The effects of catalyst conductivity and loading of dielectric surface structures on plasma dynamics in patterned dielectric barrier discharges

Dielectric barrier discharges (DBDs) are promising tools for air pollution removal and gas conversion based on excess renewable energy. Catalyst loading of dielectric pellets placed inside the plasma can improve such processes. The effects of such metallic and dielectric catalyst loading on the discharge are investigated experimentally. A patterned DBD is operated in different He/O2 mixtures and driven by a 10kHz pulsed rectangular voltage waveform. Hemispherical dielectric pellets coated by different catalyst materials at different positions on their surface are embedded into the bottom grounded electrode. Based on phase resolved optical emission spectroscopy the effects of different catalyst materials and locations on the streamer dynamics are investigated. The propagation of cathode directed positive volume streamers towards the apex of the hemispheres followed by surface streamers, that move across the structured dielectric, is observed for positive applied voltage pulses. Coating the apex with a conducting catalyst results in attraction of such streamers towards the apex due to charging of this surface, while they avoid the apex in the presence of a dielectric catalyst. Surface streamers, that propagate across the hemispheres, are stalled by conducting catalysts placed on the embedded pellets as rings of different diameters, but propagate more easily across dielectric coatings due to the presence of tangential electric fields. Reversing the polarity of the driving voltage results in the propagation of negative streamers across the patterned dielectric and attenuated effects of catalytic coatings on the streamer dynamics.


Introduction
In the light of climate change, a variety of industrial processes need to be de-carbonized and made more energy efficient.Ideally, such processes can be operated based on excess renewable energy sources, which requires electrification and short switch on/off times.Plasma technology can contribute to this important societal goal in many ways.A prominent example is plasma catalysis [1,2].
Volume and surface dielectric barrier discharges (DBDs) [3][4][5] can be used for volatile organic compound (VOC) removal from exhaust gas streams [3,[6][7][8].VOCs are present in exhaust gas streams in industry, especially in paints and inks [9,10].They are a human health hazard [11,12].Conventional methods to decompose VOCs are based on thermal techniques, which are typically expensive, consume a lot of energy, and cannot be switched on/off quickly [13].DBDs can overcome these limitations [14][15][16][17].Schücke et al [3,18] used a twinsurface DBD to remove VOCs efficiently.Böddecker et al [19] combined multiple of such surface DBDs into arrays to increase the throughput of such pollution remediation processes at high gas flows.
Plasma technology and catalysis can also be used for gas conversion into value-added products.For instance, Kreuznacht et al [20] used atmospheric pressure microwave plasma torches and gliding arcs for the conversion of methane into hydrogen and solid carbon.Another promising application is ammonia synthesis based on plasma catalysis, often realized in DBDs [21].
For many such applications, volume DBDs are filled with dielectric pellets loaded with catalytic materials to form packed-bed plasma reactors (PBPRs).The most common design of a PBPR is a coaxial volume DBD filled with a high number of spherical pellets as shown in figure 1(a).The plasma is generated in the voids between the pellets and on their surfaces, typically in the form of volume and surface streamers.At the contact points of adjacent pellets electric field enhancements were observed [22] and, thus, the plasma is primarily generated at these locations.
Previous investigations have shown high conversion using such a plasma source design experimentally and computationally [23][24][25][26][27][28][29][30][31].Van Laer and Bogaerts [28,29] found an increase of the electric field and electron temperature close to the contact points of such pellets as a function of the dielectric constant of the spheres.Based on 2D fluid simulations using Mark Kushner's nonPDPSIM, Kruszelnicki et al [30] and Babaeva et al [32] studied the fundamental physics of plasma generation between dielectric particles located in the volume of classical DBDs.For such a PBPR setting, they observed the propagation of cathode directed volume streamers in between adjacent pellets.As such positive streamers approach a dielectric pellet, the positively charged streamer head polarizes the pellet.This leads to charging and electric field enhancement at the poles of the pellet.As such a streamer arrives at the pellet surface and depending on its size as well as dielectric constant, it is converted into a surface ionization wave (SIW) that propagates along the pellet's surface and envelops it.Such SIWs propagate based on tangential electric fields that result in electron acceleration towards a strong charge on the pellet's surface, much like a positive streamer moving along the surface.On the other side of the pellet, possibly based on seed electrons generated by photoionization, a new volume streamer is generated that propagates towards the next adjacent pellet based on electric field enhancements at the pellets' poles.
For plasma catalysis, such spherical pellets are typically made of or are coated by catalyst materials to improve conversion and selectivity [1,33].The details of the synergistic effects between plasma and catalyst are not yet understood in many cases.Thus, a knowledge-based development of plasma catalytic processes is often not possible.
The catalyst material loaded onto such pellets can be highly conductive (metallic) or dielectric.Prominent examples of such catalysts are Cu, e.g. for CO 2 conversion [1], and MnO 2 for VOC removal from gas streams [8,[34][35][36][37], respectively.The presence of metal catalysts was demonstrated to result in larger surface areas covered by the plasma and in higher plasma densities [38,39].Based on simulations, Kruszelnicki et al [40,41] revealed an enhancement of the local electric field close to metallic particles loaded onto dielectric surfaces.This was found to be caused by local maxima of the surface charge density at the triple-points, where metal and dielectric surfaces as well as the plasma meet.Such local electric field enhancements were found to result in higher fluxes of reactive species generated in the plasma towards boundary surfaces and in a reduction of the breakdown voltage due to enhanced electron power absorption.These effects can improve the energy efficiency of such plasma processes.In many cases detailed information on the effects of different catalyst materials on the electron dynamics, especially on the generation of energetic electrons, that dissociate the neutral gas and generate process relevant reactive neutrals, are missing.This is particularly true for experimental investigations and largely caused by inherent problems of classical PBPRs for diagnostic access to the electron dynamics.First of all, the packing of the plasma volume with pellets is random and sometimes even unstable, i.e. the pellet arrangement can change during the process.This results in different dimensions of the voids and distances between adjacent pellets and, thus, in different and complicated streamer dynamics.Mujahid et al demonstrated that the streamer dynamics are sensitive to the gap between adjacent pellets, as two streamer heads can repel each other in case of short gaps [42].Overall, there is typically no pulse-to-pulse repeatability of the streamer dynamics for consecutive pulses of the applied voltage waveform so that diagnostics such as phase resolved optical emission spectroscopy (PROES) [43], that require such repeatability for averaging over multiple pulse periods to obtain good statistics, can hardly be used.Moreover, optical diagnostic access is generally limited by the presence of the pellets themselves, since the plasma emission resulting from streamers is blocked.Gaining scientific insight into the effects of the pellet arrangement on the process performance is challenging, since the pellet position cannot be controlled precisely in such plasma sources.Finally, the presence of the pellets in the volume between the electrodes blocks the gas flow, which is detrimental for gas conversion applications.
To overcome these limitations of PBPRs, Mujahid and Oteef developed a modified version of a volume DBD, where dielectric structures are embedded into one of the electrodes at controlled locations, the so-called patterned DBD (p-DBD) [44].The p-DBD design is shown schematically in figure 1(b).Both electrodes are covered by dielectrics, but the presence of structured/patterned surface topologies at one electrode allows precise control of the plasma.By embedding hemispheres into one electrode, as done in this work, volume streamers will propagate between the apex of such surface structures and the flat counter electrodes where the distance between dielectrics is the smallest.Similar to PBPRs these volume streamers will be converted into SIWs upon arrival at a boundary surface.However, in contrast to PBPRs, the location of the pellets (hemispheres) is precisely controlled.This results in pulseto-pulse stability and repeatability of the streamer dynamics which enables the option of using certain plasma diagnostics.For example, in contrast to conventional PBPRs, PROES can be used as a way to study the electron dynamics.Any insights into the effects of surface materials, shape, as well as pellet dimensions and locations can be tested and used easily in such reactors, as the pellet arrangement is controlled.The gas flow and optical diagnostic access to the gap between the apex of the dielectric structures and the counter electrode are not blocked.This reactor design can be easily scaled up to larger dimensions by enlarging the electrode surface areas and embedding more dielectric surface structures for larger scale industrial applications.
In the absence of catalytic coatings, the discharge dynamics in such p-DBDs were investigated previously [44][45][46][47].Figure 2 shows a schematic illustration of the streamer evolution when applying a pulsed rectangular (positive) voltage waveform with a peak-to-peak voltage in the kV-range, a repetition rate of the order of 10 kHz and a duty cycle of about 1% to the top electrode, while the bottom electrode is grounded.After applying the positive potential, residual electrons from previous cycles of the applied voltage waveform will be accelerated towards the top electrode, that corresponds to the anode (figure 2(a)).This results in ionization and an electron avalanche in front of the anode.As the positive ions cannot react to the applied potential as fast as the electrons, they will be left behind and a positive space charge is generated in front of the anode.Electrons will be accelerated from the volume below this positive space charge towards it.This results in the formation of a positive streamer that propagates towards the cathode as energetic electrons generate strong ionization below the positively charged streamer head, that results in the formation of a new region of positive space charge underneath the original streamer head (figure 2(b)).Due to the presence of the dielectric hemisphere at the cathode, the electrode gap is minimal above the apex of this surface structure so that the electric field is maximal above it.This leads to the propagation of the positive streamer from the flat anode towards the apex of the patterned cathode.As the positively charged streamer head approaches the apex of the hemisphere, it polarizes it (figure 2(c)).This results in negative surface charging of the hemisphere and in an electric field enhancement in front of the apex, which results in strong electron acceleration and finally in high ionization as well as plasma emission at this location.Additionally, photoionization was found to have an influence on the streamer propagation [4,48,49].When the positive volume streamer arrives at the apex of the hemispherical dielectric, it is converted to a surface streamer that continues propagating towards the cathode in the form of a ring around the pellet (figure 2(d)).This propagation of the surface streamer is based on tangential electric fields generated by the positively charged streamer head.Simulations have shown that the surface streamer floats above the dielectric at a small distance from it [4].When the positive volume streamer reaches the apex, an anode directed negative streamer can be generated by electrons generated at the apex and accelerated towards the anode [4,46].
Based on these fundamentals and advantages of p-DBDs, we will study the effects of catalytic coatings of the embedded dielectric hemispheres on the streamer dynamics in such a plasma reactor experimentally by PROES.Different catalytic materials, both conducting (Cu) and dielectric (MnO 2 ), will be placed onto the dielectric hemispheres at different locations on this surface.Qualitatively understanding the effects of such variations of the catalyst material and location on the electron dynamics is essential for the knowledge based development of applications of DBDs, such as gas conversion, since energetic electrons largely determine the conversion by dissociating the neutral gas.Based on such fundamental understanding, the discharge could be controlled in a way to beneficially impact the local electron energy distribution function and conversion rates could be optimized by selecting the ideal catalyst material and location.
The manuscript is structured in the following way: section 2 introduces the experimental setup including the diagnostics.Experimental results are presented and discussed in section 3, while the concluding remarks are provided in section 4.

Experimental setup
The experimental setup including all diagnostics is shown in figure 3. Two versions of the setup are shown for PROES measurements from the side to provide a side-view onto the embedded dielectric pellets (figure 3(a)) and from the top for a topview onto the embedded dielectric structures (figure 3(b)).The p-DBD itself consists of two planar electrodes, of which the top electrode is powered and the bottom one is grounded.A reactor chamber is placed in between the electrodes.Two different versions of the top electrode are used: (i) an electrode made of stainless steel (figure 3(a)) and (ii) an electrode that consists of a soda lime glass plate (ε = 7.75 at room temperature) coated by indium tin oxide (ITO, figure 3(b))).While the stainless steel version corresponds to the standard reactor design, it does not provide optical access to the plasma from the top and, thus, does not allow to view the bottom patterned electrode from above.For this purpose it is replaced by a transparent electrode made of soda lime glass and ITO.The bottom electrode is made of stainless steel for both versions of the setup.The reactor placed in between the planar electrodes, the dimensions of the dielectric pellets embedded into its bottom surface and the different locations of catalytic coatings on such pellets are shown in more detail in figure 4. The reactor is rectangular and made of quartz (ε r = 3.75) to ensure good optical access to the plasma.It is x = 100 mm long, y = 10 mm high, and z = 36 mm wide.Two holes with a diameter of 6 mm are located at its side walls for gas inlet and outlet.These holes are centered within the left and right sidewalls and face each other.In this way gas flows through the reactor along the x-axis.Previous work indicated, that discharges generated at adjacent pellets behave similarly, although the gas flow and the carriage of excited species by the gas might not be exactly the same for adjacent pellets [42,47].The planar electrodes are placed directly above and below this reactor covering its top and bottom surfaces in the xz-plane.
The main feature of the p-DBD is the presence of a patterned/structured dielectric at the bottom electrode.The patterned dielectric is located at the bottom of the quartz cell and consists of a flat Macor plate, which is x = 100 mm long, y = 1 mm high, and z = 36 mm wide, and three embedded dielectric pellets.For this work, segments of custom-built Macor hemispheres are used.Macor, with a dielectric constant of ε r = 6.03 at room temperature, is used, as it provides good thermal stability, a smooth surface and easy manufacturing.The hemisphere segments have a height of 6 mm and a diameter of 24 mm and were manufactured from a full hemisphere with a diameter of 30 mm.The dimensions of one pellet are shown as a grey surface in figure 4(b).To fix the pellets in place, small holes are drilled into the flat Macor plate which match the size of a small cylinder (diameter of 4 mm, height of 1 mm) that sticks out of the bottom of each pellet.These cylinders are put into the holes to hold the pellets in position.The gap between adjacent pellets is fixed to 2 mm based on previous work to avoid repulsion of adjacent SIWs, that occurs for smaller gaps [42].Overall, both electrodes are covered by dielectrics, while the one located at the top electrode is flat and the bottom one is patterned.As described in section 1, this ensures a stable volume streamer propagation between the flat top dielectric and the apexes of the patterned dielectric at the bottom electrode as well as a stable propagation of SIWs.
Catalytic materials are loaded onto the patterned dielectric surface at three different locations on the surface of each pellet (3 mg • cm −2 , see figures 4(c) and (d)).Firstly, a dot of 2 mm diameter is placed centrally at the apexes of all pellets.Secondly, a small ring with an inner diameter of 5 mm and a thickness of 0.5 mm is placed onto these surfaces centered around their apexes.Finally, a large ring with an inner diameter of 10 mm and a thickness of 0.5 mm is placed onto each pellet centered around its apex.Both rings include two gaps at opposite sides for simultaneous investigations of surface streamers facing catalysts and gaps under the same conditions.Those gaps have a width of approximately 2 mm.
The top electrode is driven by an approximately rectangular voltage waveform generated by a high voltage (HV) unipolar pulse generator (DEI PVX-4110) in combination with one of two HV DC power supplies (AU-30P40 and AU-30N40) for positive and negative pulses, respectively.These voltage waveforms have a repetition rate of 10 kHz, a duty cycle of 1%, and rise-/fall-times of 5 ns.As an example, a driving voltage waveform with a peak-to-peak value of 4 kV is shown in figure 5.These plots also include the resulting current waveform.The repetition rate is chosen to match previous experiments and simulations of p-DBDs [45,46].
The discharge is operated in helium with a purity of 99.999% and a flow rate of 2 slm (standard liters per minute) at atmospheric pressure with a variable admixture of up to 1% of oxygen (99.998% purity).For the highest reactive gas admixture, the total gas flow will be increased to 2.02 slm.As described in detail in [47], helium gas is used to ensure the presence of a stable discharge, with high pulse-to-pulse repeatability.Moreover, realistic simulations can be performed in this noble gas with small admixtures of O 2 to obtain more detailed insights into the plasma physics [50,51].In more complex gases, more complicated mechanisms play an important role such as vibrational and rotational excitation, electronegativity and photoionization, which are not present in helium.Typically, such plasmas cannot be described accurately by simulations and the understanding of fundamental phenomena such as the streamer dynamics is more challenging.Thus, using a simple chemistry is ideal for the fundamental studies performed in this work.Moreover, the choice of helium gas is consistent with past experiments on patterned DBDs [45,46].Small amounts of oxygen are added to generate reactive oxygen species (ROS) such as O and O 3 , which are relevant for some applications of DBDs.It was shown that especially small concentrations of oxygen lead to the desired ROS concentrations [52].Adding O 2 enhances the quenching of excited states of helium [53] and, thus, shortens their effective lifetime.This, in turn, enhances the temporal resolution of optical diagnostics.In this work, admixing O 2 to He was found to allow a more precise determination of the streamer velocity from optical measurements of the position of maximum plasma emission as a function of time.The presence of O 2 can also enhance the role of photoionization for streamer propagation [54], which can cause volume and surface streamers to propagate faster compared to pure He.
During plasma operation the driving voltage waveform is measured time resolved by a HV probe (Tektronix P6015A) at the top electrode.The current is recorded using a current probe (MagneLab CT-0.5-BNC) on the grounded side.An intensified charge coupled device (ICCD) camera (Stanford Computer Optics 4Picos SR) is synchronized with the measured voltage waveform to do PROES for a side-and top-view of the embedded pellets, respectively (see figure 3).For the present measurements, a recording and accumulation of 2 000 000 periods is used.The camera is combined with a bi-telecentric lens (TC23016) and an optical filter (710 nm with a full width at half maximum of 10 nm) to monitor the helium emission line at 706.5 nm, which results from the following electron transition: The He(3 3 S 1 )-state is primarily populated by electron impact excitation of ground state helium atoms.The electron threshold energy for this excitation process is 22.7 eV.Thus, monitoring this emission line with high spatial and temporal resolution allows monitoring the spatio-temporal dynamics of energetic electrons with energies above 22.7 eV.Such electrons play a crucial role for the dissociation and ionization of the background gas.Correspondingly, such PROES measurements are the basis for knowledge-based plasma process development.Based on a camera gate width of 2 ns and the short lifetime of the excited state, the temporal resolution of the PROES measurements is approximately 2 ns.More detailed information on this diagnostic can be found elsewhere [43,[55][56][57].At each timestep a 2D (x-y) spatially resolved image of the plasma emission at 706.5 nm is taken (see figure 6).The maximum of the emission always appears on the side of the pellet where the line of sight is not blocked by the presence of the pellet itself and, therefore, more emission can be detected.By tracing the change of the position of maximum emission as a function of time within the applied voltage pulse in the 2D spatially resolved plots of the plasma emission, the streamer speed can be determined time resolved.Each of the 2D spatially resolved images of the plasma emission can be binned horizontally in x-direction to obtain horizontally averaged emission vectors, whose rows correspond to spatial resolution in perpendicular (y) direction.Such emission vectors are obtained for each time-step and can be combined to a spatiotemporal emission matrix to yield plots of the emission with temporal resolution along the horizontal and perpendicular spatial resolution along the vertical axis (see figure 6).From the resulting (normalized) spatio-temporal plots of this emission important conclusions on the streamer dynamics can be drawn.The time scales of all PROES plots approximately match the time axis shown in figure 5.This is done by calibrating the time axes of the PROES and current/voltage measurements relative to each other in the following way: For the current/voltage measurements, t = 0 is defined as the time, when the voltage starts to rise.The ICCD camera used for the PROES measurements is triggered on the falling flank of the voltage pulse.An additional delay is then set so that the first 2D spatially resolved image is taken at the beginning of the voltage pulse.This time is called t = 0, similar to the voltage/ current measurements.Then the delay is increased gradually to resolve the streamer dynamics temporally.This additional delay (with respect to t = 0) is indicated in each 2D spatially resolved PROES result, e.g. in figure 7. Instrumental errors and physical effects, however, can lead to inaccuracies of this relative calibration of the time axes of the PROES and current diagnostic.For instance, time shifts can be caused by the triggering of the ICCD camera, since the falling flank has a temporal width of approximately 50 ns which can lead to a corresponding delay in time when comparing different PROES measurements to each other, especially for different applied voltage signals.Therefore, such time scales have to be considered as a time relative to the triggering time of the ICCD camera.Additionally, physical phenomena can lead to changes of the time of streamer generation with respect to the rising flank of the driving voltage waveform.These time shifts can depend on external control parameters.Before streamer generation, electrons are accelerated towards the anode and electron avalanches happen.This results in charge formation and separation between electrons and ions and initiates the formation of a streamer head [4].This charge separation is crucial for the generation of the strong electric field that sustains the propagation of the streamer.However, this process introduces a certain time delay that depends on external control parameters such as the driving voltage and gas mixtures.
The influence of residual He metastables from one period of the driving voltage waveform on the next via Penning ionization can be neglected, since the corresponding reaction time is much shorter than the pulse period.Penning ionization will happen for both gas admixtures (pure He and He/O 2 ).In the case of pure helium, Penning ionization is caused by impurities of the gas itself (e.g.O 2 , H 2 O, N 2 ) and the reactor which is estimated to be in the range of 0.01%.One possible reaction is He * + O 2 → He + O + 2 + e − .The corresponding reaction time is estimated to be much faster than the pulse repetition interval (using a rate constant of 2.5 • 10 −10 cm 3 • s −1 [58]) and, hence, an effect of residual metastables on the following pulse is unlikely.When adding O 2 , Penning ionization will even happen on the same time scale as the ionization of the background gas.Using pure helium, previous simulation results indicated that the spatio-temporally resolved emission of the He(3 3 S 1 )-state is in a reasonable qualitative agreement with the spatio-temporal ionization dynamics of helium [42].
It is important to note that the dynamics of the streamer cannot be determined from the 706.5 nm-line alone.As mentioned, it only provides information on highly energetic electrons and, therefore, conclusions on the dynamics of all electrons cannot be drawn.However, in the following, the term 'streamer' is still used to describe the dynamics of the highly energetic electrons caused by the emission resulting from electron de-excitation from the He(3 3 S 1 )-state as it resembles the dynamics of energetic electrons in actual streamers in DBDs.
In this work, the peak-to-peak voltage of the driving voltage waveform is varied from 4 kV to 6 kV and the catalyst conductivity (conductive or dielectric), as well as the location of the catalyst on the dielectric pellets, are changed systematically according to figures 4(c) and (d).A minimum peakto-peak voltage of 4 kV is used to ensure the generation of efficient plasma breakdown and streamer generation.Voltages higher than 6 kV are not used, since the speed of the streamers increases as a function of voltage and cannot be traced accurately by PROES at higher voltages due to the limited temporal resolution.Most measurements are performed for positive rectangular voltage waveforms to study the propagation of positive streamers towards and across the patterned dielectric.To study the same phenomenon for negative streamers, ) was chosen as an insulating catalyst.While the copperbased coating is simply sprayed onto the pellet using an aluminum mask, the coating process of MnO 2 is based on a more advanced procedure described in detail by Peters et al [36].
Placing the catalyst at the apex of the embedded pellets as a dot primarily affects the volume streamer, when it approaches the pellet.In contrast to this, ring shaped catalysts placed on the pellets according to figures 4(c) and (d) interact with SIWs propagating across the pellet.Using all three catalyst variants (no catalyst, conductive catalyst and dielectric catalyst), a qualitative statement on how the volume and surface streamer behavior is affected by the catalyst conductivity can be made.

Results
In this section the interaction of volume and surface streamers propagating towards and across patterned dielectrics will be discussed in the presence of conducting (Cu) and dielectric (MnO 2 ) catalysts located at different positions on the structured surface.Hemispheres embedded into the grounded electrode of a p-DBD are used and the results are obtained experimentally by PROES.While three adjacent hemispheres are used in the experiment (see figure 4(a)), only results for the central pellet will be discussed.The section is divided into three parts corresponding to (i) volume streamers approaching the patterned dielectric and (ii) surface streamers propagating across the patterned dielectric in the presence of positive rectangular driving voltage waveforms.In the third subsection, (iii) the volume and streamer dynamics for a reversed (negative) polarity of the driving voltage waveform are discussed.

Positive volume streamers approaching patterned dielectrics loaded with catalysts
Figure 7 shows 2D spatially resolved ICCD camera images of the plasma emission at 706.5 nm at different times within the period of the applied rectangular and positive driving voltage waveform.The peak-to-peak voltage is 4 kV and the plasma is operated in He gas.The images are taken from the side in the presence of blank pellets (a)-(c), pellets with a Cu dot at the apex (d)-(f) and with a MnO 2 dot at the apex (g)-(i).A positive and cathode directed volume streamer is observed as it propagates from the planar anode at the top towards the patterned dielectric at the cathode, which is located at the bottom.As the position of the maximum measured emission is located on the cathode side of the positively charged streamer head, it indicates the instantaneous position of the streamer head.The plots also indicate that the streamers for all three conditions reach the pellet 290 ns after the positive flank of the voltage waveform was applied (c,f,i).Based on the time of maximum current (see figure 5), we conclude that the streamer is generated approximately 200 ns after the start of the voltage pulse.
Different streamer dynamics are observed for different catalytic coatings of the apex, i.e. the catalyst material is found to affect the electron dynamics.In case of a blank Macor pellet, the strongly charged positive streamer head polarizes the apex of the embedded pellet, which the streamer targets during propagation through the volume.Further, this results in a local electric field enhancement at the tip of the pellet so that strong electron power absorption and maximum plasma emission are observed close to the apex.Upon arrival of the volume streamer (figure 7(c)), the region of high emission spreads out horizontally over a distance of about 4 mm (2 mm to each side of the apex).This is caused by the fact that the apex region of the hemisphere is relatively flat so that the approaching volume streamer polarizes a relatively large surface area efficiently.
In the presence of a Cu dot at the apex that acts as a floating metal electrode, the approaching volume streamer is focused to a smaller area at the apex with a radius of only 1 mm, that corresponds to the area covered by the conducting catalyst (figure 7(f)).According to simulation results of Kruszelnicki et al [40,41], this is caused by the presence of triple points at the edge of the Cu dot, where this conductive material meets the more dielectric material of the pellet (Macor) and the plasma.At such triple points, electric field enhancements due to geometrical effects and charging of the metal surface induced by the approaching volume streamer are most pronounced, so that strong ionization and emission happen close to these positions.Such electric field enhancements can also lead to field emission of electrons at the corresponding locations on the surface.This can result in an additional local increase of the ionization and emission [41].For conducting materials the electric field at the surface is purely perpendicular to the surface and no tangential fields are present.Thus, these effects lead to a focusing and acceleration of the approaching volume streamer towards the conducting catalyst located at the pellet's apex.This results in an increase of the streamer speed, while it propagates in the volume above and along the surface of the dielectric, compared to scenarios where the apex is dielectric (Macor or MnO 2 dots).This is shown in figure 8 for two different peak-to-peak driving voltages of 4 kV and 6 kV.These results are obtained by tracing the position of maximum emission as a function of time based on the PROES measurements.To obtain more accurate results, 1% of O 2 is admixed, since this results in higher time resolution due to the shortening of the effective lifetime of the excited He-state monitored by PROES by collisional quenching with oxygen.Similar results as shown in figure 7 for He gas were also obtained with 1% O 2 admixture.For consistency with the results discussed in the next section, we show 2D plots of the emission only for He in this section.At the beginning of the positive voltage pulse, figure 8 clearly shows a higher streamer speed for both peak-to-peak voltages in the presence of the Cu dot as compared to the blank Macor pellet and the MnO 2 dot at the apex.The streamer speed is generally increased as a function of the driving voltage so that the volume streamer arrives at the apex faster at 6 kV compared to 4 kV due to higher plasma densities and electric fields at the streamer head.Once it arrived at the apex, it continues to propagate towards the cathode across the pellet's surface as a SIW.The speed of the surface streamer decreases as a function of time as the SIW moves from the apex (thick dielectric) to the cathode (thin dielectric).This is caused by the increase of the capacitance of the dielectric underneath the surface streamer as it moves from the apex to the bottom of the pellet due to the changing dielectric thickness.Therefore, the time required to charge the dielectric locally is increased as the SIW approaches the bottom of the pellet and, thus, the streamer speed decreases.It is worth mentioning that the speed oscillations, which occur for 4 kV applied voltage, lie within the error bars of the speed measurement.
In the presence of a catalyst with a high dielectric constant at the apex of the embedded pellet (MnO 2 ), the volume streamer dynamics change significantly, although the surface area covered by MnO 2 is identical to the one covered by Cu in the previous scenario.As shown in figure 7(g), the surface area covered by high plasma emission is larger compared to the previous cases.Now a circular area with a radius of more than 3 mm is covered by strong plasma emission.In fact, the approaching volume streamer branches upon arrival at the apex.One part propagates towards the center of the apex similar to the blank pellet and the Cu dot, while a second part propagates more to the side indicated by the increased emission observed at positions further to the right along the x-direction before its arrival at the pellet surface.10 ns later (figure 7(h)) the plasma emission caused by the central streamer branch is attenuated, while the emission caused by the other streamer branch remains strong.These observations are explained as follows: Similar to the blank pellet and the Cu dot, the dynamics of the central streamer branch is determined by the polarization of the MnO 2 dot.As MnO 2 is a porous material with a large surface area and a high dielectric constant, the capacitance of the coated pellet at the apex is large and a lot of charge is lost to it by the approaching streamer to charge it.Thus, the central streamer branch extinguishes quickly.The propagation of the second streamer branch to the side is caused by the strong positive space charge located at the MnO 2 dot at the apex center.In contrast to the metal dot, this leads to the generation of tangential electric field components at the surface that cause the generation of a second streamer branch that propagates to the side.Those tangential electric field components can only be generated in this way above dielectric surface materials.Unlike conducting metal surfaces the electric field at the surface does not have to be fully perpendicular to the surface, but can contain parallel/tangential components.This effect is described in detail in the work of Kruszelnicki et al [40,41] and was also confirmed by models for this reactor [45].In case of MnO 2 at the apex, this effect is more pronounced compared to the blank Macor pellet due to the higher dielectric constant (and, thus, stronger polarization) and the larger surface area of MnO 2 .
Overall, these effects lead to a propagation of the streamer to the side for dielectric catalysts located at the apex, while it is focused towards a conductive catalyst dot.These findings are important since they show that the plasma-surface interactions might be enhanced for conducting catalysts compared to dielectrics in case of volume streamers approaching catalytic surfaces.For the Cu as well as the MnO 2 dot at the apex, the surface streamer arrives at the bottom of the patterned dielectric, i.e. at the contact point of two adjacent embedded pellets, earlier compared to the blank hemisphere at 4 kV applied voltage.This is the reason why data on the streamer speed is only shown until 360 ns for MnO 2 and until 370 ns for Cu in figure 8(a), which correspond to the respective arrival times of the surface streamers at the contact point.For the Cu-dot this is caused by the initial acceleration of the volume streamer towards the conducting catalyst, which results in a higher speed of the surface streamer, when it reaches the apex.In case of MnO 2 this effect is explained by the branching of the volume streamer into two components as discussed previously.One component propagates to the side, while the streamer is still in the volume, and, therefore, crosses a shorter distance until it arrives at the contact point so that it arrives earlier compared to the blank pellet, although its speed is similar.

Positive surface streamers propagating across patterned dielectrics loaded with catalysts
After the positive volume streamer has arrived at the apex of the pellet embedded into the bottom grounded electrode, it continues propagating downwards towards the cathode along its surface.In contrast to placing a catalyst at the apex as a dot, where it interacts primarily with the incoming volume streamer (see section 3.1), it can now be placed onto the surface of the hemispheres as a ring of variable radius and centered around the apex.Figures 9 and 10 show 2D spatially resolved PROES results at different times within the pulse of the applied positive and rectangular voltage waveform for a peak-to-peak voltage of 4 kV in He and He with 1% admixture of O 2 , respectively, in the presence of a small Cu ring on the dielectric surface with an inner diameter of 5 mm and a thickness of 0.5 mm (represented by the dotted black lines).As indicated in the figures, the ring has a small gap, where no catalyst is present.Results are shown for a side-view and a top-view onto the central of three adjacent pellets.Similar findings are obtained for rings with larger radii (not shown).2D-spatially resolved ICCD camera images of the plasma emission at 706.5 nm at different times within the period of the applied rectangular and positive driving voltage waveform for a peak-to-peak voltage of 4 kV in He gas and in the presence of a small Cu ring on the dielectric surface with an inner diameter of 5 mm and a thickness of 0.5 mm (indicated by the dotted black lines).The ring has a small gap, where no catalyst is present.The left column shows results obtained for a side-view onto the pellets, while the right column shows results from the top-view.
The surface streamer is initially accelerated towards the Cu ring due to the presence of triple points and enhanced electric fields at the streamer-facing side of the conducting ring.This is shown in figure 11(a) based on the local maximum of the streamer speed at 290 ns, which corresponds to the time, when the SIW approaches the ring (see also figure 10(d)).A similar effect is observed for a ring made of MnO 2 , as will be discussed later, but not for blank Macor pellets.This effect vanishes at higher applied voltages of 6 kV according to figure 11(b), presumably because of an increase of the plasma density and electric field at the streamer head, which lead to a higher streamer speed and less influence of enhanced electric fields at the catalyst ring on its propagation.
In pure He, figure 9 shows that the surface streamer is stalled, when it reaches the Cu ring, while it continues propagating through the gap included in the ring (see figures 9(e) and (f)).This is caused by the high conductivity of Cu and the absence of tangential electric fields at its surface, that could ensure a propagation of the SIW across the ring.Similar phenomena were observed computationally by Kruszelnicki [40].Our experiments validate these computational findings.At the position of the gap, where no conducting, but a dielectric Macor surface is present, the electric field is not purely perpendicular to the surface and tangential electric fields are present.Therefore, the SIW continues to propagate along the surface of the pellet towards the cathode.In order to cross the Cu ring, the streamer needs to find a way to seed electrons to the other side of the ring, where tangential electric fields are present again at the surface and can accelerate such electrons to ensure the generation of a new SIW.One such way might be streamer propagation through the volume above the ring, where tangential fields can be generated.This is, however, inefficient, since the positive streamer needs to propagate upwards in the direction of the anode to cross the ring.Correspondingly, little plasma emission is observed above the ring in He gas.Nevertheless, a weak generation of a new SIW on the other side of the Cu ring is observed in figure 9(f).
According to the results shown in figure 10, admixing 1% of O 2 allows the SIW to propagate across the Cu ring more efficiently.This could be caused by oxidation of the Cu surface caused by the admixture of O 2 so that the surface behaves more like a dielectric and tangential electric fields are present that drive the streamer across the catalyst ring.The observed effect of adding O 2 might also be caused by more efficient photoionization due to the presence of O 2 via ionization of oxygen by photons emitted by excited He atoms and/or higher plasma densities that lead to higher electric fields at the streamer head, which could facilitate streamer propagation through the volume above the ring.Such photoionization might result in the seeding of electrons above and on the opposite side of the Cu ring by the approaching surface streamer.Such seed electrons could provide the basis for a continued propagation of the SIW through the volume above and along the surface on the other side of the Cu ring driven by tangential electric fields as observed at the top right of figure 10(f).Generally, admixing O 2 is found to result in earlier breakdown and faster streamer propagation.
Figure 12 shows 2D spatially resolved measurements of the plasma emission at 706.5 nm at different times within the applied voltage pulse in the presence of a small ring made of MnO 2 at 4 kV peak-to-peak voltage of the positive pulse applied to the top electrode in He gas.Similar to the Cu ring, also this ring contains a small gap as indicated in the figure .In contrast to the Cu ring, where the streamer cannot propagate over the catalyst due to the absence of tangential electric fields, the SIW is now able to cross the dielectric ring and to continue propagating towards the bottom electrode at high emission intensity.This is caused by the presence of tangential electric fields at the ring's surface that allow the SIW to propagate more easily across a ring made of a dielectric compared to a conducting material.Nevertheless, the emission intensity decreases above the MnO 2 ring, while it remains high at the gap of the ring, where the SIW does not face a change of surface material along its path.This is assumed to be caused by the polarization of the MnO 2 ring, which has a higher dielectric constant than the pellet material (Macor), by the approaching surface streamer.Similar to floating dielectric pellets in volume DBDs [30,32], electric field enhancements due to polarization are expected to occur on both sides of the dielectric ring, but not above it.Moreover, significant charge is lost to the ring due to its high capacitance and large surface area as a consequence of its porous structure.
Overall the results presented in this section show that different catalytic materials and their location on dielectric pellets can have different effects on the propagation of surface streamers.While conductive materials tend to stall approaching SIWs due to the absence of tangential electric field components, they can cross dielectric catalysts more easily because, unlike for metallic catalysts, there are tangential electric fields at the surface of dielectric materials.Gaps included in catalytic coatings can facilitate the continued propagation of SIWs through such holes.As active plasma regions such as SIWs lead to the generation of active neutral species that interact with catalytic coatings, such insights can be of key importance for the performance of plasma catalytic processes.

Streamer dynamics for negative polarity of the driving voltage waveform
The results presented in the previous sections were obtained for a positive polarity of the rectangular driving voltage waveform.In this case, the cathode is located at the bottom patterned electrode and the anode corresponds to the flat top electrode.A positive streamer was observed to propagate from the flat anode downwards through the volume towards the apex of the structured dielectric, where it is converted into a SIW that continues propagating towards the cathode across the patterned dielectric surface.This is also illustrated by the spatiotemporal plot of the plasma emission at 706.5 nm shown in figure 13(a) for 6 kV peak-to-peak voltage in He gas with 1% admixture of O 2 in the presence of a large Cu ring on the pellet's surface as indicated by the horizontal dashed black lines.When the surface streamer reaches this ring, it is stalled and the emission is decreased in the area covered by the ring as discussed in section 3.2 for the smaller Cu ring.
Figure 13(b) shows results for the reversed (negative) polarity of the driving voltage waveform under otherwise identical conditions.Now, the cathode corresponds to the flat electrode at the top, and the propagation of a cathode-directed positive volume streamer from the apex of the hemisphere at the bottom electrode towards the flat electrode at the top is observed.When it reaches the flat dielectric, it continues propagating along this as a SIW.In parallel, an anode-directed negative streamer propagates downwards across the structured dielectric towards the bottom electrode.Significant emission caused by such a negative streamer is only observed at relatively high voltage [4,[61][62][63].At even higher voltages the streamers propagate so quickly that their dynamics cannot be resolved temporally by the PROES diagnostic.To resolve the motion of the negative streamer well, when it interacts with the Cu ring, the large ring is used, since it is located close to the bottom electrode, where the structured dielectric is thin and the streamer, thus, propagates slowly enough. Figure 13(b) shows that the negative streamer is not stalled and there is no decrease of plasma emission, when it reaches the catalyst ring in contrast to the positive streamer in case of the positive polarity of the driving voltage.This different behavior of positive and negative streamers upon interaction with the Cu catalyst might be related to the role of photoionization.While the propagation of positive streamers can require photoionization to seed electrons in front of the streamer head, this is less important for negative streamers [4].This might allow negative streamers to cross the Cu ring more easily.
Such different behavior of positive and negative streamers in interactions with catalysts can have important consequences for plasma catalysis, since it is expected to affect the flux of neutral radicals generated within active plasma regions towards the catalyst surface.

Conclusions
Based on PROES the interaction of volume and surface streamers with conducting (Cu) and dielectric (MnO 2 ) catalytic surfaces located at different positions along the streamer path was investigated in a p-DBD.Depending on the catalytic material and its location different streamer dynamics were found with potentially important consequences for plasma catalysis due to its effect on the fluxes of radicals generated in active plasma regions towards catalyst surfaces.
The p-DBD reactor design allows to control the streamer dynamics and ensures the propagation of cathode directed positive volume streamers towards the apex of hemispherical pellets embedded in the bottom electrode for positive rectangular driving voltage waveform applied to the top planar electrode in helium gas with O 2 admixture.Upon arrival at the apex the volume streamer is converted into a positive surface streamer that continues propagating downwards towards cathode.Placing a conducting Cu dot at the apex results in focusing and acceleration of the approaching volume streamer towards this dot due to the formation of triple points where the electric field is enhanced locally and due to the absence of tangential electric fields at conducting surfaces.If such a dot of identical surface area is made of MnO 2 as an example of a dielectric catalyst material, no such focusing and acceleration is observed.Instead the incoming volume streamer branches into two parts, one of which propagates towards the apex and the other one propagates to the side due to the presence of tangential electric fields at the dielectric catalyst surface, which is charged strongly due to its high dielectric constant and large (porous) surface area.
If the catalyst is placed on the embedded pellets as a ring centered around the apex, it is found to interact with surface streamers propagating across this surface.When a positive surface streamer faces a conducting Cu ring, the streamer is stalled and can hardly propagate across it due to the absence of tangential electric fields at such surfaces.This is different for dielectric rings made of MnO 2 , for which positive surface streamers can hop over such rings more easily due to the presence of tangential fields.At low voltages positive surface streamers are found to be accelerated towards such rings due to charging of their surfaces as a consequence of the formation of triple points and polarization.In both cases, Cu and MnO 2 , positive surface streamers can propagate more easily through gaps included in the catalyst rings.
In case of reversed (negative) polarity of the driving voltage waveform, negative surface streamers propagate across the patterned dielectric and are found to propagate across conducting (Cu) rings more easily as compared to positive streamers.This might be related to the less pronounced role of photoionization for the propagation of negative as compared to positive streamers.
The obtained fundamental insights into the effects of catalyst conductivity and location on the dynamics of volume and surface streamers in DBDs are expected to be important for applications in the field of plasma catalysis.Based on these findings catalytic coatings can be designed so that their interactions with species generated in the active plasma (streamer) regions is In this way a variety of applications could potentially be improved, e.g.gas conversion and VOC removal from exhaust gas streams.

Figure 1 .
Figure 1.Schematic of a classical packed bed DBD (a) and a patterned DBD (b) used for gas cleaning.Purple regions indicate the presence of plasma.While the classical reactor includes randomly arranged pellets in the volume, such pellets are embedded into boundary surfaces at controlled locations in the case of the patterned DBD.In both scenarios, the pellets can be loaded with catalysts.

Figure 2 .
Figure 2. Qualitative illustration of positive streamer dynamics in a patterned DBD for positive voltage pulses applied to the flat top electrode (anode), while the patterned bottom electrode is grounded (cathode).Grey and yellow indicate dielectric surfaces.

Figure 3 .
Figure 3. Experimental setup used for PROES measurements from the side (side-view of the embedded dielectric pellets, (a)) and from the top through a transparent top electrode (top-view onto the embedded dielectric pellets, (b)).

Figure 4 .
Figure 4. Patterned DBD reactor design (a), dimensions of a single embedded dielectric pellet (b), and top/side view of embedded pellets loaded with catalytic coatings at different locations (c) and (d).

Figure 5 .
Figure 5. Exemplary driving voltage applied to the top electrode of the p-DBD and current waveform for a peak-to-peak voltage of 4 kV.

Figure 6 .
Figure 6.Schematic illustration of the procedure used to generate spatio-temporal plots of the measured plasma emission at 706.5 nm with spatial resolution perpendicular to the planar electrodes.The curved black line in the left plot marks the surface of the embedded pellet at the bottom electrode, while the horizontal black line in the right plot indicates the position of the apex of the embedded pellet.

Figure 7 .
Figure 7. 2D-spatially resolved ICCD camera images of the plasma emission at 706.5 nm at different times within the period of the applied rectangular and positive driving voltage waveform for a peak-to-peak voltage of 4 kV in He gas.The images are taken from the side in the presence of blank pellets (a)-(c), pellets with a Cu dot at the apex (d)-(f) and with a MnO 2 dot at the apex (g)-(i).The black line indicates the surface of the pellet.The catalytic coatings are indicated by black rectangles located at the apex.

Figure 8 .
Figure 8. streamer speed as a function of time in the presence of blank hemispheres (Macor) and hemispheres with a Cu as well as a MnO 2 catalyst at the apex in the form of a dot.The results are obtained for a rectangular and positive driving voltage waveform with peak-to-peak voltages of 4 kV (a) and 6 kV (b) in He with 1% admixture of O 2 .

Figure 9 .
Figure 9.2D-spatially resolved ICCD camera images of the plasma emission at 706.5 nm at different times within the period of the applied rectangular and positive driving voltage waveform for a peak-to-peak voltage of 4 kV in He gas and in the presence of a small Cu ring on the dielectric surface with an inner diameter of 5 mm and a thickness of 0.5 mm (indicated by the dotted black lines).The ring has a small gap, where no catalyst is present.The left column shows results obtained for a side-view onto the pellets, while the right column shows results from the top-view.

Figure 10 .
Figure10.resolved ICCD camera images of the plasma emission at 706.5 nm at different times within the period of the applied rectangular and positive driving voltage waveform for a peak-to-peak voltage of 4 kV in He gas with 1% admixture of O 2 and in the presence of a small Cu ring on the dielectric surface with an inner diameter of 5 mm and a thickness of 0.5 mm (indicated by the dotted black lines).The ring has a small gap, where no catalyst is present.The left column shows results obtained for a side-view onto the pellets, while the right column shows results from the top-view.

Figure 11 .
Figure 11.Surface streamer speed as a function of time in the presence of blank hemispheres (Macor) and hemispheres with a small ring made of Cu as well as MnO 2 , respectively.The results are obtained for a rectangular and positive driving voltage waveform with peak-to-peak voltages of 4 kV (a) and 6 kV (b) in He with 1% admixture of O 2 .

Figure 12 .
Figure12.resolved ICCD camera images of the plasma emission at 706.5 nm at different times within the period of the applied rectangular and positive driving voltage waveform for a peak-to-peak voltage of 4 kV in He gas and in the presence of a small MnO 2 ring on the dielectric surface with an inner diameter of 5 mm and a thickness of 0.5 mm (indicated by the dotted black lines).The ring has a small gap, where no catalyst is present.The left column shows results obtained for a side-view onto the pellets, while the right column shows results from the top-view.

Figure 13 .
Figure 13.Spatio-temporal plots of the measured plasma emission at 706.5 nm in the presence of a positive (a) and negative (b) rectangular driving voltage waveform with a peak-to-peak voltage of 6 kV in He gas with 1% admixture of O 2 and in the presence of a large Cu ring on the dielectric surface with an inner diameter of 10 mm and a thickness of 0.5 mm.The horizontal solid black line indicates the position of the apex of the embedded hemispherical pellet, while the dashed horizontal black lines indicate the position of the catalyst ring.
[59]erature, at which the coating is made.Assuming that the original material composition contains 50% dimethyl ether, the final coating will, therefore, contain approximately 50% of Cu and is a conductor.In contrast to this, MnO 2 (ε r ≈ 68.1[59]