Sustainable Multifunctional Biface Sensor Tag

In this article, a sustainable, multifunctional, low‐cost, wireless sensor tag is presented. The sensor tag combines three different environmental sensors in one single platform for the dedicated purpose of wireless structural health monitoring of a variety of applications. However, the adaptive design allows the integration of different sensors depending on the specific sensing task. The material consumption is minimized by double‐sided printing, resulting in compact, resource‐efficient sensor solutions. On one side, the tag is equipped with a printed antenna, a fully passive silicon‐based near‐field communication chip and a carbon‐based strain sensor, while the environmental sensors for humidity and temperature are printed on the other side. Due to its low cost, the usage of environmentally friendly materials and the absence of a battery, the biface sensor tag is a milestone in the field of wireless, sustainable electronics for ubiquitous sensing applications. The fabrication itself comprises a series of processes with a focus on efficient additive manufacturing. The characterization of the three sensors shows sensitivity values and characteristics comparable to those found in literature and industrially manufactured sensors. The utilization of a smartphone for reading out the sensor signals further emphasizes the sustainable approach of this sensor system.


Introduction
The fastest-growing waste stream worldwide is electronic waste; in 2019, every person in the world generated an average 7.3 kg of e-waste, which is expected to rise to 9 kg per capita by 2030. [1] Concurrently, the number of sensors deployed increases at a high pace. For instance, the industrial sensors market size is expected to grow at a compound annual growth rate (CAGR) of 9.8% by 2025, which is mainly driven by the growing popularity of industry 4.0 and internet of things (IoT). [2] Those developments rise an urgent demand for sustainable and green sensor solutions for the future. In an approach to address the aforementioned challenges, a multifunctional biface sensor was developed. The sustainability of the fabrication process and the device itself was taken into account by employing additive manufacturing technologies, hybrid integration of silicon-based electronics, resource-efficient design and sustainable substrate material, namely uncoated paper substrate. On one of its faces, the sensor tag is equipped with a near-field communication (NFC) antenna and a silicon-based fully passive NFC chip, while humidity, temperature and strain sensors are printed on the other side.
The benefits of this design are on one hand the sustainable use of resources; e.g., uncoated paper substrate is a renewable, low-cost, easily available, and biologically degradable raw material and its usability for printed electronics applications has been well-studied. [3][4][5] On the other hand, the bifacial design further reduces the consumption of resources and increases the compactness of the sensor tag. In general, additive manufacturing is considered to have a low environmental impact, provided that also eco-friendly materials are employed. [6] Considering the cost efficiency in the manufacturing of printed electronics, it is widely agreed that the cost per function can be higher for certain applications compared to the production of silicon electronics. [7] The performance and yield of fully printed electronic devices, especially thin film transistors, [8] and memories, [9] are inferior compared to conventional silicon components, and the fully additive manufacturing of entire microcontroller units including data readout and analysis can neither be considered as feasible nor reasonable in the near future. Hence, the demand for hybrid electronic systems arises, which creates novel areas of applications combining the best of both worlds. [10][11][12][13] Exploiting such a hybrid approach, a commercial, passive NFC chip has been integrated into the printed sensor tag. Due to the fully passive design, no external battery is required. Instead, the energy for the sensor read-out and data transmission is harvested by the chip from the interrogating NFC field. The read-out can be performed handily using a contemporary smartphone.
Aside from environmental considerations, the sensor is highly versatile. The platform is designed in an adaptive way, which means that it can be equipped with different sensors, depending on the individual sensing requirements. Digital manufacturing technology offers large freedom of design as no screens or special equipment is required, which results in efficient manufacturing of highly customized electronic devices even at small batch sizes. At the same time, the device costs are low, due to the use of inexpensive and widely available paper substrates in combination with inkjet printing, which enables cost-efficient manufacturing even at small batch sizes.
In order to benefit from those design advantages, several challenges need to be addressed. Electrical vias through the paper substrate are required, whereby any kinds of burrs need to be avoided at the edges, as they would represent a barrier for the inkjet-printed tracks. Furthermore, the diameter of the hole should be small to keep the surface area small compared to the circumference. This increases the mechanical stability of the via, since the paste, which fills the hole, only adheres to the circumference. In addition, the method for creating the hole should enable precise alignment with the interconnects and avoid costly process steps. Kujala et al. [14] reported the fabrication of highly flexible screen-printed vias in 125 μm thick PET foil using silver microparticle inks for applications in energy harvesting and storage modules. As part of their work, the substrate was perforated employing a laser cutter and the holes were subsequently filled with screen-printing ink. They demonstrated the functionality of this approach by fabricating a flexible printed supercapacitor. In a similar approach, Jansson et al. [15] developed a full roll-to-roll manufacturing process of through-substrate vias with stretchable thermoplastic polyurethane (TPU) enabling double-sided wearable electronics.
For the sensor readout and wireless data transfer, the tag is equipped with a fully passive commercial silicon-based NFC chip. The integration of silicon components onto (thermally) sensitive carrier substrates, such as paper, poses another challenge. Conventional soldering or thermocompression bonding methods are not applicable, due to the low-temperature tolerance of paper. A possible solution to this is thermosonic fine-pitch flipchip bonding, as demonstrated by Roshanghias et al. [16] Alternatively, anisotropic conductive films can be employed. As an example, Yoon et al. [17] reported on the reliability of the bonding of bare silicon dies with thicknesses of 30 μm and 730 μm to screen-printed paper and PET substrates. Yet another approach is mechanical bonding using non-conductive adhesives, which are generally cheaper than conductive adhesives.
In the current configuration, the sensor tag is equipped with three environmental sensors for the measurement of temperature, humidity, and strain, for the dedicated purpose of structural health monitoring of natural fiber-reinforced lightweight components. Today's environmental challenges require innovative solutions for the sustainable use of resources including increased employment of renewable raw materials. In the area of lightweight construction, composites of biopolymers and natural fibers, such as flax, hemp, or cotton, have been studied throughout recent years and their ecological compatibility as well as advanced material performance have been demonstrated. [18][19][20] For future material development to improve the performance and reliability of such composites, structural health monitoring in the fields can be considered as key enabling. [21,22] To gain invaluable sensing data for a deeper understanding of degradation modes due to environmental influences the sensor tag could be fully integrated becoming an inherent part of the composite material to monitor, as previously presented for a simple paper-based temperature sensor in. [5] However, the use of wired external devices for the readout of sensor data is not convenient for the proposed application, especially when monitoring mobile components, such as rotor blades for small wind turbines. Consequently, wireless read-out options are desired, as implemented in the presented wireless multifunctional biface sensor tag.

Sensor Tag Design
In order to minimize the substrate material consumption, the sensor design utilized both faces of the substrate. The designs of the different components had to take into account the requirement of the chip. These are summarized in Table 1 The total area of the sensor tag was 50 × 50 mm 2 and was determined by the size required to accommodate the antenna (see Section 2.5). The sizes and designs of the three sensors were tailored to match both the antenna size and the chip's requirements in terms of impedance and footprint. Therefore, the temperature and humidity sensors were placed on the backside of the substrate. The temperature sensor consisted of 26 parallel meander-shaped tracks with a width of 250 μm, resulting in a full track length of 2.4 meters and, using silver ink, an ohmic resistance of 1 kΩ. This structure occupied about two-thirds of the area of the back side (see Figure 1) and was chosen to match the upper limit of the 250 to 1000 Ω input resistance range of the chip, where the resistivity measurements could be made with about 1% accuracy. The resistance of the sensor was maximized within this range in order not to lose any resolution of the measurement signal. The remaining area of the sensor tag's backside was used to implement a humidity sensor with a capacitance of 9.4 pF in order to match the measurement range of the chip. The humidity sensor was designed as an interdigitated comb structure with dimensions of 26.6 × 11.6 mm 2 . The twenty interdigital fingers were 800 μm wide and had a length of 10 mm. The gap between the fingers was 400 μm. Due to the electric field distribution the resulting capacitor, therefore, used the uncoated paper substrate as well the surrounding atmosphere as a dielectric [23] with relative permittivities of ≈3.7 and 1, respectively. [24] The strain sensor was placed on the front side of the substrate. Since the conductivity of the silver ink was too high to achieve sufficiently high resistance values in a strain gauge-typical structure, carbon paste was chosen for this sensor, resulting in a simple strip with dimensions of 17 × 1 mm 2 and a resistance of 250 Ω.

Antenna Design and Simulation
The ISO14443A RF interface of the SIC4340 RFID IC (radiofrequency identification, RFID; integrated circuit, IC) provides 50 pF of on-chip matching capacitance. Since there was no external matching capacitor, the corresponding antenna coil needs to have an inductance of 2.76 μH to provide matching at 13.56 MHz. To meet the goal of reading the tag with a smartphone, the size of the antenna had been chosen to be in the range of 50 × 50 mm 2 , since this roughly marked the size of smartphone antennas, which was necessary to maintain an optimal coupling factor, while providing spatial alignment flexibility. To satisfy the inductance requirement, the design resulted in six turns within the 50 × 50 mm 2 at 800 μm width and spacing between the turns. This design had been simulated with the fullwave EM analysis tool Ansys HFSS to confirm the targeted inductance of 2.76 μH. The final geometry of the NFC antenna coil could be seen in Figure 2.
As an excitation, a lumped port was applied, and a quarter wavelength radiation boundary condition was used to simulate at 13.56 MHz. The required inductance was achieved with a deviation of 0.07%. The conductivity of the conductors was 5.13 MS m −1 , i.e., a typical value obtained for inkjet-printed silver in previous experiments, while the relative permittivity and permeability were very close to 1.

Sensor Tag Fabrication
A schematic representation of the manufacturing processes is shown in Figure 3. The humidity and temperature sensors, as well as the connections and antenna structures, were fabricated using inkjet printing of an Ag-nanoparticle ink (Sicrys 150-TM119, PVNanocell, Israel) on an uncoated paper substrate. A PIXDRO LP50 (Süss Microtec SE, Germany) system with a Spectra SE-128 AA 128 (Fujifilm Dimatix Inc., USA) 30 pL print head assembly at a resolution of 400 × 400 dpi was used. Due to the absorbing nature of the porous paper substrate, two layers of ink had to be applied to achieve sufficient conductivity. To align the structures of the two faces to each other, the substrate was perforated with a 75 μm precision needle at three corners of the antenna, which was printed first. With the help of the camera integrated into the printer, the position for printing the back side structures could be found in this way. The printed structures were subsequently sintered by means of intense pulse light annealing (Pulse Forge 1200, Novacentrix) at an overall energy of 2.1 J cm −2 . Following this procedure, a sheet resistance of 60 mΩ ϒ −1 equalling to a specific resistivity = 12 μΩ cm (7.6 times the value of bulk silver) at a layer thickness of 2 μm could be achieved. [25] The antenna, temperature and humidity sensor structure were inkjet-printed with silver nanoparticle ink in two layers with a resolution of 400 × 400 dpi. The strain sensor was fabricated by means of screen printing of Carbon screen-printing paste (Dycotec DM-CAP-4511S, UK). The used screen had a mesh count of 63 threads per cm and 63 μm thread diameter, the mesh opening was 93 μm. The sensor design corresponded to a simple straight line with dimensions of 17 × 1 mm 2 and a thickness of 6 μm. To prevent the strain and temperature sensor from being affected by humidity, they had been coated with a clear hydrophobic overcoat (Dycotec DM-OC-6020S, UK) by screen printing on one specimen.

Fabrication of Through-Substrate-Vias
Due to the bifacial design, through-substrate connections had to be realized. The individual holes were manufactured using a Nd:YAG-Laser at a wavelength of 1064 nm, perforating the substrate with a diameter of ≈300 μm. To avoid burning of the 125 μm thick paper during perforation, Argon inert gas flushing was employed. The vias were subsequently filled with silver flake screen-printing paste (Novacentrix Metalon HPS-FG32, USA) using a micro-dispenser device and then thermally cured at 80°C for 30 min. Optical microscope examination confirms that there were neither fractures in the via material nor that the via was detached from the substrate. For testing of functionality for electrical connection, the contact tips of a digital multimeter were positioned at the inkjet-printed terminals of each vias on both sides of the paper substrate. Vias with resistance values below 700 mΩ were comparable to inkjet-printed structures and were classified as electrically conductive. Furthermore, such low resistances contributed <1% of the nominal value of the resistive sensors and were therefore negligible.

Hybrid Integration of Packaged Chip
In this instance, the thermos-compression flip-chip technique was utilized, resulting in a pad-to-pad connection, shown in Figure 4. The flip-chip technique was carried out by using a microassembly station (Fineplacer, Finetech GmbH, Germany). The chips (SIC 4340, Silicon Craft Technology, Thailand) were attached employing a non-conductive adhesive (Heat-curing Epoxy Delo DA-255, Germany) at a force of 15 N, resulting in a mechanical connection between the surfaces of the chip pads and substrate electrodes and realizing an electrical contact between them. The tool was first aligned to the surface plane of the chip, so the chip received the same pressure overall. Next, the pads of the chip were optically aligned with the substrate. After the electrodes were aligned, the chip was placed on the substrate and bonded with a stage temperature of 170°C.

Sensor Characterization
The temperature response of the silver nanoparticle-based temperature sensor was tested inside an environmental testing chamber (Vötsch Industrietechnik, Germany), where the specimen was first heated from 10°C to 80°C in five steps and then cooled down to 10°C. The temperature sensors were fixed in a Perspex frame, shown in Figure 5. The resistance was recorded with a 2-probe measurement setup using a digital multimeter  (Keithley 2700, USA) set to an output current of 1 mA and the temperature-dependent change in resistance R(T) was linearly approximated by with R 0 corresponding to the nominal resistance at a temperature T 0 = 10°C. is the temperature coefficient of resistivity (TCR), which is a material-dependent constant and can be either positive or negative. The strain sensor was characterized using a Zwick & Roell (Germany) load test stand using an automated cyclic test procedure with a maximum elongation of 1240 μm m −1 , at a maximum force of 7.1 N. In parallel, the resistance of the strain gauge was recorded with a two-probe measurement setup using a digital multimeter (Keithley 2700) set to an output current of 1 mA. The gauge factor (GF) was calculated from the experimental data by where ΔR corresponds to the change in resistance due to strain, R is the unstrained resistance value of the strain gauge, and is the actual strain.
Analogous to the temperature sensors, the humidity sensors were also fixed in a Perspex frame and measurements were carried out in the environmental testing chamber. The samples were exposed to 25°C and 10% relative humidity constantly for 60 min in the environmental testing chamber. Then, the relative humidity was increased in steps of 10% up to 60% and further decreased again at a constant temperature. Each humidity level was held constant for 15 and 30 min in two measurement cycles in order to achieve a stable condition and to bring the sensor substrate into climatic equilibrium with its environment. At the same time, the capacitance of the humidity sensor was measured using an impedance analyzer (Keysight E4990A, USA) set to 1 V AC at a frequency of 50 kHz according to the chip. Additionally, the complex relative permittivity of materials also showed a frequency dependency and behaved most stably at frequencies in the kHz range and above. [26] The capacitance of an interdigitated comb structure was calculated according to where N represents the number of finger electrodes and C I and C E corresponds to the internal and external capacitances of the respective capacitor regions inside and outside the substrate, which further depend on the relative permittivities of the substrate and the surrounding atmosphere. [27,28] Finally, the second multiplication term in Equation (3) resulted in a quadratic dependence of the total capacitance on the relative permittivity and, thus, also on the relative humidity. For the investigation of cross-sensitivities, the other two sensors were also measured during the individual characterization measurements. To obtain the humidity sensor's sensitivity toward strain the force was applied in parallel to the orientation of the interdigitated fingers. By analyzing the obtained data, measurement errors could further be corrected.

Wireless Sensor Measurements
Commercially available cell phones were nowadays usually equipped with NFC modules and were widely and easily available. Therefore, one (Huawei P40, China) was used to read out the response of the sensors via the chip. For this purpose, a smartphone App was adapted to control the chip used on the sensor tag. The strain gauge was tested using a Zwick & Roell tensile testing machine, reading out the resistance of the strain gauge wirelessly with the smartphone (see Figure 6). Strain values from 0 to 0.12% were applied in steps of 0.02%. All seven strain values were maintained for 30 s each and the resistance was measured five times in succession via the smartphone at a distance of 5 mm. The current was set to 504 μA at a measurement frequency of 50 kHz. This would from here on be referred to as static measurement conditions.
For testing the wireless measurement of the humidity and temperature sensors, they were exposed to changing climatic conditions in the environmental testing chamber and the signal was read out with the smartphone. Since the smartphone would not withstand the conditions in the environmental testing chamber, the customized near-field communication-based sensor platform described in [29] was employed. This platform consisted of three individual modules, namely the antenna module, the microcontroller module, and the sensor module, which provided the connectors to attach resistive or capacitive sensors to the inputs of the controller for characterization purposes. The sensor module was placed inside the climate chamber and directly connected to the antenna module that was fixed at the outer wall of the chamber. Temperature and humidity responses were recorded from the NFC tag outside of the chamber wirelessly via the smartphone without any disturbance of the climate chamber conditions.

Statistical Methods
The characterization measurements for the respective sensors were carried out for n = 1 to 3 samples. The exact sample size is mentioned in the corresponding figure legends. In the case of multiple sensor samples, average measurement values are plotted, and error bars represent the standard deviation from the average.

Results and Discussion
An example of a printed biface sensor is shown in Figure 8. The sensor tag is illuminated to provide a view of the entire structure. The front side contains the NFC antenna and the RFID chip, as well as the carbon-based strain sensor. The back side consists of inkjet-printed structures for temperature and humidity measurements. The inkjet-printed layers have an average thickness of 4 μm, while the screen-printed carbon strain gauge is 38 μm thick. The exact printing parameters for silver ink and the resulting properties of the printed layer, such as layer thickness and homogeneity, are presented in a previous publication. [25] The average nominal sensor values measured at 25°C and 30% relative humidity are as follows: 9.4 pF for the capacitive humidity sensor, 250 Ω for the strain gauge and 1 kΩ for the resistive temperature sensor. The antenna had an inductance value of 2.47 μH and an electrical resistance of 78 Ω, values that are close to the ones obtained from the simulated model. In accordance with the methods described above for testing the through-substrate vias, twenty of these were fabricated on five different sensor tags and tested for their functionality, i.e., electrical conductivity. The twenty-five samples (five vias on each of the five sensor tags under test) measured had an average resistance value of 580 mΩ, with a standard deviation of 17%. In addition, it was found, that the vias changed their resistance in the range of only a few percent during the strain tests. Thus, the electrical resistance contribution of the vias to the overall system is small and their effect is negligible. A cross-section view of silver via in a paper substrate is shown in the microscope image in Figure 7. www.advancedsciencenews.com www.advsensorres.com Wired temperature measurements in the range from 10°C to 80°C show a linear response of the sensor with a coefficient of correlation R 2 ≈ 0.997 as shown in Figure 9a. The TCR of 0.17% per Kelvin varies by 1.2% around the average value with three measuring sensors, shown here with error bars, and corresponds to the values commonly found in the technical literature for similar sensors. [30] The repeatability error is 0.2% of the nominal sensor resistance at the center of the measurement range (here equivalent to 45°C). We assume that this deviation results from thermal post-curing of the metallization layer. Figure 9b shows the temperature-dependent measurement signal (10°C -80°C) using the wireless measurement setup. Again, a linear trend is observable with a calculated TCR of 0.19% and 0.16% for rising and falling temperatures, respectively. Compared to the wired measurements presented in Figure 9a, the goodness of the linear fit is lower with coefficients of correlation R 2 ≈ 0.909 and 0.914 for rising and falling temperatures, respectively. As presented in, [29] a microcontroller-induced deviation of the measurement values from the theoretical values of up to 1.35% can be expected when measuring the temperature with a reference Pt1000. During the characterization of the printed temperature sensor, as illustrated in Figure 9b, larger deviations of up to 2.7% were observed.
The results of the strain sensor characterization are illustrated in Figure 10a. According to Equation (2), the gauge factor is GF = 2.6, comparable to commercial metallic strain gauges, which have a typical gauge factor of GF = 2. The gauge factor of GF = 2.6 is also in good agreement with values presented in the literature before. [31] The results of wireless strain measurements using a smartphone are shown in Figure 10b. The wire-bonded and wireless measurements are comparable in terms of linearity and sensitivity. Figure 11 shows the results of the measurement data for the printed humidity sensor. In accordance with the literature and Equation (3), there is a quadratic dependence of the capacitance on the relative humidity. [28] The sensor signal results from the humidity absorption of the paper substrate and the surrounding atmosphere and the resulting changes in the relative permittivities. The result in Figure 11a shows a hysteresis of 8%. In this case, the humidity was kept constant for 15 min at each measurement point. The hysteresis decreases to 1.5% for measurement intervals of 30 min (Figure 9b). This indicates that the uptake of water by the paper is faster than its release. Figure 12 shows the wirelessly measured humidity response of the capacitive sensor. The result is in good agreement with the wired measurement presented in Figure 11b with a quadratic dependence of the capacitance on the relative humidity and a similar hysteresis.
With regard to cross sensitivities, humidity, in particular, would naturally have a strong influence on the measurement signal of the temperature and strain sensors, due to a swelling of the paper fibers of the substrate. To prevent this influence, the two sensors were protected with a clear overcoat which prevents   the penetration of humidity, while the layer, which is only 30 μm thick, does not affect the performance of the two sensors.
The cross sensitivities of the strain and humidity sensor to temperature are shown in Figure 13a, while the influence of strain on the temperature and humidity sensor is shown in Figure 13b. The strain sensor is quite sensitive to temperature as Carbon is also well-known as a sensing material for the realization of resistive temperature sensors. [32] The capacitive humidity sensor shows some sensitivity to temperature as well. Although the relative humidity is kept constant during the test, the absolute amount of humidity (water vapor in air) depends on the ambient temperature and pressure. Consequently, at higher temperatures, there is more humidity present in the sensor's environment penetrating the substrate fibers of the paper substrate, which has an influence on the dielectric permittivity and hence the measured capacitance. The cross-sensitivity of the humidity sensor to strain in the direction of the electrode fingers can be explained by the elongation of the fingers which increases the overlapping regions and consequently the area of the interdigitated capacitor.

Conclusion and Outlook
This work presents the development and characterization of a multifunctional, wireless, and battery-free biface sensor tag, de-signed and manufactured with a focus on sustainability by minimizing material usage, utilization of sustainable materials as well as resource-efficient manufacturing techniques. The tag consists of three different sensors, a resistive inkjet-printed silver temperature sensor, a capacitive inkjet-printed silver humidity sensor and a resistive screen-printed carbon strain gauge. In addition, it features an inkjet-printed planar coil, which together with a near-field communication chip enables energy harvesting and wireless data transfer. The substrate's environmental impact is minimized by using uncoated paper and printing the structures on both sides to reduce the required surface area. A process for through substrate vias has been developed, which establishes electrical connections between the two sides of the substrate.
The wired characterization of the individual sensors showed performances, which are comparable to industrially manufactured sensors or similar work reported in literature. The temperature sensor showed a linear response with a TCR of 1.7 × 10 −3 per K the humidity sensor a quadratic dependence with 1.5% hysteresis from the nominal value and the strain gauge is characterized by a GF = 2.6.
Investigations of the cross sensitivities revealed that the influence of humidity on temperature and strain measurements could be avoided with a protective coating. Temperature and strain, however, showed an effect on the other parameters, which exhibit linear dependencies on those two. In order to be able to  . a) Response of the humidity and strain sensors to temperature and b) response of the humidity and temperature sensors to strain take these into account, further investigations will be necessary to reduce the cross-sensitivity or consider them with a correction algorithm. Still, the wireless read-out via the chip enables a wide range of applications for the sensor tag without the requirement for a battery. The ease of use, small weight and compactness meet the requirements for applications such as smart process monitoring or construction monitoring. This paper-based biface multi-sensor tag could be of particular importance in lightweight construction and in the case of natural fiber-reinforced composite materials, due to its sustainable fabrication. Although previous work has shown that the integration of paper-based sensors per se does not negatively affect the mechanical integrity of lightweight elements, [5] but the 3D geometry of the packaged chip could. To counteract this, bare dies or even ultra-thinned chips [33] could be employed.