Self-aligned, inkjet-printed resistors on flexible substrates with excellent mechanical stability, high yield, and low variance

Resistors are basic yet essential circuit components that must be fabricated with high precision at low cost if they are to be viable for flexible electronic applications. Inkjet printing is one of many additive fabrication techniques utilized to realize this goal. In this work, a process termed self-aligned capillarity-assisted lithography for electronics (SCALE) was used to fabricate inkjet-printed resistors on flexible substrates. Capillary channels and reservoirs imprinted onto flexible substrates enabled precise control of resistor geometry and straightforward alignment of materials. More than 300 devices were fabricated using poly(3,4-ethylene dioxythiophene):poly(styrene sulfonate) (PEDOT:PSS) as the resistive material and silver as the electrode material. By varying PEDOT:PSS ink formulation and resistor geometry, resistances spanning from 170 Ω to 3.8 MΩ were achieved. Over 98% of devices were functional and the relative standard deviation in resistance ranged from 3% to 18% depending on resistor length and ink composition. The resistors showed no significant change in resistance after 10 000 cycles of bend testing at 1.6% surface tensile strain. In summary, this work demonstrated a fully roll-to-roll compatible process for inkjet printing resistors with superior properties.


Introduction
Flexible printed electronics (FPE) has drawn significant attention in recent years due to the unique mechanical properties and additive vacuum-free manufacturability of these technologies.The ability of FPE devices to operate while bent and conform to surfaces with complex topographies makes them suitable for wearable medical sensor applications.For example, FPE sensors have been used to monitor heart rate [1], body temperature [2], and blood pressure [3], as well as other important rehabilitation metrics [4][5][6][7].The additive quality of FPE allows scaling to large areas (many m 2 ) without material waste [8][9][10].Additionally, the printing processes employed to fabricate FPE devices are typically roll-to-roll (R2R) compatible, allowing high-throughput production at low cost per unit area.
Inkjet is one of several printing techniques used for FPE.This printing method has many advantages including simultaneous printing of multiple materials, non-contact deposition, and high commercial availability [11].Furthermore, features less than 5 µm in size have been achieved using superfine inkjet printing technology [12][13][14].However, aligning features of this size remains a challenge when multiple layers of material are required (e.g. in transistors, capacitors, and resistors).Multi-layer alignment of small features becomes extremely difficult at a R2R manufacturing scale, as compensating for substrate velocity and stretching becomes an additional obstacle.
Self-aligned capillarity-assisted lithography for electronics (SCALE) is a new process developed to address the alignment issue described above and achieve micron-level resolution [15,16].The SCALE process starts by fabricating multi-level microfluidic channels and large circular reservoirs into a silicon wafer using traditional photolithographic techniques.This wafer is used as a master template for the creation of many elastomeric stamps, which are used to imprint the pattern into an ultraviolet (UV)-curable polymer layer coated on a flexible substrate.Then, inkjet printing is used to deposit inks into the large circular reservoirs and capillary flow spontaneously wicks the inks into the microfluidic channels.The use of photolithography to create the master template enables sub-micron resolution to be achieved.Additionally, containment of ink inside the channels results in very sharp edges and high aspect ratios [17], qualities that are unattainable using inkjet printing alone.Crucially, SCALE solves the layer-to-layer alignment issue by implementing all the alignment controls in the multi-level imprinting step; delivery of ink to the easily targetable circular reservoirs in the correct sequence automatically creates aligned multi-layer material stacks.Additionally, SCALE is fully R2R-compatible, making it well-suited for high-throughput production [17].
Inkjet-printed active circuit components such as diodes and transistors have been abundantly reported in the literature using traditional printing techniques [18][19][20] as well as SCALE [21][22][23] due to the roles these devices play in important circuits, such as full-wave rectifiers and logic gates.Passive circuit components, specifically resistors, play an equally essential role in circuits by limiting currents and dividing voltages but are often overlooked due to their perceived simplicity.A previous SCALE resistor design with good device metrics was reported in 2018 by Cao et al [24].However, in that work there were several device and processing characteristics that were sub-optimal including the resistance variance, the difficulty in controlling precise resistor dimensions, the non-linear modulation of resistance with length, and the non-uniformity of the deposited resistor material, poly(3,4-ethylene dioxythiophene):poly(styrene sulfonate) (PEDOT:PSS).The new resistor design reported in this study shows improvement in all these areas while also improving performance in other aspects, such as mechanical stability.While several other publications have also demonstrated fully inkjet-printed resistors on flexible substrates [25][26][27][28], we believe that the combination of high yield, low variance, mechanical/electrical stability, film uniformity/repeatability, and predictable geometric modulation shown in this study represents a new milestone (see supplementary table S1 for full comparison).Moreover, selfaligned printing processes are rare but are advantageous for high-throughput R2R fabrication.Overall, we present here a reliable self-aligned inkjet printing process for resistors that is compatible with R2R processing and achieves favorable figures of merit across a wide range of mechanical and electrical tests.

Experimental section 2.1. Silicon master template fabrication
Two rounds of photolithography were employed to fabricate the silicon master template.In the first round, a 4-in silicon wafer was subjected to a 3 min dehydration bake at 150 • C. All baking steps for both rounds of photolithography were completed on a hotplate.The wafer was placed in a hexamethyldisilazane atmosphere for 3 min to improve photoresist adhesion.A thin layer of positive photoresist (AZ 1512, MicroChem) was spin-coated onto the wafer at 3000 rpm for 30 s with an acceleration of 3000 rpm s −1 (figures S1(a) and (b)).A 70 s soft bake at 100 • C was completed, followed by a 5 s exposure to UV light (MA6, Suss MicroTech) through a photomask using the machine's hard contact mode.The exposed wafer was removed from the aligner and developed for 30 s in a solution containing a 1:5 volume ratio of developer (AZ 340, MicroChem):deionized (DI) water, followed by a 2 min DI water rinse (figure S1(c)).The exposed features were etched to a depth of 15 µm using a deep reactive ion etcher (Rapier, SPTS Tech) (figure S1(d)).After etching, the wafer was washed with acetone to remove most of the photoresist and then sonicated in an acetone bath for 30 min to remove the remainder (figure S1(e)).
For the second round of photolithography, a negative permanent resist was used (SU-8 2025, MicroChem) to build features on top of the wafer.The negative resist was spin-coated onto the wafer at 500 rpm for 10 s at an acceleration of 100 rpm s −1 , followed by 2000 rpm for 30 s at an acceleration of 300 rpm s −1 (figure S1(f)).Buildup of photoresist at the edge of the wafer (edge bead) was carefully removed using a razor blade.The coated wafer was soft baked at 65 • C for 3 min, then 95 • C for 6 min.A second photomask was aligned to the wafer and the photoresist was exposed to UV light for 12 s using the same aligner listed previously.Postexposure, the wafer was baked at 65 • C for 1 min, then 95 • C for 6 min.The coated photoresist was placed in developer (SU-8 developer, MicroChem) for 2 min, then washed with isopropanol (IPA).Lastly, the wafer was hard baked for 20 min at 150 The final thickness of the SU-8 layer was measured to be 30 µm using a stylus profilometer (P16 Surface Profiler, KLA Instruments).

PDMS stamp fabrication
Poly(dimethylsiloxane) (PDMS), an elastomer, was used to create stamps from the master wafer.To prevent adhesion of PDMS to the silicon master template, the wafer was placed in a vacuum chamber containing Trichloro(1H,1H,2H,2Hperfluorooctyl)silane (Sigma Aldrich) for 2 h.
A 10:1 weight ratio of PDMS monomer to curing agent (Sylgard 184, Dow) was measured.After stirring for several minutes, the mixture was placed in a vacuum chamber to remove bubbles and poured onto the silicon master template (figure S1(h)).The template and PDMS were then placed in an oven for 2 h at 75 • C. The cured PDMS was peeled off the silicon master template and placed back in the oven at 125 • C for 2 h, then 175 • C for 2 h to improve its durability (figure S1 (i)).

Imprinting of flexible substrates
A 127 µm thick film of polyethylene terephthalate (PET) (Melinex ST505, DuPont Teijin) was used as a substrate.To improve adhesion of the UV-curable polymer, the PET film was subjected to an O 2 plasma (PDC-001-HP, Harrick Plasma) for 3 min at 0.8 Torr of pressure and 45 W of power.The UV-curable polymer (NOA 73, Norland Products) was poured onto the PET film and the PDMS stamp was placed on top (figure S1 (j)).The stack of materials was transferred into a nanoimprinter (B200, Nanonex) where the NOA 73 was cured using a high intensity UVsource for 3 min while 1 psi of pressure was applied.The materials were removed from the nanoimprinter and the PDMS stamp was peeled off the NOA 73.The final result of this process was a PET film with an imprinted layer of NOA 73 adhered to it (figure S1(k)).The maximum thickness of the NOA 73 layer was about 45 µm.

Ink preparation
For printing the conductive silver electrodes, a reactive silver ink was used (EI-011, Electroninks).The silver ink was filtered through a 0.22 µm pore filter before printing to remove any precipitates.Two PEDOT:PSS inks (Clevios PH1000 and Clevios AI4083, Heraeus) were used as the resistive materials.Ethylene glycol (Sigma Aldrich) was added to some of the inks to increase conductivity.IPA was added to all PEDOT:PSS inks to decrease surface tension and improve flow in the microfluidic channels.Additives were thoroughly mixed into the ink and sonicated for 1 min.Before printing, the PEDOT:PSS inks were filtered through a 0.45 µm pore filter.

Device fabrication
The reactive silver ink was deposited using a dropon-demand inkjet printer with an 80 µm diameter nozzle (MJ-ATP-01-80, MicroFab).A unipolar waveform with a rise of 5 µs, dwell time of 20 µs, fall time of 5 µs, a driving voltage of 60 V, and frequency of 1000 Hz was used to deliver 60 drops of silver ink into both inner reservoirs of the resistor (figure S1(l)).Details on the layout of the resistor design are provided in section 3.1.Deposited silver ink was annealed for 10 min at 120 • C in air on a hotplate.The printed silver electrodes were submersed in a 1 mM ethanol-based solution of 11mercaptoundecanoic acid (Sigma Aldrich) for 2 h to increase the water wettability of the electrode surface.The same inkjet system and printing parameters were utilized for depositing PEDOT:PSS ink, with the exception of the driving voltage, which was increased to 70 V.Four hundred drops of PEDOT:PSS ink were delivered into both outer reservoirs of each resistor unless otherwise specified (figure S1(m)).PEDOT:PSS ink was allowed to dry at room temperature in air and then annealed for 10 min at 120 • C in an N 2 atmosphere.Printing of both materials was conducted in a humidity-controlled enclosure at 55% relative humidity.

Characterization
Plan-view micrographs of devices were acquired using a digital optical microscope (KH-7700, Hirox).Current-voltage (I-V) curves of devices were measured in air using a source measurement unit (2611B, Keithley).A microtome (EM UC7, Leica Microsystems) was used to cross-section devices after they were embedded in an epoxy (Poly/Bed 812, Polysciences) and cured for 24 h at 60 • C. Crosssectional micrographs were obtained using a scanning electron microscope (SEM) (SU-8230, Hitachi) at an acceleration voltage of 5 kV.
Bend testing was performed using a dynamic test instrument (ElectroPuls E3000, Instron) with custom grips at a cycling frequency of 2 Hz.The bending radius was controlled by changing the distance between grips.Surface tensile strain estimations were made using the following relation: ε = t/2r [29], where ε, r, and t are strain, bending radius, and substrate thickness, respectively.
Surface tension was measured on a drop shape analyzer (DSA30, Kruss) at 25 • C using the pendant drop method with a 1.85 mm diameter needle.Thirty datapoints were collected over the course of 10 s and averaged for each measurement.The reported surface tension values represent the average of four separate measurements.A Couette rheometer (HR 20, TA Instruments) was used to measure the viscosity of each ink at 25 • C. Shear rate was swept from 0.1 s −1 to 1000 s −1 using a relaxation time of 0.1 s.Viscosity versus shear rate data are shown in figure S2.

Device fabrication sequence
Figure 1 provides an overview of the SCALE process for the fabrication of resistors, beginning with the removal of a PDMS stamp from the silicon master template after curing (figure 1(a)).Prior to this process, a silicon master template was fabricated using photolithography as described in the     (15 µm deep feature).Silver ink was deposited into the inner, deeper section of the reservoirs and wicked into the connected narrow channels.Capillary flow halted when the liquid front was pinned at the end of the narrow channels (figure 2(b)).After thermal annealing of the silver, PEDOT:PSS ink was printed into the reservoirs, filling the entire reservoir, covering the silver electrode and then flowing into the wider, shallower channel between the two reservoirs.
The PEDOT:PSS ink flows from both reservoirs to complete the resistor by connecting the silver electrodes (figure 2(c)).Figures 2(d

Resistor performance
Having the ability to modulate resistance over many orders of magnitude without significantly changing the device footprint is desirable for circuit design purposes.In this study, more than four orders of resistive magnitude were demonstrated by utilizing two different PEDOT:PSS inks (PH1000 and AI4083) with different conductivities and through the addition of varying amounts of ethylene glycol.Ethylene glycol has been shown to increase conductivity of PEDOT:PSS by inducing phase separation of PEDOT and PSS, creating a more connected conductive path [30,31].Figure 3(a) shows five sets of resistors prepared with different ink formulations spanning resistances from 170 Ω to 3.8 MΩ.For each ink formulation, resistors with lengths of 100 µm, 200 µm, 300 µm, and 400 µm were fabricated.Each data point represents the average resistance of 16 devices, totaling to 320 devices for this plot.Table 1 provides the composition, viscosity, and surface tension for each ink as well as information about device yield.In this study, resistor failure was a result of misprinted silver ink that spilled out of the inner reservoirs and into the resistor channels, leading to very low resistance (short circuit).Yield was defined as the percentage of resistors within a given batch where shorting did not occur.The total yield for all devices was high at 98.4% and the relative standard deviation in resistance for the 20 datasets shown in figure 3(a) ranged from 3% to 18%, with 14 of the 20 achieving standard deviations less than 10%.A more detailed statistical summary for the data shown in figure 3(a) can be found in table S2 of the supplementary information.The low variance in resistance is primarily attributed to high precision in resistor length and width.This geometric precision is a unique benefit of the SCALE process and is achieved by confinement of ink within capillary channels, which are patterned using photolithography during the master template fabrication.The imprinted features also enabled high yield by simplifying the printing process, as ink only needed to be delivered into the circular reservoirs for devices to be completed.Devices also exhibited excellent electrical stability, showing no noticeable current hysteresis when swept from −5 V to 5 V (figure 3(b)).Furthermore, no change in resistance was observed when a device was subjected to 1000 sweep cycles (figure S5(a)).The maximum power that resistors made with each ink could withstand before a permanent change in resistance occurred was also tested.Resistors made with inks 1-4 all withstood more than 20 mW of power, while resistors made with ink 5 showed no change in resistance at 400 V.The results of this test are shown in figure S5(b).
While adjustments to ink formulation were used to easily alter resistance on an order of magnitude scale, more fine adjustments were made through modulation of resistor length and width.The ability to do so in a predictable manner is desirable, so that higher accuracy in resistance can be achieved.The scaling of resistance with a resistor's length and width in the ideal case is predicted by Pouillet's law where R, ρ, L, W, and t are the resistance, resistivity, length, width, and film thickness of the resistor, respectively.Figure 4(a) shows resistance as a function of resistor length for devices made using ink 3, demonstrating excellent linear adherence to this relation.Figure 4(b) shows resistance versus resistor width at all four resistor lengths for devices made with ink 4. Resistors of each length show inverse scaling of resistance with width, as expected in accordance with Pouillet's law.Thickness uniformity within the resistor channel was investigated through cross-sectional SEM imaging.It was found that the PEDOT:PSS films were uniform in thickness across the channel width (figures S2(a) and (b)).This result can be attributed to confinement of the ink within the channel walls in combination with a reduction in ink surface tension through the addition of IPA, as reported by Xing et al [32].
As a way of testing film thickness repeatability, resistors with four increasing volumes of PEDOT:PSS were printed using ink 4 (400 drops, 800 drops, 1200 drops, 1600 drops).Figure S2(a) shows a plan-view optical micrograph at each volume.The increase in darkness and decrease in iridescence within these micrographs indicated PEDOT:PSS was thickening within the circular reservoirs as ink volume increased.Interestingly, resistance of devices showed no dependence on ink volume over the range investigated (figure S2(c)).Additional cross-sectional imaging revealed that the PEDOT:PSS thickness remained constant within the channel regardless of the volume of ink deposited into the reservoirs (figure S2(d)).While unexpected, this phenomenon likely played a role in reducing the variance of resistors by enabling excellent repeatability of film thickness.Additional experiments are required to understand this effect mechanistically and to determine its applicability to other materials systems and device designs.
To determine their mechanical stability, resistors were subjected to bend testing using the setup shown in figure 5(a).Devices were bent facing outward, so the applied strain on the resistors would be  tensile.Only devices in the center of each sample were considered tested.The bending radii were measured by taking photographs of samples in their bent state, using the dimensions of the grips for scale (figure 5(b)).Three sets of tests were conducted at bending radii of 5.4 mm, 3.4 mm, and 2.1 mm corresponding to surface tensile strains of 1.6%, 2.6%, and 4.2%.Figure 5(c) shows a plot of resistance change versus bending cycles for devices made with ink 4. When 1.6% surface strain was applied, devices exhibited exceptional resilience to bending, withstanding 10 000 cycles with only a 2% increase in resistance.SEM imaging showed there was no visual difference in these resistors after testing when compared to untested resistors (figures S6(a)-(c)).When surface strains of 2.6% and 4.2% were applied, however, a decrease in resistance was observed.SEM imaging of the resistors subjected to surface strains of 2.6% and 4.2% showed cracks near the silver/PEDOT:PSS junction (figures S6(d) and (e)).However, because the resistance did not increase, the silver remained electrically connected to the PEDOT:PSS.A possible reason for the reduction in resistance is a morphology change in the PEDOT:PSS, similar to that reported by Lee et al [33], who showed PEDOT-rich cores grew in response to an applied tensile strain (when the PEDOT:PSS ratio was 1:2.5).
The ability of this resistor design to withstand 10 000 bending cycles at 1.6% surface strain without a significant increase in resistance can be attributed to the PEDOT:PSS layer printed on top of the silver electrodes.Crack formation and propagation are known to cause electrical failure in inkjet-printed silver lines during bend testing [34].The PEDOT:PSS printed on top of the silver electrodes in this device design improved mechanical durability by acting as a support layer that prevented crack growth and by providing an additional conductive path for current if cracking in the silver did occur.

Compact resistor designs
While the circular ink reservoirs are quite large in comparison to the resistor channels, reservoirs can be used to supply ink to multiple devices simultaneously.Figure 6(a) shows a plan-view optical micrograph of two parallel devices with ink supplied from one set of reservoirs.Similarly, figure 6(b) shows four resistors in parallel.Figures 6(c) and (d) display the equivalent circuits of these devices with labels R1, R2, R3, R4 corresponding to resistors of length 100 µm, 200 µm, 300 µm, and 400 µm.The resistances of these multi-channel devices were in good agreement with their expected values, which were calculated mathematically using Kirchhoff 's laws and data from single-channel resistors.Using this strategy, the area per resistor was decreased significantly from around 1.4 mm 2 to less than 0.35 mm 2 .Additional channels could be attached to further reduce device footprints.Figure S7 shows an illustration of this resistor design incorporated into a band-pass filter with two interdigitated capacitors, demonstrating the design's ability to conserve space even when combined with other circuit components.

Conclusions
Resistors with low variance and high yield spanning several orders of magnitude in resistance were demonstrated in this work.The resistors exhibited linear I-V curves, and no degradation in current after 1000 repeated measurements up to 5 V. Resistor length and width modulated the resistance as expected, enabling precise resistance control through two design variables.Excellent uniformity and repeatability in PEDOT:PSS film thickness was shown through cross-sectional imaging.Bend testing of the printed resistors revealed excellent mechanical stability compared to previously reported printed resistors.Lastly, it was shown that a single pair of reservoirs could supply ink to several devices, enabling a higher density of resistors.The combination of these findings demonstrates these resistors are well-suited for real applications, while the self-aligned fabrication process used in this work enables full R2R compatibility.

Figure 1 .
Figure 1.Illustrations of the SCALE process showing (a) removal of a cured PDMS stamp from the master template, (b) imprinting of a UV-curable polymer using the PDMS stamp, (c) removal of the PDMS stamp after UV-curing, (d) inkjet printing of silver ink into the inner circular reservoirs, and (e) inkjet printing of PEDOT:PSS ink atop silver into the outer reservoirs.Note: a single device is shown for illustration.

Figure 2 .
Figure 2. (a)-(c) Plan-view illustrations of the SCALE resistor before printing, after depositing silver, and after depositing PEDOT:PSS.(d)-(f) Cross sectional schematics at the dotted red line for the same three steps.(g)-(i) Magnified plan-view optical micrographs at each step.

Figures 2 (
Figures 2(a)-(c) illustrate a plan-view of the resistor design before depositing ink, after printing silver, and after printing PEDOT:PSS, respectively.The color contrast in figure 2(a) represents the different depths of the imprinted structure with the lightest tone corresponding to the surface, the middle tone corresponding to a depth of 30 µm, and the darkest tone corresponding to a total depth of 45 µm

Figure 3 .
Figure 3. (a) Over four orders of resistive magnitude achieved using 5 different ink compositions.(b) An I-V plot showing the forward and reverse sweep for a single resistor made with ink 4.
)-(f) provide a crosssectional view of the dotted red line at each printing step.An SEM micrograph of the cross section illustrated in figure2(f) is provided in figureS3.The narrow silver-containing channel was 20 µm wide, while the wider PEDOT:PSS-containing channel was 40 µm in width.Figures2(g)-(i) show plan-view optical micrographs at each step.The silver thickness within the narrow channel was measured to be approximately 1 µm (figureS3(d)), while the thickness of the PEDOT:PSS within the wider channel was measured to be 2 µm (figures S4(b) and (d)).

Figure 4 .
Figure 4. (a) Plot of resistance versus resistor length with a linear fit.(b) Plot of resistance versus resistor width at four different resistor lengths.Error bars represent 95% confidence intervals on the mean.

Figure 5 .
Figure 5. (a) Photograph of the mechanical testing apparatus.(b) Side-view photograph of a bent sample, showing the bending radius of the devices.(c) Change in resistance as a function of bending cycles for three different applied strains.Error bars represent 95% confidence intervals on the mean.

Figure 6 .
Figure 6.(a), (b) Plan-view optical micrographs and (c), (d) equivalent circuit diagrams of a 2-channel resistor and a 4-channel resistor, respectively.(e) Comparison of measured and expected resistance values of both multi-channel designs using ink 4. Error bars represent 95% confidence intervals on the mean.

Table 1 .
Ink compositions and yield metrics for the resistors shown in figure3(a).Reported viscosities were measured at a sheer rate of 10 3 s −1 .See figureS2for viscosity versus shear rate data.Surface tension error values represent the standard deviation of four measurements.