4NCl, go = oo in 0.1 mol dm-3 KN03 and4 go in 0.1 niol dm-3 LiN03 or NaN03.In order to try and push the surface charge to higher values, a titration wascarried out in Mg(NO& solutions.The results are shown in fig. 4. The oo againstpH curves are much steeper and reach higher negative surface charges than wereobserved in equivalent concentrations of the 1 : 1 electrolytes. The P.Z.C. determinedA preliminary titration i14 TiO2-ELECTROLYTE INTERFACEfrom the intersection point of the and moldm-3 curves was againpH 5.8 k0.1. However, the P.Z.C. shifted with increasing Mg(N03)2 concentrationand the intersection point of the and 10-1 mol dm-3 curves was at pH 5.6A0.1.A further titration in 1.0 mol dm-3 Mg(N03), showed an inflection point and hencea probable P.Z.C. at pH 5. The maximum oo observed in 1.0 mol dm-3 Mg(N03)2was 73 pC cm-2 at pH 8.3 but the hydrolysis of aqueous Mg2+ made this titrationunreliable at pH values above 8.The shift in the P.Z.C. and the high surface chargesobserved both indicate that Mg2+ is specifically adsorbed and more strongly so thanLi+, which did not shift the P.Z.C. until the concentration was > 10-1 mol dm-3.Breeuwsma l7. l9 also found that Mg2+ shifted the P.Z.C. of a-Fe,03, although theshift was larger and occurred at lower concentrations [< mol dm-3 Mg(N03),]than the shift observed here on Ti02.In general the shape of the go against pH curves for Mg(N03)2 is similar to theshape in the 1 : 1 electrolytes except that the change in pH necessary to keep thesurface charge constant if the activity of the electrolyte is increased by tenfold, i.e.,the Esin-Markov ~oefficient,~~ is smaller in Mg(N03)2 than in the 1 : 1 electrolytes.This is in good agreement with the theoretical Esin-Makov coefficients calculatedby Lyk1e1n.a.~~I 1 I 1 1 I5 6 7 8 9PHFIG.4.-Variation of the surface charge density of TiOa as a function of pH in aqueous solutions ofMg(NO&. V, 10-1 ; A, ; 0, mol dnr3D. E. YATES AND T. W. HEALY 15Comparison of the cr, against pH curves described here with reported results forboth natural 26 and synthetic uncalcined rutile 2 3 9 27 reveal good qualitative agree-ment, although the magnitude of the cr, values found in the present study appears tobe slightly lower. A precise compaxison is not possible because different workershave used different electrolytes.The slightly lower cr, values may be a result of thecalcining of the present sample. Further comparison of the present cro against pHcurves with those of other oxides shows that the present sample also exhibits similardouble layer properties to most other metal oxides, in particular to a-Fe203,178 2oy 28* 29E - F ~ O O H , ~ ~ ZnO 12* l3 and Al2O3.l69 30 As with the other oxides, the cro valuesand differential capacities at high electrolyte concentrations are considerably greater(w 3-5 times) than the corresponding values for the classical mercury and AgIinterfaces. Of course, these high values were the reason for the proposal 24* 31 ofthe gel layer theory of oxide-water interfaces in which it is proposed that the surfacecharge and part of the counter-charge can be accommodated behind the surfaceproper.However, in this case our extensive studies of the nature of the Ti02surface revealed no evidence of either physical porosity or the presence of gel layersat the surface. It might be argued that a gel layer does exist while the oxide is insolution but collapses on out-gassing to form a non-porous crystalline surface so thatour surface characterization would not have revealed it. This is unlikely for tworeasons; firstly, the out-gassing was carried out under very mild conditions (i.e.,room temperature) and secondly, the gel layer must reform unexpectedly rapidlywhen wetted with aqueous solutions. This follows because Ahmed’s ’* 26 studiesshow clearly that the high surface charges are developed as soon as dry Ti02 and otheroxides axe wetted with electrolyte solutions. In view of the slow rates of crystal-lization and dissolution of most metal oxides, it is unlikely that a gel layer on rutilewould crystallize so readily or reform so rapidly.With regard to the maximum possible surface charge for Ti02, it is interesting tolook also at the work of Boehm 32 and Schindler and Gamsjager 3 3 on a pyrogenicsample (Degussa P25).These workers obtained maximum charges of w -65 and- 48 pC cm-2 in NaOH and NaC10, solutions at high pH compared to our maximummeasured value of -73 pC cm-2 in 1 mol dm-3 Mg(N03)2. However, none of thesevalues is expected to correspond to full ionization of the surface groups, as no definiteplateau has ever been observed on go against pH isotherms.In fact, in om previouspaper it was shown that the rutile surface carries z 6 OH groups per nm2 plus3 coordinated H20 molecules per nm2, which gives a maximum surface charge ofw - 190 pC cm-2 if all these groups were fully dissociated. Therefore the maximumobserved cr, values are all considerably less than the maximum possible charge, whichwas to be expected if the surface consists of exposed impenetrable crystal planes.The gel layer model of oxides would have allowed much greater cr, values.Therefore the present rutile sample exhibits typical oxide surface charges (i.e.,high), even though there is no independent evidence of any surface porosity. Henceit appears that while surface porosity to ions may, in some cases,** 24 contribute tothe high surface charge densities observed for oxides, it cannot be the general explana-tion.The present results have been used in two separate theoretical papers 3 4 y 35where site dissociation models of the e.d.1. of oxides have been used to analyse thedata.ION ADSORPTION SEQUENCESFinally, it is necessary to consider the important experimental approach of study-ing “ ion adsorption ” or ‘‘ ion specificity ” sequences.This approach was not use16 TiO,-E LE CTROL Y TE I N T ERF A CEextensively in the present study; only a brief comment is needed. The oo againstpH curves for 0.10 mol dm-2 K+, Li+ and Mg2+ nitrates are compared in fig. 5where it is clearly seen that the order of adsorption on the negative side iswhich reveals the importance of both the chemical nature and charge of the cation.It is also apparent from the agreement of the positive oo values in fig.5 that, withinexperimental error, there is no significant co-ion effect.P HFIG. 5-Variation of the surface charge density of TiOz as a function of pH in aqueous solutions of0.10 mol dm-3 K+(A), Li+(O) and Mg2+(V) nitrates.For oxides the order in the specificity of adsorption of ions depends on the particu-lar oxide. For example, the adsorption of alkali-metal ions, given by the presentand other studies on Ti02 23 and a-Fe203,17* 2 1 ~ 2 8 increases in the orderLi+ > Na+ > K+ > Cs+, but on Si02 surfaces 24p 36 the sequence is reversed. Theorder observed for both mercury and AgI interfaces is the same as observed onSO2, i.e., Cs+ > Li+.The observed adsorption sequence of tht; halide ions onTiOz 24 and ZnO 12* l3 is C1- > Br- > I-, which is the reverse order of that foundon mercury. The most thorough study of ion specificity in 1 : 1 electrolytes has beenmade on a-Fe203 sols by Dumont and Watillon 37 using a coagulation technique.Where comparison is possible, their stability sequences are in agreement with corD . E. YATES AND T . W. HEALY 17responding adsorption sequences,17* 9 9 * higher adsorption corresponding to lowerstability. Dumont and Watillon 37 considered that the ion-surface interactionswere governed by Gurney 3 8 type “ structure making ” and “ structure breaking ”properties of the ions and of the surface itself.In solution, ions of like orderingattract, those of opposite ordering repel each other. Therefore their a-Fe203 and thepresent TiQ, behave as a structure-promoting surface, as “ structure making ” ionssuch as Li+ are adsorbed more strongly than “ structure breakers ”. Surfaces suchas mercury, AgI or SO,, which preferentially adsorb “ structure breaking ” ions likeCs+, act as “ structure breaking ” surfaces. Hence ion-surface interactions areregarded as analogous to bulk electrolyte ion-ion interactions. There are, however,several general hypotheses for explaining ion sequences on various surfaces, for theproblem is a difficult one and more fundamental knowledge of the relative and absoluterole of factors such as electrostatic attraction, specificity of ion interaction, ionhydration and water structure is required.CONCLUSIONSIn summary, the go against pH curves presented here clearly show that thepresent sample of TiO, is characteristic of TiO, samples studied by other workersand exhibits similar double layer properties to most other metal oxides, in particularto a-Fe2O3,l79 19* 2 8 * 29 WF~OOH,,~ ZnO l2.l3 and Al2O3.l6’ 30 As with theseother oxides, the go values at high electrolyte cuncentrations are considerably greater(= 3-5 times) than the cr0 values for mercury and AgI interfaces at correspondingpotentials and conditions. Nevertheless, although no limiting value of go has beenobserved, all the experimental go values are considerably lower than the possiblesurface charge produced by full dissociation of all the surface groups ; in particular,the values observed for go and the differential capacity were much larger than thecorresponding values for the classical mercury and AgI interfaces.Thus highsurface charges occur for rutile-aqueous electrolyte interfaces even when there is noindependent evidence of either physical porosity or the presence of gel layers at thesurface.It was not possible to completely exclude the presence of a gel layer when therutile is in solution. However, if such a gel layer is present in solution then it mustcollapse to form a non-porous crystalline surface under conditions as mild as roomtemperature outgassing. Furthermore, if a gel layer is responsible for the highsurface charges, it must reform rapidly when wetted with aqueous solutions. This isbecause Ahmed’s ‘9 26 studies clearly show that the high surface charges are developedas soon as dry TiO, and other oxides are wetted with electrolyte solutions.In viewof the slow rates of crystallization and dissolution of most metal oxides, it is unlikelythat a gel layer on rutile would crystallize so readily or re-form so rapidly. A sitedissociation model 349 35 appears more useful at present.This work was supported by the Australian Research Grants Committee.D. E. Y . acknowledges the award of a Commonwealth Postgraduate Research Award.D. E. Yates, R. 0. James and T. W. Healy, J.G.S. Furuduy I, 1980, 76, 1.G. R. Wiese and T. W.Healy, J. Colloid Interface Sci., 1975, 51, 434.M. J. Sparnaay, The Electrical Double Layer (Pergamon Press, New York, 1972).J. Lyklema and J. Th. G. Overbeek, J. Colloid Sci., 1961, 16, 595.G. H. Bolt, J. Phys. Chem., 1957, 61, 1166.G. A. Parks and P. L. de Bruyn, J. Phys. Chem., 1962,66,96718 TiO,-E LE CTRO LY TE INTERFACE' S. M. Ahmed, Oxides and Oxide Films, ed. J. W. Diggle (Marcel Dekker, New York, 1972),vol. 1, chap. 4, p. 230.D. E. Yates and T. W. Healy, J. Colloid Interface Sci., 1976, 35,9.G. Y. Onoda and P. L. de Bruyn, Surface Sci., 1966,4,48.lo Y. G. BCrubC, G. Y. Onoda, Jr. and P. L. de Bruyn, Surface Sci., 1967, 8,448.l 1 Y. G. BCrubC and P. L. de Bruyn, J. Colloid Interface Sci., 1968,27, 305.l2 L. Blok, Ph.D. nesis (State University, Utrecht, 1968).l 3 L. Blok and P.L. de Bruyn, J. Colloid Interface Sci., 1970,32, 518, 527 and 544.l4 F. J. Hingston, R. J. Atkinson, A. M. Posner and J. P. Quirk, 9th Int. Congr. Soil Sci. Trans.,l5 F. J. Hingston, A. M. Posner and J. P. Quirk, J. Soil Sci., 1972, 23, 177.l6 C. P. Huang, Ph.D. Thesis (Harvard University, 1971).l7 A. Breeuwsma and J. Lyklema, Disc. Faraday SOC., 1971, 52, 324.l8 H. C. Li and P. L. de Bruyn, Surface Sci., 1966, 5, 203.l9 D. E. Yates, Ph.D. Thesis (University of Melbourne, Australia, 1974).2o A. Breeuwsma, Thesis (Agricultural University, Wageningen, The Netherlands, 1973).21 R. G. Bates, Determination of pH, Theory and Practice (J. Wiley, New York, 1964).22 G. A. Parks, Chem. Rev., 1965,65, 177.23 Y. G. BCrubC and P. L. de Bruyn, J. Colloid Interface Sci., 1968, 28, 92.24 Th. F. Tadros and J. Lyklema, J. Electroanalyt. Chem., 1968, 17, 267."J. Lyklema, J. Electroanalyt. Chem., 1972, 37, 53.26 S. M. Ahmed and D. Maksimov, J. Colloid Interface Sci., 1969, 29, 97.27 H. J. Wright, Ph.D. Thesis (University of Sydney, 1972).28 A. Breeuwsma and J. Lyklema, J. Colloid Interface Sci., 1973,43,437.29 R. J. Atkinson, A. M. Posner and J. P. Quirk, J. Phys. Chem., 1967,71, 550.30 C. P. Huang and W. Stumm, J. Colloid hterface Sci., 1973,43,409.31 J. Lyklema, J. Electroanalyt. Chem., 1968, 18, 341.32 H. P. Boehm, Adv. Catalysis, 1966, 16, 249 ; Angew. Chem. Int. Edn, 1966,5, 541 ; Disc.Faruday SOC., 1971,52,264. ; M . Herrmann and H. P. Boehm, 2. anorg. Chem., 1969,368,73.33 P. W. Schindler and H. Gamsjager, Disc. Faraday SOC., 1971,52,286 ; Kolloid-Z., 1972, 250,759.34 J. A. Davis, R. 0. James and J. 0. Leckie, J. Colloid Interface Sci., 1978, 63,480.35 T. W. Healy and L. R. White, Adv. Colloid Sci., 1978, 9, 303.36 R. P. Abendroth, J. Colloid Interface Sci., 1970, 34, 591.37 F. Dumont and A. Watillon, Disc. Faraday SOC., 1971, 52, 352.38 R. W. Gurney, Ionic Processes in Solution (Dover Publications, New York, 1953).1966, 1, 669.(PAPER 7/502