Control Soot Emitted From Acetylene Diffusion By Electric Field

Control of Soot Emitted from Acetylene Diffusion Flames by Applying an Electric Field MASAHIRO SAITO, TOSHIHIRO ARAI, and MASATAKA ARAI*

Department of Mechanical System Engineering, Gunma University, Tenjin-cho, Kiryu 376-8515, Japan This paper deals with the control of soot emission from acetylene diffusion flames by applying an electric field. The effects of applied voltage, polarity, and spacing of electrodes on soot emissions were investigated experimentally. The results showed that the shape of the flame changed remarkably with increasing applied voltage. The polarity of the applied voltage influenced the shape of the flame and the soot emissions. When a positive voltage was applied to the nozzle electrode, the flame length became shorter and the width at the flame tip was spread at high voltages. More than 90% of the soot emission was suppressed at over 200 kV/m of electric field intensity. Also, the flame temperature increased with increasing applied voltage. In particular, in the case of voltages above 200 kV/m, the temperatures at the flame tip were about 500°C higher than in the absence of an electric field. The rise of flame temperature was caused by the air entrainment promoted by an ionic wind. It was concluded that the soot reduction by applying an electric field was due to the oxidation of soot particles. In contrast, when negative voltages were applied to the nozzle electrode, the efficiency of soot control was limited to about 70% because the flame temperature, even at high applied voltages, was comparable to that in the absence of the electric field. © 1999 by The Combustion Institute

NOMENCLATURE F h I Lf Mf MS MS0 Tf Wf Z ␭

Electric field intensity, kV/m Spacing of electrodes, mm Electric current, ␮ A Flame length, mm Mass flow rate of fuel, mg/s Soot emission at applied electric field, mg Soot emission at nonapplied electric field, mg Flame surface temperature, °C Flame width, mm Distance from nozzle, mm Wavelength, nm

INTRODUCTION Generally, it is well known that soot emission occurs when fuel is burnt in insufficient oxygen. In order to control soot emission in the case of a premixed flame of gaseous fuel, it is necessary to supply enough oxygen for combustion. Also, promotion of mixing between fuel and oxygen is necessary in the case of a diffusion flame. Many studies on soot formation, nucleation and control of soot emission have been conducted. Faeth et al. [1, 2] studied soot formation, nucle*Corresponding author. E-mail: [email protected] 0010-2180/99/$–see front matter PII S0010-2180(99)00065-6

ation, and growth in laminar acetylene/air diffusion flames. They reported that soot formation began when temperatures reached roughly 980°C and fuel decomposition yielded acetylene. Also, soot nucleation rates were correlated with a first-order acetylene reaction. Significant levels of soot nucleation and growth require temperatures higher than roughly 980°C and fuel-equivalence ratios larger than 1.14. With regard to the control of soot emission by additives, Du et al. [3] reported the effects of various gaseous additives on the soot particle inception limit. The addition of H2 increases flame temperature and is effective at suppressing soot inception in flames. Chung et al. [4, 5] conducted experimental studies on the suppression of soot by metal additives during the combustion of polystyrene. They reported that the combination of two metals (i.e., K and Ca, Sr, or Ba) is much more effective than each metal singly at the same addition rate and that the maximum percentage of soot suppression reaches approximately 90%. Concerning soot reduction by applying electric fields, it has been found by Weinberg et al. [6 – 8] that the quantity of soot emitted from a flame decreased considerably when an electric field was applied to a flame. They investigated the effects of applied electric fields on the process of soot formation in diffusion flames: COMBUSTION AND FLAME 119:356 –366 (1999) © 1999 by The Combustion Institute Published by Elsevier Science Inc.

CONTROL OF SOOT BY APPLYING ELECTRIC FIELD nucleation, growth in the pyrolysis zone, and deposition. They reported that the rate of generation of particles depends on both the polarity and magnitude of the ion flux through the pyrolysis zone. The positive ions act as nuclei for the formation of soot particles. Negative charges do not appear to act as nuclei but tend to reduce soot formation by neutralizing positive-ion nuclei. In order to study the process occurring during the very early (nucleation) stage of formation of carbon particles in flames, they measured particle mobility and carbon particle size when a potential is applied across flat, counter-flow diffusion flame. They reported that particles have mobilities ranging from 10⫺3 to 3 ⫻ 10⫺2 cm2 s⫺1 V⫺1, depending on the applied potential. From detailed size analyses by electron micrographs, carbon particles with 50 nm diameter with no applied field reduced to about 10 nm in diameter with a small electric field of several kV. Also, Lawton and Weinberg [9] reviewed the electrical aspects of combustion. They have demonstrated theoretically and experimentally the effects of electric field on flames, ions, electrons, charged particles, and on electrically induced ionic velocity. They investigated flame distortion in uniform or nonuniform electric fields and observed soot particle size using electron micrographs. Bradley et al. [10 –12] reviewed the effect of electric fields on combustion processes, and studied the blow-off characteristics of premixed methane–air flames when coronas are generated around the rim of a cylindrical burner. The application of DC electrical fields showed that a negative corona at the flame tip produced an increased flow rate at blow-off. The blow-off flow rates were increased by the electric field even before the onset of corona. Ohisa et al. [13] reported the effect of DC or AC corona discharges on soot emission from a propane turbulent diffusion flame. A considerable reduction in soot emission was observed in the case of both DC and AC corona discharges when the coronas were applied across the lower part of the flame, where soot inception occurred. When a DC corona discharge was applied to the flame, the entrainment due to corona wind played a role on the soot reduction. Mizutani and Nakahara [14] investigated the effect of alternating

357

Fig. 1. Schematic diagram of experimental apparatus.

electric fields on diffusion flames of propane/air in coaxial flow. The size, structure and concentration of the soot particles were measured in both cases with and without an electric field. When an alternating electric field was applied, oxidation of soot particles was enhanced because the electric field controlled the formation and the growth of soot particles. However, few quantitative investigations regarding the effects of applied voltage, polarity, and spacing of electrodes on soot emission have been reported. Also, the mechanism of soot reduction by applying electric fields is not fully understood at present. In this study, the influences of electric field variables such as applied voltage, polarity, spacing of electrodes on the flame shape, and the soot emission for acetylene diffusion flames were examined. Furthermore, the flame temperature was measured in order to elucidate the mechanism of soot reduction by applying an electric field. EXPERIMENTAL APPARATUS AND METHOD A schematic diagram of the experimental apparatus is shown in Fig. 1. Acetylene was used as a

358

M. SAITO ET AL. particles was 1 minute in each run. The flame behavior in an applied electric field was photographed with a 35-mm camera and a video camera, and the changes in flame shape were measured on a monitor screen. Flame surface temperatures were measured by means of a two-color pyrometer (Siguma Electron Co., C-120). The two selected wavelengths of the pyrometer were ␭1 ⫽ 610 nm and ␭2 ⫽ 750 nm. In order to investigate the behavior around the flame when applied voltage increased, flow pattern was visualized by means of a smoke injection method. In addition, gas velocity at postflame was measured using sparkler as a tracer.

Fig. 2. Types of applying method of electric field between electrodes.

fuel gas in the experiment. The flow rate of acetylene, M f , was varied between 0.98 and 3.09 mg/s. After regulating the flow rate by a needle valve, acetylene was discharged from the nozzle (stainless steel pipe, o.d. 2.0 mm, i.d. 1.4 mm, and 100 mm length) which served as an electrode. The details of the nozzle electrode are shown in Fig. 2. A laminar diffusion flame was formed on the nozzle electrode. The plate electrode which imitates Rogowski’s electrode is inlaid with a nozzle at the center position of the electrode. The tip of the nozzle was cut and ground at a right angle. The tip face of the nozzle is flush with the plate electrode face. On the other side, a brass ring (o.d. 76 mm, i.d. 52 mm, thickness 12 mm) was used as a ground electrode. DC high voltages up to E ⫽ ⫾15 kV were applied between the electrodes using a DC high-voltage supply device (Spellman Co., SL-150). The spacing of electrodes, h, was varied stepwise at 25, 50, and 100 mm. The electric current was measured by using an ammeter attached to the ground side. The soot particles generated in a flame were sampled using a soot collector with a glass fiber filter. The weight of soot particles collected on the glass fiber filter was weighed using an electronic balance (Shimazu Co., AEG-45SM) with an accuracy of 0.01 mg. The soot emission was determined by the weight difference of the glass fiber filter before and after the sampling. The sampling time of soot

RESULTS AND DISCUSSION Method of Applying Electric Fields to Electrodes There are four combinations of applying electric fields between the nozzle electrode and the ground electrode: (A) nozzle electrode (positive)/ring (ground) (B) nozzle electrode (negative)/ring (ground) (C) nozzle electrode (ground)/ring (negative) (D) nozzle electrode (ground)/ring (positive)

electrode electrode electrode electrode

First, we tried to examine the fundamental characteristics of how these four methods of applying an electric field influence flame shape and soot emission. From the results, it was confirmed that the behavior of (A) coincided with that of (C), and (B) with (D), respectively. It meant that the fuel gas was not charged and that the flame was affected only by the electric field applied to it. Accordingly, we decided to proceed with two methods: (A) nozzle electrode (positive)/ring electrode (ground), and (B) nozzle electrode (negative)/ring electrode (ground). Variation of Flame Shape by Applying Electric Fields The variations of flame shape with increasing applied voltage for the cases of (a) positive

CONTROL OF SOOT BY APPLYING ELECTRIC FIELD

359

Fig. 3. Variation of flame shape with increasing applied voltage (M f ⫽ 3.09 mg/s, h ⫽ 50 mm). (a) Positive potential. (b) Negative potential.

potential and (b) negative potential are shown in Fig. 3. The mass flow rate of acetylene, M f , was 3.09 mg/s and the spacing of electrodes, h ⫽ 50 mm. At first, in the absence of an applied voltage (E ⫽ 0), a laminar diffusion flame formed on the nozzle electrode. When a positive voltage of 3 kV was applied to the nozzle electrode, the tip of flame started to spread slightly. The width of the flame tip increased with increasing applied voltage and sharply pointed beaks appeared in the flame tip at high voltages above E ⫽ 9 kV. Also, the flame length gradually decreased with increasing applied voltage and was accompanied by deformation of the flame tip. The shape of the flame tip was drastically changed,

but the base of the flame was unchanged even when high voltages were applied. On the other hand, when a negative voltage was applied to the nozzle electrode, the flame shape was similar to that in the absence of an applied voltage up to E ⫽ ⫺6 kV. The flame tip did not widen even if the negative applied voltage increased over E ⫽ ⫺6 kV. At E ⬎ ⫺12 kV, the base of the flame opened over the nozzle electrode because the flame was crushed, and the soot particles deposited on the nozzle electrode were burnt. Furthermore, as a characteristic phenomenon of negative potentials, a circular blue flame of unique flame shape was observed under the condition of E ⫽ ⫺15 kV, Mf ⫽ 0.98 mg/s, h ⫽ 50 mm. A photograph of

360

Fig. 4. Photograph of a circular flame forming at negative high voltage (E ⫽ ⫺15 kV, M f ⫽ 0.98 mg/s, h ⫽ 50 mm).

the typical circular blue flame is shown in Fig. 4. Effect of Electric Field on Flame Shape As shown in these photographs, the flame shape varied with the magnitude of applied voltage and its polarity. The variations of the flame length, L f , and width at the flame tip, W f , with applied voltage were measured. The result in the case of positive potential is shown in Fig. 5a. The parameters L f and W f were defined as illustrated in the figure. The flame length L f begins to shorten when the applied voltage exceeds several kV. The threshold voltage, that is the voltage at which the flame length begins to shorten, decreased with decreasing spacing of electrodes. On the other hand, the flame width at the tip, W f , spread gradually with increasing applied voltage. In the case of negative potentials applied to the nozzle electrode on the contrary, the flame length, L f , decreased gradually with increasing applied voltage (Fig. 5b) because the flame was

™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™3

Fig. 5. Variations of the flame length and width with increasing applied voltage. (a) Positive potential. (b) Negative potential.

M. SAITO ET AL.

CONTROL OF SOOT BY APPLYING ELECTRIC FIELD

361

applied voltage to that in the absence of an applied voltage. The value M S /M S0 was used to evaluate the efficiency of soot reduction by the electric field. As is evident from the figure, the soot emission decreased steeply at several kV of applied voltage. At E ⫽ 15 kV, the value of M S /M S0 reached 0.1, i.e., 90% efficiency of soot reduction. Also, in the same voltage range, the efficiency of soot reduction decreased with increasing spacing of electrodes. The effects of negative potential on soot emission are shown in Fig. 6b. The decrease of soot emission with increasing applied voltage was gradually compared with the positive potential. The efficiency of soot reduction was limited at about 70%. Payne and Weinberg [15] measured the mass of carbon deposited on the grounded plate electrode when a positive or negative of 2 kV was applied to the nozzle electrode. They reported that the flame was constricted and carbon deposited on the grounded plate electrode was bulky in the case of negative potential. The mass deposited with time was small compared with that of no electric field. On the contrary, in the case of positive potential, the amount of carbon deposited upon the plate was small and did not continue to increase with time. It was found that the effect of polarity on the soot emission characteristics was essentially similar to the result of Payne et al. Relation between Soot Emission and Strength of Electric Field Fig. 6. Effect of applied voltage on soot emission. (a) Positive potential. (b) Negative potential.

crushed and the value kept constant. The voltage at beginning of the constant flame length decreased with decreasing electrode spacing. In contrast, the flame width at the tip, W f , did not extend over the whole range of applied voltages up to E ⫽ ⫺15 kV. Effect of Applied Voltage on Soot Emission The effects of the polarity of the electric field on the soot emission from flames are shown in Figs. 6a and 6b. Figure 6a shows the effect of positive potential on soot emission. The soot emission was defined as the ratio of mass of soot at the

Figure 7 replots these results, replacing the applied voltage of Fig. 6 by the mean intensity of the electric field, F (⫽ E/h [kV/m]). The relation between the efficiency of soot reduction and the mean intensity of the electric field was classified into two groups according to the polarity of applied voltage. In the case of positive potential, the efficiency of soot reduction exceeded 90% at electric field strength above F ⫽ 200 kV/m. In the case of negative potential, the limit of the soot reduction was 70% efficiency. Relation between Applied Voltage and Electric Current When there is no flame between the electrodes, the electric current was almost zero over a wide

362

M. SAITO ET AL.

Fig. 7. Relation between soot emission and intensity of electric field.

range of the applied voltages up to ⫾15 kV. Figure 8 shows the relation between applied voltage and electric current measured with a flame between the electrodes. The presence of a flame leads to an electric current between the electrodes even if the flame tip does not reach the ground electrode, because of an electric conductivity of the hot burned gas and the flame which contains carbon particles and ion species. In the case of positive potential (a), the electric current increased linearly with increasing applied voltage. In the case of negative potential (b), the electric current increased in proportion to the applied voltage in the range E ⬍ ⫺7 kV, but no increase in electric current was observed in the range at E ⬎ ⫺7 kV and ⫺9 kV for h ⫽ 25 and 50 mm. This is due to the increase of resistance between electrodes because the flame length became extremely short in such high electric fields, leading to saturation. In the range of E ⬍ ⫾7 kV, the electric current in the negative voltage case was larger than in the positive case. This seemed to be due to the difference of mobility of positive and negative ions. This will be mentioned later in the detailed discussion.

Fig. 8. Relation between applied voltage and electric current. (a) Positive potential. (b) Negative potential.

Variation of Flame Surface Temperature by Applying Electric Field In an attempt to clarify the mechanism of soot reduction by applying electric fields, the temperature of the luminous flame, T f , was measured by means of a two-color pyrometer. Figure 9a shows flame temperature when positive voltages were applied to the nozzle electrode. In the case of E ⫽ 0, the value of T f at the base of flame (near the nozzle) was about 1800°C, but T f

CONTROL OF SOOT BY APPLYING ELECTRIC FIELD

Fig. 9. Variation of flame surface temperature with increasing applied voltage. (a) Positive potential. (b) Negative potential.

decreased gradually with increasing distance from the nozzle and T f at the flame tip of a sooting flame was about 1100°C. Gomez et al. [16] reported that soot nucleation occurs on the centerline of a laminar diffusion flame and the characteristic temperature of soot onset was 1077°C. In the present experiment, it was confirmed that the temperature of the sooting flame at the tip of the flame in the absence of an applied electric field (E ⫽ 0) was nearly equal to the characteristic temperature of soot onset. The

363

flame surface temperature rose as the applied voltage increased. In particular, at high voltages above E ⫽ 10 kV, T f at the flame tip was about 1800°C, and the whole flame maintained the high temperature above 1800°C. It was concluded that the rise of flame temperature by applying a positive potential caused the oxidation of the soot particles formed in the flame. As a result, more than 90% of soot emission was suppressed at 200 kV/m of the intensity of electric field. On the other hand, when a negative potential was applied to the nozzle electrode, the temperature at the tip of flame was only slightly higher than that for nonapplied electric field as shown in Fig. 9b. Thus the T f values at the flame tip at E ⫽ ⫺5 and ⫺10 kV were 1350 and 1600°C, respectively. These temperatures were 200 ⬃ 450°C lower than that for the positive potentials. Nevertheless, the mechanism of soot control for negative potentials is thought to be as follows. In the region of E ⬍ ⫺5 kV, the flame length was similar to the case of no electric field but the light intensity of the flame was lower [17]. It is conjectured from the decrease of light intensity of the flame that the soot nucleation in the pyrolysis zone is suppressed. Also, at E ⬎ ⫺5 kV, the flame is crushed because the ionic wind blows from the ground electrode toward the nozzle electrode. The combination of the mechanism of the negative electric field and the ionic wind forced soot particles to move toward the nozzle electrode. It is considered that soot particles moving toward the nozzle electrode or soot particles deposited on the nozzle electrode are burnt within the flame. As a result, a 70% soot reduction was achieved for the negative potential. Next, to explain ionic wind for both cases of positive or negative potential, the illustrations are shown in Fig. 10. Payne and Weinberg [15] reported theoretically the movement of flame ions by an applied electric field and the forces acting on flame gases due to ion movement. They concluded that positive and negative ions are formed in the reaction zones of flame. The former are probably of carbon and the latter electrons but, whatever their nature initially, the mobility of the negative ion is greater than that of the positive ion. In the case of positive potentials (Fig. 10a), as

364

M. SAITO ET AL. wind along the flame which enhanced the mixing of the fuel gas and surrounding air. On the other hand, in the case of negative potentials (Fig. 10b), negative ions and electrons are forced to the ground electrode. The carbon particles and positive ions are conversely moved toward the nozzle electrode. The directions of soot particles and positive ions become opposed to the fuel gas flow. As a result, it is thought that the soot particles that were moved close to the nozzle electrode could be burned in the flame because the base of flame had enough oxygen to become a premixed flame as shown in Fig. 4. However, the electric current which is mainly carried by electrons was larger than that of positive potential case, because the electrons moved in the absence of the burned gas. Fig. 10. Illustration of ionic wind.

a high voltage was applied to the nozzle electrode, the electric field lines concentrate the flame where ions are produced. Consequently, the positive ions and soot particles are moved to the ground electrode. As a result, an ionic wind flows from the nozzle electrode toward the opposite ground electrode. Since the flow directions of positive ions and soot particles are the same as that of fuel gas flow, these ions and carbon particles act to accelerate the gas flow toward the opposite ground electrode because positive ions and soot particles are larger than the electrons. It is considered that the flame surface temperature was raised by the ionic

Visualization around a Flame in an Electric Field Figure 11 shows the variation of flow patterns around the flame in an electric field. As is evident from the photographs, in the absence of an applied voltage (E ⫽ 0), the smoke moved by natural convection. When 3 to 6 kV was applied between electrodes, the flow of the smoke was rectified like a laminar flow. As high voltages larger than 9 kV were applied, turbulence occurred in the postflame region. The turbulence grew with increasing applied voltage. In the case when high voltage was applied to a flame, the flame itself acts as a needle electrode because the flame has electroconductivity. In such cases, the electric field of the nozzle elec-

Fig. 11. Variation in flow pattern around the flame with applied voltage (Acetylene, M f ⫽ 0.98 mg/s, h ⫽ 50 mm).

CONTROL OF SOOT BY APPLYING ELECTRIC FIELD

365

CONCLUSIONS Control of the soot in acetylene diffusion flames by applying electric fields was studied experimentally. From the experimental results the following conclusions were drawn:

Fig. 12. Relation between gas velocity at postflame and electric field intensity.

trode was concentrated on the flame tip, and a false nonuniform electric field (needle/ring electrode) will be formed. In the case of nonuniform electric fields, it is known that an ionic wind blows from needle electrode toward opposite plate electrode. Thus, it is considered that the change in the flow pattern is caused by an ionic wind due to the electric field. Relation between Gas Velocity at Postflame and Electric Field Intensity In order to investigate the gas velocity caused by ionic wind, photographs of sparks from a sparkler were taken using 35 mm camera with shutter speed of 1/1000 seconds. The gas velocity was determined by the shutter speed and the length of sparks trajectories. Figure 12 shows the relation between gas velocity measured at 60 –75 mm from the nozzle and electric field intensity. From the figure, the gas velocity induced by ionic wind increased with increasing applied voltage. Lawton and Weinberg [18] reported that the maximum ionic wind velocity was found to be 5.5 m/s. It was confirmed that the ionic wind velocity was of the order of 1 to 4 m/s, and the ionic wind caused the temperature rise of the flame.

1. When a positive voltage was applied to the nozzle electrode, the tip of flame spread and the flame length shortened with increasing applied voltage. On the other hand, the flame was crushed when a negative high voltage above about E ⫽ 10 kV was applied to the nozzle electrode. 2. The soot emission decreased with increasing applied voltage, and the efficiency of soot suppression exceeded 90% in the region of intensity of electric field over F ⫽ 150 kV/m. In contrast, the efficiency was limited around 70% in case of negative potential. 3. In the absence of an applied voltage (E ⫽ 0), the flame temperature was about 1800°C at the base of flame and 1100°C at the tip of flame. On applying a positive voltage to the flame, the flame temperature increased with increasing applied voltage and the whole flame maintained a high temperature above 1800°C at applied voltages above E ⫽ 10 kV. In contrast, the flame temperatures with a negative potential were 200 ⬃ 450°C lower compared with that for positive potential. 4. The application of an electric field created an ionic wind towards the ground electrode. The ionic wind enhanced the mixing of the fuel gas and surrounding gas, resulting in the high temperature of the flame. It was thought that the consequent increase in the flame temperature for positive potentials caused oxidation of soot particles produced in the flame. 5. It was verified by visualizing the flow patterns that a flow due to ionic wind formed around the flame. The gas velocity in the postflame increased in proportion to the electric field intensity and was on the order of 1 to 4 m/s. The authors express their thanks to Mr. Norihiko Ueno for his assistance in conducting the experiments.

366

M. SAITO ET AL.

REFERENCES 1. 2. 3. 4. 5. 6.

7. 8. 9. 10.

Sunderland, P. B., and Faeth, G. M., Combust. Flame 105:132–146 (1996). Lin, K. C., Sunderland, P. B., and Faeth, G. M., Combust. Flame 104:369 –375 (1996). Du, D. X., Axelbaum, R. L., and Law, C. K., Combust. Flame 102:11–20 (1995). Chung, S.-L., and Tsang, S.-M., J. Air Waste Manage. Assoc. 41:821– 826 (1991). Chung, S.-L., and Lai, N.-L., J. Air Waste Manage. Assoc. 42:1082–1088 (1992). Place, E. R., and Weinberg, F. J., Eleventh Symposium (International) on Combustion, The Combustion Institute, Pittsburgh, 1967, pp. 245–255. Lawton, J., Mayo, P. J., and Weinberg, F. J., Proc. Roy. Soc. Lond. A 303:275 (1968). Mayo, P. J., and Weinberg, F. J., Proc. Roy. Soc. Lond. A 319:351–371 (1970). Lawton, J., and Weinberg, F. J., Electrical Aspects of Combustion, Clarendon Press, Oxford, 1969. Bradley, D., in Advanced Combustion Methods (Weinberg, F. J., Ed.), Academic Press, London, 1986.

11. 12. 13. 14. 15. 16. 17. 18.

Bradley, D., and Gupta, M. L., Combust. Flame 40: 47– 63 (1981). Bradley, D., and Nasser, S. H., Combust. Flame 55: 53–58 (1984). Ohisa, H., Horisawa, H., and Kimura, I., Trans. Japan Soc. Mech. Engin. 63:255–262 B (1997). Mizutani, Y., and Nakahara, M., Trans. Japan Soc. Mech. Engin. 49:1530 –1538 B (1983). Payne, K. G., and Weinberg, F. J., Proc. Roy. Soc. Lond. A 205:316 –336 (1959). Gomez, A., Littman, M. G., and Glassman, I., Combust. Flame 70:225–241 (1987). Arai, T., Saito, M., and Arai, M., Trans. Japan Soc. Mech. Engin. 64:3881–3887 B (1998). Lawton, J., and Weinberg, F. J., Proc. Roy. Soc. Lond. A 277:468 – 497 (1964).

Received 25 August 1998; revised 12 April 1999; accepted 24 April 1999