Aerosol Science 37 (2006) 1069 – 1080 www.elsevier.com/locate/jaerosci
Characterization of unipolar electrical aerosol chargers—Part II: Application of comparison criteria to various types of nanoaerosol charging devices A. Marquard∗ , J. Meyer, G. Kasper Institut für Mechanische Verfahrenstechnik und Mechanik, Universität Karlsruhe (TH), Germany Received 3 June 2005; received in revised form 26 August 2005; accepted 5 September 2005
Abstract The framework presented in Part I for comparing the performance of electrical aerosol chargers is applied to published data as well as to new measurements with unipolar corona chargers of very different designs in the size range between 10 and 60 nm. Variables include the internal flow conditions, the ion production arrangement, and the type of electrical field in the mixing zone between ions and particles. Comparisons are made in terms of losses of charged particles vs. charging efficiency and/or vs. average charge per particle. It was found that each charger has a characteristic curve of losses vs. charging effectiveness, comprising data points for a broad range of electrical and ion generation conditions. The results further suggest that the high charge levels and low losses can be achieved by a charger design combining short residence times (associated with turbulence) and high ion concentrations. A carefully adjusted sheath flow is somewhat helpful (at the cost of diluting the aerosol), whereas an AC electric field in the mixing zone of ions and particles does not contribute to improving the charge–loss relationship, even under laminar flow conditions. The uniformity of charging conditions for a weakly charged nanoaerosol in a given charger can be characterized without the need to determine n·t-products, from a plot of measured intrinsic average charge vs. intrinsic charging efficiency and comparing with predictions from charging theory. 䉷 2005 Elsevier Ltd. All rights reserved. Keywords: Nanoaerosols; Unipolar aerosol charging; Charger comparison; Particle losses
0. Introduction The literature contains a large number of design variations for aerosol charging devices, which differ substantially both with respect to purpose and performance. Selecting the right design and operating mode for a specific application requires rigorous, consistent and as far as possible unambiguous performance criteria. In Part I of this paper we have reviewed a variety of such criteria from the angle of particle losses vs. charging effectiveness (the latter expressed as efficiency and average charge per particle), and we have discussed the underlying assumptions made in deriving them from measurements. We also introduced a set of charge vs. loss diagrams for characterizing and comparing ∗ Corresponding author. Tel.: +49721608 6565; fax: +49721608 6563.
E-mail address:
[email protected] (A. Marquard). 0021-8502/$ - see front matter 䉷 2005 Elsevier Ltd. All rights reserved. doi:10.1016/j.jaerosci.2005.09.002
1070
A. Marquard et al. / Aerosol Science 37 (2006) 1069 – 1080
Fig. 1. Summary graph of (partly recalculated) charge and loss data from various nanoaerosol charging studies in the size range of 10–30 nm. Data of Büscher et al. (1994) and Hernandez-Sierra et al. (2003) are for several charger voltages; data of Chen and Pui (1999), Kruis and Fissan (2001) and Biskos et al. (2005) are for a single operating point each. The diagonal represents the limiting case intr = 1 = extr + lossc (Eq. 22).
chargers objectively without requiring internal charger information such as ion concentrations or particle residence time distributions. We now proceed to apply this framework to published data for unipolar nanoaerosol chargers, as well as to measurements of our own with several types of chargers some of which have been presented at earlier conferences (Marquard, Ehouarn, & Kasper, 2002; Marquard, Bredin, Meyer, & Kasper, 2004a, b), almost all of which are based on ion production by corona discharge. The literature yielded only five charger studies (four of them with corona) in a comparable particle size range, for which loss information could be derived and which use the same charging parameter, namely intrinsic and extrinsic charging efficiencies. Unfortunately the literature on photoelectric chargers could not be used, because loss information is generally missing. Rather than focus on completely understanding and perhaps even modeling a particular type of charger, these data will mostly be discussed in the light of comparative information gained with regard to the advantages or disadvantages of a particular charger design in terms of charging effectiveness vs. losses of charged particles. A second objective is to highlight the advantages of the framework suggested in Part I for such performance comparisons. 1. Re-evaluation of published nanoaerosol charger data The concept presented in Part I is first applied to published unipolar charging experiments from five different studies, by Büscher, Schmidt-Ott, and Wiedensohler (1994), Chen and Pui (1999), Kruis and Fissan (2001), Hernandez-Sierra, Alguacil, and Alonso (2003) and Biskos, Reavell, and Collings (2005). Common to all these studies is a particle size range of approximately 10–30 nm and the fact that implicit or explicit loss information was given along with extrinsic efficiencies. Büscher et al. and Hernandez-Sierra et al. investigated the charging of sodium chloride particles, Kruis and Fissan and Chen and Pui conducted experiments with silver particles, while Biskos et al. used an unspecified combustion aerosol. All studies were done with chargers employing the diffusion charging mechanism; three of them (Büscher et al., Kruis and Fissan and Biskos et al.) with different types of modified Hewitt chargers (Hewitt, 1957) utilizing indirect corona (ion production separated from particle charging) and AC-fields; one study used direct DCcorona (charging and ion production within the same chamber, Hernandez-Sierra et al.), and one generated ions with a Po210 radioactive source (Chen and Pui: ion and particle movement collinear). Sheath air was used by Büscher et al. (dilution factor f = 1.25), Chen and Pui (f = 4) and Biskos et al. (f not given, but probably not far from unity). Some of the data had to be recalculated according to Eqs. (A3), (A5), (A13) and (A15) from Part I in order to obtain the desired parameters. Because only one study contains particle charge information (Biskos et al.), it was not possible to use the plots recommended in Part I, Section 3. Instead, the comparison is in terms of lossc vs. extr , as shown in Fig. 1 for 10 nm particles (black symbols), 20 nm (grey symbols) and 30 nm (open symbols). Data points of a particular symbol which are connected by a line result from variations of the charger voltage (Büscher et al. and Hernandez-Sierra
A. Marquard et al. / Aerosol Science 37 (2006) 1069 – 1080
1071
et al.), with increasing voltages generally corresponding to higher efficiency and/or higher losses. It has to be noted also that recalculation of the Büscher data according to Eqs. (A13) and (A15) (from Part I) with the published dilution factor f = 1.25 resulted in negative loss values. Since this is not reasonable, particle production within the charger being unlikely, we proceeded with the assumption f = 1. Neither Kruis and Fissan nor Chen and Pui provide explicit loss information. The loss of charged particles, lossc , was therefore estimated on the assumption of complete charging according to Eq. (A7), except for the 10 nm data where is definitely less than unity. When interpreting Fig. 1, it should be kept in mind that extr is not independent of lossc because intr = extr + lossc : according to Eq. (22). Hence data points for complete charging (intr = 1) will lie on the diagonal, while for partial charging they will be located below. In other words, the more intrinsically efficient a charger, the closer upwards to the diagonal the data points will be positioned. There are very large differences in the loss-efficiency relations of the five charger studies represented in Fig. 1, which are best discussed by separating the data into a group around 10 nm, and a group of 20–30 nm particles. The 10 nm data are generated by three chargers, which all have roughly the same extrinsic charging efficiency. The device investigated by Biskos et al. (with an AC field in the mixing zone) has by far the highest losses, although it had the highest intrinsic efficiency before the particles got lost; the other two are about equal in performance. Since the device of Büscher et al. uses an AC field as well, it would seem that this approach does not offer any advantage over a DC field in preventing losses. Comparing the 20 and 30 nm data, the charger by Chen and Pui (triangular symbols in Fig. 1) has by far the highest charging efficiency and the lowest losses, however, at the expense of the highest dilution factor. Again we note that the ‘AC chargers’ (Biskos et al., Büscher et al., Kruis and Fissan) are inferior in performance, although two of them probably had an intrinsic efficiency of 100%. (Chen and Kruis give no explicit loss information, and thus the 10 nm data of these studies could not be included in the graph.) Comparing the devices with AC (Biskos et al., Büscher et al., Kruis and Fissan) and DC (Hernandez-Sierra et al., Chen and Pui) electric fields inside the charging zone, Fig. 1 shows that AC chargers do not reduce losses of nanoparticles, despite efforts (especially by Biskos et al.) to optimize the laminar aerosol flow. The AC field seems to only affect the amount of ions reaching the charging region. This is at first somewhat unexpected, since charging experiments in an AC field with larger particles (e.g. 4 m; Lackowski, 2001) have shown that the control of particle trajectories by an AC field works well. Its lack of success for nanoparticles probably has two explanations: Nanoparticles are more susceptible to hydrodynamic forces from the ion wind or flow disturbances due to the short inlet geometry, which often makes transport to the walls unavoidable, irrespective of an additional electric field. Secondly, the electric forces of the ion cloud drive charged and highly mobile particles to the wall. Recent data by Krupa, Jaworek, Lackowski, Czech, and Luckner (2005) with rather large particles (Sauter mean diameter 9 m) seems to confirm the overriding effect of EHD mechanisms for particle losses, because they show that the loss reducing effect of an AC field is somewhat inversely proportional to the charging intensity. Fig. 1 clearly illustrates the value of a presentation of loss of charged particles vs. efficiency in comparing charger performance. Had one plotted only the extrinsic efficiencies of these chargers against the particle size, many results would have appeared about equal, especially for the 10 nm group, although the losses vary from 20% to 70%. (Of course one could have also plotted losses vs. size, but that would have required a separate diagram for comparisons. And not all studies provide loss information.) Fig. 1 further shows that discussing charger data in a plot of losses vs. extrinsic efficiency is not that straightforward. Since the intrinsic efficiencies were often 100% (i.e. data points were on or near the diagonal), it would have been even more revealing to compare losses vs. average charge. Unfortunately, particle charge data were only reported by Biskos et al. and the available information therefore does not support further quantitative and objective comparisons. Since Biskos and others found that even particles as small as 20 nm can have multiple charges, the charger of Biskos et al. might have represented a better compromise in such a plot, despite its rather high losses.
2. Comparative experiments for several corona chargers in a low-diffusion-loss scenario New experiments were carried out in order to compare three different charger concepts on the basis of their chargevs.-loss performance for a variety of electrical set-ups, arrangements of ion generation and hydrodynamic operating conditions. When presenting these results here it should be kept in mind that the goal is not charger optimization, but
1072
A. Marquard et al. / Aerosol Science 37 (2006) 1069 – 1080
Fig. 2. Set-up for measuring average charges q, ¯ charging efficiencies and total losses with monodisperse DEHS-aerosol of a fixed size.
to identify the relevance of certain design parameters such as AC electric fields, gas flow conditions, or direct/indirect corona, on their overall performance. Due to the many relevant experimental variables, the number of experiments was reduced by working at a single, fixed particle size of 61 nm, and by keeping the chargers geometries unchanged. Working in this size range has the advantage of negligible diffusive losses (with attendant consequences for the charger performance parameters), while having the freedom to simulate the weak charging of much smaller particles by an appropriate choice of the experimental conditions. 2.1. Experimental set-up and chargers investigated The experimental set-up consists of an aerosol generator, the charger, and analytical equipment to characterize the aerosol (Fig. 2). A polydisperse DEHS droplet aerosol (dg = 55 nm, g = 1.5) is generated via an evaporation/condensation process according the well-known Sinclair–LaMer principle, then classified in a mobility analyzer to narrow the size distribution (DMA 3071, TSI Inc. with 85 Kr neutralizers) followed by re-neutralization. Charged particles are removed by parallel plate electrostatic precipitators (ESPs) without corona electrodes. The remaining, uncharged aerosol is mixed with additional carrier gas at STP (filtered air, dried to RH < 4%), to assure low concentrations (c = 104 cm−3 ) of a narrow, approximately log-normal aerosol with dg = 61 nm and g 1.2. Diffusion is negligible for these particles during the residence times encountered here. Aerosol samples are taken at the inlet and outlet of the charger to determine particle concentrations (by CPC 3022, TSI Inc.), size distributions (by SMPS 3934, TSI Inc.) and mean particle charges q, the latter from the ratio of current transported by particles (measured with a sensitive Faraday cup electrometer ‘FCE’ of our own design) and the particle flow rate (by CPC 3010, TSI Inc.). An ion trap consisting of two concentric electrodes, one grounded, the other connected to an AC voltage (5 V, 50 Hz), is placed at the inlet of the FCE. Uncharged particles leaving the charger are detected by means of a second ESP and the third CPC. Care was taken to first calibrate the CPCs by charge measurements after the DMA, and then to operate them at concentrations below 104 cm−3 , the operating range with highest accuracy. For the DC corona, a negative or positive glow discharge regime was adjusted and controlled by means of an oscilloscope (Mod. TDS 210, Tektronix Inc.); for the AC corona this could not be verified. Current measurements were performed with a variety of electrometers (Mod. 617, Keithley Inc., HV electrometers, own design). A schematic of the first charger is shown in Fig 3a. It is a twin corona module charger consisting of a cubic chamber made of polyethylene (w × h × l: 70 × 90 × 120 mm) with two charging modules placed on opposing side walls perpendicular to the main flow (11 or 20 l/min, resp.). Within these modules, ions are generated in a point-ring corona configuration (1 cm gap, ring diameter 15 mm) based on the ion gun concept of Whitby (1961) and transported by humidified air (50% RH; 1, 10 or 20 l/min, resp.) through the ring into the particle charging chamber. We refer to this arrangement where ion generation and ion-particle mixing zone are spatially separated as an ‘indirect corona’. Inside
A. Marquard et al. / Aerosol Science 37 (2006) 1069 – 1080
1073
metal insulator
Qelectrodes
Qelectrodes DC HV needle ring
DC HV
AC HV
plate
Qaerosol
Qtotal
CHARGING ZONE
Q aerosol
Qtotal
CHARGING ZONE
AC HV: φ +180°
DC HV
DC HV
(a)
(b)
Fig. 3. Chargers for own experiments and electrical connections. (a) Twin corona module charger. (b) EAN 581 corona neutralizer, operated in unipolar mode (Topas GmbH).
the charger, additional metallic plates are placed around the corona module holes. Particle residence times are approx. 3 s < t < 10 s, dilution ratios resulting from the corona flows are 1.1 < f < 1.9. With this charger design, it is possible to establish different electrical fields within the charging zone, somewhat independent of the ion production process. The abbreviation FOP (‘flow operating point’) used in the following, refers to the main flow rate and to the sum of both electrode flow rates expressed in units of l/min: e.g. FOP 11/1. The second charger is a device used commercially for aerosol neutralization (EAN 581, Topas GmbH) of more or less highly charged gasborne particles (> 100 nm) by a bipolar ion atmosphere. Here, however, it was operated as a unipolar charger. The EAN 581 also consists of an ‘indirect corona’ arrangement. It is essentially a corona charger with two corona modules mounted on opposite sides of a stainless steel tube (inner diameter 0.012 m, distance between first module and outlet 0.07 m, distance between modules 0.03 m), through which the aerosol passes (Fig. 3b). Each module comprises a needle-ring electrode configuration with supporting air flow to draw ions through the ring. The modules are similar to those in the first charger (gap 1 mm, ring diameter 2 mm), but they are connected to the aerosol zone via an additional intermediate chamber separating them from the aerosol tube. This chamber is connected to the aerosol zone via six circumferentially placed holes (1 mm) operated as critical orifices. Aerosol and electrode flows were 21 and 50 standard-l/min, respectively, at 1 bar operating pressure, yielding a Reynolds number Re?2300 in the aerosol tube, a dilution ratio of f = 3.4 and aerosol residence times of less than 0.05 s. The mixing zone between aerosol and ions is thus moderately turbulent. The third charger is an straightforward electrostatic precipitator design (ESP) consisting of a cylindrical tube (L = 26 cm, ∅i = 4 cm) with a coaxial tungsten wire (∅w = 0.1 mm). In this ‘direct corona’ arrangement ions are generated and mixed with particles within the same chamber. By shielding both ends of the wire, the length of the corona was set exactly to L = 24 cm and end effects were eliminated, thereby creating a very uniform discharge zone (see Marquard, Kasper, Meyer, & Kasper, 2005). The average carrier gas flow velocity was fixed at 1.9 m/s, well inside the turbulent flow regime, with no dilution (dilution factor f = 1). 2.2. Charger performance vs. charger operating parameters (in other words: raw data) Sample data series for the three devices are presented at first in Fig. 4 for directly accessible performance values q¯exit , exit and lossc which are plotted against the charger operating parameters. The data of the twin corona module
A. Marquard et al. / Aerosol Science 37 (2006) 1069 – 1080 2
80
1.5
60
qexit
1
40 0.5
20
lossc
0 0
200
400
600
0 1000
800
frequency ε / Hz
(a)
exit charging efficiency / % loss of charged particles / %
100
2
EAN 581 80
1.5
εexit
60 40
0.5
20 c
loss 0
0 10
20
30
40
50
60
corona current Ic / µA
(b)
exit charging efficiency / % loss of charged particles / %
1
qexit
0
(c)
average charge / e
twin corona module charger
εexit
average charge / e
exit charging efficiency / % loss of charged particles / %
100
2
100
εexit
tube ESP 80
1.5
qexit
60
1 40 0.5
20 0 4500
lossc 4600
4700
4800
average charge / e
1074
0 4900
corona voltage Uc / V
Fig. 4. Directly measurable charger performance parameters q¯exit , exit and lossc vs. operating parameters of the three chargers. (a) Twin-module corona charger: needles on −10 kV, square wave AC voltage on rings, plates at 4 kV; gas flow FOP 11/10, variable frequency . (b) EAN 581 with one module active; variable corona needle current. (c) Cylindrical ESP: variable corona wire voltage.
charger are shown in Fig. 4a for a ‘symmetrical’ operating point with needles on −8.8 kV DC; the plates and rings were both on ±2 kV square wave AC of variable frequency; the AC voltages of the two modules had a phase shift of 180◦ . The flow rates were fixed at a flow operating point (defined earlier) FOP 11/1. The charge data of the EAN 581 are plotted vs. the corona currents (Fig. 4b) and the ESP data are given for increasing corona voltage (Fig. 4c). Apart from the EAN581, the data in Fig. 4 shows that higher particle charges and charging efficiencies generally correspond to increased losses. The experiments with the EAN 581 gave the lowest charging and showed no significant losses at all, both presumably due to the short particle residence time. A detailed discussion of the relevance of the experimental parameters frequency , corona current Ic and corona voltage Uc is omitted here, since this study does not address details of charger design, but the comparison of different charging modes and chargers. Nevertheless, it becomes obvious from these three diagrams that finding the best compromise between charging and losses with this kind of data presentation would be tedious, especially when additional measurements series increase the extend of data and one has to go through a lot of diagrams to get an overview.
twin corona module charger
loss of charged particles lossc / %
100 1 kV AC amplitude
80 60 sinus signal
40
rings grounded & plates on square wave AC
1 module active 20 0 0
0.5
1
2
-40 rings & plates on square wave AC
-60 -80
-100
4 kV amplitude: rings & plates on square wave AC
(a)
twin corona module charger
4 kV AC amplitude
80 rings grounded & plates on square wave AC
60 40
sinus signal
20 0 0
2.5
average charge q / e
-20
intrinsic efficiency εintr / %
1.5
1075
100
intrinsic efficiency εintr / %
loss of charged particles lossc / %
A. Marquard et al. / Aerosol Science 37 (2006) 1069 – 1080
0.5
-20
1
1.5
2
2.5
average charge q / e rings & plates on square wave AC
-40 -60 -80
-100
rings & plates on square wave AC
(b)
Fig. 5. Twin corona module charger performance comparison for different AC voltage set-ups (aerosol flow 11 l/min, module air flow 10 l/min (FOP 11/10)). (a) 1 kV AC amplitude voltage. (b) 4 kV AC amplitude voltage.
2.3. Loss vs. charge characteristics Plotting lossc and intr vs. q¯ (= q¯exit ) removes the explicit dependence of these parameters on operating conditions and offers a more complete way of comparing different chargers and operating modes, as compared to the diagram of Fig. 1 which uses only lossc and extr . In the following diagrams full information about the charge and loss results is contained and therefore, all other efficiency, loss and charge parameters can be derived by the definitions and conversions given in Part I (Sections 2 and 3, Appendix A). However, in this section focus lays only on the loss vs. charge part of the diagrams; the efficiency vs. charge plots will be discussed in Section 2.4. Since the EAN 581 neutralizer was operated practically without losses, its performance will be discussed only in Section 2.4. At first, the twin corona module charger is studied for different electric and hydrodynamic conditions. The advantage of this charger is its flexibility to emulate different electrical and hydrodynamic situations resembling the multitude of charger concepts known from the literature. On the other hand, the distribution and concentration of ions in the charging zone is very sensitive to small changes in such factors as humidity, temperature, misalignment of the corona needles, or wall charging effects, because only a small fraction of the ions produced in the corona modules enters the particle chamber. However, absolute reproducibility is not needed here as long as relative reproducibility can be guaranteed. This was achieved by the following strategy: All experiments were divided into main groups each consisting of individual series of measurements (e.g. a group at a fixed gas flow rate containing series with different electrical operating modes). At the end of each main group the first series of experiments was repeated. Only series, which passed this reproducibility test, were used for comparison. Relative reproducibility was generally very good when the entire main group was done without interruption. The impact of different AC set-ups on the charge-vs.-loss performance was investigated at one fixed flow condition FOP 11/10 and a constant DC corona voltage (−9 kV). The frequency of the voltages on the rings and plates was varied (usually from 10 to 1000 Hz; square wave voltage with 25 s rise time; 180◦ phase shift between modules). With these conditions one can test for different kinds of ion production, the relevance of symmetry in ion generator module operation and the shape of the AC signal. The variation of the ion production was achieved by connecting ring and plate of each module to the same AC voltage on one hand (squares in Fig. 5) and connecting the rings on ground and the plates on AC on the other hand (circles). For the investigation of symmetry one module was turned OFF electrically: plate, ring and needle on same AC (crosses in Fig. 5). Finally, the influence of the signal shape on the charger performance was studied by replacing the square wave power source by a supply giving sinusoidal voltage signals (triangles). These three series were carried out with two AC amplitude voltages (1 and 4 kV; corresponding to 0.25 and 1 kV/cm, respectively). For better readability, the data are split into Fig. 5a and b. Additionally, one 4 kV series is included in Fig. 5a to allow for easier comparison of the two measurement groups.
20
loss of charged particles lossc / %
100 twin corona module charger 80 DC field
60
UDC
AC field
40 20 0 0
0.5
40 60 80
1
1.5
average charge q / e
no field
intrinsic efficiency εintr / %
loss of charged particles lossc / %
A. Marquard et al. / Aerosol Science 37 (2006) 1069 – 1080
intrinsic efficiency εintr / %
1076
twin corona module charger 80
UDC 60
AC field
40 20
DC field no field
0 0 20
0.5
1
1.5
2
2.5
3
average charge q / e
40 60 80 100
100
(a)
100
(b)
Fig. 6. Twin corona module charger performance comparison without external electric field, with DC and with AC field in the ion-particle mixing zone. Flow regimes (a) aerosol flow 11 l/min, total module air flow 1 l/min (FOP 11/1); (b) aerosol flow 11 l/min, total module air flow 10 l/min (FOP 11/10).
It is obvious from these plots that all data points of both diagrams follow the same curves with regard to losses and efficiency. Obviously, neither the variation of ions production, nor the symmetry of operation, nor the nature of the AC set-up have an influence on the charge loss performance. Presumably, hydrodynamic forces and interactions with charged insulating walls overwhelm the AC-field totally. The next group of experiments compares electrical conditions in the ion-particle mixing chamber (Fig. 6) for two fixed flow conditions FOP (11/1) and FOP (11/10) and different corona voltages, DC fields and AC frequencies. FOP (11/1) represents a very small flow rate through the ion generator modules (u < 5 cm/s) and thus fewer vortices in the mixing zone; FOP (11/10) on the other hand with an average flow velocity in the modules around 0.5 m/s causes more flow distortion by turbulence and is therefore likely to induce higher losses. The cases investigated are charging without external electric field (rings and plates grounded; triangular data points), with DC field (one module OFF with all three electrodes on positive voltage; triangles) and with AC field (symmetric setup with 4 kV square wave voltage, as described above; circles). Note in Fig. 6, how conveniently all these conditions can be represented in one graph. Looking at Figs. 6a and b we again see that the data points follow a rather well-defined common curve. The only significant deviations from that curve are observed on the right sides of the lossc -vs.-q¯ curves corresponding to high DC voltages. At high external DC fields in the charger zone losses are found to increase while the average particle charge is reduced (Fig. 6a) or remains constant (Fig. 6b). This behavior cannot be explained by the stronger E-fields alone, because in both cases, AC and DC, equal maximum field strengths are applied (≈ 1 kV/cm). Presumably, the increased losses at high DC voltages are due to interactions with charged walls (wall charging could be proven by an electrostatic voltmeter, Mod. 541, Trek Inc.) and/or any loss reducing action of the AC compared to the DC starts at this 1 kV/cm and would improve for higher electric field strengths. However, higher AC-voltages could not be adjusted, because the threshold of sparking within the module electrodes was reached. When no external field is applied, on the other hand, average charge and charging efficiency drop to their lowest levels compared to the other operating modes, because there is no electric field to assist ion penetration through the rings and plates. Comparing Fig. 6a and b, again, the AC field offers no benefit compared to the DC field, irrespective of the flow conditions in the mixing chamber (FOP 11/1 vs. FOP 11/10). The impact of different hydrodynamic parameters on the charge-vs.-loss performance is illustrated in more detail in Fig. 7 for a fixed electrical set-up. For this purpose, combinations of two aerosol flows rates (11 and 20 l/min) and two ion generator air flow rates (1 and 10 l/min) were combined to establish different mixing motions and particle residence times together with different ion rates penetrating the rings and plates. The ion generator modules were operated with a square wave voltage of variable frequency on rings and plates.
intrinsic efficiency εintr / %
loss of charged particles lossc / %
A. Marquard et al. / Aerosol Science 37 (2006) 1069 – 1080
1077
100 twin corona module charger 80 60 11/1
40
11/10
20 0 0
0.5
20
1
1.5
average charge q / e
40 20/10 60 20/1 80 100
Fig. 7. Twin corona module charger performance comparison for different flow rate combinations of aerosol and corona module (needle voltages −10 kV, 4 kVAC voltage on the rings and plates, variable frequency).
From Fig. 7 we see that the losses measured at the longest particle residence time (tav ≈ 9.5 s) corresponding to FOP 11/1 are on the highest level with over 20% for q¯ = 0.5 e, while for FOP 11/10 and 20/10 (in both cases tav ≈ 4.8 s and tav ≈ 2.5 s) they lie on a common curve at the lowest loss level. The values of FOP 20/1 (tav ≈ 4.6 s) fall between the other two cases. Thus, losses appear to be residence time dependent, but accentuated by the degree of mixing turbulence at comparable residence times. The influence of the aerosol flow on losses is rather weak within the studied range of flow rates. As expected, an increased aerosol flow at comparable module flow rates (comparing FOPs 11/10 with 20/10 and 11/1 with 20/1) reduces the charge level due to the decreased residence time. Increasing the module flow also reduces the charge level (compare FOPs 11/1 with 11/10 and 20/1 with 20/10). In the wire-tube ESP ion production and mixing of ions with particles take place in the same chamber. This ‘direct corona’ arrangement is generally considered more efficient with respect to ion utilization than the ‘indirect corona’ of the previous chargers. Fig. 8 shows the charge-loss-efficiency plots for different DC and square wave AC corona voltages for the following operating parameter variations: (a) increasing DC voltage on the wire (triangles); (b) variation of frequency at fixed pulse duty factor, df (switching between ground and +10 kV at frequencies between 10 and 1000 Hz and pulse duty factors df of 1:1 and 1:3; diamonds and circles, respectively); (c) variation of pulse duty factors at fixed frequencies (pulse duty factors from 1:1 up to 1:5 for 100 Hz and 1 kHz; squares and crosses, respectively). Varying the pulse duty factor is expected to reduce losses on one hand due to a lower average deposition electric field, but also to reduce the average ion concentration on the other hand (e.g. df = 1:1 and df = 1:4 correspond to corona activities of approximately 50% and 20%, respectively, of the signal period). The aim of this series is to learn whether one of these effects would be dominating. As in the experiment with the twin corona module charger, all data points lie on a common curve, except that the slope of the loss curve is lower than for the twin corona module charger, mainly due to the shorter residence time (here ≈ 0.02 s). It can be concluded from all considered variations of the DC corona and the unipolar AC corona experiments, that we again have a system where the investigated changes of the electric conditions affects both charging and deposition conditions to the same extent. In order to explain this result for the series with varied pulse duty factors one shall consider the underlying time scales. Whereas the considered ion residence times are in the range of a few 100 s, the investigated frequencies correlate to periods down to 1 ms. Therefore, different pulse duty factors act as switches for both ion concentration and electric field for particle deposition; e.g. half the ions generated with df = 1:1 correspond to half the time available for deposition forces to act. 2.4. Analysis of non-uniform charge conditions with efficiency—charge diagrams Comparisons of particle charge levels with diffusion charging theory are often based on the assumption of a uniform charging history for all particles, whereas particle residence times and ion concentrations inside a charger will most likely both be distributed. As a result, theoretically predicted charge distributions for averaged charging conditions
A. Marquard et al. / Aerosol Science 37 (2006) 1069 – 1080 100 tube ESP 80 60 ν = 10-1000 Hz df = 1:3 DC corona
40 20
ν = 10-1000 Hz df = 1:1
0
intrinsic efficiency νintr / %
loss of charged particles lossc / %
1078
0
0.5
1
1.5
2
2.5
20
3
3.5
4
average charge q / e
40 ν = 100 Hz df = 1:1 - 1:5
60
ν = 1 kHz df = 1:1 - 1:4
80
100
100
100
intrinsic charging efficiency εintr / %
intrinsic charging efficiency εintr / %
Fig. 8. Wire-tube ESP charger performance comparison for different wave forms of the unipolar corona voltage; variable frequencies and pulse duty factors df.
Fuchs, t - distributed 80 Fuchs, n.t uniform 60 40 20
twin corona module charger
0
ESP 60 EAN 581
40 20 0
0
(a)
Fuchs, n.t uniform 80
0.5
1
1.5
intrinsic average charge qintr / e
2
0
(b)
0.5
1
1.5
2
intrinsic average charge qintr / e
Fig. 9. Plots of intr vs. q¯intr compared with predictions by the Fuchs model for homogenous n·t-products (solid line) and for non uniform n·t-products with particle residence time distribution (dashed line). (a) Twin corona module charger. (b) EAN 581 (Topas GmbH) and wire tube ESP.
(n·t-product, E-field) will be narrower than those measured under actual non-uniform conditions, and the averages for charge and efficiency will also be different. The real situation is usually too complex to be modeled; even the assumption of independent distributions for particle residence time and ion concentration will not quite resolve the problem, because both distributions are in reality connected via the distribution of particle trajectories. (Furthermore, this would require information from within the charger, which we have for good reasons excluded from consideration in the context of this paper.) On the other hand, one can obtain a measure for the uniformity of a charging process without actual insider information: for any given n·t-product, diffusion charging theory separately predicts an average intrinsic charge and an intrinsic charging efficiency. By plotting intr vs. qintr one can compare these theoretical predictions for uniform n·t-products with experimental data for a real charger, without having to associate the experimental data pairs with specific values of the n·t-product. This approach is taken in Fig. 9 for the three devices already discussed in the preceding sections. Obtaining intr is fairly straightforward (see Part I, Eq. (12a) is identical to Eq. (B6)) and these data were already presented in an earlier section. The intrinsic average charge is not directly accessible and was estimated from the measured exit average charge, assuming that the average charge of the particles lost inside the charger is equal to that of the charged particles measured at the exit (Part I, Eq. (18)). This assumption is reasonable for the two chargers with turbulent flow, but only a rough approximation for the twin corona module charger where mixing strongly depends on the operating mode. The solid curves in Fig. 9 are calculated for n·t-products in the range of 109 .1012 s/m3 according to Fuchs (1963), Natanson (1960) and Adachi, Kousaka, and Okuyama (1985), using an ion mobility value of Z+ = 1.4·104 m2 /Vs and
A. Marquard et al. / Aerosol Science 37 (2006) 1069 – 1080
1079
ion mass of m+ = 109 amu. (The curves for Z− = 1.9·104 m2 /Vs and m− = 50 amu are almost identical and therefore not shown.) Comparisons for the twin corona module charger are shown in Fig. 9a, and for the EAN 581 charger and the directcorona ESP in Fig. 9b. As we see, the efficiencies of the twin corona module charger are significantly offset to the right of the theoretical curve for constant n·t-products (solid line), while the data points of the other two chargers closely follow the theoretical curve. A shift to the right is typical for non-uniform charging conditions, which tend to increase the average charge more strongly than the charging efficiency. Hence the data show that the twin corona module charger is much less uniform than in the other two. For a rough estimate of the inhomogeneity in the twin corona module charger, the intrinsic -q-curve was re-calculated assuming distributed particles residence times in a homogeneous ion atmosphere (dashed line in Fig. 9a is representative for all FOPs investigated, because all these curves lie very close to each other). The particle residence time distribution was estimated from particle pulse measurements with the charger turned off. As we see, the new curve is somewhat offset from the homogeneous Fuchs curve, but nevertheless lies well above the measured data points. This clearly shows that other factors such as the spatial distribution of ions have not yet been accounted for. This is especially pronounced in the low charging range, where the model predicts an average charge around 0.5 e while only 20% of the particles are actually charged; i.e. more than half of the charged particles must be multiply charged. Further analysis of the ion distribution with an ion current density probe could be done (Meyer, Marquard, Poppner, & Sonnenschein, 2005). It is omitted here, since the qualitative result would be trivial and the quantitative result is of minor relevance.
3. Summary and conclusions Five published charger studies in the particle size range of about 10–30 nm were evaluated with respect to loss of charged particles vs. extrinsic charging efficiency. Pairing these two parameters in one compact diagram permits a differentiated performance comparison, which revealed substantial differences with regard to losses at comparable extrinsic efficiencies. Comparisons with respect to losses vs. average particle charge were not possible because this information was usually not given. Loss information had to be derived indirectly for several studies, in some cases from the (justified) assumption of 100% intrinsic efficiency. New performance data are presented for three types of corona chargers of very different design. Parameters varied include internal flow conditions, the ion production arrangement, and the type of electrical field in the mixing zone between ions and particles. The measurements were made at a single particle size of about 60 nm, where diffusion losses are negligible for the given particle residence times. By placing losses, average charge and intrinsic efficiency into a single diagram, particularly compact presentation of the large number of results was obtained in accordance with the framework suggested in Part I. It was found that each charger had a characteristic loss–charge curve, comprising the data points for a broad range of electrical and ion generation conditions at a given flow. Chargers differed mainly with regard to the steepness of these curves, i.e. the compromise between losses and average charge. The strongest influence resulted from the flow conditions inside the charger, with short residence times (and pronounced turbulence) and high ion densities giving the highest charge levels and the lowest losses. Charger designs employing AC electric fields to bring ions into contact with the particles do not generally improve the compromise between charging and particle losses, at least for nanoaerosols, as compared to DC charging. This is supported both by the reanalyzed literature data and the new experiments. There is evidence that the ion wind will dominate particle motion in such a highly mobile aerosol system and overwhelm the theoretically predicted beneficial effects of the AC field and laminar flow on particle motion. An additional sheath air can further reduce electrostatic losses effectively, but at the cost of more or less unwanted aerosol dilution. (There may be other cases such as photoelectric chargers, where an AC field could be more beneficial, because the ions are generated throughout the space. However no such data were found in the literature.) By plotting data pairs q¯intr vs. intr (with q¯intr estimated according to Eq. (18), Part I) and comparing with predictions from charging theory it is further shown in Section 2.4 that one can obtain a measure for the uniformity of a charging process without the need to associate the experimental data with specific values of an n·t-product. This aspect is mainly relevant for nanoaerosols and weakly charged aerosols with larger particles.
1080
A. Marquard et al. / Aerosol Science 37 (2006) 1069 – 1080
References Adachi, M., Kousaka, Y., & Okuyama, K. (1985). Unipolar and bipolar diffusion charging of ultrafine aerosol particles. Journal of Aerosol Science, 16, 109–123. Biskos, G., Reavell, K., & Collings, N. (2005). Unipolar diffusion charging of aerosol particles in the transition regime. Journal of Aerosol Science, 36, 247–265. Büscher, P., Schmidt-Ott, A., & Wiedensohler, A. (1994). Performance of a unipolar “square wave” diffusion charger with variable nt-product. Journal of Aerosol Science, 25, 651–663. Chen, D.-R., & Pui, D. Y. H. (1999). A high efficiency, high throughput unipolar aerosol charger for nanoparticles. Journal of Nanoparticle Research, 1, 115–126. Fuchs, N. A. (1963). On the stationary charge distribution on aerosol particles in a bipolar ionic atmosphere. Geofisica Pura e Applicata, 56, 185–193. Hernandez-Sierra, A., Alguacil, F. J., & Alonso, M. (2003). Unipolar charging of nanometer aerosol particles in a corona ionizer. Journal of Aerosol Science, 34(6), 733–746. Hewitt, G. W. (1957). The charging of small particles for electrostatic precipitation. Transactions of the AIEE, 76, 300–306. Kruis, F. E., & Fissan, H. (2001). Nanoparticle charging in a twin Hewitt charger. Journal of Nanoparticle Research, 3, 39–50. Krupa, A., Jaworek, A., Lackowski, M., Czech, T., & Luckner, J. (2005). Efficiency of particle charging by an alternating electric field charger. Journal of Electrostatics, 63, 673–678. Lackowski, M. (2001). Unipolar charging of aerosol particles in alternating electric field. Journal of Electrostatics, 51–52, 225–231. Marquard, A., Bredin, A., Meyer, J., & Kasper, G. (2004a). Efficiency–loss-relations of unipolar nanoaerosol chargers. In S. Keping, Y. Geifei (Eds.), Recent developments in applied electrostatics (pp. 199–202), Amsterdam: Elsevier. Marquard, A., Bredin, A., Meyer, J., & Kasper, G. (2004b). Unipolar charging of nanoparticles at inhomogeneous N·t-products and charge–lossrelations. Proceedings of the EAC 2004 (pp. S429–S430), Budapest, September 6–10. Marquard, A., Ehouarn, P., & Kasper, G. (2002). Designing a unipolar corona charger with maximum efficiency and minimum losses of nanoparticles. Proceedings of the sixth international aerosol conference (pp. 1163–1164), Taiwan, September 8–13. Marquard, A., Kasper, M., Meyer, J., & Kasper, G. (2005). Nanoparticle charging efficiencies and related charging conditions in a wire-tube ESP at DC energization. Journal of Electrostatics, 63, 693–698. Meyer, J., Marquard, A., Poppner, M., & Sonnenschein, R. (2005). Electric fields coupled with ion space charge. Part 1: Measurements. Journal of Electrostatics, 63, 775–780. Natanson, G. L. (1960). On the theory of the charging of amicroscopic aerosol particles as a result of capture of gas ions. Soviet Physics-Technical Physics, 5, 538–551. Whitby, K. T. (1961). Generator for producing high concentrations of small ions. The Review of Scientific Instruments, 32, 1353–1355.
View publication stats
