Home Search Collections Journals About Contact us My IOPscience DC glow discharge in air flow at atmospheric pressure in connection with waste gases treatment This content has been downloaded from IOPscience. Please scroll down to see the full text. 1993 J. Phys. D: Appl. Phys. 26 1630 (http://iopscience.iop.org/0022-3727/26/10/014) View the table of contents for this issue, or go to the journal homepage for more Download details: IP Address: 194.67.73.234 This content was downloaded on 02/08/2015 at 14:47 Please note that terms and conditions apply. J. Phys. D: Appi. Phys. 26 (1993) 1630-1637. Printed in the UK I I I I I DC glow discharge in air flow at atmospheric pressure in connection with waste gases treatment Yu S Akishev, A A Deryugin, I V Kochetov, A and N I Trushkin P Napartovich Troitsk Institute for Innovation and Fusion Research, Troitsk. Russia Received 17 July 1992, in final form 10 June 1993 a DC glow discharge in fast air flow at a pressure of up to two atmospheres are presented. A high efficiency in the production of chemically active palticles is achieved in this kind of discharge. For dry air the ozone yield was as high as 80 g kwh” at a concentration of 0.05%. The resuits of a successful application of this discharge for SO2 and NO removal from polluted air are described. A full kinetic model for the glow discharge at atmospheric pressure of dry air has been developed which explains experimental results with satisfactory accuracy. Abstract. The results of an experimental investigation of 1. Introduction 2. Experimental devices One of the global problems currently encountered is environmental pollution and many steps should be taken to diminish and dispose of hmnful pollutants. An important and promising way to remove air pollutants is the plasma technology developed over recent years. To apply this technology one needs an efficient and powerful plasma source at atmospheric pressure. The well developed arc discharge technique produced thermal plasmas with equilibrium temperature up to 10000 K. Employment of the non-thermal plasma technique is a promising way to improve the energy and, hence, cost efficiency of pollutant removal. Techniques involving corona and silent discharges and electron beam systems have been used in the past. For example, a pulsed corona and an e-beam were employed to remove SO2 and NO, from stack gases [1-5] and a silent (or barrier) discharge i s widely employed for ozone production [61. A glow discharge in a fast gas flow has also been considered as an efficient non-thermal plasma source for applications in pollution control [7]. However, experimental results [SI did not promise this system to be effective. The aim of this paper is to present the results of experimental and theoretical research into a glow discharge in air flow at atmospheric and overpressure. The successful application of this discharge for the removal of SO2 and NO, from air will be described in brief and compared with pulsed corona and e-beam systems. The configuration of the discharge installation in our experiment was quite similar to that of multi-pin corona [SI. The gas flowed through a rectangular channel, the top side of which was the anode plate, and the multipin cathode was built in the bottom side of the channel. Each pin was individually ballasted by a I MC2 resistor. The discharge gap length could be vaned up to I cm. The air pressure was varied from 1 to 2 atmospheres, and the gas flow velocity from 70 to 200 m s-’. A DC power supply provided voltages up to 20 kV, which was sufficient to sustain the steady state glow discharge. In the experiments the partial currents on selected groups of the cathode pins and the total current were measured. The electric probe method was developed to measure the electric field strength distribution in the discharge gap. The experimental set-up for the removal of SO2 and NO, is sketched schematically in figure 1. Dry air, which was contained in a high-pressure balloon ( I ) was heated in the heater (2). then mixed in (3) with water vapour, SO2 or NO, or dust and was finally blown through the discharge channel (4). The air flow after the discharge treatment was slowed down to 0.5-1 m s-’ in the chemical reactor (5). The gas temperature in the reactor could be varied by premixing of cooled air. The length of the reactor was 1 m. A filter was situated at the reactor output. The heater provided gas temperatures up to 100°C. The mixer (3) allowed one to add a fixed dose of NH,, SOa, NO, (up to 0.3% vol), water vapour (up to 10% vol) and dust (up to N 150 mg m-3, where N means normal conditions). The evaporator, ducts and 0022-3727/93/101630+08$07.50@ 1993 IOP Publishing Ltd oc glow discharge at atmospheric pressure ANODE CAWQDE 20, Figure 1. Schema of the experimental set-up for gas cleaning: 1, dry air balloon: 2, heater: 3, mixer: 4, glow discharge; 5, chemical reactor: 6, water vapour; 7,ammonia; 8,sulphur oxide: 9, dust; 10, power supply: 11, bag filter. mixer were all thermally isolated and could be heated individually. SO2 and NH3 could also be premixed at the chemical reactor. The gas temperature was measured at many sites of the tract. The process of the plasma assisted NO, removal was studied in an installation placed in a real stack gas duct at the local boiler house. This installation consisted of the discharge in the fast stack gas flow and the chemical reactor with the slow gas flow. The concentrations of admixtures and some products of plasma treatment were measured by optical or chemical diagnostics. The system for the data collection and mathematical treatment was automated. 201 Figure 2. Spatial distribution of the potential and electric field in the discharge for dry air. P = 1.1 atm, v=180ms-’. U,-c 14 , kV . L 3. Experimental results 3.1. Discharge characteristics Measurements of the discharge voltage and current revealed that the average current density of about 10 mA cm-* is much higher than that typical for a barrier or corona discharge. The reduced electric field strength averaged over the discharge gap is lower than the breakdown one; however, it is sufficient ( E / N = (810) x V cm2), the electron energy being in the range 2-3.5 eV. The electrode configuration in our experiments was similar to that of multi-pin-to-plane negative corona in air. However, the electric power density in our discharge is about IO0 times higher than in the usual corona. This effect was achieved primarily by applying a fast gas flow and ballasting each cathode pin. The most important difference from a usual corona is that according to simple estimates the current camers in our case are free electrons. Visually on the each cathode pin surface there is a bright glow. The diffuse channel extends from the pin to the anode and its diameter grows slowly in the range 3-5 mm. The potential profile along the discharge gap measured by the electric probe technique is shown in figure 2, and the electric field strength calculated from this curve is also presented. The discharge voltage is an increasing function of the current (see figure 3). The glow discharge is very stable against the addition to air of water vapour up to a concentration of 0 3 6 12 9 j. mA/cm2 Figure 3. Discharge voltage against average current density for dry air. P = 1 atm, v = 150 m s-’, discharge gap length is 5 mm. 10% vol, of small admixtures such as NH,,NO,, SO,, hydrocarbons (concentrations of about 1000 ppm) and of dust. The discharge voltage increases with gas ROW velocity, water vapour concentration and gas pressure. To verify this discharge as an efficient source of chemically active species, special experiments were carried out. We had employed our discharge in dry air to produce ozone, which has many applications in industry as a very strong oxidizer. The experimental dependence of ozone concentration produced in the discharge on the energy deposition reduced to the mass of treated air is shown in figure 4. The superlinear growth of ozone concentration is the consequence of voltage increasing with current (see figure 3). The ozone concentration reached a value of 0.05% at a gas total pressure (the sum of static and dynamic pressures) of 1.3 atm and a gas flow velocity of 200 m s-’. The energy cost for the ozone production is equal to 12 kWh kg-’ of ozone in units common to ozone technology (in units more appropriate for a physicist it is 21.6 eV per mol). This is quite a good value-typical for a barrier discharge ozonizer when using air. To make a concentration of ozone of about I%, which is suitable 1631 ~ Yu S Akishev et a/ No3, (11016, E,% Cm-3) / 2 1.2 20 10 0 0.4 / 0 ' 0 20 40 60 10 30 20 80 9. J l g Figure 4. Ozone concentration as a function of energy deposition for dry air. P = 1.3 atm, v = 200 m s-' . for most consumers, it would be necessary to increase the length of the discharge along the gas flow and to cool the gas-keeping the energy cost on the same level is anticipated to be difficult. The capability of the discharge to activate the gas was also checked by its efficiency in oxidizing hydrocarbons such as butane or propane. The experiment demonstrated that the energy cost for this process in the discharge is a third of that for thermal oxidation. These experiments have confirmed that a glow discharge in a gas flow promises to be effective in the removal of harmful pollutants such as SO2 or NO, from the flue gas. Q,I/€ .. Figure 5. SO2 removal efficiency against energy deposition wilhout NH3 injection. [SO& = 1000 ppm, T = 65 C, [HzO] = 6%. E, % 32. Removal of SO2 The experiments for SO?removal from air were carried out on the installation described above (see figure I). It is well known at present that in the processes of SO? removal with addition of ammonia, only about 20-30% of the removal efficiency is due to the discharge or e beam action. The purely thermal removal efficiency is more than 70% at a gas temperature of 70°C and water vapour concentration of about 6%. The process of SO2 removal by the discharge treatment was studied, at first, without NH3 addition. The resultant dependence of the removal efficiency on the energy deposition is shown in figure 5. The initial SO2 concentration was 1000 ppm, the gas temperature was nearly 65 "Cand the water vapour concentration was 6%. The increase of the water vapour content is beneficial for the quantity of SO2 removed. For example, at an energy deposition of 15 J g-' the sulphur dioxide decrement at the same initial SO2 concentration changed 1.65 times by variation of the water content from 1.5% to 4.5%. The experiments for SO2 removal with ammonia injection to the discharge were carried out for a gas 1632 0 0.2 0.4 0.6 0.8 1 1.2 RNH3 Figure 6. SOp removal efficiency as a function of NH3 stoichiometrics. [SO2j0= 1000 ppm, Q = 20 J g-', [H20] = 6%, T = 65 'C. temperature of 60-65 "C and water vapour concentration of 6%. The dependence of the total SO2 removal efficiency on the ammonia stoichiometry is shown in figure 6. We have revealed the positive influence of ammonia on the removal efficiency by discharge treatment. For example, the discharge removal efficiency increased from 11% without ammonia to 26% with the injection of ammonia. The energy cost for SO2 removal equals 8 J g-' per 100 ppm SOz, which is comparable with the best results for e-beam technology [4] and the pulsed DC glow discharge at atmosphericpressure Table 1. [H20] = 1%. 0 (J g-i) E / N (x10i6 V cm2) ~'(~cj € (%) 7.8 23 11.7 23.5 15 23.8 23 24.5 72 74 75 78 20 28 35 50 € I N ( ~ 1 0VcmZ) '~ T ('C) E (%) ($10- '6V.cm2) 80 60 Table 2. [H20] = 9.5%. 0 (J vi) . E IN. ~. NO 6.8 22 96 15 14 23 101 26 23 23.5 106 42 31.4 24 111 63 corona technique [5]. The total removal efficiency in our experiments exceeded 96%. 40 20 33. Removal of NOx The processes of plasma assisted NO, removal were studied on the flow of real stack gas at the output of a local boiler fired by natural gas. This flow did not contain a measurable amount of SO2 and the concentration of nitrogen oxides equalled 65-75 ppm. The temperature of the flue gas varied from 85 to 120°C. The water vapour concentration varied from 8 to 12%. Ammonia could be injected before the discharge and after the discharge as well. Because the mean reduced electric field strength E / N is very important for the processes studied, tables 1 and 2 demonstrate the correlation of this quantity with parameters such as the gas temperature T , the reduced energy deposition Q, and the moisture content; the NO removal efficiency E for these conditions is also presented. Spectroscopic measurements had shown that in all cases the concentration of the nitrogen dioxide was negligible in the initial stack gas. This result was confirmed by chemical diagnostics ma& by our colleagues at the Gas Cleaning Technology Research Institute. The concentration of nitrogen dioxide at the chamber exit was not measured and the removal efficiency was defined by diminishing the concentration of NO only. At first, we studied the NO removal efficiency as a function of the gas flow velocity, the energy deposition and the moisture content without added ammonia. The results are presented in figures 7 and 8. By increasing the gas flow velocity, the discharge voltage and the removal efficiency both increased; this may be explained by more effective dissociation of molecules in the discharge. Figure 8 demonstrates that the removal efficiency is nearly proportional to the energy deposition and increases with water vapour concentration (the effect of diminishing flow velocity is overwhelmed by an increase in water vapour content). This behaviour is evident because the most effective way of removing NO is by the radical OH. It should be noted that the NO concentration in our experiments is much lower than is typical [2-51. As a result, an essential proportion of OH produced in the discharge may be destroyed in other reactions before meeting an NO molecule. We 0 100 50 Y, m / s Figure 7. NO removal efficiencyand reduced electric field strength as functionsof the gas flow velocity with no NH3 injection. [Nola= 65 ppm, Q = 35 J g-', [HzO] = 9.5%, T = 12O'C. €,% 0 10 20 30 40 9, J l g Figure 8. NO removal efficiency against energy deposition with no NH3 injection. [Nolo= 65 ppm; 0, v = 109 m s-', [HpO] = 9.5%; 0 , v = 96 m SMi, [HpO] = 12%, T = 12O'C. expect that the removal efficiency will increase with NO concentration, at least in the range of several 100 ppm. As in studies on NO removal from gases by virtue of e-beam treatment [3,4], the negative effect of increased gas temperature on removal efficiency was observed. We did not detect any other discharge parameters change while varying the gas temperature from 85 to 12OoC, so the negative temperature effect is of a purely chemical nature. Bearing in mind the problem of simultaneous removal of sulphur and nitrogen oxides we explored 1633 Yu S Akishev et a/ E,% 6o 1. 0 O 10 20 30 40 I 0.l 1 10 R N H ~ Figure 9. NO removal efficiency as a function of NH3 stoichiometrlw with injection to discharge. [Nolo = 65%, 0 = 2 2 J g - ’ , ~ = 1 0 9 m s - ’ , [ H ~ 0 ] = 9 . 5 ~T=95’C. ~, the effect of ammonia injection before and after the discharge treatment of the gas. A decrease in the removal efficiency with ammonia addition was observed (see figure 9) when it was injected before the discharge. In our opinion this effect is due to the variation of discharge characteristics with the ammonia concentration. Figure IO demonstrates that the post discharge injection of ammonia does not have a strong influence on the removal efficiency. Comparing figures 8 and IO one can see the negative influence of the high gas temperature. 4. Modelling of the discharge The experimental measurements of the electric field distribution (see figure 2) gave a striking result: a quasineutral plasma with electronic conductivify exists under an electric field essentially lower than that of the breakdown. In this case attachment processes are dominant in the plasma, so the question of ionization of the gas flowing into the discharge region must be answered. This problem may be adequately treated only by a two- or three-dimensional model, involving the solution of Poisson’s equation. However, the principal problem of finding the conditions under which the charged particle production is greater than their loss is kinetic in its nature. So, for the simulation of a glow discharge in an air flow we adopt a zero-dimensional kinetic model. The evolution of a small portion of the gas transported by the flow was described as the temporal evolution. Our kinetic model is similar in some respects to the recently published model 191, aimed at explaining the glow phase of the air spark breakdown. 1634 9 , 518 Figure 10. NO removal efficiency against energy deposition, NH3 injection after the discharge. [NO10 = 65%. v = 101 m s-l,[H20] = 8.4%, T = 95’C. Ammonia stoichiometry is: 0, 0; A , 0.5; 0 , 0.75; 0 , 1.5. Let us explore the evolution of plasma in dry air considered as the mixture N2 : O2 = 4 : 1 under an electric field, which is calculated by solving the equation for the electric circuit, including the power supply, the ballast resistor and the discharge. Our kinetic model consists of several parts. The first part is to solve the Boltzmann equation. For the case of the glow discharge considered this part may be greatly simplified. Because of very low-energy deposition into the gas the variation of the gas and vibrational temperatures may be neglected. The electron-electron, electron-ion collisions and any collisions of the second kind may also be neglected. In this case all the kinetic coefficients for the processes involving electrons are functions of the reduced electric field strength E / N and the gas composition. They may be calculated in advance and kept in the computer memory. The second part includes a rather complete set of kinetic equations describing ion- and electron-molecule interactions and chemical reactions of neutral and charged particles. Although the vibrational excitation of molecules has little influence on the processes, our model includes the standard equations of vibrational kinetics (see, for example, [IO]). Our model allows calculation of the evolution of many important plasma components like the electrons, ions (Nl,O:, 02, 0-, 0;. 0;).electronically excited particles (N2(A3Z$), N;, Oz(a’A), O2(b1Z), O(’D)), atoms (N, 0),ozone (03). nitrogen oxides (NO, NzO,NOz, NO3, NzOs). Here N; denotes the electronically excited nitrogen molecule at any level except A3E:. Omitting here the list of more than one hundred reactions for these species, let us discuss some of the most important reactions in understanding the atmospheric glow discharge. At the experimentally measured electric field strength the attachment rate is certainly greater than the ionization oc glow discharge at atmospheric pressure rate by electron impact. Another electron loss process is electron-ion recombination. The only process that may, in principle, compensate for electron losses is the detachment of electrons from the negative ion. However, there is no reliable information available about the rates of detachment processes in air plasmas at atmospheric pressure and under an electric field. The most important attachment process for dry air is the dissociative attachment to 0 2 . So the main primary negative ion is 0-; furthermore, it may be converted to the 0; and 0; and so on. This ion may also be destroyed in collisions with molecules, atoms and excited particles. The most important process, in our opinion, is collisional detachment in the reaction 0N2 --f N 2 0 e. The rate of this process has been measured as a function of the electric field strength [ I l l . Previous data [12,13] allow one to conclude that the rate of the conversion of 0- to other oxygen negative ions is of the same order as that of the detachment, and both are much higher than the ionization and recombination rates. As a result, the concentration of 0- is quickly established to be near the electron concentration, and both are much less than the concentrations of positive ions and ions O;, 0;. In the steady state the concentration of positive ions is determined by the balance of ionization and ion-ion recombination became the electron-ion recombination is negligible. Because of the quasi-neutrality condition the concentration of the sum of ions 0; and 0; must be nearly equal to the positive ion concentration. Considering all these points several inequalities and approximate formulae may be derived as follows. Let us call n I the concentration of 0-, n2 the total concentration of 0; and O;,nP the concentration of positive ions and ne the electron concentration. Then, obviously, a short time after the discharge start the inequality ne,n , << n2, np would be established. Denoting the dissociative attachment rate to 0 2 by v,, the collisional detachment rate from 0- by vdl and from 0; 0; by VdZ, the negative ion conversion 0- -+ O;,0; rate by v,, the simple relation between the 0- and e concentrationsmay be derived as + + + nl = neva/(vdl f vc). (1) The kinetic equation for nz has the form where pii is the effective ion-ion recombination coefficient. One can see that employing relation (I) allows one to eliminate the concentration of 0- and to bring in an apparent attachment rate for the negative ions of second type v a e = ~ v,vc/(v, va,). The evolution of the positive ions is described by the approximate equation dn= ~ vine - piinznp (3) + df where Q is the electron impact ionization rate and the electron-ion dissociative recombination is neglected. So equations (2) and (3) together with the equation for the electric circuit are suficient to describe the evolution of air plasmas under an electric field somewhat lower than the breakdown field. At low charged particle concentrations the recombination processes may be neglected and two distinct stages of plasma evolution may be determined. In the first stage the electrons are mainly attached, resulting in a negative ion concentration n 2 increasing to the equilibrium between the attachmentdetachment processes. At the second stage all the concentrations increase due to the ionization process according to the approximate equations: n2 = newa d V d 2 2 np This stage may be described by the apparent ionization rate vi = %vrlZ/va eff. Finally, this effective plasmas multiplication must be limited by the ionion recombination processes according to equation (3). It is no problem to find the resulting quasi-stationary concentrations of charged particles when using the equation for the electric circuit. One should remember, however, that the gas composition varies due to the reactions induced by the discharge. In particular, the ozone concentration increases. As a result, new reactions of negative ion conversion and electron attachment become important. These effects may be described by the same simplified model, but the variation of the appropriate kinetic processes should be included adequately. The results of numerical calculations by using the full kinetic model confirmed the simplified picture developed above. The approximate theory demonstrated that uncertainties in the data about collisional detachment from negative oxygen ions, negative ion conversion and ion-ion recombination are equally important. More thorough research on these processes is therefore required. Figure 11 presents a comparison between the experimentally measured discharge current distribution over four sections along the gas flow and calculated evolution of the discharge current. Two scales, spatial z and temporal are coupled through the relationship z = V t , where V is the gas flow velocity. The agreement between theory and experiment is quite satisfactory. The calculated electron and ion currents are also shown in figure 11. Figure 12 shows the calculated (full curves) dissociative and apparent attachment rates and the ionization rate and the experimental values for the attachment rate as functions of the reduced electric field. Noting a great scattering in the data the decline in the apparent attachment rate is similar to the experimental one for high electric fields. The apparent ionization rate gives a characteristic time of the order of 5 ps. Using this value the ionization length along the gas flow is less than I mm, and along the electric field 2 mm. These lengths are shorter than the size of the discharge gap and the cathode electrode 1635 Yu S Akishev el al 2,cm 0 2 1 3 J, mA/cmZ model additional gas components such as HzO, CO*. NHJ and so on. Obviously, the problem of describing all the important processes in such mixtures is much more difficult because of complex kinetics, though this problem is solvable in principle. 5. Conclusions IO 0 0 100 200 t. PS Figure 11. Comparison of experimental current distribution over electrode sections with the calculated value. Calculated discharge voltage and electron and ion currents are also shown. Dry air, P = 1 atm, v = 150 m s-', discharge gap length d = 5 mm. E/N, ( ~ 1 0 - 1 6 V.cmZ) Figure 12. Calculated dependence (full curves) of ionization (cuwe I), dissociative attachment (2) and apparent attachment (3) coefficientsagainst the reduced electric field strength. Experimental data for the attachment coefficient are shown by the symbols: a,f r a n [16]; 0, from [lq;<:I, from [9]; 0 , from [18]; A, from [19]. length. This fact is the argument for the applicability of our zero-dimensional model. The model formulated in this section demonstrates the possibility to predict the main characteristics of a glow discharge in air flow at atmospheric pressure with reasonable accuracy. To describe the discharge in real gases to be cleaned it is necessary to include in the 1636 Our research has proved that a classical glow discharge, previously obtained only at an air pressure less than 101) Torr, may be sustained at air pressures up to 1520 Torr. The spatial discharge current and electric field distributions can be regarded as being reasonably homogeneous. The electric field strength, lower than the breakdown one, is high enough for the effectiveness of the discharge in producing active species to be beneficial for many applications. The main discharge characteristics may be described by the kinetic model developed, at present, only for a discharge in dry air. In future this model may be easily extended to consider actual polluted gases such as stack gas. It would be of great interest to compare the techniques for pollution control based on the e-beam [2,3], pulsed corona [1,2, 51 and barrier discharge [ 14,151 with the technique based on the DC glow discharge described above. One of the important parameters for comparison of different pollution control technologies is the energy cost for the removal, for example, 100 ppm of pollutants. For SO2 typical value of this cost is IO J g-' for the e-beam technique [3], 25 J g-' for the pulsed corona technique [5] and 15-30 J g-I for the barrier discharge technique [I51 (these figures are the result of computations based on a room-temperature experiment) and IO J g-' for the glow discharge technique. For NO, the typical values of this cost are: 15 J g-' for the e-beam treatment, 1530 J g-' for the pulsed COIUM technique and 60 J g-' for the glow discharge technique. The last figure was obtained for a very low initial NO concentration, and we anticipate that the results would be better for greater NO concentrations. Taking into account the simplicity of the technique, the glow discharge technology described seems to be advantageous for environmental applications. Acknowledgments We thank Luigi Civitano for useful discussion. We are also grateful to L V Degaeva and Yu M Afanasiev for help in this work. References Mizuno A and Ciements J 1987 US Parent N4.695,358 Jen-Shih Chang 1989 Pmc. 2nd Inr. Symp. on Hieh Pressure Low Temp, Plasma Chemisfry.Hacone 2 (Poland) Sepfember 12-14 1989 pp 103-8 Jordan S 1990 Radial. Phys. Chem. 35 N 1-3, p 409-415 Willibald U, Piatzer K H and Winig S 1990 Radiat. Phys. Chem. 35 N 1-3, p 422-6 DC G , Civitano L and Rea M 1990 IEEE Trans. Ind. 26 535 [61 Eliasson B and Kogelschatz U 1986 J. Chem. Phys. 83 151 Dinelli .~ Appl. 279 [71 Kan& I E and Hundstad R L 1987 US Palent N 4,657,738 181 Yamammo T, Lawless P and Parks W 1988 IEEE Tram. Ind. Appl. 24 934 [91 Rodriguez A E, Morgan W L, Touryan K J and Moeny W M 1991 1.Appl. Phys. 70 2015 [IO] Cordiets B E Osipov A I and Shelepin L A 1980 Kinetic Processes in Cases and Molecular Lasers (Moscow: Nauka) [I 11 Rayment S W and Moruzzi J L 1978 Inr. J. Mass. Spectinsc. Ion. Physics 26 321 glow discharge at atmospheric pressure (121 Kossyi I A, Kostinsky A Yu, Matveev A A and Silakov V P 1992 Plasma Sowces Sri. Technol. 1 207 [13] Samoylovich V 0, Gibalov V I and Kozlov K V 1989 Pkysical Chemisriy of Barrier Dischmge (Moscow: Moscow State University) [ 141 Sardja I and Dhali S K 1990 Appl. Phys. Len. 56 21 [IS] Chang M 9, Balbach J H.Rood M I and Kushner M J 1991 J . Appl. Phys. 69 4409 E161 Frommhold L 1964 Forrschr.irt der Physik 12 597 1171 Warren R W and Parker J H 1962 Pkys. Rev. 128 2661 [IS] Crunberg R 1969 2.Name A 24 1039 1191 Ryzhko H 1965 Proc. Phys. Sac. 85 1283 1637