Electron Hopping Across Hemin‐Doped Serum Albumin Mats on Centimeter‐Length Scales

Exploring long‐range electron transport across protein assemblies is a central interest in both the fundamental research of biological processes and the emerging field of bioelectronics. This work examines the use of serum‐albumin‐based freestanding mats as macroscopic electron mediators in bioelectronic devices. In particular, this study focuses on how doping the protein mat with hemin improves charge‐transport. It is demonstrated that doping can increase conductivity 40‐fold via electron hopping between adjacent hemin molecules, resulting in the highest measured conductance for a protein‐based material yet reported, and transport over centimeter length scales. The use of distance‐dependent AC impedance and DC current–voltage measurements allows the contribution from electron hopping between adjacent hemin molecules to be isolated. Because the hemin‐doped serum albumin mats have both biocompatibility and fabrication simplicity, they should be applicable to a range of bioelectronic devices of varying sizes, configurations, and applications.

the structural protein together with the cytochromes. However, restrictions for genetically expressing and purifying these pro teins in sufficiently large quantities have prevented the oppor tunity of integrating these proteins in bioelectronic devices.
In this work, we present a new type of freestanding bio logicalmoleculebased hybrid material that serves as an effec tive electron conductor over centimeterlength scales with enhanced ET relative to the bacterial pilus. We recently dem onstrated that electrospun bovine serum albumin (BSA) mats are efficient proton conductors. [20] One of the major advantages in using BSA mats is the ability of BSA to noncovalently bind a variety of small molecules. [21] It was previously shown that binding small molecules to proteins (such as BSA) affects their ability to mediate electrons in solidstate device structures, [22][23][24] and that tethering ferrocene to redox enzymes can result in an electron mediating relay between the enzyme redox site and the electrode in an electrochemical setup. [25][26][27] In analogy with solidstate device nomenclature, this is usually termed a "doping" process, with the selected molecules as "dopants"; one might equally characterize this process as the formation of a hybrid or composite material. Herein, we use the term "molec ular doping" for simplicity of acknowledging this fact.
Hemin (an Fe containing porphyrin as shown in Figure 1) was used as the dopant in this work because it is one of the main electron mediators in nature and is also capable, as already noted, of supporting long range ET in the bacterial pilus. [11,12] Also, it was shown that the electrochemical proper ties of organic and inorganic films can be improved by mixing them with hemecontaining proteins. [28][29][30] Figure 1 shows how simply placing a BSA mat in hemin solution for 24 h results in a robust hemintoBSA binding and a resultant colour change from white to black. UV/Vis absorption was used to quantify the molecular doping of the mat by measuring the absorption of the solution before and after immersing the mat ( Figure S1a, Supporting Information). We found an average of 3.93 ± 0.61 nmole of hemin (or equivalently Fe, as there is one Fe atom per hemin) for 1 mm 3 of BSA mat, which corresponds to 2.37 × 10 18 atoms of Fe per cm 3 of mat. The absorption spectrum of the hemindoped mats (Figure S1b, Supporting Information) confirms that the oxidation state of the hemins within the mat corresponds to the one in solution. Importantly, while placing the hemin doped mats in a fresh aqueous solu tion we could not observe any hemin leaching for at least 24 h, indicating that although the BSA is in its unfolded state in the mat, it retains its ability to strongly bind hemin. We used Xray photoelectron spectroscopy (XPS) to further study the effect of hemin doping on the BSA electronic structure, and specifi cally the valence band structure of the protein mat ( Figure S2, Supporting Information). It can be seen that in the nondoped mat there are no electronic states present in the gap, above the valence band maximum up to 0 eV binding energy, whereas upon doping we find additional electronic states present in the gap. This might be expected to lead to enhanced conduction, with at least a greatly reduced (or no) energy gap between occu pied electron transport levels and the empty energy levels. The XPS survey scans also confirm the presence of Fe in the mats, and enabled us to quantify the amount of remaining salt in the mats upon doping ( Figure S3 of the Supporting Information and discussion within).
Recently, we showed that (nondoped) hydrated BSA mats can conduct via proton transfer over millimeter distances with associated conductivity of σ Η ≈ 50 µS cm −1 . [20] The large proton conductivity value was attributed to the abun dance of oxoaminoacids in the mat, aided by a waterbased hydrogenbonded network. We also concluded that bulk proton conduction was the dominant process contributing to the high frequency AC electrochemical impedance spectroscopy (EIS) and DC current-voltage (I-V) responses, which exhibited the same distance and temperature dependences. Here we inves tigate the effect of hemin doping on inplane mat conductivity, across larger distances than in our previous study but using an otherwise identical geometry (see Figure S4, Supporting Information). In a similar fashion to the case of the nondoped hydrated mats [20] here also the measurements were conducted in the waterswollen state, i.e., at ≈150 wt% hydration. Fre quencydependent AC EIS measurements are highly informa tive as they enable deconvolution of charge transport processes having different time constants. While the highfrequency (≳1 kHz) AC domain is mainly governed by charge carrier bulk transport, the lowfrequency (≲100 Hz) domain is governed by traprelated carrier capture and release and/or by carrier accu mulation next to the contacts due to contact resistances. [31,32] In DC I-V measurements, on the other hand, it is hard to distin guish between these processes.
Unlike our previous study where we used only symmetric Au finger array electrodes, [20] in this study we use both Au and Ti electrodes, since (as will be discussed below) the chloride ion of the hemin (Figure 1) can electronically interact with the Au electrode. As shown by Bode (absolute impedance as a func tion of frequency) and Nyquist (imaginary part of the imped ance as a function of the real part) plots of the EIS measure ments (Figure 2; Figure S5 of the Supporting Information, respectively), the impedance at highfrequency of the doped mat is around 35-45fold lower than that of the nondoped mats (with the same measurement geometry) for all of the measured electrodeseparation distances (Table S1, Supporting Informa tion). [33] This leads to fitted conductivity values of ≈2 mS cm −1 for the doped mats. Accordingly, while the resistance of the nondoped mats allowed us to measure only over distances up to 2.5 mm, the lower resistance of the doped mats allowed us to increase the electrode separation distance up to 24 mm (Figure 2c,d). Importantly, the measured highfrequency EIS response across the mats was very similar between the two electrode materials ( Figure S6  We conclude that doping the mats with hemin significantly increases the conductance across the mat, which is not expected to be due to an increase in proton conductivity (vide infra) but rather due to electron transport mediated by the hemin moi eties. As observed by the XPS results ( Figure S3, Supporting Information), the doped BSA mats do contain some sodium, which can be attributed to the presence of ≈30 × 10 −3 m NaCl in the doping solution. Hence, to support our electron transport hypothesis, and to exclude the possible contribution of ionic conduction of higher mass ions to the measured high conduct ance across the doped mats, we measured the EIS conductance of nondoped mats that were placed in 50 × 10 −3 m and 0.5 m NaCl solution ( Figure S7, Supporting Information). As shown in Figure S7 of the Supporting Information, placing the non doped mats in 50 × 10 −3 m NaCl solution, which is of slightly higher concentration that what was used in the doping solu tion, resulted in only 20%-40% increase (1.2-1.4fold) in con ductance for the different measured distances. Furthermore, placing the nondoped mats in 0.5 m NaCl solution induced only a further minor increase in conductance in the order of 50-70% (1.5-1.7fold) compared to the nondoped mats in deionized water. These results strongly support our conclusion that the hemin molecules induce the high conductance of the mat. We also measured the EIS across a thin film of only hemin mole cules that was formed by dropcasting a similar hemin solution that was used to dope the mats ( Figure S8, Supporting Informa tion). The measured conductance across the dropcast hemin film was comparable to the one measured for the hemindoped BSA mats, and importantly, the complex shape of the Nyquist plot for the hemindoped BSA mats is also present in the drop cast hemin films.
In order to complement the highfrequency EIS measure ments, we compared the DC I-V response of the mats before and after doping (Figure 3). It can be clearly seen that meas ured currents across the nondoped mat are significantly lower than the ones across the doped mats for both the Au and Ti electrodes (Figures S9 and S10 of the Supporting Information compare, pairwise, the I-V curves of the nondoped and doped mats for electrode separations in the range 0.25-2.5 mm for the Au and Ti electrodes, respectively). However, while the non doped mats have monotonically varying I-V curves measured with both electrodes, the doped mat exhibited peaks at around ±0.4 V only when measured with the Au electrodes, and a monotonically varying I-V curves when measured with the Ti electrodes. The presence of the peaks in the forward and reverse sweeps ( Figure S11, Supporting Information), together with the change of the peaks' locations as a function of the distance between the electrodes (Figure S12 of the Supporting Informa tion and text within), leads us to speculate that the observed redox peaks are due to an oxidation and reduction electro chemical process involving the hemin, and specifically the one between the chloride ion of hemin to the Au electrode (since the peaks were observed only by using the Au electrodes). [34][35][36] Nevertheless, it is important to note that the measured DC con ductance (current at low bias) for the doped BSA mats is very similar between the Au and Ti electrodes (Figure 4; Figure S13, Supporting Information), suggesting a similar physical origin. We also followed the DC I-V response of the dropcast hemin films ( Figure S14  speculation that the peaks can be associated to the hemin mole cule, but more importantly, their different locations imply that the hemin is indeed tightly bound to the BSA scaffold, which changed its electrochemicalassociated potential compared to free hemins in the dropcast configuration. The differences between the EIS and I-V results for the non doped and doped mats can be explained by the differences in their charge transport processes. For the nondoped mats, the charge carriers are predominantly protons. The highfrequency AC EIS measurement is then dominated by bulk proton con duction via hopping through a water mediated network of pro tein oxoaminoacids. [20] The nondoped mat DC I-V data have the same (bulk proton conduction) origin, with no evidence for deep traps and a DC contact resistance (<10 6 Ohm) that is negligible in comparison to the bulk resistance (>10 8 Ohm) ( Figure S13a,b, Supporting Information). The similar dis tance, L, dependence of the AC and DC conductance, G, which approximately follows Pouillet's law (G ∝ L −1 ) supports this per spective (Figure 4, empty and filled circles, respectively). The best fit in both cases gave G ∝ L −x where 0.86 < x < 0.91; the deviation in the exponent from 1 being most probably due to the effects of a finite contact resistance and/or the fractal nature of diffusion across the fibrillar network in the mat.
For the hemin doped mats, charge transport is more com plicated. Just as for the nondoped mats, the doped mats are good proton conductors. The two oxoacids per hemin molecule ( Figure 1) might contribute to this proton conductivity but in comparison to the amount of oxoaminoacids in the protein, and especially compared to the amount of water in the mat, we expect this contribution to be very low, and moreover, due to the bulkiness of the hemin moieties we would expect even a reduction in oxoacids volume density. The large decrease in the measured highfrequency EIS impedance must then be due to the existence of an additional charge transport process, namely ET mediated by the hemin molecules. Metalloporphyrines and metallophthalocyanines are frequently used for the formation of electron conducting molecular semiconductor networks. [37] The conduction through such molecular networks could be considered as akin to band conduction, which is due to the for mation of an ordered crystalline network of the cyclic molecule. By contrast, in our study we "dope" an insulating layer (the BSA mat) with hemin, which binds strongly to the aminoacid chain of the protein. Hence, we believe we cannot describe the observed ET conduction via a similar bandlike process. Accord ingly, we will describe the ET conduction as an intermolecular hopping conduction mechanism (vide infra). In addition to the ET process across the mat, the doped mat also exhibits a sig nificant charge injection (contact resistance) barrier at the inter face between the electrodes and the BSA mat ( Figure S13c,d, Supporting Information). Since the electronic resistance across the doped BSA mats is significantly lower in comparison to the nondoped mats, the contact resistance (≈3 × 10 6 Ohm) domi nantes the measured DC current, especially for the short dis tance regime (Figure 4 right panels, filled squares, and insets of Figure S13c,d, Supporting Information). At longer distances, the DC current displays a shallow distance dependency due to the higher resistance of the mat as the distance increases. Accordingly, unlike the highfrequency EIS measurements of the doped mats that follow Pouillet's law and exhibit a power law decay (Figure 4, empty squares), the DC measurements of the doped mats do not follow bulk ET across the doped mats, i.e., do not exhibit a power law decay as a function of distance ( Figure 4, filled squares)  electrode, which is an electrochemical reaction most probably between the Au electrode and the chloride ion of the hemin molecule. This type of interface ET should not be related to the distance between electrodes. Nevertheless, since the lowbias DC resistance is very similar between the samples measured with the Au electrodes to the ones measured with the Ti elec trodes ( Figure S13c,d, Supporting Information) we can con clude that the same type of contact resistance determines the resistance in both cases.
In short, in addition to the proton conductivity measured across nondoped mats, doped mats exhibit both bulk electron conductivity and a highresistance interface electron injection. They also yield different temperature dependences ( Figure S15, Supporting Information). In our previous study on nondoped mats we demonstrated that both the AC EIS and DC conduc tivities are thermally activated, with an identical activation energy ≈0.3 eV, as expected for a single proton hopping pro cess. We find, however, for the doped mats that the activation energy for the AC (EISmeasured) conductance is 0.2 ± 0.04 eV (inset to Figure S15a, Supporting Information), while that for the DC (I-V measured) conductance is 0.15 ± 0.02 eV (inset to Figure S15b, Supporting Information), both of which are dif ferent from that for the nondoped mats.
Given our AC and DC findings above, the bulk electron con ductivity of the doped mats can be described using the same model proposed for electron hopping between the decaheme cytochromes along the bacterial pilus. [38,39] The model is based on the nonadiabatic electron transfer rate (k ET ) equation, also known as the Marcus-Hush formalism: where λ is the reorganization energy, k B the Boltzmann con stant, T the absolute temperature, H DA the electronic coupling matrix between adjacent (donor and acceptor) molecules, and ∆G 0 the energetic driving force. Polizzi et al. [38] thereby derived  where 10 13 is the maximum charge transfer rate (for direct contact between hemes), β the distance decay constant, r nn the intermolecular separation, and r 0 the van der Waals contact dis tance (0.35 nm).
The conductance of a sample can be described by: where A is the crosssectional area between electrodes sepa rated by distance L, N the charge carrier density, e the electron charge, and µ the charge carrier mobility. The latter can be expressed, via the Einstein relation, as: where the diffusion constant: Taken together, G can then be expressed as: Using this model and our EISmeasured conductance values (which follows bulk ET as discussed above), the values for λ, r nn , and k hop can be estimated for the doped mat. Due to the small deviation of our measured bulk conductance from Pouillet's law (G ∝ L −1 ) we have used our measured values (G ∝ L −0.79 and G ∝ L −0.85 for Au and Ti electrodes, respec tively, Figure 4) in Equation (6) to yield a better fit to the results. The driving force can be expressed as ∆G 0 = ∆G total /N, where ∆G total is the driving force between the electrodes and N is the number of sites. Since a low bias (0.05 V = ∆G total ) was distrib uted along the numerous hemin sites located between the elec trodes, ∆G 0 is very close to 0 eV. We adopted the β = 1.4 Å −1 value calculated by Breur et al. for the distance decay between two heme molecules [40] and took a charge density value of N = 2.37 × 10 18 per cm 3 (i.e., the Fe atom density obtained from UV-Vis measurements, see above). We considered a range of reorganization energies, λ = 0.7-1.1 eV, as found in the bacte rial decaheme cytochrome. [40] For each value of λ we calculated the required distance between two adjacent hemin molecules (r nn ) that would match our measured conductance, yield a rea sonable hopping rate (k hop ), and be in line with theoretically calculated interheme ET rates (see Figure 2 in ref. [40]). Using these constraints, the best fit (R 2 ≥ 0.98) was found for a r nn value of 5.5 Å and λ values of 0.84 and 0.75 eV for the meas urements with Au and Ti electrodes, respectively (Table S2). The resulting k hop between adjacent hemin molecules then lies between 1.9 × 10 9 s −1 (for λ = 0.84 eV) and 4.3 × 10 9 s −1 (for λ = 0.75 eV). In relation to the theoretical study of Breuer et al., [40] our calculated interhemin distances and ET rates sug gest that the average spatial orientation of the hemin porphyrin ring planes is either Tshaped or coplanar (rather than cofacial). We can use the obtained values for λ to calculate the activa tion energy of the process These values correlate well to our measured value for the EIS measurements (≈0.2 eV, Figure S15 of the Supporting Informa tion), which further implies that the rate limiting step in the EIS measurements is indeed bulk ET, and not interface ET. It is important to note that since the BSA mats are very complex structures, the different binding sites for hemin in the structure will result in different energetic properties of the hemin (redox potential) and accordingly, the λ, ∆G 0 , and H DA will be different between different pairs of adjacent hemin molecules. The values extracted here are set accordingly to be an average for all the different binding sites. Nevertheless, while taking into account the possible wide distribution in the redox potential of the hemin molecules within the BSA mat, in comparison to the Fermi energy of the electrodes, the energy alignment does not restrict the suggested ET mechanism and allows the injection of electrons to the material from the electrodes ( Figure S16 of the Supporting Information and text within).
In summary, we have presented a study of the transport properties of hemindoped BSA mats that show favourable electron conduction over centimeter distances, hence rep resenting the first of its kind proteinbased conductive free standing material that is not being dissolved in an aqueous solution. Furthermore, unlike common conductive organic polymers, where the dopant (usually acids) frequently leaches out of the polymer while placing it in an aqueous solution, our chosen dopant (hemin) has a large affinity to the BSA scaffold and remains within the material in an aqueous solution for months. The resulting conductivity ≈2 mS cm −1 is much higher than measured for bacterial pili in a macroscopic configuration (5 µS cm −1 ), [41] which until now was the only reported protein based long range electron conductor. ET in the doped mat is proposed to have a similar origin to one of the mechanisms suggested for the bacterial pili, [38] namely charge carrier hop ping between hemin molecules. However, unlike the evolution optimized system of the bacterial pili, our proteinbased system is completely artificial. Although bacterial pili have been sug gested for integration within bioelectronic devices, protein expression of a "clean" system is a very difficult and costly task. By contrast, hemindoped BSA mats can be produced much more cheaply, over large areas with significant quantity and in a range of configurations and geometries. Finally, considering both their biocompatibility and ability to sustain cell growth, [42] hemindoped BSA mats appear to represent an excellent ready touse platform for bioelectronic devices.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author. The raw data supporting this paper are available upon request from m.stevens@imperial.ac.uk.