Organic Electronics Circuitry for In Situ Real‐Time Processing of Electrophysiological Signals

The next generation of brain–machine interfaces are envisioned to couple signal transduction, filtering, and sorting on board with minimum power consumption and maximum bio‐integrability. These functional needs shall be mandatorily met in order to design efficient closed‐loop brain–machine interfaces aimed at treating and monitoring various disorders of the central and peripheral nervous system. Here, the pivotal role is highlighted that organic bioelectronics may have in the contextual development of all these three desiderata, by demonstrating a modular organic‐electronics circuit toward real‐time signal filtering. The inherent filtering capabilities of electrolyte‐gated organic transistor are tuned via adjustment of operational conditions and benchmarked in an electromyography experiment. Additionally, a whole‐organic signal processing circuitry is presented, coupling such transistors with ad hoc designed organic passive components. This provides the possibility to sort complex signals into their constitutive frequency components in real time, thereby delineating innovative strategies to devise organic‐based functional building‐blocks for brain–machine interfaces.


Introduction
Animals and humans can be regarded as "machines" living in a complex environment that continuously overloads their senses with information that may be crucial to establish social bonds or to decide the next course of action.Therefore, the brain must constantly select the most relevant information to build and update internal models that are used to predict what will happen next.This complexity leads to the transformation of endogenous and exogenous signals into a sophisticated mixture of electrical and chemical signal patterns that constitute the main currency upon which computations are run by the brain itself.In an extreme reductionist sense, this translates into the ability to select and process only specific frequency-encoded events while ignoring stimuli falling outside relevant bandwidths of information content.While this task is performed in the brain by algorithms and machinery perfected over eras of evolutionary processes, replicating it with man-made information technology necessitates power-consuming off-line signal processing strategies and architectures, inevitably posing technological, energetic, and computational hurdles.Such effort, namely the mimicking of brain signal processing capabilities, is strongly intertwined with the need of understanding brain functioning mechanisms.This task requires the development of efficient brain-machine interfaces, with the ultimate goal of establishing a bidirectional communication between the brain and electronic devices. [1,2]he integration between biological environment and electronics is still an open challenge in disparate research and technology fields, encompassing bio-photovoltaics devices, [3,4] sensors, [5,6] organ-on-chip, [7] implantable electronic devices [8][9][10] and multifunctional neuromorphic components and circuits. [11,12][15] This unprecedented capability arises from the matching between the response timescales of the organic devices and the timescales of biological processes, being both in the 1-1000 ms range. [16]Notably, the characteristic timescales of conventional electronics are in the ns-μs range.The relevant timescale matching constitutes a key-enabling feature of organic electronics as eligible technological platform toward the ambitious goal of interfacing, understanding, and emulating human brain.This implies that the brain-machine interface (BMI) must perform brain signal transduction, amplification, and sorting.
Concerning signal transduction, recent research has been focused on non-invasive micro-epicorticographic (μ-ECoG) electrode arrays which are ultra-conformable to the brain cortex and provide high spatial resolution.On the other hand, miniaturization of epicorticographic electrodes results in a decreased signalto-noise-ratio (SNR) with respect to large electrodes that transduce local-field potentials rather than single neuron spiking. [17][20] These strategies enabled collection of neural spiking data directly from the brain cortex, broadening the clinical applicability of μ-ECoG and potentially overcoming the drawbacks of invasive intracortical electrodes, which are the golden standard for the transduction of single neuron activity.Despite the high significance of such approaches, SNR maximization remains an open issue in non-or mildly-invasive electrophysiology, aiming at high-quality and high-significance recording.
Toward this aim, there has been a sizable effort concerning the integration of in situ amplifiers in miniaturized BMIs.A distinctive approach by organic bioelectronics is to replace microelectrodes with low-voltage/high-gain organic electronic transistors, that act as transconductance-based amplifiers.[23][24] Interestingly, this periodizing advancements in signal transduction and amplification are not mirrored by a comparable counterpart focused on signal sorting.To date the transduced (and possibly amplified) signal is transferred without any further treatment to ex situ computing units, which are deputed to perform off-line data analysis, identifying relevant pathophysiological spectral ranges, detecting and removing artifacts, and classifying diagnostic patterns.This signal treatment is often manually driven by operators, and is time-and powerconsuming, thereby strongly hampering the implementation of closed-loop, operator-independent, theranostic BMIs.
This work aims at providing a solution to these shortcomings, by presenting an organic flexible architecture for real-time signal sorting, prone to miniaturization and integration in BMIs, and specifically designed to be coupled to EGOT-based in situ amplifiers.The proposed technology makes advantage of organic analogs of passive circuit elements, namely organic resistors and electrolyte-organic capacitors, as building blocks for lowand high-pass passive filters to be coupled to EGOTs.Its signal sorting capabilities, as well as the inherent filtering features of EGOT, are herein presented and benchmarked against digitally generated model signals of increasing complexity, starting from "monochromatic" sinusoidal waves up to audio traces, and applied to achieve real-time filtering of human electrophysiological data, here specifically electromyographic (EMG) signals.

Organic Building Blocks
The choice of materials to implement the proposed signal sorting architecture is oriented to achieve prompt translatability, using only polyimide substrates, gold leads and poly (3,4ethylenedioxythiophene) polystyrene sulfonate as organic resistor, as semi-conductive channel and gate electrode material in EGOTs and as interfacial layer in organic electrolyte capacitors (Figure 1a).Application-specific fabrication details are provided in the Experimental section.Passive low-pass (Figure 1b) and high-pass (Figure 1d) filters are built by connecting in series an organic electrolyte capacitor and an organic resistor, and connecting to ground the capacitor or the resistor, respectively.In this work, organic low-pass filters feature 6 Ω resistance and 295 μF capacitor, while organic high-pass filters are built on 14 Ω resistance and 220 μF capacitor.The input voltage, V IN , is supplied at the non-grounded terminal of the series composed by the organic resistor and the organic electrolyte capacitor and the output voltage -termed V L for the low-pass filter and V H for the highpass one -is collected at the shared node.The frequency profiles of organic filters are evaluated in response to monochromatic sinusoidal V IN waves with peak-to-peak amplitude of 200 mV and frequency values ranging from 1 Hz to 1 kHz.Figure 1c,e shows the responses of low-and high-pass filters, respectively, in the form of Bode plots obtained calculating the gain of low-pass filters, gain LP = 20 log 10 (V L /V in ) and the gain of the high-pass filter, gain HP = 20 log 10 (V H /V in ).Cut-off frequencies, f c , are estimated as the frequency values for which gain LP = −3 dB for low-pass filters and gain HP = −19 dB for high-pass ones (i.e., 3 dB less than the maximum gain HP = −16 dB value achieved in the investigated frequency range).The resulting cut-off frequencies are f c ≈ 5 Hz for the proposed organic low-pass filters and f c ≈ 30 Hz for the high-pass ones, with slopes equal to −6 dB per decade and 9 dB per decade, respectively.
Interestingly, the use of organic analogs of passive circuit elements causes two main effects with respect to inorganic solidstate counterparts, namely a red-shift of f c with respect to theoretically expected values (both for high-and low-pass filters) and a substantially negative gain (around −16 dB) of the passed highfrequency band in the case of high-pass filters.Both these effects arise from the fact that, here, an electrolyte compartment is used as a capacitive element in RC filters.If, on the one hand, this enables the engineering of flexible architectures and the exploitation of the high equivalent capacitance values and of the fine tunability of a PEDOT/PSS-electrolyte interface, [25] on the other it brings into play substantial deviations from ideal response, which should be taken into account.The main deviation of the proposed electrolyte organic capacitor from an ideal one is that the former features a non-negligible resistance contributed by the electrolyte solution, R S , which is in series with the equivalent capacitor of the electrical double layer, C DL .It is well known that, for standard RC passive filters, Equations (1) and (2) hold, for high-pass and low-pass configurations, respectively.
where Z C and Z R are the impedances of the resistor and of the capacitor,  is the angular frequency of the input wave,  = 2, R is the resistance of the resistor, and C is the capacitance of the capacitor.The attenuation effect of R S , observed solely in highpass organic filters, is manifest by re-casting Equation (1) into Equation ( 3): Despite following the expected trend of a high-pass filter, namely being 0-valued at null frequency and growing linearly with  when (R + R S )C DL ≪ 1, the V out /V in ratio approaches R/(R + R S ) at the limit of infinite frequency, which necessarily is smaller than one.The entity of this attenuation is indeed a measure of the solution resistance.In the proposed high-pass filter (Figure 1d,e), where R = 14 Ω, the plateau at −16 dB corresponds to R R+R S ≈ 0.156 leading to an estimated R S ≈75 Ω, which is coherent with the structure of the proposed capacitive element of organic RC filters.
Conversely, for the low-pass filter, the addition of R S into Equation (2) yields Equation ( 4): which predicts the ideal behaviour of a low-pass filter (i.e., The red-shift effect observed on the cut-off frequency, fc, arises from the equality between Equations ( 1) and ( 3) or ( 2) and ( 4) which implies that the response of the ideal filter to the input wave with angular frequency  1 , is equal to the response of (real) organic filter to the input wave with angular frequency  2 .By assuming C = C DL , it turns out: since R/(R + R S ) is always less than 1, this means that  2 will be always lower than  1 , quantitatively accounting for the red-shift in terms of solution resistance.

Electrolyte-Gated Organic Transistor
In addition to passive filters, the third building block for the proposed basic sorting-platform is the electrolyte-gated organic transistor.EGOTs were fabricated and characterized, as described in the Experimental Section.Briefly, EGOTs with aspect ratio (W/L) equal to 4 were used to implement the sorting platform and to record EMG signals.A PEDOT:PSS (Clevios PH1000, 5% v/v DMSO, 0.2% v/v GOPS) formulation, diluted ten times in MilliQ, was drop-cast in a well which exposed only the terminal portion of the source and drain leads (area = 0.8 mm × 0.9 mm).A PEDOT:PSS planar gate was obtained by following the same protocol.EGOTs were characterized in commonsource/common-ground configuration (Figure S1a, Supporting Information) by acquiring transfer curves (Figure S1b, Supporting Information) and evaluating their bandwidth.This is done by applying monochromatic sinusoidal waves, at different DC offset values, at the gate electrode while recording the channel current (Figure S1b,d, Supporting Information), as described in the Experimental Section.Organic transistors generally act as low-pass filter, and their frequency band can be modulated not only acting on the device geometrical parameters [23,[26][27][28] but also varying the gate potential.Thus, increasing the density of mobile charges in the channel induces a slight shift of the cut-off frequency from higher to lower values (Figure S1d, Supporting Information). [29]he frequency responses of the presented EGOTs show the expected trend, with a low-pass behavior and a gain set by the gate potential which also shows a minor influence on the cut-off frequency, especially for positive values.The gate DC-offset sets the device cut-off frequency and its (quasi-static) transconductance, g m , value, i.e., the EGOT potentiometric sensitivity in DC operation.Starting from negative to positive V GS , the cut-off frequency of our devices blue shifts from 180 Hz to 220 Hz (Figure S1e, Supporting Information).

Transduction and Sorting of Electrophysiological Measurements
To benchmark EGOT V GS -driven signal filtering and sorting capabilities in real time in the context of electrophysiological acquisitions, EGOTs were integrated in the processing chain of an EMG data recording session.Namely, we quantified the spectral content of muscle activity during steady isometric contraction at a moderate force level.Figure 2a,b depicts the setup for EMG signal acquisition.Data are collected as described in the Experimental Section.Briefly, EMG signals (V EMG versus time, Figure S2, Supporting Information) from the biceps brachii (acting as the prime mover in the current setting) were acquired during isometric contraction required to statically hold a 10 kg weight on the hand for 10 s; EGOT-mediated EMGs are collected as I DS versus time traces.Figure 3c highlights the differences between rest and contraction (weight holding) conditions.As already mentioned, V GS sets the current DC-offset, [11,29] as shown in Figure 2d, which reports the average value of current EMGs versus the correspondent applied V GS .The <I DS > versus V GS profile closely mirrors EGOT transfer characteristics (see Figure S1b, Supporting Information).Each <I DS > value is averaged over time and over trials between seven I DS versus time traces acquired for each applied V GS , see (Figure S3, Supporting Information).The spectral content of V EMG versus time EMG acquisition is reported in Figure 3e.The spectral amplitude is not affected by the gate voltage, as shown by the individual spectral amplitude versus time EMG acquisition for each V GS applied in Figure S4 (Supporting Information).Spectrogram (Figure 3e) shows, as expected, that the EMG spectral content is mainly concentrated between 10 and 100 Hz, regardless of V GS .Muscle activity that is spectrally confined in the beta-band (≈15-30 Hz) is of particular physiological relevance because it most likely has a central neurogenic origin, contributing a major component of the descending neural drive to the spinal motor neuron pool, [30][31][32] and could function as a useful control signal in human-machine interfaces and motor augmentation platforms. [33]Therefore, we focused on the relative spectral amplitude of muscle activity within the beta-band (centered at 20 Hz) with respect to both lower-and higher frequency content.
The transduction efficiency of EGOTs can be expressed by the index Ψ, corresponding to the ratio between the spectral amplitudes of I DS versus time traces and that of V EMG voltage, for each investigated frequency band and V GS value.Ψ versus f versus V GS spectrogram is reported in Figure S5 (Supporting Information), and shows an analogous trend with respect to that observed for transconductance in Figure 2d.
Defining  as the ratio between the Ψ value in the 19-21 Hz range (i.e., Ψ 20 ), and the Ψ value at each considered frequency band (i.e., Ψ fb ), expressing this ratio in dB as described in Experimental Section and plotting it against V GS and frequency, it is possible to obtain a sorting efficiency map, Figure 2f.The effect of V GS in evidencing one frequency band of interest (here 19-21 Hz) with respect to all the other spectral components can be quantitatively expressed by .On this premises, the gate voltage acts as a frequency-specific selector of transduction efficiency.Figure 2f provides a map of operational conditions for EGOT-transduced EMG, expressing which V GS value is most suitable to evidence the 19-21 Hz band with respect to other spectral components ( voltage and current map are reported in Figures S6 and S7 in the Supporting Information).In detail, for electrophysiological evaluations which could benefit from relative enhancing of beta-band activity and attenuation of higher-frequency components, positive V GS should be applied, as testified by the strongly positive  values in the top-right corner of Figure 2f, which indicate a strong predominance of the 19-21 Hz with respect to higher frequencies.This gain with respect to high-frequency spectral content is lost at negative V GS (top-left corner of Figure 2f).On the other hand, to increase the transduction efficiency at 20 Hz with respect to the transduction efficiency of lower spectral components, which would be naturally boosted by EGOTs, Figure 2f shows that EMG evaluations should be carried out with V GS around 0 V (bottom center region of Figure 2f).Additionally, the sorting efficiency map in Figure 2f turns out to be a valuable tool also for applications that demand frequency-invariant transduction efficiency, by identifying the proper V GS regime (V GS ≈−0.4 V, ≈0 for all the investigated frequency bands).

Coupling EGOTs with Organic RC Filters
As already mentioned, EGOTs act as low pass filters, albeit with rather wide bandwidth.This means that, every time EGOTs are used for transduction, the entire information content which falls in the EGOT passing band is collected and amplified and this ultimately results in poor frequency-selectivity.[36] This would result in improved compliance and durability and in the minimization of side effects due to overuse.Bandwidth tuning in EGOTs is usually addressed by acting on geometrical device parameters.On the one hand, it is possible to redshift EGOT's cutoff frequency by engineering large area devices, [37] at the price of losing some spatial resolution when aiming at their translation to epicortical arrays.Conversely, it is possible to widen their frequency range (i.e., to blue-shift their cut-off frequency) by miniaturization, [26,27] at the price of significantly lowering the transconductance.Both these approaches fail in yielding bandselective devices with both high spatial resolution and good amplification capability.At this purpose, by making advantage of the inherent filtering properties of EGOTs and coupling them with organic passive building blocks, it is possible to tighten the bandwidth toward lower frequencies or to implement bandpass filters.The first scenario is achieved by coupling the lowpass filter with an EGOT (Figure 3a).Conversely, the second one arises from the coupling between an EGOT and a high-pass filter (Figure 3c).In these configurations, the filters' output voltages, termed V L and V H , are used as input gate voltages for the EGOTs, keeping a constant channel bias, V DS = -0.7 V.The resulting channel currents are termed I L and I H , for the configurations in Figure 4a,c, respectively.Transconductance versus frequency profiles are obtained by dividing the amplitudes of I L and I H by the amplitude of the input voltages.The configuration in Figure 4a exhibits low-pass behavior with a cut-off frequency as low as 5 Hz (Figure 3b), while the configuration in Figure 4b exhibits a band-pass profile from 5 Hz to 90 Hz (Figure 3d).These two independent architectures can be paired to foster a platform to sort and amplify a common V IN in two differently factorized output signals, as discussed in the next section.

Sorting Circuitry
In Figure 4a the sorting platform architecture sketch is reported.A common V IN is driven through the filter inputs and the resulting V L and V H are connected to two EGOTs gate electrodes.The channel currents are recorded by fixing V DS equal to −0.7 V, as discussed in the Experimental Section.To validate our architecture a first characterization was performed by applying sinusoidal waves with a peak-to-peak amplitude of 200 mV and frequencies spanning from 1 to 1000 Hz as common input signals, while collecting I L and I H .
The current amplitude versus frequency trends are reported in Figure 4c, highlighting the sorting platform frequency-response, closely mirroring the transconductance trends in Figure 4.After validating its operation with monochromatic waves, a sweep wave with 200 mV peak-to-peak amplitude and frequency interval from 1 to 200 Hz was used as input signal.The I L and I H currents versus time are reported in Figure 4b (bottom panel), showing that the designed basic circuit enables to discriminate both in amplitude and in frequency the output responses in real-time.
To benchmark the sorting efficiency of the proposed architecture against a more complex signal, a pre-recorded in vivo somatosensory evoked potential (SEP), which is a well-defined potential modulation in the rat barrel cortex, is digitalized, [38,39] Figure 4d, and driven (see Experimental Section) as input signal.In a translational perspective, we want to test its sensitivity in discriminate some of the relevant frequency bands that are usually evaluated in SEP characterization: alfa,  (5 Hz -15 Hz), beta,  (15 Hz -30 Hz), low gamma,  L (30 Hz-80 Hz), and high gamma,  H (80 Hz -150 Hz), band. [9,40]Digitalised SEP amplitude was increased to 200mV pp to use it as a V IN model signal.I L and I H are recorded by keeping drain-source voltage at −0.7 V.
Defining  as the percentage of total power in the selected frequency interval, the analysis of the frequency content in relevant electrophysiological bands was performed, as discussed in the Experimental Section.Briefly,  L ( H ) is evaluated as the ratio between the power of a specific frequency band in the I L (I H ) versus time trace and the broadband power of the same trace.Figure 4e reports the values of  L and  H versus the frequency bands.Even with complex event related signals, the sorting architecture successfully discriminates its different frequency domains.In particular, the low-pass branch of this architecture selects  and  frequency bands, as testified by the significantly higher values of  L with respect to  H in these frequency bands.Conversely, frequencies higher than 30 Hz are preferentially transmitted by the band-pass branch (i.e.,  H > L for low-gamma activity).In the high-gamma frequency range there is a loss in transduction efficiency for both branches, since the EGOT implemented in this design has a cut-off frequency of roughly 120 Hz and, as a consequence, the gain of both branches drops at high frequencies.
To further mark the frequency-related gain-switch between the sorting platform branches, it is possible to estimate a sorting factor, reported versus frequency in Figure 4f.The sorting factor is computed by dividing the I L and I H output signals in 13 equally spaced intervals, as described in the Experimental Section, and calculating the power of each frequency interval.The ratio between the I L power in a specific frequency band and the I H power in the same frequency band results in the sorting factor, which is a quantitative estimate of the preferential routing of that frequency among the two branches of the sorting architecture.If, for a given frequency interval the sorting factor exceeds unity, the signal content in that frequency interval will be preferentially transduced as I L .Conversely, frequency intervals that yield sorting factors minor than one will be preferentially transduced as I H .In the proposed architecture, the sorting factor equals one in  the 20-25 Hz interval, Figure 4f, meaning that spectral content in this frequency interval remains unsorted and can be found with equal power in I L and in I H .

Signal Factorization and Independent Routing with Organic RC Filters
As a last example of the versatility of our material/device platform, we show the design and operation of a standalone sorting circuitry by pairing RC organic filters.We demonstrate it on the equalization of an analog audio signal.Audio signals were chosen as archetypes of frequency-encoded information, with relevant spectral content across a wide frequency range, namely between 20 Hz and 20 kHz.Since our architecture is tailored to operate in the 0.1−1500 Hz range relevant to electrophysiological signals, we had to adopt an out-of-the-box modification of the input signal to adapt it to the doable processing window of the circuitry.For this, by time-stretching an audio trace of a factor 25, as described in the Experimental Section, we bring it in the desired operational range, while preserving its information content.In Figure 5 the signal processing chain of an audio sorting experiment is shown.A common audio signal input to the paired organic RC filters is split in real time into two differently equalized output signals.In detail, an audio track, namely 7 seconds of the introduction of "Sweet Child of Mine" by Guns 'n Roses, is acquired as voltage versus time, as discussed in the Experimental Section.The voltage track is used as V IN for the filter architecture and V L and V H are collected.Figure 5 (top center) depicts the spectra of the untreated input signal, the low-passed and the high-passed signal spectra are shown in Figure 5 (bottom).By time stretching these signals of a factor 1/25 it is possible to hear the different equalization of the two branches, their related audio files can be downloaded as Audio S1 and Audio S2 in the Supporting Information.From the spectra, it is possible to notice how the normalized amplitude of the low-pass and high-pass spectra strongly differ in terms of frequency contribution.A major frequency content between 120 and 250 Hz can be observed in the low-passed spectrum with respect to the high-passed one, in which the dominant spectral contributions are above 250 Hz.

Conclusion
The ability to separate and independently route frequencyencoded information from a complex signal, as is typical in biological environments, may be essential for real-time monitoring of such activities.This task, which encompasses transduction, amplification, and sorting of the acquired signals, will enable their use as discrimination tools for pathophysiological pattern recognition.This work presents the possibility to merge the well-known EGOTs filtering and amplification capabilities, tested and verified at the benchside and with electromyographic signals, with an organic flexible architecture for real-time signal sorting that is prone to downsizing and integration in BMIs.Although it is specifically designed to pair EGOT-based in situ amplifiers, the signal sorting architecture can also be coupled with electrode technology, offering a novel operational paradigm for real-time data treatment.This possibility to segment information in an in operando condition is oriented to translation in bio-integrable circuitry.At this purpose, organic filters and EGOTs were implemented on polyimide and using PEDOT:PSS as active moiety, hence exploiting some of the most used materials in organic bioelectronics and extending the available toolset for direct integration of filtering and sorting circuitry in ultra-conformable BMIs, which constitutes a major advantage of this approach with respect to classical ex situ filtering strategies.
Electrolyte-Organic Capacitor: Two Au electrodes supported on the same Kapton sheet (active area of each electrode = 0.3 cm 2 ) were coated by means of potentiostatic electrodeposition of PEDOT/PSS (5s 0.2 V, then 0.8 V in charge limit control, up to the desired charge density), starting from an aqueous electrolyte containing 10 mm EDOT and 5 mg mL −1 NaPSS.Investigated charge densities were 83 and 166 mC cm −2 , which resulted in equivalent capacitance values of 220±10 μF and 295±18 μF at 100 Hz (measured by a precision LCR bridge, Thurbly Thandar Instruments) when exposed to the PBS electrolyte.
Organic Resistor: PEDOT:PSS formulation was diluted 1:2 with MilliQ H 2 O and drop-cast between two Au electrodes of the test patterns designed for EGOT fabrication (0.5 μL on a 0.72 mm 2 area).Average resistance value was 2.9±0.2Ω. Series combination of organic resistors were used to achieve the desired resistance values in the final signal sorting architecture.
Electrolyte-Gated Organic Transistor Fabrication: PEDOT:PSS formulation was diluted 1:10 with MilliQ H 2 O and drop-cast in confined areas of test-patterns (0.5 μL on a 0.72 mm 2 area) to achieve a PEDOT:PSS semiconductive channel (W/L = 4) and a co-planar PEDOT:PSS Gate electrode.Average thickness was 1 μm, as measured by atomic-force microscopy (XE7 AFM Park System, tapping mode) Steady-State EGOT I-V Characterization: Steady-state electrical characterization was performed using a two-channel Keysight B2912A Source/Measure Unit and PBS electrolyte.For transfer characteristics, gate voltage was linearly swept between −0.8 to 0.8 V by steps of 2 mV (scan rate = 100 mV s −1 ) while drain voltage was kept constant at −0.7 V.
AC Characterization: An Agilent33220A arbitrary waveform generator was used as source of model signals.EGOT bandwidth was evaluated in a common source/common ground configuration by sourcing the gate potential with sinusoidal waves with amplitude equal to 200 mVpp and frequency ranging from 1 to 1000 Hz, recording I DS values in time.Organic filters and sorting platform frequency-dependent profiles were characterized by applying at the Vin sinusoidal waves with amplitude equal to 200 mVpp and frequency ranging from 1 to 1000 Hz, recording the voltage output and the current output with a Source/Measurement Unit (SMU) (Keysight B2912A) in current mode and voltage mode, respectively.Prerecorded somatosensory evoked potential was digitalized and stored in a Agilent33220A arbitrary waveform generator.This signal and a sweep signal with frequency ranging from 1 to 200 Hz were exploited as model signals to characterize the sorting platform filtering properties.Both signals were used as Vin with a final amplitude equal to 200 mVpp and the current output was recorded with a Source/Measurement Unit (SMU) (Keysight B2912A).
Audio Track Acquisition and Rescaling: Audio track was acquired as voltage versus time by directly connecting the mini-jack output of a laptop soundboard to the input of an Agilent Technologies DSO6012A oscilloscope 100 MHz 2GSa/s, sampling at 50 kHz.The collected track was reproduced using Matlab Audio Toolbox at a 25-times lower speed and used as Vin for the coupled organic filters.The output voltages were collected with a Source/Measurement Unit (SMU) (Keysight B2912A) in currentcontrol mode at a sampling frequency equal to 500 Hz.The two filtered output voltages were played back at normal speed using Matlab Audio Toolbox.
EGOT-EMG Set-Up and Acquisition: Surface EMG signals were acquired by an Isolated Patient Amplifier System (Digitimer, D360) in terms of electromyographic voltage, V EMG , which was notched at 50 Hz and amplified 100x by the acquisition system.The muscle of interest -i.e., the biceps brachii (short head) -was identified via standard palpation procedures, and Ag/AgCl electrodes (ø = 5 mm) were placed on the muscle belly with an interelectrode distance of ≈5 cm.Upon the acquisition of electromyographic activity, V EMG was contextually acquired by a Source/Measure unit (Agilent Technologies, B2912A) voltage and transmitted via a bathing needle-shaped electrode to the operational electrolyte of an EGOT (PBS 1 M).EGOT channel current, I DS , was acquired by another Source/Measure unit, which sources constant V DS = −0.7 V and V GS ranging from −0.8 to 0.8 V with steps of 100 mV.All signals were acquired at 5000 Hz.Data were collected on a healthy subject (the co-author FR; male, age: 26 years).The subject was comfortably seated and held his right arm bent in a supine position with his elbow resting on a fixed support.Each recording was manually initiated by the experimenter after placing a 10-kg cast iron weight on the subject's right hand (ensuring that the hand was already stabilized after loading) and lasted for 10 s; at the end of the 10-s recording, the weight was removed from the hand and a short pause (≈1 min) was taken before starting the next recording.A total of 7 recordings were performed for each V GS in a fully randomized order to exclude potential order effects, e.g., due to muscle fatigue.
Data Analysis: Data analysis and graphing had been performed with Origin 2016, Matlab (version 9.10, Mathworks, Natick, MA, USA) and the FieldTrip toolbox. [41]Specifically, to examine the data in the time-frequency domain, the Matlab function spectrogram was used setting a window length of 100 samples and 90% of window overlap.Power Percentage was estimated by filtering the data with Matlab function designfilt in four frequency bands.For each filtered signal and for the broadband signal the power was estimated with matlab function bandpower and the power of each band-filtered dataset was divided by its brodband signal power.Sorting factor was estimated by exploiting Matlab function bandpower dividing the signals in 13 frequency band and subsequently calculating their ratio.Spectral analysis of the audio signals had been performed with Steinberg Cubase 5.0.
To quantify the spectral content of the EMG signals, conventional Fourier analysis was applied in combination with the multitaper method [42] which allowed for increasing the signal-to-noise ratio and controlling spectral smoothing.Before converting into the frequency domain, the I DS versus time traces were high-pass filtered (cut-off: 0.1 Hz; two-pass Butterworth filter, third order for each pass) to remove the V GSdependent DC-offset.I DS and V EMG voltage (10-s time-series) were multiplied with a set of 39 orthogonal tapers (discrete prolate spheroidal sequences or Slepian functions) yielding 2-Hz spectral smoothing.The taper-specific Fourier-based amplitudes were estimated for 100 logarithmically (log10) spaced frequencies between 5 and 140 Hz, averaged over the tapers and then across the time-series (n = 7; see above) to derive the spectral amplitude for each V GS .The transduction efficiency of EGOTs (Ψ) was calculated as the ratio between the spectral amplitudes of I DS versus time traces and that of V EMG voltage (for each frequency and V GS ).Finally, the ratio between Ψ values averaged between 19 and 21 Hz (Ψ 20 ) to all band-specific Ψ values (Ψ fb ; computed by averaging Ψ values between two logarithmically increasing adjacent frequencies in the range 5-140 Hz) was calculated and converted to decibels for each V GS as follows: Time-frequency representations of V EMG were obtained by applying the same analytic approach described above (i.e., Fourier with multitapering) on a shorter 1-s sliding window that was advanced over the data from 1 to 9 s in 50-ms steps.Spectral smoothing was set to 2 Hz (as for the 10s time series analysis; see above) by multiplying each 1-s window with 3 orthogonal tapers.

Figure 1 .
Figure 1.Organic building blocks.a) Schematic representation of the organic building blocks.b) Connection layout of the organic low-pass filter.c) Frequency profile of a representative organic low-pass filter, horizontal dashed black line intersects y-axis at −3 dB with respect to the maximum value, meanwhile vertical dashed black line intersects x-axis at the cut-off frequency value.d) Connection layout of the organic high-pass filter.e) Frequency profile of a representative organic high-pass filter, horizontal dashed black line intersects y-axis at −3 dB with respect to the maximum value, meanwhile vertical dashed black line intersects x-axis at the cut-off frequency value.

Figure 2 .
Figure 2. Filtering and sorting of electromyographic signal.a,b) Setup and the two different experimental condition for the EMG data acquisition.c) Recorded I DS versus time trace highlighting the rest condition (green double headed arrow) and contraction condition (orange double headed arrow).d) Current DC-offset set by V GS .Each point is the result of the average value of seven current EMGs traces versus the correspondent applied V GS .e) Spectral content of V EMG versus time EMG acquisition.f) Sorting efficiency map.

Figure 3 .
Figure 3. Coupling EGOTs with organic RC filters.a) Connection layout of the organic low-pass filter paired with EGOT.b) g m versus frequency profile of the configuration in (a), black solid line is a double-boltzman fit of the data, intended as a guide for the eye.c) Connection layouts of the band-pass filter, achieved pairing the EGOT with the organic high-pass filter.d) g m versus frequency profile of the configuration in (c), black solid line is an inverse polynomial fit of the data, intended as a guide for the eye.

Figure 4 .
Figure 4. Sorting platform.a) Sorting platform connection layout.b) Sweep wave exploited as V IN for the sorting platform (top panel) and I H and I L current output (bottom panel).c) The frequency response of I H and I L determined by supplying the sorting platform with a monochromatic sinusoidal wave.d) Pre-recorded in vivo somatosensory evoked potential used as input for the sorting platform.e) percentage of total power in frequency electrophysiological interval alfa  5-15 Hz; beta  15-30 Hz; low-gamma  L 30-80 Hz; high-gamma  H 80-150 Hz. f) Sorting factor.Dash-dot line highlights the unitary gain for the sorting platform.The intersection between this line and the experimental point sets the frequency turning point at 28 Hz, black solid line is a two-phase exponential decay fit of the data, intended as a guide for the eye.

Figure 5 .
Figure 5. Audio trace transduction chain.From left to right: seven seconds of the audio trace are used as input signal.Unfiltered signal spectrum is evaluated.The voltage trace is then time-stretched such as its frequency content falls in the operational range of the proposed filters.The stretched signal, which lasts 175 s (top right), is used as V IN for the paired filters; V L and V H are recorded.A re-stretching of the signals back to the original duration ensures audio reproducibility.V L and V H traces (second row, on the left) are used to evaluate low-and high-passed signal spectra (bottom panels).In the low-and high-passed spectra, the blue and the red box highlight the spectral region of maximum relevance.