Screen Printed Passives and Interconnects on Bio-Degradable Medical Hydrocolloid Dressing for Wearable Sensors

The healthcare system is undergoing a noticeable transformation from a reactive, post-disease treatment to a preventive, predictive continuous healthcare. The key enabler for such a system is a pervasive wearable platform. Several technologies have been suggested and implemented as a wearable platform, but these technologies either lack reliability, manufacturing practicability or pervasiveness. We propose a screen-printed circuit board on bio-degradable hydrocolloid dressings, which are medically used and approved, as a platform for wearable biomedical sensors to overcome the aforementioned problems. We experimentally characterize and prepare the surface of the hydrocolloid and demonstrate high-quality screen-printed passive elements and interconnects on its surface using conductive silver paste. We also propose appropriate models of the thick-film screen-printed passives, validated through measurements and FEM simulations. We further elucidate on the usage of the hydrocolloid dressing by prototyping a Wireless Power Transfer (WPT) sensor and a humidity sensor using printed spiral inductors and interdigital capacitors, respectively.

In this work, we propose the bio-degradable medical hydrocolloid dressings as a platform for screen-printed passives and interconnects and the integration of ICs, to completely replace the conventional hard printed circuit boards (PCBs). Medical hydrocolloids are flexible dressings that are medically used and approved. A hydrocolloid has a self-adhering surface on one side and a water-resistant surface on the other side, allowing an extended time of usage up to 7 days. Hydrocolloids come in different shapes, sizes and thicknesses. We physically and electrically characterize a hydrocolloid and prepare its surface before printing. A medical hydrocolloid with an Interdigital Capacitator (IDC) screen-printed on its surface is shown in Fig. 1b (photo taken in the authors' laboratory). We also verify the capability of hydrocolloids serving as substrates by designing, fabricating and characterizing printed passives (L, C, R) and integrating them in applications such as Wireless Power Transfer (WPT) and smart adhesive bandage. Appropriate mathematical models are selected and modified to aid in the design process of the printed passives [17][18][19][20][21] . The proposed hydrocolloid dressing platform solves the problems above of the current wearable technologies: it offers an excellent replacement of stiff or unreliable substrates without adding restrictions on the wearer's clothing. It is patient-friendly and practicable in the circuit fabrication process. The shape and the size of the hydrocolloids have negligible effects on the fabricated passives.

Results
physical and electrical characteristics of the hydrocolloid dressing. Before printing, it is critical to investigate and analyze the physical properties of the hydrocolloid substrate including its smoothness and thickness. Information about smoothness shows whether any surface preparation is needed, while the thickness of the substrate is required for accurate passives design. In this study, hydrocolloid dressings from 3 M (Tegaderm TM ) with planar dimensions of 10 cm × 10 cm is selected. Fig. 1c shows an SEM micrograph of the printing surface, which is the water-resistant non-sticky side of the dressing. It shows that the surface is rough with uneven micro-features in which their dimensions are comparable with the patterns appearing after applying the silver paste, shown in Fig. 1c; to print a robust and an accurate passive on it, we apply a transparent layer of Polyvinyl Acetate (PVA) with few microns thickness to fill in the pores and thus to smoothen the surface.  which was presented in our previous work 17 shows the SEM image before and after the smoothening. The substrate's electrical permittivity, measured as a function of frequency, is shown in Fig. 1e. It can be noticed that the permittivity declines as the frequency increases. The decline can be explained by the different processes involved at the microscopic level (ionic and dipolar relaxation) in determining the value of the complex permittivity and thus indicating the amount of energy lost and stored. This measurement essential for designing screen-printed inter-digital capacitors. The dielectric constant is around 4.7 at 2 MHz. printed passives fabrication and characterization. We use screen-printing as a fabrication method because it is a thick-film technology leading to low resistance traces and interconnects, and most importantly, is mass-production friendly. Serpentine resistors, Interdigital capacitors (IDC) and spiral square inductors were designed and screen-printed onto the smoothened hydrocolloid's surface. The general layout of the three passives and their design parameters are shown in Fig. 2a. Designs were created such that extreme cases of each design parameter were selected while considering the area limitation of the utilized hydrocolloid. Designs were created on a CAD tool and exposed on a mesh for printing. Hydrocolloids were smoothened by applying a transparent layer of Polyvinyl Acetate (PVA) to fill the micro-features. Conductive silver paste (Paron90) was utilized for screen-printing and Fig. 2b shows a sample of the fabricated passives. The thickness of metallization was measured by imaging a cross section of a screen-printed sample. Fig. 2c shows an SEM cross-sectional micrograph showing a metallization thickness of around 20 µm. This thickness can be mainly controlled by the mesh size or the number of screen-printing iterations.
Fabricated samples were characterized: resistance was measured for serpentine resistors, the capacitance for IDCs and inductance for spirals. These measurements were compared against FEM simulations and mathematical models' prediction for the same passives' designs considering the hydrocolloids properties.  measurements was replicated three times at 2 MHz, which is sufficiently below their self-resonance frequencies (SRF). As an example, SRF of a large inductor with N = 11 was measured to be around 58 MHz as shown in Fig. 2d.
Reliability tests were also performed on the printed passives by assessing their impedance while mechanically deformed. This test is crucial for confirming the reliability of the fabricated wearable electronics as it resembles a realistic wearable environment where convexity is present. The test was applied on three different passives (one spiral inductor, one interdigitated capacitor and one LC tank). The bending parameter, denoted herein as R, is the radius of a virtual circle created by extending the arc which is formed by the bandage convexity as depicted in Fig. 3a. Flat bandage (no convexity) has R = infinity and more convexity is indicated by a smaller R. The impedance change in the passives with respect to the change in R is shown in Fig. 3b-d for a bandage of inductor, capacitor and an LC tank, respectively. It can be seen that the pattern and behavior of the passive is conserved under the mechanical stress with a minor difference in the impedance measurement. Additionally, the measurement highlights the resonance frequencies of the inductor and LC tank; however, the resonance frequency of the capacitor is beyond the capability of the utilized instrument.
printed passives modeling and tuning. Layouts of passives shown in Fig. 2a are readily borrowed from passives printed on printed circuit boards (PCBs) in the field of microwave integrated circuits (MICs). Several models in the literature describe those passives (resistance R, capacitance C, and inductance L) in the domain of PCB and thin-film fabrication processes and it was of a target to validate if those models can be straightforwardly applied to screen-printed passives on hydrocolloids. Screen-printing usually yields structures with thicker metallic films as compared to other thin-film methods like sputtering. Both measurements and FEM simulations were utilized as validators for the best model that can be used for our purpose.
Resistance. Serpentine resistors are easily modeled using their design parameters by the Eq. (1) below: where l, b, w and t are length, number of bends, width and, thickness respectively. www.nature.com/scientificreports www.nature.com/scientificreports/ Capacitance. In screen printing technology, l and w are in the range of mm while t is in µm. Thickness of the substrate can greatly affect the capacitance. The capacitance increases with the substrate thickness; however, in case the substrate t is much greater than the finger w of the IDC, its capacitance saturates. Capacitance of IDCs is due to the adjacent interdigitated fingers (C N ) and due to the fringing fields at the end of each finger (C end ), as shown in Fig. 2a. The first model in literature that described IDCs is by Alley et al. 18 which is insensitive for cases where w ≠ s, while Bahl el al. 's model 22 always overestimates the capacitance value according to simulations. Igreja-Dias et al. 19 and Gevorgian et al. 23 models were found to underestimate capacitance value though the latter considers all capacitive effects. Our proposed capacitance model for IDCs on hydrocolloids is based on the model in 19 for calculating C N , but additionally considering two very important effects: the non-negligible thickness (t) of the screen-printed IDC and the capacitance due to the fringing fields at the end of each finger (C end ) 24 . The finite thickness t is taken into account by introducing the effective width in C N : The proposed model is now a function of C N and C end and is described by: proposed N e ff r e nd Table 1 summarizes the accuracy of those models with respect to FEM simulations regarding the Mean Percentage Error (MPE).
Inductance. Three models describing square spiral inductors have been compared to FEM simulations considering hydrocolloid substrate. Table 2 summarizes the performance of the three models concerning the wide range of simulated spirals. Corrected Current Sheet (CCS) model shows the best accuracy with Mean Percentage Error (MPE) of around 6%. The quadratic dependence of inductance on N was found to overestimate the inductance considering the layout shown in Fig. 2a; this is because the contribution of the first turn is always incomplete. The proposed model suggests utilizing the CCS model while changing N 2 dependence to N 1.97 .
where d avg and ρ are the average diameter and fill ratio respectively, L correction is a correction term to account for the finite thickness of the spiral and c 1 , c 2 , c 3 and c 4 are constants equal to 1.27, 2.07, 0.18, 0.13 respectively 20 .
Wireless power transfer using printed spirals on hydrocolloid. The design and characterization of printed spirals on hydrocolloid have been thoroughly explained in our previous work and summarized in Fig. 4  www.nature.com/scientificreports www.nature.com/scientificreports/ parallel position. It can be observed from Fig. 4b that the octagonal inductor exhibits the best performance with the efficiency of around 20% at 5 mm distance. For square inductors, the higher the value of the inductance the better the performance. To further improve the performance, resonance mechanism has been examined in which the inductor is attached to a capacitor forming an LC tank with a controlled operating frequency. A 22 pF capacitor in hydrocolloid was selected and attached to both L 1 and L 2 . The resonance frequency was experimentally measured for both tanks as shown in Fig. 4c. The measurement shows that the octagonal inductor has a lower resonance frequency and a sharper peak indicating a higher Q-factor. It is shown in Fig. 4d that repeating the power transfer experiment at a power carrier of V pp = 5 V boosted the performance of both inductors, L 1 to 20% while L 2 to 40%. Fig. 4e shows the experimental setup that was used while conducting the WPT measurements.

Smart adhesive bandage using printed iDc on hydrocolloids.
We have also demonstrated the possibility of fabricating a smart bandage that is capable of detecting the level of humidity for wet wounds. A 43 fingers IDC was screen-printed and tested as a sensing element for humidity; Fig. 5a. Humidity was controlled by adding drops of water on the surface of the IDC and measuring the capacitance value after each droplet. The nominal capacitance value was measured to be around 44.9 pF, and each droplet contributed ~4-5 pF additional capacitance. The accumulation of water drops on the top of the printed IDC alter the effective permittivity and thus, capacitance changes by up to 105% for 11 drops of water as shown in Fig. 5b. The impedance change is also due to the change in the fingers conductance attributed to the ionic conduction, however in our application, we are interested in the capacitance change solely 25 . This screen-printed sensor was integrated with off-the-shelf components such as resistors, capacitors, a 555-timer, and a counter to form a complete circuitry for humidity sensing. The sensed capacitance is connected to a 555-timer taking part in controlling the output frequency of the timer. This frequency is then divided using an 8-bit counter reducing the frequency for easy visualization and hence, an LED can be used. The higher the humidity, the higher the capacitance and hence the less the output frequency and the less the flashing frequency. Connection points and traces were carefully designed on AutoCAD considering the dimensions of the ICs and their pins. The design was screen-printed, and off-the-shelf components were attached using the same silver paste used in printing. Fig. 5c shows the screen-printed design before placing components. When water drops are placed on the surface of the IDC, we can observe the frequency changes accordingly. For high levels of humidity, the LED saturates to ON state. Fig. 5d shows the smart adhesive bandage while in operation.

Discussion
parameter space of the selected mathematical models. Parameters space defines the range of values for each design parameter for which a particular mathematical model is valid. Parameters space can be bound due to some assumptions the model is built on or due to physical limitations. In the context of passives printed on hydrocolloids, physical limitations can be due to the screen printing technology and the size of the hydrocolloid itself. The Minimum Feature size that can be achieved using a typical screen printer is around 200 um ∼ 500 um, www.nature.com/scientificreports www.nature.com/scientificreports/ while the size of the square hydrocolloid patch under this study is 10 cm × 10 cm. Igreja-Dias model of IDC was selected and tuned to serve as an accurate model for IDCs printed on Hydrocolloids. The model was developed on the basis of a few assumptions that will define the parameter space of the model: 1. The periodical structure of fingers is infinite. 2. Length of fingers is infinite or much larger than the feature size of the IDC λ > + = l 2(s w) .
These two assumptions indicate that potential lines are assumed to be perfectly flat in between the adjacent differently polarized electrode fingers as shown in Fig. 6a. Also, the model is developed for IDCs where N > 3. Taking these assumptions into consideration, the parameter space of the IDC model is: where N max can be calculated for a given λ and η, or in other words, for a given s and w given the equation below: Mohan's corrected current sheet approximated model 20 was selected after tuning to serve as an accurate model for square spirals printed on hydrocolloid. The current sheet approximation, though corrected, has a limitation on s w ratio, where it is not preferable to go beyond 3. Additionally, the model assumes the length of every straight conductor must be larger than its width. The smallest conductor length of a square spiral is equal to the inner diameter, d in . Considering these two assumptions, the model's parameter space is: www.nature.com/scientificreports www.nature.com/scientificreports/ where N max is largest possible number of turns giving s and w. N max can be realized as follows: Fig. 6b is an example design plot for a hydrocolloid generated using a 10 cm × 10 cm size. This contour plot studies the relationship between λ in mm, η and capacitance density in pF/mm 2 . It is clear that there are many realizations for the same value of capacitance density. However, there can be some factors that will make some realizations better than the others as mentioned in 19 . Capacitance density is used instead of capacitance to remove any area-related constrains. Similarly, it is possible to generate a design contour plot for the inductance in Henry as shown in Fig. 6b . As the spiral inductor is square in shape, it is expected to allocate an area for the inductor such the L = W. www.nature.com/scientificreports www.nature.com/scientificreports/ Finally, we demonstrated the use of medical hydrocolloids to serve as a platform for printed passives and electronics integration in the wearable electronics domain. Medical hydrocolloids are adhesive, flexible and medically approved. Hydrocolloids were physically and electrically characterized prior to printing. Several models for planar passives were explored and compared to COMSOL simulation results while considering the hydrocolloid properties. It was found that Igreja-Dias model 19 after the proposed modification serve as an accurate model for printed IDC while the Current sheet approximation by Mohan 20 was found to be an accurate model for square printed spiral inductors after including the proposed modifications. Proposed models gave mean percentage error of <6% as compared to measured fabricated samples. Two demonstration has been discussed in this work: IDC as a humidity sensor and spiral inductors for WPT. Lastly, limitations of the proposed models have been discussed.

Methods
Hydrocolloid surface characterization. Quanta 250 SEM by FEI was used to acquire several micrographs for the surface of the 3 M Tegaderm hydrocolloid. The surface micrographs were taken under specific conditions of beam energy, chamber pressure, and magnification. It can be noticed from Table 3 that the beam energy used is relatively low to prevent charges from accumulation as the hydrocolloid is a non-conductive material. Higher energy beam however was used for observing the flake size of the silver paste.
The cross-sectional micrograph needs more powerful SEM and thus used Nova NanoSEM 650 by FEI. This microscope provides higher magnification and enabled us to measure the thin layer of printed silver which was approximated to around 20 µm. The beam energy used in acquiring Fig. 2c is 1 kV at a magnification of x690. passives fabrication. Fig. 6c shows the methodology followed in this work to fabricate the passives on hydrocolloids. All passives in this work were fabricated and printed on top of square 3 M Tegaderm Hydrocolloid dressings with a side length of 10 cm and thickness of around 1 mm. Prior to printing, hydrocolloids were laminated with a thin layer of Polyvinyl Acetate (PVA) which is dense enough to smoothen the un-even features of the surface while being sparse enough to maintain flexibility. Paron-910 silver paste which is distinguished by its low resistivity (15~50 mΩ/ϒ/mil) was utilized in screen-printing the passives on top of the prepared hydrocolloids. Silver paste flakes were characterized and their size was examined by Quanta250 SEM as shown in Fig. 6d. All passives were first designed on AutoCAD and exposed on a mesh with an opening of size 90 µm. A 1000 Watts mercury lamp was used to expose designs on the emulsion layer. Printing ink reducer by Sunglo is mixed with the silver paste to ease the process of printing and to reduce the amount of paste sticking on the mesh surface. The fabrication process is illustrated in Fig. 6e. passives feM simulations. COMSOL simulations were conducted spanning wide range of IDC and spiral inductors geometrical configurations. These simulations were specifically adapted for the hydrocolloid's characteristics. As shown in Fig. 7, the simulation environment consists of 3 geometrical elements: IDC/Spiral structure, the hydrocolloid itself and the surrounding air. Thickness of the hydrocolloid was set to 1.155 um as reported earlier. Simulation was run for 45 structures of IDCs and 45 structures of spirals which their geometrical parameters are listed in Tables (4,5).
It was found that both, the size of the air box and the thickness of the hydrocolloid have noticeable effect on the simulated capacitance and inductance. Hydrocolloid's thickness was already measured under the Scanning Electronic Microscope (SEM) while the size of the air box was set to the minimum size possible after which the simulated capacitance/inductance doesn't change. This is mainly because we aim to reduce the time of meshing and hence the computation time.
Materials were assigned to each of the three geometries: air, silver paste and hydrocolloid. Hydrocolloid is not available in the existing materials library, therefore, a new material was created. Electrical properties of the hydrocolloid were provided as part of the material's definition. Table (6) below summarizes electrical properties of each of the materials that were used during simulations. The Electric Currents (ec) and the Magnetic Field (mf) submodules under the AC/DC module were used for running all IDCs and spirals simulations. Meshing is physics-defined and is usually set to Finer or Extra Fine depending on the structure. Fig. 7 shows COMSOL Multiphysics 3D geometry for both, the IDC and the spiral inductor. electrical characterization of passives. Printed resistors, IDCs, and spiral inductors were characterized by E4980A LCR meter from Keysight to measure their resistance, capacitance, and inductance respectively. The tool was calibrated for open and short circuit tests, and measurements were conducted at a frequency of 2 MHz and the voltage level of 1 V peak to peak. LCR meter ports were hooked to two XYZ 500TRS micro-positioners from Quarter Research and needles of 3.5 µm tip diameter from Signatone for stable and accurate measurement. Needles tips were bent to avoid the destruction of the soft hydrocolloid surface and the fabricated passive. The permittivity measurement, shown in Fig. 1e, was performed on a hydrocolloid sample through the self-parallel plate capacitance in which we cover both sides of the hydrocolloid sample with silver paste forming a parallel plate   www.nature.com/scientificreports www.nature.com/scientificreports/ capacitor. The capacitance of this parallel plate was measured as we scan the frequency up to 2 MHz and the permittivity was calculated using the dimensions of the sample, both the area (6.15 cm × 5.3 cm) and the thickness (1.155 mm). On the other hand, the resonance frequency measurement shown in Figs. 2d and 4c was obtained using E4990A Impedance Analyzer from Keysight. The tool was calibrated, and the same setup of probes and needles was used. Frequency scan was set between 500 kHz −83 MHz with a frequency step of 63.460 kHz. For reliability tests, passives where measured using Keysight 4294 A Impedance Analyzer having a frequency limitation upto 110 MHz.
Wireless power transfer measurements. This experiment required a pair of inductors at a time for sending and receiving power. Four pairs of printed spiral inductors were fabricated, and their dimensions and geometry are shown in Fig. 4a. Both, the TX and the RX inductors were fixed perpendicular to the surface and parallel to each other. The TX inductor was connected to a Tektronix AFG3022B signal generator, and wires were stuck to the TX inductor pads by carbon tape which is intrinsically conductive. The RX inductor pads, on the other hand, were connected to DSOX3034A Digital Oscilloscope to monitor the wirelessly received signal. Each of the fabricated pairs of spiral inductors has been tested under inductive coupling mechanism using a power carrier of   while being exactly parallel to each other. The distance between both inductors was varied between 5 mm-70 mm. The magnetic coupling experiment for L 1 and L 2 on the other hand were conducted after attaching a discrete capacitor of 22 pF to both the TX and RX inductors. To teste each LC tank pairs on their own resonance frequency, the resonance frequency was measured for each of the tanks (parallel connection) using E4990A Impedance Analyzer. The WPT experiment was conducted while the LC components are connected in series.
Smart adhesive bandage measurements. The circuitry of the hydrocolloid-based smart bandage made use of two ICs: NE555P timer from Texas Instrument and 74LS393 dual 4-bit binary counter from Fairchild. Two 10 MΩ resistors and a 220 nF capacitor were also used for a successful operation of the timer. This circuit was powered by a 5 V from E3648A DC power supply by Agilent. This can be potentially replaced by a disc or flexible battery for easy integration and portability.