Method for Inkjet-printing PEDOT:PSS polymer electrode arrays on piezoelectric PVDF-TrFE fibers

Abstract

We present a method for printing conductive polymers onto P(VDF-TrFE) nanofibers to create all-polymer piezoelectric devices. Inkjet printing is an attractive fabrication approach for rapid prototyping of flexible electronics, but until now with limited applications in developing P(VDF-TrFE) nanofiber-based devices. We have demonstrated an approach to infill the void space within a piezoelectric nanofibrous matrix to allow for the inkjet printing of aqueous inks while avoiding leakage that typically leads to electrical shorting and without significant loss of voltage output. This was done using a diluted PDMS solution and a commercially available conductive ink. The 1 cm² devices showed a 254 mV/N sensitivity to impact as well as a sensitivity to bending. The device was shown to be able to detect breathing and pulse rate when placed superficially to the carotid and radial arteries. Using these techniques, flexible piezoelectric sensing can be done in an array format, shown with applications in foot movement sensing.

I. Introduction

AS wearable and implantable technologies for health monitoring have become more popular, functional materials such as piezoelectric and conductive polymers have been developed as building blocks for more flexible and conforming devices. Due to its outstanding flexibility while maintaining a great piezoelectric constant, poly(vinylidene fluoride-cotrifluoroethylene) (PVDF-TrFE) is one of the most studied piezoelectric materials. It is often fabricated in a nanofibrous structure due to its superior voltage output when compared to thin films [1].

Inkjet printing is an attractive fabrication approach in the field of flexible sensing, which allows for low-cost, rapid prototyping, and scalable fabrication of many of these devices. The printing approach can provide large area fabrication and is being developed for continuous high-speed process in roll-to-roll fabrications of electronics [2]. One of the key benefits of inkjet printing is its ability to rapidly iterate through many patternable designs. Fabrication of flexible piezoelectric devices in a patternable, array format could lead to the development of improved physical biomarker sensors, as similar non-printed array structures have been shown to detect pulse wave velocity [3]. Poly(3,4-ethylenedioxythiophene) polystyrene sulfonate (PEDOT:PSS) is a conductive polymer that has been of great interest due to its high conductivity, ease of use, and flexibility [4]–[8]. It has recently been developed as an electrode for bioelectronics that shows great stability [9], [10]. It has been inkjet printed as a conductive layer in many technologies, including on PVDF thin films [11], [12]. However, the porous nature of the PVDF-TrFE nanofiber matrix causes leakage of the conductive layer through the fibrous matrix, leading to electrical shorting. To circumvent this, we developed a method for filling polydimethylsiloxane (PDMS) in the void spaces within the PVDF-TrFE nanofiber matrix to prevent ink leakage. The method has enabled the fabrication of all-polymer arrayed piezoelectric devices using inkjet printing.

In this study, we first develop a method for inkjet printing conductive polymer electrodes onto piezoelectric polymer fibers. We then characterize the devices fabricated with this method, including single element and array structured devices. Finally, we show applications made available by these new all-polymer piezoelectric devices, including those for cardiovascular monitoring and foot pressure detection.

II. Materials and Methods

The key development to allow for inkjet printing on PVDF-TrFE nanofibers and the fabrication of the devices studied here is the infill of the void spaces in the PVDF-TrFE nanofibrous matrix. Fig. 1a shows a scanning electron microscopy (SEM) image of typical sub-micron diameter PVDF-TrFE nanofibers and the high level of porosity is evident. The general approach for fabrication is to infill these void spaces with a diluted polydimethylsiloxane (PDMS) solution (Fig. 1b), which allows for inkjet patterning of conductive polymer (PEDOT:PSS) and the formation of the basic all-polymer piezoelectric device (Fig. 1c). In Fig. 1b, a cross sectional SEM image of the PDMS infilled PVDF-TrFE nanofiber matrix shows that the PDMS infill fills the gaps between the fibers, creating a solid film with embedded fibers protruding from the freeze fractured cross section. The schematic for the fabrication process of the all-polymer piezoelectric devices is demonstrated in Fig. 2 with step 1) electrospinning of PVDF-TrFE nanofibers, step 2) spin coating of diluted PDMS, step 3) inkjet printing of conductive PEDOT:PSS electrode layer, and step 4) PDMS encapsulation of the all-polymer piezoelectric devices.

Fig. 1.

Schematic overviewing the flexible all-polymer printed piezoelectric device. (a) SEM image of PVDF-TrFE nanofibers. (b) SEM image of a cross-section of the PVDF-TrFE/PDMS composite where the PVDF-TrFE fibers protruding from the PDMS is evident. (c) Cross-section schematic of the final printed device structure with printed PEDOT:PSS electrode layers on the PVDF-TrFE/PDMS composite substrate, encapsulated in PDMS.

Fig. 2.

Schematic showing the fabrication steps of the all-polymer piezoelectric device. The first step is to electrospin PVDF-TrFE nanofibers, followed by spin coating of a filler layer of PDMS diluted in TBA, and finally, inkjet printing of conductive polymer, PEDOT:PSS, electrode layers. The final device’s electrode layer can be easily patterned into complex designs through this process.

A. Electrospinning of PVDF-TrFE Nanofibers

In step 1, PVDF-TrFE nanofibers were fabricated through a standard electrospinning process. In brief, a solution of 18 wt% 70:30 PVDF-TrFE (Piezotech) was mixed into a 1:1 ratio of dimethylformamide (DMF) and methyl ethyl ketone (MEK). The solution was electrospun using a 20-gauge needle, a flow rate of 600 μl/hr, and an applied voltage of 12kV. A rotating drum collector was used to create aligned fibers [1], [13], [14], at a distance of 9 cm from the needle, and a rotation speed of 3,000 rpm. The needle was mounted on a linear motor to create a more highly uniform fiber mat. Due to the in-situ poling, no further electrical poling or thermal annealing processes were required [15], [16]. The electrospinning process forms fibers from 500 nm to 1 μm in diameter, with large bulk alignment [1]. The as-electrospun nanofibers were characterized using X-ray diffraction (Rigaku) in Fig. 3a. The key peak at 19.8° represents the piezoelectric β-phase of the semi-crystalline polymer, PVDF-TrFE [17], [18].

Fig. 3.

Characterization of PVDF-TrFE/PDMS substrate used to allow for inkjet printing of PEDOT:PSS electrodes to form all-polymer piezoelectric sensors. (a) XRD pattern of the as-electrospun PVDF-TrFE nanofibers with a peak at 19.8° representing the piezoelectric β-phase. (b) A comparison of SEM cross sectional images of PVDF-TrFE nanofibers with spin cast PDMS (top) and a 1:1 PDMS:TBA dilution (bottom). With the diluted PDMS the fiber’s porous matrix is infilled without forming an additional excess PDMS layer. (c) A stress-strain curve comparing the mechanical properties of the PVDF-TrFE fibers (black), PVDF-TrFE fibers infilled with PDMS:TBA (red), and pure PDMS (blue). (d) Dielectric constant of PVDF-TrFE nanofibers infilled with 4:0, 3:1, and 2:2 dilution of PDMS:TBA over a range of 20 Hz to 20 kHz. The dielectric constant increases with more dilute PDMS.

B. Diluted PDMS Infill

Initial attempts to print conductive layers onto the PVDF-TrFE nanofiber substrates to form electrode-piezoelectric-electrode sandwich structured devices, led to electrical shorting across the parallel electrodes. The shorting occurred even in thin, single layered prints. The electrical shorting was due to the leakage of the ink through the porous structure of the PVDF-TrFE fiber matrix. To counteract this leakage, in step 2, a filler layer of spin-cast PDMS was used.

To reduce the thickness of this filler layer and to avoid signal loss due to the increased insulating layer, by increasing the resistance for charge transfer across the mat [19], the PDMS was diluted in tert-Butyl alcohol (TBA) [20]. 1.5 mL of a 4:0, 3:1, and a 1:1 solution of PDMS diluted in TBA was dropped onto the PVDF-TrFE nanofiber substrate and allowed to settle for 1 minute. The fibers were then spin coated for 1 minute at 5000 rpm. The PVDF-TrFE/PDMS composite was then heated in the oven for 3 hours at 60 °C to cure the PDMS. The final PVDF-TrFE/PDMS composite forms a transparent film, changing from the standard opaque white color of electrospun PVDF-TrFE nanofiber mats. Similar spin-cast PDMS layers that have been used to infill porous PVDF thin films have been shown to enhance piezoelectric output due to a higher incompressibility of the infill materials [21]. Fig. 3b shows the effect that PDMS dilution in TBA has on the excess filler layer by examining SEM cross section images of fibers infilled with non-diluted (4:0) and diluted (2:2) PDMS:TBA dilution. The SEM cross sectional image of the diluted PDMS:TBA infill shows a fiber matrix that has been filled in to form a solid film with embedded fibers. In the SEM cross sectional image of the non-diluted PDMS, we see an additional PDMS film layer of ~20 μm form. This is due to the increased viscosity of the non-diluted PDMS. In the 2:2 PDMS:TBA dilution the decrease in viscosity leads to thinner films of PDMS during spin coating. The measured thickness of the fibers infilled with 4:0, 3:1, and 1:1 PDMS:TBA solutions were 75, 60, 45 μm, respectively.

The mechanical properties of the fiber substrate made with the diluted PDMS were investigated using a tensile test (Instron), the results are shown in Fig. 3c. The tensile extension curves are also compared with that of the PVDF-TrFE fiber matrix and a pure PDMS thin film of similar thickness. As expected, the material made with diluted PDMS was more highly comprised of the stronger PVDF-TrFE nanofibers than the pure PDMS material and thus had a larger Young’s modulus. Due to the addition of the high-strain capable PDMS infill, the composites were able to undergo much larger strains than that of the regular PVDF-TrFE fiber matrix.

Using the lower viscosity PDMS dilution as an infill also effects the electrical properties of the infilled fibers. The dielectric constant of the fibers infilled with 4:0, 3:1, and 2:2 PDMS:TBA dilution were measured over a frequency range of 20 Hz – 20 kHz (IET Labs). The 2:2 dilution had a dielectric constant of 6.32, decreasing to 5.34 and 4.72 for the 3:1 and 4:0 dilutions, respectively. PVDF-TrFE typically has a higher dielectric constant of 13 [22], and the additional thicker layers of PDMS decreases the composites to approach that of pure PDMS of 2.4 [23].

C. Inkjet Printing

The conductive polymer electrodes were inkjet printed using a Fujifilm Dimatix DMP-2850 and a 10 pL nozzle head with 0.8 wt% PEDOT:PSS ink (Millipore-Sigma) with a drop spacing of 20 μm. The ink was first sonicated for 10 minutes at room temperature and then any aggregates were removed using a 40 μm filter. Nozzle voltage was set to 40 V and the substrate was heated to 50 °C during printing. Before printing the surface of the substrate was treated using an oxygen plasma (Harrick Plasma) to improve its surface properties. A handheld laboratory corona discharge treater was also shown to work (Electro-Technic Products). The effects of printing with and without oxygen plasma treatment can be seen in Fig. 4a. Without any treatment the surface shows very poor wettability and the aqueous ink forms many individual droplets and does not form the desired continuous film layer. After printing of the top electrode layer, the device was dried on a hot plate at 60 °C for 30 minutes. The device was then flipped over, and the same plasma treatment and printing steps were repeated to form the bottom electrode layer. The electrode patterning could be easily and rapidly tuned using the printer, allowing us to print any design of interest. For initial characterization we started with a simple 1 cm² square device and a 2 × 2 array structure of 1 cm² square elements. The entire device was finally coated in a layer of PDMS to serve as an encapsulation layer.

Fig. 4.

Characterization of the inkjet printing of PEDOT:PSS electrodes on PVDF-TrFE/PDMS substrate. (a) Shows PEDOT:PSS inkjet printed on to a standard PVDF-TrFE/PDMS substrate (left) and with oxygen plasma treatment and the formation of the desired film structure (right). (b) (left) Optical image of the lowest width PEDOT:PSS line capable with the inkjet printer/substrate. (right) The print design for testing the resolution of the printer, down to a single-drop width. (c) Plotting resistance/cm of various printed layers of PEDOT:PSS. Each layer adds 200 nm to the electrode structure. After 10 layers, the printed electrodes drop from over 1 kΩ/cm to 71Ω/cm. Inset image shows the printed pattern for the test electrode. (d) Testing a 2 cm × 0.25 cm PEDOT:PSS electrode resistance over 30,000 bending cycles undergoing a 1.3 cm linear deflection.

III. Results

A. Inkjet Patterning of PEDOT:PSS Electrodes

The PDMS infill procedure was developed to allow for successful patterning of conductive polymer electrodes by inkjet printing on piezoelectric polymer nanofibers. We then characterized the process by determining the printer’s resolution, and the conductivity and robustness of printed electrodes.

The printing resolution of the electrode ink on the PVDF-TrFE/PDMS substrate was characterized using an optical microscope to measure line width in accordance with droplet width. We printed an array of lines with increasing width, starting with a width of 1 drop and increasing to 20 drops. After printing we then imaged the printed lines to show that all the printed lines remained continuous, demonstrating that with the PEDOT:PSS ink we have printer resolution of at least a 1 drop width (~100 μm), Fig. 4b. This gives a boundary condition to pattern development, with PEDOT:PSS electrode traces to be a minimum of 100 μm in width.

The conductivity of the printed electrodes was characterized in a per layer basis. The resistance of the printed PEDOT:PSS was measured across a 1 cm gap with increasing number of printed layers in 3 separate printed electrodes (Fig. 4c). Each layer was allowed to dry for 5 minutes and the final printed film was dried on a hot plate at 60 °C for 30 minutes before testing using a 2-probe multimeter (Agilent). The thickness of each additional printed layer was found to be 200 nm using interference microscopy (ADE Phase Shift MicroXAM-100 Interferometric Surface Profiler), for a total thickness of 2 μm in a 10-layer electrode. The single layer of PEDOT:PSS had a resistance of 1,096 Ω/cm while 10 layers reached 71 Ω/cm. Fig. 4c shows that at 5 layers, 146.2 ± 1.3 Ω/cm, the resistance no longer is decreasing as drastically with each successive printed layer, so the rest of the devices we printed for testing used 5-layer electrodes.

The robustness of the printed electrodes was tested under long-term cyclic bending in Fig. 4d. A 2 cm × 0.25 cm electrode was mounted to a stepper motor and underwent a 1.3 cm linear deflection at a rate of 0.5 Hz. The resistance was measured periodically and the average percentage change in resistance across the 2 cm electrode was plotted against the number of bending cycles. In the first 10,000 cycles there is very little change in resistance. The resistance began to drastically increase after 30,000 cycles after which it was considered that the device had failed, and measurements stopped. The cyclic testing shows as a proof-of-concept that the printed polymer electrodes can undergo many deformations without severe degradation in performance.

B. Single Element Piezoelectric Device Characterization

To determine the viability of piezoelectric sensor devices created through this inkjet printing process we printed a 1 cm² square shaped electrode (schematic seen in the inset of Fig. 5a). For comparison we fabricated a standard PVDF-TrFE nanofiber device of the same size with magnetron sputtered gold electrodes (Cooke) and no PDMS infill. Fig. 5a shows that a 1.25 N impact force at 3 Hz on the inkjet printed PEDOT:PSS device has a similar response as the standard gold electrode device of ~0.8 V peak-peak, demonstrating that the printed polymer electrode piezoelectric devices could show performances similar to that of standard metallic electrode piezoelectric nanofiber devices.,

Fig. 5.

(a) Voltage output from a 1.25 N force of a 1 cm² PEDOT:PSS inkjet printed device compared to a standard gold electrode device. (b) Test to determine relationship between voltage output of the PEDOT:PSS inkjet printed device and force. Two fabrication lots of fibers were tested using two 1 cm² devices from each lot. Linear fits of the slope for device 1 and 2 of fiber lot 1 were 109 and 96 mVpp/N, respectively, and for devices 3 and 4 of fiber lot 2 were 254 and 256 mVpp/N, respectively. Showing that within a fiber lot devices perform similarly but due to lot variations, calibrations will be needed in the development of future force sensors using these materials. The standard deviation of the measured voltage values is represented for each device by the shading surrounding each line, while the standard deviation of each applied force is represented by the horizontal error bars. (c) Testing the charge output of the device over long-term impact. The output showed good stability. (d) Testing the effect of bending on the charge output of the device from 2 mm to 12 mm of linear displacement. The device showed an increasing response with increasing bending length with a response of 0.687 pC/mm of linear displacement. (e) Testing the charge output of the devices over long-term bending. The output showed good stability but was noisier than that of the long-term impact testing.

To determine the printed PEDOT:PSS single element’s voltage sensitivity to force inputs, we measured peak to peak voltage outputs over a variety of forces from 0–7 N from direct impacts with a 1 cm² square impact area at a frequency of 1 Hz. The testing included two devices from two different fabrication lots of the PVDF-TrFE nanofibers. Device 1 and 2 were from fiber lot 1 and device 3 and 4 were from fiber lot 2. The testing set up consisted of a shaker (2025E from the Modal Shop) for controlling of impact force and frequency, a force transducer (PCB Piezotronics), a fixture frame for mounting the devices, and a low-noise voltage preamplifier (Stanford Research Systems) with a gain of 1 and a low-pass filter cutoff of 30 kHz. Fig. 5b shows the voltage response of all 4 devices. From the plot we see that the devices from the same fiber lot performed similarly, but there is a significant difference between fiber lots. A linear fit of the response of device 1 and 2, from fiber lot 1, was 109 mVpp/N and 96 mVpp/N with an R² of .924 and .921, respectively. A linear fit of the response of device 3 and 4, from fiber lot 2, was 254 mVpp/N and 256 mVpp/N with an R² of .971 and .959, respectively. These results suggest that variations in lot-to-lot fabrication of the PVDF-TrFE nanofibers have a significant impact on the output of the force sensors, which will thus require a calibration step in order to accurately detect force. The piezoelectric response of one of these sensors was tested over a longer term of 1,000 cycles at a 1 Hz frequency and showed reasonable consistency (Fig. 5c).

To test the effect of bending on the piezoelectric response of the devices, a device with a 2 cm² substrate and a 1 cm² active area electrode was clamped to a fixture and the shaker was used to control linear displacement, causing the device to bend. The device was connected to a charge amplifier (Measurement Specialties) with a 100 pF feedback capacitor and a gain of 1. The signal from the amplifier was digitized by an analog to digital convertor (National Instruments) and processed in LabView (National Instruments). In Fig. 5d the 5 different compressive linear displacements tested, 1, 2, 4, 8, 12 mm are shown with their representative charge outputs. The charge output increased with increasing linear displacements to a maximum of 8.9 pC. A linear fit of the response of the device was found to be 0.687 pC/mm of linear displacement with an R² of .958. The response to bending was tested over the long-term of 800 cycles at a rate of 2 Hz (Fig. 5e) and held steady around 8 pC charge output at a linear displacement of 8 mm.

C. Application: Physiological Signal Monitoring

The developed method allows us to fabricate all-polymer piezoelectric devices that are responsive to both bending and impact forces. The mechanical flexibility of the devices makes them an ideal candidate for biomedical applications, such as measuring pulse rate from near surface arteries. As a proof-of-concept a 1 cm² device was tested on a healthy 27-year-old male, in accordance with Study00032043 approved by the Dartmouth College Internal Review Board.

The left carotid artery was the first location tested, where the device was mounted with Kapton tape to the skin superficially to the artery. The mounted sensor had no external pressure applied other than the mounting tape. Outputs from the device were passed through a charge amplifier with a 100 pF feedback capacitor, and a low pass filter of 10 kHz. The amplifier was always set to a 40 dB gain and the data was scaled back during post processing. The test subject was in a relaxed, seated position and was asked to avoid movement and to perform various types of breathing exercises.

In Fig. 6a we can see the effect of different types of breathing on the output of the device on the carotid artery. In Fig. 6a (i) we see the output from heavy breathing after exercise, while in (ii) we see the output from normal breathing. The shape and frequency of the waveform during normal breathing is similar to that of heavy breathing but of lower amplitude. The breathing rate is approximately 0.33 Hz during normal breathing and slightly faster during heavy breathing.

Fig. 6.

(a) The effects of breath on the inkjet printed device placed superficially to the carotid artery. (i) Heavy breathing shows a similar waveform but higher amplitude when compared to (ii) normal breathing. (iii) When the breath is held, the pulse waveforms become apparent and a large spike in output is seen after a gasp for breath. (b) We can see the output from both an inhale and exhale of each breath, as well as make out the pulses from each heartbeat.

We also see during both heavy and normal breathing that there are smaller peaks appearing at a frequency near 1 Hz that appear to correspond with the heart pulse. This is verified in (iii) in which the breath is held, and the smaller amplitude peaks still appear at a nearly 1 Hz frequency (60 beats per minute), in the healthy range of the average adult. In (iii) we also see the change in output of the device during a gasp after the breath is held.

Fig. 6b shows a zoomed in version of the normal breathing in Fig. 6a (ii). We can see that at this breathing frequency each breath has approximately 3 heart beats and we can distinctly tell the difference between an inhale and exhale. The inhale is comprised of a lower slope increase, peaking at 0.3 pC, while the exhale is a much sharper decline, falling to −0.3 pC.

The detection of the pulse waveform from the carotid artery is of great interest as the waveform can offer many clinically relevant measurements [24], [25]. Due its importance, the radial artery was also tested for the ability to detect the pulse waveform. Fig. 7a is a schematic showing the locations of the sensor placement, the left carotid artery and the right radial artery.

Fig. 7.

(a) Schematic showing the placements of the inkjet printed devices for pulse detection, the left carotid artery and the right radial artery. (b) (i) Pulse signal from the carotid artery with a pulse rate of approximately 60 beats per minute. (ii) A single pulse over a 1 second time span from the dashed-box in (i). The dicrotic notch is shown in the shaded oval. (iii) Pulse signal from the radial artery with an applied pressure from a wristband. The sample is much noisier than the carotid artery signal. (iv) A single pulse over a 1 second time span from the dashed-box in (iii). The dicrotic notch, in the shaded oval, is not as distinguishable as in in the carotid artery signal.

For the carotid artery, further analysis of the pulse output over a 10 second window in Fig. 7b (i) again verifies the pulse rate of approximately 60 beats per minute. Further analysis of a single pulse waveform over a 1 second time window is shown in (ii). The full pulse waveform can be seen, with the maximum charge output due to the systolic pressure and the minimum charge output due to the diastolic pressure. The sensor is sensitive enough for the detection of the separation of systolic and diastolic phase, marked by the dicrotic notch.

The device was also tested on the radial artery, which has a noticeably weaker pulse. The use of a wrist band was needed to maintain a steady pressure on the device against the radial artery to increase signal output. No pulse was easily discernable without the applied pressure. However, after the wrist band was applied, an output of similar frequency (60 beats per minute) to the carotid artery placed device was found (Fig. 7b (iii)). The output from the radial artery was noticeably noisier than the carotid artery. In Fig. 7b (iii) we see that the output amplitude is higher than that of the device on the carotid artery (i), this is due to the applied external pressure of the wrist pressure output by the radial artery. Similar to (ii), in (iv) a plot of a single pulse wave from the radial artery is shown. Here, the dicrotic notch is less distinguishable than in the carotid artery pulse.

We are able to demonstrate the ability of the inkjet printed sensor to detect pulse rate from both the carotid and radial artery. The conformability of the device allows it to form to the skin more easily than standard a ceramic piezoelectric device.

D. Array Structure Characterization

One of the key benefits of developing a fabrication method that utilizes an inkjet printer is the rapid prototyping ability and easy pattern formation. To show the most basic application of this patterning ability we printed a 2 × 2 array (schematic seen in the inset of Fig. 8a). The 2 × 2 array was made up of four 1 cm² square elements spaced 0.3 cm apart, all on the same PVDF-TrFE/PDMS composite substrate. Fig. 8a shows individual outputs of each element from a 5 N impact at 1 Hz. The test shows that each of the elements performs similarly to equal force inputs with outputs near 8 pC.

Fig. 8.

(a) 2 × 2 square array proof of concept device (schematic shown in inset) made up of four 1 cm² square electrodes spaced 0.3 cm apart, all on the same 3 cm × 5 cm PVDF-TrFE/PDMS composite substrate. Charge outputs from a light tapping on each element shows device functionality. (b) Charge outputs from two adjacent elements (1 and 2) as an acrylic square impacts element 2 and moves to the right until it is fully impacting element 1. Both elements show large charge output decreases while the square is directly impacting the opposite element. (c) Testing shape detection of the 2 × 2 array by impacting elements 2, 3, and 4 with an L-shaped piece of acrylic, while element 1 remained untouched. Element 1 showed a much lower output compared to the other 3 elements, suggesting that these devices could be used for shape detection.

The array was also tested to determine the relationship between outputs of adjacent elements within the array. A 1 cm² acrylic square was impacted onto one of the elements, element 2, at a frequency of 1 Hz. The square was then slowly moved towards another element, element 1, and the output of both elements was measured. In Fig. 8b we see that while the elements were both under full impact of the square, they had similar outputs of 6 pC. While the square was impacted on the opposite element the output fell to 1/10 of the direct impact value, below 0.6 pC. While element 2 was directly under the impact it had an output of 6.15 pC versus 0.11 pC of element 1. While element 1 was directly under the impact it had an output of 6.55 pC versus 0.10 pC from element 2. This represented an average of a 60.7x increase in signal. The result suggests that the array structure can be used in applications with spatial sensitivity to determine location of various forces.

This was validated using various shapes applied to the array while testing the output of all 4 elements. Shapes tested were single 1 cm² squares, a 2 × 1 element-width rectangle, a square that covered all 4 elements at once, and an L-shape that covered 3 of the 4 elements. The L-shape was determined to be the most complex shape and in the inset of Fig. 8c is shown impacting elements 2, 3, and 4. Fig. 8c shows the outputs from each of the individual elements from a 1 Hz impact of the L-shape. We can see that the 3 elements being impacted by the device show similar outputs around 8 pC, while element 1, not impacted by the L-shape, shows an output of only 3 pC. This validates that the inkjet printed array structures can be used for shape detection and opens the sensor to further applications involving spatial sensitivity to force.

E. Application: Foot Motion Detection

To further explore applications of the inkjet printed array structures, we printed a 2 × 2 array of sensor to fit on the sole of a shoe. A photograph of the shoe and the inkjet printed sensor array on the shoe sole can be seen in Fig. 9a. Each element of the sensor array had an active area of 2 cm by 5 cm.

Fig. 9.

(a) Photograph of the inkjet printed array sensor placed on a shoe sole for an athletic shoe. Each sensor element had an active area of 2 cm × 5 cm. (b) Testing toe-to-heel (i) and heel-to-toe (ii) movements while wearing the inkjet printed array sensor on a shoe sole. The inset shows the labeling of the elements within the shoe sole sensor, 1–4. In (i) we see that the time difference between the front element 4 and the back element 1 is +0.331 seconds, while in (ii) it is a −0.298 seconds, suggesting that we can detect foot strike direction. The error bars represent the standard error of the mean.

To test the ability of the sensor to determine spatiotemporal inputs we measured outputs from the 4 elements during a toe-to-heel step (Fig. 9b (i)) and a heel-to-toe step (Fig. 9b (ii)). The time value of the peak outputs from each element were subtracted from the time values of the peak outputs from element 1 at the right heel. In doing this we are able to determine the time difference between pressure inputs at various parts within the shoe.

In Fig. 9b we can see the results of the time differences of outputs from element 4 (left front) and element 2 (left heel and next to element 1). During a toe-to-heel foot movement in (i) we see that element 2 has a mean time difference of 0.003 seconds. This shows good agreement between both sensor elements in the heel. At the front of the foot, element 4, has a mean time difference of +0.331 seconds, which is expected because the front of the foot is striking the ground before the heel during a toe-to-heel movement and so the time difference between peak outputs from element 1 and 4 should be a positive number.

During a heel-to-toe foot movement in (ii) we see that element 2 has a mean time difference of −0.030 seconds, again in good agreement with element 1. At the front of the foot, element 4, has a mean time difference of −0.298 seconds. This again is expected, as the time value should be opposite of the toe-to-heel movement. This shows that we are able to determine the spatiotemporal difference between the front and back of the shoe, allowing us to determine the difference between a heel-to-toe or a toe-to-heel foot strike.

IV. Discussion

The devices developed here are capable of measuring both the pulse rate and breath rate, and the heaviness of breathing appears to have an impact and may be measurable. The devices were also used for detecting the type of foot movement. These devices were fabricated using electrospinning and inkjet printing which are both highly scalable fabrication methods [2], [26], [27]. The ability to make use of inkjet printing of conductive polymer PEDOT:PSS electrodes on electrospun piezoelectric PVDF-TrFE nanofibers made possible through the methods developed here may allow for fabrication of easily patternable all-polymer piezoelectric devices at large scale.

One of the key benefits of this novel fabrication method, for inkjet printing polymer electrodes on PVDF-TrFE nanofibers, is the ability to rapidly fabricate a variety of electrode patterns. This allows us to easily develop devices for a variety of applications. In the future we will investigate more advanced patterning structures for their applications in biomedical sensing. To do this, the arrays will need to be further scaled down by approximately an order of magnitude. The array devices tested here showed great promise in the ability to accurately detect force location, but this still needs to be tested at smaller scales to determine the array distance effect on element’s sensitivity to force. With the ability of a single device to detect pulse waveforms in the wrist and neck, a further optimized array structure could offer the ability for further pulse wave analysis such as pulse wave velocity [3].

V. Conclusion

Through the addition of a physical separation layer of PDMS, we are able to inkjet pattern PEDOT:PSS electrodes on PVDF-TrFE nanofibers to create an all-polymer piezoelectric array device, which previously was not achievable due to ink leakage. The fabrication methods were characterized, and as a proof of concept, single element and simple arrays were fabricated and characterized. The device’s applications were demonstrated for use in cardiovascular sensing in the detection of both radial and carotid arterial pulses, as well as variations in the breath. Further array designs were also capable of detecting foot movement with spatiotemporal inputs. The fabrication method shows great promise as a highly scalable method for fabricating sensitive and flexible piezoelectric sensors.

Biography

Andrew Closson received the B.Sc. degree in bio-engineering from the University of Maine in 2016. He currently is Ph.D. candidate at Thayer School of Engineering at Dartmouth College where he is a part of the PhD Innovation Program. His research interest includes flexible materials for biosensing and energy harvesting.

Haley Richards received her bachelor’s degree in Engineering Sciences Dartmouth College in 2020 and competed a senior honors thesis under the supervision of Professor John XJ Zhang. Her research focuses on the design and analysis of self-powered wearable cardiac sensors.

Zhe Xu received his M.S. (2014) degree and Ph.D. (2017) degree from Erik Jonsson School of Engineering and Computer Science at University of Texas at Dallas, Dallas, Texas. Currently he is a research associate of Professor John X.J. Zhang in Thayer school of Engineering at Dartmouth College, Hanover, New Hampshire. His research interest lies at piezoelectric composite materials, bio-inspired materials, advanced manufacture, Implantable energy harvesting application, and bio-integration of soft electronics.

Congran (Billy) Jin is currently a Ph.D candidate at Thayer School of Engineering at Dartmouth College, Hanover, NH, USA. He obtained his B.S. and M.S. degree in Mechanical Engineering from Rensselaer Polytechnic Institute (RPI), Troy, NY in 2016. His research interest include piezoelectric energy harvesters, sensors, functional nanoparticle synthesis and soft robots.

Lin Dong is an Assistant Professor at the Department of Mechanical & Industrial Engineering at New Jersey Institute of Technology. She received her Ph.D. from Stevens Institute of Technology, where she was awarded the university’s Innovation and Entrepreneurship Doctoral Fellowship. She was a Research Associate at Dartmouth College. Her research interests lie in the intersection of mechanical engineering, material science and biomedical engineering, covering the topics of functional nanomaterials design, soft materials and robots, and energy harvesting/sensing devices.

John X.J. Zhang received the Ph.D. degree in electrical engineering from Stanford University, CA, USA, in 2004. He was a Research Scientist in systems biology at the Massachusetts Institute of Technology in 2005. He was a tenure Associate Professor in the Department of Biomedical Engineering, University of Texas of Austin. He is currently a Professor with the Thayer School of Engineering, Dartmouth College, Hanover, NH, USA. He is a fellow of the American Institute for Medical and Biological Engineering, and a recipient of the 2016 NIH Director’s Transformative Research Award. His group has authored over 160 peer reviewed publications, presented over 70 invited seminars worldwide, and filed more than 50 patents (seven U.S. patents and over 30 international patents issued). He has a track record for developing well-funded research programs and his research has been sponsored by the NIH, NSF, DARPA, the Wallace H. Coulter Foundation, the British Council, and several other agencies. His research focuses on exploring bio-inspired nanomaterials, scale-dependent biophysics, and nanofabrication technology toward developing new diagnostic devices and methods on probing complex cellular processes and biological networks critical to development and diseases. Both multi-scale experimental and theoretical approaches are combined to investigate fundamental force, flow, and energy processes at the interface of engineering and biomedicine. In particular, his laboratory is leading the development of integrated microfluidic and photonic microsystems (MEMS, micro-electromechanical systems), semiconductor chips, and nanotechnologies critical to healthcare, defense, and environmental applications. His research findings were licensed to two companies: CardioSpectra (acquired by Volcano, Nasdaq: VOLC), and NanoLite Systems for developing successful products designed to diagnose cancer through blood screening, tissue imaging, and cell transformations at the point-of-care. He received the Wallace Coulter Foundation Early Career Award for developing handheld microphotonic imaging scanners and microsystems for early oral cancer detection, the NSF CAREER Award for the invention of plasmonic scanning probes design for controlled perturbation and imaging at sub-cellular level, and the DARPA Young Faculty Award for patterning plasmonic surface on MEMS for biomarker sensing applications. He was also a recipient of the NIH Director’s Transformative Research Awards in 2016, to develop implantable energy harvesting devices enabled by flexible porous polymer films integrated on multi-stable structures. He is an alumnus of the NAE Frontiers of Engineering Programs, an Editor for the IEEE/ASME JOURNAL OF MICROELECTROMECHANICAL SYSTEMS, and has authored a textbook for undergraduates: Molecular Sensors and Nanodevices: Principles, Designs, and Applications in Biomedical Engineering.