SURFACE AND INTERFACE ANALYSIS, VOL. 19, 139-144 (1992) Adsorption of Hydrogen Fluoride on Alumina R. G. Haverkamp,’ J. B. Metson,’ M. M. Hyland’ and B. J. Welch’ Department of Chemistry, University of Auckland, Private Bag, Auckland, New Zealand * Department of Chemical Engineering, University of Auckland, Private Bag, Auckland, New Zealand In the smelting of aluminium, HF fumes are produced that are subsequently trapped by absorption onto alumina. The factors that affect the adsorption capacity of alumina have been studied previously and are weU established, but the mechanism by which H F adsorbs onto the alumina surface is not well understood. In this study, x-ray photoelectron spectroscopy (XPS)was used to investigate the nature of the surface adsorp tion of HF on alumina. XPS is particularly weU suited to the study of this type of gas adsorption process. Laboratory-prepared samples were studied with particular interest in evidence for AI-F bonding, or in fluoride species formed by reaction with -OH or -0. Also of interest was the role of sodium, since it is segregated to the surface of the alumina during calcination. AI-F bonding was observed on only one sample type. An AI-F interaction was identified when the alumina had been predried and dry HF was adsorbed. When moisture was present no AIF, formation was observed. This suggests that under conventional conditions (i.e. moisture present) the adsorption of HF involves a weak interaction, probably hydrogen bonding, with intermediate layers of water. After heating the samples containing weakly bound HF to 500 “C, no Al-F interaction was observed. Much of the HF was desorbed at 700 O C . An Na-F interaction was observed in all fluoride-adsorbed samples; however, this can only account for a small proportion of the total fluoride adsorbed. INTRODUCTION Aluminium-smelting cells produce emissions of fluoride consisting of hydrogen fluoride and particulate sodium aluminium fluorides. The hydrogen fluoride is formed by the reaction of either water or hydrogen with cryolite. Particulate fluorides, volatile at cell temperatures, are emitted and then condense in the off gas stream. These emissions a can be captured by dry scrubbing systems in which the reactive gases and particulates are adsorbed onto alumina. The efficiency of the scrubber will depend to a large extent on the fluoride adsorption capacity of the alumina. Cochran et al.’ studied the adsorption of H F on alumina. They found that the chemisorbed fluoride is initially ‘amorphous’ but forms AIF, on heating. Cochran’ later postulated that the chemisorption of a monolayer of H F on Al’O, takes place by the formation of F(O), tetrahedrons with H + as a central cation. He suggested that at high temperatures this is converted into AlF, . He also commented on the role of sodium in the adsorption and concluded that there was insufficient total sodium present to account for more than a small proportion of the fluoride to be adsorbed by reaction with sodium. More recent studies on the mechanism of H F adsorption have generally concluded that water is involved in the adsorption process. Lamb3 found that multilayers of HF are adsorbed and that water vapour has a large influence on the amount of HF that can be held. Initial adsorption of HF on alumina takes place by reaction with surface sodium and with surface hydroxyl groups, with the hydroxyl groups being substituted by fluoride. Further adsorption then takes place by another mechanism to 0142-2421/92/240139-06 $08.00 0 1992 by John Wilcy & Sons,Ltd. form multilayers of HF. He suggested that this could be explained by hydrogen bonding between the fluoride molecules already adsorbed and the incoming HF molecules or, in the presence of water, with hydrogen bonded HF-H’O chains. In the ‘bimolecular’ model of Baverez and de Marco4*’it is proposed that multiple layers of HF molecules are bound to each other and to the alumina surface via adsorbed water molecules. This model is based on the observed dependence of H F adsorption on relative humidity. Coyne et have suggested that H F is directly bound to the surface hydroxyl groups since surface hydroxyl content (as determined by moisture on ignition) has a greater effect on HF capacity than humidity. Coyne et al.’ found that the H F adsorption capacity of alumina increased slightly with increasing residual sodium content. One of the commercial processes for manufacturing AlF, is by the reaction of H F with Al(OH), at high temperatures (500-600 “C).’ In the HF adsorption studies referred to above,’-’ no direct evidence of the formation of AIF, has been found at lower temperatures (20-120 “C)comparable to those in dry scrubber units. It appears that formation of AIF, by the reaction of H F with alumina has not been identified except at elevated temperatures. This paper reports new studies using x-ray photoelectron spectroscopy (XPS)to determine whether AlF, can be detected in samples prepared by the adsorption of H F on alumina under conditions of varying humidity. Evidence of an Na-F interaction was also sought. Al-F interactions were identified only when the alumina had been predried. When moisture was present, as in the earlier studies, no AIF, formation was observed. Na-F bonding was observed in all fluoride-adsorbed samples; however, this can only account for a small proportion ~ 1 . ~ 9 ’ R. G. HAVERKAMP ET AL. 140 of the total fluoride adsorbed. The work reported here deals only with adsorption of H F and H F / H 2 0 gas mixtures but not with aluminium reduction pot off gases, which also contain particulate sodium aluminium fluorides. In an earlier paper9 we reported XPS studies on fluorinated alumina where Na-F bonding was identified and HF was shown to desorb on heating rather than form AIF, . EXPERIMENTAL allowed variable gas concentrations of hydrogen fluoride in nitrogen to be produced. Warming of the H F cylinder was required to maintain sufficient pressure of HF gas. The exhaust gas was bubbled into a solution of TISAB with a fluoride electrode monitoring the presence of dissolved fluoride. Two samples were prepared from a batch of commercial-grade aluminium-smelting-feed alumina. For the first of these, DF1, the alumina was not pretreated. This alumina had a weight loss due to adsorbed water of 1.0% on heating to 300°C for 1 h. For the second sample, DF2, the alumina was predried for 12 h at 300°C before HF adsorption. HF adsorption Desorption of fluoride. The fluorinated aluminas were Wet HF adsorption. A system for adsorbing fluoride onto alumina using aqueous hydrogen fluoride was set up based on the procedure of Coyne et aL6 The set-up is illustrated in Fig. 1. A metered flow of preheated nitrogen was mixed with 3.7% hydrofluoric acid solution delivered by a calibrated syringe pump. This mixture then passed through a Teflon coil in a heated block before passing through the heated alumina bed. The heated block was maintained at 120"C.The adsorption behaviour was followed by monitoring the exhaust gas fluoride content. The exhaust gas was bubbled into a solution of TISAB (total ionic strength adjustment buffer) with a fluoride electrode monitoring the presence of dissolved fluoride. The alumina bed held up to 1 g of alumina, which was sufficient for XPS and XRD studies. Several samples were prepared by this method and the one referred to in this paper is labelled WF1. Dry HF adsorption. An apparatus for adsorbing hydro- gen fluoride gas, without added water, onto larger (80 g) samples of alumina was developed. As illustrated in Fig. 2, this consisted of a vertical bed of alumina fluidized by a nitrogen flow. Into this nitrogen flow, a small flow of hydrogen fluoride gas was bled. This arrangement 1 -- X-ray photoelectron spectroscopy X-ray photoelectron spectra were collected using a Kratos XSAM 800 spectrometer with an Mg Ka x-ray source. The presence in the analytical chamber was typically in the 10-9-10-'0 Torr range during analysis. Wide scans (0-1100 eV) of the alumina samples were collected under analyser conditions that gave good intensity but low-energy resolution. Narrow scans over the appropriate Na, F, 0, C and A1 peaks were subsequently collected under high-energy resolution conditions (20 eV pass energy). Binding energies (20.1 ev) are quoted using the 2p peak of A1 and the 1s peak for 0. For Na and F, the Auger parameter (a')is reported. This parameter is the difference between the kinetic energy (KE) of the major Auger line and the binding energy (BE) of the major photoelectron peak (the 1s level for both Na and F). It is more sensitive to changes in the Na and F chemical environments than the BE of (d) / ~ subjected to heating in an electric furnace in air for 500, 700 and lo00 "C for 1 h. The objective was to determine the effect of heating on the surface fluoride levels and on the fluoride bonding. \ 1 ressure 'gauge (0-SOokPa) flow meter (0.2-2 Ilmin) 1 alumina bed teflon tTbe needle valve 'teflon tee syringe pump (HF solution) TISAB solution Figure 1. 'Wet' hydrogen fluoride adsorption apparatus. ADSORPTION OF HYDROGEN FLUORIDE ON ALUMINA 141 (7 ressure 'gauge ' i \/' i , p i valves alumina fluid bed 3 i TlSAB solution Figure 2. 'Dry' hydrogen fluoride adsorption apparatus. the 1s peaks." Binding energies are referenced to adventitious hydrocarbon, which is present on all samples and set at 285.0 eV. The possibility of HF desorption under vacuum conditions was checked by monitoring the relative intensities of the A1 2p and F 1s peaks over a 90 min residence period in the analytical chamber. There was no detectable decrease in the F/Al intensities and therefore H F was not being desorbed during this phase of the analysis. It is possible that weakly bound H F may be desorbed during the initial pumpdown, but this could not be monitored. The XPS spectra of the laboratory-prepared samples of HF-adsorbed alumina, before and after heating, were collected. In addition, XPS spectra of commercialsmelter-grade alumina used as the starting material were collected. This was done using a pressed pellet of the sample before and after heating at 300°C under a vacuum for 2 h and for 18 h in the sample-holding chamber of the XPS instrument. Also, the spectra of AIF, , synthetic cryolite (Na3AlF,), NaF and y-Al,03 were run as BE standards. X-ray diffraction X-ray diffraction spectra were run on the fresh alumina and dry scrubber alumina samples and on the laboratory-fluorinated samples using a Philips x-ray diffractometer, scanning from 10" to 70" 28 using a Cu Ka x-ray source. RESULTS AND DISCUSSION The surface elemental concentrations obtained by XPS, for samples and standards, are listed in Table 1. Standard alumina The 0 1s spectra of the unreacted alumina shows a small asymmetry on the high-binding-energy side of the peak, which is due to some surface -OH groups on the alumina. This is still present in the spectrum of alumina Table 1. Surface elemental composition Mass concentration Atomic concentration F Na %Al %O %F % Na Al 0 Na,AIF, (syn) NaF 42.3 31.4 11.1 0 54.9 6.1 5.2 0 1.1 62.1 45.7 47.2 1.7 0.5 38.1 52.8 1 1 1 0 2.19 0.33 0.79 0 0.04 2.81 5.85 1.08 0.05 0.02 4.03 1 WF1 WF1 500°C 1 h WF1 700°C 1 h 52.7 38.5 38.3 36.9 46.3 52.4 10.1 13.2 5.8 0.3 2.0 3.6 1 1 1 1.18 2.03 2.31 0.27 0.49 0.22 0.1 0.06 0.11 DF1 DF1 500°C 1 h DF1 700°C 1 h 38.9 39.4 40.6 46.9 41.4 51.6 13.4 17.4 5.7 0.8 1.7 2.1 1 1 1 2.03 1.77 2.14 0.49 0.63 0.20 0.02 0.05 0.06 DF2 DF2 500°C 1 h DF2 700°C 1 h DF2 1000°C 1 h 29.3 28.8 34.5 44.6 25.6 12.5 40.2 53.1 44.8 55.9 21.4 1.4 0.3 2.7 3.9 0.9 1 1 1 1 1.47 0.73 1.97 2.01 2.17 2.76 0.88 0.04 0.01 0.11 0.13 0.02 Sample no. A1203 Al F, R. G . HAVERKAMP ET AL. 142 after heating at 300°C for 2 h and even after 18 h of heating. This confirms that this asymmetry is due to terminal hydroxyl groups rather than to adsorbed water, which would be removed at this temperature. A thermogravimetric study suggested that the hydroxyl groups are removed at a temperature of 500-600 "C. a) HF adsorption levels For the vaporized-HF-solution-adsorbed sample, WF1, the adsorption behaviour was similar to that reported by other^.^*^.' Adsorption continues quantitatively until saturation is achieved. Once this equilibrium H F adsorption capacity of the alumina under the particular humidity conditions is reached, no further H F is adsorbed. A surface H F level of 10.1 wt.% F was achieved under the above conditions. This corresponds to a surface F/Al atomic ratio of 0.3 : 1 atomic ratio. For the HF-adsorbed sample using the apparatus illustrated in Fig. 3, with a dry atmosphere but with the alumina not predried (DFl), a surface HF level of 13.4 wt.% F was obtained, corresponding to a surface F/Al atomic ratio of 0.5: 1. For adsorption on alumina that had been dried prior to adsorption (DF2), a very high level of surface fluoride was achieved. The 44.8 wt.% F corresponds to a surface F/Al atomic ratio of 2.2: 1. Fluorination of the sample was stopped before equilibrium was achieved. The reaction of HF with this dried alumina was markedly exothermic, with a temperature of 120 "Cbeing reached. b) -,-Ad /hh c) / \ / 4f " I , , , .,-i - 4 A1 2p binding energies For the samples and the standards, Table 2 lists the binding energies of the A1 2p and 0 1s peaks, as well as the Auger parameters for the Na and F lines taken from the narrow scans. For the fluoride-adsorbed aluminas WF1 and DF1, the A1 2p binding energies (summarized in Table 2) do not show significant shifts from the value of 74.3 eV in unreacted alumina. The binding energies of compounds AlF, and Na,AlF6 , containing A1-F bonds are higher at 77.0 and 76.0 eV, respectively. This suggests that in the fluoride-adsorbed samples WF1 and !3F2 there is Figure 3. Al 2p spectra of: (a) standard alumina; (b) wet HF adsorbed alumina (WF1); (c) HF adsorbed predried alumina (DF2); (d) AIF,. Table 2. Binding energies of aluminium, oxygen, sodium and fluorine Al 2p 0 1s AIF, Na,AIF, NaF 74.3 77.0 76.0 - 531.5 533.0 531.8 - WFl WF1 500°C 1 h WF1 700°C 1 h 74.5 74.5 74.3 DF1 DF1 500°C 1 h DF1 700°C 1 h DF2 Sample A1203 DF2 500°C 1 h DF2 700°C 1 h DF2 1000°C 1 h Binding energy F a' (eV) Peak width (eV) Fls Na 1s Na a' A l2p 01s 1340.5 1339.9 1339.7 1339.8 2062.4 2060.3 2061 .O 2.2 2.1 2.1 - 2.8 3.3 3.3 - 1.4 2.4 2.9 2.2 1.8 2.5 2.2 531.8 531.8 531.4 1340.2 1340.8 1340.8 2061.0 2062.1 2062.0 2.3 2.4 2.4 2.8 2.7 2.7 2.6 3.1 3.0 1.7 2.4 74.7 74.7 74.9 531.8 531.8 531.8 1340.5 1340.4 1340.8 2060.6 2061.3 2061.7 2.2 2.3 2.2 2.7 2.6 2.5 3.1 3.2 2.7 2.0 2.2 75.7 76.4 75.1 74.4 532.7 533.7 531.8 531.4 1340.2 1339.5 1340.3 1340.9 2060.2 2060.0 2060.7 2062.1 3.1 4.3 2.4 1.8 4.1 3.5 2.5 2.2 3.0 3.3 3.8 1.9 2.8 2.7 - ADSORPTION O F HYDROGEN FLUORIDE ON ALUMINA not a strong interaction between the H F and the surface A1 atoms. On the other hand, in sample DF2, which was dried prior to HF adsorption, there is a shift to higher BE of the A1 2p peak of 1.4 eV from the unreacted alumina value, This is clear evidence of A1-F bond formation. The A1 2p BE peak for the DF2 sample can be separated by computer curve-fitting into two unresolved doublets, with 2p3,, peaks at 74.9 eV and 76.3 eV (Fig. 4). The area under each of these peaks is 49% and 51%, respectively. The intensity of the 76.3 eV peak is evidence that about half of the surface A1 atoms are involved in a strong interaction with F. The x-ray diffraction patterns of the wet fluorinated alumina, WF1, and one dry fluorinated alumina, DF1, showed no peaks other than those due to alumina. This indicates that the absorbed fluoride in these samples does not form distinct crystalline compounds in SUEcient quantity to be detected. It therefore supports the evidence from XPS that no AlF, is formed. The x-ray diffraction pattern of the predried alumina that was fluorinated (DF2) did show the presence of major quantities of 8-AlF, 3H20, with lesser amounts of a-AlF, 3H,O and possibly AlF,(OH). The predried alumina may be able to react with the dry H F to form AlF, because the otherwise protective layer of surface-adsorbed water has been removed. When this layer of water is present, the fluoride is hydrogen bonded to the water layer and is not able to approach the aluminium or the oxygen atoms of the alumina closely enough to enable reaction and the formation of AlF, . - 0 Is binding energies The 0 1s peak position and the widths also provide information on changes taking place with fluorination of alumina. For DF2 there is a shift in the 0 1s peak, relative to unreacted alumina, of 1.2 to 532.7 eV (Table 2). The peak width is 4.1 eV, which is considerably larger than the 2.8 eV of fresh, undried alumina. + 80 78 76 74 Binding Energy 72 Figure 4. Al 2p of dry HF-adsorbed predried alumina (DF2) with two peaks fitted. 143 From the width and shape of the 0 1s peak it is clear that this envelope is composed of several peaks. The peak at 531.8 eV is assigned to oxygen in Al,O, as well as to adsorbed water, which also lies at -531.8 eV. Another broad peak, which makes up 50% of the area, is centred at 534.0 eV (Fig. 5). This second peak is consistent with the compound 8-AIF, . 3H,O identified by x-ray diffraction in DF2, in which the oxygen contained in the lattice water is moved to higher BE by the adjacent fluoride. For the HF-adsorbed samples where water was present, i.e. WF1 and DF1, the 0 1s peak does not shift from the value for pure alumina, indicating that there is no strong F-0 interaction (or F-OH). For compounds AlF, and Na,AlF,, which nominally do not contain any oxygen, surface oxygen attributed to adsorbed water was detected. The 0 1s binding energies for these compounds were, respectively, 533.0 and 531.8 eV (Table 2). The approximate position for adsorbed water is 531.8 eV. For AlF, the large shift in the 0 1s peak suggests that water is structural rather than just surface adsorbed as with Na,AlF,. The fluorine Auger parameters for the samples studied are listed in Table 2. There is little change in this parameter between the different standards, e.g. NaF and AlF, , and between the samples. Fluoride desorption The desorption of fluoride from the fluorinated aluminas can be seen from the data in Table 1. Heating at 500°C for 1 h does not result in a reduction in the surface fluoride levels. After heating at 700°C for 1 h a significant removal of fluoride occurs, while at lo00 "C essentially all the fluoride is removed. The desorption behaviour is similar for all the samples. For the WFl and DF1 samples there is no observable increase in the higher binding energy component of the A1 2p peak with heating to 500"C, suggesting that AlF, is not formed. With further heating this surfaceadsorbed HF is volatilized. For the DF2 sample, where 538 536 534 532 Binding Energy 530 528 Figure 5. 0 1s of dry HF-adsorbed predried alumina (DF2) with two peaks fitted. 144 R. G . HAVERKAMP ET AL. an Al-F interaction is already apparent, the average BE of the A1 2p peak shifts to a higher value with heating, indicating an increase in the proportion of aluminium atoms involved in the A1-F interaction. The 0 1s peak also shifts to higher BE as the dehydration of P-AlF, . 3H20 takes place to form AlF, . 2 H 2 0 and AlF, * H 2 0 , with a consequently stronger influence of fluoride on the oxygen orbitals. Sodium fluoride interaction Sodium is involved with the bonding of some of the adsorbed fluoride. Coyne et aL7 found that the HF adsorption capacity of alumina increased slightly with increasing residual sodium content. In an earlier paperg we reported XPS studies on fluorinated alumina where Na-F bonding was identified in fluorinated aluminas. The sodium Auger parameters for the samples studied here are listed in Table 2. There is a progressive shift in the sodium Auger parameter in the order: unreacted alumina > WF1 > DF1 > DF2. This shift in the Auger parameter is a result of the sodium present on the surface of the alumina reacting with the adsorbed HF to form Na-F bonds. Despite the strong surface segregation of sodium, even if all of the surface sodium is involved in HF bonding this can account for only a small proportion of the adsorbed fluoride. CONCLUSIONS We have investigated hydrogen fluoride adsorption processes occurring at the surface of alumina. The alumina surface contains terminal hydroxyl groups (Al-OH), surface adsorbed water and surface segregated sodium. X-ray photoelectron spectroscopy was used to look for an A1-F interaction in samples prepared by the adsorption of H F on alumina under a variety of humidity conditions. Evidence of an Na-F interaction was also sought. An Al-F interaction was observed in only one of the sample types, i.e. the case of the alumina that had dry HF adsorbed after being predried. When moisture was already present on the alumina, or water was present in the H F gas stream, as in the earlier no AlF, formation was observed. As the previous sample has shown that when an Al-F interaction occurs we can detect this in the XPS spectrum, it can be confidently said that there is no significant Al-F interaction in these samples adsorbed with moisture present. This suggests that the adsorption of H F involves a weak interaction, probably hydrogen bonding with layers of water, as proposed by other workers.,-' On heating, no evidence of the formation of AlF, was seen for the WF1 and DF1 samples up to 500"C, and by 700°C the HF was being desorbed. For the DF2 sample, both the A1 2p and the 0 1s BEs increased with heating to 500 "C, indicating more Al-F interaction and the dehydration of the P-AlF, 3H2O. The fluoride was desorbed >700 "C for all these samples. Na-F bonding was observed in all fluorideadsorbed samples; however, this can only account for a small proportion of the total fluoride adsorbed. The work reported here deals only with the HF and HF/H,O gas mixtures but not with aluminium reduction pot off gases, which also contain particulate sodium aluminium fluorides. REFERENCES 1. C. N. Cochran, W. C. Sleppy and W. B. Frank, J . Met. 22,54 (1970). 2. C. N. Cochran, Environ. Sci. Techno/.8, 63 (1974). 3. W. D. Lamb,./. Met. 31,32(1979). 4. M . Baverez and R. De Marco, Light. Met. 1979,891 (1979). 5. M . Baverez and R. de Marco, J. Met. 32,lO (1980). 6. J. F. Coyne, M . S. Wainwright, M . P. Bruigs and A. N. Bagshaw, Light Met. 1987,35(1987). 7. J. F. Coyne, P. J. Wong, M.S. Wainwright and M . P. Brungs, Light Met. 1989,1 1 3 (1989). 8. H. Sauer, J. N. Anderson and F. Kaempf, Ger. Offen. DE 3,405,452(1985). 9. M . M . Hyland. J. B. Metson. R. G. Haverkamp and B. J. Welch, Light M e t 1992,1323 (1 992). 10. C. D. Wagner, W. M . Riggs, L. E. Davis, J. F. Moulder and G. E. Muilenburg, Handbook of X-ray Photoelectron Spectroscopy. Perkin Elmer Corporation, Eden Prairie, M N (1979).
© Copyright 2026 Paperzz