Alternative Catalyst Supports for HCN Synthesis and NH3 Oxidation

Transcription

1 Ind. Eng. Chem. Res. 1993,32, Alternative Catalyst Supports for HCN Synthesis and NH3 Oxidation Introduction Daniel A. Hickman,+ Marylin Huff> and Lanny D. Schmidt' Department of Chemical Engineering and Materials Science, University of Minnesota, Minneapolis, Minnesota We have examined several alternative catalyst supports consisting of various materials, geometric configurations, and catalyst loadings (Pt and Pt/Rh) for the Andrussow process for HCN synthesis and for the Ostwald process for NH3 oxidation. With increasing residence time, the HCN hydrolysis reaction decreases the amount of HCN formed. Experiments also show that the addition of Rh to a Ptcoated support slightly reduces the selectivity of HCN synthesis and that catalytically active supports can adversely affect the reaction selectivity. For the Ostwald process, increasing residence times cause the decomposition reactions of NO and NH3 to contribute and thus decrease the net NO, formation. In both HCN synthesis and NH3 oxidation, turbulent, wellmixed flow improves the product selectivity. Based on these tests, guidelines are presented for the design of an alternative Ptcoated catalyst support which gives competitive selectivities and conversions. HCN Synthesis. HCN synthesis is carried out industrially in one of two ways. In the Degussa process (Koberstein, 19731, Pt is deposited on the walls of ceramic tubes and CH4 and NH3 react endothermically on the inside of the tubes to form HCN, CH, + NH3 HCN + 3H, AH,,, = 61 kcal/mol (1) while the temperature of the tubes is maintained at 1500 K by the combustion of CH4 outside the tubes. Approximately 90% of the NH3 is converted to HCN. The more common approach, the Andrussow process, employs a gauze catalyst (Satterfield, 1991; Twigg, 1989). The commercial catalyst, consisting of 2050 layers of woven Pt10 % Rh wires, is typically only a few millimeters thick and up to several meters in diameter. In this process, a mixture of CH4, "3, and air is passed over the gauze catalyst at about 2 atm. This process is exothermic with an adiabatic temperature of about 1400 K. Linear velocities of the gases are on the order of 1 m/s, giving a total contact time of about 1 me with a negligible pressure drop. The desired overall reaction for this process is CH, + NH3 + 3/20, HCN + 3H,O AH,,, = 115 kcal/mol (2) The feed composition (typically about 1:l:l) is chosen to be slightly fuel rich since HCN synthesis involves formation of a CN surface species from the two fuels (Hwang et al., 1987; Hwang and Schmidt, 1989) and so that the losses in selectivity due to oxidation of NH3 and CH4 are reduced: AH = 75 kcal/mol NH3 + 3/40, '/,N2 + 3/2H,0 (3) CH, + 1/20, CO + 2H, AH = 8.5 kcal/mol (4) CH, + 20, CO, + 2H,O AH = kcal/mol (5) The oxidation of CHI over Pt (Hickman and Schmidt, 1992a) and Rh (Hickman and Schmidt, 1992b) has been addressed elsewhere. ~~~~~~~~~~ ~ ~ ~ ~ ~ * To whom correspondence should be addressed. + Current address: The Dow Chemical Company, Midland, MI * Supported by DARPANDSEG Graduate Fellowship. The gauze catalyst is supported from below by a ceramic structure which is immediately followed by a heat exchanger which cools the products before the highly unstable HCN can react according to the reaction HCN + H,O NH3 + CO oaaa5aa519312~320a09~04.00/ American Chemical Society AH298 = 12 kcal/mol (6) The HCN synthesis process operates adiabatically at 1400 K, with HCN yields around 6070% based on NH3 fed (Satterfield, 1980). Ammonia is usually recovered from the product gas and recycled. An early study of a pilot plant reactor (Pan and Roth, 1968) examined the effect of feed composition on reaction selectivities. These studies demonstrated that the optimal feed composition is a compromise between selectivity of HCN formation based on NH3 reacted and overall HCN production. For example, for a fixed fueltoair ratio, the maximum selectivity is obtained at a higher CHJNH3 ratio than the maximum overall HCN production. These maxima in selectivity and HCN production occurred near minima in the surface and product gas temperatures. At higher fractions of fuel ("3 and CHI) for a given NH3/ CHI ratio, the amount of unreacted NH3 (ammonia leakage) increased. However, the amount of NH3 converted to Na by reaction 3 decreased with increasing fuel levels. In the case of NH3 oxidation, in the absence of CHI, virtually all NH3 was oxidized. The addition of CH4 inhibited this reaction, but in fuellean conditions this inhibition is less pronounced than in fuelrich conditions. The kinetics of HCN synthesis over polycrystalline Pt and Rh foils have been examined at low pressures, with the rates being fit to a form of LangmuirHinshelwood kinetics (Hasenberg and Schmidt, 1985,1986,1987). These rate expressions have been used in an atmosphericpreseure model that simulates quite well the performance of an Andrussow reactor using 13 simultaneous reactions (Waletzko and Schmidt, 1988). Several studies have addressed the microstructural properties of gauze catalysts used for HCN synthesis (Pan, 1971; Schmidt and Luss, 1971; Cowans et al., 1990). The gauze catalyst takes about 60 h to reach its maximum conversion of NH3 to HCN and then slowly deactivates with time (Pan, 1971). The initial activation is accompanied by a roughening of the catalyst surface which increases the catalytic activity. Presumably, as carbonaceous deposits accumulate, the number of active sites decrease and the gauze deactivates. Although this process operates at significantly higher temperatures than the Ostwald process for NH3 oxidation (1100 K), the loss of

2 810 Ind. Eng. Chem. Res., Vol. 32, No. 5, 1993 Pt from the gauze is significantly smaller in the Andrussow process because the formation of volatile PtO2 requires a strongly oxidizing environment (Satterfield, 1980). The literature is generally limited to studies of the gauze catalysts used in industry, with few studies of alternative supported catalysts (Hutchings, 1986). The purpose of this study was to test several alternative catalyst configurations as substitutes for the gauze catalyst in the Andrussow process. This search for a viable alternative was motivated by several factors. 1. Because the gauze is composed of pure noble metals, the capital cost associated with a gauze pack (Pt10% Rh) for a reactor about 1 m in diameter is on the order of $1 million, and a large portion of this cost is due to the Rh content of the gauze. Platinum currently costa about $380 per troy ounce, while the cost of Rh in recent years has beenabout 10 times that of Pt (Espinosa, 1991). However, Rh is added to the gauze because it increases the mechanical strength of the gauze, not necessarily because of its catalytic properties. Thus, by coating an inexpensive substrate with a thin layer of pure Pt, one should expect toreduce the catalyst capital cost by an order of magnitude. 2. The conversion and selectivity obtained in the current Andrussow process leave much room for improvement. In fact, a large portion of the capital and operating costs for this process are associated with separation of NH3 from the product stream, which typically contains 13% "3. Thus, an alternative catalyst would find favor if it were able to give equivalent or superior HCN conversion levels while reducing the amount of NH3 in the product stream. 3. The examination of catalyst supports having different geometries, material properties, and catalyst loadings should yield valuable information pertaining to the influence of these properties on the performance of an HCN synthesis reactor. The various geometries and materials of the catalyst supports result in a wide array of characteristic flow patterns and thermal properties. In addition, samples can be prepared with a variety of noble metal loadings and compositions. Examining the performance of these various catalysts and supports should yield some valuable insights into the relative importance of the physical and chemical phenomena that occur in the Andrussow process. 4. Finally, the geometry of a catalyst support can have a significant impact on the selectivity of a heterogeneous process by affecting the masstransfer rate from the gas phase to the catalyst surface (Hickman and Schmidt, 1992~). Comparing the HCN synthesis selectivities of several catalyst geometries should yield additional insights into this effect. NH3 Oxidation. The Ostwald process, the superoxidation of NH3 to NO, is the key step in the production of nitric acid. Nitric oxide and NO2 are then absorbed in HzO to form HN03 (Satterfield, 1980). The current industrial process uses Pt10% Rh gauze as the catalyst. The gauze pack is typically 4 m in diameter and may contain as many as 40 layers of gauze. The reaction takes place at 1100 K and atmospheric pressure. With a feed composition of 10 % NH3 in air, the process is as high as 97 % selective to NO, with nearly complete conversion of "3. This reaction is extremely fast with a typical surface contact time of 1 ms (Satterfield, 1991; Honti, 1976). The desired overall reaction for this process is the superoxidation of NH3 to NO. The excess 0 2 is consumed by homogeneous reaction with NO to form NO2. NH, + 5/40, NO + 3/2H,0 AH = 54 kcal/mol (7) NO + '/,O, NO, AH = 14 kcal/mol (8) There are several reactions that occur on the catalyst which compete with these to reduce the overall NO, selectivity. '/,N, +,/,H,O AH = 75 kcal/mol (9) NO '/2N, + '/,02 AH = 22 kcal/mol (10) NH, + 3/40, NH3 + 3/2N0 5/4N, + 3/2H20 AH = 104 kcal/mol (11) In order to maximize the production of NO,, the contributions of these reactions must be reduced. We compare the effectiveness of a foam monolith supported Pt catalyst to the industry standard Pt10% Rh gauze catalyst. We evaluate the catalyst effectiveness as a function of feed composition, reaction temperature, and flow rate. Experimental Section Catalysts. Catalyst configurations of four basic types were examined: Pt10% Rh gauzes, Ptcoated A1203 foam monoliths, Ptcoated cordierite (a magnesium aluminosilicate) extruded monoliths, and Ptcoated metal monoliths. Systematic studies of these four groups permitted the evaluation of the effects of several parameters on the production of HCN. For the NH3 oxidation studies, the gauze and foam monoliths were compared. As shown in Table I, important parameters include the thermal conductivity of the substrate, the flow patterns in the catalyst support, and the dimensions of the channels within the catalyst. In addition, samples containing a variety of total loadings and noble metal compositions were prepared and tested. The gauze catalysts were 40 and 80mesh (40 and 80 wires per inch) woven wire samples which were cut into 18mmdiameter circles and stacked together to form a single gauze pack 110 layers thick. They were typically sandwiched between two extruded ceramic supports. The foam monoliths were aal203 samples with an open cellular, spongelike structure. We used samples with pores per inch (ppi) which were cut with a core drill into 17mmdiameter cylinders 220 mm long. A coating of Pt or Pt/Rh was then applied directly to the alumina by a technique involving organometallic deposition. Relatively high loadings of noble metals were used. The foam monolith samples used in this work had loadings which varied from 2 to 20 wt % noble metal. Scanning electron microscopy (SEM) micrographs of these catalysts before and after use revealed that the catalyst formed large crystallites ( N 1 pm) on the support with the metal covering a significant fraction of the support surface. The cordierite extruded monoliths, having 400 square cells/in2, were similar to those used in automotive catalytic converters. However, instead of using an alumina washcoat as in the catalytic converter, these catalyst supports were loaded directly with 1214 wt % Pt in the same manner as the foam monoliths. Because these extruded monoliths consist of several straight, parallel channels, the flow in these monoliths was laminar (with entrance effects) at the flow rates studied. The metal monoliths were prepared by electroplating a strip ( 1 cm wide) of the metal with Pt. The Pt loadings were typically 15 wt %, and the metal support compositions included various ironbased alloys (such as

3 Table 1. Characteristics of Catalysts Ind. Eng. Chem. Res., Vol. 32, No. 5, thermal characteristic channel flow support material conductivity geometry diameter (mm) characteristics gauze extruded monolith Pt10% Rh wires cordierite high low woven mesh uniform; straight < ( cells/in.*) turbulent laminar channels foam monolith alumina low cellular (3050 ppi) turbulent metal monolith various iron and high corrugated, about 1.0 laminar cobalt based concentric alloys, nickel cylinders Table 11. Comparison of "Best" HCN and NO, Selectivities and Conversions for the Four Types of Catalysts HCN synthesis NH3 oxidation support % NH3 breakthrough 7% CHI breakthrough HCN/NH3 in NO,/NH3 in geometry or material gauze foam monolith extruded monolith 59 metal monolith 12 stainless steel 316 and 3041, acobalbbased alloy, and nickel. The metal strips were then coiled to form a monolithic structure with a diameter of 18 mm and wall spacings of approximately 1 mm. Apparatus and Procedure. The performance of each substrate was evaluated by reacting "3, CH4, and air at various compositions and flow rates. Typically, the linear flow rate of the cool feed gases was 1030 cm/s. The product gases for HCN synthesis were analyzed using gas chromatography (GC). Gasphase infrared spectroscopy was used to analyze the product gases for NH3 oxidation since NO and NO2 react with H2O in the GC column to form HN03, which destroys the column packing. Mass flow controllers were used to set and control the flow rate of each reactant. The wellmixed gases were fed to a quartz tube with an inner diameter of 18 mm. This tube contained a preheating section followed by the reaction zone. The reaction zone contained the catalyst sample placed between radiation shields, usually noncatalyticceramic monoliths. The outside of the reaction tube was insulated and the reactor was operated autothermally in order to better simulate the adiabatic operation of the industrial process. Heated lines (maintained at "C to prevent condensation of H2O and polymerization of HCN) carried the product gases from the reactor to the gas chromatograph or infrared spectrometer and then to the process incinerator. HCN Synthesis Product Analysis. The gas chromatograph performed two separate analyses. In all cases, a thermal conductivity detector was used. In the first analysis, 1 ml of the gases was injected into a He carrier gas stream which passed through a 6 ft. X l/a in. column containing 60/80mesh Hayesep R. Temperature programming was used to quantitatively separate "3, H20, and HCN, while the remainder of the components eluted very quickly. After this analysis was complete, a second 1mL sample was injected into a second column 11 ft. X '/a in. column containing 50/80mesh Porapak T. The outlet of this column was connected to a 10 ft., X '/a in. column with 80/100mesh Molecular Sieve 5A. Hydrogen, the first component to elute, had a thermal conductivity very similar to helium, resulting in low sensitivity, nonlinearity, and peak reversals with thermal conductivity detection. Thus, the hydrogen was transferred from the helium carrier gas into a nitrogen stream using a palladium tube heated to 580 "C. After H2 eluted, a valve was switched so that flow through the Molecular Sieve column was stopped. C02 was then allowed to elute from the Porapak T column. Finally, flow through the Molecular Sieve 5A column was resumed, and 02, N2, CH4, and CO layers of gauze ppi X 6 mm (HCN) 30 ppi X 9 mm ("3) csi X 2 mm 0.04 Haynes alloy 188 (Cobased alloy) eluted from the column while the remaining components were removed from the system by reversing the flow of carrier gas through the Porapak column. In order to convert the integrated peak areas into mole fractions, calibration runs were performed periodically. Calibration gases were used for all gases except HCN and H20. HCN and H2O calibrations were then estimated by performing mole balances on carbon and oxygen, respectively, for several GC analyses. The mole number change due to reaction was simultaneously determined by an atomic mole balance on nitrogen. The GC results were quite reliable and reproducible, with all of the atomic mole balances (C, H, 0, and N) typically closing to within f5%. NHs Oxidation Product Analysis. Ammonia, NO,, and H2O interact strongly following reaction, and conventional analysis by gas chromatography is difficult. We analyzed the products using infrared spectroscopy by flowing them through a 10cm heated (200 "C) IR cell with ZnSe windows. The concentrations of H20, "3, NO, and NO2 were determined by measuring the intensity of transmitted light at ,3334.1,1913.3, and cml, respectively, and applying the BeerLambert law for mixtures. Since some Components absorb at more than one of these wavenumbers, the transmissions of these species were calibrated at each of these wavenumbers. The concentrations of NZ and 0 2 were determined by mole balances. Due to systematic inaccuracies caused by the overlap of the IR spectra of the product components, the calculated concentrations are accurate to within approximately &5 % of their values. HCN Synthesis Results The data in Table I1 compare the HCN synthesis performance of the four substrates. In each case, the results are shown for the best HCN synthesis catalyst within that category. The HCN selectivity and NH3 and CH4 conversions measured for the gauze are essentially the same as in industrial reactors. As we will demonstrate in this paper, the foam monolith is the best of the three alternative configurations because of the wellmixed, turbulent flow that results from its cellular structure and the inactivity of the catalyst support. The results of the experiments which examined the three alternative catalyst systems are presented in this section. Foam Monoliths. In order to determine the optimal configuration of a foam monolith for HCN synthesis, the effects of the following parameters were examined: (a)

4 812 Ind. Eng. Chem. Res., Vol. 32, No. 5, % Pt 130 ppi foams HCN 0 1 " " " C CH,lair length (cm) Figure 1. Effect of residence time on HCN production Ptcoated foam monoliths. All three monoliths were 30 ppi foams with 13.8 f 0.2 wt % Pt. The feed consisted of 3 slpm (standard litersper minute) total flow with molar ratios of NHdair = and CHJair = Multiple points for a given length are for separate analyses and result from a combination of experimental error and changes in catalytic activity with time. the length of the foam, (b) the average cell size in the foam, and (c) the noble metal composition and loading on the foam. Figure 1 shows the effect of monolith length (and thus residence time) on the product composition. Foams of three different lengths but identical Pt loadings (13.8 f 0.2 wt %) were tested. Since all three foams contained 30 ppi, the flow patterns should be identical at all lengths. As illustrated in this figure, increasing the residence time significantly reduces the amount of HCN produced, while CO production increases and the amount of NH3 in the product gases passes through a minimum. This result reveals that HCN is formed primarily at the front end of the catalyst, while the HCN hydrolysis reaction (6) reduces the amount of HCN produced and increases the amount of NH3 in the product stream. Such behavior is consistent with the belief that a major fraction of the overall reaction occurs on the first few layers of gauze in the industrial Andrussow process. Since the NH3 minimum does not coincide with an HCN maximum, an optimal length exists somewhere between these two extrema and would be a function of the economics of NH3 separation from the product gases in an industrial process. A comparison between the product gas composition from a gauze pack four layers thick and the 30 ppi (pores per linear inch) X 6 mm foam monolith from Figure 1 is shown in Figure 2. In comparison to the gauze, the 30 ppi foam gives lower selectivities for HCN production based on NH3 and CH4, where the selectivities (or yields) are defined as 'CH, = ACH, (13) where nhcn is the molar rate of production of HCN and Ai is the molar rate of conversion of species i. At CHdair = 0.19, the 30 ppi foam gives SNH~ = 0.58 and = 0.35 compared to 0.83 and 0.61, respectively, for the gauze. For the foam, as the feed ratio CH$air was increased to 0.24, S":, increased to As outlined in an earlier paper (Hickman and Schmidt, 1992c), the selectivity of formation of an intermediate product (such as HCN) in a series reaction is maximized C CH,/air Figure 2. (a) Nitrogen and (b) carboncontaining species in the product gases for 6mmthick 3Oppi wt % Pt monolith compared with four layers of 40mesh Pt10% Rh gauze at 2 slpm air and NHs/air = Hollow symbols with dashed lines represent foam monolith data, and filled symbols with solid lines represent gauze data. by increasing the rate of mass transfer of reactants and products through the boundary layer near the catalyst surface and optimizing the residence time. Thus, monoliths with smaller pore sizes were examined. Foams with high Pt loadings (>lo wt 7%) and various lengths (ranging from 2 to 7 mm) were examined, and the best HCN selectivity and NH3 conversion, defined as ~HCN NH3 conversion = n"3 in (14) were obtained for the shortest sample, with results for this 50 ppi foam compared to the gauze and 30 ppi foam in Figure 3. As shown in Figure 3a, decreasing the pore size increased SNH~ but had no significant effect on SCH~. The improvement in SNH~ is probably due to the decreased residence time in the 50 ppi foam, which results in a smaller fraction of NH3 converted to Nz by the oxidation reaction (3). On the other hand, since SCH~ did not change significantly, these data suggest that the CH4 oxidation reactions, (4) and (5), may be more significant on the supported catalysts than on the gauze. This may be a consequence of CHI oxidation reactions taking place on the alumina support or a result of differences in microstructure for supported Pt compared to gauze catalysts. The noble metal composition and loading were also varied. Experiments in which the total loading of Pt (between 3 and 14% of the total monolith weight) was varied showed that different Pt loadings do not significantly affect the HCN selectivities with the foam monoliths. However, experiments with 30 ppi foam monoliths having a wide range of total loadings revealed that the

5 Ind. Eng. Chem. Res., Vol. 32, No. 6, P w I A/ 0.41 I I I I I I I ~ weight % Rh d cc) 0.40 ib)/ 0 gauze 30 ppi 0.25 I I I I I CH,/au Figure 3. Selectivities for HCN formation based on (a) NHs and (b) CHI converted for a 60 ppi X 2 mm foam with wt % Pt compared to the four layers of gauze and the 30 ppi foam in Figure 2. addition of Rhreduces the HCN conversion and selectivity relative to a sample coated only with Pt. The product gas compositions for these different samples are shown in Figure 4. The sample at 0% Rh is the same 30 ppi foam (13.9 wt % Pt) used in earlier experiments (Figures 2 and 3). The remaining samples were of identical size and geometry, but had various total loadings of catalyst ranging from 0.8/0.2to 17, wt % PtIRh. For these samples, a constant 4/1 Pt/Rh ratio was used and the total metal loading was varied. While varying the total loading of Pt (with no Rh present) had no significant effect on the product gas composition, Figure 4 shows that the presence of Rh has a strong effect on reaction selectivity. As the amount of Rh increases, CH4 conversion increases and CO, COZ, and H2 are formed more selectively while HCN production decreases. As we will discuss later, Pt is a better catalyst than Rh for H2O formation. Since Rh makes less H20, more oxygen is available to form the undesirable oxidation products CO, C02, and N2. This observation is consistent with the fact that Rh is included in the industrially used gauzes only because it increases the mechanical strength of the gauze, not because of its catalytic behavior. In all of the above experiments, the foam monoliths gave qualitative behavior quite similar to the industrially used gauzes. Similar product distributions were obtained for a variety of feed conditions. In addition, many of the apparent shortcomings of the foam Catalysts could be improved by varying geometrical parameters (the length and average cell size). Thus, it is apparent that the catalytic behavior of the Pt coated on the ceramic I I I I I weight % Rh Figure 4. Effect of Rh on HCN Synthesis over 7mmthick 30 ppi foam monoliths. For all samples, a constant Pt/Rh weight ratio of 4/1 was used, so increasing Rh loadings are accompanied by proportionally increasing Pt loadings. Data shown are for 2 slpm air with feed molar ratios of NHa/air = 0.16 and CHdair substrates is quite similar to the catalytic behavior of the pure noble metal gauze catalysts. The foam catalysts are able to produce S":, values comparable to those found on the gauze catalyst. However, a higher CH4 to air ratio is required to achieve this. Data show that increasing the pore density causes S":, to increase and become closer to S":, obtained with the gauze catalyst. This trend is not observed for Under no circumstances was SCH, for the foam catalyst comparable to S C for ~ the gauze catalyst. This is the primary difference between the foam and gauze catalysts. Extruded Monoliths. Ptcoated extruded cordierite monoliths having a wide range of lengths and channel diameters were also teated for their usefulness as HCN synthesis catalyst supporte. Since the only significant difference between the extruded and the foam monoliths should involve the nature of the flow patterns within the catalyst support, these experiments were useful in determining the effect of the flow patterns on HCN synthesis conversions and selectivities. As shown in Figure 6, all of the extruded monoliths produced significantly less HCN than foam monoliths of simii channel dimensions and lengths. However, increasing the cell density (and thus decreasing the average cell diameter) improved the selectivity of HCN formation from NHs and gave higher C& conversions. Nonetheless, even the best extruded monolith gave lower NHs conversions to HCN than the foam monoliths. These results are probably a direct consequence of the laminar flow that exists in these straightchanneled monoliths. Laminar flow is accompanied by low mass

6 814 Ind. Eng. Chem. Res., Vol. 32, No. 5, & 8 5 'ij a& $8 4 E; a % g2 1 HCN '0 5iO Id00 15bO 2doO 2500 cells/ii2 1 Y H 4 X cells/in2 Figure 5. HCN synthesis over 6mmlong extruded cordierite monolithscoated with 1213wt % Ptandhavingvaryingcelldemitiee. Data shown are for 2 slpm air with feed molar ratios of NHdair = 0.16 and CHdair = transfer rates and higher levels of axial dispersion (relative to that obtainedwith the gauze and foam catalysts). Thus, an increased residence time is required to obtain reasonable conversions of the reactants. However, longer residence times and broader residence time distributions also allow the HCN hydrolysis reaction to have a greater impact, resulting in reduced selectivity and conversions to HCN based on NH3 fed to the reactor. By these arguments, reducing the channel diameter should improve the selectivity of HCN formation, as observed in Figure 5, by reducing the thickness of the masatransfer boundary layer. Metal Monolith. The final catalyst supports studied were metal monoliths electroplated with Pt. Because the resulting monoliths consisted of essentially concentric, annular channels, the flow should be laminar. Unlike the two groups of ceramic supports, the metal supports have a very high thermal conductivity. Several different metal supports were tested, with all exhibiting similar behavior. As shown in Table 11, the metal monoliths produced the least HCN of the four types of substrates. In addition, the product gases from the metal monoliths contained significantly more CH4 and less NH3 than the product gases from the other catalysts. This uniquely different behavior suggests that the metal support competed catalytically with the Pt by decomposing "3. With the stainless steel catalysts, a reddishbrown color (presumably iron oxide) was observed on the surface of the catalyst support after reaction. Even small amounts of iron are extremely detrimental to the selectivity of HCN formation. Experiments with gauze catalysts have shown that as little as 22 ppm surface iron will reduce SNH3 by 5 % (Pan, 1971). I h U e ' E i; C Mole Fraction NH3 in Figure 6. Catalyst exit temperature as a function of reactant composition for NH3 oxidation. In addition, a Pt foil monolith with Pt electroplated on it (in the same manner as the other samples) gave significantly higher yields of HCN, although only about as good as the best extruded monolith because of the laminar flow through the Pt foil monolith. These results show conclusively that metal monoliths with electroplated Pt fail to give good HCN selectivities because of (a) the undesired catalytic activity of the support and (b) the geometry of the catalyst. If a catalytically inactive metal support were discovered, the best geometry for high rates of mass transfer would still probably be that of a gauze catalyst. NH3 Oxidation Results Table I1 also compares the NH3 oxidation performance of the Pt10% Rh gauze and Pt on foam monolith. In both cases, the results are shown for 10 % NH3 in air fed to the reactor at 4 slpm. The results of experiments studying the selectivity of NH3 oxidation to NO and NO2 are presented in this section. Ammonia oxidation was examined over the two types of catalyst at 2,3, and 4 slpm for inlet compositions from 8 to 14% NH3 in air. The preheat temperature was varied to alter the reaction temperature so that all comparisons of catalyst effectiveness at each feed composition were made at the same catalyst exit temperature. This temperature varied with feed composition as shown in Figure 6. For the gauze catalyst, the product gas stream contained approximately 1 % "3; for the foam monolith catalyst, the product gas stream contained only 0.5% NH3 for feeds up to 12% NH3 in air. Since this NH3 composition does not change with temperature or feed composition, but increases slightly at higher flows, the incomplete conversion of NH3 was attributed to breakthrough around the edges of the Catalyst. For "3rich feed compositions, additional NH3 breakthrough was observed. The selectivity, S, and yield, Y, of the superoxidation reaction to form NO, are defined as nno + nno, snoz NH, *NH, nno + nno, Y= IZ"3in 5 (15) (16) The change in number of moles due to reaction is ignored because this change is typically less than 5 %, which is less than the accuracy of the concentration measurements.

7 The selectivity and yield of NO, products are always higher for the gauze catalyst than the foam monolith catalyst as shown in Figure 7. The selectivity and yield decrease with increasing NH3 feed concentration over the range of this study. In excess "3, the competing reactions proceed more readily. NH3 + 3/zN0 5/4Nz + 3/zH20 (17) NH3 1/2N2 + 3/zHz (18) Discussion At higher fuel concentration, the decomposition of NH3 increases in importance. Also, the excess NH3 reacts with the NO to form NZ and H20. The selectivity and yield should also decrease in the very lean region ("3 < 8%) because the excess oxygen adsorbed on the catalyst surface promotes the reduction of NO by NH3 reaction (Pignet and Schmidt, 1974). The selectivities found here are somewhat lower than those achieved industrially. The most probable cause of the discrepancy is an inaccurate calibration of the NH3 mass flow controller. However, this error is systematic and does not affect the comparison between the two catalysts. Figure 7 also shows that the selectivity increases with increasing flow rate. This effect is shown more explicitly in Figure 8 for 10 % NH3 feed at a catalyst exit temperature of 700 "C. It is seen that the gauze catalyst is always more selective than the foam. As the flow rate and catalyst temperature increase, so does the NO, selectivity. The NO, selectivity is plottedversus the gauze and foam monolith catalyst exit temperature for various NH3 feed compositions in Figure 9. The NO, selectivity of the gauze Ind. Eng. Chem. Res., Vol. 32, No. 5, catalyst and that of the foam monolith catalyst are compared at a high flow rate of 4 slpm. The two catalysts behave comparably, and some general trends can be noted. There is an optimum catalyst exit temperature where the NO, selectivity is maximized for each situation. In each case, the NO, selectivity is better over the gauze catalyst than over the foam monolith catalyst. These studies of alternative catalyst supports for HCN synthesis have demonstrated several important parameters that must be considered in the design or operation of an HCN synthesis reactor. This complex reaction system has both series and parallel reaction components, with HCN being the intermediate in a series reaction where HCN hydrolysis is the final step, and with much of the observed N2, CO, and C02 being formed by reactions parallel to the HCN synthesis reaction. Experimental evidence shows that the reaction selectivity in this complex reaction scheme is affected by the catalyst (Pt, PtRh), the catalyst support (gauze, alumina, cordierite), and the mass transfer of reactants and products through the gasphase boundary layer near the catalyst surface (flow rate, pore density). The thermal conductivity of the catalyst support may also affect reaction selectivity, although this effect was not explicitly observed independent of other effects. NH3 oxidation exhibits similar characteristics. In this case, NO, is an intermediate product where NO, is lost by both decomposition and reduction by "3. We show that "t IO Mol Fraction NH, in Md Fraction NH, in 15aoIslpm( # Mol Fraction N% in 55 50t s %OS Mol Fraction NH, in Figure 7. NH3 oxidation selectivity (a, b) and yield (c, d) to NO, over gauze and foam monolith catalyst as a function of reactant composition at a flow rate of (a, c) 2 slpm and (b, d) 4 slpm. I

8 816 Ind. Eng. Chem. Res., Vol. 32, No. 5, 1993 a5 80 t 8 I I I I Flow Rate (slprn) Figure 8. NO, selectivity over gauze and foam monolith Catalysts at 10% NH3 feed composition and 700 OC exit temperature as a function of flow rate. (a> 8% Gauze '!?! 6dO 7dO 8;0 Y;O IAO Temperature ("C) _j Temperatwe ("C) Foam Figure 9. NH3 oxidation Selectivities for NO, formation over (a) gauze catalyst and (b) foam monolith at 4 slpm as a function of reaction composition and catalyst exit temperature. the mass transfer through the boundary layer and the thermal conductivity of the catalyst support affect the NO, selectivity. Mass Transfer. HCN hydrolysis can significantly impair the performance of an Andrussow process reactor. As shown by Pan (1971), industrial PtRh gauze reactors must have contact times short enough to avoid HCN loss by hydrolysis. The effects of this reaction must be minimized by optimizing the reactor length for maximum HCN production and by maximizing the masstransfer rate of reactants to the catalyst surface. The HCN hydrolysis reaction essentially imposes very tight restrictions on the geometric parameters used in designing a supported HCN synthesis catalyst. The reduction of NO by NH3 hinders the production of NO, in the Ostwald process when the rate of mass transfer is low. Nitric oxide is formed on the surface, and if it is not removed from the surface quickly, it will be reduced by NH3 to Nz. The reactor and reaction conditions must be designed to minimize this effect. Because HCN and NO, are essentially intermediate products in a series reactions, high masstransfer rates are required to maximize the selectivities (Hickman and Schmidt, 1992~). Besides operation at high flow velocities, the masstransfer rate can be improved by choosing a catalyst geometry that induces better mixing of the gases and, thus, thinner masstransfer boundary layers. These experiments have shown that the cellular, spongelike structure of foam monoliths is superior to the straightchanneled structure of the extruded and metal monoliths. However, the foam monoliths are still inferior to the gauze. Catalyst and Support Composition. These experiments have shown that Pt/Rh supported catalysts give lower HCN conversions than pure Pt catalysts. In addition, the catalyst support may catalyze undesirable reactions. Although only the metal supports exhibited an obvious effect on the reaction selectivity, some questions remain regarding the activity of the aluminaand cordierite supports. Since SCH~ was significantly lower for both the foam and extruded monoliths, the difference in HCN synthesis activity may be due to the catalytic activity of the catalyst support or the difference in catalyst microstructure between the supported and gauze catalysts. The adverse effect of Rh on HCN selectivities is especially interesting since similar studies of direct oxidation of CH4 to synthesis gas over catalytic monoliths (Hickman and Schmidt, 1992b) have shown that Rh is superior to Pt for the production of H2 from a fuelrich CH4/02 feed. Those results have been attributed to the relative rates of formation of OH surface species on Pt and Rh. Lowpressure studies of the H202H20 reaction system on both Pt and Rh (Williams et al., 1992; Zum Mallen et al., 1993) have shown that water formation must proceed via an OH, intermediate, so the superiority of Rh over Pt for production of H2 by methane oxidation is at least partially due to the higher activation energy for OH, formation, H, + 0, OH,, on Rh (20 vs 2.5 kcal/mol on Pt). Since OH, formation from 0, and H, adatoms is slower on Rh than on Pt, more 0, adatoms are available for oxidation of other surface species. Consequently, on Rh, the surface reaction step whereby CH4 and NH3 fragments combine on the surface to form HCN must now compete with more predominant oxidation reactions. Support Thermal Conductivity. Because of the catalytic activity of the metal supports, a comparison between the highly conductive metal monoliths and the ceramic monoliths to determine the effect of substrate thermal conductivity was not possible. Because the HCN synthesis reaction (1) is endothermic, temperature gradients that may influence reaction selectivity could be supported in a ceramic structure. However, because the reactions in an Andrussow reactor are so fast, the axial length over which reactions occur is quite short, and the catalyst in an insulated reactor can be assumed to be isothermal. Hence, the thermal conductivity of the catalyst support would probablynot be a significant factor in choosing an alternative HCN catalyst system. The isothermal assumption is not valid for NH3 oxidation. The shape of the temperature profile along the length of the catalyst depends on the Lewis number. On the basis of "3, the limiting reactant, Le = 0.8 at 800 OC. When the Lewis number is less than 1, the surface temperature of the catalyst can exceed the adiabatic reaction temperature as discussed by Hegedus (1976). Because of the thermal properties of the A1203 support, the foam monolith can support a much steeper temperature gradient than the metal catalysts can. Therefore, when the exit temperatures coincide, the actual surface temperature where the bulk of the reactions take place is much higher in the foam catalyst than in the gauze catalyst.

9 This fact contributes to some of the differences in performance of the two catalysts. At higher catalyst temperatures, the decomposition reactions may be favored (Satterfield, 1980): NH, + l/znz + 3/2H20 NO '/,N2 + 1/202 (20) An increase in the importance of these two reactions would reduce the selectivity to NO,. Also, high surface temperatures can cause the catalyst to melt or sinter if the catalyst temperature exceeds the maximum adiabatic temperature (Lee and Farrauto, 1989). There is no evidence of either melting of the gauze or sintering of the Pt particles on the foam catalyst. Summary The performances of an HCN synthesis reactor and an NH3 oxidation reactor are governed by both mass transfer and chemistry. By comparing the behavior of various catalyst geometries, one can see that turbulent, wellmixed flow is necessary in order to maximize the rate of these synthesis reactions. In HCN synthesis, this maximizes the synthesis reaction (1) which is mass transfer limited. This is necessary because the HCN hydrolysis reaction (6) begins to reduce the concentration of HCN almost as soon as the HCN is formed. Consequently, the geometry of an HCN synthesis catalyst should have two basic features: (a) a geometry that maximizes the rate of mass transfer of the reactant species to the catalyst surface and (b) a length that minimizes the effect of the hydrolysis reaction. Of course, before optimizing the geometry of a catalyst support, one must choose a catalyst that is highly selective for HCN synthesis and a support that does not reduce the selectivity of the catalyst. As explained earlier, tests with various loadings of Pt and Rh deposited on geometrically identical foam monoliths show that the addition of Rh reduces the selectivity of HCN formation. In addition, the tests of the metal monoliths reveal the importance of choosing a support material that does not catalyze undesirable reactions. On the basis of these results, an alternative catalyst support for HCN synthesis may be difficult to find. The woven mesh wires of the gauze catalyst appear to have the best masstransfer characteristics, and the lack of another catalytically active material is avoided by using pure noble metal wires. However, if an inexpensive and catalytically inactive metal support were discovered, significant savings might be achieved by electroplating the cheaper metal with Pt. For Pt supported on a ceramic substrate, the foam geometry is superior to the extruded monoliths. However, if Ptcoated foam monoliths are to replace Pt Rh gauzes, the cause of the lower selectivity of HCN formation from CHI (SCH~) must be determined and addressed. The results for NH3 oxidation are somewhat similar. Turbulent, wellmixed flow maximizes the NO, production (7) and reduces the effect of NO reduction by NH3 (11). The reaction temperature should also be optimized to maximize the rate of NO production without promoting the decomposition reactions. For NH3 oxidation, the thermal characteristics of the catalyst material are highly important. More study is needed in this area before the feasibility of replacing Pt Ind. Eng. Chem. Res., Vol. 32, No. 6, % Rh gauzes with Ptcoated foam monoliths can be determined. The shape of the temperature profile along the length of the catalyst should be determined, and the NO, selectivity at high flow rates must also be addressed. Acknowledgment The authors would like to thank Dr. L. Campbell of Advanced Catalyst Systems Inc. for preparing many of the foam catalyst samples and Dr. W. B. Retallick for preparing the metal monolith samples. Literature Cited Cowans, B. A.; Jurman, K. A.; Delgass, W. N.; Li, Y. Z.; Reifenberger, R.; Koch, T. A. Scanning Tunneling Microscopy of Platinum Rhodium Gauze HCNA Catalysts. J. Catol. 1990,125,501. Espinosa, J., Ed. Metal Statistics; American Metal Market: Hasenberg, D.; Schmidt, L. D. HCN Synthesis from CHI and NH3 on Clean Rh. J. Catal. 1985,91,116. Hasenberg, D.; Schmidt, L. D. HCN Synthesis from CH4 and NH3 on Platinum. J. Catal. 1986,97, 156. Hasenberg, D.; Schmidt, L. D. HCN Synthesis from CH4, "3, and 02 on Clean Pt. J. Catal. 1987,104, 441. Hegedus, L. L. Temperature Excursions in Catalytic Monoliths. MChE J. 1975,21,849. Hickman, D. A.; Schmidt, L. D. Synthesis Gas Formation by Direct Oxidation of Methane Over Pt Monoliths. J. Catal. 1992a, 138, 267. Hickman, D. A.; Schmidt, L. D. Hydrogen and Carbon Monoxide Formation by Oxidation of Methane over Rh Monoliths. Catol. Lett , in press. Hickman, D. A.; Schmidt, L. D. The Role of Boundary Layer Mass Transfer in Partial Oxidation Selectivity. J. Catal. 1992c, 136, 300. Honti, J. D., Ed. Nitric Acid. In The Nitrogen Industry; Akademiai Kiado: Budapest, 1976; pp Hutchings, G. J. A Study of Supported Platinum Catalysts for the Production of Hydrogen Cyanide by the Ammoxidation of Methane. Appl. Catal. 1986,243, 7. Hwang, S. Y.; Schmidt, L. D. Surface chemistry of CN Bonds on Rh(ll1): I. CzNz and CH3NH2. J. Phys. Chem. 1989,93,8327. Hwang, S. Y.; Seebauer, E. G.; Schmidt, L. D. Decomposition of CH3NH2 on Pt(ll1). Surf. Sci. 1987, 188, 219. Koberstein, E. Model Reactor Studies of the Hydrogen Cyanide Synthesis from Methane and Ammonia. Znd. Eng. Chem. Process Des. Dev. 1973, 12 (4), 444. Lee, H. C.; Farrauto, R. J. Catalyst Deactivation Due to Transient Behavior in Nitric Acid Production. Ind. Eng. Chem. Res. 1989, 28, 1. Pan, B. Y. K. Characteristics of PtRh Gauze Catalyst and Kinetics of the HCN Synthesis. J. Catal. 1971,21, 27. Pan, B. Y. K.; Roth, R. G. Optimization of Yield Through Feed Composition: HCN Process. Znd. Eng. Chem. Process Des. Deu. 1968, 7 (l), 53. Pignet, T.; Schmidt, L. D. Selectivityof NH3 Oxidation on Pt. Chem. Eng. Sci. 1974, 29, Satterfield, C. N. Catalytic Oxidation. In Heterogeneous Catalysis in Practice; McGrawHik New York, 1991; pp Schmidt, L. D.; Luss, D. Physical and Chemical Characterization of PlatinumRhodium Gauze Catalysts. J. Catal. 1971,22,269. Twigg,M. V., Ed. CatalyticOxidations. In Catolys(Handbook; Wolfe Publishing Ltd.: London, 1989; pp Waletzko, N.; Schmidt, L. D. Modeling Catalytic Gauze Reactors: HCN Synthesis. AIChE J. 1987, 34 (7), Williams, W. R.; Marks, C. M.; Schmidt, L. D. Steps in the Reaction Hz + Oz HzO on Pt: OH Desorption at High Temperatures. J. Phys. Chem. 1992,96,5922. Zum Mallen, M. P.; Williams, W. R.; Schmidt, L. D. Steps in Hz Oxidation on Rh OH Desorption at High Temperatures. J. Phys. Chem. 1993,97,625. Received for review September 4, 1992 Revised manuscript received January 14, 1993 Accepted February 2,1993