The photocurrent performance of the protein films was studied using the M_cI_bS_pM_a device architecture, either in the on-tip on-dip, planar or stacked architecture, depending on the measurement requirements (see Figure 1A). While the on-tip on-dip architecture allows the easy exchange of insulating buffer solution I_b, the planar architecture allows highly controlled studies with non-transparent counter electrodes M_c and the stacked architecture a close and well-defined electrode spacing between the active electrode M_a and M_c. For the planar architecture, electrodes strips of 2 × 26 mm were separated by 4 mm gap. For the stacked architecture the bulk electrodes were separated by a 1 mm thick glass spacer (allowing the comparably easy removal of the liquid medium) bonded using thermally curable spacers (60 μm, 110°C, Solarex). In the case of the temperature-dependent studies, a sealed device architecture was used to minimize evaporation of the I_b layer upon heating. For this, the bulk M_a electrode was covered with a film of 500 nm parylene-c around the active layer deposition site, into which the protein film was subsequently deposited as usual. A patterned Au electrode was used as M_c, and adhered onto the parylene-c film using the above mentioned thermally curable spacers. After filling with the I_b solution, the device was sealed with a glue gun.

UV-Vis absorption spectroscopy was used to confirm the purity and stability of the protein samples upon storage, buffer exchange and film deposition. The influence of salt type, salt concentration and pH on the absorption characteristics was also investigated, and results compared to literature. UV-Vis spectra were recorded on a Shimadzu UV-2450 or UV-3100 PC spectrometer in the range of 200–800 nm against buffer solution. In the case of reconstituted samples of TR and SyHR, or in the case of BR PM, an integrating sphere was used to remove the effect of scattering.

The UV-Vis spectra of films on glass or ITO substrates were recorded through a perforated plate (d = 6 mm) fixed, normal to the beam, on the cuvette holder of the standard set-up of the UV-Vis machine, against clean reference substrates. Due to the absorption of the ITO and glass substrates at wavelengths below 280 nm, no information in this spectral range could be obtained. UV-Vis spectra of selected films were measured on the day of film fabrication, after storing and rehydration, as well as before and after photocurrent measurements.

Where necessary, spectral deconvolution of the absorption spectra was performed using a combination of 3 log-normal curves following the strategy reported in [52].

Photocurrent measurements were performed, following the procedures described in [34, 53]. The modulated light of an LED of 530 nm was directed through a fibre optic cable (d = 1 mm, NA = 0.39) onto the active layer under study through a collimator, ensuring a distance-independent beam of about 5 mm diameter. The light power was regulated using a home-built setup and calibrated using a commercial silicon photodiode. In the case of the on-tip on-dip measurement system, no collimator was used, resulting in a linear distance dependence of the diameter of the beam on the sample (see Supplementary Figure S3). The distance was optimized prior to measurement to ensure the illumination of a large (or the entire) area of the protein film. The frequency f_light of the square-wave modulated light was regulated by a function generator (Hewlett Packard 8116A or Texio FG-274). The measured transient photocurrent signals I_PC of the MISM devices were amplified using a low-noise transimpedance amplifier (Femto DLPCA-200, 10⁸–10⁹, DC mode, unless otherwise stated), and recorded on a digital oscilloscope (Tektronix TBS1052B or TDS5104B).

Each device was tested for time-dependency (long-term stability), as well as dependence on f_light. It is well established that upon increasing f_light the transient waveform changes to a square waveform (defined as 100% signal strength for bandwidth calculation [31]) and then a triangular signal, until the bandwidth f_BW of the device is reached and the light modulation speed exceeds the speed of response of the MISM device. For comparison, most data were acquired at a modulation frequency low enough to allow sufficient discharge of the device during the light-OFF period.

The equivalent circuit of liquid systems must be described using a combination of capacitive and resistive components [54]. This leads to a frequency dependent complex impedance, with the magnitude Z(f) and the phase shift θ(f), between the real component Re(Z) (f) = Z(f)cos(θ(f) π/2), the resistance, and the imaginary component Im(Z) (f) = Z(f)sin(θ(f) π/2), the reactance of the system.

The frequency dependent impedance Z(f) and θ(f) of the electrolytic buffer solutions used for I_b were studied between two parallel bulk ITO electrodes (40 ¹ x 9 mm), separated by a spacer of 4 × 1 mm (sealed with the above mentioned thermally curable spacers) within the range of 4 mHz–100 kHz in 100 steps (on the log scale) using a home-written labview program to control the LCR Meter ZM2371 with ZM2392 low-impedance test leads (NF corp., speed vslow, voltage 100 mV, trigger delay 1s), after correcting for contributions of the cables and electrodes. The dependence of the curves on the concentration of NaCl or Na₂SO₄ was measured through titration of a concentrated salt solution in a 4 mL UV cuvette (filled with 3.5 mL solution) in the range between 0 and 1 M in several steps on the log scale. The dependence on pH was measured for buffers of 10 mM MOPS with and without 10 or 25 mM NaCl or 12.5 mM Na₂SO₄ within the pH range of 4.9 – 8.2, confirmed prior measurement using a pH meter (Seven Compact, Mettler Toledo). The LCR ITO electrodes were washed with DI water and ethanol, and dried before exchanging the sample medium. For the temperature dependent measurements in selected I_b (25 mM NaCl, with and without 10 mM MOPS), the cuvette was placed in an oil-bath on a stirrer hotplate, which was equilibrated for 10 min between measurements (using a stirrer) in the range between 20°C and 70°C in 5°C–10°C steps (both heating and cooling cycles), after degassing the sample through centrifugation to avoid the formation of air bubbles during the heating process.

The light induced pH changes of a suspension of E. coli cells expressing SyHR were studied depending on the wavelengths of light and salt content. For this, the cultured E. coli cells were washed three times in a solution containing 300 mM NaCl or 100 mM Na₂SO₄ to remove residual culture medium and resuspended in the same solution until OD₆₆₀ = 7. pH changes were monitored using a LAQUA F-72 pH meter equipped with a micro ToupH electrode (Horiba Ltd., Japan) at 25°C controlled by a thermostat. Each cell suspension was kept in the dark, followed by a 3 min long illumination with a 300 W Xe light source MAX-302 (Asahi Spectra Co. Ltd., Japan) through bandpass filters of either 400 ± 10 nm or 520 ± 10 nm. Measurements were repeated under identical conditions after the addition of the protonophore carbonyl cyanide-m-chlorophenylhydrazone (CCCP, final concentration, 10 µM) and further addition of tetraphenylphosphonium (TPP⁺, final concentration, 20 mM).

The incorporation of TR as the photoactive layer S_p in the differential MISM photodetector resulted in a clear photocurrent response, with responsivity improved by an order of magnitude compared to the devices reported in 2015, based on SrSRI [34] (see Figure 3A). This can be partly explained by the improved overlap between the used LED and the absorption maximum of TR at 530 nm, which gives TR in solution, as well as in the deposited films, a characteristic pink appearance. Furthermore, TR has been reported to be a proton pump of comparably high efficiency and of high thermal stability (in 2014 TR was record holder with stability over 3 h at temperatures as high as 75°C [45]).

The magnitude of I_PC, as well as its high stability with time, allowed the conductance of a full characterisation of the TR MISM response. The characteristic MISM PD frequency dependence of I_PC can be observed, reaching bandwidths f_BW exceeding the 100 Hz range (see Figures 3C, D) depending on the exact device architecture. The rise and decay of I_PC could be well fitted with the introduced single exponential rise and single exponential decay function (see Figure 1B), with τ_decay in the 10s of ms and τ_rise in the ms range. It should be noted that the rise time is already affected by the rise of our amplifier, therefore discussion will be mostly limited to the decay times. When comparing the light-ON and light-OFF peaks, it becomes evident that τ_decay is lower under illumination, with concomitant increase in amplitude I₀, leading to a balanced charge Q_I being moved back and forth during the light-ON-OFF cycles.

For organic MISM devices, the τ values could be associated with the resistive and capacitive elements in the equivalent circuit [32]: The photovoltage source in the S layer V_S could be shown to be in series with a resistor R_S, corresponding to the active layer, followed by a capacitive element C_S of the S layer, connected in parallel with the series R_I and C_I elements, as well as R_ec corresponding to the insulator and the electronic connections, respectively. Although the interpretation of R_S and C_S is not trivial for protein films (especially in the case that no charge-transfer but chemical doping of the film occurs), as would be the assignment of Fermi levels (found essential for the description of the polarity of response in organic MISM devices [53]), certain aspects of this model will be considered and terminology of it used, before considering the effect of the photocycle kinetics of the proteins under light illumination. According to the model [32], the total series resistance R_S + R_I + R_ec strongly affects the signal magnitude, while both τ values are associated with the RC time constants of the parallel circuits, i.e., (R_S + R_I + R_ec)*C_I and R_x*C_S, with R_x being the smaller of R_S or R_I. The amount of charge Q = Q_S + Q_I stored internally in the organic device is directly proportional to the sum capacitance of the layers Q = V_s*(C_S + C_I). In contrast, for a chemically doped film, the effective charge Q_S should determine the photovoltage V_S (through V_S ∼ Q_S), responsible for charging I_b capacitor C_I: Q_I = V_S*C_I ∼ Q_S*C_I, thus being proportional to the charge of the protein film, and likely to increase with protein layer thickness. The time-constant associated with the charging of the S layer in organic MISM devices, is likely to be not only related to the electronic components of the S_p layer, but also of the kinetics of the charging process itself (influenced by the time constants k₁ and k₂ as illustrated, e.g., inFigure 3B) (vide infra).

Consistent with previously reported data on organic MISM devices [34], the signal intensity increased asymptotically, when increasing the salinity of I_b, and therefore reducing the resistance in the active layer R_I (see Supplementary Figure S5). A similar increase could be obtained upon increasing the salt content in S_p during film deposition. However, this hampered homogeneous film formation and led to a lower stability of M_a even for the relatively inert Ag. In general, salt was therefore added to I_b and the film was deposited with the minimum amount of salt required for ensuring protein stability. Concomitantly to the increase in I_pkpk the speed of the PC response increased (and thus τ decreased, as expected from a reduction in R_I or R_S). The speed did (as expected) also increase when decreasing the thickness of the S_p layer (see Supplementary Figure S6). Simultaneously, however, I_pkpk was reduced, likely due to the reduced number of TR proteins in the film, and a possible increase in C_s, lowering V_s.

Furthermore, the characteristic sub-linear light intensity dependence of MISM devices [32, 53] could be observed. However, in contrast to devices having an organic S layer, no significant changes in signal shape with increasing light intensity could be observed (see Supplementary Figure S7), within the resolution of the TR I_PC traces, suggesting that the electronic parameters (R and C), or the corresponding lifetimes τ, are not affected by the increasing number of charges within the protein film [32], and the sublinearity of the increase in responsivity with increasing light intensity can be directly correlated with a reduced charging efficiency of the S_p layer.

As can be seen in Figure 2E, the wavelength dependence of the photocurrent response for the TR MISM device shows a high overlap with the absorption spectrum of TR, confirming the that optical activity of TR is the primary reason for the observed photocurrent signal. The slight red-shift in the position of maximum responsivity to λ_{Resp_max} = 540 nm could be a result of the strong linear increase in light power (towards longer wavelengths) of our light-source, and the associated even higher increase in the number of available photons for longer wavelengths.

The polarity of the photocurrent response corresponds to the negative component of the EDL building up at the I_b/M_c interface (see Figure 1C upper case). The photocycle of TR [44] consists of a number of intermediate states with corresponding rate constants k_i. In two of these transitions, a proton is either moved in or out of the TR protein moiety. For TR it has been reported that photoactivation results first in a proton uptake, followed by a proton release [44], as illustrated in Figure 3C. Under constant illumination protons would thus accumulate in the TR film, inducing a positive doping of the S_p layer, drawing negative charge to M_a via the external circuit, and counterbalancing the negative polarity of the EDL at M_c. Whether or not the doping leads to any charge transfer between the active metal M_a and adjacent S_p layer cannot be extracted from our data. However, we have previously reported that in the case of dry films of SrSRI, the signal polarization could be inverted, essentially inducing an MISIM architecture, where the S_p layer is electronically separated through an additional insulating (here non-responsive protein film) layer [34]. In the case of organic films, the MISIM architecture prohibits any charge transfer from S to M_a, which greatly enhanced the speed of organic MISIM devices, compared to their MISM counterparts [56]. In the case of SrSRI, it was shown that under such conditions, the polarity of response can then rather be explained by the lower case depicted in Figure 1C, which will be further discussed when considering the BR photocell.

The good thermal stability of TR, also reported for elevated temperatures, facilitated a temperature dependent measurement of the photocurrent signal (see Figure 4). Upon increasing the operation temperature of the TR-MISM photocells gradually to 75°C, a linear increase in signal magnitude (I_pkpk and I₀) and a concomitant increase in signal speed was observed (see Figure 4C–lower trace). When comparing the values between RT and 66°C, I₀ increased linearly almost 3x, τ_decay decreased 5x, while τ_rise less than 2x. The quasi-linear decrease in τ can be explained by a power relationship of its corresponding rate constant 1/τ with T (see Supplementary Figure S9), which strongly increases especially at elevated temperatures. The comparably small increase in I₀ compared to that in τ_decay results in a more then 2fold decrease in “moved” charge Q_I in the TR MISM device with T (see Figure 4C–middle trace), which could correspond to either an increase in insulating layer capacitance C_I or decrease in photovoltage V_S in the protein layer (and thus its accumulated charge Qs).

In order to correlate the observed changes with the behaviour of the protein film and/or the photocycle kinetics of TR, the effect of temperature on the RC time constants of the buffer solution I_b needs to be considered and subtracted. Therefore, the impedance of the isolated I_b layer was studied in more detail using impedance spectroscopy.

The impedance spectra of the studied electrolyte or buffer solutions of moderate concentration, measured in the frequency range relevant for low-speed MISM devices with f_BW < kHz, could be well described with the proposed equivalent circuit [32], composed of a parallel RC section, associated with bulk effects (according to the Debye model describing the bulk capacitance of the cell and the ionic conductivity in the bulk), a CPE related to the EDL and a resistor associated with the resistances of the electrical contacts and cables (see Figure 2).

For TR, no information on the pH dependence of its photocycle kinetics is available, and therefore another well studied microbial rhodopsin protein Bacteriorhodopsin BR was chosen to study the dependence on pH on the photocurrent signal. Since its early discovery in 1971 [46], BR has evolved to a model protein for light-driven proton pumps, found in the membrane of microbes, fueling the synthesis of ATP. Its photocycle has been studied in great detail (see ref in [47]). In nature, BR occurs in the form of a two-dimensional hexagonal lattice of BR-trimers in the so-called purple membrane (PM), which is a distinct section within the cell membrane of the archae Halobacterium salinarum, in which BR is the only protein present. BR in PM can be isolated directly from the Hs culture, leading to a comparably high rhodopsin concentration within the membrane (protein:lipid ratio approximately 3:1 [58]) compared to reconstituted samples (as used in the study of TR, SrSRI and SyHR), and thus resulting in a comparably high quantum yield of deposited films. However, the quantum yield at 530 nm is reduced by about 50%, due to a red-shift in the absorption maximum of BR towards 568 nm at ambient pH values (see Supplementary Figure S12) [59], compared to TR. For low pH, a purple to blue transition can be observed, related to the protonation of an amino-acid Asp85, located close to the Schiff base (SB) and the retinal chromophore. Asp85 acts in its deprotonated form as primary proton acceptor from the SB during the photocycle. The protonation of Asp85 at low pH, and thus the occurrence of a shoulder in the absorption spectrum of BR, located at around 600 nm (see Supplementary Figure S12), strongly reduces the light-induced proton transport [59]. The pKa of Asp85, with its value of approximately 2.6 [59], sets therefore the low limit of pH for BR functioning as a proton pump.

For BR featuring a deprotonated Asp85 a photocycle can be triggered by light absorption through the retinal chromophore which initialises proton translocation. This light induced proton translocation is known to be a multistep process driven by changes in the proton affinities (i.e., their pKa values) of crucial amino acid side groups. In general, for pH below the pKa of the side groups (i.e., pH with comparably high H⁺ concentration in solution) the majority of amino acids are found in their neutral protonated state (A-H) or eager to receive a proton if being deprotonated by a previous reaction, while for high pH values, the amino acids are deprotonated (A⁻) or eager to deprotonate. In analogy to TR, two steps in the photocycle are crucial for the MISM applications, the first involving the proton release into the medium from a so-called proton release complex (PRC), and the second involving the proton uptake from the medium by an amino acid, Asp96 in the case of BR, featuring pKa values of 5.8 and 7.2–7.5 during the corresponding steps, respectively. Changing the pH of the surrounding medium has therefore pronounced effects, not only on the order of proton uptake and release, but also on the kinetics of the photocycle (see Figure 6) [59]: In the case of a pH ensuring a deprotonated PRC (or facilitating the deprotonation of the PRC) (i.e., pH > 5.8), a proton is released from BR via the PRC to the solution at the extracellular side shortly after photoisomerization of the retinal. A slow proton uptake from the intracellular side follows before the activated protein fully decays to its original state (see left side in Figure 6). The rate of proton uptake (and thus k₂) is further slowed down upon increasing the pH above the pKa value of the proton acceptor Asp96 [59, 60]. In contrast, at pH below 5.8 (see right side in Figure 6), the slow proton uptake takes place before the unfavourable proton release, which nevertheless happens shortly after the proton uptake by Asp96 from the PRC-H complex, and thus the sequence is inverted, changing the charge of the light activated BR PM film from being negative to positive. The proton uptake can be further delayed (and thus k₁ decreased) when decreasing the pH below the pKa of an amino-acid (pKa_Asp35 = 4.5) which slows the required deprotonation of Asp96 by only slowly transferring its proton to the deprotonated SB, which is required prior to uptake of the proton from solution. It should be noted that conformational changes in the proteins, e.g., the flip of Asp96 after deprotonation from an orientation facing the SB to one facing the cytoplasmic surface or relocation of mobile internal water molecules during transitions between photocycle intermediates ensure the directionality of the photocycle reaction [59].

As expected, a stable photocurrent response can be observed in BR MISM devices (see Figure 7B), along with their characteristic dependence on light illumination frequency associated with the differential device architecture (see Supplementary Figure S13). At pH ≥ 6.2, the device shows a positive light-ON peak, which increases in magnitude with increasing pH. The signal intensity observed at high pH values is increased compared to TR by another order of magnitude, likely due to the denser protein film, and fast I_PC response. The device bandwidth can approach the kHz range, and thus also increased by an order of magnitude compared to TR. This is consistent with the reported increase in photocycling rates of BR at room temperature being about 10-fold faster compared to TR, but similar to TR at elevated temperature [44]. The increase in I_pkpk with pH is accompanied with an increase in the τ_decay (see Figure 7C–top trace), while τ_rise remains stable, ensuring the pH independent bandwidth. It should be noted that an increase in τ_decay would concomitantly cause an increase in I_pkpk until the decay of I_PC is slow enough to not intercede its signal rise [32], explaining the steeper increase of I_pkpk with pH compared to I₀ (see Figure 7C–bottom trace). The simultaneous increase in peak height and width leads to a significant overall increase in charge Q_I moved between the electrodes for a given light signal (see Figure 7C–middle trace).

When lowering the pH below the critical value of 5.8 to, e.g., 4.5, the polarity of the photocurrent signal is inverted (see Figure 7B red trace). The observed signal inversion would agree with the reported change in the order of proton uptake and release [59]. The comparably low signal to noise ratio and large fluctuations at pH values below 5.8 hamper reliable characterisation of its I_PC. However, it can be seen that the bandwidth of the PC signal is significantly reduced, caused by an increase in τ_rise which approaches the value of τ_decay. When considering τ_rise, its sudden increase for low pH values seems surprising considering the almost constant R_IC_I values of the I_b layer, obtained from the impedance studies (see Figure 5C). However, the rise can only be rate limited by the comparably fast reacting I_b layer, if k₁ (or the contribution to k₁ happening directly after light illumination, which is most of it, in the case of low pH values – illustrated in the position of the “lightening” in Figure 6) is faster than the I_b response. This can be the case for BR at elevated pH, but at low pH the initial charging of the BR film happens with slow kinetics. In such a case, the increase in charge in the S_p layer controls both the rise and decay of the PC signal. Furthermore, the comparably fast k₂ does not control the time of reaching an equilibrium state under illumination, leading to close values for τ_rise and τ_decay, as observed in the I_PC traces. This would decrease the measured Q_I values, and also is consistent with the observed large discrepancy between I₀ and I_pkpk.

At elevated pH the increase in τ_decay with pH agrees with a slowing of the overall photocycle kinetics, caused by a reduction in k₂, as discussed before. The greater imbalance in τ_decay to τ_rise leads to an increase in induced charge in the BR PM film at equilibrium state. However, it should be considered that the BR film at a specific pH value consists of a mixture of states, determined by the balance in protonation state of some amino acids (according to their pKa values). At pH 6.2, already a significant percentage of BR molecules will feature a protonated PRC group, locally inverting the induced charge of the film. This would attenuate the light induced Q_S of the equilibrium state at pH values approaching 5.8, and would add to the decrease in Q_I associated by the changing photocycle kinetics. This also explains, why in some cases close to the pKa value of the PRC only an unstable signal can be observed, featuring double-peak behaviour–known also from organic MISM devices, in which two competing processes occur [53].

In contrast to the shape of I_PC, the shape of the wavelength dependent photocurrent does not change with pH within the full measurement range (see Figure 7D). The spectrum overlies well with the absorption spectrum of BR at elevated pH values. The appearance of the shoulder peak in the absorption spectrum at low pH values (around 630 nm) is not visible in the wavelength dependence photocurrent spectra. This confirms that blue shifted BR does not contribute to proton translocation, as reported previously [59], and thus has no photocurrent activity. The partly purple to blue transition of BR will therefore be another factor contributing to the observed reduction in I_PC observed at low pH values by deactivating some of the BR proteins from proton pumping.

While the change in polarity upon decreasing the pH value below 5.8 can be easily explained by the reversal of the proton uptake and release steps in the BR PM film, the observed photocurrent polarity is contradictory to that observed for TR MISM devices. A negatively doped film would be expected to induce a positive charge in the M_a layer by moving e⁻ to M_c, resulting in a negative light-ON peak (see Figure 1C, upper case). However, in the case of BR a positive peak is observed. As already mentioned, a second scenario can be used to explain the direction of charge movement between the electrodes (see Figure 1C, lower case): here, the released ions would stay in the protein film, close to the M_a electrode, drawing charge of opposite polarity to that obtained in the solution (and not the proteins). We can only speculate on the reason this might occur in BR PM films, which could include their greater density, or a lower ability of water molecules to move in and out of the film. The unified orientation of BR molecules in the PM membrane (in contrast to the “random” orientation in reconstituted samples) and the thereby unidirectional light-activated H⁺ translocation could be another reason.

In the next step, we aim to investigate factors affecting the photocurrent response of MISM devices, based on a novel microbial rhodopsin protein, SyHR [48]. Interestingly, SyHR was the first reported microbial rhodopsin protein which supposedly transports divalent ions, despite their large size. The device would, therefore, allow an investigation of the influence of valency and ion size on the photocurrent response, when comparing I_PC in the presence of Cl⁻ or SO₄ ^2- ions.

As can be seen in Figure 8A, a positive photocurrent peak upon illumination with green light can be observed in the presence of Cl⁻ ions in SyHR MISM devices. The characteristic change in shape with f_light can already be seen at comparably low light modulation frequencies with f_BW in the 10 Hz region (see Supplementary Figure S14), and thus about an order of magnitude slower than those obtained for TR. The magnitude of I_PC is similar to that of TR at high pH values, and increases 2fold when decreasing the pH in the solution (see Figure 8B– bottom trace) (in contrast to BR, where a strong decrease of I_pkpk is observed with decreasing pH). By analogy to BR, we see a strong increase in τ_decay with increasing pH (see Figure 8B– top trace), however, the extracted values are in average about 5x larger. This suggest comparably slow photocycle kinetics in SyHR compared to BR. At pH values below 7.0 SyHR shows similar rate constants compared to TR at RT. The very slow decay of the photocurrent peak especially at elevated pH values, results in large Q_I values (see Figure 8B– middle trace), about 10 to 20x those measured for TR, or for BR at low pH values. Only at pH8.0 does Q_I measured for BR exceed the values of SyHR, due to the considerably increased magnitude in BR’s PC response.

During the photocycle a Cl⁻ release from a binding site close to the retinal chromophore has been reported to occur, with a subsequent Cl⁻ uptake at a later intermediate of the photocycle [48] (illustrated in Figure 8D). This would result in a positive doping of the protein film, drawing e⁻ to the M_a electrode, as evident by the positive PC peak upon light illumination. Therefore, a similar model as used for TR can be used to explain the polarity of PC response. The same polarity (as expected by the conserved order in anion release/uptake [48]), but for a much weaker (about 2 orders of magnitude) photocurrent response can be observed when exchanging Cl⁻ for SO₄ ^2- ions in the I_b solution (see Figure 8C). It was previously reported that performance of SO₄ ^2- pumping by SyHR is comparably low [48], which would be consistent with the low responsivity of our devices. The strong decay in I_pkpk is accompanied by an about 10fold increase in τ_decay, suggesting very slow photocycle kinetics in the presence of SO₄ ^2- ions. As both ions have an influence on the absorption maximum of its deeply buried retinal chromophore, compared to that reported in salt free solution, an anion binding site close to the retinal chromophore has been suggested [48]. Thus, the observed slow kinetics of moving a large ion deeply into the SyHR moiety compared to a small ion seems consistent with this understanding.

When measuring the wavelength dependence of the photocurrent response (see Figure 8E) in the presence of Cl⁻ ions a strong peak can be observed with a maximum at 560 nm slightly red-shifted compared to the absorption spectrum (see also Supplementary Figure S15). The position and shape of the wavelength dependence does not change with pH in the solution.

In the case of SO₄ ²⁻ the wavelength dependent photocurrent spectrum features 2 peaks, one at 575 nm and the other at 370 nm. The peak at 575 nm can be expected, and agrees with the 20 nm red-shift in the absorption spectrum upon binding of SO₄ ^2- ions (see Supplementary Figure S15) [48]. It should be mentioned that a second peak around 400 nm has been reported for elevated pH values in the absorption spectra of SyHR in the presence of both Cl⁻ and SO₄ ²⁻ ions. This peak has been attributed to the deprotonation of the SB in SyHR, with pKa values of 10.6 in the presence of Cl⁻, which is lowered to 8.2 in the presence of SO₄ ^2- [48]. However, these values were measured for SyHR in DDM micelles. In the case of SyHR reconstituted into PC lipid membranes, the pKa of this transition seems to have dramatically decreased in the presence of SO₄ ²⁻ ions (see Supplementary Figure S15). Upon addition of Cl⁻ ions to the solution, the peak disappears, and the main peak resembles that measured in pure Cl⁻ solutions, showing that SyHR has a much larger affinity to Cl⁻ compared to SO₄ ²⁻ ions, and suggesting that the exchange to SO₄ ²⁻ facilitates the deprotonation of the SB even at moderate pH values.

Although no full characterisation of the PC behaviour at 370 nm could be made with our current setup, we could verify that the polarity of the PC peak remains positive upon blue light illumination. The observation of the comparably strong peak for blue light, inspired the wavelength dependent investigation of the light-induced pH changes of SyHR in E. coli cells, depending on halide ion type. Upon blue light irradiation a light-induced pH decrease was observed in a solution containing only SO₄ ²⁻ ions, which disappeared through the addition of the protonophore, CCCP. This could suggest an initial proton uptake during the photocycle of SyHR under blue light illumination, which would agree with the positive photocurrent response observed during the acquisition of the wavelength dependent photocurrent spectrum. In the presence of Cl⁻ ions, this activity was greatly suppressed. It should be noted that a similar H⁺ pumping activity upon blue light illumination (above 360 nm) was reported in the Cl⁻ pump HR by [61]. A more thorough investigation of the origin of the peak, ideally combining measurements on photocycle kinetics upon blue light illumination with FTIR data providing information of the protonation state of crucial amino acids, would be, however, necessary to confirm this preliminary data.

Furthermore, the relative high ratio of peak height at 370 nm in the wavelength dependent photocurrent spectrum, compared to that at 576 nm corresponding to the pumping of SO₄ ^2-, might be misleading. A strong change in τ values of the corresponding I_PC traces, can considerably falsify relative I_{peak, lock} values due to the use of the integrating lock-in amplifier in the measurement setup. The lock-in amplifier would generally generate larger values for square wave signals compared to transient ones (of similar magnitude). Simultaneously, the I_pkpk value of the square wave signal would be reduced compared to its lower frequency response. Therefore, a detailed study on I_PC under blue-light illumination should be included in a follow-up study.

Building on the first demonstration of a microbial rhodopsin MISM photodetector in 2015 [34], the observation of a stable photocurrent signal could be extended to three additional microbial rhodopsin proteins TR, BR and SyHR. In all cases, the performance of photodetection was greatly improved, compared to the prototype SrSRI MISM device, allowing detailed studies under a range of conditions. As expected, the wavelength dependent responsivity of the devices did well align with the corresponding absorption spectra of the protein films. Furthermore, the changes in the kinetics of the decay of the photocurrent peak (τ_decay) followed the changes in photocycle kinetics reported in previous studies. Although the bandwidth of the protein MISM devices are still low compared to MISM devices based on optimised organic semiconductors (despite the increase from the 10 Hz to kHz range upon changing from SyHR to BR), the order of magnitude of the device’s responsivity (reaching 5 μA/W and 2 μA/W, for BR and SyHR PD at their respective absorption maximum when calculated at full decay, respectively), the specific detectivity (approx. 1 × 10⁶ cm Hz^1/2/W and 3 × 10⁴ cm Hz^1/2/W, for BR and SyHR, respectively) as well as their f_BW approach those observed for low mobility organic semiconductors [62].

The extraction of protein film dependent rate constants from our devices is facilitated by the nearly constant electronic parameters of the I_b layer with pH and temperature, featuring comparably low time constants which barely influence the shape of I_PC. Nevertheless, the extraction of photocycle kinetics from I_PC data would require a combination with spectroscopic techniques, providing detailed information of the state of the protein film, and should be combined with an investigation of the effect of induced electric fields on photocycle lifetimes of the film in equilibrium as well as during device operation. Furthermore, although it could be shown that the polarity can be inverted when changing the chemical doping of a given protein film with e.g., pH, the direction also depends on the model which needs to be used for the interpretation of the results, which greatly depends on the film condition, as has been already shown for our prototype devices [34].

Using proteins instead of organic semiconductors provides us with the possibility of adding secondary functionalities to our MISM devices. As has been shown, if the proteins can considerably change their photocycle kinetics and/or the order of ion uptake or release, depending on external conditions other than wavelength of light, I_PC of the corresponding MISM device will also change. If a significant change in one of the parameters extracted from the I_PC curves can be obtained (as observed here for changes in pH or T, for τ_decay, I_pkpk and/or their combination i.e., Q_I), the devices could be calibrated for each condition. This calibration also benefits from the reported quasi-insensitivity to the exact content of the I_b layer in respect to e.g., ion or buffer concentration, once a few mM of ions are present in the solution. Once inserted into an I_b of e.g., unknown pH or T, the obtained I_PC or extracted parameters could provide information about the pH or T, respectively.

The study on SyHR, has also shown that the investigation of I_PC using the MISM device architecture of a novel protein, can provide suggestions on the influence of external parameters on the photocycle response and associated kinetics. Compared to common experiments used in the characterisation of photoactive biomolecules, the MISM experiment is simple and fast, and could thus be used to provide an initial screening tool of the biomolecules. Furthermore, when considering the information obtained upon increasing the temperature, it should be noted that those are only limited to the range of ensuring liquid buffer solutions. In contrast, if the proton pumping activity of rhodopsin proteins is measured in a living cell (e.g., in the membrane of E. coli), a considerably narrower temperature range can be measured (ensuring intact cells) [63, 64]. pH dependent experiments are also limited to relatively natural pH values, while no limitations to the pH values are given in MISM devices.

Especially, once the understanding on the effect of an electric field across the MISM device on the photocycle kinetics has increased and a methodology developed allowing the reliable extraction of relevant values of photocycle kinetics from the rates obtained from the I_PC trace, the use of screening will increase for gaining understanding of the proteins themselves. In future it might also be used, in combination with other methods, to gain an improved interpretation of the (electro)chemical potential [65] of the proteins in the presence and absence of light.