Aerosol Jet Printing Conductive 3D Microstructures from Graphene without Post-Processing

Abstract

Three-dimensional (3D) graphene microstructures have the potential to boost performance in high-capacity batteries and ultrasensitive sensors. Numerous techniques have been developed to create such structures; however, the methods typically rely on structural supports, and/or lengthy post-print processing, increasing cost and complexity. Additive manufacturing techniques, such as printing, show promise in overcoming these challenges. In this study, we employed aerosol jet printing for creating 3D graphene microstructures using water as the only solvent and without any post-print processing required. The graphene pillars exhibited conductivity immediately after printing, requiring no high-temperature annealing. Furthermore, these pillars were successfully printed in freestanding configurations at angles below 45° relative to the substrate, showcasing their adaptability for tailored applications. When graphene pillars were added to humidity sensors, the additional surface area did not yield a corresponding increase in sensor performance. However, graphene trusses, which add a parallel conduction path to the sensing surface, were found to improve sensitivity nearly 2x, highlighting the advantages of a topologically suspended circuit construction when adding 3D microstructures to sensing electrodes. Overall, incorporating 3D graphene microstructures to sensor electrodes can provide added sensitivity, and aerosol jet printing is a viable path to realizing these conductive microstructures without any post-print processing.

Graphical Abstract

Graphene ink is printed with an aerosol jet printer to form 3D microstructures. The conductivity and strength of the stack-printed graphene without post-processing is highlighted, achieving pillars with less than a 45° angle with respect to the substrate. After extensive parameter optimization, appropriately designed graphene 3D microstructures are shown to boost sensitivity in electronic humidity sensors.

1. Introduction

Additive manufacturing of microscale three-dimensional (3D) structures has gained recent attention for a range of applications. Although cleanroom facilities have historically been utilized to create microstructures, additive manufacturing has grown more prevalent due to its accessibility and ease of use. A wide variety of techniques, including powder bed fusion^[1] and direct-write printing^[2,3], have been explored to realize microstructures composed of various materials, such as gold^[4], indium tin oxide^[5], and even ice^[6,7]. Given the broad applicability of three-dimensional microscale structures for electronics^[8], microfluidics^[9,10], and biomedical applications^[11,12], interest in these structures has grown extensively. In electrical applications, microstructures are often utilized to increase the active sensing area of electrodes (where footprint dimensions may be limited) to improve the sensitivity of sensors^[13], efficiency of solar cells^[14], and areal capacity of batteries^[15] and supercapacitors.^[16] Since this field is still in its infancy, the continued development of materials and compatible techniques to form such microstructures is valuable to keep pace with the rapid advancement in potential use cases.

Among the wide range of materials suitable for microstructure formation, graphene has been one of the most prevalent due to its high mechanical strength and electrical conductivity^[17,18]. Growth mechanisms and additive manufacturing techniques for graphene are well studied, with several methods showing the ability to create graphene microstructures, including inkjet printing^[19], 3D printed template-directed assembly^[20], and light-cured 3D printing^[21]. While these techniques have demonstrated the feasibility and utility for fabricating 3D graphene microstructures, all current additively manufactured graphene systems are limited by the requirement for post-print processing techniques to achieve desired electrical properties. These ink formulations necessitate various forms of post-processing, such as baking in an oven^[18,22], heating on a hot plate^[23], photonic annealing^[24], freeze drying^[19], etc. The necessity for post-print processing decreases the application readiness of such printed materials, allowing only chemically or thermally resistant materials to be utilized as the substrate or previously deposited layers.

Aerosol jet printing (AJP) has been used with broad success for printing conductive materials without the use of post-print processing, with demonstrations of graphene^[25], AgNWs^[26,27], and PEDOT:PSS^[28]. Specifically, AJP has been utilized to print conductive graphene in a variety of 2D electronic structures, such as interconnects^[29,30], transistor electrodes^[25,31,32], and sensors^[33–36]; however, AJP has yet to be harnessed for the use of printing 3D graphene structures. Previously, AJP has been utilized to create microstructures composed of perovskites^[37], silver nanoparticles^[38,39], photopolymer,^[40] and gold^[41]. Notably, Saleh et al. created Ag nanoparticle 3D angled structures with the aerosol jet printer for use in battery electrodes^[38,39]. Demonstrating the potential for AJP to yield 3D graphene microstructures is needed in order for this additive manufacturing approach to be competitive with other techniques for enabling a diverse assortment of custom printed structures from 2D materials without extensive post-print processing.

In this work, we developed an aerosol jet printing procedure to additively manufacture 3D graphene microstructures without the use of support structures or post-processing. Since 3D aerosol jet printing of inks is still in its infancy, this study provides insights into how to develop an ink capable of stacking to achieve vertical 3D structures. Key printing parameters were studied, including print speed that was found to create high-aspect ratio, uniform pillars at 0.3 mm/s. Furthermore, given the outstanding structural integrity and rapid drying of the graphene ink, we found that pillars can be printed below a 45° angle with respect to the substrate without altering the angle of the substrate or the nozzle. The microstructures exhibited relatively high conductivity with negligible contribution from any over-sprayed graphene that accumulated around the base of the structures. Additionally, 3D structures were functionalized onto a graphene humidity sensor, providing information about the effects of structural topology on sensor performance, showing that the use of 3D graphene trusses yields a nearly two-fold improvement in device sensitivity. With the development of this simple approach to fabricating 3D conductive graphene microstructures without post-processing, the application space of additively manufactured 3D microstructures can be further expanded beyond sensors to fields such as wearable electronics.

2. Results and Discussion

2.1. Aerosol jet printing of 3D graphene pillars

Aerosol jet printing functions via the atomization of liquid ink material dispersions. Utilizing the ultrasonic atomizer, the ink is aerosolized while an atomizer flow carries the ink from the vial to nozzle, where a sheath flow keeps the aerosolized column in a right stream as it exits the nozzle to minimize both clogging and overspray (Figure 1A). The graphene concentration of the ink was maximized to 2.3 % w/w in water. At concentrations above 2.3 % w/w of graphene, the ink is too viscous to readily print utilizing the ultrasonic atomizer and susceptible to clogging of the nozzle. This concentration of ink produces the fast graphene deposition with rapid solvent evaporation that is required to produce graphene pillars with high structural integrity. To achieve aerosol jet printing of 3D graphene pillars, the ultrasonication power is finely tuned (from 300–500 mA) to balance the ink deposition rate, as too rapid or too slow a rate will result in poorly defined 3D structures. Ultrasonic atomization did not significantly alter the graphene composition during printing, as confirmed by Raman spectroscopy of a single-pass AJP graphene rectangle and drop-cast graphene ink (Figure 1B, S1, Supplementary Information). The G peak magnitude (Ig) is sufficiently larger than that of the D peak magnitude (Id) for the printed graphene with only a slight increase in the Id/Ig ratio from 0.24 for drop-cast graphene to 0.38 for the printed sample, indicating only a small number of impurities were introduced due to ultrasonic atomization and printing^[42–44]. Without any post-processing, the final printed line at a print speed of 0.3 mm/s, with a width of ~25 μm is about 1 μm in thickness and is sufficiently conductive, 0.143 mS/cm, with only a single print pass (Figure 1C). This print speed yields a sufficient line thickness to build up 3D graphene structures that are both uniform and reproduceable, which enables the in-place printing of large arrays.

Figure 1. Aerosol jet printing of 3D graphene pillars.

(A) Schematic illustrating the graphene ink flow in an aerosol jet printer to form 3D graphene pillars. (B) Raman of as-printed graphene, indicating the reasonably low defect density in the film. (C) Optical image of a single print pass of graphene at a print speed of 0.3 mm/s with optimized printing parameters. (D) SEM images of a 5-by-5 array of printed pillars with images from left to right iteratively focusing in on a single pillar to demonstrate the effects of overspray on the top of the pillars. Pillars were printed using print path of 75 μm diameter circles at a speed of 0.3 mm/s.

Without moving the print nozzle during graphene printing, vertical graphene pillars can be formed upwards of 2 mm tall, with a variable diameter ranging from 10 to 50 μm in a single structure, producing a pillar with an aspect ratio of 40–200. The height of these pillars was only limited by the printer nozzle distance from the substrate (kept constant at 3.5 mm) rather than the structural integrity of the graphene pillar itself (Figure S2, Supplementary Information). While printing without moving the nozzle or stage does produce a 3D structure, these pillars are non-uniform and have low reproducibility. In order to produce more uniform structures, the nozzle must move in a repeated motion while the structure grows from the bottom up (Video S1, Supplementary Information). For instance, wider and more uniform graphene pillars can be formed by successive circular depositions directly atop one another.

To examine the repeatability of printing pillars using successive circular depositions at a print speed of 0.3 mm/s, a 5 × 5 pillar array of 128 stacked circles per pillar were fabricated (Figure 1D). Designing the nozzle print path to be 75 μm diameter circles revealed that the resultant pillars had a diameter of about 100 μm due to the 25 μm deposition line width. Further, despite the built-in 75 μm gap between the printed lines, the pillars appear closed at the top due to excess ink deposition and limited water evaporation, which generates two distinct surface characteristics on the pillars. The smoother portion of the microstructure is where ink is coherently deposited as a jet to form pillars while the rough surface, on the top and around the bottom of the pillar, is individual graphene flakes due to overspray during printing, which is a phenomenon where small errant particles exit the controlled deposition jet stream and deposit already dried ink constituents in an area around the intended deposition location (Figure 1D). The graphene flakes seen in the overspray are consistent with the flake distribution provided by the supplier, 500 – 1500 nm, therefore showing that there is no appreciable reduction in size due to sonication. Surprisingly, the surrounding overspray pattern at the base of each pillar is nearly identical, which is attributed to the condition of the nozzle, be it partially clogged or not (Figure S3, Supplementary Information). This largely uniform deposition was achieved through several optimization steps. Slight overspray-induced roughness notwithstanding, the resultant pillar dimensions and appearance are repeatable over the duration of printing time required to produce 25 pillars at 0.3 mm/s— ~1.6 minutes per pillar, ~42 minutes total. To achieve these prints, accelerating solvent evaporation was pivotal. To accomplish this, the platen and ink bath temperatures were raised from room temperature to 120 °C and 50 °C, respectively. Additionally, the atomizer current was closely monitored throughout the print process to ensure there is no excess solvent deposition (Note S1, Supplementary Information).

To understand the impact of printing speed (defined as X-Y platen movement on AJP) on pillar structure, pillars created from printing speeds of 0.1 mm/s to 2 mm/s were characterized (Figure 2 A–B). Given that the ink deposition rate was unchanged, there is insufficient graphene deposition at faster print speeds to deposit a uniform line; rather, a significant fraction of the deposition was formed of overspray, which concentrated in specific locations, causing thin spires. As the print speed decreases to 0.1 mm/s, the pillar height increases since the total graphene deposition increases, enabling greater total pillar height. The pillar height showing an inversely proportional trend to print speed is consistent with the theory that the thickness of the printed line is inversely proportional to printing speed, highlighting that the graphene is indeed stacking to form these pillar structures. Although exploring slower speeds would be insightful to learn more about the observed trend, the Optomec system utilized in this study has a minimum print speed of 0.1 mm/s. It is evident that the 0.1 mm/s speed produced the tallest pillar, however, these pillars are not as uniform as those printed with speeds of 0.3 to 0.5 mm/s which may be attributed to the increased spatial deposition causing excess water accumulation and post-printing ink reflux (Figure S4, Supplementary Information). In addition to forming a jagged pillar with less-uniform sidewalls, decreasing the print speed from 0.5 to 0.1 mm/s in turn causes the prints to take ~5 times longer. The diameter variation of the pillars printed at speeds of 0.3 to 0.5 mm/s is ± 10 μm, a ± 10% change in the desired pillar diameter of 100 μm, and may be attributed to excess ink spilling over the sides of the pillar while drying. To alter the surface roughness of the pillars for specific applications, such as the smooth Ag nanoparticle 3D microstructures for battery electrodes seen in Saleh et al. ^[39], the addition of a binder, which may decrease graphene conductivity without post-processing, or a more-volatile solvent than water may be considered for future work. Throughout the rest of this study, a print speed of 0.3 mm/s was utilized to optimize the trade-offs between time, height, and pillar uniformity.

Figure 2. Tuning of print parameters for graphene pillars.

(A) SEM images of pillars with 60 printed layers at print speeds of 1 mm/s, 0.5 mm/s, and 0.1 mm/s. (B) Pillar height as a function of print speed after 60 printed layers were deposited, as measured through SEM, indicating an increase in pillar height as print speed decreases due to an increase in dwell time of the nozzle around the loop at each layer. Inset showing three pillars printed at 0.3 mm/s, demonstrating the uniformity and repeatability of printing at this speed. (C) Lateral shift distance, d, in the printing file and corresponding printed layer. (D) Angle of the pillar with respect to the substrate as a function of d with the light orange showing standard deviation, highlighting the increased variability in the printed pillars at d greater than 3 μm. (E) SEM images of pillars with 60 printed layers with lateral shift distances, d, between each layer at 1 μm, 4 μm, and 7 μm, highlighting how the pillar leans further with each increase in spacing. All pillars, vertical and leaning, used a print path design of 75 μm diameter circles.

2.2. Angled 3D graphene pillars and trusses

Although vertical pillars are a basic building block in 3D, creating pillars at various angles with respect to the substrate increases the number of achievable structures, such as letters^[7] and lattices^[38], expanding the applications for this technique. To determine the minimum angle a graphene pillar can achieve by this print method, pillars were printed at 0.3 mm/s with lateral shift distances, d, ranging from 0 to 7 μm between each printed layer (Figure 2C). In the resulting microstructures, the corresponding angle, α, between the pillar and substrate was measured using a scanning electron microscope, SEM (Figure 2D–E, Figure S5, Supplementary Information). As the printed circles are spaced further apart (i.e., as d increases), the angle α decreases nearly linearly, achieving down to a 36.6° slanted pillar. Saleh et al. showed that the minimum angle range for aerosol jet printing Ag nanoparticle 3D structures is 30 to 40°, therefore confirming that these graphene structures, using only water as the solvent, are able to achieve this angle range necessary to create high areal capacity battery electrodes^[38,39]. From 0 to 3 μm spacing, the pillars were easily reproducible without fine tuning of the print parameters. In contrast, increasing the spacing from 4 μm to 7 μm, the pillars were very difficult to repeatedly produce standing structures, (Table S1, Supplementary Information). In the set of pillars with d = 4 μm spacing, only four out of nine pillars were successfully printed. Attempting to print above a spacing of 7 μm at a print speed of 0.3 mm/s, we were unable to form any pillars because the volumetric deposition of graphene was not sufficient to form the necessary layers for vertical stacking. Thus, it was determined that beyond a 3 μm lateral shift distance, which yields a 70 ° angle, an increase in dwell time or drying time is necessary to create reproducible, uniform structures with a diameter variation below ±20%. One such method is creating truss structures where the ink deposition alternates between the right and left sides of the structure, allowing for an increased drying time between each layer deposition and more structural support to reduce pillars from falling over post-printing.

In addition to adding structural integrity and improving the yield of leaning pillars, creating truss structures allows for investigation of the electrical conductivity of the 3D graphene prints – a characteristic unable to be measured in standing pillars. This is advantageous since a probe cannot be manually placed at the top of a 100 μm diameter pillar to measure the conductivity without damaging the pillar itself. To measure the electrical resistance of 3D graphene structures, a graphene truss was printed between two printed graphene electrodes with a lateral spacing distance of 4 μm (Figure 3A). Aerosol jet printing of graphene pillars produces considerable overspray below the trusses (Figure 3B, Figure S6A, Supplementary Information). This overspray appears to bridge the gap between the two ends of the truss, potentially causing a parallel conduction pathway (Figure 3C). To verify that this overspray is not contributing to the measured resistance, the trusses were mechanically removed without disrupting the overspray on the substrate after an initial resistance measurement; then the remaining overspray resistance was measured (Figure S6B–D, Supplementary Information). The overspray resistance is nearly six orders of magnitude higher than that of the measured truss complex, thus confirming that the measurements were dominated by the truss resistance. The average resistance of over 100 printed trusses is 2.87 kΩ, thus confirming that the truss structures are in fact conductive. The relatively high resistance of the deposited graphene can be attributed to the sodium deoxycholate that is added to the ink to achieve a graphene dispersion, which reduces inter-flake conduction. Post-print processing may be used to further improve upon the resistance; however, it is not necessary for demonstrating the utility of graphene trusses in application to humidity sensors, which rely on the relative change in graphene’s resistance to measure humidity.

Figure 3. Graphene trusses printing procedure and electrical characterization.

(A) Printing process for graphene trusses, printing the furthest right circle, then the furthest left circle, then repeats with a lateral shift distance inwards of 4 μm for both the right and left sides, repeated 63 times until the truss is formed. (B) SEM image of graphene truss printed at a print speed of 0.3 mm/s and d of 4 μm. (C) Simplified equivalent circuit of as-printed truss demonstrating the parallel resistances of the printed graphene truss and overspray.

2.3. Humidity sensors functionalized with 3D graphene structures

After confirming the conductivity of the 3D structures, humidity sensors were fabricated from printed silver nanoparticle (AgNP) interdigitated electrodes (IDEs) with a graphene resistor printed overtop the electrodes (Figure 4A). The printed graphene was enhanced with either fifteen pillars or twelve trusses that bridge the gap between the silver IDEs to determine the influence of such structures on the sensitivity of the sensors (Figure 4B–D). The stability of the control humidity sensor, without any 3D structure, is relatively steady over a 15-hour period under ambient conditions, 20 °C and 39 % RH, with only a 2 % change in resistance that may be attributed to slight fluctuations in relative humidity over time (Figure 4E). The change in resistance (ΔR/R₀) of the three humidity sensor designs was measured between 40 and 50 % RH, which was monitored by a commercial humidity sensor (Figure 4F–G). The control sensor yielded an average sensitivity of 3.4 ± 1.0 %Ω / %RH, which was increased to 5.2 ± 2.0 %Ω / %RH when functionalized with trusses. Interestingly, the average sensitivity decreased slightly to 3.2 ± 0.6 %Ω / %RH when pillars were added to the graphene, indicating that increased surface area of sensing electrode does not necessarily correlate to a higher sensor sensitivity for gas-based sensors, contrary to what is seen in electrochemical sensors^[41,45]. As the humidity is increased to 75 % RH, the truss devices continue to have a higher sensitivity than the control and pillar devices, with about a four times higher percent change in resistance (~2300 %) than the control device (~600 %) at 75 % RH (Figure S7, Supplementary Information). Even under high RH environments, the trusses have a distinct advantage over the control and pillar humidity sensors.

Figure 4. Humidity sensors functionalized with 3D graphene microstructures.

Schematic of (A) control, (B) pillar-functionalized, and (C) truss-functionalized humidity sensors. (D) Optical image of control humidity sensor with the addition of silver epoxy for enhanced electrical connection. (E) Stability of control sensor over a 15-hour period at room humidity, ~39 %. (F) Percent change in resistance of each humidity sensor from 40 % RH to 50 % RH, with the truss-functionalized device exhibiting the best sensitivity. (G) Testing setup for humidity sensor measurements showing the sensor connected to a parameter analyzer and placed in an environmental chamber for precise humidity control. (H) Percent change in resistance over time with respect to 40 % RH resistance value at varying RH values, from a low humidity of 15 % RH up to 60 % RH which was achieved by (i) drying out the chamber, (ii) opening the chamber to room humidity (~39 % RH), then (iii) adding a humidifier in an enclosed space.

When thinking about the pathways that electrons can flow through each sensor, it is evident that trusses provide an additional current path whereas pillars do not. Furthermore, electrons that flow through the upper current pathway in the truss are highly exposed to humidity on all sides, whereas the electrons that flow underneath the base of the pillars are actually shielded from feeling the effects of the change in humidity. Therefore, as water accumulates on the surface of the truss and graphene underneath it, both current pathways are affected, hence why the trusses experience such a sharp increase in sensitivity. In contrast, the pillars do not contribute useful sensitive surface area in this device design, and in fact cover up some of the sensitive current pathways in the graphene (including with non-conductive overspray) which ultimately reduces the sensitivity of the device. Ultimately, pillars decreased the sensor’s performance, but by adapting the pillars into 3D truss microstructures we demonstrate a substantial increase in air-based sensitivity to boost humidity sensor performance.

To further characterize the response of the graphene humidity sensor enhanced with graphene trusses, the sensor was exposed to relative humidity ranging from 15 to 60 % (Figure 4H). Under nitrogen, the relative humidity was stable at 15 % and the corresponding sensor response was also considerably steady. When the chamber is opened to ambient air, ~39 % RH, the commercial sensor and graphene sensor track the relative increase in humidity within the chamber. As the humidity is increased at a faster rate with the addition of a humidifier, up to 60 % RH, the graphene sensor tracks this change more quickly than that of the commercial sensor, demonstrating a more rapid reaction time. However, as the humidity rose to 60 % RH, water droplets started to form on the sensor, inducing a large spike in resistance. Finally, as the chamber was removed from the humid environment back to ambient humidity, the sensor resistance decreased and stabilized quicker than that of the commercial sensor, plotted as the %RH in Figure 4H, further confirming that this sensor has a faster reaction than that of the commercial sensor. The improved sensor performance and fast response time to change in humidity demonstrates the utility of graphene trusses for use in fully printed air-based sensors. Although the repeatability of these sensors may be improved upon in future work through ink formulation and optimized printing conditions, it is evident that simple 3D aerosol jet printed microstructures have the potential to significantly improve sensor performance without compromising on device footprint or power consumption.

3. Conclusion

We have reported the development of an aerosol jet printing process to create 3D graphene microstructures, which can reach up to 2 mm in height with aspect ratios as high as 200. These microstructures are highly customizable, achieving angles down to 45° with respect to the substrate without any additional support. Furthermore, through the use of truss structures, it was confirmed that the printed graphene provides a conductive spire, without the need for post-processing, using a maximum temperature of 120 °C applied to the printer platen during printing. To demonstrate the utility of these 3D microstructures, fully printed humidity sensors were fabricated and enhanced with 3D graphene pillars, revealing that an increase in surface area alone does not immediately correlate with increased sensitivity; however, the addition of trusses creates a high surface area with parallel resistance path that increased sensitivity by 150%. By enabling the development of microscale 3D structures that can be incorporated directly onto thin film printed electronics, our print process offers a practical approach to improve sensor performance and a simple, yet effective, way to fabricate 3D microstructures for target applications.

4. Experimental Section

Graphene Ink Preparation and Printing:

Graphene ink (Sigma-Aldrich 80556–10ML at 7 wt.% graphene concentration in water with 5–10 wt.% sodium deoxycholate) was diluted with DI water to a 2.33 wt.% graphene concentration. This ink has flake sizes ranging from 500 −1500 nm, which is crucial for printing with the aerosol jet printer as flakes larger than several microns may be bigger than an aerosolized water droplet, causing the graphene to not be deposited during printing. The ink was printed with an Optomec AJ300 aerosol jet printer using a 150 μm nozzle and the ultrasonic atomizer. The distance between the nozzle and the substrate was kept at a constant 3.5 mm throughout each print. The sheath flow was set at 25 SCCM and the carrier gas flow was kept in the range of 35 – 42 SCCM. The platen and ink bath temperatures were 120 °C and 50 °C, respectively. The atomizer current ranged from 300 mA – 500 mA. The printing speed varied depending on the print.

Silver Nanoparticle Ink Preparation and Printing:

Silver nanoparticle ink (Ag40 x, UT Dots Inc. USA) was diluted with terpinol in a 9:1 ratio. The ink was printed with an Optomec AJ300 aerosol jet printer using a 150 μm nozzle and the ultrasonic atomizer at a current of 350 mA. The sheath and atomizer gas flow rates were set at 25 SCCM and 20 SCCM, respectively. The ink bath temperature was 20 °C and the platen temperature was 60 °C during printing. A print speed of 9 mm/s was used for all silver ink printing. Post printing, the printed AgNP electrodes were annealed in an over at 200 °C for 1 hour to improve electrical conductivity.

Humidity Sensor Fabrication:

The humidity sensors were printed on glass slides. All substrates were cleaned by rinsing with isopropyl alcohol (IPA), rinsing with deionized (DI) water, and blowing dry using nitrogen. The interdigitated AgNP electrodes were printed with the above printing parameters with a gap of 450 μm between the 5 electrodes with dimensions of 1.96 mm × 170 μm. A 1.25 mm × 4.25 mm graphene electrode was printed between the interdigitated electrodes with the above printing parameters using a 2 mm/s print speed. Finally, to functionalize the sensors with 15 pillars or 12 trusses, the graphene was printed with the same parameters but at a print speed of 0.3 mm/s.

Humidity Sensor Testing:

Ag epoxy was added to the AgNP electrodes of the humidity sensors to expand the electrode to allow for an alligator clip to be attached to the electrode. The measurements of the sensors were carried out by a SMU (Keysight B2902A). The relative humidity was monitored by a commercial sensor (SensorShare SDAWIR01/02), retrieving data from the sensor every 10 seconds. The sensors were placed in a chamber so that humidity can be easily altered by the addition of nitrogen or water vapor.