Inkjet‐Printed Flexible Thin‐Film Thermal Sensors for Detecting Elevated Temperature Range

All inkjet‐printed thermal sensors are manufactured based on a metal–insulator–metal (MIM) interface or capacitor architecture, for the adapted device size ranging from 16 to 36 mm2 active area. Two different material inks, namely a nanoparticle conductive silver ink and an inorganic‐polymer‐based hybrid insulator ink, are applied layer by layer on a thin flexible polyimide substrate, for developing the printed MIM devices. To ensure the desired electronic conductivity and insulation from the layers, the manufacturing process steps and parameters are tuned, accordingly. The results show that the inkjet‐printed MIM devices could constitute up to 15 μm thickness and demonstrate average detection of a change in electrical capacitance ranging from 20 to 100 pF, when the temperature is varied between 100 and 300 °C. The investigations also summarize that the change in the electrical response is enough to detect an increment of 50 °C. The printed sensors also display high operational stability and repeatability, when subjected to thermal cycling.


Introduction
Inkjet printing technology due to its micrometer-scale deposition accuracy, industry-relevant upscalability, print-transfer adaptability from 2D to 3D surfaces, and versatile additive manufacturing processability (digitally controlled) fits perfectly toward the deposition of thin-film-based printed electronic components.During the past decade, a wide variety of printed electronics-based applications were demonstrated using inkjet printing technology, including realization of piezoresistive sensor matrix, [1] inkjetprinted photodetectors for detecting X-ray or UV radiation, [2,3] sensors for detecting dissolved oxygen concentration, [4] and even development of electronic chemical platforms for detecting phosphates. [5][8] These publications not only illustrate the development of electronic circuits for applications such as signal filtration, amplification, and power switching, and realization feasibility using inkjet-printed TFTs, but at the same time describe the importance of selecting the right printable functional material stack and process parameters, for desired electronic application as well.In addition to this, high exploitation potential is also experienced in the application field of high radio frequency identification communication, [9] deposition of catalytic layers in the proton-exchange membrane fuel cell, [10] performance enhancement of photovoltaics and their components, [11,12] and 3D-conformable heating surfaces. [13,14]Finally, the technological upscalability of the manufacturing process for active and passive electronic devices and its tolerance with the performance evaluation are also investigated. [15,16]Based on these, one could conclude that the inkjet printing is a key enabling technology for developing microelectronic components and circuits, with the benefit of low-cost investments, fair trade-of between performance and device designing with respect to applicability, reduced material wastage, and high manufacturing sustainability.
In this article, we want to focus on the topic of sensors which are actually able to detect change in temperature and that too, over an elevated range (more than 100 °C).The temperature sensors introduced by Wang et al. were found based on the principle of surface plasmon resonance with the help of multimode fiber-photonic crystal fiber-multimode structure coated with gold thin films.The sensors were able to detect a temperature change between 35 and 100 °C, with a sensitivity of À1.551 nm °CÀ1 . [17]ao et al. focused on a temperature range of 25-80 °C, where the optical fiber sensor comprised polydimethylsiloxane (PDMS) and silica composite filling, helping the single-mode system to detect the modulation in emerging resonance wavelength upon temperature change. [18]Rajasekar et al. on the other hand developed nanosensing platforms manufactured on the basis of the photonic crystal ring oscillators, with the principle of refractive index modulation, which contributed in detecting temperature change from 5 °C to an elevated range of 540 °C. [19]Similar R&D work was also accomplished by Pan et al., where the main focus was put forward on TiO 2 thin film coated on the optical fiber tip (diffraction grating mechanism) temperature sensors operating between À10 and 180 °C. [20]Besides the classical way, that is, optical mechanism for detecting temperature change, Sarma et al. demonstrated the development of temperature sensor working in the range of 22-200 °C, on the basis of change of resistance and specially grown carbon nanotube functional electronic layers on Si-based substrate. [21]A similar principle was also used by Liu et al., where the sensors were fabricated with the reduced graphene material and could detect the temperature change in range of 30-100 °C, for the implementation in robot skin and Internetof-Things (IoTs). [22]Once again the same resistive principle was exploited by Sivagnanapalani et al., who were successful in obtaining powder precursor material Nd 2 Ti 2 O 7 (NTO) and its processing technique, with the characteristic to be utilized in high-temperature sensors based on perovskite-like layered structured, illustrating a resistance drop from 10 14 to 10 6 Ω cm with an increase in temperature from 100 to 900 °C. [23]In contrast to the abovementioned sensor specification, Alpert et al. showed rigid solid-state sensors manufactured using InAlN/GaN and AlGaN/GaN heterostructures, where the increase of temperature from room temperature to 576 °C was detected by the decreasing voltage-scaled magnetic field, as an electronic effect arising from the decreasing electron mobility due to the scattering effect. [24]ome of the above mentioned works cannot be compared directly to our research focus in this manuscript, as we concentrate on a broad range of temperature, represent printed sensors, and intend to follow the principle of electrical capacitance for our measurements.
In contrast to this, the use of printing technologies for manufacturing temperature/thermal sensors on flexible substrates has been also widely addressed by several investigators, for example, Soni et al. showing printed temperature sensors based on poly (3,4-ethylenedioxythiophene) polystyrene sulfonate (PEDOT:PSS)-graphene oxide composite and Su et al. reviewing on similar materials for sensing temperatures between 25 and 100 °C. [25,26][32][33][34] With this research paper, we intend to exhibit that with the right choice of functional electronic materials, deposition equipment, printing-, pre-& post-treatment steps, and tuned deposition process parameter sets, we are able to achieve fully inkjet-printed thin-film-based thermal sensors, for detecting temperature variations over the elevated range from 100 to 300 °C, on 50 μm thin flexible polymeric substrate.It is known from the work presented by Rajasekar et al. [19] and Alpert et al. [26] that they were able to develop sensors operating at the high-temperature range on the principle of change in optical refractive index modulation and voltage-driven magnetic field modulation, respectively.These devices are bulky, rigid and stiff, coming from solid-state electronics that cannot be implemented for specific applications, and show their infeasibility toward direct 3D integration.Moreover, we want to investigate printed thermal sensors that are based on the principle of inverse correlation of capacitance versus temperature changes, occurring due to the change in dielectric permittivity modulation once polymer reaches its glass transition temperature, for example, epoxies, and also the connected change in thermal conductivities. [26][37] The main scientific focus here is to investigate and establish the correlation between the electrical sensitivity of the printed electronic devices, with respect to several variables such as the deposition parameters, geometrical dimensions, and materials, concerning the variation in high-temperature changes.The printed thermal sensors were based on a MIM interface architecture (capacitive type), while the active areas are investigated between the range of 16 and 36 mm 2 .For manufacturing these MIM-printed devices, two different ink materials namely a nanoparticle based silver (for electrodes) and an inorganic-polymer based hybrid insulator ink were implemented layer by layer on a thin polyimide (PI) substrate.The abovementioned inks were jetted out individually through dedicated drop on demand (DoD) inkjet printheads along with the manufacturing process adapted deposition parameters, for example, printing resolution or drop space, jetting waveform, jetting frequency, ink meniscus etc., optimized for achieving planar layer topologies.Moreover, the specified electronic conductivity and insulation from the layers were ensured by tuning the thermal post-treatment method, that is, temperature & duration.When compared to the work done in literature, we are able to state that the kinds of thin-film inkjet-printed MIM devices are noble, and they for the first time addressed in the research field, showing a considerable change in the electrical capacitance, when temperature is varied between 100 and 300 °C, along with a change in the electrical response that is enough to sense an increment of 50 °C.Also, these inkjet-printed MIM devices display high operational stability and repeatability, when subjected to thermal cycles, for which similar work has not been pursued before.The outlook of transferring the manufacturing process to 3D object is also very innovative, where the sensors and nodes can be adapted directly for a specific 3D mechanical component, with a possible adaptation of its device sensitivity.

Experimental Section
For realizing metal-insulator-metal (MIM) capacitive devices using inkjet technology, it was necessary to plan and design the digital patterns correctly for the printing process.This would help in obtaining the required electronic characteristics of a capacitor, when the conductive and insulating materials are deposited in a specific sequence.The fundamental of a MIM device stack architecture is shown in Figure 1A.The planning of the inkjet-printed devices started with the design of the top electrode, which were formulated in such a way that a squareform resulted into the effective device active area, which was then extended lengthwise with the conductive track for establishing the electrical contact during the characterization process.
Experiments were planned, so as to see the influence of change in capacitance over the variations that are dependent on the change in dimension of the squareform regarding the top electrode.Hence, following active areas were now defined for the top electrode layer, that is, 4 mm Â 4 mm leading to an active area of 16 mm 2 , 5 mm Â 5 mm for the devices with 25 mm 2 , and 6 mm Â 6 mm for the MIM devices with 36 mm 2 , respectively.The intermediate insulator layers were formulated with squareforms with dimensions of 6 mm Â 6 mm for the devices corresponding to 16 mm 2 active area, 7 mm Â 7 mm for the devices corresponding to 25 mm 2 active area, and 8 mm Â 8 mm for the devices corresponding to 36 mm 2 active area, respectively.The design and exact allocation of the bottom electrode was accomplished very precisely below the top electrode, such that the effective dimension of the electrode was slightly more than that of the top electrode, that is, 4.25 mm Â 4.25 mm for the MIM devices with 16 mm 2 active area, 5.25 mm Â 5.25 mm for the MIM devices with 25 mm 2 active area, and 6.25 mm Â 6.25 mm for the MIM devices with 36 mm 2 active area, respectively.This increase in the dimension of bottom electrode was accomplished, so as to confirm a favorable overlap between the top and bottom electrodes, thereby generating the desired electric field.
As depicted in Figure 1B, the pattern layouts were printed using two inkjet-printable commercial functional inks, that is, a nanoparticle-based silver ink with up to 45 wt% particle loading and a hybrid insulator ink constituting of SiO x -based nanoparticles embedded to the polymer matrix.At first, the digital layout with regard to the bottom electrode layer was printed on top of a 50 μm thin polyimide (PI) Kapton film from DuPont.A laboratory relevant "DoD"-based piezoelectric inkjet printer DMP-2831 from Fujifilm Dimatix Inc. was used along with 10 pL-based DMC cartridges, for the deposition of conductive and insulator inks.Two discrete jetting waveforms were developed and utilized, dedicated to the individual inks based on their rheology and printing condition-related adaptations, to obtain the most optimal layer quality.The printing of the silver-based bottom and top electrodes was accomplished using printing resolutions of 1016 dpi (25 μm drop space) and 1270 dpi (20 μm drop space), respectively, and implementing only 1 layer.In contrast to this, for acquiring better insulation properties, the printing of insulator ink was performed with two layers, once with two times the same printing resolution 1270 dpi i.e., (20 þ 20) μm drop spaces or with combination of printing resolution of 1693 dpi 1 layer and 1270 dpi 1 layer i.e., (15 þ 20) μm drop spaces, respectively.While using the two different printing strategies for the insulator ink, layers were always obtained out of "wet-on-wet" deposition from the two layers collectively, either with (20 þ 20) or (15 þ 20) μm drop spaces, which were then followed up with post-treatment method.The printing of both the inks was performed using 5 kHz jetting frequency, 12-16 nozzles, %1 mm printing distance, 28 °C substrate temperature, 35-40 °C printhead temperature, and equivalent underpressure of 4.5 inch H 2 O leading to a favorable negative ink meniscus.Soon after the printing of two insulator layers was accomplished, they were thermally cured using a vacuum oven at 150 °C temperature for a duration of 30 min.In comparison to this, the printed conductive layers, that is, bottom and top electrodes were individually sintered in the same vacuum oven and also with the same temperature and time duration.This meant during the deposition of the printed MIM device stacks, the first printed silver layer, that is, the bottom electrode experienced a thermal posttreatment at 150 °C, for three times and in total for 90 min.Next to this, the second printed insulator consisting of two layers experienced a thermal post-treatment at 150 °C, but for two times and in total of 60 min, and finally the last printed silver layer, that is., top electrode containing only 1 layer, experienced only single post-treatment process.
For the characterization of surface topology regarding the printed layers, that is, layer thickness, layer homogeneity, roughness, edge sharpness, spread-out, etc., an optical microscopy Axio Imager.M2m from Carl Zeiss Microscopy GmbH and a mechanical tactile-based surface profilometer Dektak 150 surface profilometer from Vecco Instruments Inc. were used.The electrical characteristics of the MIM devices such as the resistance and the capacitance were measured at 10 kHz, using PM5 probe station from Süss MicroTec coupled with rounded pressure-sensitive measurement needles and an Agilent E4980A Precision inductance capacitance resistance (LCR) meter.The sensor's functional properties were evaluated using a simulated in situ measurement setup environment, where the MIM devices were subjected to high elevated temperatures and the respective change in the capacitance was recorded.Here, the printed MIM samples were placed on top of a conventional heating plate and the respective capacitance values were recorded using the LCR meter coupled to the measurement probe station.With any increase of the temperature at the heating plate, the localized temperature at the MIM devices was also increased proportionally, and at this point, the corresponding capacitance was recorded using the LCR meter coupled with the probe station.

Optical & Topological Characterization
The photo of the inkjet-printed MIM devices deposited on PI substrate can be seen in Figure 2, where one out of the ten representative devices (active area of 36 mm 2 ) is exemplified to show further details.It can be seen from the magnified image that the MIM structure shows a straightforward-printed pattern layout, where the bottom and top electrodes possess identical dimensions that overlap each other, generating the active zone of the MIM device.
The insulating layer separating the electrodes is semitransparent in appearance and was designed geometrically to exceed the dimension of the active zone for the MIM structures, that is, overlapping area of the conductive electrodes.From the microscopic image, it is also seen evidently that the printed silver ink was managed to allocate from the level of insulator down to the PI substrate, hence defining the top electrode contact, besides the other already existing bottom electrode contact extending out beneath from the insulator layer.When the surface of the top electrode at the active area is focused, some minor material distribution inhomogeneities can be detected, occurring due to the sintering process and difference between the thermal coefficient of expansion between the insulator and silver layers.The impact of this irregularity was not found to be very dramatic, as the electrical conductivity of the entirely printed pattern over the insulator layer was found to be consistent.The next step was to measure and evaluate the morphological and topological properties of the individual printed conductive and insulating layers, as a function of the change in process parameters and material.For the purpose to execute characterization, 5 mm Â 5 mm digital patterns were printed on substrate for both the conductive and insulating inks, using the optimized deposition parameters.As mentioned before, the metal electrodes were deposited using a nanoparticle-based silver ink with 25 μm drop space and optimized parameters, followed by sintering in the oven at 150 °C for 30 min.However, the insulator layer was printed with two layers, using two different print resolutions/drop space, that is, twice 20 μm drop space and combination of 15 and 20 μm drop spaces.Soon after the layers were printed, they were cured thermally in the oven at 150 °C for 1 h.Then, the solid layers were measured for layer thickness, surface homogeneity, and material distribution properties, using a Dektak surface-profilometer. Using the measurement system, several linear scans (5 times in and 5 times counter print directions) were performed to evaluate the cross-sectional property of the layers.The average surface profiles of the printed conductive and insulating layers are shown in the Figure 3.Although the printed silver layer exhibited a relatively low thickness, that is, 1.2 AE 0.1 μm, it showed comparatively high roughness 100 nm but good material distribution and layer homogeneity.The relatively high roughness of the silver layer is expected to be the result of high particle size and loading, and also the sintering process as a whole, which is specific to the ink chemistry for a defined sintering temperature and duration.Compared to the silver ink, the printed insulator layers showed even higher layer thicknesses, for example, 10 AE 4 and 14 AE 4 μm, for the layers deposited using two times 20 μm drop space and combination of 15 and 20 μm drop spaces, respectively.This summarizes the fact that lower layer thicknesses can be achieved, when the deposition process is executed using two times μm drop space, rather than combination of 15 and 20 μm drop spaces.This relates directly to the justification that with a utilization of high print resolution (low drop space), more material is deposited and thereby higher layer thickness and vice versa.Despite the high layer thickness, the topology of the printed layers was found not to be very planarized, although for both the deposition strategies, similar cross-sectional profiles as a function of scan length were obtained.Depending on the implemented printing drop spaces, the rightmost section (top part of pattern) of the layer profile showed thickness ranging between 14 and 18 μm, whereas left section (bottom part of pattern) of the profile was as low as 5 μm.This difference in the layer thickness can only be explained with the phenomenon of gradient type in situ drying behavior, with regard to segmentwise deposition of materials within a typical slow laboratory-scale inkjet printing process, where the number of printing nozzles is limited.Here, as the material is deposited, the solvent starts to evaporate quickly, leaving behind the semisolidified insulator material that generates a high surface tension edge, thereby forcing the ink to accumulate more rapidly toward the print start location rather than at the very end of the print pattern finishing.It is also interpreted that the impact of this phenomenon is very subjective to the implemented geometrical dimension of the print pattern, that is, as the pattern size is increased, so does the effect and vice versa.

Electrical Characterization
In order to evaluate the electrical performance of the printed MIM devices, several devices (up to five per parameter setting) were manufactured using the two different parameters, that is, deposition of 2 layers, each composed of either 20 μm drop space or combination of 15 and 20 μm, respectively.Then, these printed MIM devices were electrically characterized for capacitance with the LCR meter, under room temperature.In Figure 4, the result of the characterization is depicted for the printed MIM devices, which differs from each other with respect to the active areas.From the graph, it could be very clearly seen that, as the area of the MIM device is increased (from 16, 25, 36, 49 to 64 mm 2 ), the capacitance is found to increase as well (from 97 to 290 pF for 2 layers 20 μm drop space and from 69 to 200 pF for 2 layers drop spaces 15 & 20 μm, respectively).
It was also seen that the MIM devices manufactured with relatively thinner dielectrics %10 AE 4 μm (consisting of 2 layers & drop space 20 μm) showed higher capacitance than the devices with layers combinedly processed with drop spaces 15 and 20 μm, respectively.Both the abovementioned outcomes of the electrical characteristics from the MIM devices, that is, variation in capacitance with respect to the change in dielectric thickness and device sizes, fit perfectly to the physics of parallel plate capacitors.Furthermore, the dielectric constant of the insulator layer used in the MIM devices was also calculated to be %5, which matches the general electrical properties of polymeric materials.As the measurement values were compared and analyzed, it was observed that in general the devices with bigger active areas showed higher tolerances in electrical capacitance, than the ones with smaller device sizes.Next to the abovementioned basic electrical characterization for measuring capacitance against the device size and layer thickness, it becomes extremely important to first focus on the usable operational range of the MIM device size and second the effect of the exposure to the elevated temperatures for the printed MIM devices.To execute such an in situ measurement process, dedicated measurement needles or probes were connected to an LCR meter, which was furthermore extended to the electrode contacts of the MIM devices, placed on top of a heating plate.From the previous investigation, the range concerning the MIM device size was narrowed down to 16, 25, and 36 mm 2 .Several devices (up to five per parameter setting) were manufactured with the mentioned active areas, which were then exposed to elevated temperature range starting from 100 to 300 °C, with a step of 50 °C.In Figure 5, graphs are shown which describe the capacitive response to the implemented elevated temperatures, for which the range is varied between 100 and 300 °C.
From both the graphs (A) & (B), it could be observed commonly that the capacitance varies proportionally with respect to the active area of MIM devices, and next to this, increase in capacitive response is observed with respect to increase in the temperature.It could be seen that the capacitance varies for all three active areas between 68 and 136 pF at 100 °C, to maximum of 95-186 pF at 300 °C, when the dielectric within is processed with two layers using 15 and 20 μm drop spaces, respectively.In comparison to this, for all MIM devices processed with 2 layers using 20 μm drop space, the difference in the capacitive response was found to be relatively high, due to thinner layer, that is, between 95 and 190 pF at 100 °C, to maximum of 130-290 pF at 300 °C, respectively.The degree of increment in capacitance was found to be higher for drop space 20 μm, compared to the ones with combination of 15 and 20 μm drop spaces.Next to this, an average capacitance of 7 AE 3, 9 AE 4, and 13 AE 7 pF per increment of 50 °C was achieved, for the devices developed using dielectric containing two layers and processed combinedly with 15 and 20 μm drop spaces, along with active areas of 16, 25, and 36 mm 2 , respectively.In contrast to this, for the MIM devices developed using dielectric composed of two layers processed with only 20 μm drop space, an average increment of 9 AE 4, 14 AE 7 and 18 AE 12 pF for the same active areas, respectively, was shown.When compared among each other, in general, it can be said that the average increment of capacitance per increase of 50 °C for thinner dielectric (2 layers with 20 μm drop space) is higher, along with the higher tolerance value observed in direct dependence to the active area of the MIM devices.However, the increment of capacitive response attained by the MIM devices manufactured using thicker dielectric is comparatively low, as well as the low tolerance offered, and stable throughout the entire high-temperature range of 100-300 °C, for all the device sizes.Besides this, in general, one could also calculate theoretically the sensitivity for the thermal sensors ranging between 0.08 and 0.26 pF °CÀ1 , 0.1 and 0.42 pF °CÀ1 , and 0.12 and 0.6 pF, for active areas of 16, 25, and 36 mm 2 , respectively.The precise measurement of such small values of capacitance over stepwise increment of temperature is technologically feasible, but has been excluded within this research work.
As next step, the same number of samples were exposed to thermal cycles, that is, the MIM device samples were in situ measured for capacitance, with one up and one down swing of temperature ranging from 100 to 300 °C, with step of 50 °C increment.During the entire in situ measurement process with the implemented elevated temperature (treated by heating plate), the samples continued to lay on the heating plate, thereby experiencing the entire temperature range.In Figure 6, results are shown where the printed MIM devices of different sizes and dielectric thickness (drop space combination) are individually varied, as a function of change in the elevated temperature ramped up from 100 to 300 °C, and finally ramped down to the initial value of 100 °C.In the figure (A), (C) & (E), the dependency is shown toward the capacitive response with respect to various active areas and against the elevated temperature changes for the MIM devices, fabricated using dielectric processed with drop spaces of 15 and 20 μm.Here, the capacitance was seen to: (A) increase from 70 AE 2 pF for 100 °C to 84 AE 1 pF for 300 °C while going up and returning to 71 AE 2 pF while ramping down to the initial temperature for 16 mm 2 active area; (C) increase from 101 AE 1 pF for 100 °C to 123 AE 1 pF for 300 °C while going up and returning to 104 AE 1 pF while ramping down to the initial temperature for 25 mm 2 active area; and (E) increase from 139 AE 4 pF for 100 °C to 179 AE 3 pF for 300 °C while going up and returning to 139 AE 4 pF while ramping down to the initial temperature for 36 mm 2 active area.Similarly, in the figure (B), (D), & (F), the dependency is shown toward the capacitive response with respect to various active areas and against the elevated temperature changes for the MIM devices, fabricated using dielectric processed with 2 layers using only 20 μm drop space.Here, the capacitance was seen to: (B) increase from 99 AE 3 pF for 100 °C to 118 AE 3 pF for 300 °C while going up and returning to 99 AE 3 pF while ramping down to the initial temperature for 16 mm 2 active area; (D) increase from 145 AE 3 pF for 100 °C to 174 AE 3 pF for 300 °C while going up and returning to 148 AE 2 pF while ramping down to the initial temperature for 25 mm 2 active area; and (F) increase from 186 AE 34 pF for 100 °C to 248 AE 17 pF for 300 °C while going up and returning to 198 AE 16 pF while ramping down to the initial temperature for 36 mm 2 active area.From the results presented here, it can be summarized as first observation that the tolerance in capacitive response (standard deviation) offered by MIM devices fabricated using 2 layers and drop space 20 μm was found higher than the ones which were manufactured using 2 layers with drop space combination of 15 and 20 μm respectively.Second, it is found evident to see a slight drop in capacitance for all the printed MIM devices, while ramping down from 300 to 100 °C.Finally, it was also found that the capacitance values for mostly all the MIM devices while ramping up and down the temperature range ended with the same initial value, indicating a minor hysteresis.Next to this, it was also concluded that the MIM device response as function of elevated temperatures was found to be very stable and robust.
Figure 7 on the other hand shows the result of the measurements performed on the printed MIM devices, manufactured with selected process parameters, that is, active areas 25 and 36 mm 2 along with layer thickness variation (14 AE 4-10 AE 4 μm) achieved from the combination of drop spaces (15 þ 20) and (20 þ 20) μm, respectively.For these considered 25 printed devices (5 device per parameter), a total of 25 thermal cycles were implemented (5 cycles per device) between the predefined two temperatures 150 and 300 °C, forward and reverse.The main objective in this case was to evaluate the repeatability of the change in the capacitance at defined elevated temperatures.For a summarized overview, the five individual sensors per device parameter (active area and dielectric thickness) were averaged; therefore, a higher standard deviation arises due to deviations in the initial capacitance of the singe devices.Within Figure 7A, it was observed clearly that as the temperature was cycled forward and reverse, on the devices with 25 mm 2 active area and manufactured with drop spaces (15 þ 20) μm, the average capacitance was found to vary between 107 pF (at 150 °C) and 119 pF (at 300 °C), with a tolerance of AE1.5 pF.In contrast to this, for the devices shown in Figure 7B with the same active area but thinner insulator layer, that is, processed with drop spaces (20 þ 20) μm, it was found to exhibit an average capacitance between 154 pF (at 150 °C) and 165 pF (at 300 °C) with a tolerance of 2 pF.When comparison was made between all the devices among the same active area of 25 mm 2 , it was summarized that, first, the value of the individual initial capacitance itself is on a higher level for the MIM devices manufactured using drop space (20 þ 20) μm (starting capacitance is %47 pF higher).Second, among the same device sizes, when the sensitivity (capacitance difference) of the thermal sensors is compared between 150 and 300 °C, then one could say that the capacitance variation is found to be 12 pF for the devices developed using drop space (15 þ 20) μm, whereas that for the devices processed with drop spaces (20 þ 20) μm it was only 11 pF.Next to this, from Figure 7A,B it can also be observed, that although the tolerances in capacitance of the devices processed with two different thicknesses were almost same, that is, 2 pF, the device performance stability was found better for the ones with drop space (15 þ 20) μm.
On the contrary, from Figure 7C, it was seen that as the temperature was cycled on the devices with 36 mm 2 active area and manufactured with drop spaces (15 þ 20) μm, the average capacitance was found to vary between 143 pF (at 150 °C) and 161 pF (at 300 °C), with an average standard deviation of 5 pF. Figure 7D shows the results from the devices manufactured with the same active area, but thinner insulator layer, that is, processed with drop spaces (20 þ 20) μm, where they exhibited an average capacitance between 206 pF (at 150 °C) and 224 pF (at 300 °C) with a standard deviation of %AE18 pF, which is high compared to the other MIM devices.An important remark here is again the representation of the average capacitance of five devices with different initial starting capacitances.Considering each single sensor, the standard deviation of the capacitance was found to vary between 1 and 2 pF over the 5 cycles at 150 and 300 °C, respectively.However, illustrating all the individual devices would overlay the measurement data, which would no longer show a clear tendency for a read out.The repeatability of the manufacturing process for the devices is shown and reveals a more stable and reliable device manufacturing for smaller sensor areas.While comparing all the devices of the same 36 mm 2 active area, it was noticed that the change in capacitance for both the drop space combinations is on a higher level, that is, 18 pF (from 150 °C to 300 °C).Finally, Figure 7C,D shows that the deviation in capacitance within a single device is comparatively low, with only 2-3 pF, but the tolerance among several devices manufactured with the same parameters is high.The range of tolerance from sensor to sensor can be considered as big, leading to the interpretation that the sensors of 36 mm 2 active area size offer lesser operational stability, and they would be difficult to adapt toward certain electronic applications, where minimum tolerance and maximum sensitivity are demanded.
From Table 1, we can summarize that with an increase of 10 °C we can achieve only 0.73-1.2pF increase of capacitance which is for certain high-precision sensor application not really relevant.However, when the temperature resolution (low) is set between 20 and 25 °C, the increase in capacitance reaches above 1 pF, which sets the measurement technique, data recognizing, and comparison, in a comfortable situation to realize electronic applications.As alternative, for realizing practical applications including the printed devices, the minimum scale of the measurement equipment can also be tuned (electrically) for extracting high sensing/processing resolutions versus small  capacitance increment, for responding to the faster change in temperatures along with high accuracy.
It can be stated clearly that from the abovementioned results that the capacitance increases as the temperature is increased on an elevated range especially between 150 and 300 °C.This relation demonstrates a direct dependency toward each other, which brings up the question of reasoning behind the occurrence of this specific electronic phenomenon with such an effect.From the literature, it can be recognized that the insulator's material physics would play a big role in the development of sensing properties for the printed devices, upon exposure to the thermal signatures of high elevated range.We believe that in our case when the printed MIM devices are subjected to temperatures above 100 °C, the insulator in hybrid form (SiO x nanoparticles embedded in polymer matrix) behaves very differently.It is interpreted that as the temperature increases beyond 100 °C, the dielectric constant of the insulator material starts to vary, due to several intrinsic electronic factors drifting the dielectric behavior upon thermal changes.][37] A possible increase in the dielectric constant would result in an increase in the charge accumulation at the electrode to insulator interface junction, for both the electrodes across the insulator layer, for the MIM devices.Besides this, we have also concluded exclusion of the possibility of geometrical/ dimensional changes developed by either the increase of active area by the expansion of electrodes or decrease of the insulator layer thickness, which are the only other ways to inhibit changes to the capacitance values, at a constant operational bias frequency.Finally, due to the mentioned effect, it can be said that there is an increase/decrease of capacitance value in a linear tendency, upon a proportional variation in temperature, that is shown in the graphs in Figure 5-7, respectively.

Conclusion
In our research, we have developed inkjet-printed thermal sensors that go beyond the commonly know thermal sensors, which mostly operate between room temperatures and 150 °C and are already reviewed for printed and flexible electronics.Here, a thermal sensor based on a capacitive working principle suitable for detecting temperature changes at an elevated range between 100 and 300 °C has been developed, by manufacturing a fully inkjet-printed MIM stack on a flexible substrate.Various parameters, such as active area and dielectric thickness, were thoroughly investigated, analyzed, and identified, where all differently designed sensors reveal a general response to a change in temperature, but a higher capacitance will increase the sensor's performance and sensitivity.In detail, this means that the sensors with a larger active area and thinner insulators will lead to the highest change in capacitance, over the temperature variations.However, the electrical fluctuation and tolerance for devices increase with higher active area and higher change in capacitance.Finally, the thermal sensors are validated by defined thermal cycling accomplished between 150 and 300 °C, demonstrating straightforward repeatability with a very low tolerance, as well as a high reliability within the single sensor.Furthermore, these findings also allow to address the integration of such thermal sensors working at the high-temperature range, directly onto 3D objects in combustion engines/automotive friction parts, security features, or in embedded electronic applications for lighting purposes.

Figure 1 .
Figure 1.A) Representation of the MIM device stack architecture on PI substrate and B) the corresponding digital images for the realization of the printed MIM structure of different active areas on the basis of the bottom electrode, intermediate insulator, and top electrode layers, respectively.

Figure 2 .
Figure 2. Photographic and microscopic representation of all inkjet-printed MIM structures shown exemplary for 36 mm 2 active area developed on PI substrate, for detecting change in temperature over high range.

Figure 3 .
Figure 3. Graph depicting surface profiles of inkjet-printed metal and insulator layers, difference in the layer thicknesses processed from the implemented drop spaces of 25 μm for the nanoparticle silver ink, and (20 þ 20) and (15 þ 20) μm for the dielectric ink, respectively.

Figure 4 .
Figure 4. Graph shows the relationship between the size/active area of the inkjet-printed MIM devices with respect to the measured capacitance at room temperature and varied deposition parameters, that is, drop spaces.

Figure 5 .
Figure 5. Graphs show the relationship between the measured capacitance with respect to the size/active area (16-36 mm 2 ) of the inkjet-printed MIM devices and variation in the exposure to elevated temperature ranging between 100 and 300 °C, when the insulator layer is deposited using A) drop spaces (15 þ 20) μm and B) drop spaces (20 þ 20) μm, respectively.

Figure 6 .
Figure 6.Graphs A-F) show the relationship between the measured capacitance with respect to the size/active area (16-36 mm 2 ) of the inkjet-printed MIM devices with respect to the variation in exposure to elevated temperature ranging from 100 to 300 °C and back (50 °C step), when the insulator thickness is varied by implementing drop spaces (15 þ 20) μm and drop spaces (20 þ 20) μm, respectively.

Figure 7 .
Figure 7. Graphs A-D) show the impact of measured capacitance in form of thermal cycles, with respect to the two sizes (25 & 36 mm 2 ) of the inkjetprinted MIM devices and variation in exposure to the elevated temperature ranging from 150 to 300 °C and back, when the insulator thickness is varied by the implementation of drop spaces (15 þ 20) μm and drop spaces (20 þ 20) μm, respectively.

Table 1 .
Addressable temperature resolution for the thermal sensors.