Influence of dielectric thickness and electrode structure on the ion wind generation by micro fabricated plasma actuators

Surface dielectric barrier discharges are investigated in order to explore the combined effects of barrier thickness and microstructure of the exposed electrode on the ion wind generation. Actuators with straight and structured high voltage electrodes with characteristic sizes of 200 and 250 µm and dielectric thicknesses of 0.5, 1 and 2 mm are compared. It is observed that: i) actuator efficiency of ion wind generation strongly depends on the applied voltage amplitude; ii) operation voltage depends on the dielectric thickness logarithmically; iii) electrode microstructure slightly increases the dynamic pressure (few percent in maximum), however the effect decreases with thicker dielectrics and smaller electrode structures; iv) the pattern of the most intensive discharge parts as well as the dielectric erosion repeats the regular structure of the electrodes down to 200 µm. Several identical samples are tested during different days to estimate the impact of the air humidity and the degradation of the dielectric. The microscale precision of the sample manufacture was accomplished by a commercial facility for printed circuit boards.


Introduction
The growing demand on passenger aircraft and the limitation of air transportation and finite fossil fuels require fuelsaving technologies. During the last decade, the flow control by non-thermal plasma with asymmetric surface dielectric barrier discharges (SDBDs) was extensively investigated as drag reduction technology for aerospace applications [1][2][3][4][5][6][7]. These Original Content from this work may be used under the terms of the Creative Commons Attribution 4.0 licence. Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI. devices are also called plasma actuators. SDBDs have typically two electrodes fixed on the opposite sites of a dielectric barrier, the so-called top or exposed electrode and the lower or embedded electrode. A discharge initiates at the edge of the exposed electrode and expands over the dielectric surface in the area above the embedded electrode. The SDBD induces an airflow above the surface named as electric or ion wind. Typical applied voltage amplitudes are between 5 and 35 kV and frequencies in the kHz range [1]. The induced flow can prevent a boundary layer separation at aircraft wings and enable higher angles of attack [8]. The use of SDBDs can also damp Tollmien-Schlichting waves [9] and thus, delay the boundary layer transition from laminar to turbulent, which leads to a local reduction of the drag by up to 50% [10]. The discharge operates successfully at reduced pressure [11], e.g. at flight altitude of commercial passenger aircrafts. The boundary layer transition delay was also tested successfully under flight conditions [12]. In addition, SDBDs looks promising for the deicing of aircraft surfaces [13].
The ion wind velocity increases with the amplitude of applied voltage up to a certain voltage threshold at which the velocity grows less significantly [14]. This saturation voltage depends on the voltage waveform, the frequency as well as on geometry of the discharge. The highest measured ion wind velocity for a single SDBD was reported by Benard et al [1] with a value of 7 m s −1 by using a 4 mm thick dielectric and 1 kHz frequency of the applied voltage amplitude of 30 kV. The same authors also reported a maximum ion wind velocity for a multi SDBD arrangement with four SDBDs in a row of 10.5 m s −1 by using a 3 mm thick dielectric and 1 kHz frequency of the applied voltage amplitude of 24 kV. A thick dielectric and a low frequency lead to higher saturation voltages and were recommended in order to reach the maximum thrust by the ion wind [1,14,15]. High voltage amplitudes and large dielectric thicknesses complicate the implementation of the actuators to flow control applications. Therefore, effective generation of significant ion wind at low voltages using thin and thus, flexible dielectric surfaces is a current demand. The optimization of the discharge geometry still provides wide opportunities for further improvements of SDBD actuators [14,16]. Two linear actuators working in opposite direction provide ion wind in the direction perpendicular to the dielectric, which can be enhanced in annular geometry [16]. Stair-wise shaped dielectric [5] and paired electrode design [5,17] can improve actuator efficiency. In the present work we focus on the shape of the high voltage electrode in combination with the dielectric thickness.
Thomas et al [14] showed that the thrust increases for structured electrodes in comparison to straight electrodes. The effect is decreasing with the voltage amplitude, but more than 30% gain of thrust was measured at a voltage amplitude of about 20 kV (40 kV PP ). In the study of Thomas et al an isosceles triangle electrode structure was investigated with a base length of 3.2 mm and a height of 12.7 mm. Debien et al [18] pointed out that the use of a 13 µm wire as exposed electrode increases the thrust by more than 25% in comparison to a 300 µm wire. Abe et al [19] also demonstrated that smaller electrode thickness and a fine structured electrode enhance ion wind generation. Also it was concluded that a microstructured electrode edge, fabricated from thin wire mesh, can increase ion wind thrust in 50% at atmospheric pressure in comparison to a straight tape electrode of the same thickness.
The decrease of the dielectric thickness leads to an increase of the electric field and consequently results in discharge ignition at lower voltage amplitudes. Forte et al showed [15] that at low voltage amplitudes thinner dielectrics result in higher maximum velocities of the ion wind. However, for a significant flow control effect a sufficiently high voltage amplitude is required. This restricts the minimum thickness because the dielectric has to resist against the electrical breakdown.
The role of both parameters-the size of the microstructured electrode and the dielectric thickness-on the ion wind generation as well as their combination are investigated in this study. Actuators with straight and structured (rightangled triangles with base lengths of 200 and 250 µm) electrodes with dielectric thicknesses of 0.2, 0.5, 1 and 2 mm are studied. Attention is given to the reproducibility of the measured effects by the examination of several microfabricated samples of the same design produced by a commercial printed circuit board manufacturer. The size of the investigated structures is significantly smaller than the one of Thomas et al [14]. The structure of rectangular isosceles triangles is characterized by a single parameter, namely, the distance between them. In addition, the manufacturing process for the structured and straight electrodes allows a gradual transition between them in contrast to Abe et al [19]. Indeed, the fabrication of the electrode microstructures by semiconductor fabrication techniques, e.g. for micro-electro-mechanical systems (MEMS) [20,21] would be feasible and would enhance the precision. However, the enhanced effort (e.g. cleanroom facilities, processing time, advanced etching techniques) is not justified and economically unreasonable for the small number of specimen.
The ion wind is characterized by means of measured dynamic pressures. The flow velocity is not used for characterization. It has a square root dependence on the dynamic pressure and therefore, it is less sensitive to the observed effects. In addition, the dynamic pressure behaves proportional to the actuator thrust. Latest studies show a strong degradation of organic materials like the often used polyimides (e.g. Kapton ® ) [22]. Therefore, and considering our own observations, the surface and electrode degradation by the plasma operation was investigated using a scanning electron microscope (SEM).

Experimental setup
Printed circuit boards (PCB) with an area of 80 x 80 mm from IBR Leiterplatten GmbH & Co. KG were used as a base for the SDBDs. The dielectric material is woven fiberglass embedded in resin (FR-4 TG135) with thicknesses of 0.2, 0.5, 1.0 and 2.0 mm. However, the samples with a thickness of 0.2 mm were not sufficiently robust for systematic measurements due to the fast degradation of the dielectric in contact with the plasma. The design of the top (brown) and bottom (blue) electrode of the PCBs are shown in figure 1. The high voltage electrodes (brown) have three different shapes at the connection line to the ground electrode (blue). One shape of the electrode is straight. The others have right-angled triangles with base lengths of 200 and 250 µm as shown in figure 1(b). All electrodes consist of a 35 µm thick copper layer with a chemical gold coating for passivation.
The power supply was a AC source (Chroma, model 61 604) providing a 1 kHz sinus shaped waveform to a custombuilt transformer (Bremer Transformatoren GmbH) supplying a maximum amplitude of up to 10 kV. The discharge was operated in ambient air at atmospheric pressure. The high voltage was measured using a Tektronix P6015A 1:1000 high voltage probe. The charge transferred in the gas gap of the SDBDs was determined by use of a Mica CD19 capacitor (Cornell-Dubilier, C m = 1 nF ± 1% tolerance with), see figure 1(a), connected to an Agilent 10 073D 1:10 probe. The voltage probes were coupled to an oscilloscope (Agilent DSO 7032B), which also averaged the charge-voltage plot of the samples over 100 high voltage cycles. The charge-voltage characteristics were integrated to obtain the dissipated power of the SDBDs. For further details on the technique of power measurements see [23] and references therein.
The ion wind was quantified by measurements of the dynamic pressure by a Pitot tube connected to a MKS Instruments 120AD Baratron, which is a high-accuracy capacitance differential manometer, and a MKS Instruments 651 C control unit. The quartz Pitot tube had an inner diameter of 1 mm and an outer diameter of 2 mm. The inner tube diameter of 1 mm covered at least four triangles of the microstructured electrodes, so that no influence by the structure in the Pitot pressure could be measured along the electrode. The Pitot tube was aligned on the dielectric surface (y = 0 mm). The ion wind above the surface is always influenced by boundary layer effects, e.g. friction. For typically investigated SDBDs friction effects are mainly obtained at wall distances below y = 0.5 mm, e.g. Jolibois and Moreau [24] or Kriegseis et al [25]. The Pitot tube in the present work integrated the ion wind within distance y of 0.5 and 1.5 mm from the dielectric surface and, therefore, should capture its maximum, except in the region at close vicinity to the electrodes.
A micrometer-screw translation stage (OWIS) was used to move the Pitot tube along dielectric surface. The dynamic pressure quickly grows with the distance from the high voltage electrode until a maximum is reached and then gradually decreases. Thus, the measurements of the dynamic pressure behind the maximum are much less sensitive to the positioning of the Pitot tube. Most of the measurements were performed at the position x = 7 mm which is just behind the maximum of the dynamic pressure for all investigated conditions as shown by the coordinate system in figure 1.
The Pitot tube measurements provide the values of the dynamic pressure p d , which is related to the ion wind velocity v by Bernoulli's equation. If the laboratory conditions are assumed as: T = 296 K and p = 1013 mbar, that leads to an ambient air density ρ air ≈ 1.19 [kg·m −3 ] and ion wind velocity can be computed as: The upper pressure limit of the used differential manometer is about 10 Pa, which corresponds to an ion wind velocity of about 4 m s −1 . This was sufficient for most measurements limited by amplitude of the applied voltage of 10 kV (≈ 7.1 kV RMS ). Forte et al [15] reports a maximum measured velocity of 3 m s −1 for a similar setup with 7 kV RMS , 1 kHz and a 1 mm thick PMMA plate as dielectric, which is inside the upper limit of the used devices and also agrees with the data of this study.

Discharge appearance
The appearance of the discharge was photographed with a Canon EOS 77D camera and a Tamron objective with 18-200 mm focal length and 10 s exposure time (vibrational compensation was turned off). Selected parts of the pictures, corresponding to 1x10 mm actuator area are presented in figure  2. At the maximum applied voltage, see figure 2(b), the discharge expands almost 7 mm outwards of the high voltage electrode edge. The most intense parts of the discharge are located in the vicinity of the electrode edge and their shape depends on the electrode structure. The edge of the straight electrode is illuminated homogeneously and it is seen as a bright straight line. The edge of the 200 µm structured electrode contains five bright spots situated along 1 mm electrode length equidistantly. The edge of the 250 µm structured electrode contains four spots with similar brightness. The positions of these spots represent the protruding tips of the triangles of the electrode structure, which have the smallest distance to the ground electrode. The discharge develops towards the electrode. Furthermore, a weak emission is also present above the electrode. In the case of the 250 µm structure the discharge in the opposite direction of the main discharge above the upper electrode is less pronounced.
In figure 2(a) discharge photographs are presented at an applied voltage of 1.8 kV RMS which is slightly above the ignition voltage of about 1.5 kV RMS . Irregular bright spots are present on the edge of the straight electrode indicative of the filamentary nature of the surface discharge. At low voltages, preferred locations of the filaments can be detected with the 10 s exposure time, whereas at high voltage the large amount of erratically distributed filaments leads to the impression of a uniform plasma. The structured electrodes have bright spots at the same positions, but, at low voltages the brightness varies from spot to spot. Probably, the electrode defects have a smaller influence at higher voltages and the discharge spreads more regular.
The SDBD has a filamentary behaviour (streamer mechanism) in general as recognized by short pulses on the current waveforms and by high-speed camera photos, see for example [2,16]. The random appearance distribution of the filaments results in an uniform appearance if the photo exposure time covers about 10 4 high voltage cycles as in figure 2. However, if the applied voltage is too high, the SDBD can turn to the saturation regime, when a further increase of the power does not enhance the ion wind and the discharge will not appear uniform anymore [14]. This saturation regime was never reached in present work since the maximum high voltage was limited to 10 kV (7.1 kV RMS ). This value is below the saturation threshold.

Electrode structure effect for 0.5 mm dielectric
In figure 3 the results of four sets of measured plasma power and dynamic pressure data are shown. Every set contains the data of three individual SDBD samples with a straight, a 200 µm and a 250 µm structured high voltage electrode, respectively. The samples in one set were investigated sequentially within four hours in order to minimize effects of the changing ambient humidity. The high voltage was applied only for a short period of time that was necessary to record the measurements (about a minute for each point). Such minimization of the operation time reduces the degradation of the dielectric. The voltage was increased stepwise from 0 up to 7.2 kV RMS and then it was decreased with the same step size of about 0.7 kV RMS . The voltage of 1.5 kV RMS is just below the discharge ignition voltage for the samples of 0.5 mm thickness. Figure 3(a) elucidates the influence of the electrode structure for every set. In one measured set the consumed power agrees within an experimental error of about 5% for all electrode structures. The dynamic pressure shows a similar trend, but the data for the 250 µm structured electrodes exhibit systematically highest values followed by the 200 µm structured electrodes and the lowest values are found for the straight electrodes.
All measured power values are plotted together in figure  3(b) indicating the day-to-day variation, which is most likely correlated with the ambient air humidity, see inset in figure 3(b). All dynamic pressure values are shown in figure 3(c) underlining that the electrode structure has a slight effect on the ion wind generation while the influence of the humidity is less pronounced, see the inset in figure 3(c).

Dielectric thickness
The data for SDBD samples with different dielectric thicknesses are displayed in figure 4. The measured powers and ion wind induced dynamic pressures for 1 and 2 mm thick PCBs are shown in figure 4(a) as a function of the applied voltage. The structured electrodes show systematically higher values of the dynamic pressure similar to the 0.5 mm samples, but, this feature is decreasing with the thickness of the dielectric. The pressures generated by micro-structured electrodes are systematically higher by a few percent compared to the straight electrodes. However, these values are within the experimental error. To show this effect more clearly, the dynamic pressure is plotted versus the consumed power in figure 4(b). The data points for the 200 µm structured electrodes were fitted by polynomial functions for all thicknesses and only the fits are displayed for better visibility. The data points for the straight electrode are below the curve or coincide with it, whereas the data points for the 250 µm structured electrode are on the curve or above it. This difference is more pronounced for 0.5 mm thick samples in comparison to 2 mm samples. The data for samples with 1 mm thickness are not presented in the figure but also confirm this trend. The absolute values for the power and the dynamic pressure decrease with increasing dielectric thickness for the same applied voltages, see figures 3 and 4(a). The 2 mm thick samples at the maximum applied voltage of 7 kV RMS generate dynamic pressures about 3± 1 Pa, whereas for 1 and 0.5 mm thick samples it requires only 6 and 5 kV RMS respectively. Thus, a thinner dielectric allows operating the SDBD at lower voltages. The direct comparison of the dynamic pressures generated by the 200 µm structured samples with different thickness is shown as a function of voltage in figure 5(c). The idea of the analysis was to compensate the effect of the different dielectric thicknesses on the ion wind generation by a re-scaling of the applied voltage value. It was found that a multiplication of the voltage with a factor of 1, 1.15 and 1.3 for 2, 1 and 0.5 mm thick samples, respectively, merges the measured data to the same function, see figure 5(d). This scaling procedure is described by the following empiric relation between the operation voltages V 0 and V d applied for the generation of the same air flow at SDBDs with thickness d 0 and d respectively: The fact that doubling of the dielectric thickness requires an increase of the operation voltage amplitude of about 15% for the generation of the same dynamic pressure can be related to peculiarities of our set up, in particular the type of dielectric and the electrode design. Thus, for our case, the logarithm was multiplied on constant 1/2. We assume that the logarithmic dependence of the operation voltage on dielectric thickness similar to equation (2) can be applied for other actuators if the constant in front of logarithm will be properly adjusted.

Efficiency of the ion wind generation
The actuator efficiency is estimated by the ratio of the dynamic pressure to the consumed electrical power (p d /P). Figure 5 illustrates the influence of the dielectric thickness on the ratio p d /P for all electrode structures. Here, p d /P is plotted as a function of power (a), voltage (b) and distance from the high voltage electrode (c). In figure 5(a), the ratio p d /P is plotted for the straight and 250 µm structured electrodes. Samples with a thicker dielectric have the advantage of a higher efficiency at     low powers. This agrees with Jolibois and Moreau [24] and with figure 4(b), where a higher value of the dynamic pressure at the same plasma power was obtained. Important to note, that in figure 5(a) the efficiency approaches a saturation above 0.3 W. Since the maximum voltage amplitude was limited in the present experimental setup, the effect could not be validated for higher electrical power consumptions for the thick dielectrics.
For further analysis, the same experimental data are plotted as a function of voltage in figure 5(b). The influence of the ratio p d /P on the dielectric thickness diminishes here in contrast to figure 5(a). This leads to the conclusion that the amplitude of the applied voltage is crucial for an efficient ion wind generation. SDBDs with thicker dielectrics require higher operation voltages, thus, providing a higher efficiency at low power where the actuator effect is low. It seems that the difference in the efficiency caused by dielectric thickness is insignificant when the generation of remarkable airflow (> 1 Pa) is achieved.
The data in figures 5 (a) and (b) were measured at one location of the Pitot tube, namely 7 mm apart from the high voltage electrode and near the dielectric surface. The space resolved Pitot tube measurements for the 200 µm structured samples are plotted in figure 5(c) for one selected power and voltage amplitude. At constant power, the efficiency of the actuator with thicker dielectric is higher for all distances from the high voltage electrode, whereas at constant voltage the efficiency of the SDBDs is similar. The deviation for the 2 mm thick SDBD near the high voltage electrode may be explained by a broader velocity profile of the ion wind over the surface. Probably, for thinner dielectrics the maximum velocity is closer to the surface. Then, the stream lines can be disturbed by the Pitot tube wall of 0.5 mm, which can lead to an underestimation of the maximal dynamic pressure for narrow velocity profiles. The velocity profiles becomes broader with increasing the distance to the high voltage, hence the efficiency curves merge.
For a more reliable comparison of the electrode structure, the efficiency as a function of applied voltage is shown for all discharge geometries in figure 6. The straight electrode provides systematically lower efficiency in comparison to electrodes with 250 µm structure. The efficiency for electrodes with 200 µm structure are always in the middle. The electrode structure has a more significant influence if the SDBD is composed by thin dielectrics.

Surface analysis and material degradation
The material degradation of the SDBDs was investigated using a scanning electron microscope (SEM; instrument: Jeol JSM 7500 F). The SEM employs a field emission gun and a secondary electron in-lens detector enabling the observation of a specimen at a maximum specified resolution of 1.0 nm. Here, the images of the samples were taken with the help of secondary electron imaging (SEI) at 2 kV without conductive coating of the surface. This allows the visualization without artificial effects caused by the sample preparation. A SEI detector taking emitted electrons at a declination angle of 40 degrees has been used. Therefore, the SEM imaging setting is appropriate for 3D samples with a complex perspective in the observation field. In particular, details of electrodes and their interfaces with the dielectric material can be imaged. This observation demonstrates the strong degradation of the organic materials under discharge operation. Removing resin near the exposed electrode can lead to an electrical shortcut. Therefore, it was problematic to use 0.2 mm thick PCB actuators for systematic investigations. Poor resistance of organic materials (polyamides, resin) against aggressive plasma environment is well known [22,26]. However, in the present work the main erosion channels were aligned to the regular pattern of the electrode microstructure. This can be important for fabrication of multifunctional surfaces where an actuator is combined with a riblet structure [21] for aerodynamic drag reduction. Namely, the most intense discharge parts can be placed away from the resin riblet structure to reduce its degradation.

Discussion
The data confirm that the electrode microstructure has a influence on the discharge performance. However, a much larger effect was observed by Thomas et al [14] for much larger structures (3.2 x 12.7 mm isosceles triangle), where even about 40% decrease of the discharge ignition voltage was obtained. The peak-to-peak ignition voltage for straight electrode of 25 kV (17.7 kV RMS ) and for the serrated electrodes of 15 kV (10.6 kV RMS ) were roughly estimated from figure 18 in [14]. This drastic difference between the present work and the work of Thomas et al [14] is most likely due to the geometric differences of the electrode structures.
The increase of the ion wind with the size of the electrode structure observed in the present work together with the fact that millimetre size structures provide a much stronger effect favours the assumption that the size of the structure plays a crucial role. Whether the structure size is large or small depends on the characteristic size of the discharge, which is the distance between the electrodes or the dielectric thickness. The diminishing of the microstructure effect with the dielectric thickness in figures 4(b) and 6 supports this assumption.
Another effecting parameter is the angle of the triangle structure. The work of [14] claims that on the base of empirical observations the optimum angle is 14 • . In the present work, an angle of 90 • is used for the isosceles triangles with a comparable thickness of the electrodes (40 µm in [14] and 35 µm in the present work) However, the angle might influence two parameters, namely the radius of curvature of the electrode edge as well as the ratio between structure length and width. The radius of curvature influences directly the electric field amplitude on the electrode tip. It is about 30 µm in present work, estimated from figure 7 and limited by the manufacturing technology anyhow. The value is comparable with the electrode thickness. The radius of curvature parameter was not given in [14]. The ratio of the structure length and width could also influence the electrode electric field configuration. However, too sharp and too thin structures lead to a concentration of the discharge footprints and consequently to a faster erosion of the electrode material.
The third crucial parameter is the gap between the upper exposed high-voltage electrode and the lower embedded ground electrode. In [14] an overlap of the electrodes about the half-dielectric thickness of 3.18 mm was applied while in the present work, the counterpart edges of both electrodes coincide without any distance.
The actuator of [19] had a comparable geometry to one of our experiment: 1 mm of dielectric thickness (glassfiber-reinforced-epoxy); electrode thickness of 20-40 µm for straight electrodes. The microstructured electrode consisted of 25 µm cutted wires, sticking out about 125 µm from a mesh. The enhancement of the ion wind by the structured electrode was about 50% . Obviously, these thin wires provide much sharper edges in comparison to our experiment in particular, and to the straight electrode of the same thickness. Furthermore, the ratio of its width and length differs significantly.
In the present work about 30% of ion wind enhancement by microstructure implementation can be stated, see figure 3(a), set 1 and set 3, figure 4. However, the scattering and reproducibility of the data points, figures 3(c) and 6, are in the same range. This might be caused by the high spatial resolution of our pressure measurements compared to [14,19] where the thrust was determined over the full width of the electrode.
Furthermore, the degradation of the electrodes and the dielectric (or status of the surface in general) with operation time, which was not considered in [14,19], has to be taking into account. In the present work the high voltage amplitude was first increased and then decreased gradually to obtain two data sets per configuration. For all Pitot tube pressure curves a slight deviation was obtained between increasing and decreasing voltage data sets. Obviously, the degradation of the organic dielectrics, see figures 7 and 8, can affect the overall performance after short operation time. Furthermore, our day-to-day and sample-to-sample variation (see e.g. figure 3(c)) gives a deviation of up to 20%. Therefore, we consider the effects of the electrode structure in [14,19] rather as qualitative then as quantitative.
Comparing the actuator efficiency for the different geometries as a function of applied voltage amplitude, see figure 5, it becomes obvious that larger dielectric thickness has some positive effect on the efficiency. This could be a result of broader velocity profile above the surface [24] and, therefore, the lower wall friction, which consumes up to 30% of the induced momentum close to the wall [25,29]. The increase of the efficiency for the structured electrodes, figure 6, should be attributed to its size as the curvatures of the 200 µm and 250 µm triangles are the same.

Summary
The effects of microstructured exposed electrodes on the power consumption and the ion wind generation of surface DBDs were investigated. The shape of the microstructured electrode and the thickness of the dielectric barrier have been varied systematically. It was shown that the actuator efficiency for ion wind generation grows with the applied voltage. The growth is similar for all investigated samples and a maximum efficiency of about 10 Pa/W was measured at a maximum power consumption per electrode length of about 40 W/m. Thinner dielectrics allow to operate actuators at lower voltages. In particular, it was observed, that a doubling of the dielectric thickness requires an increase of the operation voltage on 15% in order to obtain the same dynamic pressure. A logarithmic dependence of the operation voltage on the dielectric thickness was suggested. The structured electrodes provide an increase of the efficiency for ion wind generation of samples with thin dielectrics of a few percent, but the effect diminishes with larger dielectric thicknesses. A reliable effect of the electrode structure can be achieved if the characteristic length of the structure is comparable with the dielectric thickness. Thus, the investigated 200 and 250 µm structures with 90 • angle does not allow to enhance the ion wind significantly in contrast to the 3.2 × 12.7 mm structure with an angle of 14 • . For this situation an increase of the thrust of more than 30% was reported by Thomas et al [14]. The electrode structure can distribute the most intense discharge parts by varying the distance between grounded and high voltage electrode. Spots with strong emission are identical with the positions of the erosion channels on the dielectric. Scanning electron microscope images indicate separate erosion sites for 200 and 250 µm electrode structures, but not for straight ones.
Commercially produced printed circuit boards provided reliable micro-scale SDBD samples for systematic studies. The use of multiple actuator samples with the identical design indicates that the reproducibility of the results can be significantly affected by the change of humidity in ambient air and the state of the dielectric surface.