Silica Decomposition Using a Transferred Arc Process - Industrial

A novel process for silica decomposition consisting of transferring a thermal arc directly to a silica anode was investigated. The effects of current ...
0 downloads 0 Views 131KB Size
Ind. Eng. Chem. Res. 1999, 38, 2299-2309

2299

Silica Decomposition Using a Transferred Arc Process Tony Addona and Richard J. Munz* Department of Chemical Engineering, McGill University, 3610 University Street, Montreal, Quebec H3A 2B2, Canada

A novel process for silica decomposition consisting of transferring a thermal arc directly to a silica anode was investigated. The effects of current (150-250 A), plasma gas flow rate (10-20 L/min of Ar), and plasma gas composition (0-2.8% H2) on the rate of decomposition were examined. The decomposition rate ranged from 0.09 to 1.8 g/min and was determined to be heat-transfer-limited, with decomposition occurring below the arc root where the anode surface attained its boiling point. The decomposition rate was independent of plasma gas flow rate, suggesting that convective heat transfer was reduced significantly by the counterflow of decomposition products (SiO(g) and O2) from the surface. Increasing the current increased the decomposition rate because heat input to the anode because of electron flow and arc radiation increased. Adding H2 to the plasma gas increased the decomposition rate because of an increase in radiative heat transfer to the anode, a reduction in the theoretical energy requirement for decomposition, and a consumption of O2 which lowered the silica boiling point. The fumed silica produced had properties typical of commercial fumed silica products. 1. Introduction Silica decomposition is the first and most energyintensive step in the thermal plasma production of fumed silica from quartz. The aim of decomposition is to generate silicon monoxide (SiO(g)), which can then be quenched rapidly using an oxidizing agent to produce fumed silica. The decomposition can be entirely thermal, producing SiO(g) and O2, or a reducing agent such as C, Si, or H2 can be used to combine with the O2 and lower the energy requirements. The energy required is supplied by a thermal plasma heat source. Because of their high energy efficiency and potential for scale-up, dc arc torches have received the most attention as thermal plasma generators for this process. The energy required to process the silica feed is a function of the heat-transfer efficiency and the type of reducing agent used. Silica decomposition using various reducing agents has been studied by various authors.1-3 The type of reducing agent chosen depends on cost, purity, availability, and ease of use. The efficiency with which energy is transferred to the silica feed depends strongly on the processing method and plasma reactor design. Previous systems have relied on convective and radiative heat transfer. For example, the use of dc nontransferred arc torches to heat the silica resulted in primarily convective heat transfer.4,5 Radiative heat transfer was more significant in the studies using transferred arc torches.6-8 However, because silica was not directly used as the anode, these systems failed to utilize important heattransfer mechanisms associated with arc impingement as well as resistive heating. Both of these are important features of the transferred arc process and the main reasons it is preferred when treating condensed phases.9 In other works where transferred arc configurations have been used, the anode has consisted of a mixture of silica and an electrically conductive reducing agent such as Si or C.10-13 Although taking advantage of true transferred arc operation, these processes suffered from * To whom correspondence should be addressed. E-mail: [email protected]. Fax: 514-398-6678.

the use of relatively expensive and possibly contaminating reducing agents. A novel technique for silica decomposition was examined in this work to overcome the above deficiencies. This consisted of using a transferred arc system in which silica alone served as the anode material. Silica decomposition using this technique is expected to be more energy efficient than previous processes. The maximum energy efficiency for a nontransferred arc torch is typically 80% while true transferred arc operation can result in efficiencies up to 95%.9 The need to mix Si or C with the silica raw material could also be avoided, and cheaper gaseous reducing agents, such as H2, could be used instead. These gaseous reducing agents could be added to the reactor separately or as part of the plasma gas. The current study examined the effect of current and plasma gas flow rate on the decomposition rate of silica in a steady-state transferred arc process. Because of the limited number of experiments which could be performed, the effect of arc length was not examined here because it was not considered as important a process scale-up variable as the plasma gas flow rate and current. The small size of the reactor and the continuous feeding of quartz particles allowed steady state to be achieved rapidly. Because the effect of reducing agents on silica decomposition is well-known, an Ar plasma gas was used to further simplify the experiments. No reducing agent was added to the reactor to limit the decomposition rates to reasonable values. Because of this, intense feeding of quartz particles was not required and steady-state operation was more easily maintained. The effect of adding H2 to the plasma gas was investigated but only at low concentrations. The results of this study were used in a subsequent work which investigated the effect of oxidation zone parameters on the properties of the fumed silica generated. The details of this work will be given in a future publication; thus, only a brief summary of the powder properties obtained are presented here.

10.1021/ie980652k CCC: $18.00 © 1999 American Chemical Society Published on Web 05/13/1999

2300 Ind. Eng. Chem. Res., Vol. 38, No. 6, 1999

Figure 1. Heat of silica decomposition for inert and reducing conditions (computed using F*A*C*T).

Figure 2. Species generated at equilibrium from the reaction of 1 gmol of SiO2 and 1 gmol of C (computed using F*A*C*T).

Table 1. Silica Decomposition Equilibrium under Inert Conditions Predicted by F*A*C*T14

Table 1 shows the results of a similar analysis conducted using the thermodynamic computer package known as F*A*C*T.14 Again, SiO(g) and O2 are predicted to be the major products of silica decomposition. Some discrepancies were found with the predicted normal boiling point and the species present at equilibrium. While ref 2 predicted the presence of small quantities of SiO2(g) and Si2O2(g), these species are absent in the F*A*C*T simulation. The discrepancies are probably due to differences in thermodynamic property estimation. 2.2. Carbothermic Reduction of Silica. Because in this work the silica anode was to be contained in a graphite crucible, silica might be reduced by C. The possible reactions occurring at the silica/anode interface were determined by examining the species present at equilibrium from the reaction of SiO2 and C. Figure 2 shows that below 1800 K the reaction between SiO2 and C is negligible. Above this temperature, the following reactions must be considered:

gas-phase composition pressure (kPa)

boiling point (K)

species

yi

101.3

3151

107

3159

112

3167

SiO O2 O SiO O2 O SiO O2 O

0.634 77 0.269 55 0.095 68 0.634 74 0.269 49 0.095 77 0.634 71 0.269 43 0.095 86

2. Background The decomposition of silica is a highly endothermic process, but the theoretical energy requirement can be reduced significantly by using a reducing agent such as C, Si, or H2 as shown in Figure 1. Hydrogen provides the greatest decrease in the heat of decomposition, but the use of Si allows the production of 2 mol of SiO(g) for every 1 mol of SiO2 decomposed. Hydrogen is an ideal reducing agent because it is relatively inexpensive and noncontaminating and can be added directly to the plasma gas. 2.1. Silica Decomposition in an Inert Atmosphere. According to ref 2, the equilibrium reached by silica and its decomposition products under inert conditions can be described by the following set of reactions:

2SiO2 T 2SiO(g) + O2

(1)

O2 T 2O

(2)

SiO2 T SiO2(g)

(3)

2SiO(g) T Si2O2(g)

(4)

The normal boiling point was estimated to be approximately 3070 K, with the major species being SiO(g) and O2. Dissociation of O2 and production of SiO2(g) were predicted to occur only to a minor extent, and the amount of Si2O2(g) was estimated to be negligible. The reaction of eq 1 was therefore determined to be the predominant reaction describing the equilibrium of silica decomposition under inert conditions.

SiO2 + C f SiO(g) + CO

(5)

SiO2 + 3C f SiC + 2CO

(6)

SiO2 + 2C f Si(l) + 2CO

(7)

2.3. Use of Silica as an Anode Material. The proposed decomposition technique required the operation of a transferred arc to a silica anode. For this to be possible, silica must be electrically conductive. At low temperatures silica is an effective insulator; however, its electrical resistivity is reduced with increasing temperature. Conduction is the result of electronic and ionic contributions, both of which increase with increasing temperature. The dominant contribution depends on the material, its purity, and temperature.15 Previous studies16,17 have shown that the electrical resistivity of silica decreases dramatically when heated to high temperatures and with impurity levels on the order of ppm. Donor or acceptor type semiconduction has been associated with impurities in oxides.18 The use of silica as an anode material at high temperatures should thus be feasible. 3. Experimental Section 3.1. Apparatus. The transferred arc reactor used during this study is shown schematically in Figure 3.

Ind. Eng. Chem. Res., Vol. 38, No. 6, 1999 2301 Table 2. Summary of Experimental Conditions for Decomposition Experiments Using an Ar Plasma Gas

Figure 3. Transferred arc reactor used for the silica decomposition study.

The reactor consisted of a water-cooled, cylindrical, stainless steel vessel with a side exit and view port, allowing a clear view of the arc and its impingement on the silica bath surface. The reactor was heavily insulated with graphite and alumina felt insulation. Within the reactor was a graphite vessel (i.d. ) 12.7 cm and height ) 15.2 cm), which served as the plasma chamber. The crucible containing the molten silica was connected to a water-cooled anode holder using a hollow graphite stem. The diameter of the conical crucible cavity decreased from 10.2 cm at the top edge to 2.5 cm at the bottom. The depth of the cavity was 2.5 cm. The plasma torch, consisted of a conical (60°), thoriatedtungsten tip cathode inserted into a copper nozzle with a contained angle of 65°, an exit orifice diameter of 6.4 mm, and a gap width of 2.8 mm. The nozzle, which also served as the auxiliary anode during arc ignition, had a length of 5.0 mm. Eight Miller SRH-444 rectifiers (Miller Electric Mfg. Co., Appleton, WI), each having an open circuit voltage of 75 V and a rating of 19.6 kW, were connected in series to supply power to the torch. A high-frequency generator (HF 2000 high-frequency arc starter, Miller Electric Mfg. Co.) was used for arc ignition. Voltage was measured using a divider circuit connected across the cathode and anode at the output of the high-frequency generator. Current was measured using a shunt in series with the anode. Quartz supplied by Baskatong Quartz Inc. (Montreal, Quebec, Canada) served as the silica raw material. The quartz particles were irregular in shape with a nominal size of 1-3 mm and a purity of 99.94% SiO2. The particles were fed into the crucible via an alumina tube connected to a vibrating bowl powder feeder (model 104B, TAFA Metallisation Inc., NH). The gases used were high-purity Ar (Matheson Gas Products, Calgary, Alberta, Canada) and H2 (Air Products Canada Ltd., Brampton, Ontario, Canada) from regulated high-pressure cylinders. The reactor pressure was measured with a bourdon-tube gauge, and temperatures were determined using type C (tungsten/rhenium) thermocouples. The thermocouple assemblies consisted of isolated bare thermocouple wires in alumina insulators. The crucible

experiment

plasma gas flow rate (L/min, 293 K, and 101.3 kPa)

current (A)

duration (min)

54 60 119 50 59 120 52 61 87 89 90 81 88 85 86 76 78 82 84 77 83 91

10 10 10 15 15 15 20 20 10 10 10 15 15 20 20 10 10 15 15 20 20 20

150 150 150 150 150 150 150 150 200 200 200 200 200 200 200 250 250 250 250 250 250 250

180 180 180 180 180 190 180 190 130 120 120 120 120 120 120 65 60 60 60 60 60 60

temperature was measured 6.4 mm below the bottom center of the silica anode. A water-cooled, stainless steel quench pipe connected the transferred arc reactor to a baghouse filter used to collect either the condensed decomposition product (i.e., SiO(s)) or fumed silica powder. The exit gas was monitored for concentrations of H2O, O2, and CO using a thermohygrometer (Cole-Parmer, Chicago, IL), an O2 meter (model GC 502, G.C. Industries, Fremont, CA), and a CO meter (model MEXA-201GE, Horiba Instruments Inc., Irvine, CA). Reactor exhaust gases were evacuated using a diaphragm vacuum pump (Gast, Benton Harbor, MI). A control valve at the pump entrance was used to regulate the reactor pressure. 3.2. Procedure. Prior to each experiment, a new crucible was weighed and its mass recorded. The charged crucible was then placed into the plasma chamber, and a known mass of quartz was placed into the particle feeder. Once the reactor was sealed and the torch lowered to within 2 cm of the top of the crucible, the system was purged with 5 L/min Ar of for 1 h. The arc was then ignited using the high-frequency generator to create an arc first between the cathode and nozzle and then from the cathode to the silica anode using a proprietary method. Following arc transfer, the plasma gas flow rate and current were adjusted to their desired levels and particle feeding was started. The time delay between arc ignition and the start of data acquisition was approximately 1 min. The decomposition rate was determined by performing a silica mass balance after each experiment. This consisted of weighing the following items: the silica left in the crucible, the quartz left in the particle feeder, and the quartz fed during the experiment which landed on the chamber floor. The crucible was also weighed to determine its mass loss during the experiment. 4. Results and Discussion 4.1. Silica Decomposition Using an Argon Plasma Gas. 4.1.1. Summary of Experiments. The three-level factorial design used to investigate the effect of plasma gas flow rate and current is summarized in Table 2.

2302 Ind. Eng. Chem. Res., Vol. 38, No. 6, 1999

Figure 4. Typical temporal crucible temperature profiles for silica decomposition experiments using an Ar plasma gas.

Argon was introduced into the reactor through the torch as the plasma gas, through the feed tube (4.1 L/min, 293 K, and 101.3 kPa) to prevent blocking due to partial melting of the quartz particles, and through the view port (3.6 L/min, 293 K, and 101.3 kPa) to prevent deposition of SiO(s) on the window. Current could be consistently regulated to within (5 A of the desired value. The arc length, defined as the distance from the cathode tip to the top of the crucible, was 2 cm for all experiments. The effect of arc length was not examined during this study because it was not considered to be as important a process scale-up variable as the plasma gas flow rate and current. Reactor pressure was maintained between 101.3 and 112 kPa to avoid air entering the system. The duration of experiments varied between 1 and 3 h depending on the current level being studied. 4.1.2. Crucible Temperature. The evolution of the crucible temperature with time for three typical experiments is shown in Figure 4. Because the crucible contents are far from isothermal, it is the variation with time and operating conditions that are important and not the absolute values of temperature shown. Steady state was assumed to have been achieved when the crucible temperature attained 90% of its value prior to shutdown. Using this criteria, steady state was achieved after 30, 20, and 15 min for current levels of 150, 200, and 250 A, respectively. These times did not vary with plasma gas flow rate and decreased with increasing current because of an increase in power, which resulted in a more rapid heating of the plasma chamber components. For experiments at 150 and 200 A, approximately 83% of the experimental duration was spent at steadystate compared to 75% for experiments at 250 A. The steady state crucible temperature is shown in Figure 5 as a function of Ar plasma gas flow rate and current. In each case, the mean and standard deviation are plotted for a given condition. Increasing current resulted in higher crucible temperatures while increasing plasma gas flow rate had essentially no effect. The trend with current was due to an increase in the heat input to the silica anode. The independence of crucible temperature to variations in plasma gas flow rate demonstrated that convective heat transfer was a minor component of the overall heat input to the anode surface.

Figure 5. Steady-state crucible temperature for decomposition experiments using an Ar plasma gas.

4.1.3. Arc and Anode Voltage. The voltage measured during an experiment included the voltage drops across the arc, silica anode, crucible, crucible stem, and other electrical connections. Calculations and measurements showed that all except those due to the arc, silica anode, and crucible stem were negligible. Because the resistivity of graphite is a weak function of temperature, the voltage drop in the stem could be computed as a function of current. The voltage drops through the stem were estimated to be 3.93, 5.24, and 6.55 V for currents of 150, 200, and 250 A, respectively. The measured voltages reported in this section have been corrected using these values. Typical voltage variations during experiments are given in Figure 6. The voltages reported at time zero correspond to those after 1 min of arc operation because data acquisition began only then. In all cases, the voltage was initially high (>60 V) and gradually decreased to its steady-state value through continued operation. This pattern is explained by a decrease in the electrical resistivity of the silica anode as it was heated. The voltage reached steady state at approximately the same time as the crucible temperature. The error in the overall voltage measurement, taking into account the error associated with the meter and the uncertainty in the estimate of the stem voltage drop, was estimated to be (0.5 V. Figure 7 shows that the steady-state voltage increased with increasing plasma gas flow rate. The data were fitted to a simple linear regression, giving

V ) 34.15 + 0.356Qplasma - 1.013 × 10-2I

(8)

where V is the voltage due to the arc and anode, Qplasma is the plasma gas flow rate (L/min, 293 K, and 101.3 kPa), and I is the current (A). The parameters in the equation are significant at the 95% confidence level. An attempt was made to separate the arc and anode voltages by measuring the voltage between the cathode and a graphite disk placed 2 cm below the cathode and using identical operating conditions of current and plasma gas; the results are shown in Figure 8. This was still only an estimate of the arc voltage during decomposition experiments because the chemistry of the arc root was different. Decomposition of silica at the anode

Ind. Eng. Chem. Res., Vol. 38, No. 6, 1999 2303

Figure 8. Estimate of arc voltage for silica decomposition experiments using an Ar plasma gas.

Figure 6. Typical voltage profiles for silica decomposition experiments using an Ar plasma gas.

Figure 9. Estimate of anode voltage for silica decomposition experiments using an Ar plasma gas.

Figure 7. Steady-state voltage for silica decomposition experiments using an Ar plasma gas.

surface results in the infiltration of Si and O species within the plasma gas; however, the effect on the overall arc voltage was expected to be small because of the low decomposition rates obtained during this study. Figure 8 shows that the arc voltage increases with both plasma gas flow rate and current, as has been observed by others.19,20 The arc voltages, shown in Figure 8, were used to compute the silica anode voltage drop and resistance

Figure 10. Estimate of anode resistance for silica decomposition experiments using an Ar plasma gas.

shown in Figures 9 and 10. The anode voltage drops were significant, ranging from 6 to 13.6 V. The decrease

2304 Ind. Eng. Chem. Res., Vol. 38, No. 6, 1999

in anode voltage drop with increasing current may be attributed to the decreasing resistance of the silica as its temperature increased. Figure 10 shows that the anode resistance remained approximately constant with plasma gas flow rate, which suggests that temperatures within the anode were not affected greatly by variations in this parameter. This agrees with Figure 5, which showed the crucible temperature increasing with increasing current and remaining approximately constant with increasing plasma gas flow rate. The computed values of anode voltage indicate that between 18 and 34% of the total power was dissipated directly in the anode as resistance heating. This fraction was almost independent of plasma gas flow rate and was highest for the lowest current. The absolute power dissipated in the anode by resistive heating remained constant at about 1.75 kW for all conditions because an increase in current resulted in a decrease in anode resistance. 4.1.4. Appearance of Silica Anode following an Experiment. The silica anodes were examined after each experiment. In all cases, the top surface was transparent and glasslike, suggesting its temperature had been well above 2000 K during transferred arc operation. For experiments at 250 A, the bottom of the anode was also glasslike but less transparent. This was due to the lower temperatures within the anode as it approached the crucible (1962-2048 K). The silica anode bottom at 150 and 200 A comprised fused particles and was completely opaque. The crucible temperature for these experiments ranged from 1700 to 1930 K. Transition to the vitreous state occurs at approximately 2000 K. The melting point of quartz has, however, been reported to be as low as 1883 K2. 4.1.5. Rate of Silica Decomposition. The mass of silica decomposed during an experiment was determined by subtracting the mass left in the crucible and on the chamber floor from the initial mass and that fed during the experiment. The feed rate required to maintain a constant anode mass was determined through trial and error. Feeding was inefficient at 150 A because particles had a tendency to bounce off unmelted quartz at the periphery of the crucible. The reaction of quartz particles with the chamber bottom was negligible because the chamber wall temperature, which approximated that of the chamber bottom, was below 1800 K for all cases (see Figure 2). In principle, the crucible could react with the silica anode to produce SiO(g), Si(l), or SiC, or it could react with the O2 produced by the thermal decomposition of silica. Measurements of crucible weight loss and examination of the crucible after experiments showed that reaction with the crucible was negligible in all cases, except at 20 L/min and 250 A. Crucible consumption was mainly due to the reaction of the exposed hot graphite surface with O2 produced at the anode surface. No consistent trend of weight loss was observed, although experiments at 15 and 20 L/min and 250 A produced higher crucible mass losses averaging 1.5 and 2 g, respectively, compared to approximately 1 g for all other cases. The examination of the crucible after experiments showed no attack at the silica/graphite interface in any of the experiments with currents below 250 A and only very minor attack with a plasma gas flow rate of 10 L/min and a current of 250 A. The crucible mass loss for low current operation was thus attributed entirely to the oxidation of the upper surface of the crucible. The crucible surface where reaction

Figure 11. Average silica decomposition rate for experiments using an Ar plasma gas.

between the silica and graphite was observed was gray in color and therefore probably the result of Si(l) formation. A green deposit associated with SiC was not present. Crucible consumption at 20 L/min and 250 A occurred on the upper portion of the crucible facing the reactor exit. Because this area was located furthest from the feeding point, it was slightly exposed because of the lower silica level. The exposed graphite surface was therefore more susceptible to heating by convective heat transfer and arc radiation which were greatest at the maximum plasma gas flow rate and current. This resulted in a slightly higher temperature in this region and an increase in the rate of silica decomposition at the anode/crucible interface. A portion of the Si(l) generated at the crucible surface probably dissolved within the silica anode and underwent further reaction, producing SiO(g). This interpretation was corroborated by the increasing presence of bubbling within the crucible for the high current runs and its absence at low current; the bubbling was probably mainly due to the production of CO from the carbothermic reduction of silica at the crucible interface. This was negligible at lower currents because of the lower crucible temperatures (