Development of polymeric blue prosthetic retina photoreceptors

Two conjugated polymers with absorption spectra like that of human short-wavelength photoreceptors, or blue cones, are characterised to act as essential components of a sub-retinal flexible prototype prothesis for people suffering from retinal disease affecting photoreceptor cells. Spectrophotometric and transient tests undertaken in unbiased photovoltaic mode in ambient conditions demonstrate that the addition of hole-blocking ZnO layer results in reliably forcing a favourable capacitive charging regime and acts to improve the photoresponse over ten times in one polymer and 45 times in another. We report that the addition of fullerene and non-fullerene acceptor molecules in bulk-heterojunction (BHJ)-active layers make an almost sevenfold measured improvement to extracellular photovoltage for devices operating in an electrolyte environment.


Introduction
Recent advances in biomedical engineering have led to a growing worldwide effort to develop a retinal prosthesis designed to restore vision to people suffering from retinal dystrophies such as age-related macular degeneration (AMD) and retinitis pigmentosa (RP) [1]. Many of the causes of these diseases are genetic, and treatments have been met with limited success, so prosthetic retinal implants are becoming a viable option [2,3].
Arguably the most prolific venture to date has been the Argus II developed by Palanker et al. for Second Sight, an epiretinally configured device with an array of 60 siliconbased electrodes [4]. Organic semiconducting materials have emerged as a viable alternative to silicon and in particular for retinal applications due to their potential biocompatibility, very high absorption coefficients, and excellent opto-electronic properties of carbon-conjugated backbones to create arrays of miniature photovoltaic pixels [5,6]. A widely quoted example is the POLYRETINA where a P3HT:PCBM bulk heterojunction was used as the active material in 967 electrodes for the first iteration [7]. More recently it was improved to over 10,000 pixels with the goal of further widening the field vision to 46.3°. Moving from monochromatic artificial vision to a full-colour prototype demonstrated the use of four conjugated small molecules to potentially replace each photoreceptor [8]. This sub-retinal placement would capitalise on the processing capabilities of the retinal network and rely on neural plasticity ability of the brain to recode the colour information.
Characterisation with these devices both in vitro and in vivo relies on threshold values for operation: for stimulation current density as well as extracellular and intracellular potentials, particularly in the bipolar cells targeted by a subretinally implanted device. In general, studies show that the membrane voltage required to simulate ON-bipolar cells is a depolarisation from the resting potential of around − 60 to − 40 mV and current level from 1 µA for 25 ms [9][10][11][12][13][14]. A specific study showed that the passive response to capacitive stimulation of neurons from a giant snail was ± 10 nA with a transmembrane voltage of − 95 mV [13]. Another investigation on rabbit retinas showed a stimulation current density of 2.1 mA cm −2 inducing a transmembrane voltage change of 15 mV [15]. Differentiation between electrical photoresponse and intracellular response was made explicit by testing an organic device eliciting a current of 400 μA and transmembrane voltage of − 35 mV from applied transient voltage of 330 mV generated from illumination of power density 60 mW cm −2 [16]. Stimulation of peripheral nerves in vivo was achieved with transferred charge density of 1-60 μC cm −2 which increased to 1 mC cm −2 with larger electrodes. Clearly, charge is an important component of neural stimulation, but particularly in a capacitive regime, it is a value ultimately driven by an electric field due to membrane voltage [16,17]. Therefore, the natural effect of a graded voltage change delivered to a bipolar cell has a significant impact on stimulation.
The maximum permissible exposure (MPE) limits incident light to a power density ten times less than that which has a 50% chance of incurring a lesion [18]. The MPE for blue light, of the inherently high-energy short-wavelength part of the visible spectrum, is low. For example, a contact area of 13 mm 2 illuminated with 405 nm light is limited to 2.45 μW mm −2 . This would seem a good motivation to explore operational parameters in the near infrared which offers much higher safety ceiling [2,4,19].
However, an upper limit to the amount of available light given by the sunlight AM1.5 illumination spectrum overlaps with the absorption bands of each photoreceptor illustrated in Fig. 1a [14,20]. The hypothetical amount of light power available to induce photo-action of each type of cone cell is the integral of the irradiance multiplied by the response function within its absorption band as shown in Eq. 1 where E( ) represents the spectral power density and R( ) is the cone response curve. Fig. 1 a Cone-level relative photoreceptor sensitivities from Wald [18], for S-cones (blue), M-cones (green), and L-cones (red). Rods response is not shown. Sunlight ASTM G173-03 reference spectra are derived from SMARTS v. 2.9.2 (AM 1.5). It is noted that the peak absorbance for blue cones occurs at 420 nm. b Schematic of hypothetical prototype device for context. In a realistic setting, the observer would see scattered light from objects rather than direct sunlight; thus, light power density reaching the retina will be a small fraction of AM1.5 spectrum intensity. For blue cones with a FWHM absorption band of about 139 nm, available power density can be estimated to be up to 7 mW cm −2 by direct sunlight and possibly a factor of ten lower for scattered light. To put this value in context, the ANSI Energy Standard for Buildings recommends a range of Lighting Design Criteria including lighting power densities [21]; some typical indoor settings for full spectrum white light are shown in Table 1.
Another important consideration of configuring and testing organic semiconductor devices to work in unbiased photovoltaic mode is to study the response on a time scale that is comparable to that of biological processes. The natural response time for human photoreceptors is about 10 ms with a latency of 80-100 ms [4,[22][23][24][25]. Transient and photospectra tests used in this work relied on pulsed illumination, and the combination of period and pulse length was chosen to be within a window from tens to hundreds of milliseconds.
Here, we focus on two conjugated polymers with absorption peaks in the blue region of the visible spectrum as part of a larger study into developing materials for a full-colour organic device. We show not only that they can elicit impressive extracellular voltage, but that the addition of fullerene and non-fullerene acceptor molecules in a bulk heterojunction can improve capacitive voltage delivery for devices operating in a bio-compatible electrolyte environment.

Materials
Organic semiconductor materials chosen as active layers for this investigation were selected with absorption spectra nearly matching the photosensitivity of human shortwavelength photoreceptors, with sufficient material solubility, utilising solution-based fabrication techniques. Here, we report on two polymers, poly(triarylamine) PTAA and indenofluorene-phenanthrene copolymer PIFPA intended to serve as blue photoreceptor replacements, and later, report on the addition of acceptor molecules as bulk heterojunctions. Polymer chemical structures are detailed in Fig. SI 1. Importantly, we employ solution processible ZnO as a hole-blocking metal oxide interlayer at the current collecting electrode to reliably elicit capacitive charge transfer and improve device performance [26]. By reducing the difference in work functions between the electrode with ZnO layer and LUMO level of the active materials, and by utilising the relatively deep valence band of the ZnO to block holes, electron collection can be maximised. As a result, two outcomes are achieved: (i) significantly higher photo-induced current is obtained with ZnO interlayer and (ii) capacitive photocurrent behaviour is demonstrated, thereby reducing the likelihood of current-induced electrochemical reactions with undesirable Faradaic photocurrent behaviour in deviceneuron interfaces [27,28].
For an additional boost to photocurrent and photovoltage, a bulk-heterojunction (BHJ) configuration was used, with conjugated polymer materials as donors and acceptor molecules including PCBM70 and two non-fullerenes ITIC and Y6 [29,30]. PCBM70 is widely known to work with many thiophene donors, and ITIC and Y6 are becoming increasingly more popular alternatives to fullerenes. The donor:acceptor ratio of all was 1:1 apart from one set of samples that were PCBM70 at 1:2 with no attempt to optimise either these ratios or the underlying micro/nanostructure beyond the feasibility study to extract maximum photovoltage offered here. All materials were dissolved in dichlorobenzene.

Experimental setup
The device structure was designed to test the photoresponse of 20-200-nm-thick active layers spin coated onto ITO/glass and ZnO/ITO/glass substrates. To simulate the biological environment, samples were immersed in electrolytic phosphate-buffered saline solution (PBS) for transient and spectrophotometric measurement. For transient measurements, a 1.5-mm-diameter blunt-end Ag/AgCl counter-electrode was immersed about 100 µm away from the semiconductor/ These power density values are comfortably below the maximum retinal exposure yet give a tangible figure to safe levels of ambient incident white light we experience daily electrolyte interface in a reservoir yielding a contact area of 13 mm 2 , as schematically shown in Fig. 1d. Ag/AgCl is widely used in electrophysiology and is known to prevent electrode polarisation in electrolyte. A blue 405 nm LED was placed 2 cm away to ensure a relatively homogeneous illumination across the sample active area. Data were recorded in photovoltaic mode (open circuit) for unamplified photovoltage measurements and in a short-circuit configuration for photocurrent measurements using a Thorlabs AMP120 transampedance amplifier. Absorption data were acquired using a Cary 5000 UV-Visible Absorption Spectrometer in optical density mode. Spectrophotometric measurements in the visible range were acquired using Bentham scanning monochromator as described previously [8] where the power ouput varied from 30 μW at 300 nm to 550 μW at 800 nm ( Fig.  SI 4), and the counter-electrode was set to a distance up to 8 mm above the surface of the sample.

Pristine and heterojunction active layers
A successful artificial retina will safely absorb light and be electronically active at roughly the same wavelengths as the photoreceptors they aim to replace. Absorption data in Fig. 2a, b are shown for pristine polymer and polymer/acceptor bulk-heterojunction samples. The pristine polymers (grey line) show a strong isolated peak of about 400 nm which is within the range of absorption for blue photoreceptors and close to their peak at 420 nm (Fig. SI 2). Absorption for BHJs (blue and red lines) includes a thin interlayer of ZnO which presents as a peak at 300 nm and additional absorption bands due to acceptors: PCBM70 contributes to the absorption between 425 and 650 nm, and ITIC has an absorption band in the deep-red part of the spectrum, as expected (Fig. SI 3).
Spectrophotometric responsivity data for the pristine polymer samples are shown in Fig. 2c, d, and for the BHJs in e, f. The spectrophotometric power profile of monochromator output is shown in Fig. SI 4. Photocurrent peaks occur at almost the same wavelengths as the absorption data for the same samples with variations noted in Table 2. We also note that the addition of an ZnO electron-collecting/holeblocking interlayer improves the photocurrent response significantly as reported previously although it might result in a slight shift to shorter wavelengths especially for lower photocurrent values [8]. The BHJ samples in Fig. 2e, f show broad-spectrum photocurrent bands for PCBM70 and a deep-red band for ITIC as expected, but due to the nature of the lock-in amplification, no information is given on the polarity or time-dependence, both of which are demonstrated in the following transient experiments.
Transient data with pulsed LED excitation reveal polarity and time-dependent behaviour of each device. A Faradaic response from ITO/polymer samples (Fig. 3a) is indicated by the positive response at the ITO electrode due to hole collection at that interface. Under illumination, electrons accumulate at the polymer/electrolyte surface leading to a reduction at the counter-electrode. Faradaic charges are known to degrade both tissue and polymer [27]. Samples with a ZnO interlayer show a capacitive response over a range of pulse durations in Fig. 3b, c as demonstrated by an initial spike of negative electron collection at the ITO electrode develops and the photocapacitor charges up. As illumination ends, a counter spike forms as the electrode is discharged indicating a capacitive regime. Given the activation time and latency period of human photoreceptor cells, a 20 ms pulse width and 100 ms period were chosen as the time-dependent parameters as they allow sufficient time to ascertain if the device operates in a Faradaic or capacitive regime. An example of typical photocurrent (blue) and photovoltage (red) traces is shown in Fig. 3d, e for ITO/ZnO/polymer samples under 40 μW mm −2 incident blue light of 405 nm. Similar capacitive transients from BHJs were gathered, and the negative peaks were used to compare trends (Fig. SI5).

Bulk heterojunctions
Bulk heterojunctions were tested under ramped incident blue light from about 0.4 to about 40 μW mm −2 . Both photocurrent and photovoltage transient peak values were recorded. Amplitude of photovoltage peaks vs incident power density showed a steep increase at incident power destinies below 10 µW mm −2 , followed by slower photovoltage increase at higher incident power densities (Fig. 4) whereas the photocurrent peaks showed a linear relationship for photocurrent in both polymers (Fig. SI 6).
Considering the importance of extracellular photovoltage, a comparison between active layers of the pristine polymers and the BHJs shows that the primary trend is that the BHJ with PCBM70 elicited significantly and consistently higher photovoltage than pristine polymer samples as shown in Fig. 4. Non-fullerene acceptors provided a small but consistent increase in photovoltage. Whilst it is true that the full range of incident power density of 405 nm light reported here exceeds the upper limit of available blue sunlight by some margin, the trend of acceptor molecules improving photovoltage remains consistent even at reasonable light levels. At 8 μW mm −2 , PIFPA:PCBM70 outperformed the pristine polymer alternative by 15%. At 40 μW mm −2 , the increase was 87%. The equivalent PTAA comparison was an impressive increase by 7 times at both benchmark power densities.

Discussion
Blue photoreceptors, s-cones, absorb light in the region of 380-450 nm, and the aim of this investigation was to characterise polymers to use as their potential replacements for photoreceptors in a retinal prosthetic device. Absorption and spectrophotometric measurements show that pristine PTAA and PIFPA polymers devices were suitable candidates; peaks were detected in both tests at around 400 nm. BHJ samples with similar thickness were compared, and for devices immersed in PBS electrolyte, photo-induced current and voltage peaks were obtained at wavelengths corresponding to polymer absorption bands and acceptor molecules. By demonstrating that our devices can deliver an extracellular voltage almost 100 mV with 405 nm excitation, we have indicated that it is possible to improve the photovoltage from  bar (a, b) indicates spectrally active region for cone cells. a Absorption data (optical density) for thin films of pristine PTAA (black) and BHJs with ZnO interlayer, b Absorption data (optical density) for thin films of pristine PIFPA (black) and BHJs with ZnO interlayer. Acceptor layers absorption curves are shown in Fig. SI 3. c-f Are spectrophotometric data: c PIFPA and d PTAA with and without ZnO interlayer showing a large responsivity improvement with ZnO as compared to ITO only devices, e PIFPA and f PTAA responsivity for polymers with a BHJ of each acceptor molecule showing spectral response from the blends polymeric devices with the addition of acceptor molecules. However, further work is required in two main areas: to investigate if the extracellular applied voltage is enough for neural stimulation and to reduce the impact of the acceptors on spectral response at longer wavelengths.
The addition of acceptor molecules was intended to improve both photocurrent and photovoltage in devices interfaced with electrolytic environment. Results show that the photovoltage increase in BHJ vs pristine polymer samples is substantial. Brightly illuminated samples at 40 μW mm −2 showed an increase of − 46 to − 86 mV in PTAA and almost seven times from − 14 to − 97 mV in PIFPA. If extracellular voltage is indeed the driver for neuronal stimulation, then the BHJs approach could be a valuable tool to improve the performance of conjugated molecules-based devices for retinal prosthesis applications.
Safe neuronal stimulation needs to be capacitive to eliminate generation of dangerous free radicals and potential destructive oxidation [17,27], as was confirmed by transient photocurrent tests on the order of the timescale of photoreceptor activation and relaxation. Whilst it has been demonstrated that capacitive and Faradaic regimes may occur simultaneously [27], it is also important to take into account the timescales of these behaviours. For example, a slow capacitive response may present as Faradaic if the measurement period is too short and a capacitive response must be confirmed over longer time scales as shown in Fig. 3b.
Finally, we note that photovoltage and photocurrent tests were carried out with a 100 μm gap between the surface of the device and the counter-electrode, exceeding the length of a rod cell, and possibly the maximum distance a device would ever be from a bipolar cell layer if it used as a subretinal implant. Shorter device-to-cell distance would result in a higher voltage induced by the implant due shorter field screening distance in the electrolyte.

Conclusion
We have shown that conjugated polymers PTAA and PIFPA are suitable semiconductor materials to replace blue photoreceptors in an artificial retina due to their favourable absorption characteristics and high photovoltage output in devices in electrolyte environment. Transient measurements show that devices with an electron-transport ZnO interlayers demonstrate capacitive response, and bulk heterojunctions with PCBM70 acceptor show that photovoltage is substantially improved, up to 7 times, reaching 100 mV, as compared to pristine polymer devices.
Future work includes an investigation for the other longer wavelength conjugated molecules materials to mimic human photoreceptors response to evaluate photo-induced voltage levels in green and red spectral regions. Patch clamp work needs to be carried out to determine the relationship between transmembrane and transient voltages. In addition, micron-sized pixelation could have a significant impact on photophysics of these devices. It would be important to test the photovoltage with a smaller diameter counter-electrode against various pixel sizes in the range of 20-50 μm or evaluate device response with smaller electrolytic contact area, imitating small pixel size.