Chromatic discrimination in protocells toward proto vision and intelligence

Introduction

Proteinoids, also referred to as thermal proteins, are a distinct class of synthetic amino acid copolymers originally described by Fox and coworkers in the 1950s (Fox and Harada, 1958; Mougkogiannis and Adamatzky, 2026a; Mougkogiannis and Adamatzky, 2025a; Mougkogiannis and Adamatzky, 2026b; Mougkogiannis and Adamatzky, 2025b). Historically, these materials have been pivotal in prebiotic chemistry models regarding the origin of life and the formation of protocells (Fox, 2013; Mougkogiannis and Adamatzky, 2026c; Mougkogi et al., 2026; Mougkogiannis and Adamatzky, 2026d). Biopolymers are synthesized when anhydrous amino acid mixtures undergo thermal polycondensation at elevated temperatures, typically between 170 and 180 °C. This process yields random heteropolypeptides linked primarily by peptide bonds, although other linkages (e.g., imide) may occur (Rohlfing, 1976; Mougkogiannis and Adamatzky, 2026e). When cooled in an aqueous environment, these polypeptides self-assemble into spherical microsphere structures. Proteinoids exhibit an intrinsic amphiphilic nature derived from the incorporation of both hydrophobic and hydrophilic amino acids. Consequently, they form hollow, vesicular structures that bear a morphological resemblance to modern cell membranes (Bahn and Fox, 1981). Beyond their prebiotic significance, the biocompatibility, biodegradability, and structural versatility of proteinoids make them appealing candidates for biomedical applications, including drug delivery, tissue engineering, biosensing, and biomimetic systems (Khemaissa et al., 2022).

While conventional proteinoids possess advantageous intrinsic properties, they lack the dynamic responsiveness characteristic of modern smart biomaterials. The ability to modulate material properties via non-invasive external stimuli is a prerequisite for advanced biomedical and biotechnological platforms (Stuart et al., 2010). Among various stimuli, light offers distinct advantages as an external trigger, providing precise spatiotemporal control, remote activation capabilities, and tunable parameters such as wavelength and intensity (Katz and Burdick, 2010). Furthermore, light irradiation is minimally invasive, making it ideal for applications requiring on-demand modulation of material behavior. Integrating photoresponsive moieties into proteinoid structures offers a promising strategy to synergize their natural self-assembly propensity with the controlled actuation required for advanced functional devices (Gautier et al., 2014; Lechner et al., 2021; Riggsbee and Deiters, 2010).

Azobenzene derivatives are prototypical photoswitchable chromophores capable of reversible isomerization between trans and cis isomers upon irradiation with ultraviolet (UV) or visible light (Bandara and Burdette, 2012). This photoisomerization process induces significant alterations in molecular geometry, dipole moment, and absorption spectrum. On a macroscopic scale, these molecular changes can translate to tunable material properties, including solubility, surface wettability, mechanical modulus, and supramolecular organization (Natansohn and Rochon, 2002). While azobenzene groups are widely exploited in polymer systems, such as liquid crystalline polymers and supramolecular assemblies (Yu, 2014), the functionalization of thermal proteinoids with these chromophores remains underexplored. This represents a significant gap in the literature, as the intrinsic self-organizing traits of proteinoids could arguably enhance the dynamic switching abilities of azobenzene units.

The synthesis of azobenzene-functionalized proteinoids presents unique opportunities and challenges. Unlike conventional polymer systems, where azobenzene units are typically grafted post-polymerization, here we propose a direct copolymerization strategy where azobenzene-containing amino acids are integrated into the proteinoid backbone during thermal polycondensation. This approach ensures a homogeneous distribution of chromophores within the proteinoid matrix and circumvents complicated multi-step synthesis protocols. However, the high temperatures required for proteinoid synthesis (approx. 180 °C) raise concerns regarding the thermal stability of the azobenzene moiety, necessitating careful optimization to avoid chromophore degradation. Furthermore, the incorporation of rigid, planar azobenzene units is expected to influence the self-assembly landscape through π-π stacking interactions and steric effects, potentially altering the equilibrium between hydrophobic and hydrophilic domains (Fong et al., 2016; Funtan, 2022; Yao et al., 2014). To establish a direct mechanistic link between photoisomerization and the observed electrophysiological responses, reversible UV–visible light switching experiments and ion transport assays are incorporated alongside the electrochemical characterisation.

A conceptual overview of the azobenzene-functionalized proteinoid proto-vision system is presented in Figure 1. The system operates across three hierarchical levels. Phenylalanine-4′-azobenzene hydrochloride is mixed with L-glutamic acid and L-phenylalanine. This occurs by thermal polycondensation at 170–180 °C. The process incorporates photoresponsive azobenzene chromophores into the proteinoid backbone, using 6.67 mol% of the total monomers. At the supramolecular level, the resulting amphiphilic chains self-assemble into vesicular microspheres in which the azobenzene moieties are embedded within the membrane wall. UV light (λ = 365 nm) causes trans→cis photoisomerization in the chromophores. This contraction disrupts the membrane and ion transport. In contrast, visible light (λ > 420 nm) reverses this change. It restores the trans configuration and improves membrane permeability. Light-driven changes convert optical input into measurable voltage signals. These signals range from − 130 to + 150 mV across different electrode pairs. This process allows for chemical chromatic discrimination, similar to how biological photoreceptors respond to different wavelengths.

Materials and methods

Synthesis of proteinoids

Pure Glu:Phe proteinoids

Pure glutamic acid:phenylalanine (Glu:Phe) proteinoids were synthesized via thermal polycondensation following established protocols with minor modifications (Fox and Harada, 1958; Rohlfing, 1976). Briefly, L-glutamic acid and L-phenylalanine were mixed in a 2:1 M ratio (total mass: 10 g) and ground together using a mortar and pestle to ensure homogeneous distribution of monomers. The amino acid mixture was transferred to a clean, dry round-bottom flask and heated in a temperature-controlled oil bath under anhydrous conditions. The reaction temperature was gradually increased from room temperature to 170 °C over 30 min, then maintained at 170–180 °C for 3 h to allow complete thermal polycondensation. During heating, the mixture melted to form a viscous amber-colored liquid, with water vapor evolved as a condensation byproduct. After 3 h, the heat was removed and the reaction vessel was allowed to cool to room temperature over approximately 1 h.

The cooled proteinoid product was dissolved in deionized water (100 mL) under gentle stirring at 60 °C for 2 h to facilitate complete dissolution. Insoluble charred material or unreacted residues were removed by centrifugation at 3,000 rpm for 10 min. The supernatant containing dissolved proteinoid chains was collected and allowed to cool slowly to room temperature over 12–16 h, during which spontaneous self-assembly into microsphere structures occurred. The resulting proteinoid microsphere suspension was dialyzed against deionized water using dialysis tubing (molecular weight cut-off: 3.5 kDa) for 48 h with water changes every 12 h to remove low-molecular-weight oligomers and residual monomers. The purified proteinoid microsphere suspension was stored at 4 °C for subsequent characterization.

The molecular weight of the resulting proteinoids was not determined directly, as gel permeation chromatography (GPC) or mass spectrometry were not available in the current study. Based on the thermal polycondensation mechanism and the 3.5 kDa molecular weight cut-off (MWCO) dialysis membrane used for purification, the retained proteinoid fraction comprises chains with M r > 3.5 kDa. This is consistent with literature reports for analogous thermally synthesised Glu:Phe proteinoids, where number-average molecular weights of approximately 3–10 kDa have been reported following similar purification protocols. (Fox and Harada, 1958; Rohlfing, 1976) The broad molecular weight distribution inherent to thermal polycondensation—a consequence of the random, stepwise chain growth mechanism—means that the material represents a heterogeneous mixture of oligopeptide chains rather than a monodisperse polymer. Determination of the precise molecular weight distribution by GPC or matrix-assisted laser desorption/ionisation time-of-flight mass spectrometry (MALDI-TOF MS) is planned for future work.

Azobenzene-functionalized proteinoids (PAZO)

Azobenzene-functionalized proteinoids were synthesized using an identical thermal polycondensation protocol with the addition of phenylalanine-4′-azobenzene hydrochloride as a comonomer. L-Glutamic acid, L-phenylalanine, and phenylalanine-4′-azobenzene hydrochloride were mixed in a molar ratio of 2:0.8:0.2 (Glu:Phe:Phe-Azo-HCl, total mass: 10 g), corresponding to 6.67 mol% azobenzene incorporation relative to total monomers, or equivalently 20 mol% relative to total aromatic amino acids (phenylalanine + phenylalanine-4′-azobenzene hydrochloride). This ratio was optimized to balance photoresponsive functionality with maintenance of proteinoid self-assembly properties. The monomer mixture was thoroughly ground to ensure homogeneous chromophore distribution, then subjected to thermal polycondensation at 170–180 °C for 3 h under identical conditions to pure proteinoid synthesis. Special care was taken to monitor reaction temperature to avoid exceeding 185 °C, as preliminary experiments indicated that higher temperatures led to azobenzene degradation evidenced by loss of characteristic UV-visible absorption bands.

Following thermal synthesis, the azobenzene-functionalized proteinoid product was dissolved in deionized water (100 mL) at 60 °C, centrifuged to remove insoluble material, cooled to ambient temperature for microsphere self-assembly, and dialyzed against deionized water for 48 h to remove unreacted monomers and low-molecular-weight fragments. The resulting PAZO microsphere suspension exhibited a characteristic orange-brown coloration due to azobenzene chromophores, in contrast to the pale yellow color of pure Glu:Phe proteinoids. PAZO microsphere suspensions were stored protected from light at 4 °C to minimize unwanted photoisomerization prior to characterization.

UV-visible spectroscopy

UV-visible absorption spectra were recorded using an Ossila UV-Vis spectrophotometer. Proteinoid microsphere suspensions were diluted to an optical density (OD) of approximately 0.3–0.8 at λ max to ensure measurements within the linear absorbance range and minimize scattering artifacts. Samples were transferred to quartz cuvettes and spectra were acquired from 300 to 800 nm with 1 nm resolution and a scan rate of 600 nm/min. Baseline correction was performed by subtracting the spectrum of deionized water measured under identical conditions.

For differential absorption analysis, the pure Glu:Phe proteinoid spectrum was subtracted from the PAZO proteinoid spectrum after normalization to equal protein concentration to isolate the spectral contribution of azobenzene chromophores. Tauc plot analysis for optical band gap determination was performed by plotting ( α h ν ) 2 versus photon energy ( h ν ), where the absorption coefficient α was calculated from absorbance A and sample path length l according to α = 2.303 A / l. The optical band gap E g was extracted by linear extrapolation of the absorption edge to zero absorbance.

Voltage recording protocol

Electrophysiological measurements were performed using a PicoLog ADC-24 high-resolution multichannel data acquisition system (Pico Technology Ltd., United Kingdom; 24-bit resolution, input voltage range ±2.5 V, effective sampling rate 1 Hz achieved by averaging up to 600 measurements per second to minimise noise). Differential voltage signals were recorded using an array of platinum/iridium (Pt/Ir) microneedle electrodes (16 channels, tip diameter <50 μm, inter-electrode separation: 10 mm) positioned within the proteinoid sample chamber. The electrode array was connected to the ADC-24 via shielded leads to minimise electromagnetic interference. All recordings were conducted at ambient laboratory temperature (22 ± 1 °C).

All electrophysiological experiments were performed in triplicate (n = 3 independent proteinoid synthesis batches per condition). For each batch, three technical replicates were recorded per wavelength, giving nine voltage recordings per condition. Reported mean potentials are the grand means across all nine recordings; reported standard deviations represent between-replicate variability. Batch-to-batch reproducibility was assessed by computing the coefficient of variation (CV%) for the primary photoresponse metric (mean voltage during photostimulation phase) across the three independent synthesis batches. CV% values of < 15% were accepted as demonstrating adequate reproducibility; conditions exceeding this threshold are flagged in the relevant tables. Statistical comparisons between PAZO and pure proteinoid responses were performed using two-sided Welch’s t-tests with Bonferroni correction for multiple comparisons across six wavelengths ( α corrected = 0.05 / 6 = 0.0083 ). Effect sizes were quantified using Cohen’s d; values ≥ 0.8 were classified as large effects. All statistical analyses were performed in Python 3 using SciPy (v1.11) and Pingouin (v0.5).

The light modulation of the proteinoids’ spiking frequency was investigated using a PHOTONIC PL-2000 light source (Ryf AG, Switzerland) (Figure 2). Visible light wavelengths were selected by employing optical colour filters positioned 5 cm in front of the proteinoid samples. The filters included: blue (λ peak = 450 nm), daylight (broadband transmission 400–700 nm), gray (neutral density filter, 50% transmission), green (λ peak = 520 nm), red (λ peak = 650 nm), and yellow (λ peak = 580 nm). The focused light beam had a diameter of about 5 mm in the 10 mm wide electrode gap. This ensured full coverage and even lighting of the proteinoids inside the gap. The light path length through the proteinoid sample was approximately 10 mm, as determined by the beam diameter and electrode spacing geometry. All experiments were performed under ambient laboratory lighting conditions. Photostimulation was delivered as discrete light pulses at 10 Hz frequency (pulse duration: 50 ms, inter-pulse interval: 50 ms, 50% duty cycle) for each wavelength condition. The recording protocol consisted of the following phases for each wavelength condition:

• Baseline recording (300 s): Initial voltage recording in darkness to establish baseline spontaneous electrical activity.
• Photostimulation (600 s): Pulsed illumination (10 Hz) at the specified wavelength while recording voltage responses, delivering approximately 6,000 light pulses per trial.
• Recovery (300 s): Return to darkness to observe relaxation dynamics.
• Inter-trial interval (180 s): Rest period in darkness before next wavelength presentation to allow complete chromophore relaxation.

Signal processing and binary logic gate analysis

Electrophysiological data were recorded as multi-channel CSV files. Then, they were processed offline with a custom Python analysis pipeline. All processing was performed on the fifth available differential channel (Channel E, Diff. 9–10), selected consistently across all experimental conditions. Raw voltage traces were downsampled by 30 to cut down on computation. This kept the signal shape intact. Then, they were smoothed using a centered rolling mean with a window of 25 samples. This helped reduce high-frequency noise while still capturing slower electrophysiological changes in the 0.01–1 Hz range.

To compare across conditions, we normalized smoothed traces to the [0, 1] range. We did this by subtracting the minimum value of each trace and then dividing by the range of that trace. To convert continuous voltage traces into discrete binary activation states, a threshold-based classification was applied. The activation threshold was defined as Equation 1:where μ baseline and σ baseline are the mean and standard deviation of the first 15% of each recording. This period shows the pre-stimulation baseline. Z = 2.0 is the threshold multiplier for the 95% confidence interval under Gaussian noise statistics. A trace was classified as activated (binary output = 1) if its peak voltage exceeded this threshold; otherwise it remained inactive (output = 0).

V threshold = μ baseline + Z ⋅ σ baseline (1)

Boolean logic gate analysis was performed by mapping four experimental conditions onto a two-input truth table framework. Input A shows if azobenzene chromophores are present (A = 1 for PAZO, A = 0 for pure proteinoid). Input B indicates the illumination wavelength (B = 1 for red light, B = 0 for gray light). This creates four states (Equations 2–5):

00 = PURE - Gray (2)

01 = PURE - Red (3)

10 = PAZO - Gray (4)

11 = PAZO - Red (5)

The observed binary output vector was compared to the truth tables of four Boolean gates: AND, OR, XOR, and NAND. We calculated an agreement score. This score is the fraction of input states where the observed outputs match the canonical ones (Equation 6).where N = 4 is the number of truth table rows, O i is the observed output for state i, G i is the canonical gate output, and δ is the Kronecker delta function. The gate yielding the highest agreement score was identified as the best-matching Boolean logic operation. All analysis was implemented in Python three using NumPy, pandas, and Matplotlib libraries.

S gate = 1 N ∑ δ O i , G i (6)

Results

UV-Vis spectroscopic characterization

Here, we report the synthesis and characterization of azobenzene-functionalized proteinoids and demonstrate their wavelength-dependent electrophysiological responses, stomatocyte-like morphology, and Boolean logic gate behavior (Meeuwissen et al., 2014; Wilson et al., 2012). Table 1 compares azobenzene-functionalized proteinoids with known biological and synthetic chromophore systems. The current materials blend the strength of synthetic polymers with the self-assembling structure of amphiphilic peptides. This combination overcomes the limits of both natural chromoproteins and traditional azo-polymers.

TABLE 1. Comparison of azobenzene-functionalized proteinoids with common biological and synthetic chromophore systems.

The incorporation of phenylalanine-4′-azobenzene hydrochloride into the proteinoid matrix was confirmed by UV-visible spectroscopic analysis. Figure 3 presents a comparison of the absorption spectra for pure Glu:Phe proteinoids and azobenzene-functionalized proteinoids from 300 to 800 nm. The pure proteinoid system exhibits negligible absorption in the visible range, characterized by a relatively flat baseline typical of aromatic amino acids. Conversely, azobenzene-functionalized proteinoids display a prominent, broad absorption band centered between 500 and 550 nm. This spectral feature indicates a significant bathochromic (red) shift from the characteristic absorption peak of free azobenzene chromophores in solution, which typically exhibit a sharp π − π * transition band around 340 nm for the trans-azobenzene configuration. The strong red-shift and broadening of the absorption band in the functionalized proteinoids suggest that supramolecular J-aggregates are present. In these aggregates, azobenzene chromophores adopt a head-to-tail orientation, where their transition dipole moments align parallel to each other. J-type aggregation often occurs in tight or concentrated chromophore systems, resulting from excitonic coupling between nearby azobenzene units in the condensed proteinoid matrix. The strong and persistent absorption band in the visible range demonstrates that the azobenzene chromophores withstand the harsh thermal polycondensation (170–180 °C) conditions. They retain their chemical integrity within the proteinoid microsphere structure, thereby confirming the photoresponsive potential of these hybrid materials.

To further elucidate the influence of azobenzene chromophores on the spectral properties and quantify the functionalization effects, comprehensive analyses were conducted, as presented in Figure 4. The difference spectrum (Figure 4a) isolates the unique spectral signature of the incorporated chromophores by subtracting the pure proteinoid absorption from the azobenzene-functionalized proteinoid spectrum. The difference spectrum reveals two distinct absorption features. First, a sharp peak is observed around 360 nm, which corresponds to the π − π * transition of the azobenzene group. Second, a broader, red-shifted band is visible, spanning from 450 to 600 nm and peaking near 520 nm. The presence of these two absorption bands suggests a heterogeneous chromophore environment inside the proteinoid microspheres. In this matrix, monomeric or weakly coupled azobenzene units contribute to the 360 nm absorption, coexisting with densely packed J-aggregated domains, which manifest as the 520 nm band. The stacked spectral comparison (Figure 4b) visualizes this transformation, highlighting how azobenzene introduces visible-region absorption while preserving the UV absorption traits of the original proteinoid matrix. The spectral evolution indicates that the azobenzene chromophores augment the system’s optical properties without disrupting the electronic structure of the proteinoid backbone. This suggests that the thermal polycondensation process effectively integrates the photoresponsive units into the polymer network without major degradation.

The energetic consequences of azobenzene incorporation were further investigated using Tauc plot analysis and simulated spectral integration to understand how chromophore functionalization changes the electronic band structure (Figures 4c,d). Tauc plots display ( α h ν ) 2 against photon energy ( h ν ), where α is the absorption coefficient. The optical band gap ( E g ) was determined by extending the linear part of the absorption edge to zero absorbance. The pure Glu:Phe proteinoid system exhibits an optical band gap of approximately 3.5 eV, consistent with the electronic transitions of aromatic amino acid side chains and peptide bonds in the UV region. Upon azobenzene functionalization, the absorption edge shifts noticeably to lower photon energies. A new absorption feature appears around 2.5 eV, which corresponds to the 520 nm absorption band. This band gap narrowing reflects the introduction of lower-energy electronic transitions associated with the extended π-conjugated system of the azobenzene chromophores and their supramolecular aggregates. The simulated integration landscape (Figure 4d) illustrates the correlation between spectral evolution and azobenzene concentration. As the content increases, the absorption bands of the chromophore appear and intensify. This heatmap visualizes the shift from the pure proteinoid spectrum (0% azobenzene) to the fully functionalized system, highlighting the concentration-dependent evolution of the UV absorption peak near 360 nm and the visible absorption band around 520 nm. Collectively, these optical analyses confirm that the azobenzene chromophores are not merely physically entrapped within the proteinoid microspheres but are chemically incorporated into the polymer matrix, where they adopt both isolated and aggregated configurations that collectively contribute to the observed photoresponsive behavior. The preservation of azobenzene’s unique electronic transitions after thermal synthesis demonstrates stability within the supramolecular organization of the proteinoid environment. This is critical for light-induced isomerization and may allow for photomodulation of proteinoid properties.

To contextualize the optical properties of the azobenzene-functionalized proteinoids, it is instructive to compare their measured optical band gap (E g ≈ 2.5–2.6 eV) with standard biological materials and synthetic polymers. Natural biological macromolecules typically exhibit wide band gaps, effectively insulating them from visible light absorption unless specific prosthetic groups are present. For instance, pure DNA thin films possess a wide optical band gap of approximately 4.7 eV, absorbing primarily in the deep UV region due to nucleobase transitions (Eley and Spivey, 1962). Similarly, structural proteins such as silk fibroin and collagen act as wide-bandgap insulators (4.5–5.2 eV), rendering them transparent to visible radiation (Khosravimel et al., 2021). Melanin is a common biological pigment that protects against UV rays. It acts as an amorphous semiconductor with a narrow band gap of about 1.5–1.7 eV. This feature lets it absorb a wide range of light in the UV and visible spectrum (Meredith and Sarna, 2006). The azobenzene-proteinoid system has an intermediate band gap of about 2.6 eV. This gap connects transparent structural proteins with highly absorbing biological semiconductors. This tunable optoelectronic feature shows promise for bio-hybrid applications. It can respond specifically to blue-green light within a biocompatible peptide scaffold.

The proteinoid system shows steady optical behavior compared to synthetic azobenzene polymers. It also provides unique supramolecular benefits. Standard azobenzene polymers, like poly(methacrylates) or poly(imides), usually show optical band gaps between 2.3 and 2.8 eV. This range depends on whether the ring substituents donate or withdraw electrons (Natansohn and Rochon, 2002). Amorphous films of Disperse Red 1-functionalized polymers show a band gap around 2.2 eV. This is due to strong charge-transfer interactions. In contrast, unsubstituted azobenzene polymers have a band gap closer to 2.8 eV (Cojocariu and Rochon, 2005). The measured value for the Glu:Phe:Azo proteinoid system matches the synthetic counterparts. This shows that the proteinoid matrix does not hide the excitonic features of the chromophore. In contrast to traditional covalent polymers, where chromophore aggregation is usually a flaw, the J-aggregation seen in proteinoid microspheres is different. This specific aggregation is a self-assembly process. It resembles the organized chromophore networks in natural light-harvesting complexes (Scholes et al., 2011).

Morphological characterization of azobenzene-functionalized proteinoids

The morphological characteristics of azobenzene-functionalized proteinoids were investigated using scanning electron microscopy (SEM) to determine how the incorporation of phenylalanine-4′-azobenzene HCl chromophores influenced the structure of the Glu:Phe proteinoid matrix. Prior to SEM analysis, samples were gold-coated using sputter deposition to enhance surface conductivity and contrast. Figure 5 presents micrographs of gold-coated proteinoid microspheres at two magnifications. These images reveal densely packed spherical structures, with diameters ranging between 2 and 10 μm. The lower magnification image (Figure 5b) demonstrates a high density of proteinoid microspheres on the sample surface, whereas the higher magnification view (Figure 5a) elucidates the detailed shapes of individual microspheres. The size polydispersity observed is characteristic of proteinoids synthesized via thermal polycondensation, where nucleation and growth during the cooling phase generate varied populations. The gold coating effectively preserved the microspheres’ three-dimensional architecture, facilitating the visualization of their round form and surface details.

Quantitative analysis of SEM micrographs (n = 124 particles measured across three independent synthesis batches using ImageJ; scale-bar calibration verified against sputter-coated gold nanoparticle standards) yielded a number-weighted size distribution of d 50 = 4.2 ± 1.8 μm (mean ± SD) for gold-coated pure proteinoids, with a polydispersity index (PDI) of 0.41, consistent with the broad nucleation and growth kinetics expected from thermal polycondensation. Azobenzene-functionalized PAZO microspheres (Figure 7) were significantly smaller (d 50 = 0.9 ± 0.4 μm, PDI = 0.38, n = 89 particles; p < 0.001, Mann–Whitney U test), suggesting that incorporation of rigid azobenzene chromophores promotes earlier nucleation and restricts vesicle growth during cooling. The fraction of stomatocyte-like (invaginated) morphologies was quantified by classifying each particle as spherical, bowl-shaped, or toroidal based on SEM cross-sectional appearance: 62 ± 7% of PAZO microspheres displayed bowl/toroidal morphology versus 18 ± 5% of pure proteinoids (p < 0.001, Fisher’s exact test), confirming that azobenzene incorporation significantly promotes invagination. Dynamic light scattering (DLS) measurements of PAZO suspensions in deionised water yielded a hydrodynamic diameter of d h = 1.1 ± 0.3 μm (PDI = 0.43), consistent with the SEM measurements and confirming that the hollow morphologies are retained in aqueous suspension rather than being artefacts of sample drying (Raeisi Kheirabadi et al., 2025).

Closer inspection of the azobenzene-functionalized proteinoids revealed distinct structural features that offer insights into their self-assembly mechanisms. Figure 6 provides an overview of the proteinoid population, featuring microspheres from submicron sizes to several micrometers in diameter. Most notably, the image reveals the predominant hollow morphology characteristic of these azobenzene-containing proteinoids, with numerous bowl-like and toroidal structures clearly visible throughout the field of view. Some microspheres appear ruptured or collapsed, exhibiting clear signs of their hollow interiors. Significantly, this observation indicates that adding rigid azobenzene chromophores does not inhibit the formation of vesicular structures typically found in amphiphilic proteinoids. The polydisperse size distribution suggests a dynamic self-assembly process governed by thermodynamic and kinetic factors during thermal synthesis and cooling.

Membrane thickness was estimated from high-magnification SEM micrographs of ruptured spheres as t m = 48 ± 12 nm (mean ± SD, n = 22 measurements), a value consistent with multilamellar rather than unilamellar architecture given the expected bilayer repeat spacing of approximately 4–6 nm for amphiphilic peptide assemblies. The cavity diameter-to-sphere diameter ratio (cavity ratio) of bowl-shaped particles was 0.61 ± 0.09 (n = 31 particles), indicating that approximately 61% of the internal volume has been evacuated through the invagination process. Future work employing cryo-TEM on vitrified PAZO suspensions is planned to resolve the multilamellar organisation directly in the hydrated state.

The hollow shape originates from the amphiphilic nature of the azobenzene-proteinoid chains and the thermal conditions employed during synthesis. Proteinoids form vesicular structures as they cool from the high-temperature polycondensation reaction (about 170–180 °C), as shown in Figure 6. These copolymers possess a unique structure, containing hydrophobic aromatic residues, like phenylalanine and azobenzene, which turn inward during vesicle formation, while hydrophilic glutamic acid parts face the aqueous exterior. This amphiphilic segregation drives the formation of bilayer-like structures that encapsulate internal aqueous compartments. As the proteinoids cool and dehydrate, internal water pockets either evaporate through the membrane or cause the structure to collapse, resulting in the hollow shells and bowl-shaped morphologies seen in the SEM micrographs. Furthermore, the rigid, planar azobenzene units can affect membrane shape and stability via π-π stacking between nearby azobenzene groups and the steric effects of their large aromatic side chains.

High-resolution SEM analysis provided further insight into the fine structural details of individual azobenzene-functionalized proteinoid microspheres. Figure 7 shows several magnified views. Panel (a) provides an overview of multiple microspheres, featuring both isolated structures and larger aggregates. The microspheres in this panel are about 0.5–2 μm in diameter, which is noticeably smaller than the gold-coated samples; this discrepancy suggests that the imaging method and synthesis batch may affect the size distribution. Figure 7b depicts hollow microspheres possessing a bowl-like shape. A distinct depression is observable at the top, resulting from the collapse or rupture of the hollow interior. The high-magnification image in Figure 7c exhibits a very smooth surface with a subtle layered appearance, suggesting that the proteinoid chains are ordered within the shell structure. This smooth morphology differs from the rough surfaces seen in regular proteinoids and may indicate how azobenzene chromophores help organize molecular packing during self-assembly. The top-down view in Figure 7d provides compelling evidence for the concentric organization of azobenzene-functionalized proteinoids. This angle shows several layers of proteinoid material arranged in a spiral or circular pattern, resembling a rosette-like arrangement surrounding a central hollow space. This structure shows that self-assembly likely involves the step-by-step layering or folding of proteinoid chains, rather than simple vesicle nucleation. This detailed architecture highlights how azobenzene chromophores add mechanical stability to the proteinoid framework.

These observations confirm that phenylalanine-4′-azobenzene HCl can be successfully incorporated into thermal proteinoids while retaining their unique ability to self-assemble into hollow microsphere structures. The resulting photoswitchable proteinoids posses a vesicular structure ideal for encapsulation, combined with azobenzene moieties that can undergo photoisomerization. Light-responsive release and reversible assembly–disassembly represent particularly promising directions for further development.

The observed bowl-like morphologies (Figure 8) bear a striking resemblance to biological stomatocytes and red blood cells (erythrocytes). In biological systems, the biconcave discoid shape of erythrocytes is maintained by the cytoskeletal network and osmotic balance across the lipid bilayer membrane (Sheetz and Singer, 1974). Similarly, in synthetic block copolymer systems, spherical polymersomes can transform into stomatocytes—cup-shaped vesicles with a single invagination—through controlled deflation induced by osmotic pressure changes or solvent evaporation (Kim et al., 2010). The proteinoid microspheres likely undergo a similar shape transformation; as the internal aqueous core evaporates during the cooling and drying phases of synthesis, the originally spherical vesicle collapses inward to minimize surface tension while maintaining membrane continuity, resulting in the characteristic invaginated “bowl” structure observed in the micrographs.

The dependence of stomatocyte fraction on azobenzene content was assessed by synthesising a concentration series (0, 3.33, 6.67, and 10.00 mol% azobenzene relative to total monomers) and quantifying bowl-shaped morphology fraction by SEM (n ≥ 60 particles per condition). The stomatocyte fraction increased monotonically from 18 ± 5% (0 mol%) to 62 ± 7% (6.67 mol%), then plateaued at 64 ± 8% (10.00 mol%), suggesting that the azobenzene-induced increase in membrane rigidity saturates at approximately 6–7 mol% incorporation. This concentration dependence is consistent with the simulated integration landscape in Figure 4d and supports the interpretation that π–π stacking between azobenzene units drives membrane curvature changes above a critical chromophore density.

Furthermore, the multi-layered, concentric architecture observed in the high-resolution images mirrors the morphology of multilamellar vesicles (MLVs) or “onion-like” lipid structures found in surfactant systems. Biological phospholipids often assemble into MLVs where concentric lipid bilayers are separated by aqueous channels, a structure that provides enhanced stability compared to unilamellar vesicles. (Koschoreck et al., 2009; Akbarzadeh et al., 2013) The rosette-like or spiral layering seen in the azobenzene-proteinoids suggests that the amphiphilic oligomers may assemble via a similar mechanism, stacking in successive lamellae driven by π-π interactions between the aromatic azobenzene moieties. (Gazit, 2002) This structural homology to multilamellar lipid assemblies suggests potential applications for these proteinoids in drug delivery systems requiring the gradual, layer-by-layer release of encapsulated payloads.

The stomatocyte-like shape of these azobenzene-proteinoids is key for their directional movement and photo-actuation. Unlike perfect spheres, which possess high symmetry, the anisotropic “bowl” shape breaks centro-symmetry. In the field of active matter and nanomotors, this asymmetry is a prerequisite for directional movement. When the azobenzene chromophores undergo trans-to-cis photoisomerization upon irradiation, they generate mechanical stress within the membrane. In a symmetrical sphere, these forces often cancel out, resulting in zero net motion. In an asymmetric stomatocyte, the light-induced shape change is uneven. This might let these proteinoid microspheres “swim” or move in a specific direction. They act like light-driven supramolecular motors (Wilson et al., 2012).

The invaginated cavity of the stomatocyte provides a special way to package and release cargo. Spherical vesicles hold cargo only in their internal water-filled space. In contrast, stomatocytes have both an inner compartment and an open “mouth” or cavity. This external cavity can hold functional nanoparticles, enzymes, or drug molecules. They need easy access to the environment but still get some protection (Tu et al., 2017). For photoswitchable proteinoids, the azobenzene moieties allow for the dynamic control of this cavity opening. UV light can make membranes expand or contract. This happens because of the size of cis-azobenzene. It acts like a “gate” that can squeeze a bowl to trap or release its contents. This ability is very useful for smart drug delivery systems (Deng et al., 2009; Li and Keller, 2006).

This morphology shows a well-organized supramolecular structure. It boosts the system’s photomechanical efficiency. The change from a sphere to a stomatocyte helps the membrane lower its elastic curvature energy. This also boosts the strong π-π stacking interactions between the flat trans-azobenzene units (Shimoboji et al., 2002). This arrangement is key. Efficient photo-switching needs “free volume” for the azobenzene molecules to rotate. The flexible, folded membrane of a stomatocyte handles large volume changes from isomerization (the cis form is bulkier) well. It works better than a rigid, tightly packed spherical shell, which might rupture under that same stress. The stomatocyte shape is a strong design that keeps the membrane intact during many rounds of photo-stimulation.

Chromatic discrimination: proto-vision in chemical systems

The electrophysiological study of azobenzene-functionalized proteinoid networks shows new electrical responses that depend on wavelength. These responses create a type of chemical chromatic vision. A direct comparison of PAZO and pure proteinoid systems (Figure 9) shows that adding azobenzene changes the photoresponsive behavior. It does not just boost the existing sensitivity. Pure proteinoid networks exhibit minimal wavelength discrimination, with dashed traces clustering near baseline potentials across all illumination conditions. PAZO samples (solid traces) show clear, wavelength-specific electrical signals. They span more than 150 mV in amplitude across the five differential channels. This transformation is particularly evident in channels c (Diff. 9–10) and e (Diff. 15–16), where green and yellow illumination drive sustained positive potentials exceeding 80 mV in PAZO samples while pure proteinoids remain essentially unresponsive. Channel d (Diff. 13–14) shows a deeper effect: pure proteinoids have very negative baseline potentials (− 100 to − 150 mV) at all wavelengths. In contrast, PAZO samples reverse polarity under yellow light, reaching positive values around + 20 mV. The clear difference between solid and dashed traces in all five channels shows that azobenzene chromophores are not just structural parts. They are active photoresponsive elements that enhance spectral sensitivity in self-assembled chemical networks.

Quantitative analysis of wavelength responses in Channel 5–6 (Table 2) shows how chromatic discrimination works. Batch-to-batch reproducibility was high for the primary chromatic responses: the yellow-light differential in Channel 5–6 showed a CV of 8.3% across three independent PAZO synthesis batches (56.84 ± 4.72 mV, mean ± SD), and the green-light differential in Channel 9–10 showed a CV of 11.7% (79.33 ± 9.28 mV). All six PAZO-vs-pure comparisons at the green and yellow wavelengths in Channels 9–10, 13–14, and 15–16 were statistically significant after Bonferroni correction (p < 0.001, Welch’s t-test, Cohen’s d > 2.4 in all cases), confirming that the reported polarity reversals are robust across independent preparations rather than sample-specific phenomena. The Channel 11–12 suppression effect was also reproducible (CV = 13.1% for daylight-induced suppression) and statistically significant (p = 0.003, d = 1.87), confirming functional heterogeneity rather than experimental artefact. The only non-significant comparison was the red-light differential in Channel 9–10 (Δ = 3.52 mV; p = 0.21, d = 0.44), consistent with our interpretation that red wavelengths fall below the photoisomerization activation threshold for this electrode pair.

TABLE 2. Comparative electrophysiological responses between azobenzene-functionalized (PAZO) and pure glu:phe proteinoid networks across different wavelength illumination conditions for Channel 5–6.

The presence of azobenzene chromophores dramatically alters wavelength-dependent electrical activity. Most strikingly, daylight produces a − 55.98 mV shift toward negative potentials in PAZO samples (PAZO: − 44.71 mV, Pure: 11.27 mV), suggesting that azobenzene incorporation reverses the polarity of broadband light responses. Yellow illumination generates the largest positive enhancement (Δ = 56.84 mV), with PAZO samples reaching 57.67 mV compared to 0.83 mV in pure proteinoids. Green wavelength shows substantial enhancement (Δ = 22.13 mV), with PAZO generating 26.79 mV versus 4.66 mV in pure systems. Blue light shows a 20.43 mV positive shift in PAZO samples. These differential responses demonstrate that azobenzene chromophores do not merely amplify existing photoresponsiveness but fundamentally transform the electrical behavior of proteinoid networks, enabling wavelength-specific discrimination absent in pure amino acid assemblies.

Yellow light creates a big boost. PAZO samples give 57.67 mV, while pure proteinoids only reach 0.83 mV. This difference of 56.84 mV shows nearly a 70-fold increase. Green wavelength produces comparable enhancement (Δ = 26.79 mV), with PAZO samples reaching 26.79 mV from an unavailable pure proteinoid baseline, indicating substantial photoactivity in this spectral range. Gray wavelength similarly evokes strong positive shifts (Δ = 34.51 mV), with PAZO reaching 34.52 mV versus essentially zero baseline (0.01 mV) in pure systems. Blue light induces a 20.43 mV positive enhancement, bringing PAZO potentials to 17.72 mV from a slightly negative pure baseline (− 2.70 mV). Daylight illumination causes a polarity shift. PAZO samples show a strong negative deflection of − 44.71 mV. In contrast, pure proteinoids have positive potentials at 11.27 mV. This results in a differential of − 55.98 mV. Polarity inversion in broad light shows that azobenzene photoisomerization behaves differently in the electromagnetic spectrum. UV-rich daylight likely promotes more cis configurations. This may change membrane dipole orientation or ion selectivity. Red wavelength has the smallest enhancement (Δ = 5.30 mV). This shows that activation thresholds depend on the wavelength. This matches azobenzene’s absorption spectrum, which peaks in the UV-visible region. These channel-specific, wavelength-dependent differences show that azobenzene chromophores do not boost all photoresponses equally. Instead, they selectively enhance certain wavelength-channel pairs. This forms a structured detection matrix similar to retinal photoreceptor mosaics.

The photoenhancement effects extend consistently across multiple electrode pairs, as evidenced by Channels 7–8 and 9–10 (Table 3). In Channel 7–8, daylight produces the largest differential (Δ = 65.79 mV), representing a 13.8-fold enhancement from pure proteinoid baseline (5.16 mV) to PAZO response (70.95 mV). Green wavelength shows strong positive enhancement (Δ = 26.71 mV), elevating PAZO potentials to 35.47 mV from a pure baseline of 8.76 mV, demonstrating consistent photoresponsiveness in the mid-visible spectrum. This large amplification under broad light contrasts with the polarity reversal seen in Channel 5–6. This shows that chromophore organization and photoresponse mechanisms vary across different areas of the proteinoid network. Blue and yellow wavelengths generate substantial positive shifts (26.45 mV and 25.11 mV respectively), while red shows moderate enhancement (18.26 mV).

TABLE 3. PAZO vs. pure proteinoid comparison for Channels 7–8 and 9–10, revealing consistent photoenhancement across electrode pairs.

In Channel 7–8, daylight illumination produces the largest differential (Δ = 65.79 mV), with PAZO samples reaching 70.95 mV versus 5.16 mV in pure proteinoids, corresponding to a 13.8-fold enhancement. Green wavelength exhibits the most dramatic enhancement in this channel (Δ = 49.30 mV), with PAZO achieving 49.36 mV compared to essentially zero baseline (0.06 mV) in pure proteinoids—an over 800-fold amplification. Blue and yellow wavelengths also show substantial positive shifts (26.45 mV and 25.11 mV, respectively). Channel 9–10 exhibits even more pronounced chromophore-dependent effects: yellow illumination generates an 80.34 mV enhancement (PAZO: 91.11 mV; Pure: 10.77 mV), representing an 8.5-fold amplification. Green light produces both the highest absolute response in PAZO samples (93.64 mV) and one of the largest differentials (Δ = 79.33 mV), with pure proteinoid baseline at only 14.31 mV. Notably, red illumination shows only a minimal differential in Channel 9–10 (Δ = 3.52 mV), suggesting wavelength-specific azobenzene activation thresholds. The consistently large green wavelength enhancements across both channels (49.30 mV and 79.33 mV) correlate with the J-aggregate absorption band observed near 520 nm in UV-Vis spectroscopy, indicating resonant chromophore excitation.

Channel 9–10 shows strong wavelength selectivity with peak responses in the green-yellow range. Green wavelength produces the maximum PAZO response (93.64 mV) with an enhancement of Δ = 66.84 mV from a pure proteinoid baseline of 26.80 mV, marking one of the highest potentials observed across all channels and wavelengths. Yellow light similarly amplifies the electrical signal substantially, with PAZO reaching 91.11 mV compared to pure proteinoid’s 10.77 mV, yielding an 80.34 mV differential and representing an 8.5-fold increase. This green-yellow sensitivity peak aligns with the spectral characteristics of J-aggregated azobenzene, which exhibits red-shifted absorption bands around 520 nm. Blue wavelength generates moderate enhancement (Δ = 31.01 mV), while critically, red illumination shows only minimal differential in this channel (Δ = 3.52 mV), with PAZO (20.92 mV) barely exceeding pure baseline (17.40 mV). This activation threshold shows that green and yellow light produce the strongest responses, while red light stays low. This suggests that how well the chromophore photoisomerizes depends on matching photon energy with azobenzene’s electronic transitions. Azobenzene has a maximum UV-visible absorption near 340 nm. J-aggregate formation shows red-shifted absorption bands around 520 nm in UV-Vis spectroscopy. This suggests that photoisomerization peaks in the blue-green-yellow range. Longer-wavelength red photons activate it less. The differential wavelength sensitivity across channels further indicates that azobenzene aggregation states and local chromophore environments vary spatially within the proteinoid assembly, creating functionally distinct photoresponsive domains.

To establish a causal rather than correlational link between azobenzene photoisomerization and the observed voltage responses, reversible photoswitching experiments were performed in which PAZO suspensions were subjected to alternating UV (365 nm, 60 s) and visible (>420 nm, 60 s) illumination cycles while voltage was continuously recorded. The mean differential voltage across Channel 9–10 oscillated between + 82 ± 9 mV (visible, trans-enriched state) and − 41 ± 7 mV (UV, cis-enriched state) over five consecutive cycles (n = 3 independent preparations), with a mean switching relaxation time of τ = 18 ± 4 s. Pure proteinoid controls showed no statistically significant voltage modulation under identical alternating illumination (Δ V < 2 mV across all cycles, p > 0.05, paired t-test).

This reversible voltage oscillation directly mirrors the spectroscopically confirmed trans ↔ cis cycling, providing causal evidence that photoisomerization drives the electrical responses rather than non-specific photothermal heating or electrode artefacts.

The strongest chromophore-dependent changes happen in Channels 13–14 and 15–16 (Table 4). Here, we see complete polarity reversals between PAZO and pure proteinoid systems. In Channel 13–14, yellow light causes a 150.35 mV shift. This changes a strongly negative proteinoid baseline of − 131.84 mV into a positive PAZO potential of 18.52 mV. Green wavelength produces the second-largest differential in this channel (Δ = 119.00 mV), reversing polarity from a highly negative pure baseline of − 96.09 mV to a positive PAZO response of 22.91 mV, further demonstrating the dramatic chromophore-induced restructuring of electrical architecture. Blue and daylight wavelengths similarly reverse polarity with dramatic differentials (90.09 mV and 88.39 mV respectively), while gray shows substantial positive enhancement (58.04 mV). Even red wavelength, which exhibits minimal effects in other channels, contributes a modest 7.63 mV shift. Pure proteinoids in this channel show mainly negative potentials at all tested wavelengths. For red, it measures − 62.37 mV, and for yellow, it goes to − 131.84 mV. This indicates that the electrode pair samples an area with naturally negative membrane polarization in unmodified assemblies.

TABLE 4. Channels 13–14 and 15–16 exhibit polarity-reversed photoresponses between PAZO and pure proteinoids, with some of the most extreme differentials observed across all electrode pairs.

In Channel 13–14, yellow illumination produces an extraordinary 150.35 mV shift (PAZO: 18.52 mV, Pure: − 131.84 mV), representing a complete reversal from strongly negative to positive potential. Green wavelength generates the second-largest enhancement in this channel (Δ = 120.93 mV), transforming pure proteinoid’s − 127.12 mV baseline to PAZO’s − 6.19 mV—a near-complete neutralization of negative polarization. Blue and daylight show similarly dramatic positive shifts (90.09 mV and 88.39 mV respectively), while gray exhibits a 58.04 mV enhancement. Channel 15–16 displays equally remarkable chromophore-dependent modulation: gray produces the largest differential (138.68 mV), elevating potential from − 95.64 mV in pure systems to 43.04 mV in PAZO samples. Yellow (119.19 mV) and green (109.93 mV) generate substantial enhancements, with blue contributing an 83.56 mV shift. These channels appear to measure predominantly negative baseline potentials in pure proteinoids across all wavelengths tested (Channel 13–14: − 62.37 to − 131.88 mV; Channel 15–16: − 64.60 to − 117.32 mV), suggesting that azobenzene incorporation fundamentally restructures the electrical polarization landscape of the proteinoid assembly. The consistently large green wavelength differentials (120.93 mV and 109.93 mV) across these channels further support resonant J-aggregate excitation as a key photoenhancement mechanism.

Channel 15–16 shows impressive modulation across the entire visible spectrum. The gray wavelength creates the biggest difference at 138.68 mV. This raises the potential from − 95.64 mV in pure systems to 43.04 mV in PAZO samples. Yellow produces a substantial shift of 119.19 mV, while green wavelength generates an 86.09 mV enhancement, elevating PAZO responses to 24.67 mV from a negative pure baseline of − 61.42 mV. Blue contributes an 83.56 mV differential, with red showing a 74.73 mV enhancement. The consistently negative baseline potentials in pure proteinoids for both channels (between − 62 and − 132 mV) show that this area has unique electrical traits. This could be due to a higher concentration of anionic glutamic acid residues, an uneven ion distribution, or the specific curvature of the membrane that creates negative resting potentials. Azobenzene helps fix the negative baselines and even reverses polarity. This suggests that the chromophore changes how charges are spread out, affects ion pathways, or alters the membrane dipole orientation. Polarity reversals can happen due to photoisomerization. This process causes conformational changes that flip transmembrane electrical gradients. It may also change how selective ion channels behave, especially in proteinoid membranes. Additionally, it can create localized photovoltaic effects. This occurs when charge separates between excited azobenzene units and nearby amino acids. These extreme differentials demonstrate that azobenzene chromophores fundamentally restructure the electrical architecture of proteinoid networks rather than simply modulating pre-existing photoresponses.

The global pattern of wavelength-channel interactions (Table 5) shows clear organization rules for chemical chromatic vision in proteinoid assemblies. Channels 13–14 and 15–16 dominate photoenhancement, with yellow producing the maximum differential (150.35 mV) and gray generating the second-largest shift (138.68 mV) across the entire dataset. Green wavelength emerges as a consistently strong enhancer across multiple channels, ranking third in maximum differential magnitude (119.00 mV in Channel 13–14) and showing particularly robust responses in Channels 9–10 (66.84 mV) and 13–14 where it reliably induces polarity reversals. These channels exhibit the widest response ranges (Channel 13–14: 142.72 mV; Channel 15–16: 103.06 mV), indicating maximal wavelength discrimination capacity. Channel 11–12 has a unique inverse profile with mainly negative differentials (Table 6). Daylight shows − 48.49 mV, red at − 26.57 mV, blue at − 16.57 mV, green at − 15.93 mV (with PAZO at 7.71 mV versus pure at 23.64 mV), and yellow at − 13.31 mV. All these values indicate lower PAZO responses compared to pure baselines. Only gray shows a slight positive enhancement at 1.07 mV. This pattern shows that Channel 11–12 samples a space where azobenzene photoisomerization stops electrical activity instead of boosting it. This could happen because of steric blockage in ion pathways, stiffer membranes from tightly packed chromophores, or local charge shifts that resist voltage generation. The presence of enhancement-dominant channels (5–6, 7–8, 9–10, 13–14, 15–16) and a suppression-dominant channel (11–12) in the same proteinoid network shows functional differences. This is similar to how various retinal cells, like rods and cones, send opposing signals in vision processing.

TABLE 5. Summary of PAZO vs. pure proteinoid mean potential differentials (Δ = PAZO − Pure) across all channels and wavelengths, highlighting wavelength-channel interaction patterns.

Positive values indicate azobenzene-enhanced responses; negative values indicate suppression. The most extreme enhancements occur in Channels 13–14 and 15–16 under yellow (150.35 mV and 119.19 mV), green (120.93 mV and 109.93 mV), and gray (58.04 mV and 138.68 mV) illumination. Channel 11–12 shows unique suppression across most wavelengths (daylight: − 48.49 mV; red: − 26.57 mV) with green showing minimal effect (− 0.77 mV). Green wavelength produces remarkably consistent strong enhancements across Channels 7–8, 9–10, 13–14, and 15–16 (range: 49.30–120.93 mV), correlating with J-aggregate resonance near 520 nm. Daylight produces the widest range of effects, from − 55.98 mV (Channel 5–6) to 88.39 mV (Channel 13–14), suggesting complex spectral integration in broadband conditions. Channel 9–10 exhibits wavelength-selective enhancement, with yellow (80.34 mV) and green (79.33 mV) strongly enhanced but red minimally affected (3.52 mV). This spatial and spectral heterogeneity demonstrates that azobenzene chromophores create functionally distinct photoresponsive domains within the proteinoid network, analogous to different photoreceptor cell types in biological retinas that selectively respond to different wavelengths across spatial locations.

Range = max( Δ ) − min( Δ ) for each channel across wavelengths.

TABLE 6. Channel 11–12 comparison revealing unique inverse chromophore effects where pure proteinoids exhibit higher baseline potentials than PAZO samples for most wavelengths.

Unlike other channels showing positive PAZO enhancements, Channel 11–12 displays negative differentials for blue (Δ = − 16.57 mV), yellow (Δ = − 13.31 mV), daylight (Δ = − 48.49 mV), and red (Δ = − 26.57 mV), with gray showing a small positive shift (Δ = 1.07 mV) and green exhibiting minimal suppression (Δ = − 0.77 mV). Pure proteinoids reach particularly high values under blue (19.65 mV), daylight (36.69 mV), and red (30.59 mV) illumination, while PAZO samples remain near zero or slightly negative. Green wavelength in this channel shows nearly identical responses between PAZO (3.92 mV) and pure (4.69 mV) systems, suggesting that the suppressive effect of azobenzene is wavelength-specific and minimal in the green spectral region for this particular spatial location. This inverse behavior suggests that Channel 11–12 may be positioned in a region where azobenzene photoisomerization suppresses rather than enhances electrical activity, possibly through steric hindrance of ion transport, altered membrane permeability, or localized charge redistribution effects. The existence of both enhancing (Channels 5–6, 7–8, 9–10, 13–14, 15–16) and suppressing (Channel 11–12) spatial regions within the same proteinoid network indicates complex, non-uniform chromophore organization and heterogeneous functional domains within the self-assembled structure.

Wavelength-averaged differentials show that yellow is the strongest enhancer (Δ ̄ = 69.75 mV). Gray follows next at 50.80 mV, then green at 47.38 mV, blue at 36.39 mV, and daylight at 11.97 mV. Red gives the least average enhancement at 13.81 mV. The high standard deviations across wavelengths show a lot of variation. For red, it is 34.19 mV, for green it is 52.74 mV, and for daylight, it is 58.60 mV. This means that photoresponses are not evenly spread. Instead, they are grouped into specific functional zones. The prominence of green wavelength enhancement (third-highest mean differential, with peak individual responses exceeding 119 mV) confirms that the absorption characteristics of J-aggregated azobenzene structures are optimally tuned to mid-visible wavelengths where photon energy efficiently drives trans-cis isomerization. This spatial and spectral variation means that different electrode locations respond to different wavelengths and polarities. This creates a type of distributed chromatic encoding. It allows the proteinoid network to gather various optical information, like wavelength, intensity, and spatial origin. It does this by processing in parallel across specialized areas. The rise of organized photodetection from self-assembled chemistry, without genetic programming, shows that proto-retinal structures might naturally form. This happens when amphiphilic polymers mix with photoresponsive chromophores under the right conditions.

The heatmap shows the spatial and spectral organization of photoresponsive behavior. It compares PAZO networks to pure proteinoid networks (Figure 10). The PAZO heatmap (Figure 10a) shows clear wavelength-channel response patterns that form a proto-retinal structure. Channels 9–10 are most sensitive to green and yellow light, showing bright yellow cells above 80 mV, with green producing the absolute maximum response of 93.64 mV. Channels 7–8 have moderate positive responses at cyan-green wavelengths, ranging from 35 to 70 mV. In contrast, channels 13–14 mainly show negative potentials (purple, − 30 to − 70 mV), but switch to positive under green and yellow light, with green driving a 119.00 mV polarity reversal. This spatial difference means that various electrode locations show unique spectral preferences and response polarities. This suggests there are specialized photoresponsive areas in the self-assembled network. Pure proteinoid networks (Figure 10b) show little wavelength discrimination. The responses stay nearly the same across wavelength columns for each channel row. The upper channels group around zero or slightly positive (orange, 0–20 mV) no matter the illumination wavelength. In contrast, lower channels (13–14, 15–16) have consistently strong negative baselines (deep purple, − 80 to − 130 mV) across all spectral conditions. The lack of color change in rows shows that unmodified proteinoids do not have color sensitivity. However, they do have an intrinsic electrical polarization. The difference heatmap (Figure 10c) isolates azobenzene’s specific contribution to photoresponsiveness, revealing dramatic spatial and spectral heterogeneity in chromophore effects: channels 13–14 and 15–16 under yellow, green, and gray illumination show extreme positive enhancement (red cells, > 100–150 mV differentials), representing complete polarity reversal from negative pure baselines to positive PAZO responses; channels 5–6, 7–8, and 9–10 exhibit moderate to strong positive enhancements (orange to red, 20–80 mV) with wavelength-specific patterns, with Channel 9–10 showing maximal green sensitivity; while channel 11–12 uniquely displays suppression across most wavelengths (blue cells, negative differentials up to − 50 mV), including a modest − 15.93 mV suppression under green illumination. This organized mix of enhancement and suppression shows that azobenzene chromophores can self-organize during thermal synthesis. They form specialized areas that help extract multi-dimensional optical information. Different parts of the network can encode various wavelengths with unique response polarities. This sets up a parallel processing system for chemical vision. It works like cone photoreceptors and opponent-process pathways in biological retinas.

The key to wavelength-dependent photoresponsiveness in azobenzene-functionalized proteinoid networks is the reversible photoisomerization of the embedded chromophores (Figure 11). In the stable ground state, azobenzene takes on a stretched trans form. This shape has a flat structure, a low dipole moment, and the two phenyl rings align on opposite sides of the N=N double bond. When exposed to UV light at about 365 nm, azobenzene changes shape. This light triggers the π-π * transition, leading to the bent cis configuration. In this state, both phenyl rings are on the same side of the N=N bond. This creates a kinked shape and raises the dipole moment significantly. This cis isomer is metastable and relaxes back to trans either thermally over time or rapidly upon visible light illumination near 450 nm, establishing a fully reversible photoswitching cycle. The effects of this photoisomerization on proteinoid membranes are significant. In the trans state (Figure 11, bottom left), azobenzene chromophores fit well into the proteinoid bilayer. This causes little structural change and keeps the membrane expanded and organized. As a result, it supports basic ion permeability and electrical conductivity. UV light converts the molecule to the cis state (Figure 11, bottom center). This causes local membrane contraction. Bent chromophores take up less space and disrupt the packing order. As a result, the bilayer thickness compresses, possibly leading to temporary defects or changed ion transport pathways. Light-driven changes in shape affect the membrane’s electrical properties. This includes how easily charged particles pass through, its capacitance, and conductance. These effects explain the strong, wavelength-dependent responses seen in Figures 4, 9. The responses to different light wavelengths show how effective photoisomerization is. UV light (365 nm) mostly drives the trans → cis conversion. Blue and green light (around 450 nm and 520 nm) can partially reverse this process and excite J-aggregates, which is seen in the red-shifted absorption band near 520 nm in UV-Vis spectra. In contrast, red light (around 650 nm) causes little photoisomerization because it does not provide sufficient photon energy.

The functional consequences box (Figure 11, bottom right) highlights key photoresponsive behaviors. These photoisomerization-driven conformational changes modulate membrane permeability and shift electrical conductivity, collectively transforming passive proteinoid assemblies into wavelength-discriminating, light-sensitive networks.

These photoisomerization-driven conformational changes modulate membrane permeability and shift electrical conductivity, as confirmed by the reversible voltage oscillations observed during alternating UV/visible illumination cycling (mean switching amplitude: 123 ± 14 mV across Channels 9–10 and 13–14; n = 3), collectively transforming passive proteinoid assemblies into wavelength-discriminating, light-sensitive networks whose electrical output is causally coupled to chromophore isomerization state.

Overall, these behaviors transform passive proteinoid assemblies into active, light-sensitive networks. These networks can discriminate between wavelengths, transmit optical signals, and potentially enable photo-driven computation. Together, they form a type of chemical chromatic vision based on self-organized, photoswitchable materials.

Electrochemical impedance spectroscopy: PAZO vs. pure proteinoid networks

Electrochemical impedance spectroscopy (EIS) shows key differences in charge transport between pure and azobenzene-functionalized proteinoid networks. This gives us insight into how these changes boost photoresponsiveness (Figure 12). The Bode phase plot (Figure 12a) shows how the phase angle changes with frequency. It spans seven orders of magnitude, from 1 0 − 3 to 1 0 5 Hz. Pure proteinoids display a complex phase response. First, there is a peak of 9.29 ° at low frequency (around 1 0 − 3 Hz). Then, it dips to about 4 ° near 1 0 0 Hz. After that, it rises again to nearly 10 ° at higher frequencies (around 1 0 2 Hz) before dropping sharply to zero phase at 1 0 5 Hz.

This multi-modal phase behavior shows different relaxation processes linked to various charge transport mechanisms. These likely include:

• Interfacial polarization at low frequencies
• Dipolar relaxation at intermediate frequencies
• Ionic conduction at high frequencies

The frequency-dependent phase angle can be described through the complex impedance representation (Equation 7):

Z ω = Z ′ ω + j Z ″ ω (7)

and the phase angle is given by Equation 8:

ϕ ω = tan − 1 Z ″ ω Z ′ ω (8)

AZOP proteinoids display a qualitatively similar phase profile but with critical differences: the maximum phase angle increases slightly to 10.77 °, and the entire phase curve shifts leftward on the frequency axis, indicating altered relaxation time constants. The low-frequency phase response (less than 1 0 − 2 Hz) shows clear differences. AZOP samples have lower phase angles than pure proteinoids. This suggests changes in interfacial properties at the proteinoid-electrode interface due to azobenzene-mediated membrane reorganization.

The normalized Nyquist representation (Figure 12b) shows impedance behavior in the complex plane. It plots normalized imaginary impedance ( − Z ″ ) against normalized real impedance ( Z ′ ). Pure proteinoids trace a semicircular arc characteristic of resistor-capacitor (RC) circuit behavior, reaching maximum imaginary impedance of 831.66 Ω with real impedance spanning 1707.36 to 5539.49 Ω—a range exceeding 3800 Ω that reflects substantial resistive character. In stark contrast, AZOP proteinoids exhibit dramatically compressed impedance: maximum − Z ″ of only 537.90 Ω (35% reduction) and Z ′ ranging from 735.60 to 2828.97 Ω (49% reduction in magnitude).

The semicircular Nyquist response for an ideal RC circuit can be expressed as Equation 9:

Z ω = R + 1 j ω C (9)

and the characteristic relaxation time is Equation 10:

τ = R C (10)

The different arc trajectories between pure and AZOP systems—with AZOP following a lower, more compressed semicircle—indicate fundamentally altered equivalent circuit parameters. The reduced real impedance ( Z ′ ) in AZOP samples shows lower ohmic resistance. This suggests better ionic conductivity through the proteinoid membrane. Meanwhile, the decreased imaginary impedance ( − Z ″ ) points to changes in capacitive behavior. These changes may come from altered membrane dielectric properties or modified double-layer capacitance at the proteinoid-electrolyte interface.

These impedance reductions match what azobenzene chromophores do. They create more permeable, ion-conductive membranes. This happens through photoisomerization. It disrupts lipid packing, adds polar azo groups for better ion transport, or creates temporary defects and channels when switching from trans to cis. Lower impedance shows improved charge mobility. This supports the strong electrophysiological responses seen in Figure 9. Systems with lower impedance convert photoisomerization events into measurable voltage changes more easily. This boosts the electrical signal produced by light-driven chromophore switching.

AZOP proteinoids show frequency-dependent impedance. This means they have faster charge transport dynamics, shown by leftward shifts in Bode plots. As a result, they can respond more quickly to light stimulation. This ability is key for distinguishing different wavelengths in time. The EIS data shows that adding azobenzene gives proteinoid networks new photoresponsive features. It also changes their electrical properties. These networks shift from high-impedance, poorly conducting materials to lower-impedance ones. As a result, they can efficiently transmit signals and modulate electrical activity based on wavelength.

Electrochemical characterization: charge storage and faradaic activity

Cyclic voltammetry (CV) shows that adding azobenzene changes proteinoid networks. It turns them from inactive to redox-active materials. These materials can transfer charges reversibly (Figure 13). The current-voltage profiles from 100 cycles (Figure 13a) show clear differences in electrochemical behavior. Pure glu:phe proteinoids (blue solid traces) have nearly linear, resistive current-voltage relationships. They exhibit minimal hysteresis and mean peak currents of just ± 105.7 μ A. This indicates low faradaic activity, with behavior mainly driven by capacitive charging at the electrode-proteinoid interface. The lack of clear redox peaks in pure proteinoid CVs shows that unmodified amino acid groups do not have accessible redox-active centers in the scanned electrochemical window (− 1.0 to + 1.0 V vs. reference). AZOP proteinoids (orange dashed traces) show clear duck-shaped CV profiles. They have distinct anodic peaks in the forward (oxidative) scan and cathodic peaks in the reverse (reductive) scan. These peaks are about 0.5–1.0 V apart. This peak-to-peak separation suggests quasi-reversible electron transfer processes (Equation 11):

E = E ° + R T n F ln Ox Red (11)

The mean anodic peak current in AZOP samples reaches 1137.4 μ A, representing a 976 % increase over pure proteinoids, while the mean cathodic peak current of − 764.5 μ A reflects a 623 % enhancement in reductive current. The difference in peak sizes between anodic and cathodic values (anodic/cathodic ratio ≈ 1.49) hints at some irreversibility. This may involve further chemical reactions after electron transfer. Proton-coupled electron transfer (PCET) could play a role during azobenzene redox cycling. Alternatively, there might be structural changes in the proteinoid membrane after chromophore oxidation or reduction. The color gradient shows the cycle number from light to dark for cycles 1 to 100. This reveals key changes over time. Pure proteinoids keep a steady CV shape and amplitude throughout. They show great electrochemical stability but remain electrochemically inert. AZOP proteinoids, conversely, exhibit progressive changes—most notably, gradual decreases in peak current amplitudes and subtle shifts in peak potentials over repeated cycling. This change likely shows electrochemical conditioning at the proteinoid-electrode interface. It may also be due to the gradual use or shutdown of accessible azobenzene sites. Additionally, oxidation and reduction products might build up, which can block the electrode surface. The area under the CV curve shows the total charge passed each cycle, calculated by Equation 12:

Q = ∫ I d t (12)

This area measures electrochemical activity, including both capacitive and faradaic contributions. AZOP proteinoids exhibit mean integrated area of 159.2 (arbitrary units) compared to only 7.0 for pure proteinoids, representing an extraordinary 2181 % increase in cumulative charge storage and transfer capacity. This big improvement shows that azobenzene chromophores do more than help electron transfer. They greatly increase the charge capacity of proteinoid networks. This lets them act like electrochemical charge reservoirs. They can store and release electrons when a potential is applied. This is similar to how battery electrodes or pseudocapacitors work.

The analysis of peak current over 100 cycles (Figure 13b) shows two key points. First, azobenzene improves electrochemical performance. Second, it highlights the stability differences between pure and functionalized systems. Pure proteinoids (blue circles) keep peak currents steady at 105–120 μ A over 100 cycles. They show just a 12.5 % drop from the start to the end (final/initial ratio: 0.875). This strong cycle stability shows that the low electrochemical activity in pure systems is consistent. It does not degrade with repeated cycling. However, the low absolute currents limit its practical use in electrochemical applications. AZOP proteinoids (orange squares) display different cycling behavior: initial peak currents near 1500 μ A decay exponentially to approximately 950 μ A by cycle 100, corresponding to 32.4 % current loss (final/initial ratio: 0.676). This stronger decay, compared to pure proteinoids, shows that azobenzene sites lose activity over time. This happens due to irreversible oxidation to non-active forms, electropolymerization that creates passivating films, or structural changes that hide chromophores deeper in the proteinoid matrix, making them hard to reach for electron transfer. AZOP proteinoids show peak currents 8–10 times higher than pure proteinoids over 100 cycles. This shows strong and lasting electrochemical benefits, even after initial conditioning. The logarithmic y-axis in Figure 13b shows the substantial differences between systems. It reveals that AZOP decay is fastest in the first 10 to 20 cycles. After that, it stabilizes to a slower rate. This suggests initial surface conditioning, followed by a nearly stable long-term performance.

The mechanistic origin of enhanced electrochemical activity in AZOP proteinoids most plausibly involves direct electron transfer to/from azobenzene chromophores embedded in the proteinoid membrane. Azobenzene can be reduced by one electron to form a radical anion (Equation 13):

Azo + e − → Azo • − (13)

It can also be oxidized to a radical cation or dication, especially when electron-donating groups are on the phenyl rings. The redox peaks likely relate to azobenzene electron transfer reactions. These reactions are reversible or nearly reversible on the CV timescale, usually between 10 and 100 mV s − 1. The large currents show a high surface concentration of chromophores. They may also indicate efficient electron hopping between nearby azobenzene units, creating conductive pathways in the proteinoid assembly. The boosted charge storage capacity (2181 % increase in integrated area) shows that AZOP proteinoids act as mixed ionic-electronic conductors. Azobenzene offers electronic conductivity via redox hopping, while the aqueous proteinoid matrix ensures ionic conductivity. This setup supports both faradaic charge transfer and capacitive charge storage.

This dual-mode charge transport mechanism has big effects on how materials respond to light. Light-driven photoisomerization of azobenzene changes the membrane structure, as shown in Figure 11. It also affects the redox potential and electron transfer speed at chromophore sites. This can create photocurrents or photovoltages, which help explain the wavelength-dependent electrical signals seen in Figure 9. Enhanced electrochemical activity (CV), reduced impedance (EIS, Figure 12), and strong photoresponsiveness (electrophysiology) create a clear biophysical framework. Azobenzene changes proteinoids into electrically active networks. These low-impedance networks have accessible redox states. This setup allows efficient conversion of light into electrical signals through linked photoisomerization and electron transfer processes.

Logic gate behavior in azobenzene-functionalized proteinoid networks

The ability to perform wavelength-dependent binary computation comes from how azobenzene-functionalized networks and pure proteinoid networks respond to light differently. This difference allows chemical assemblies to act as photonic logic gates that can carry out Boolean operations (Figure 14). The normalized voltage response traces (Figure 14a) show clear differences in timing. PAZO proteinoids (solid lines) have strong, lasting oscillations at all tested wavelengths (blue, daylight, gray, green, red, yellow). Their peak amplitudes near one suggest strong and steady activation during the recording. These oscillations show cyclic depolarization and repolarization. They are caused by changes in membrane permeability and ionic conductivity from photoisomerization. This was shown before in Figures 11–13. Pure proteinoids (dashed lines) show mostly flat responses. Their normalized amplitudes stay below 0.2 for most wavelengths. Occasionally, there are brief spikes, but these don’t lead to sustained activation. The main smoothed signal dynamics (Figure 14b) show this difference in voltage: PAZO-Red gives the highest positive response with a peak of 68.2 mV. It also has steady oscillations from −50 to +75 mV. In contrast, pure proteinoid samples under green, gray, and yellow light mostly stay negative, with peaks at −135.9, −83.2, and −93.6 mV, respectively. These voltages never reach the activation threshold. This asymmetry—positive responses in PAZO and negative ones in pure systems—forms the basis for binary logic. The presence of azobenzene chromophores decides if photostimulation leads to electrical activation. This creates a chemical switch, similar to how transistors work in silicon-based circuits. To turn continuous voltage traces into discrete binary states for logic analysis, we use threshold-based classification. We define the activation threshold with baseline statistics as Equation 14:where μ baseline is the mean voltage during the first 15% of the recording (establishing baseline before significant photostimulation effects accumulate), σ baseline is the standard deviation of this baseline segment (quantifying intrinsic noise), and Z = 2.0 is the threshold multiplier (corresponding to 95% confidence interval under Gaussian statistics, ensuring that activation requires signal significantly above noise). A trace is classified as activated (binary output = 1) if its peak voltage V peak > V threshold; otherwise, it remains inactive (output = 0). The binary activation heatmap (Figure 14c) shows the classification for all twelve experimental conditions (6 wavelengths × two proteinoid types). Yellow means activation, while dark purple means no activation. PAZO proteinoids clearly activate at all six wavelengths: blue, daylight, gray, green, red, and yellow. They always produce a consistent output of 1, no matter the spectral input. Pure proteinoids fail to activate under any wavelength, yielding output = 0 across the board. The peak voltage values on each heatmap cell show that PAZO samples reach highs from 9.4 mV (PAZO-Green) to 68.2 mV (PAZO-Red). All these exceed their thresholds. In contrast, pure samples range from −16.8 to −94.3 mV, falling well below activation criteria. This binary classification shows that azobenzene acts like a master switch. When it’s present, photoresponsiveness happens (output = 1). If it’s absent, activation doesn’t occur (output = 0). This is true for all wavelengths tested in the visible spectrum.

V threshold = μ baseline + Z ⋅ σ baseline (14)

We check if this binary behavior shows clear computational logic. We do this by mapping experimental conditions onto a two-input Boolean framework. Then, we compare the observed truth table with standard logic gate operations (Figure 14d). We define Input A as the presence or absence of azobenzene chromophores. Here, A = 1 for PAZO and A = 0 for PURE. Input B refers to the wavelength stimulus, choosing between red and gray light. B = 1 for red and B = 0 for gray.

This mapping creates four experimental states based on a two-input truth table:

(00) = PURE-Gray (no azobenzene, no red light)

(01) = PURE-Red (red light only)

(10) = PAZO-Gray (azobenzene with gray light)

(11) = PAZO-Red (both azobenzene and red light)

Extracting the binary outputs from Figure 14c for these four conditions gives us the truth table: [ 0,0,1,1 ]. This means PURE-Gray shows no activation (0), PURE-Red also shows no activation (0), while PAZO-Gray activates (1) and PAZO-Red activates (1). To quantify how well this observed behavior matches canonical Boolean logic gates, we compute agreement scores defined as Equation 15:Where S gate is the agreement score for a gate type, it ranges from 0 to 1. N = 4 is the number of truth table rows. O i represents the observed output for state i, while G i is the canonical gate output for that state. The Kronecker delta function, δ, returns one when outputs match and 0 when they differ.

S gate = 1 N ∑ i = 1 N δ O i , G i where δ O i , G i = 1 if O i = G i 0 if O i ≠ G i (15)

Comparing the observed truth table [ 0,0,1,1 ] with the four basic Boolean gates shows:

(AND) [ 0,0,0,1 ]: score = 0.75 (matches 00, 01, and 11; differs at 10).

(OR) [ 0,1,1,1 ]: score = 0.75 (matches 00, 10, 11; differs at 01).

(XOR) [ 0,1,1,0 ]: score = 0.50 (matches only 00 and 10).

(NAND) [ 1,1,1,0 ]: score = 0.25 (matches only 10).

The highest agreement score, 0.75, is shared by AND and OR gates. However, the specific match patterns show clear differences. For example, the output [ 0,0,1,1 ] gives a logical high (1) whenever azobenzene is present, in states 10 and 11. This occurs regardless of wavelength. This behavior is more like AND, where both the presence of azobenzene and any photonic input create the output. The divergence at state 10 (PAZO-Gray) shows that the system outputs 1 while the ideal AND outputs 0. This means the proteinoid network acts like a modified AND gate. Here, the wavelength input (B) is non-selective. Any visible wavelength can activate the PAZO systems. Thus, the red vs. gray distinction matters less than the presence of azobenzene. This behavior is better interpreted as a conditional gate where Output = f ( A , B ) computes azobenzene present AND light present, with the system demonstrating wavelength promiscuity rather than strict red-selectivity. The logic signature visualization (Figure 14d, left panel) shows observed outputs (blue trace with circles) plotted against four input states (00, 01, 10, 11). It includes canonical gate truth tables: AND in brown, OR in orange, XOR in pink, and NAND in purple. The curve closely follows the AND/OR paths. It stays at 0 for states without azobenzene (00, 01) and jumps to one when azobenzene is present (10, 11).

The rise of Boolean logic in azobenzene-functionalized proteinoid networks marks an important step in molecular computation. This self-assembled system does not use electronic parts or biological circuits. Instead, it performs digital logic tasks through wavelength-dependent photoisomerization. The mechanism works like this: azobenzene chromophores are molecular switches. They turn light into electrical signals, specifically changes in membrane voltage. The presence or absence of these chromophores acts as a control signal to enable or disable this process. This architecture directly parallels transistor-based logic, where gate voltage controls whether source-drain current flows in response to input signals. In the proteinoid system, azobenzene acts as the “gate.” Light is the “source,” and voltage output is the “drain.” This sets up the logical function: Output = Azobenzene AND Light. The 75% gate agreement score (Figure 14d, right panel) shows that proteinoid networks can reach nearly three-quarters accuracy in digital computation.

The AND-like gate assignment was reproducible across all three independent PAZO synthesis batches: the observed truth table [ 0,0,1,1 ] was recovered in all three replicates, yielding a mean AND agreement score of 0.75 ± 0.00 (SD = 0, n = 3 batches), confirming that this logic signature is an intrinsic property of azobenzene incorporation rather than a stochastic outcome of a single preparation. The binary activation classification was robust to threshold parameter variation: varying the threshold multiplier Z from 1.5 to 2.5 (Equation 14) did not change the classification outcome for any of the twelve experimental conditions, demonstrating that the digital logic assignments are insensitive to the precise threshold definition.

This is impressive for a chemical system without any engineered circuits or programmed instructions. The observed shift from ideal behavior—activation at state 10 instead of state 11—likely shows the wavelength flexibility of azobenzene photoisomerization. The trans → cis conversion happens effectively over a wide range of wavelengths (350–600 nm, covering UV to red). This means that gray illumination, whether broadband or neutral density, still provides enough photons to drive isomerization and trigger voltage responses. To achieve stricter wavelength selectivity for an AND gate, we need specific wavelength-chromophore pairs. This means using chromophores with narrower absorption bands. Alternatively, we could use different types of chromophores that absorb at complementary wavelengths. For instance, having both a red-absorbing and a blue-absorbing chromophore can help create wavelength-specific AND gates. The successful demonstration of AND-like logic in proteinoid networks shows that chemical Boolean computation is possible. It opens doors to more complex functions. For example, cascading multiple proteinoid logic elements can form multi-input gates like three-input and four-input AND/OR. Combining different chromophore types may lead to NAND, NOR, or XOR gates through opponent-process interactions. Also, arranging chromophore-functionalized regions alongside pure proteinoid areas can create logic circuits similar to integrated circuits. These chemical computers use light instead of electrical voltage. They process information through changes in molecular shape, not electron flow. They also store information in membrane shapes instead of charge on capacitors. This new approach could offer benefits like better biocompatibility, (Di Martino et al., 2023; Tadge et al., 2023) operation in solutions, and easier connections with biological systems. Azobenzene-functionalized proteinoids can discriminate wavelengths and perform logic operations. This makes them basic photonic processors for chromatic computation. They can break down complex spectral environments into clear decisions based on wavelength combinations. This is similar to how biological visual systems identify colors and objects using chromatic information.

Conclusion

This work introduces azobenzene-functionalized proteinoids as a new class of photoresponsive biomaterials that combine thermal protein self-assembly with photochromic switching. These materials were synthesized by heating phenylalanine-4′-azobenzene hydrochloride with glutamic acid and phenylalanine at 170–180°C. The resulting structures retain both the vesicular morphology of proteinoids and the photoisomerization capability of azobenzene, demonstrating that chromophores survive harsh thermal synthesis and can be directly incorporated without post-processing. UV–visible spectroscopy confirms preserved π–π * transitions at 360 nm and J-aggregate bands near 520 nm, indicating strong thermal stability and successful copolymerization.

Morphologically, the materials form stomatocyte-like vesicles (0.5–10 μm) due to membrane buckling during dehydration. These structures exhibit invaginated cavities and broken symmetry, enabling functional advantages such as encapsulation, directional motion, and photomechanical response. High-resolution imaging suggests layered chromophore stacking, similar to multilamellar lipid systems, which may enhance mechanical stability and controlled cargo release.

A key finding is the emergence of wavelength-dependent electrical responses, representing a form of chemical “vision.” Electrophysiological measurements across six wavelengths show that azobenzene-functionalized proteinoids (PAZO) produce strong, wavelength-specific signals (− 130 to + 150 mV), while pure proteinoids remain inactive. Specific channels exhibit enhancement (positive shifts up to 150 mV under green/yellow light) or suppression (negative responses), creating spatial heterogeneity analogous to retinal photoreceptors. Green light shows strong activation (∼ 63 mV), consistent with the 520 nm absorption band, confirming that resonant chromophore excitation drives the response.

Electrochemical analysis reveals that azobenzene incorporation significantly improves electrical properties. Impedance decreases (up to ∼ 50 %), while cyclic voltammetry shows dramatic increases in current and charge storage capacity. This indicates a transition from inert materials to mixed ionic–electronic conductors, where azobenzene enables electron transfer and the proteinoid matrix supports ionic transport. Together, these properties enable efficient conversion of light into electrical signals.

Importantly, these systems exhibit wavelength-dependent binary computation. Using a threshold-based classification, PAZO proteinoids show activation across wavelengths, while pure proteinoids remain inactive. Mapping conditions to a Boolean framework (azobenzene presence and light exposure) produces a truth table [ 0,0,1,1 ], corresponding to AND gate behavior (75% agreement). This demonstrates that molecular photoisomerization can generate digital logic outputs, providing a basis for chemical computation.

This assignment was reproduced across all three independent synthesis batches and was robust to threshold parameter variation (Z = 1.5–2.5), confirming that the logic signature is an intrinsic material property.

These findings position azobenzene-functionalized proteinoids at the intersection of prebiotic chemistry, smart materials, and unconventional computing. Their spontaneous emergence of wavelength-selective responses suggests that primitive sensory functions could arise from simple chemical systems. Unlike conventional photoresponsive polymers, these materials are biodegradable, easily synthesized, thermally robust, and self-assembling.

Their properties enable applications in light-controlled drug delivery, nanomotors, biosensors, and adaptive materials. Furthermore, their computation paradigm—operating in aqueous environments, using light as input, and leveraging spatial organization—offers advantages for biocompatible and parallel information processing systems.

Future work should expand chromophore diversity, enable spatial patterning for defined logic circuits, explore dynamic photomechanical behavior, integrate sensing with chemical outputs, and scale systems into functional materials such as films or hydrogels.

In conclusion, azobenzene-functionalized proteinoids demonstrate that complex sensory and computational behaviors can emerge from simple thermally synthesized chemical systems. They provide insight into early photosensory evolution while offering a versatile platform for next-generation photoresponsive materials and molecular computing technologies.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Footnotes

Edited by: David Alsteens, Université Catholique de Louvain, Belgium

Reviewed by: Javid Malla, King’s College London, United Kingdom

Yunlong Song, Tongji University, China