Monolithic in-plane integration of gate-modulated switchable supercapacitors (ipG฀Cap) Monolithic in-plane integration of gate-modulated switchable supercapacitors ( ip G‐Cap)

Monolithic integration of iontronic devices is a key challenge for future miniaturization and system integration, in particular the interconnection of multiple functional elements as required for ion computing. The G-Cap, a novel iontronic element, is a switchable supercapacitor with gating characteristics comparable to transistors in electronic circuits, but switching relies on ionic currents and ion electroadsorption. We report here the first monolithic in-plane G-Cap integration through 3D-inkjet printing of non-toxic nanoporous carbon precursors. The printed G-Cap has a three-electrode architecture integrating a symmetric "working" micro-supercapacitor (W-Cap) and a third "gate" electrode (G-electrode) that reversibly depletes/injects electrolyte ions into the system, effectively controlling the "working" capacitance. The printed precursor structures were directly converted into nanoporous carbon materials with a specific surface area of 544 m 2 g −1 . The symmetric W‐Cap operates with a proton conducting hydrogel electrolyte based on PVA/H 2 SO 4 and shows a high capacitance (1.3 mF cm −2 ) that can be switched “on” and “off” by applying a DC bias potential (- 1.0 V) at the G-electrode. This effectively suppresses AC electroadsorption in the nanoporous carbon electrodes of the W-Cap, resulting in a high capacitance drop from an "on" to an "off" state. Printing offers far-reaching freedom of design for varying the device structure achieving superior device performance. The new monolithic structures achieve high rate performance, reversible on-off switching with an off-value as low as 0.5 % surpassing values reported so far. Establishing technologies and device architectures for functional ionic electroadsorption devices is crucial for diverse fields ranging from microelectronics and iontronics to biointerfacing and neuromodulation.


Introduction
State of the art computing architectures predominately rely on semiconductors integrated into electronic circuits. A major concern in recent years is the ever increasing energy consumption caused by the widespread of handhelds, digitalization and modern communication. 1 Biogenic nerve systems instead use a complex interplay of ionic and electric currents/potentials as well as chemically defined signalling via neurotransmitters for logic operation and memory. [2][3][4] Such devices are genetically optimized to operate via minimal energy consumption. In this context, recently neuromorphic architectures and iontronic devices have received tremendous attention. [5][6][7][8] Such iontronic devices are explored for future information processing as well as biointerfacing. [9][10][11][12] Supercapacitors (electric double layer capacitors, EDLCs) are energy storage devices relying on purely physical ion adsorption in highly porous carbons. [13][14][15][16] They have been optimized over decades to achieve higher and higher energy and power density. 17,18 In particular rapid charging and discharging is an important characteristic in their commercial application for break-energy recuperation or as power boosters for electric engines in cranes 19 , elevators 20 and trains. 21,22 Developing supercapacitors (EDLCs) as active functional elements represents a change in paradigm.
The integration of more complex functionality, in particular switchability, may pave the way towards advanced ion-based computer architectures, logic architectures or switchable devices. A first step in this direction is the ion-based asymmetric supercapacitor diode (CAPode) enabling unidirectional energy storage via pore size engineering and an ion-sieving mechanism effectively blocking the negative bias charging direction. 8 More recently we presented the first switchable EDLC architecture with gating characteristics (G-Cap) resembling a transistor architecture based on ion electroadsorption ( Figure 1). 23 The structure consists of a simplified symmetric carbon electrode design as an AC-driven supercapacitor (working supercapacitor, W-Cap) controlled by a third porous carbon gate electrode (G-electrode; Figure 1a) arranged in a co-planar sandwich assembly. The charge storage capacity of the W-Cap is controlled by stimulated cation adsorption in the gate electrode, effectively depleting the electrolyte concentration. By applying a positive or negative bias potential on the G-electrode, the W-Cap can be effectively switched "on" or "off" (Figure 1c). 23 Switching elements such as the G-Cap are the basis of logic families to establish transistor-transistor or diode-transistor logic gates (AND, OR, NAND, NOR). 6,[24][25][26] However, a key requirement for assembling active iontronic elements such as CAPode and G-Cap into more complex circuits, e.g. networks of iontronic elements, is an in-plane realization suitable for in-plane interconnection into iontronic circuits and the respective technologies for manufacturing. The original assembly of the gate electrode by sandwiching (Figure 1a) is inappropriate for such a development. A conceptual design of an in-plane G-Cap architecture is required without reducing the ion-depletion capability of the gate. The interplay of conceptual design and rapid prototyping is essential to achieve a high on-off ratio and rapid switching characteristics. Additive manufacturing technologies such as inkjet printing provide such a high degree of freedom in terms of architecture design. [27][28][29] In the following, we demonstrate the realization of highly effective in-plane G-Cap architectures (ipG-Cap) by direct printing of a liquid carbon precursor using a nanopiezoelectric pipet on a 3Dprinter. The resulting active material is a nanoporous additive-free carbon with high specific surface area.
By tailoring the characteristic dimensions of the gate and collector electrodes we achieve ideal balancing for a high on-off ratio. A hierarchical ion-transport channel arrangement is demonstrated to be essential for a rapid switching process.
The newly developed in-plane switchable supercapacitors may not-only pave the way towards ionbased computing architectures. They are also valuable devices for selectively analysing electroadsorption kinetics of electrolyte ions, complex biomolecules or even neurotransmitters and may be applicable for drug delivery and neuromodulation in future. [30][31][32]

Results and Discussion
In the following, we develop a novel in-plane design for planar G-Cap devices. Within this development process three generations of architectures were explored and compared (Figure 2a

st generation G-Cap architecture (ipG -Cap)
For development of the first generation using piezoelectric printing, we printed a large area of nanoporous G-electrode as required for overbalancing. The latter is essential for effective depletion of Before measuring the switching behaviour, we evaluate the performance of the G-electrodes and the W-Cap in the symmetrical two-electrode setup using a proton-conducting PVA/H 2 SO 4 -based hydrogel electrolyte. Device wetting is essential to achieve rapid switching and high on-off ratios. This wetting process is realised with several cycles of adsorption with a negative bias of -1.2 V and desorption of ions with a reversal of the potential to + 1.0 V. We expect an excess of specific capacitance of the gate electrode to be necessary in order to obtain a good switching behaviour. Cyclic voltammograms (CVs) of the simplified interdigitated W-Cap show the typical rectangular shape for different scan rates and a capacitance of 0.2 mF (5 mV s -1 ) is achieved (Table 1a) (Figure 3a). W-Cap is continuously cycled at a scan rate of 50 mV s -1 between 0 and 0.5 V with a capacitance of 100 % before G-switching. After polarising the gate electrode to -1.0 V, the W-Cap capacitance initially increases and then decreases very slowly down to a residual capacitance of 49 %. This result demonstrates effective proof-of-concept but the on-off ratio is inacceptable presumably due to high diffusion limitations. The first cycles even show an increase in the capacitance retention possibly through an increase in the potential difference. Due to the large area of the gate, ions must diffuse long distances to adsorb on the entire inner surface of the G-electrode which limits the electrode utilization. The ion diffusion process effectively takes place in the printed G-electrode (thickness = 220 nm) and the area above (electrolyte gel thickness above = 1 µm). In addition, the G-electrode remains polarised after disconnecting. Complete desorption of the protons from the nanoporous G-electrode and switching back is realised by applying a positive bias potential of + 1.0 V. This forces protons back into the gel electrolyte and the W-Cap capacitance returns nearly to the initial "on" value of 100 % after some cycles. The corresponding CV-curves of different states are shown (Figure 3b). In the cyclic measurement with a scan rate of 50 mV s -1 , the interdigitated W-Cap has a narrow CV-shape with no redox reactions. The CV signal in off-state clearly demonstrates the reduced capacitance and the principle of switching properties in a G-Cap device. After desorption, CV-curves of the W-Cap are similar to the initial curve, proving reversible on/off switching.
However, since the capacitance on-off ratio in this ipG 1 -Cap is not sufficient (only 51 % capacitance reduction), the focus for geometry optimisation has to address diffusion limitations to achieve lower W-capacitance in the off-state. These diffusion limitations are also evident after analysing the rate capability of the G-electrode. The capacitance values decrease significantly by increasing the scan rate (Table 1a). a CV-curves in the on-and off-state using a scan rate of 50 mVs -1 . b W-capacitance evolution during switching.

nd Generation with reduced gate electrode footprint (ipG -Cap)
In order to enhance the utilization of the G-electrode we designed the second generation (ipG 2 -Cap) by reducing the gate area and increasing the height leading also to a slightly higher electrode volume ( Figure 2b) (Supplementary Figure 1c).
For the new gate electrode geometry, cyclic voltammetry measurements were performed in the same setup as for ipG 1 -Cap (Supplementary Figure 2b). Compared to ipG 1 , the capacitance of the G 2electrode is similar at 5.2 mF (5 mV s -1 ) ( Table 1b). The resistance of this optimised setup shows higher values than the area-expanded gate electrode (Supplementary Figure 6).
The characterisation of the switching behaviour for the second geometry is shown in a capacitance retention profile (Figure 4a). Switching measurement is performed under the same settings as illustrated for ipG 1 -Cap and after the same wetting process at -1.  a CV-curves in the on-and off-state (scan rate: 50 mVs -1 ) 20 mVs -1 . b Capacitance retention evolution (scan rate: 50 mVs -1 ). c CV-curves in the on-and off-state (scan rate: 20 mVs -1 ). d Capacitance retention evolution (scan rate: 20 mVs -1 ).

rd Generation ipG -Cap with hierarchical gate electrode for enhanced ion transport
In the foregoing generations a crucial limitation was the rate of electroadsorption in the G-electrode. A further adjustment of the G-electrode geometry considers hierarchical channels to reduce adsorption rate limitations and prevent resulting capacitance loss of the device. For this purpose, the total Gelectrode volume is increased by printing several layers. In addition, structuring of the electrode is performed to introduce wider transport channels and increase the external surface area and ion transport (Figure 2c). The electrode is structured by small electrode fingers with a width of 550 µm and short distances of 100 µm separating each finger (Supplementary Figure 1c, d).
In cyclic voltammetry measurements, small changes in capacitance are observed as a result from patterning of the G-electrode. Compared to the ipG 2 -Cap geometry, the capacitance of the G 3electrode is lower at 3.0 mF (5 mV s -1 ) ( Table 1c). The electrical resistance of the G 3 -electrode is very similar to the G 1 -electrode. Compared to previous G-electrodes, the CV curves are rectangular as desired for a typical supercap (Supplementary Figure 7).
The capacitance switching profile for ipG 3 -Cap shows excellent performance and the switching behaviour ( Figure 5a). Before measurement, a wetting process at a negative bias -1.2 V was applied for this architecture. The ipG 3 -Cap switches off rapidly after the bias electrode is polarised to -1.0 V, in particular as compared to the other geometries. We attribute this rate increase to the hierarchical pattern of the gate G 3 . Thus, the rate of the ion adsorption and utilization of the gate could be increased clearly. Moreover, a very low off-capacitance of only 1.8 % can be achieved in ipG 3  Hence, for improving the reproducibility, the G-electrode was partitioned into two separate electrodes, (Figure 2d) in order to perform an independent electrowetting process by cycling the two parts of G 3electrodes against each other. With improvement of the wetting process, the capacitance switching behaviour should become more consistent and reproducible.
After dividing the G 3 -electrode, an electrowetting process at a lower negative bias voltage of -1.0 V could be used. Subsequently, the switching behaviour was characterised as described above (Figure 5c).
The positive effect of the patterned electrodes is demonstrated by a better accessibility of the ion adsorption. Due to the improved wetting process, a suppression of residual W-capacitance down to 0.5 % can be achieved (Figure 5d). Thus, the structure adaptation and the lower bias potential results in an increase of reproducibility of a fast switching characteristic and a high on-off ratio. a CV-curves in the on-and off-state at a scan rate of 50 mVs -1 for the hierarchical G 3 -electrode.
b Capacitance retention evolution for the hierarchical G 3 -electrode. c CV-curves for the divided hierarchical G-electrode. d Capacitance retention evolution for the divided hierarchical G-electrode.
By optimizing G-electrode architectures for in-plane G-Cap devices in the different generations, the switching behaviour was significantly improved and diffusion limitations were eliminated.
For a quantitative analysis of the switching rate we evaluated the time ( 1/2 ) needed to reduce the Wcapacitance to 50 % of the initial value. While the ipG 1 is slow ( 1/2 = 658 s) the optimized ipG 3 switches rapidly ( 1/2 = 23.3 s) ( Figure 6, Table 2).

Conclusion
A novel switchable supercapacitor architecture (ipG-Cap) was produced by 3D-piezoelectric inkjet printing as a planar device. The essential elements are three nanoporous high surface area carbon electrodes obtained from a non-toxic liquid carbon precursor. An in-plane nanoporous gate electrode effectively controls the capacitance of a working capacitor (W-Cap) by regulating the ion concentration of a gel electrolyte. Rapid prototyping of various gate electrode geometries gives insight into ion adsorption mechanisms, rate-determining transport and adsorption phenomena in the electroadsorption device to optimise the capacitance switching characteristics by electrolyte depletion between on-and off-states, in particular rate and on/off ratio. The target of structural design was to achieve reversible switching with the highest possible on-off-ratio and a high capacitance regeneration. With the lowest off-value of 0.5 % residual capacitance the ipG 3 -Cap even outperforms earlier reports. 23 Due to purely physical ion adsorption of supercapacitors, a fast and repetitive switching behaviour without degeneration is enabled. In particular, pore accessibility at reduced diffusion limitations of ions appeared to be important for the in-plane G-Cap mechanism. Our work is a key step towards interconnecting discrete iontronic elements such as G-Caps and CAPodes into more complex iontronic circuits. Despite the switching rates are not comparable with those of electronic transistors further miniaturization is expected to enable higher frequency operation. We envision multiple G-Caps interconnected in-plane in architectures for realizing logic gates. Such logic gates could be used for the power management of autonomous microelectronic devices or even lead to more complex ion-computing architectures. For in-plane printing CAPodes it will be necessary to also print nanoporous carbons with narrow pore size distribution, which can lead to selective ion screening for precise adjustment of the pore size. 8 Further work is necessary to access also a wider voltage window using organic electrolytes. 33

Carbon precursors
The sucrose precursor was obtained by mixing 1.

Piezoelectric inkjet printing
The piezo-printing for the preparation of G-Cap structures was realised using a BioScaffolder 3.2 and a piezoelectric micropipetting tip Nano-Tip J from GeSiM mbH. The pipetting tip has a drop volume of approx. 400 pL, a chamber volume of 0.7 μL and a maximal operating frequency of 1,000 Hz. The parameters (frequency, pulse width, voltage) were adjusted to form drops at the piezoelectric micropipetting tip (150 Hz, 90 μs, 55 V). The adjustments were made using a stroboscopic camera.  27 . After printing, the dimensions were analysed using a light microscope from Breukhoven Microscope Systems BMS.
The precursor was absorbed through the tip by vacuum and subsequently electrical voltage causes the formation of precursor droplets at the tip. The voltage is modulated by a pulse width and a certain frequency. As a result, drops are formed at the tip in the micron-size range (200 µm). The pyrolysis of the printed structures was carried out at 900 °C for 2 h under argon with a constant heating rate of 150 K h −1 .

Preparation of G-Cap structure
The pyrolysed G-Cap electrodes were contacted with silver conducting paste (Ferro GmbH) on the contact pads. Furthermore, an area of 6 × 9 mm around the electrode was sealed with Kapton ® foil to define the reservoir for the gel electrolyte. Hereafter, the electrode was activated with cold-inductive coupled argon plasma, which was generated with a plasma finger KINPEN (Neoplas Tools) and 10 µL PVA/H 2 SO 4 hydrogel electrolyte was deposited on top. After drying at room temperature, the electrochemical characterisation was carried out.

Electrochemical characterisation
Symmetric interdigital supercapacitors were analysed with a VMP3 potentiostat from BioLogic Science Instruments in a two electrode setup. Cyclic voltammetry was measured with scan rates of 5 and 100 mV s -1 in a potential range of 0 to 1.0 V. The capacitances of the simplified interdigitated W-Cap and G-electrode architectures in three generations were normalized to the device area (Supplementary Figure 8, Supplementary Table 1-3).
The G-Cap was analysed in a CE to ground setup synchronising two ports with a combined working electrode. Cyclic voltammetry during the switching process between W-Cap and G-Cap was carried out in a voltage range between 0 to 0.5 V and a bias of -1.0 to 1.0 V was applied using chronoamperometry.
Dr. Lydia Galle -Faculty of chemistry and food chemistry, Department of Inorganic Chemistry I, Dresden University of Technology, 01069 Dresden, Germany, e-mail: lydia.galle@tu-dresden.de Dr. Julia Grothe -Faculty of chemistry and food chemistry, Department of Inorganic Chemistry I, Dresden University of Technology, 01069 Dresden, Germany, e-mail: julia.grothe@tu-dresden.de