J. CHEM. SOC. FARADAY TRANS., 1994, 90(21), 3335-3339 Thermal Behaviour and Physico-chemical Characterization of Synthetic and Natural Iron Hydroxyphosphates Dominique Rouzies, Jean Varloud and Jean-Marc M. M. Millet lnstitut de Recherche sur la Catalyse, CNRS, Associe 4 l'universite Claude-Bernard, Lyon I, 2 Avenue A. Einstein, F-69626 Villeurbanne Cedex, France The thermal behaviour of five hydroxyphosphates, lipscornbite [Fe,(PO,),(OH),], barbosalite [Fe,(PO,),(OH),], giniite [Fe,(PO,),(OH), * 2H,0], rockbridgeite Fe,(PO,),(OH),] and Fe,(PO,),(OH), , has been studied. Ther- mogravimetric analysis revealed the hydroxyphosphates to have distinct behaviours : some underwent dehy- droxylation between 623 and 893 K like other phosphates [lipscombite, barbosalite, Fe,(PO,),(OH),], while rockbridgeite and giniite differed.Rockbridgeite, which appeared to contain only iron(iii) ions, underwent a first partial dehydroxylation at only 438K, while giniite totally dehydroxyated only at between 623and 823K. Compari-son of the results obtained by thermogravimetric analysis with those obtained by other techniques, including chemical analysis, X-ray diffraction and Mossbauer spectroscopy, allowed us to propose the existence of a new solid solution for the giniite compound and to confirm that previously observed for rockbridgeite. The solid solution of the former corresponds to a variation in composition due to dehydroxylation-hydration : Fe~'~,Fe~'~,(PO,),(OH),~,(2+x)H,O with 0 <x < 1. The latter corresponds to a variation in the composition due to dehydroxylation-oxidation : Fe~'~,Fe~'~,(P0,)3(0H),_,0,with 0 <x < 1.In both cases, Mossbauer spectros- copy indicates that Fe2+ or Mn2+ substituents may be localized on the different crystallographic sites of the structures. A recent study' has shown that several hydroxyphosphates are potential oxidative dehydrogenation catalysts. For cata- lytic reactions conducted between 573 and 723 K, it appeared of interest to study the thermal behaviour of these com-pounds in this temperature range. This paper reports the thermal and thermogravimetric analyses conducted in argon of four synthetic and one natural hydroxyphosphate: the two polymorphic forms of Fe,(PO,),(OH), , (i) lipscombite, (ii) barbosalite, (iii) Fe,(PO,),(OH), , (iv) Fe,(P0,)4(0H)2 -2H20 called giniite and (v) Fe,(PO,),(OH),, a natural sample called rockbridgeite.Although several published studies have dealt with these hydroxyphosphates, which are well known mineral compounds, none of them have reported in detail their thermal behaviour in a neutral atmosphere. Experimental Sample Preparation Fe,(PO,),(OH), ,lipscornbite and barbosalite samples were prepared as described previou~ly.~.~ They were obtained by hydrothermal synthesis using vivianite [Fe,(PO,), * 8H20] and an amorphous iron(@ phosphate gel as precursors. The hydrothermal syntheses were conducted at 473 K for 5 h. The samples obtained were filtered, washed with water and dried at 373 K overnight.The giniite sample has been prepared by hydrothermal synthesis using a mixture of 5 g of synthetic vivianite Fe,(PO,), -8H,O along with 5 g of phosphoric acid and 1 ml of distilled water per g of solid. The hydrothermal treatment was conducted at 383 K for 120 h. The solid obtained was washed with water and air-dried at 373 K. To our knowledge, this is a novel method for giniite synthesis. The rockbridgeite was kindly supplied by the Museum National &Histoire Naturelle (Paris, France). The chemical analysis of this mineral showed that it contained a relatively large amount of manganese (Table 1). Sample Characterization The chemical composition of the solids was determined by atomic absorption. X-Ray diffraction analysis was performed on the samples using a Siemens D500 diffractometer and Cu-Ka radiation.Differential thermal analyses (DTA) and thermogravimetric (TG) analyses were performed simulta- neously using a SETARAM TGA-DTA 92 thermobalance coupled to a Balzers 420 QMC mass spectrometer. 30-60 mg of the samples were placed in a platinum crucible suspended from one arm of the balance. The analyses were conducted at atmospheric pressure under a deoxygenated and dehydrated argon flow (1 1 h-') with a heating rate of 5 K min-' and a temperature limit of 1023 K. Mass spectrometry was used to verify the nature of the eventual gaseous reaction products, namely 02,H, , H,O, CO and CO,; the precision of the thermogravimetric analyses was ca. 2%. Mossbauer spectra were recorded at room temperature, using a 2 GBqS7Co-Rh source and a conventional constant acceleration spectrometer, operated in triangular mode.The samples were diluted in A120, in order to avoid a too high Mossbauer absorption, and pressed into pellets. The isomer shifts (6) were given with respect to a-Fe and were calculated, as the quadrupole splittings (A) and the linewidth (W), with a precision of ca. 0.02mm s-'. Results Physico-chemical Characterization The main characteristics of the solids studied are presented in Table 2. The P :Fe ratios were calculated from chemical analyses and the values of Fe3+ :(Fe3++ Fe2+) were deter- mined from Mossbauer spectroscopy. The values obtained Table 1 Chemical analysis of the natural rockbridgeite sample ~~ element ~~ ~ wt.% Fe 30.2 P 13.0 Mn 5.5 A1 0.2 Ca 0.1 Mg 0.06 J.CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 Table 2 Main characteristics of the hydroxyphosphates studied ;theoretical values are given in parentheses compound P :Fe Fe3+:(Fe3++ Fe2+Y Fey'Fe"(PO,&(OH), barbosalite 0.69 (0.66) 0.66 (0.66) Fe~1Fe11(P0,)2(OH)2 lipscombite 0.69 (0.66) 0.65 (0.66) Fe!?(PO,),(OH), 0.76 (0.75) 1.00 (1.00) Fey'Fe"(PO,),(OH), .2H20 giniite 0.80 (0.80) 0.65 (0.80) Fe~'Fe"(PO,),(OH), rockbridgeite 0.64 (0.60)b 0.87 (0.80)b a Determined by Mossbauer spectroscopy. In the case of rockbridgeite P :(Fe + Mn) and Fe3+ :(Fe3++ Mn2++ Fe2+) are given. for the P :Fe ratios were in good agreement with the theo- retical ratios.The calculated values for the Fe3+: (Fe3++ Fe2+) ratios corresponded to the theoretical ones except for the giniite and rockbridgeite samples which both indicated an excess of iron(@ cations. The X-ray dif- fraction spectra all revealed pure phases.' Tbermogravimetric Analyses Fe~'Fen(PO,),(OH), ,Barbosalite and Lipscombite The weight loss of the samples occurred at 715, 802 and 823 K for lipscombite and at 743, 793 and 868 K for barbosalite [Fig. l(a) and (b)] (Table 3). The mass spectrometric analyses performed simultaneously allowed the mass losses to be attributed to the dehydration of the solids. The mass losses of the lipscombite and barbosalite are 0.97 and 1.05 mol, respec- tively, per mol of solid.These values are very close to the theroetical values, corresponding to a total content of water, of 1.00. In both cases the dehydration process began at ca. 623 K, the first weight losses were the most important and corresponded approximately to half of the water contained in the compounds. The other half corresponded to the two final smaller water departures and seemed to be equal. Results from DTA analyses have not been presented in all cases. They confirmed the weight losses characterized by poorly defined endothermic peaks. After heating to 923 K, the X-ray diffraction patterns of the samples were totally modified, showing a complete transformation of the solids after dehy- dration. Fe',"(PO,),(OH), The thermogravimetric analysis of this compound showed three weight losses at 802, 818 and 849 K [Fig.l(c)] (Table 3). These weight losses, which were attributed to water depar- tures, corresponded to 1.54 mol of water per mol of solid. This value was again very close to the theoretical value of 1.50. The first two water departures, which were very close in temperature, corresponded in total to ca. 1mol of water. Giniite, Fe',"FeU(PO,),(OH), 2H20 Thermogravimetric analysis of this compound showed three water departures at 473, 658 and 773 K [Fig. 2(a)].The first weight loss, which was the most important, corresponded to + Q,C 2-0.0 --0.1-:-I-2.0 400 500 600 700 800 900 T/K Fig. 1 DTG and TG curves of the two polymorphic forms of Fe','1Fe1'(P0,)2(OH)2,namely (a) barbosalite and (b)lipscombite and (4 Fe,(PO,),(OH) 3 Table 3 compound Fe~1Fe11(P0,)2(OH)2 Fe~'Fe"(PO,),(OH),.2H20 Fe~'Fe"(PO,),(OH), temperature/K barbosalite 743 793 868 lipscombite 715 802 843~~ 802 818 849 giniite 573 633 763 rockbridgeite 434 658 823 Details of the DTA and TG analyses of the hydroxyphosphates loss of water /mol of H20 mol-' ~~ 0.54 0.24 0.27 0.48 0.24 0.24 1.05 0.49 2.60 0.60 0.20 2.20 total loss of water" /mol of H20 mol-' 1.05 (1.00) 0.97 (1.00) 1.54 (1.50) 3.20 (3.00) 2.40 (2.50) a Theoretical values are given in italic.J. CHEM.SOC. FARADAY TRANS., 1994, VOL. 90 2.6 mol of water per mol of giniite and should correspond to the hydration water (Table 3).The two other water depar- tures, which took place between 623 and 827 K, should corre- spond to the departure of the hydroxy groups. These weight losses correspond to 0.6 mol of water per mol of giniite. This value was lower than that expected from the stoichiometry, whilst that corresponding to the first loss was higher. Rockbridgeite Fe~'Fe''(PO,),(OH), The thermogravimetric analysis of the rockbridgeite sample showed three water departures [Fig. 2(b)]. The first occurred at 434 K and the other two at much higher temperatures, 658 and 823 K (Table 3). Of these last two, the first was sharper and more significant than the second. The first departure cor- responded to 0.20 mol of water per mol of rockbridgeite and the others to 2.20 mol of water per mol of rockbridgeite, which gave a total of 2.40 per mol of rockbridgeite.Miissbauer Spectroscopic Study The Mossbauer spectroscopic analyses of the barbosalite, lip- scombite and Fe,(PO,),(OH), samples, have been p~blished.~.~The results obtained were in good agreement with both the results of the thermogravimetric and chemical analyses. The Mossbauer spectrum of the giniite sample, which is presented in Fig. 3, shows four doublets (Table 4). These doublets, two of which correspond to iron(II1) cations and the other two to iron(@ cations, could not be directly attributed since the detailed crystallographic structure of the compound is not known. However, it was observed that the Fe3+:(Fe2++ Fe3+) ratio was equal to 0.65 instead of 0.80 as deduced from the stoichiometry. Characterization of the sample of rockbridgeite by Moss- bauer spectroscopy showed that the sample contained only iron(II1) cations (Table 4).The spectrum presented in Fig. 4, shows three doublets. These doublets could correspond to the three crystallographic sites known' to exist in the structure of rock bridgei t e { [Fe;'] [Fey] [Fe"] (PO,),( 0H) 1. The unit framework of the structure of rockbridgeite consists of a cluster of three face sharing octahedra Fe(2)-Fe( 1)-Fe(2). These clusters and other FeO, octahedra [Fe(3)] are con- nected to each other by corner sharing (Fig. 5). The crystallo- graphic sites [Fe(l), Fe(2), Fe(3)] have a relative ratio equal to 20 :40 :40.From bibliographic data and by considering the relative areas of the doublets, the doublet with the quad- 0.0 --0.0 --1 .o -4.2 7 -2.0 -.-c -4.4-3.0-E\ --. --v) -0.0--0.00 3 z-1 .o --4.05 3 tu -c -2.0 -4.10 300 500 700 900 TIK Fig. 2 DTG and TG curves of the (a)giniite and (b) rockbridgeite sample -2 -1 0 1 2 velocity/mm s-' Fig. 3 Experimental Mossbauer spectrum of the giniite sample, recorded at 295 K. Solid lines are derived from least-squares fits. rupolar splitting equal to 0.66 mm s-could be attributed to Fe3+ in the cluster [Fe(2) sites] and that with the smaller quadrupolar splitting (0.38 mm s-') to Fe3+ in the Fe(3) sites. The last doublet with the larger quadrupolar splitting (0.99 mm s-') should correspond to the Fe(1) sites.The occupation of these last sites by Fe3+ instead of Fe2+ cations 3 2 v)4-C 2 0 82 veloci ty/mm s-' Fig. 4 Experimental Mossbauer spectrum of the rockbridgeite sample, recorded at 295 K. Solid lines are derived from least-squares fits. Table 4 Mossbauer parameters computed from the spectrum of the synthetic giniite and natural rockbridgeite sample, recorded at 295 K Mossbauer parameters/mm s-' relative compound site 6 W A intensities (YO) giniite Fe3'(l) 0.42 0.26 1.05 42 Fe3+(2) Fe2'(2) 0.44 1.24 0.26 0.24 0.41 2.37 24 11 rockbridgeite Fe2'(3) Fe3'(l) 1.15 0.42 0.25 0.34 2.61 0.99 23 45 Fe3'(2) 0.43 0.29 0.66 33 Fe3'(3) 0.44 0.24 0.38 22 a4 Fig.5 Fe-0 polyhedra connection in the rockbridgeite structure' should give rise to some local distortions responsible for the large quadrupolar splitting observed. This seemed acceptable, but note that a discrepancy between the relative intensities of the observed doublets (22 : 45 : 33) and the theoretical ratio exists. Discussion Lipscombite, barbosalite and Fe,(PO,),(OH), compounds which adopt relatively similar structures, were shown to undergo dehydroxylation in the same temperature range. Two main weight-loss stages could be distinguished. The first corresponded to a sharp DTG peak and the second to a broad and complex DTG peak due to at least two distinct losses of water molecules. If the temperature range of dehy- droxylation of the three basic phosphates cited above (Table 3) is compared with the temperature range of departure of constitutional water of acid phosphates (Table 5), it can be seen that these temperature ranges are approximately the same.Nathan et aL6 showed that natural lipscombite lost its constitutional water between 603 and 843 K and Gleith' reported the stabilization of the lipscombite structure up to 823 K. These results which seem incompatible may be explained by our results. We confirmed the results obtained by Nathan et d6on natural lipscombite as we observed that the weight loss occurs before 843 K. We observed two very distinct weight losses for the synthetic lipscombite, the first one at ca.715 K and the second between 802 and 843 K. An X-ray diffraction study performed on a sample heated to 773 K (i.e.a temperature intermediate between the two water departures) and quenched at this temperature, allowed us to show that the first water departure, which corresponds to the Table 5 Temperatures of water departure for several acidic and condensed phopshates temperature of compound water departure/K FePO, . 2H20 423-413 Fe(H2P04)3 423-473 FeH2P207 823-923 FeH2P,010. 2H20 313-423 613-823" 613-113FeH2P3010 FeH2P3010.5H20 313-413 613-113" Fe3(P0,0H), .4H20 473-503 673-723" Fe3(P04)2(0H)o.8 7 ' 7.13H20 313-413 ' First temperature ranges correspond to the dehydration process. J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 departure of half of the water, did not affect greatly the struc- ture of the lipscombite.8 This stabilization of the structure independently reported by Gleith7 may presumably be explained by the formation of OH vacancies up to 803 K: Fe,(PO,),(OH), -+ Fe3(P0,),0,(OH)2 -,,+ xH,O (where x 6 0.5).There are differences between our results and those of Nathan et al.6 for natural samples. First, the natural samples appear to lose water at a lower temperature since the dehy- dration begins at 603 K, and four poorly defined peaks at 628, 683, 783 and 833 K are also observed. Secondly, we did not observe the oxidation of the iron(I1) ion in lipscombite in an inert atmosphere. This oxidation observed at ca. 473 K, as in air, is proposed to take place without the addition of oxygen, but by removing the H ions from the OH groups, thus forming H, .The Mossbauer data obtained for the giniite sample showed that the solid is in a more reduced form than would be expected from the stoi~hiometry.~ However, the chemical analysis of the sample showed the expected P : Fe ratio and DTG analysis showed two water departures which should correspond to the hydration and hydroxylation water. The first water departure corresponded to 2.6 mol of water per mol of giniite, whereas the second corresponds to 0.6 mol per mol of giniite. This last result, together with the result from Mossbauer spectroscopy, allowed us to propose that the syn- thesized sample giniite corresponds to a reduced sample with the formula Fe\~3Fe:'.7(P0,),(OH)l.3-2.6H20.Therefore, it is postulated that the giniite can correspond to a solid solu-tion in which reduction of the iron is possible and is accom- panied by the substitution of hydroxy groups by water molecules: Fe~~,Fe~-,(PO,),(OH),-, -2 + xH,O. This oxido-reduction process is the fifth process known to occur to iron hydroxyphosphates: type 1 corresponds to a variation in composition with formation of iron vacancies; this is the case for lipscombite, as has been postulated by Gleith:7 Fe',"Fe"(PO,),(OH), -+ Fe:", ,,Fey- ,,(PO,),(OH), ; where 0 < x < 1 Type 2 corresponds to a variation in composition due to dehydroxylation and oxidation. This is the case for rock- bridgeite:'' Fe',"Fe"(PO,),(OH), -+ Fey:,Feft-x(P04)2(OH)2 -,ox; where 0 < x < 1 Type 3 is a combination of types 1 and 2; it has been shown to take place for Fe~'(PO,),(OH), :4 FeT(P0,)3(0H)3 + Fef'- 3,Fe:,(P04)3(OH)3 -3,03,; where 0 < x < 1 /?Fe'"Fe'~PO,)O; when x = 1 Type 4 corresponds to an auto-oxidation by decomposition of crystal water.This process has been shown, by Hanzel et d.," to proceed for vivianite upon heating under vacuum: Fe~(P0,),~8H20-+Fe~~,Fe~(P0,),(0H),~(8-x)H,O + H, Type 5 corresponds to a variation in composition due to dehydration and oxidation. It has also been shown to occur for vivianite upon heating in the presence of oxygen:12*13 Fe','(PO,), . 8H,O + 1/20, -+ Fe~-,Fe~'(PO,),(OH), * (8 -x)H,O + H,O The existence of the giniite solid solution is related to this last process.A solid solution has previously been reported J. CHEM.SOC. FARADAY TRANS., 1994, VOL. 90 which contained only pure iron(@ giniite.14 The latter was believed to vary in content of iron cations and hydroxy groups as well as hydration water molecules from Fe?~2(P04)4(OH)i.s6 '2.75H2O to Fe?!00(P04)4(OH),.,,O,. 4.60H20. It is, however, improbable that a solid solution with such a variation in composition could really exist. Note that the temperature corresponding to the departure of the hydration water molecules is exceptionally high for giniite compared with other phosphates. As shown in Table 5, the temperature of dehydration of either condensed, acidic or basic phosphates is generally between 373 and 473 K.In contrast, the departure of the hydroxy groups begins at rather low temperatures compared with other hydroxy-phosphates. This phenomenon may be linked to the fact that after the departure of the hydration water molecules the structure of giniite is weakened by the empty channels and collapses immediately, resulting in its dehydroxylation. Our results indicate that the structure of the giniite presents three crystallographic sites in a 40 :40 : 20 ratio which resembles rockbridgeite. The two most numerous sites [Fe( 1)and Fe(2)] would be occupied by Fe"' and the less common one Fe(3), by Fe". With this hypothesis, the results obtained by Moss- bauer spectroscopy could be interpreted as a partial oxida- tion of one of the two more numerous sites [Fe(2)].The doub- let characterized by 6 = 0.42 mm s-' and A = 1.05 mm s-' would correspond to Fe" occupying site Fe(l), that with 6 = 1.15 mm s-' and A = 2.61 mm s-' to Fe" occupying site Fe(3) and those with 6 = 0.44 mm s-' and A = 0.41 mm s-', and 6 = 1.24 mm s-l and A = 2.37 mm s-' would corre- spond, respectively, to Fe"' and Fe" occupying site Fe(2). The formula of giniite obtained from DTG and chemical analyses could then be rewritten as: [Fe~'][Fe',l,Fe&] [Fe"](PO,),(OH),,, * 2.6H20. Such a formula leads to a ratio between the four doublets equal to 40: 26: 14: 20, which is not far from what is observed 42 :24 : 11 :23 (Table 4). The Mossbauer analysis of the rockbridgeite sample showed that it contained only iron(rI1) cations.This is pos- sible since it has been shown that the rockbridgeite corre- sponds to a solid solution in which the oxidation level of the solid could change with the variation in the hydroxy group content: Fe~~,Fe~-,(P04),(OH),-x0,with 0 < x < 1." However, the observed water loss, equal to 2.4 mol per mol of rockbridgeite, is much higher than that of 2.0, weighted for a totally oxidized sample. At this point the presence of a large amount of Mn2+, shown by chemical analysis (Table l), has to be considered. Mn2+ is known to substitute for Fe2+ in the natural hydroxyphosphates. The stoichiometry of our compound has been determined from chemical analysis results (Table 1) as Fe~!2Mnb'.,(PO,),(OH)4.800.2. In this 3339 case the sample would then have a water content correspond- ing to 2.4 mol of water per mol of solid which is in better agreement with our TGA experimental results.Normally the Mn2+ ions should substitute the Fe2+ ions in the structure and thus occupy the less numerous sites. This was not shown in the Mossbauer spectroscopy results. The three sites observed for the iron(Ir1) cations give a relative ratio which did not agree with the substitution of Fe2+ by Mn2+ which would lead to a 48 :48 :4 ratio. The discrepancy between the relative intensities of the observed doublets (45 : 33 : 22) and the relative ratio of the crystallographic sites (40:40 : 20) can only be explained if the Mn2+ ions occupy the Fe(3) sites of the structure normally occupied by Fe3+ ions.In such a case, the ratio calculated from Mossbauer spectroscopy and chemical analysis would be 39 : 42 : 19, which is approx- imately the theoretical ratio. The Fe2+ site Fe(1) would then be occupied by Fe3+ cations and the Fe3+ site Fe(3) by Mn2+ cations: [Fe~1][Fe~f2Mn~,8][Fe111](P04)3(oH)4~800~2. This interpretation does not fully agree with the results of the crystal structure refinement which points to the occupation of the Fe(3) site only by trivalent cations.' However, note that a possible substitution of the iron cations in this site by Mn2+, +CaZ or Fe2 + cations has already been proposed. ' Finally one may suggest that the first water depature, which occurs at a lower temperature and corresponds to 0.20 mol of water per mol of rockbridgeite, could correspond to the formation of water from hydroxy groups and neighbouring oxygen ions already present in the structure.References 1 D. Rouzies, J. M. M. Millet, D. Siew Hew Sam and J. C. VCd-rine, Appl. Catal., submitted. 2 D. Rouzies and J. M. M. Millet, Hyperfine Interact., 1993, 77, 11. 3 D. Rouzies, Ph.D. Thesis, Lyon, 1992. 4 M. Ijjaali, M. Malaman and C. Glietzer, Eur. J. Solid State lnorg. Chem., 1989,26,73. 5 P. B. Moore, Am. Mineral., 1970, 55, 135. 6 Y. Nathan, G. Panczer and S. Gross, Thermochim. Acta, 1988, 135, 259. 7 M. Gleith, Am. Mineral., 1953, 38, 612. 8 J. M. M. Millet, to be published. 9 P. Keller, N. Jb Miner. Abh., 1980, H12,561. 10 M. L. Lindberg, Am. Mineral., 1949,34, 541. 11 D. Hanzel, W. Meisel, D. Hanzel and P. Gutlich, Solid State Commun., 1990,76, 307. 12 R. F. Vochten, E. de Grave and G. Snoops, N. Jhar. Miner. Abh., 1979, 137, 208. 13 J. L. Dormann and J. F. Poullen, Bull. Miner., 1980, 103, 633. 14 J. L. Jambor and J. E. Dutriziac, N. Jb. Miner. Abh., 1988, 159, 51. 15 G. Amthauer and G. R. Rossman, Phys. Chem. Miner., 1984, 11, 37. Paper 4/03 19 1A ;Received 3 1 st May, 1994