Nickel Catalyst Regeneration Thermodynamics
The Journal of Physical Chemistry, Vol. 83, No.
21, 1979 2771
Thermodynamic Model for the Regeneration of Sulfur-Poisoned Nickel Catalyst. 1. Using Thermodynamic Properties of Bulk Nickel Compounds Only A. R. Chughtai and J. R. Rlter, Jr." Chemistry Department, University of Denver, Denver, Colorado 80208 (Received May IO, 1979) Publication costs assisted by the University of Denver
By the use of the modified computer programs of Gordon and McBride for the determination of heterogeneous phase and chemical equilibria at preassigned temperatures (300-1100 K) and atmospheric pressure (101325 N m-2),the oxidation with O2 of sulfur-poisonedRaney nickel catalyst and subsequent reduction with Hz have been modeled thermodynamically by using the properties of bulk nickel compounds. An alternate process, the direct reduction with Hz of the sulfidized nickel, has also been modeled and arguments are advanced for the further investigationof this thermodynamically favored second process. In both processes the mole ratios of reactants, H2/NiSO4and H2/Ni3S2,respectively, for complete disappearance of the last product to be reduced, Ni3S2,increase markedly as the desired temperature for complete thermodynamic reduction decreases. These ratios and the equilibrium activity quotient PH2/PH2shave been determined as quantitative functions of this critical reduction temperature. A complete thermodynamic hierarchy of oxidation processes for the reaction of O2 with mixtures of Ni and NiaS2is developed. From the equilibrium calculations it is brought out that Ni3S2is relatively more stable both to oxidation with O2than is Ni and to reduction with H2than is NiO. One point of modest connection with experiment is presented for the reduction processes.
1. Introduction Renewed interest has been shown in recent years in the use of Raney nickel as a hydrogenation catalyst.l Its utility as a catalyst for the gasification of coal is well established as is its eventual poisoning by the ubiquitous coal sulfur. In this work the poisoning step and two alternative regeneration processes are modeled thermodynamically, with the bulk nickel compounds and their thermodynamic properties serving as approximate representations of the surface species of eventual interest. Without belaboring the point it should be emphasized that all results of this investigation are thermodynamic ones with the underlying assumptions of complete heterogeneous phase and chemical equilibria. These assumptions are certainly suspect, particularly with condensed phases as both reactants and products. Nevertheless the experimental mole ratios and temperatures have been determined at which sulfidation, oxidations, and reductions would occur if adequate kinetics prevailed. With these limitations in mind the following four steps have been separately modeled as batch processes in the range 300-1100 K a t 100 K intervals and at atmospheric pressure (101325 N m-2): (i) the poisoning of nickel with various amounts of sulfur; (ii) the oxidation of various compositions of nickel sulfide/nickel charge with a range of oxygen/charge ratios; (iii) the reduction by hydrogen of the nickel sulfate and nickel oxide produced in step ii; (iv) the direct reduction of the nickel sulfide, produced in step i, by hydrogen without the intervening oxidation step. The minimum reactant mole ratios n~ / ~ N ~ S (step O iii) and nHz/nNi3Sz (step iv) required for complete reduction of the charges to metallic nickel have been determined as functions of temperature. 2. Methods The heterogeneous phase and chemical equilibria programs of Gordon and McBride2 were used with only minor modifications on the Burroughs B6700 at the University of Denver. The most important of these modifications were a substantial increase in the number of condensed phase compounds which could be simultaneously considered and a large increase in the sensitivity
TABLE I: Compounds Considered in the Thermodynamic Calculations H 0 3 Ni NiS(s,I) S Ni(1) NiS(s,II) HOZ Ni(s) NiSO,(l) H, S(1) NiO NiSO,(s) S(S,I) HZ 0 NiS,(s) H,0(1) S(s,II) NiO(1) HZO, SH NiO(s) Ni3sZ(1) so Ni(OH), Ni, S, (s,I) H,S 0 so2 NiH Ni3S,(s,II) OH so3 NiS Ni 3 s,(s) 0 2 SZ NiS(1) Niz5Sz I (s)
to very minor gaseous compounds so that a broader class of equilibrium constants might be hand-~hecked.~ These programs assume that the gas phase is ideal and that each liquid and solid condensed phase compound is present as a separate immiscible phase at unit activity thus completely neglecting condensed phase solution effects. By fixing the final temperature and total pressure the equilibrium products are completely determined from a given elemental reactant ratio by assuming the availability of an infinite heat source or sink at this final temperatures3 The final enthalpy of the equilibrium products then becomes a dependent variable in this case in contrast to the more familiar one of determining the adiabatic flame temperature as a dependent variable with the assigned enthalpy and pressure as independent variable^.^ Within the limitation of the nickel-rich portion of the nickel-sulfur phase diagram5 as discussed below, all possible reactions were considered involving products for which entropy, heat of formation, and heat capacity data were available; these compounds are listed in Table I. Given the restriction to the bulk compounds only, no significant omissions are thought to occur. The source of the thermodynamic data used for all but one of the nickel compounds, including the metal, is the work of Mah and Pankratz6 The single exception is the data for the minor vapor phase molecule NiS.7 All pertinent compounds of sulfur, oxygen, and hydrogen have their thermodynamic data taken from the JANAF tables.* The tendency toward euphoria regarding the apparent precision and presumed accuracy induced by computer
0022-365417912083-2771$01.00/0 0 1979 American Chemical Society
2772
A. R. Chughtai and J. R. Riter
The Journal of Physical Chemistty, Vol. 83, No. 21, 1979
printout is well known. This was overcome in the present work by extensive hand checking of scores of the results of all types of the phase and chemical equilibria with the original thermodynamic data.6-8
7,51 65
3. Results The reactants for step i were varying ratios of nickel to sulfur. Since the vapor pressures of the solid products were less than atmospheric a small amount of helium was added as a reactant as a computational trick to maintain a gaseous pressure of 1 atm. The calculated results for this step can be succinctly summarized. For less than 26.7% w/w sulfur the products are Ni and Ni3Sz at all temperatures considered. For reactants with between 26.7 and 35.3% sulfur the products are Ni3S2and NiS, as expected from the phase diagram: at all temperatures through 1000 K. At 1100 K the NiS (mp 1066 K) breaks down somewhat into Ni3S2(mp 1064 K) and S2 vapor. A variety of other condensed phase nickel sulfides were allowed as possible products (cf. Table I) and would have been important at lower Ni/S ratios. In the experimental situation nickel would always be expected to be in larger excess, so that without loss of generality the only products from the step i computation considered as reactants for the computations of step ii and iv were Ni and Ni3S2,corresponding to a sulfur content of less than 26.7% w/w. In step ii charges ranging in composition from pure nickel to pure Ni3S2were allowed to computationally react with varying amounts of oxygen. As always in these calculations only the elemental reactant ratios and the final temperature and total pressure are fixed. No other premises are needed and indeed if introduced would act as a constraint on the search for a minimum in the total free energy subject to all possible phase and chemical equilibria. Thus the conclusions described immediately below are results of the equilibrium computations and not assumptions. If the oxygen in the step ii computations was more than enough to convert all of the Ni to NiO and all of the Ni3Sz to 2NiS04 NiO, these two products plus the excess oxygen formed an invariant mole fraction set from 300 to 1100 K, and for the reasons outlined below gave the reactants for step iii. If a deficit of oxygen, with respect to complete conversion to NiO and NiS04, is used in step ii then the following priorities are found to apply for oxidation of Ni and Ni3S2in the range 300-600 K: (a) oxidize as much as possible of the Ni to NiO, leaving the Ni3S2unchanged; (b) if all the Ni NiO, then O2begins to attack Ni3S2: Ni3S2 1/202 NiO + 2NiS
+
+
-
-
(c) if all of the Ni3S2has reacted as above, then 02 attacks NiS: NiS + 2 0 2 NiS04
(d) if all of the NiS has been oxidized as above, then and only then would any O2 remain in the gas phase as an equilibrium product. At about 700 K all NiS reacts with NiO in about a 3:l ratio to form SOz and Ni3S2. NiS04 can serve as either a minor reactant or minor product of this reaction, and reacts itself somewhat at about 800 K and above to form NiO, 02,SO2,and SOs. Except for this gradually increasing decomposition to the vapor phase and NiO these products persist to 1100 K. Again, all of the preceding arguments are based upon a deficiency of oxygen with respect to xNi + Ni3S2+ [(x + 9)/2]Oz 2NiS04 + (x + 1)NiO
-
Thus we see that treating the Ni/Ni& charge in step ii
\
\
7.01
I
\ (nHz/nNNi3S2) AS STEP I V REACTANT
50t
1
At Complete Reduction
25
2?oo
For Both Steps I11 and I V
,
,
500
T
p
~
~
~
700
~
\, ~
900
~
n
g
e
f
o
I100
TEMPERATURE / K
Figure 1. Reactant ratios for step iii and step iv reductions as functlons of temperature at which reduction to metallic nickel is just complete.
with this deficiency of oxygen does not drive sulfur completely to the sulfate, but that the nickel sulfides change oxidation states as a function of temperature. One must then take, together with H2, as step iii reactants mixtures of NiO and NiS04because of priorities a-d above. For computational steps iii and iv the equilibrium data are presented quantitatively in Figure 1. The ordinates reare the reactant ratios, nHz/nNiSO4 and nHz/nNis~z, spectively, at which the last product to be reduced, Ni3S2, completely disappears as a function of temperature. These results of Figure 1 were obtained by using the GibbsHelmholtz equation to get the limiting equilibrium activity quotients at temperatures bracketed by the upper and lower values of the 100-K intervals, with reaction enthalpies determined at these same intervals. The ordinate for step iii is varied over two orders of magnitude and that for step iv over almost five orders. One sees that the lower curve for step iii is displaced from the upper one by log 2 = 0.30 which corresponds to the ratio of S atoms in Ni3Szand NiS04. The mole ratio of NiO/NiS04 reactant in step iii varied from 0.5 (pure Ni3S2reactant in step ii) to 18.6 ( n N i / n ~ = i ~36.2 ~ ~ in step ii reactant) at each of the points whose locus forms the lower curve, with no resulting displacement of this curve. The reason is that the amount of H2necessary to reduce NiO is stoichiometric and thus entirely negligible in comparison with that necessary to reduce NiS04 to Ni3S2 and hence to Ni in step iii. The lower curve also gives the ratio H2/H2Sat the critical temperature at which the last of the Ni3S2is reduced in both steps iii and iv. 4. Discussion It is seen in Figure 1that the predicted limiting values of the H2/H2Sratios at 360 and 400 OC are 4 X lo4 and 2 x lo4, respectively. From poisoning testsg on Raney nickel catalysts at these temperatures at the U.S. Bureau of Mines Albany Metallurgy Research Center essentially all of the H2Sin a flowing gas containing 8% Hz, 1% Co, and about 2 ppm of H2Swill be removed by the nickel until all of the catalytically active sites have been poisoned. Approximately one HzSmolecule is required to poison one active site, and further reaction of the nickel with the HZS was not ob~erved.~ Thus the input H2/HzSratio is about 4 x lo4 which is close to the above computed values for
~
~
T
m
Adsorption of NCS-, OCN-, and CN- by A1,03
The Journal of Physical Chemistry, Vol. 83, No. 21, 1979
complete reduction. However, the observed experimental reaction was complete sulfidation of the active sites, the opposite of the reduction reaction of step iv. Furthermore, assuming "essentially all" means at least 90%, the H2/H2S ratio for complete sulfidation could be at least 4 X lo5 and the ratio of H2/H2Sfor complete reduction of the active surface could well be 106-107. Nonetheless these extrapolated values connected with experimental results indicate that the equilibrium values are lower bounds as one would expect. Also the thermodynamic properties for the bulk materials are likely somewhat different from those of the surface, particularly a catalytically active surface. Comparing the two cycles i ii iii and i iv on the basis of the calculations presented here leads to the conclusion that the combination oxidation/reduction process is wasteful because of the necessity of carrying the unsulfidized nickel completely through the oxidation and subsequent reduction steps as discussed above. One would expect to lose the characteristic high surface area of the nickel in the process. The direct reduction of the Ni3S2 by H2 is seen to be thermodynamically possible. Such a reduction would not be expected to affect the unpoisoned Raney nickel. One would also have the option of pre- or post-reduction separation of the unaffected nickel from the poisoned/reduced material. I t should be emphasized that the modeling summarized in this work is for batch processes only, with some gains to be made in driving the reduction reaction by removing the H2S as it is generated. The possibility of checking these ideas in a set of bench scale experiments seems attractive.
+ +
2773
Finally a short discussion of the pervasiveness of Murphy's law may not be out of order. Its applicability in these calculations was unforeseen by the authors. Just as oxygen in step ii reacts preferentially with nickel metal over Ni3S2,so does the hydrogen in step iii reduce NiO to Ni but NiS04 only to Ni,Sz until a large excess of H2 is present. In brief the compound Ni3S2at all temperatures considered is more stable to oxidation with O2 than is metallic nickel and is also more stable to reduction with H2 than is NiO.
Acknowledgment. It is a pleasure to acknowledge the original suggestion of steps i, ii, and iii of the thermodynamic model by Arne Landsberg and a very helpful critical review of this work by James Russell, both of the Albany Metallurgy Research Center, US. Bureau of Mines, and support for this research from that agency to the University of Denver through Grant No. G0122101.
+
References and Notes See, e.g., "Symposium in Commemoration of the 50th Anniversary of the Discovery of Raney Nickel", 174th Meeting of the American Chemlcal Society, Chlcago, Ill., Sept 1, 1977. S.Gordon and B. J. McBride, US. National Aeronautics and Space Admlnistration Document SP-273, National Technical Information Service, Springfield, Va., 1971. A. R. Chughtai, H. M. Harris, and J. R. Riter, Jr., Met. Trans., 8B, 507 (1977). See, e.g., D. H. Yean and J. R. Riter, Jr., J . Chem. Educ., 51, 505 (1974). G. Kullerud and R. A. Yund, J. Petrol., 3, 126 (1962). A. D. Mah and L. B. Pankratz, U . S . Bur. Mlnes Bull., No. 688 (1976). M. W. Chase, private communication. D. R. Stull et al., "JANAF Thermochemical Tables", 2nd ed, Natl. Stand. Ref. Data Ser., Nafl. Bur. Stand. (1971). J. H. Russell, private comrnunicatlon.
An Inelastic Electron Tunneling Spectroscopy Study of the Adsorption of NCS-, OCN-, and CN- from Water Solution by AI2O3 Ursula Mazur and K. W. Hlpps" Department of Chemistry and Chemical Physics Program, Washington State Unlverslty, Pullman, Washington 99 164 (Received April 23, 1979) Publication costs assisted by the National Science Foundation
The technique of inelastic electron tunneling spectroscopy (IETS) is applied to the study of the adsorption of small inorganic ions from water solution on aluminum oxide. The ions show a relative strength of adsorption in the order NCS- > OCN- >> CN-. At room temperature, two forms of surface bonded NCS- ion have been observed, whereas only one form of OCN- is detectable and no measurable adsorption of CN- from solution could be found. In sharp contrast with infrared studies of HOCN adsorbed on Al,03 from the gas phase, we observe only the oxygen bonded OCN- species at all temperatures in the range from 0 to 200 "C. In the case of the NCS- ion, the method of oxide formation affects the observed spectrum. We infer from the data presented that hard bases will be preferentially adsorbed from water solution while soft bases will be either poorly or not at all adsorbed.
Introduction This project was undertaken to address three problems. The first was the general question of the applicability of IETS to the study of ionic inorganics. The vast majority of existent IETS studies have used organic or organometallic compounds which are uncharged. These neutral molecules can be directly adsorbed to the insulating layer of the tunnel diode from the gas phase. Many inorganic ions, however, are not stable with respect to the gas phase. It has been demonstrated1v2that organic molecules can be 0022-365417912083-2773$01 .OO/O
adsorbed on the A1203layer of the standard A1-A1203-Pb diode by solution phase doping. Since many inorganic ions are water solution stable, we chose to attempt incorporating inorganic ions into tunnel diodes by this method. Thiocyanate, isocyanate, and cyanide ion were chosen as simple representatives to initiate the above study. The second question that prompted this research involves the stability of inorganic transition metal complexes with respect to aluminum oxide. We have encountered serious decomposition of metal cyanide and thiocyanate 0 1979 American Chemical Society