Langmuir 1988,4,463-469 apply to our data at higher concentrations of trivalent and tetravalent ions, since the polyion is known to undergo a conformational collapses and thus can no longer be modeled as a linear polyelectrolyte. The most important point to be made from the data of Figure 5 is that NBD-Sp condenses onto the PGC as a trivalent ion but with somewhat higher affinity than the analogous spermidine ion. The additional binding energy is presumably a result of the interaction of the attached nitrobenzoxadiazole system with its binding site. The FPR data for the tracer diffusion coefficient of NBD-Sp ions are perhaps the most interesting and certainly the most general of this report. The unique feature of these data is that bound NBD-Sp do not contribute fluorescence, so we are able to measure the diffusion coefficients of the free NBD-Sp without the complicating signal from the bound species. The single-exponential form of the data indicates that the free NBD-Sp ions, which are in different environments and may have different mobilities at any instant, are rapidly sampling a range of potential on the time scale of the measurement (seconds), so that we determine a single time-averaged diffusion coefficient as a function of increasing concentration of PGC. Theories for the reduction of the mobility of counterions in solution as a result of the interaction of the counterion with a polyion have been advanced by Manning%and by Yoshida.ls Although neither theory can be applied exactly as formulated to our experimental conditions, the limiting value for the model of Yoshida is interesting for comparison to our results. Yoshida employs a model in which counterions in the vicinity of a polyion are restricted by the electrostatic potential in their motion
463
radial to the polyion but are free to diffuse along the isopotential lines of the polyion. Within the context of such a model, the lower bound for the tracer dfiusion coefficient of the counterion is the dimensional limit of D0/3. The data of Figure 7 may be viewed as the strongest evidence yet reported in support of that model. Our data may also be related to a controversy regarding the nature of binding of multivalent polyamines such as spermine to nucleic acids. In a statement that proved to be controversial, Manning26has argued that the binding may be understood as a delocalized binding rather than an ion-pair formation, stabilized by hydrogen bonds. Direct evidence for this distinction has not been forthcoming, but our data may be the most pertinent data yet to show that at least some of the spermine molecules are attached in a nonspecific way. To be sure, the fluorescent derivative associates with nucleic acid in an interaction that probably involves specific bonding and results in localization and fluorescence quenching. However, the high mobility of those spermines that remain fluorescent is a clear indication that a nonspecific mode of binding does exist.
Acknowledgment. We are grateful to Dr. Lindsay Plank for useful discussions in the analysis of these data and to Professor Grant Krafft for consultation with chemical issues and for constructing Figure 1. This work was supported by Grant No. GM-33786 from the National Institutes of Health and by Grant No. DMB-8607843from the National Science Foundation. Registry No. NBD-C1, 10199-89-0;NBD-Sp, 112763-58-3; PGC, 25512-84-9;spermine, 71-44-3; spermine4HC1, 306-67-2.
Adsorption of Ferricyanide Ion on Alumina K. W. Hipps,* Erik Dunkle, and Ursula Mazur Department of Chemistry, Washington State University, Pullman, Washington 99164-4630 Received June 16, 1987. I n Final Form: October 8, 1987 The adsorption of potassium hexacyanoferrat411) on three commercial alumina samples was studied by IR, neutron activation,and visible absorption spectroscopy. As previously observed on thin-fii alumina, we find that under neutral or basic conditions ferricyanide is reduced (in part) to ferrocyanide on these commercial alumina samples. The extent of the reduction, however, depends on several factors including postadsorption handling of the alumina. Above pH 7, the extent of adsorption is small and the uptake of iron cyanide from solution is dominated by pore filling. Below pH 7, the adsorption of ferricyanide increases with decreasing pH and reaches a plateau near pH 5.
Introduction During the last few years, we have been concerned with the application of inelastic electron tunneling spectroscopy to the study of species adsorbed on thin-film alumina. In several cases, we have observed significant redox chemistry'+ as signaled by the vibrational modes observed in the associated tunneling spectra. In the case of TCNE and TCNQ, the reductive adsorption indicated by tunneling ~
(1) Hipps, K. W.; Mazur, U.; Pearce, M. S. Chem. Phys. Lett. 1979, 68,433. (2) Hipps, K. W.; Mazur, U. J. Phys. Chem. 1980, 84, 3162. (3) Hipps, K. W.; Mazur, U. J. Phys. Chem. 1982, 86, 5105. (4) Mazur, U.; Hipps, K. W. J. Phys. Chem. 1984,88, 1555. (5) Hipps, K. W.; Mazur, U. Rev. Sci. Instrum. 1984,55, 1120. (6) Hipps, K. W.; Dunkle, E.; Mazur, U. Langmuir 1986, 2, 528.
0743-7463/88/2404-0463$01.50/0
s p e ~ t r o s c o p yis~ ~ not surprising in the context of the well-known radical formation observed when TCNE or TCNQ adsorb on many metal oxides.'-13 The rapid conversion of diphenyl-2-picrylhydrazylto diphenyl-2-pi(7) Korman, C. S.; Coleman, R. V. Phys. Reu. B. Condens. Matter 1977,15,1877. (8) Flockhart, B. D.; Leith, I. R.; Pink, R. C. Trans. Faraday SOC. 1970, 66, 469. Jpn. 1971, (9) Hosaka, H.; Fujiwara, T.; Meguro, K. Bull. Chem. SOC. 44, 2616. (10) Meguro, K.; Esumi, K. J. Colloid Interface Sci. 1977, 59, 93. (11) Flockhart, B. D.; Scott, T. A.; Pink, R. C. Trans. Faraday SOC. 1966, 62, 730. (12) Che, M.; Coluccia, S.; Zecchina, A. J. Chem. SOC.,Faraday Trans. 2 1978, 74, 1324. (13) Che, M.; Naccache, C.; Imelik, B. J. Catal. 1972, 24, 328.
0 1988 American Chemical Society
Hipps et al.
464 Langmuir, Vol. 4,No. 2, 1988
The third alumina sample was obtained by removing the oxide crylhydrazine on thin-film alumina was also observed by from Baker-flex thin-layer chromatography (TLC) sheets (catalog tunneling and was shown to occur on bulk alumina by no. 4-4467, aluminum oxide IB-FO). X-ray diffraction indicated conventional adsorption experiments and by IR specthat this material was also partially amorphous and that several troscopy.6 phases, including K and a,were present. Multipoint BET data In the present article, attention is directed to the elecprovided a value for the surface area of 126 m2/g, a mean pore tron-transfer reduction of an inorganic ion, hexacyanoradius value of 3.16 nm, and a total pore volume value of 0.21 ferrate(II1) (ferricyanide), to hexacyanoferrate(I1) (ferromL/g. These three oxides will be identified as y alumina, BB cyanide). The reduction of ferricyanide by thin-film alumina, and TLC alumina, respectively. A slurry composed of (native oxide) alumina was first observed by tunneling 1g of oxide and 10 mL of water stirred for 1 h provided values of 9.3, 7.6, and 9.5 for the pH of y, TLC, and BB alumina, respectroscopy.lV2 We found that the tunneling spectrum spectively. Thus, the y alumina and BB alumina used were obtained from ferricyanide adsorbed on alumina was esformally basic alumina, while the TLC alumina was neutral. sentially the same as for ferrocyanide adsorbed on alumina. Potassium ferricyanide, K3Fe(CN),, was purchased from RiFurther, the spectrum was that of ferrocyanide with less del-deHaen (no. 31253). This material had an analyzed maximum than 20% of ferricyanide present. Our initial efforts to concentration of 0.02 % ferrocyanide ion. From our experience, confirm that reduction occurs in the case of adsorption on this was the purest ferricyanide available. The potassium ferbulk alumina were partially successful.14 The infrared rocyanide, K4Fe(CN),.3H20,was Mallinckrodt analytical reagent spectrum of ferricyanide on basic alumina reported in that grade (no. 6932). paper showed significant amounts of both ferricyanide and Instruments. Infrared spectra were collected on an IBM instruments IR-98 Fourier transform spectrometer utilizing a ferrocyanide ions. However, the data were taken on a Barnes diffuse reflectance accessory. Diffuse reflectance specdispersive instrument in transmission mode, and only a tralS2' were taken from samples lightly pressed (hand packed) very narrow range of concentration was studied. Thus, we into a 2-mm-diametersample cup. A liquid nitrogen cooled MCT could say that ferricyanide on basic alumina was reduced detector was used, and data were collected at 4-cm-' resolution. to ferrocyanide, but we could not evaluate the extent of Kubelka-Munk transforms were performed with the software conversion. supplied with the instrument. In a few cases, where very small Previously, reports on surface-catalyzed teduction of amounts of sample were run diluted in KBr, the spectra are ferricyanide ion had centered on activated c h a r ~ o a l and ~ ~ J ~ presented in absorbance units. In all cases, KBr was used as the paper" surfaces. The reduction of ferricyanide to ferroreflectance reference and all spectra represent the sum of at least cyanide ion by hydroxyl with the formation of 02, as first 512 scam. TLC and y alumina samples were directly hand pressed reported by Barnes,'8 is catalyzed by activated ~ h a r c ~ a l . l ~ J ~while BB oxide and residue solids were hand mixed with KBr prior to hand pressing. Feigl reports that ferricyanide is partially reduced to Because the absolute intensities of diffuse reflectance spectra ferrocyanide on paper" but does not suggest a possible depend on particle size,lS2' and because the alumina types used mechanism. Thus, there is precedent for surface-induced here had widely different mean particle sizes, absolute intensities reduction. We also note that the general study of cyanide cannot be used. Relative intensities of bands close in energy are complexes as adsorbates on oxide surfaces is especially useful, however, and indicate the relative concentrationsof species attractive because the CN stretching frequency ( 1 ~ 2 4 O O on the same oxide sample. Diffuse reflectance calibration runs cm-l) falls within the transmission window of most oxides were performed on samples of various concentrations of K3Feand near the maximum sensitivity of MCT and InSb de(CN), and K4Fe(CN),, as single components and as mixtures, diluted in KBr. We find that while the peak intensities of the tectors. Hence, metal cyanides may be used as alternatives features near 2118 and 2042 cm-' are roughly equal for the two to carbonyls as models for the chemistry of transitioncomplexes, the integrated areas differ by a factor of 4.0 f 0.1. metal complexes on oxides surfaces. Linearity and reproducibility with concentration based on peak The adsorption of ferricyanide on three commercial areas were quite good in the range from 0.1% to 6% by weight. alumina samples will be considered. Adsorption of ferThe band positions reported are given as the position of maximum ricyanide ion from aqueous solution (near pH 8) on all peak height. three oxides will be considered in detail. We will show that Visible spectra of solutions were measured on a Perkin-Elmer reduction of ferricyanide occurs on all three alumina 320 spectrometer. Aqueous ferricyanide concentration measamples but that the extent of the reduction varies with surements were made at 420 nm. Neutron activation analysis of the oxide and post adsorption hand1ir.g. In the case of y oxide samples for iron and potassium was performed at the WSU radiation center. Iron and potassium contents for the ferricyanide alumina, we will also provide data on the pH dependence free (reference) BB alumina were 180 f 22 and 15 f 10 ppm, of the adsorption process.
Experimental Section Materials. The oxides used were purchased commercially and were heated under vacuum at 140 "C for >12 h immediately prior to use. This activation step had little effect on the extent of adsorption or reduction yield. I t was performed solely to avoid any possibility of the oxides "remembering" past handling. Alfa Products supplied the y alumina (catalog no. 89372) having a multipoint BET surface area of 91 m2/g, a mean pore radius of 5.22 nm, a total pore volume of 0.24 mL/g, and a purity of >99%. Reagent grade Baker basic aluminum oxide (1-0539) was found to have a multipoint BET surface area of 168 m2/g, a mean pore radius of 3.37 nm, and a total pore volume of 0.29 mL/g. This material was a t least partially amorphous and had X-ray diffraction peaks consistent with either (or both) .y and q phases. (14)Hipps, K. W. J.Electron Spectrosc. Relat. Phenom. 1983,30,175. (15) Puri, B. R.; Satija, B. R. J.Ind. Chem. Sot. 1968, 45, 298. (16) Nagarajatah, H. S.;Rao, P. S. Ind. J. Chem. 1971,9, 490. (17) Feigl, F. Spot Tests in Inorganic Analysis; Elsevier: New York, 1958. (18) Barnes, E.J.Ind. Chem. SOC.1939, 16, 308.
respectively. The corresponding values in the case of the TLC alumina reference were 115 f 25 and 16 f 5 ppm. The y alumina reference samples had 140 h 30 ppm of iron. Adsorption. Typically, 2 g of oxide was removed from the vacuum oven and 10 mL of the adsorption solution and a Teflon-coated stirring bar were added. The flask was then sealed by a septum and placed on a magnetic stirrer for well-defined periods ranging from 1.0 to 24 h. Adsorption periods of 1-24 h gave very similar results. The above slurry was then filtered through a medium glass frit and the oxide allowed to dry for 0.2-0.5 h. Most of the filtrations were performed in air, but a few of the y alumina/ferricyanide adsorption experiments were performed by using deaerated solvents and Nz flow during the fitration operation. It should be emphasized,however, that none of the experiments was performed in such a fashion as to rigorously exclude oxygen. Our primary interest in this study was to compare (19) Childers, J. W.;Palmer, R. A. Am. Lab. (Fairfield, Conn.) 1986, 128. ~ _
.
(20) Griffith, P. R.; Fuller, M. P. In Advanc. Infrared and Raman Spectr., Clark, D., Heater, H., Eds.;1982; Vol. 9, p 63. (21) Fuller, M. P.; Griffiths, P. R. Anal. Chem. 1978, 50, 1906.
Langmuir, Vol. 4, No. 2, 1988 465
Adsorption of Ferricyanide on Alumina
oxideTLC TLC BB BB BB Y Y Y
M, mol/L 0.045 0.100 0.009 0.047 0.096 0.063 0.125 0.034
Table I. Potassium Ferricyanide Adsorption Dataa ni, pmol/g u, A2/molecule f l N , pmol/g 17 (3) 25 (3) 5 (1) 14 (2) 19 (3) 7 (2) 11 (3) 80 (6)
1230 840 5950 1990 1470 2190 1370 190
mL/g
va,
pH (slurry)
35 (2) 48 (3)
0.40 (0.10) 0.23 (0.06)
7.4 7.9
32 55 38 77
0.38 0.38 0.49 0.53
9.0 8.7 9.0 9.1 4.1
(1) (2) (2) (4)
(0.06) (0.05) (0.07) (0.06)
M is the molarity of the ferricyanide solution in contact with the alumina sample; ni and flN are the amount of iron on 1 g of oxide as determined by visible spectroscopy and by neutron activation, respectively; V , is the volume of the equilibrium ferricyanide solution required to account for the difference in measured iron uptake. Values in parentheses are standard deviations determined from 3 to 6 measurements. bulk and thin-film oxide adsorption results. Since the latter studies were carried out in the presence of oxygen, we did not exclude it in these experiments. A portion of the above filtrate was diluted appropriately and ita absorbance at 420 nm measured in order to determine the amount of ferricyanide adsorbed. Another portion of the fiitrate was tested for the presence of ferrocyanide ion by use of the ferric chloride test,” wherein 20.5% (mol) ferrocyanide in 0.05 M ferricyanide gives a recognizable positive test. At this stage, the oxide was either placed in a vacuum oven for further drying or washed with water and then vacuum dried. Some of the resulting solid samples were then studied by IR spectroscopy and/or neutron activation analysis. In a few cases, the wash water from the first, second, or third wash was retained and evaporated to dryness, and the IR spectrum of the solid residue was measured. Because the uptake of iron from solution while the oxide is in contact with the solution (adsorption) can diffef from the total iron loading of the oxide found after the filtration process, two different type of iron analysis were employed. The concentration of ferricyanide ion in solution was determined by measuring the absorbance of the solution at 420 nm. The equilibrium surface coverage, assuming that the ferrocyanide concentration was small (vide infra), was given by the moles of ferricyanide lost from solution divided by the grams of oxide. This number we identify as ni. The final oxide loading after filtration measures both the total amount of iron adsorbed and the total iron trapped or entrained on the oxide. This value of moles per gram of alumina was measured by neutron activation and is identified as nN. In order to determine the pH dependence of the adsorption process, samples were prepared as follows: 2.000 g of y alumina, 0.1646 g of K,Fe(CN),, and a varying amount of HC1 or NaOH in a total solution volume of 10 q L were mixed in a septa sealed flask a t 25 OC for 10 h. After the adsorption period, the pH of the slurry was measured. The mixture was then filtered, the solution diluted 1:100, and the absorbance a t 420 and 500 nm recorded. The solid fractions were washed twice with 2 mL of water and placed in an LN2 trapped vacuum oven to dry at various temperatures for various times. IR spectra were recorded for some of these solids. In addition, the wash water was collected, the solvent was rapidly stripped by vacuum drying, and the resulting water soluble solids extracted from the oxide samples were hand mixed in KBr for IR studies.
Results The adsorption of ferricyanide from aqueous solution decreases significantly with increasing pH, as demonstrated in Figure 1. This is qualitatively consistent with the variation of surface charge of the alumina as a function of pH.22 At the lowest pH values the extent of adsorption reaches a plateau. For the equilibrium ferricyanide ion solution concentration shown in Figure 1 at pH 4 (0.034 M), and assuming single-layer surface adsorption, each ferricyanide ion occupies about 1.9 nm2. Taking the density of K3Fe(CN)6to be 1.84, one would estimate an effective area of about 0.5 nm2/molecule. Thus, a significant fraction of a monolayer can be deposited near pH (22) Brunelle, J. P. f i r e Appl. Chem. 1978, 50, 1211.
k 8
4
p~
of
12 slurry
Figure 1. Equilibrium amount of ferricyanide ion adsorbed per gram of y alumina as a function of slurry pH. Values on the y-axis have been multiplied by lo5.
K4
1
n 2100
:
IO
ENERGY (CM-’1
Figure 2. Diffuse reflectance spectra of KSFe(CN)Band K,Fe(CN)g3H20in KBr.
4. Our primary interest in this study, however, lies in the region near pH 8. As indicated in Figure 1 and further documented in Table I, the alumina samples studied act as weak adsorbents above pH 7. Of interest for later comparison are the spectra of t h e solid residues resulting from vacuum drying the stock 0.1 M solutions of K4Fe(CN),.3H20 and K,Fe(CN), used in this study. Figure 2 presents the diffuse reflectance spectra of these residues mixed in KBr. These spectra are essentially the same as for the respective “out of bottle” cyanide complexes. Figure 2, therefore, provides the spectra of authentic potassium femo- and ferricyanides and also assures that the solution handling and drying tech-
Hipps et al.
466 Langmuir, Vol. 4, No. 2, 1988
BE-alumina
!
t
> Y c
> u c
I
v1 W
ffl
z z u
z
c ii
c
z
Y
1900
2100 SNEHGY
1900
2300
~ ~ - 1 1
Figure 3. Diffuse relectance spectra of three different alumina samples following ferricyanide ion adsorption, filtration, and vacuum drying (A) y alumina, (B)BB alumina, and (C)TLC alumina.
niques (in the absence of alumina) do not chemically alter either complex. The structure seen in these bands is due to interaction of the CN stretches of each of the several complexes per unit cell with each other and the surrounding lattice. The observed structures are characteristic of the potassium salts and are quite different for other metal If an initially neutral K3Fe(CN), solution is stirred with oxide for 1.5 h, the 420-nm absorption of the filtrate decreases such that ni mol of iron/g of oxide appears to be adsorbed. If, following filtration, the solid fraction is vacuum dried at 35 "C for 18 h, the solid residue has an iron loading of nN mol/g of oxide as measured by neutron activation and (3 f 0.3)nNmol/g of potassium. This same solid material could be studied by diffuse reflectance in the CN stretching region of the IR. Typical results for the case of 0.1 M K,Fe(CN), adsorbed on TLC, BB, and y alumina are presented in Figure 3. Values of ni,n ~and , the equilibrium pH of the slurry are given in Table I for certain combinations of oxide and K,Fe(CN), solution concentration. The V , values reported in Table I were obtained by dividing(nN- ni) by molarity, M, and correspond to the amount of adsorption solution that must be retained by the oxide to account for the differences in iron loading reported by the two methods. That is, it is a measure of how much adsorption solution the oxide must retain on filtration in order to account for the excess iron loading seen on the dry oxide. The ferric chloride test for ferrocyanide was applied to the filtrates of the adsorption slurries resulting from the addition of neutral aqueous solutions of ferricyanide to the various alumina samples. For ferricyanide concentrations of 0.06 M and below, consistently negative test results where observed. A weakly positive test was occasionally obtained from 0.1 M ferricyanide on y alumina near pH 9. The effects of various postadsorption treatments on the IR spectrum obtained from ferricyanide supported on BB (23j)Xharitonov, Y.; Gol'danskii, V.; Seifer, G.; Borshagovskii, B.; Stukan, R.Izu. Akad. Nauk SSSR, Ser. Khim. 1970, 271. (24) Nekrasov, B.; Seifer, G.; Kharitonov, Y. Izu. Akad. Nauk SSSR, Ser. .. Khim. 1970. 219. (25) Ayen, J. RlWaggoner, W. H. J. Znorg. Nucl. Chem. 1971,33,721. (26) Herrington, E. F.; Kynaston, W. J. Chem. SOC.1955, 3555. (27) Haberditzl, W.; Schleinitz, K. D.; Bartel, H. G. 2.Naturforsch. 1968, 23, 891.
2100 ENERGY (cm-11
2300
Figure 4. Diffuse reflectance spectra of variously treated BB oxide samples following ferricyanide adsorption from aqueous solution: (A) damp oxide, (B) vacuum dried oxide, (C) washed and vacuum dried oxide. Spectrum C has been expanded by a factor of 4 for ease of comparison. K O Fe (CN),
I
on 7-A12O3
x
0.0
1900
0.10
2100
2 IO
ENERGY (CM-l)
Figure 5. Diffuse reflectance spectra of variously treated y alumina samples following ferricyanide adsorption from aqueous solution: (a)vacuum dried oxide, (b) partially washed and vacuum dried oxide, and (c) further washed and vacuum dried oxide. alumina (0.1 M K,Fe(CN), stirred with BB alumina at pH 8.7) can be seen in Figure 4. In figure 4A the oxide was placed in a 25 "C vacuum oven for 2 h immediately after filtration. In Figure 4B this same sample was then placed in a 45 "C vacuum oven for 12 h and the spectrum collected. If the oxide is vacuum dried for the same period at 25 "C,then the ratio of the 2048- to 2114-cm-l peaks is intermediate between p& A and B Figure 4 but closer to part A. Thus, mild heating also plays a significant role in the conversion process. The spectrum of Figure 4C is obtained when 1 g of the solid oxide resulting from the adsorption of 0.1 M K,Fe(CN), on BB alumina is washed with two 3-mL portions of water and then dried for 12 h at 45 "C. The spectrum in Figure 4C is expanded by a factor of about 4 relative to traces A and B. If y alumina is brought into contact with 0.1 M K3Fe(CN), solution for 4 h at pH 9, filtered, and vacuum dried for 12 h at 25 "C, the diffuse reflectance spectrum shown in Figure 5a results. Figure 5b is of a fraction of the solid taken from the above adsorption experiment but washed twice with 3.5 mL of water prior to vacuum drying. Figure
Adsorption of Ferricyanide on Alumina
Langmuir, Vol. 4, No. 2, 1988 467 I
O
I
I
I
I
2042
3
1s
ENERGY (CM-l)
ENERGY (CM-')
Figure 6. Diffuse reflectance spectra of solution residue following adsorption of 0.1 M K3Fe(CN),on y alumina (a) original filtrate, (b) second wash residue, and (c) final wash residue. 3.0
,
,
,
, . ,
,
,
1.5
0.0 11
2100
I
3
ENERGY (CM-')
Figure 7. Diffuse reflectance spectra of variously treated y
alumina samples following ferrocyanide adsorption from aqueous solution: (a) vacuum dried oxide, (b)partially washed and vacuum dried oxide, and (c) further washed and vacuum dried oxide. 5c is of a fraction of the above oxide further washed with an additional 3 mL of water. Note that parts b and c of Figure 5 are drawn on the same scale but shifted for ease of viewing. Figure 5a has been decreased in intensity by a factor of 10. Thus, the intensity of all bands decreases with washing, but the band near 2115 cm-l decreases most rapidly. About 1mL, each, of the liquid fractions from the above experiments (i.e.* 0.1 M filtrate, the second 3.5-mL wash, and the f i i 3-mL wash) was placed in agate mortars and vacuum dried to solids. These solid residues were hand ground with KBr, and diffuse reflectance spectra were measured as shown in Figure 6. Figure 6a is the spectrum of the initial filttate, 6b is that of the second wash, and 6c is the spectrum obtained from the last wash residue. The y-axis scales are arbitrary. IR spectral data on the adsorption of 0.1 M K,Fe(CN), on y alumina near pH 9 are provided in Figure 7. Parts a-c of Figure 7 are spectra that were obtained by successive washing of the oxide. In all three cases, 0.1 M K,Fe(CN), was stirred with y alumina for 4 h and then the alumina removed by filtration. Spectrum 7a resulted when the oxide was vacuum dried for 12 h at 25 "C and its neat diffuse reflectance spectrum measured. If 1 g of the fil-
Figure 8. Diffuse reflectance spectra: (a) K,Fe(CN), adsorbed on y alumina followed by water washing and vacuum drying, (b) K4Fe(CN)6adsorbed on y alumina followed by water washing and vacuum drying, and (c) dried 0.1 M K4Fe(CN)6.3H20 stock solution in KBr.
tered oxide is washed twice with 3 mL of water and then vacuum dried as above, spectrum 7b is obtained. If 1 g of the filtered oxide is washed 3 times with 3 mL of water, spectrum 7c results. Note that spectra 7b and 7c are plotted on the same absolute scale but that 7a has been attenuated by a factor of 3. Figure 8 contrasts the CN stretching region spectra obtained from different samples. Trace 8c is the spectrum of the dried 0.1 M K,Fe(CN), stock solution used in the adsorption experiments of Figure 7. Figure 8b is an expanded version of Figure 7c; the oxide resulting when 0.1 M K,Fe(CN), is adsorbed on y alumina and the solid is multiply washed with water. Trace 8a is an expanded version of Figure 5b: the the washed solid resulting from the adsorption of K,Fe(CN), on y alumina at pH 9. Discussion Consideration of Table I clearly shows that, above pH 7, the amount of iron on the alumina surface after drying is considerably greater than is present when the oxide is in contact with the solution. Much of this difference can be accounted for by assuming that the filtration process leaves the oxide pores filled with solution. Thus, the volume V , = (nN- nJ/M should be roughly equal or exceed the pore volume, V,. Considering the small amounts of material studied (2 g of oxide and 10 mL of solution), V, and V,, are in reasonable agreement. Thus, it is not surprising that the ratio of K/Fe seen in neutron activation is essentially that of the adsorption solution. This retention of adsorption solution also dramatically affects the observed IR spectra of the dried oxides. The intensity of the 2115-cm-' band is enhanced by a factor of 2-3 by the excess ferricyanide present. The common features of the alumina-supported ferricyanide spectra obtained from pH >7 adsorption solutions are bands near 2050, 2115, and 2160 cm-'. The sharp strong band near 2115 cm-' is primarily due to Fe(CN)63but may also have a contribution from pentacyanoaquo ferrate(III), which has a strong band at 2120 and a shoulder near 2152 cm-1.26*26 The 2160-cm-l band is not due to the same species that produces the 2050-cm-' band since it can be selectively removed with washing as shown in Figure 5. The band near 2050 cm-l is certainly in the range of iron(I1) cyanide complexes and outside the region expected for iron(II1) cyanides?*% The width and lack of structure
468 Langmuir, Vol. 4, No. 2, 1988 of these bands are consistent with the ions being located in an inhomogeneous environment. The CN stretch in free cyanide also falls in this region (near 2050 cm-1),28but we eliminate it from consideration because CN- adsorbs weakly on alumina near pH 8. The most likely species responsible for the 2050-cm-' band, therefore, are reduced iron cyanides, most notably ferrocyanide and pentacyanoaquo ferrate(II), which also absorbs radiation near 2050 cm-1.26The complex removed from the oxide surface after extensive washing (Figure 6c) is easily identified as the potassium salt of ferrocyanide by the details of the band structure, which are unique to this salt.23-28Thus the 2050-cm-' band seen on alumina following ferricyanide adsorption near pH 8 is due primarily to ferrocyanide. In what follows we will identify it as such, but contributions to this band by the monoaquo species cannot be ruled out at this time. The 2160-cm-' band appears to be associated with a component of intensity near 2115 cm-l, as may be seen in Figure 4C. It is therefore plausible that the 2160-cm-' band is due to pentacyanoaquo ferrate(II1). The spectra presented in Figure 3 clearly show that, following extensive vacuum drying at slightly elevated temperatures, ferricyanide ion is converted to ferrocyanide ion on all three alumina surfaces. The negative to weakly positive results for the ferrocyanide test by the FeC1, method (performed on the filtrate obtained from all the aluminas studied) indicate that either (a) the ferrocyanide ion formed is strongly held by the alumina or (b) only a small amount of ferrocyanide is formed while the alumina is in contact with the adsorption solution and it is the subsequent filtration/drying operation that produces significant amounts of ferrocyanide. Comparison of parts A and B of Figure 4 clearly shows that a significant part of the reduction on BB alumina is occurring during vacuum drying. Comparison of parts A and C of Figure 4 shows that heavy washing of the oxide surface prior to vacuum drying does not dramatically change the ferrocyanide coverage although the total amount of ferricyanide decreases markedly. Whether or not the BB or TLC oxides are washed prior to vacuum drying, the intensity near 2048 cm-l always increases with drying. The amount of increase is smaller for the washed samples and greater when the drying occurs above 35 "C. Similar behavior is seen in the case of TLC alumina, although the reduction is more effective in all phases of the sample treatment and the increase in ferrocyanide CN band intensity with heating is not so dramatic as in the BB alumina case. We interpret the above results to mean that reduction of ferricyanide ion to ferrocyanide ion on BB and TLC alumina occurs to some extent while the adsorption solution is in contact with the oxide but that further reduction takes place during the filtration/drying operations. It also appears that both ferro- and ferricyanide ions adhere to BB and TLC alumina with roughly equal strength in water. The ferro- to ferricyanide intensity ratios are not inverted for washed and unwashed samples. The behavior of K,Fe(CN), on y alumina is somewhat different than for adsorption on BB and TLC alumina. As can be seen from Figure 5, washing of the postadsorption y alumina produces a complete reversal in the intensity ratio of the bands near 2050 and 2115 cm-'. This suggests that (a) much of the adsorbed ferricyanide has been reduced and the large ferricyanide CN stretch seen in Figure 5a is due to material trapped in pores, (b) the oxide surface sites that most strongly bind ferricyanide are most sus(28) Nakamoto, K. Infrared and Raman Spectra of Inorganic and Coordination Compounds; Wiley: New York, 1978.
Hipps et al.
ceptible to reduction during the filtration/drying operations, or (c) y alumina is most effective at reducing ferricyanide during the filtrationldrying steps. It is interesting to note that further washing (Figure 5c relative to 5b) does not greatly affect the Fe(II)/Fe(III) ratio. Figure 7 portrays the spectra obtained from y alumina samples on which K,Fe(CN), has been adsorbed and, in the case of Figure 7b and 7c, washed to varying extents. Note that while much of the ferrocyanide is reversibly bound, a small amount is converted to ferricyanide ion. This is especially clear in Figure 8, wherein the 2115-cm-l band of K,Fe(CN), appears in the washed oxides resulting from both K,Fe(CN), and K,Fe(CN), adsorption on y alumina. The presence of ferricyanide ion on the ferrocyanide-treated alumina may be due to either a redox equilibrium at the surface or air oxidation during the vacuum filtration and suction drying steps. The latter possibility seems most likely since the solid oxide supported ferrocyanide nearest the center of the frit sometimes develops a yellow tint while air passes over it. We can use the positions of the CN stretching bands in the washed oxides to address the question of the local environment of the metal hexacyanide ion. Kharitonov and others2,-% have compiled CN stretching frequencies for a very wide range of ferricyanide and ferrocyanide salts. While u(CN) for AlFe(CN), falls close to that for K,Fe(CN),, this is not the case for the ferrocyanide salts. A shift of between 70 and 100 cm-' to higher energy is expected as all the potassium ions in K,Fe(CN), are replaced by aluminum ions. Thus, it is quite clear from all the spectra obtained that a significant fraction of the counterions surrounding the adsorbed ferrocyanide ion is K+ ions. On the other hand, some ion exchange must be occurring since (a) the pH of the solution changes as the adsorption of K,Fe(CN), proceeds and (b) the presence of both oxidation states of iron requires that some charge compensation occur. The results observed in the IR spectra of ferricyanide or ferrocyanide ion adsorbed on alumina near pH 9 and followed by washing and vacuum drying are consistent with the spectral results obtained by IETS2 That is, in both cases the production of ferrocyanide ion is observed and a significant portion of the ionic environment is due to potassium. The extent of ferrocyanide production, however, appears to be considerably greater in the tunneling data. This difference is due, in part, to the large amount of ferricyanide ion held in the pores of the dried oxides. It is also likely that the vacuum drying and P b deposition steps performed as part of the IETS sample fabriction also affect the ferro- to ferricyanide ratio. A 20-30 "C temperature rise during lead deposition is clearly sufficient to strongly affect the Fe(II)/Fe(III) ratio, as can be seen by comparison of parts A and B of Figure 4. In the case of organic adsorption from nonaqueous solvents, it is conventional to assign hydroxyl ion (OH-) as the reducing agent on hydrous oxides,7-13with the more loosely bound hydroxyls being the most active. As we pointed out in the Introduction, charcoal is known to catalyze the reduction of ferricyanide to ferrocyanide with the production of water and oxygen.15J6 It is possible, therefore, that alumina is acting as a catalyst for the reduction of ferricyanide by hydroxyl ion. Unfortunately, the situation is complicated by the large increase in reduction associated with vacuum heat drying. Ferricyanide salts can be reduced to ferrocyanide in t h e solid state by simple vacuum heating.29 Two different mechanisms have (29) Raj, D.; Danon, J. Radiochem. Radioanal. Lett. 1973, 13, 83.
Langmuir 1988,4, 469-474 been suggested. In anhydrous salts, such as K,Fe(CN),, the reduction occurs above 200 "C and produces K4Fe(CN),, KCN, and C2Nzgas.29 When water is present in the solid, as in the case of the lithium salt of ferricyanide, the reduction occurs in the 100-160 "C range and has the pentacyanoaquo ferrate(II1) ion as an intermediate. The increased reduction seen on alumina a t elevated temperature, therefore, may also be due to catalyzed reduction of the type seen in pure solids and not requiring hydroxyl ion.
Conclusions The adsorption of potassium ferricyanide on three commercial alumina samples having surface areas of the order of 100 m2/g and pore volumes of about 0.3 mL/g was studied. We find that a significant fraction of the oxidesupported material originates from solution held in the pores after filtration. Thus, spectral and analytical studies
469
performed on the dried oxides are quite misleading. In the case of all the alumina types studied, a significant amount of reduction occurs during the drying process, with very little release of ferrocyanide into solution during the adsorption process. Several possible mechanisms for the reduction of ferricyanide were presented. While reduction by hydroxyl ion probably does occur, other reduction processes similar to those occurring in crystalline salts of ferricyanide also may be significant. The thermally induced reduction of alumina-supported ferricyanide begins near 40 "C while the reduction of alkali metal salts of ferricyanide occurs above 100 OC.
Acknowledgment. We gratefully acknowledge the National Science Foundation and the Division of Materials Research for their support in the form of grants DMR8414566 and DMR-8320556.
Surface Charge Development at the Goethite/Aqueous Solution Interface: Effects of C 0 2 Adsorption Walter A. Zeltner and Marc A. Anderson**+ Water Chemistry Program, University of Wisconsin, Madison, Wisconsin 53706 Received July 16, 1987. I n Final Form: October 14,1987 The effect of C02adsorption on the zero point of charge (ZPC) of aqueous goethite (a-FeOOH) suspensions was studied for both unpurged and N2-purgedsuspensions. Unpurged goethite had a ZPC of 8.1 f 0.1, while goethite purged for 2 months displayed a change in ZPC to 9.0 f 0.3. Electrophoretic mobility measurements performed on both suspensions gave identical isoelectric points of 9.7 f 0.2. These results were explained as being due to C02adsorption in the groove on the goethite (100) face, a conclusion supported by cylindrical internal reflection-Fourier transform infrared spectroscopic studies. Also, the measured goethite ZPC would depend on the relative contribution made to the total goethite surface area by the (100)face. Because goethite grows in the [001] direction, the (100) face makes a smaller contribution to the overall surface area in goethite samples with a higher surface area if the other dimensions remain constant. Thus, one would expect to observe higher ZPC values for higher surface area samples. This conclusion is qualitatively consistent with results of other studies.
Introduction Experimental observations of physicochemical phenomena occurring at solid/aqueous interfaces are more readily interpreted if the solid surfaces are "pristine" in the sense that all adsorbed species are known to the investigator. For studies of aqueous/metal oxide interfaces, it has been suggested that a sensitive test for the absence of adsorbed species (other than the potential-determining H+ and OHions) is to obtain agreement between the zero point of charge (ZPC) measured by potentiometric titration and the isoelectric point (IEP) measured by electrophoretic m~bility.l-~The application of this agreement between ZPC and IEP as a purity test for metal oxide suspensions is gaining wider a~ceptance."'~ Because of its importance in soil fertility, in corrosion control, and as a magnetic tape precursor, goethite (aFeOOH) has often been used as a model colloid in electrical double layer and adsorption s t ~ d i e s . ~ ~ ' ~Goethite J " ~ ~ is readily synthesized in crystalline and has very low s ~ l u b i l i t y . In ~ ~many of these studies, the different goethite samples were characterized by their ZPC t Current address: Water Chemistry Program, University of Wisconsin-Madison, 660 N. Park St., Madison, WI 53706.
values-values which ranged from 7.020 to 8.6.19 When both ZPC and IEP values were given, reasonable agree(1)Parfitt, G.D. Progress in Surface and Membrane Science; Cadenhead, D. A., Danielli, J. F., Eds.; Academic: New York, 1976; Vol. ll, p 181. (2)Hunter, R.J. Zeta Potential in Colloid Science;Academic: New York, 1981;p 229. Parks, G. A. Surf. Colloid Sci. 1982,12, 119. (3)James, R. 0.; (4)Lyklema, J. Adsorption from Solution at the SolidlLiquid Interface; Parftt, G. D., Rochester, C. H., Eds.;Academic: New York, 1983; p 223. (5)Kavanagh, B. V.; Posner, A. M.; Quirk, J. P. Chem. SOC.Faraday Discuss. 1975,59,242. (6)Kavanagh, B. V.; Posner, A. M.; Quirk, J. P. J.Colloid Interface Sci. 1977,61, 545. Stiglich, P. J.; Healy, T. W. Adsorption from (7)James, R. 0.; Aqueous Sohtions; Tewari, P. H., Ed.; Plenum: New York, 1981;p 19. (8)Sprycha, R. J. Colloid Interface Sci. 1984,102, 173. (9)Barringer, E. A.; Bowen, H. K. Langmuir 1985,1, 420. (10)Hsi, C.-K. D.;Langmuir,D. Geochim. Cosmochim. Acta 1985,49, 1931. (11)Sprycha, R. J. Colloid Interface Sci. 1986,110, 278. (12)Kallay, N.; Babic, D.; Matijevic, E. Colloids Surf. 1986, 19,375. (13)Jang, H. M.;Fuerstenau, D. W. Colloids Surf. 1986,21, 235. (14)Penners, N. H. G.; Koopal, L. K.; Lyklema, J. Colloids Surf. 1986, 21, 457. (15)Atkinson, R.J.; Posner, A. M.; Quirk, J. P. J.Phys. Chen. 1967, 71, 550. (16)Hingston, F. J.; Posner, A. M.; Quirk, J. P. J.Soil Sci. 1972,23, 177.
0743-746318812404-0469$01.50/0 0 1988 American Chemical Society