Powder aerosol deposition method — novel applications in the field of sensing and energy technology

The Aerosol Deposition (AD) method is a dry spray coating process for the production of dense ceramic coatings at room temperature directly from the ceramic raw powder. In order to avoid confusion with liquid aerosol technology, the term powder aerosol deposition (PAD) is introduced here, to highlight that the aerosol consists only of ceramic powder and carrier gas. Especially in the field of functional ceramics, PAD is a promising alternative to conventional sinter-based production processes. This review focuses on the PAD of functional ceramics in the field of sensing and energy technology. In this context, especially current developments and trends are presented. On the part of the sensors, gas and temperature sensors are especially considered, whereas in the field of energy technology, the focus is on vibration energy harvesting, thermoelectric generators, superconductors, and solar cells as well as on all solid-state batteries and fuel cells. Besides the different applications of PAD films, this review also highlights opportunities for influencing the film properties by the used powder or the process parameters.


Introduction
High temperatures, usually above 1000 ○ C, are typically required to produce high-quality ceramic components or films with sufficiently high density. 1,2 Ceramics, especially functional ceramics, which decompose at low temperatures or undergo a phase transformation, can only be produced with considerable effort. In particular, problems such as porosity or incomplete intergranular contact have to be handled. Also, a joining of ceramic films to low-melting substrate materials such as glasses, metals, or polymers is usually not possible. Novel processes are necessary to meet the increased requirements of the ceramics industry in terms of flexibility, integration capability in electronic circuits, and/or miniaturization.
A promising alternative to the established, sintering-based ceramic processes is the aerosol deposition (AD) method, sometimes also referred to as vacuum cold spraying (VCS) or vacuum kinetic spraying (VKS). In order to avoid confusion with liquid aerosol technology, the term "powder aerosol deposition" (PAD) is introduced here. It should highlight that the aerosol consists only of ceramic powders and an appropriate dry carrier gas. The powder AD or PAD is an innovative spray coating process, in which dense ceramic films are produced at room temperature directly from ceramic raw powders. [1][2][3][4][5] Thereby, single powders as well as powder mixtures of ceramic/ceramic, [6][7][8] ceramic/metal, 9 and ceramic/ polymer 10,11 (known as aerosol co-deposition, AcD) can be used. 1 Furthermore, the processing of reactive powders (e.g. moisture, oxygen) is possible. Powder aerosol deposited films typically have film thicknesses from the sub-micrometer to several hundred micrometer range and are characterized by a high density, low surface roughness, an excellent substrate adhesion, and a nanocrystalline film structure. 1,3,4 A summary of the unique features of the powder AD in terms of process and film characteristics is shown in Fig. 1.
The film characteristics can differ considerably, which is due to the controllability of the film characteristics by the used powder and the process parameters.
A general prerequisite for successful PAD of mechanically stable films is the optimal correlation between process parameters, powder properties, and substrate characteristics. Due to intensive research in the past years, it is now possible to deposit a large number of ceramic powders on various substrate materials (ceramics, 12,13 metals, 1,14 polymers, [15][16][17]. Besides familiar ceramics for the AD such as Al 2 O 3 , 5,18-20 PZT, 2,17,21 BaTiO 3 , [22][23][24] and TiO 2 , 25 further functional ceramics, [26][27][28][29][30][31][32][33][34] and partly exotic materials, such as lunar regolith, 35 can be deposited today. In addition, more and more powder mixtures are being used. 6,7,[36][37][38][39] Due to the large variety of ceramic raw powders, powder mixtures and substrate materials, powder aerosol deposited films show diverse properties and thus are used or suggested for many applications these days. In this overview, different applications of powder aerosol deposited film are presented with special focus on the field of sensing and energy technology.

Setup of a PAD Device and Process Principle
A PAD device typically consists of three main components: a deposition chamber, a vacuum pump and a unit to generate the powder aerosol. 1 Figure 2 shows a schematic setup of a PAD device. A carrier gas (e.g. O 2 , N 2 , or He), 40 is fed into the powder aerosol generation unit. In the powder aerosol generation unit, an aerosol is generated by mixing the carrier gas with the ceramic raw powder using various techniques (e.g. fluid bed or brush generator). 29,41 At the same time, a constant pressure difference between the aerosol generation unit and the deposition chamber is generated by a vacuum pump. 3 As a result, the generated aerosol is transported into the deposition chamber and is accelerated to a velocity of 100-600 m/s via a nozzle. 1 In the deposition chamber, the particles collide with the substrate and form a firmly adhering ceramic film.
The mechanism of film formation has not yet been conclusively clarified and is only briefly summarized here (for details, see e.g. Hanft et al. 1 ). It is currently assumed that the film formation takes place in two stages: formation of an anchoring layer and subsequent film buildup. 1,3 The anchoring layer is formed by the first impinging particles deforming the substrate surface and increasing substrate roughness. However, this effect has only been observed on substrate materials such as metals, glasses or polymers, but not on harder ceramic substrates (e.g. Al 2 O 3 ). 12 The subsequent film formation takes place via fracturing and plastic deformation of the impinging particles, known as Room   Temperature Impact Consolidation (RTIC), proposed by Akedo. 2,3,16 This assumption is supplemented by Lee et al. 14 and Exner et al. 4 Lee et al. attribute the main mechanism of film formation to a kind of hammering effect where the film is densified by subsequent colliding particles. 14 Exner et al. postulate the fracturing of crystallites and the generation of fresh and unsaturated surfaces as a prerequisite for successful film formation. 4 For a better understanding of the deposition mechanism, it is also important to know the exact flow conditions during deposition. Accordingly, several studies are currently dealing with flow simulations in the AD process. [42][43][44][45][46] Successfully deposited films feature an excellent substrate adhesion, a crack-and pore-free film structure as well as properties that are similar to those of classical bulk ceramics. However, the identical functional properties of the bulk ceramics (electric, piezoelectric, dielectric, magnetic) are usually not achieved. Therefore, thermal post-treatment of the films often is conducted. During this process, the film stress decreases, the crystallite size increases or the nanoporosity diminishes, so that bulk properties can be achieved.

Novel Applications
Based on the large number of oxide and non-oxidic ceramics that have been deposited by means of PAD so far, the number of possible applications of PAD films increases rapidly. Some of these current application trends in the field of sensing and energy technology are presented in Secs. 3.1 and 3.2.

Gas sensors
Powder aerosol deposited films were already used as the main functional component in a variety of gas sensing applications. Up to the year 2015, mainly studies on PADbased conductometric gas sensors were published. Here, the electrical resistance of the powder aerosol deposited functional film depends directly on the concentration of an analyte in the surrounding gas atmosphere. Examples are oxygen sensing films with a temperature independent characteristics like tantalum-doped barium ferrate (BaTi 0:7 Ta 0:3 O 3Àδ , BFT) 47 and heavily iron-doped strontium titanate (SrTi 0:7 Fe 0:3 O 3Àδ , STF), 48 as well as composite films of STF and Al 2 O 3 . 7 When BFT films were operated at a lowered temperature, also hydrocarbons could be analyzed. 47 Hsiao et al. demonstrated that ZnO PAD films could be utilized to detect carbon monoxide in the 100-1000 ppm range, however an annealing procedure by laser radiation was necessary. 49 All mentioned sensor application were already summarized in the general overview about PAD from 2015. 1 In recent years, new sensor technologies emerged in combination with powder aerosol deposited films. While some applications still utilize the conductometric working principle, also capacitive and thermoelectric properties were introduced as gas sensor parameters. A new dynamic measurement method based on alternating voltage pulses (pulse-polarization) was demonstrated for yttria stabilized zirconia films. The ongoing progress in this field is summarized and discussed in this section.
A novel application type for PAD films are humidity sensors. Kumar et al. and Liang et al. demonstrated that the capacitance of hygroscopic BaTiO 3 films is highly sensitive to moisture in the air atmosphere. 50,51 Kumar et al. investigated also the influence of the electrode geometry and film post-treatment on the sensitivity toward humidity while retaining a constant film thickness of 1.5 μm. 50 Four different electrode setups were used: three conventional interdigital electrodes (IDE) with varying electrode line widths and one setup with a spiral design. For untreated films, the spiral structure exhibited the highest sensitivity and low hysteresis. A thermal annealing at 500 ○ C significantly increased sensitivity of the BaTiO 3 films up to a factor of 80 on all tested electrode setups, possibly due to a slightly increased crystallite size. When annealed, films on conventional IDE with a line width and spacing of 100 μm become favorable in terms of highest sensitivity and low hysteresis. It was concluded that these films are promising for the application as gas sensors or in the bio-medical field in addition to their humidity sensing properties. Liang et al. intensively studied the influence of the film thickness as well as of both microstructure and surface morphology on the resulting humidity sensitivity and response time of BaTiO 3 AD films. 51 Furthermore, the sensor behavior was described using a comprehensive model also enabling the prediction of the resulting sensitivity. The film morphology is described as a specific transitional-density structure where the density gradually changes throughout different layers of the film. Here, the density at the bottom of the film close to the substrate is highest and decreases continuously to the top layer (film surface), as observed by TEM (Fig. 3).
This change in the microstructure was linked to the hammering effect, where subsequently impacting particles consolidate the already deposited film. Consequently, subjacent layers of the film closer to the substrate received intensified hammering with a denser grain arrangement while the top section of the film experienced fewer impacts, featuring more and larger voids as well as larger grains. Films were applied as well on IDE and subsequently annealed at 400 ○ C for 2 h. Within the investigated thickness range of 0.1-10 μm, the BaTiO 3 films with 1.5 μm and 3 μm showed the highest humidity sensitivities (calculated as quotient of capacitance change and relative humidity change within Powder aerosol deposition methodnovel applications in the field of sensing and energy technology the complete measured range) of 179 (pF/%RH) and 300 (pF/%RH), respectively. All other films outside this narrow range showed significantly smaller sensitivities of 0.1-25 (pF/%RH). Response times were also indirectly dependent on the film thickness, since the (rms) surface roughness and the open-pore ratio are affected. Thicker films with higher surface roughness and increased open-pore ratios exhibited longer response and recovery times, starting from a couple of seconds up to times > 1400 s. Taking all characteristics into account, the 1.5 μm thick film featured high, commercially usable sensitivities in combination with short response and recovery times of 5 s and smaller. The mesoporous microstructure promoted an improved moisture absorption and desorption combined with enhanced detection limits.
In a different approach, n-type semiconducting CuO on flexible polyimide (PI) substrates was characterized in respect to its resistive humidity sensing properties. 52 Sensor preparation included the formation of a 200 nm thin (metallic) copper film by nanoparticle deposition (modification of PAD), followed by a low temperature annealing at 250 ○ C in air in order to oxidize the copper to cupric oxide (CuO). In dry conditions, oxygen molecules absorb to the film surface and thereby trap electrons thus increasing the electrical resistance. In humid atmospheres, oxygen bound on the surface is substituted by water molecules, associated with the release of trapped electrons and decrease of the resulting electrical resistance. Produced films exhibited semitransparent properties and a high bending capability. Even after 1000 bending procedures with a bending radius of 0.5 cm, the CuO film is still free of delamination, yet some microcracking is observed. Independent of the bending state, a stable change in electrical resistance of 16-18% occurs when switching between dry ($ 2% RH) and humid (83% RH) conditions. A response and recovery time of 17.8 s and 5.5 s, respectively, was measured. The authors conclude that this type of sensor could be used as a low cost, metal oxide flexible sensor for various display or window applications. It should be noted that using classical ceramics processing technologies, such CuO layers cannot be deposited, since PI decomposes during firing.
Another new research field are planar capacitive soot sensors based on electrically insulating alumina films, as investigated by Hagen et al. 53,54 PAD alumina films exhibit a high electrical resistivity even at elevated temperatures 5,18,19 and are therefore ideally suited for capacitive measurements. Soot is produced in almost all combustions processes, i.e. diesel or gasoline internal combustion engines, and has to be continuously monitored. Dense alumina films were applied directly onto platinum IDE by PAD. Thus, the platinum electrodes are fully embedded without an electrical contact to the surrounding atmosphere. The sensor is additionally equipped with a platinum heater on the reverse side. When electrically conducting soot attaches the surface of the alumina film, the electrical field line distribution changes, leading to an increase in the measured capacitance C from 10 pF up to 50 pF. Therefore, C is a suitable measure to monitor the soot loading of the sensor. This sensor is operated in a cycle consisting of two modes: (1) soot loading until a certain amount of soot is attached (a certain value of C is reached) and (2) thermal regeneration at 600 ○ C to remove the soot and restart the cycle. These sensors were tested with real exhaust gases produced by a diesel engine in an engine test bench. During the 1 st mode, the soot deposition on the alumina surface is boosted by an applied voltage between 15 V and 180 V at the IDE, leading to shorter blind times until the first change in C occurs. Higher soot concentration in the tested exhaust gases resulted in a steeper incline of the capacitance in accordance to the FEM simulations. This indicates that powder aerosol deposited alumina films are suitable for an application as soot sensor even in harsh automotive environments. An advantage of this capacitive soot sensor over conductometric sensors may be a reduced influence of the operating temperature which is of high importance for high dynamic engine operation with widely differing exhaust gas temperatures.
Conventional gas sensors, used to detect and quantify various gaseous components like harmful carbon monoxide or nitrous gases as well as hydrogen or oxygen, were also realized using powder aerosol deposited functional ceramic films. Hanft et al. used thin PAD films (1 μm in thickness) of semiconducting tin oxide (SnO 2 ) in the conductometric operation mode at temperatures in the range between 100 ○ C and 500 ○ C to analyze reducing gases like NO, NO 2 , CO, and H 2 with concentrations up to 1000 ppm. 55 In presence of CO and H 2 , the electrical conductivity increased by two decades, showing an easy-to-measure sensor signal. Responses to NO and NO 2 are smaller, yet measurable. While at lower sensor operation temperatures below 300 ○ C, the sensitivity toward NO x was predominant, it diminished above 350 ○ C. However, CO and H 2 responses peaked between 400 ○ C and 450 ○ C. This behavior is beneficial to adjust the sensor to measure especially the intended gases with small cross-sensitivities toward other gaseous components, just by optimizing the operating temperature.
Combined resistive and thermoelectric oxygen sensors with almost temperature-independent characteristics in the range of 600-800 ○ C were realized by Bektas et al. 27,56 Films of Al-and Fe-doped barium tantalate (BaFe x Al 0:01 Ta 0:99Àx O 3Àδ , BFATx) with x ¼ 0:1 and x ¼ 0:3 were tested at oxygen partial pressures pO 2 between 10 À20 bar (reducing atmosphere) and 10 0 bar (pure oxygen). BFT exhibited a transition from n-type behavior at low pO 2 to p-type conduction at high pO 2 . BFT30 (BaFe 0:7 Ta 0:3 O 3Àδ ) was found to be a good candidate as oxygen gas sensor, since its resistivity as well as its thermoelectric Seebeck coefficient exhibited a strong dependency on the oxygen partial pressure, especially between 10 À5 bar and 10 0 bar, while being nearly independent on the operating temperature. The used measurement setup of this initial study was restricted in terms of the measuring frequency. As a result, response times of several minutes occurred that are still too long for many practical applications.
Besides these conventional approaches with a resistive or capacitive sensor mechanism, also highly dynamic methods could be adapted to PAD films. 26 An example is the nitrogen oxide (NO x ) sensor based on the pulse-polarization method. This technique is usually applied to commercially available thimble-type lambda probes and can be used to detect small concentrations of NO x in the ppm range. The mechanism involves alternating polarization pulses of an oxide ion conducting membrane (like 8YSZ) equipped with platinum electrodes, interrupted by self-discharge pauses. The amount of NO x within the surrounding atmosphere significantly affects the course (rate) of self-discharge, even when only sub-ppm concentration are present. Exner et al. used a miniaturized sensor with a planar setup consisting of an aerosol deposited 8YSZ film on an alumina substrate and screen printed and sintered platinum IDE, considerably smaller than thimble-type lambda probes. A sensor temperature of 425 ○ C enabled the quantitate detection of NO, NO 2 , and mixtures of both for small concentrations of 3-48 ppm. High sensor output voltages of 50-300 mV with small noise were observed, that could be easily measured. When the 8YSZ films were produced by screen-printing instead of PAD, the sensitivity toward NO x reduces and the noise increases, again underlining the great potential of PAD to produce functional ceramic films. Another interesting finding is the decrease in sensitivity toward NO when the sensor is operated in a unidirectional pulse mode instead of alternating pulses. Since the sensitivity of NO 2 is not affected, a mixed operation mode periodically switching between alternating and unidirectional pulses was suggested. Here, a simultaneous determination of the NO and NO 2 may become possible.
PAD offers a high variety and potential to produce functional sensor components, as described by the prior examples. The high integrity and adhesion of aerosol deposited films in combination with near bulk-like functional properties provides benefits for ceramic gas sensors.

Temperature sensors
Also in the field of ceramic temperature sensors based on negative temperature coefficient thermistor materials, PAD is a promising alternative to conventional sinter-based manufacturing processes. 1 Thus both Ryu et al. 57,58 and Baek et al. 59 showed that it is possible to deposit a spinel-based NiMn 2 O 4 raw powder into dense films (no pores or cracks) with excellent substrate adhesion. The films are characterized by a nanocrystalline film structure and the typical NTCR behavior, i.e. they show an almost exponential decrease of the electrical resistance R with increasing temperature. It was also noticed that the characteristic NTCR parameters (specific resistance at 25 ○ C 25 and thermistor constant B) are slightly higher than those of conventionally sintered components but can be reduced by subsequent film tempering. In addition, Ryu et al. 60 showed that the temperature sensitivity of the PAD-NTCR film could be improved by doping the NiMn 2 O 4 powder with Co or both Co and Fe. In this way, mechanically stable PAD films with a B value above 5600 K could be produced. A further modification of the NTCR Powder aerosol deposition methodnovel applications in the field of sensing and energy technology 1930005-5 characteristic was possible by depositing a NiMn 2 O 4 powder with dispersed conductive LaNiO 3 particles, as described in Kang et al. 61 Besides the dense film structure and the excellent substrate adhesion, the PAD films also showed a higher temperature sensitivity (B-value).
In the past few years, the development of ceramic NTC thermistors has focused primarily on two approaches. First, the existing approach of PAD of a completely processed powder has been further developed. Jung et al. showed that the PAD of perovskite-based YBa 2 Cu 3 O 6þx powders to mechanically stable films with typical NTCR characteristics is also possible. 62 Schubert et al. carried out further investigations on the PAD of spinel-based NiMn 2 O 4 ceramics. 30,63,64 Besides the development of chip-based components (see Fig. 4), 63,64 these authors were focusing on tempering 30,63 and aging 63 of NTCR films. They attributed the known tempering effect to a reduction of microstrain and crystallite growth and showed a significant improvement in aging stability. Films tempered at 400 ○ C exhibit higher aging stability than conventionally sintered NiMn 2 O 4 bulk ceramics. 63 In the second approach, a thermally untreated powder mixture of the starting oxides is deposited and the deposited films are then thermally treated (in situ calcination 38 ). It was shown that mechanically stable films with cubic spinel structure and typical NTCR characteristics can be produced by powder aerosol co-deposition of a NiO-Mn 2 O 3 powder mixture followed by an in situ calcination of the composite films. 38 The authors also developed a novel process, in which the approach of powder aerosol co-deposition with subsequent in situ calcination is combined with the sintering of screen-printed electrodes. 39 The advantage of this approach is only one single temperature treatment at 850 ○ C is required to produce mechanically stable NTCR sensors with the same 25 À and B-values as classically sintered bulk ceramics. Due to the possibility to produce fully dense and mechanically stable films with typical NTCR characteristics by PAD or by AcD with subsequent in situ calcination, both approaches are very interesting for a possible utilization in NTC thermistor applications. Especially the absence of the complex sintering step above 1000 ○ C simplifies the process (no problems with porosity or material decomposition [65][66][67] ) and saves energy.

Energy technology
In addition to the sensor field, applications of PAD films in energy technology are currently in vogue. In the following, some new findings in the fields of vibration energy harvesting, thermoelectric generators (TEGs), superconductors, solar cells as well as batteries and fuel cells are overviewed.

Vibration energy harvesting and actuation
The unique feature to produce high-density ceramic films at room temperature also motivates the application of the PAD method for piezoelectric materials. 2,68 Due to its superior piezoelectric properties, especially Pb(Zr,Ti)O 3 (PZT) has been object under investigation. 68 The requirement of large grain sizes (> 1 μm) for high piezoelectric activity is, however, in strong contrast to the typical nanocrystalline and amorphous microstructure of PAD films. 1,69 Therefore, deposited films are usually heat-treated to promote grain growth. 68,70,71 Conventional annealing limits the selection of substrate materials to chemical inert materials with a high thermal stability. 71 Approaches to overcome this challenge include laser annealing or separation of the deposited films from the original substrate. Palneedi et al. used a continuous-wave ytterbium fiber laser (560 nm, approximately 50 μm effective beam size) to locally anneal PZT-films on heat sensitive amorphous Metglas foils (FeBSi) by raster scanning. 72,73 Hwang et al. described a technique called inorganic-based laser lift-off (ILLO) to separate PAD films from their original substrates. 74 Deposition and conventional annealing in a furnace have been conducted on the high melting sapphire substrate. Subsequently the film has been separated by irradiating the reverse side of the sapphire substrate with a XeCl excimer laser (308 nm, photon energy of 4.03 eV). Because of the wide band gap of sapphire (10 eV), the energy initially couples in the interface between PZT and sapphire and locally vaporizes the PZT material (band gap energy 3.3 eV). [75][76][77] After separation, the PZT film has been attached to a flexible polyethylene terephthalate (PET) substrate.
Akedo et al. emphasize the main advantages of piezoelectric systems for vibration energy harvesting. They include low cost, compact design, and the potential to realize wearable structures. 2,78 Recent works propose the benefits of piezoelectric PAD films for energy harvesting. Especially the utilization for wireless sensing applications seems to be promising. 21 Vibration energy harvesting devices based on cantilever beams (unimorph or bimorph structure) have been reported. 2,21 Here, an oscillating weight (seismic mass) on the tip of a cantilever beam deforms a piezoelectric layer periodically in order to generate an output voltage. For example, a maximum power output of 423 μW at an excitation frequency of 143.4 Hz and excitation acceleration of 1.5 g for a bimorph structure (9 Â 6 Â 0:09 mm 3 ) has been achieved by Kuo et al. 21 On the contrary, the composite structure from Hwang targets lower frequencies. 79 In a long-term test encompassing 115,000 bending cycles an open-circuit voltage of 75 V and short-circuit currents of 14 μA at a frequency of 0.4 Hz have been achieved.
Actuators are another field of application for powder aerosol deposited piezoelectric layers. Two possible applications were mainly investigated in scientific papers. These are on the one hand optical microscanners 2,80-82 and on the other hand piezoelectric micromachined ultrasonic transducers (pMUT). 2,83 Regarding the optical microscanners, two different substrate materials were used: silicon wafer and stainless steel. On the Si wafer, a 6 μm thick PZT film was deposited and annealed at 600 ○ C for 1 h. With this device, a scanning angle of 25.9 ○ was achieved, using a driving voltage of 60 V and a resonance frequency of 33 kHz. 80 The scanners made on stainless steel could reach similar values. Using PAD-PZT films with a thickness between 5 μm and 20 μm (also annealed at 600 ○ C for 1 h), a scanning angle up to 41 ○ was possible, with a driving voltage of only 60 V and a resonance frequency at roughly 20-30 kHz. 81,82 By adjusting the mirror size and hinge length, the scanning frequency and angle could easily be tailored.
Another reported application field of a piezoelectric actuator device is a pMUT. 83 For this ultrasonic transducer, an 8 μm thick PZT film was deposited via Granule Spray in Vacuum (GSV) on a Pt/Ti/SiO 2 /Si substrate and structured by subsequent processes to produce a membrane with a diameter of 400 μm. GSV is a derivative of PAD using granulated powder in order to allow stable powder feeding since granules show better flow properties than the primary powder particles. However, at GSV higher speeds are required, as part of the kinetic energy is needed for breaking up the granules before the actual RTIC mechanism occurs. Again, the sample had to be heat-treated at 600 ○ C to obtain sufficient piezoelectric properties. A displacement of 0.8 μm was achieved by a driving voltage of only 5 V at a resonance frequency of 300 kHz. By utilizing the GSV for this transducer setup, the most time-consuming step of piezolayer deposition was accomplished within minutes and represents a significant improvement over conventional processes.

Thermoelectric Generators (TEGs)
The utilization of TEGs can help to convert excess heat from different processes into electrical energy. TEGs are usually assembled from bulk materials. The drawbacks of these type of TEGs are low productivity of the bulk process and the high volume of the devices. Film manufacturing techniques show low deposition rates. They need high temperatures and/or the control of the chemical composition is difficult. 84 P-type Bi 0:5 Sb 1:5 Te 3 was deposited having a film thickness of 276 μm and n-type Bi 2 Te 2:7 Se 0:3 with 557 μm. The films showed power factors of 0.62 mW/mK 2 and 0.21 mW/mK 2 , respectively, which were comparable to the bulk properties. 84 The assembly of a TEG by PAD has also been demonstrated. 85,86 Here, p-type bismuth antimony telluride is used. Due to its nanocrystallinity, the thermal conductivity is 20-30% reduced compared to bulk samples. 86 As a low thermal conductivity is favorable for a more efficient power generation, this is a benefit of the PAD method. It is reported that with this material 200 μm thick coatings can even be formed on PET. 85 The maximum output power reported in this publication is about 54 μW.
Materials to prepare thermoelectric sensors are also mentioned in the section sensing. As an example, the delafossite copper iron oxide (CuFeO 2 ) owns a high Seebeck coefficient and is therefore also a promising thermoelectric material. 87 Depending on the oxygen partial pressure this material can be used in the p-or n-type state. The RTIC process can reduce the grain size, leading to a lower thermal conductivity which has a positive impact on the thermoelectric performance. Nevertheless, the power factor is lower compared to bulk samples. 33 The reason is yet unclear.

Superconductors
The possibility to produce superconducting coatings, especially of magnesium diboride (MgB 2 ), via PAD has already been mentioned several years ago by Akedo et al. 88 Since then, several studies investigating this material class have been published. 29,[89][90][91][92] State-of-the-Art to prepare films, bands, or wires of MgB 2 are the so-called powder-in-tubetechnique magnesium diffusion method and the conventional physical vapor deposition. 90 Both techniques need either a well-aligned atmosphere or high temperatures. A successful production of long tapes has not been reported up to now. As indicated by Arndt, 93 the PAD method may overcome this due to its capability as a reel-to-reel manufacturing process. Since MgB 2 is nearly isentropic and has a large coherence length, 94 it is thought to be appropriate to be used as a polycrystalline PAD-film. A critical current density of J C % 5 kA/cm 2 (at 4.2 K, 0 T) was achieved by Kauffmann-Weiss et al. 90 They also report a critical temperature T C0 of 18.1 K for these films. 90 The average thickness of MgB 2 films can be in the range of 70-100 μm. Due to stresses in the film and the high amount of amorphous regions in PAD films, the transition temperature is reduced and the transition Powder aerosol deposition methodnovel applications in the field of sensing and energy technology width is increased compared to bulk samples prepared from the same precursor powder. 91,92 A heat treatment of MgB 2 PAD-films leads to a better intergrain connectivity and the T C is enhanced as well as the transition width is reduced.

Solar cells
Depending on the optoelectronic material system, many processes for solar cell fabrication suffer from low deposition rates, coupling of material synthesis and film formation and/ or from poor suitability for large area production. For instance, an established alternative to silicon-based solar cells are copper indium gallium selenide (CIGS) solar cells, which are often co-evaporated or sputtered onto a heated substrate. 95 Characteristic for the process is the need for a high vacuum (very low pressures) and slow deposition rates making the production expensive.
Park et al. simulated the deposition of CuInGaSe 2 -particles onto Mo-substrates to optimize the deposition process for this material system. 96 Their results suggest a rather low particle velocity of maximum 200 m/s to avoid substrate damage that would lead to a poor electrical connection to the back contact. Their simulations also indicate local temperatures of 1000 K which would allow local sintering. In a following study, Cu(In, Ga)(S,Se) 2 solar cells made by PAD with a conversion efficiency of up to 5.5% are demonstrated for the first time. 97 However, they had to use various post-treatment steps to remove oxygen incorporated by spraying with dry air and to increase film quality and crystal size. An SEM image of an improved CIGSSe-film for solar cell fabrication is shown in Fig. 5.
The emerging lead-halide perovskite solar cells with the archetypical methylammonium lead (II) iodide (CH 3 NH 3 PbI 3 , MAPbI) are usually processed on solution basis, for instance by spin coating or by co-evaporation. 98 Both techniques have in common that they couple perovskite synthesis and film formation. This makes film formation hard to control and complicates fine-tuning of film properties. In contrast to that, the PAD method allows for decoupling material synthesis from film formation by first manufacturing and optimizing a hybrid perovskite powder, e.g. mechanochemically in a ball mill, 99 which can then serve as the powder for the PAD process. Therefore, Panzer et al. deposited MAPbI onto TiO 2 coated glass substrates. The perovskite structure and optical properties of MAPbI are transferred from the powder to the film paving the way for a solvent-free synthesis route. 32 Before the rise of perovskite solar cells PAD of dye sensitized solar cells components (DSSC) were an intensively researched field. For that purpose, the PAD of TiO 2 as an electron extraction layer has been extensively studied. An overview on the deposition of TiO 2 for DSSC can be found in Hanft et al. 1

Batteries
Applications for the PAD process can be found in every field, where its unique features (see Fig. 1) provide advantages compared to the current state of the art in process technology. This situation can be found in the field of battery production. The future demand for storage capacity driven by electromobility will exceed the current capacities many times over. The present research on lithium batteries aims on finding new ways to increase energy and power density of battery cells and stacks, which seems to require a rigorous change in materials and process technology. Several studies predict a "solid future" for the battery 100,101 since all-solid-state battery (ASSB) architectures might enable new anode (lithium metal) and cathode materials (high voltage material and high capacity materials) with high energy density by replacing liquid electrolytes with solid electrolytes (SEs), e.g. ceramic lithium ion conductors. Different film forming process technologies such as tape casting, screen-printing and extrusion are worth to be considered for fabrication of such cells. 102,103 Promising concepts on ASSBs see lithium metal (approximately 3860 mAh/g) as potential anode material. Apart from lithium metal, different kinds of anode materials such as TiNb 2 O 7 (TNO), 104 Li 4 Ti 5 O 12 (LTO) 105 were deposited by PAD and tested concerning their cycling behavior. However, especially when using lithium metal as anode material, the dense structure of the SE is necessary in order to prevent from dendrite shortening of the cell. Ceramic or glass-like Li þ conductors can be used to provide a high ionic conductivity as SE and to guard the anode from a composite cathode of e.g. NCM (Fig. 6).

1930005-8
Considering the described ASSB architecture, the AD process might have certain tremendous advantages for all different cell components. In the following, we will discuss some ideas of PAD application.
To achieve high energy densities, both mass and volume of passive components like e.g. the SE, have to be kept small. This means light materials are favorable. 106 In general, materials with high ionic and low electronic conductivities are relevant. The aim is to achieve homogeneous, dense, and flawless membranes in the μm-thickness-range with sufficient mechanical stability at the same time. A variety of SE materials was already shown to be processable as dense and μm-thick films by PAD, e.g. phosphates as Li 1:7 Al 0:3 Ti 1:7 (PO 4 ) 3 (LATP or LAGP), 107-109 titanates (Li 0:35 La 0:55 TiO 3 , LLT), 110 or garnet-type structures (Li 7 La 3 Zr 2 O 12 , LLZO). 31,111,112 Although the crystal phases of the deposited materials remain unchanged, usually severe lattice distortion and nanocrystallinity (they stem from the RTIC mechanism-related particle fracture) reduce the ionic conductivity. By a thermal post-processing, the original electrical bulk properties may be recovered. 31,108 Compared to conventional sinter processes, the much lower recovery temperatures are beneficial. This is a general advantage of the PAD process. 3 Switching from common liquid Li þ -ion cells to ASSB has major consequences on the cathode side. Deposition of cathode films of several μm thickness was shown for LiMn 2 O 4 , 113 LiNi x Co y Mn z O 2 , 114,115 LiFePO 112;116;117 4 and LiV 3 O 8 , 118 although, considering high energy cells, the thickness of the cathode layer has to be increased. 106 However, it becomes even more complicated, since besides the storage component, solid ion-and electron-conducting components are necessary. All components need to be dispersed homogeneously in order to guarantee high charge and discharge current densities (Fig. 6). Several examples proved that mixtures can be deposited with homogeneous dispersion and intimate contact in general. 119 This could provide additional advantages, since there is no need for binder additives to achieve stable composite films.
However, the distortion of the crystal structure turns out to be a major issue that results in a decay of the electrochemical behavior of the films: An annealing post-treatment of films turns to be difficult, due to a possible interdiffusion and corresponding degradation of interfaces between the different materials like e.g. NCM and LLZO in the composite cathode. 120,121 Considering this, it would be preferable to control the crystallinity in the film by controlling the PAD process parameters. Since the densification of the film relies on fracture and deformation of the particles, one of the effects (a full densification or preservation of initial crystallinity) would be impeded. However, it might be possible to control this effect by a certain degree through variation of the kinetic energy of the particles, e.g. by the carrier gas flow, 3,122 species, 3,40,123,124 and nozzle design. 125 Furthermore, particles size and distribution of powder mixtures can decide over the final film properties. Studies show that agglomerates of nano-sized primary particles show a negligible reduction in particle and crystallite size after deposition, which is expressed in an unchanged width of diffraction reflexes of the films compared to the initial powder. 105 The resulting electrochemical properties remain (mostly) unchanged. As another attempt, consolidated films with less strain and microstructural distortion could be created by the codeposition of different materials. This can generate interesting composite microstructures as shown for the nanoparticle deposition (NPD) process in Fig. 7. 126 Among others, the effect can be caused by the different powder particle sizes. Besides that, it can also be achieved by embedding a harder fraction into a softer material. 6 Up to now, this effect has not been shown for battery materials yet.
Finally, as there are several examples to fabricate the individual functional components by PAD, one should also consider the direct processing of subsequent components of a battery cell e.g. by the direct deposition of the cathode on the SE or vice versa as well as the deposition of gradients through subsequent deposition of different materials mixtures. 127 In comparison to high-temperature processes that will lead to significant diffusion of elements between the different components of the cathode, 128 PAD offers a valuable alternative due to the room temperature process conditions. This can prevent from reaction and deliver intimate contact between the different components. However, effects such as diffusion that might arise from an additional annealing treatment even at reduced temperatures to recover the functional properties have to be taken into account. The described process can indeed reduce the temperature for setting the coating properties. Ultimately, however processing of components might not be possible without any thermal post-treatment.
As mentioned in this section, the PAD process gives a variety of opportunities to process the different components of a future ASSB. The versatility of the process results from Powder aerosol deposition methodnovel applications in the field of sensing and energy technology 1930005-9 the targeted variation of the process parameters such as particle size distribution, co-deposition, and particle energy during deposition.

Fuel cells
Fuel cells are an alternative way to convert chemical into electric energy. High-temperature solid oxide fuel cells (SOFCs) are not only for energy generation through the recombination of hydrogen and oxygen but also for the conversion of common hydrocarbon fuels and thus for forming an ideal transition technology (e.g. for decentralized energy supply). The core of that fuel cell-type is a ceramic electrolyte that conducts oxide ions (e.g. yttria-stabilized zirconia, YSZ), whereby proton-conducting ceramic electrolytes are currently also under investigation, since one envisages reduced operating temperatures of only approximately 700 ○ C. As mentioned in the battery section, one of the major challenges in the operation of SOFCs are temperature-related interdiffusion processes. They can occur during co-sintering processes of the components during cell production, but also during (high temperature) operation. In particular, insulating intermediate phases (e.g. SrZrO 3 and La 2 Zr 2 O 7 ) are formed between the cathode (e.g. LSCF) and the electrolyte (e.g. YSZ). They reduce the cell performance and hinder its commercial realization. One way to meet this challenge is to introduce diffusion barrier layers between the electrolyte and the cathode to prevent this kind of cell aging. These protective layers mostly consist of gadolinium-or samarium-doped cerium oxide (1). Another conceivable option is to reduce the processing temperatures. This can also go hand-in-hand with the choice of other electrolyte materials (2). The PAD method at room temperature is predestined for both approaches.
(1) By PAD, flawless ceramic coatings on the SE can be produced. The requirements for the powder already have been investigated. Even if particles with a particle size of approximately 2 μm are required for the PAD process, 28,129 layers with a thickness of approximately 1.5 μm 130 (see Fig. 8) can be produced due to the impactbased consolidation process of the particles. 3 The layers have a high adhesive strength and are gas-tight. In addition, it has already been proven that the cathode could be built up on the PAD-processed protective layers by screen-printing and that functional cells could be produced. 131 (2) In addition to the formation of diffusion barrier layers to allow high-temperature operation, SEs can be directly deposited on porous electrodes. 132 In order to kinetically prevent interdiffusion during operation, it is necessary to lower the operating temperature of the fuel cell. Further adjustments are necessary to ensure that the performance does not drop, since YSZ conducts ions significantly lower at lower temperatures. This can be achieved by alternative electrolyte materials with sufficient conductivities at lower temperatures, e.g. some proton-conducting ceramics (barium stannates, barium cerates, or barium zirconates). Another approach is to significantly reduce the electrolyte layer thickness and thus the oxide diffusion length (< 15 μm) as one major contributions to  the overall cell-resistance. In that case, the mechanical stability of the cell must be provided by other components (electrode or interconnector).
In principle, the structure of the porous electrodes is also conceivable. As already shown, the co-deposition of several powders is also possible. If a soluble/combustible second component (e.g. a polymer) is selected, a porous PAD electrode-layer can also be produced in a subsequent step.

Summary
This small review deals with current trends and developments in the field of PAD of functional ceramics for applications in the field of sensing and energy technology. It is shown that PAD offers an enormous potential for various applications in the field of gas and temperature sensors, as well as for vibration energy harvesting and actuation, TEGs, superconductors, solar cells, batteries, and fuel cells. This potential is based on the unique features of the PAD. Various features offer ample advantages over the state-of-the-art coating techniques like physical vapor deposition (PVD), chemical vapor deposition (CVD), chemical solution deposition (CSD), screen-printing and thermal spraying for functional ceramics. From the side of process technology (left-hand side in Fig. 1), processing at room temperature, high deposition rates, or solvent-free processing have to be mentioned. As shown in Fig. 9, the heat loading on the substrate during PVD processes, such as vaporization or sputtering, is similarly low. However, the deposition rate of the PAD processes is about 30 times higher than for PVD or CVD processes. 3 All other processes take place at higher substrate temperatures (CVD) or require a thermal treatment for solvent removal (CSD) or for film sintering (screen-printing). Compared to PAD, thermal spraying or screen-printing achieve significantly higher deposition rates. However, the surface roughness and the film density (pores, cracks) are usually higher, too. 3,134 Changes in the composition or oxidation state of the functional ceramics can also occur during thermal spraying, screen-printing or PVD processes. 134 This is different with the PAD process. In this process, films with high density, transparency or controlled porosity can be produced without any change in composition, stoichiometry or crystal structure. Consequently, the PAD process is also advantageous from the side of film characteristics in many applications.
In future, the process of the PAD could play an increasingly important role for functional ceramics when applied in the field of sensing and energy technology.