Room-temperature tunnel magnetoresistance across biomolecular tunnel junctions based on ferritin

Charge transport through self-assembled monolayer (SAM)-based tunnel junctions generally occurs through quantum mechanical tunneling, and enables novel electronic function potentially complementary to devices based on semiconductors [1–3]. For applications in spintronics, it is fundamental to efficiently inject spin-polarized carriers and transport them across the molecules inside the junctions. Achieving such goal is, however, highly challenging owing to the presence of structural imperfections and defects at the ferromagnetic (FM)/molecule interfaces (e.g. step edges, pinholes, or grain boundaries, or oxide formation) [4–6], impedance mismatch [7], or due to the complexity in controlling the nanoscale interface and the energy level alignment of the molecular junction [7]. Consequently, examples of magneto-transport across (bio)molecular tunneling junctions are rare, while the obtained performances have remained very modest to date [8, 9]. Here, we succeed in fabricating biomolecular tunnel junctions that incorporate a SAM of highly symmetrical, thermally stable (up to 80 °C) protein cages derived from a two alanine-modified Archaeoglobus fulgidus ferritin (AfFtn-AA) [10] loaded with ferrihydrite. The ferritins are anchored on ultra-smooth and oxide-free FM surfaces with EGaIn top-contacts (see figure 1 below). The resulting ferritin-based tunnel junctions display a remarkable 60% tunnel magnetoresistance (TMR) at 200 K (25% TMR at room temperature). We attribute this good TMR performance to a new mechanism of spin polarized charge transport that involves long-range tunneling across ferritin, possibly assisted by magnons excited in the ferrihydrite nanoparticles (NPs), while the protein cage efficiently decouples the magnetic ferrihydrite core from the FM bottom electrode, providing a clean tunneling barrier eliminating the need for impedance matching.

Zoom In Zoom Out Reset image size

Spin transport phenomena across MTJs are studied in several types of device geometries. Usually, two FM electrodes are separated by a thin insulating layer, which serves as the tunneling barrier. A SAM provides a well-established means to chemically and physically modify a metal surface by lowering the energy barrier for charge injection as well as reducing the role of impedance matching [4, 14, 15]. Electrons tunnel across the SAM if a bias voltage is applied between the two FM electrodes. The resistance R of the junction depends on the magnetization direction of the two FM electrodes, which can be tailored by an external magnetic field, resulting in the TMR signal. The resistance of the junction is in principle smaller when both electrodes have parallel magnetization, R_P, rather than antiparallel magnetization R_AP. The magnitude of the TMR is defined by equation (1) [6]:

$$\begin{equation}{\text{TMR}} \equiv \frac{{\left( {{R_{{\text{AP}}}} - {R_{\text{P}}}} \right)}}{{{R_{{\text{AP}}}}}} \times 100\% .\end{equation} \tag{ 1 }$$

Typical TMR values in this class of MTJs lie within the 1%–10% range at room temperature [9].

TMR can also be obtained in spin-filter tunnel junctions [16], where a FM insulator is inserted as a spin-filtering tunneling barrier in between a FM and non-magnetic electrode. The spin-filter effect is an interlayer-controlled process, with the ability to selectively allowing for the transfer of one type of spin (up or down) through the interlayer. The performance of TMR depends on interface roughness [7], surface oxidation of the FM electrode [17], and electron spin phase coherence across the junction [18], amongst other factors [19–22].

Although TMR is well-known in junctions with insulating tunnel barriers sandwiched between two FM electrodes, studies of junctions with a spin-filter tunnel barrier are scarce, with generally poor performances at room temperature (typically few percent) [8]. Experiments have shown that biomolecules [8, 23], single-molecule magnets [1, 24, 25], and magnetic NPs [26, 27], can act as spin filter when incorporated in the tunneling barrier. The shape, size, and stability of the magnetic NPs play an important role in their magnetic properties and, consequently, in their performance. To date, values of TMR of 60%–300% have been reported at low temperatures (<10 K) for tunnel junctions incorporating these nanoscale magnetic systems, but performances usually decrease very fast with increasing temperature [27, 28]. For instance, Naaman et al recently reported magnetoresistance of only 2% at room temperature across biomolecular MTJs of the form Au/linker/biomolecule/MgO/Ni with bacteriorhodopsin [23]. Alternatively, it has also been recently shown that chiral biomolecules such as DNA may mediate spin-filtering based on 'chiral-induced spin selectivity', with spin filtering of up to 60% [29], although this requires ultrahigh-vacuum conditions and optical detection, whereas the transfer mechanism at play remains obscure. Compared against traditional MTJs with applications in, for instance, hard disk drive read-heads, magnetic random access memory, and sensors which typically consist of 8–64 layers with 70%–604% of magnetoresistance at room temperature [30–32], the advantage of our devices is in their relatively simple structure consisting of only three layers (the two electrodes and the biomolecular layer) with a reasonable TMR of 25% at room temperature. Some distinctive features of ferritin are its highly symmetrical structure with innate capability to store and release iron in a controlled fashion. The hollow globular cage-like structure (with an outer diameter of 12 nm and an inner diameter of 8 nm) is formed by the self-assembly of 24 protein subunits. Iron is stored in the form of ferrihydrite NPs inside the protein cage. Therefore, the ferrihydrite NPs are separated from each other by at least 4 nm formed by the 2 nm thick protein shell of two adjacent ferritins. The magnetic properties of ferritins have been widely studied in powder form and depending on the iron oxide loading conditions, iron oxide loading, and types of ferritin, they exhibit a rich variety of controllable magnetic properties including superparamagnetism [33, 34], antiferromagnetism [35, 36], or ferrimagnetism [33, 37], and have manifested one of the first observations of quantum tunneling of the magnetic moment [38]. We recently reported that tunneling across ferritin is highly efficient as evidenced by a low tunneling decay coefficient of 1.30 ± 0.02 nm⁻¹ [13]. For these reasons, we studied the magnetic properties of ferritin incorporated in large-area molecular junctions of the form of FM-linker-ferritin//GaO_x /EGaIn junctions, where the FM electrode is ultra-flat template-stripped nickel (Ni^TS). In this work, we replaced the Au^TS bottom electrode in our previous work [38] with a FM Ni^TS electrode. This replacement results in a ferritin-based junction that shows TMR characteristics as a consequence of magnetic polarization of the conduction (or tunneling) electrons. The detailed monolayer characterization of ferritin and fabrication of the MTJ are described in Section 4 and SI (figures S1–3, and table S1 (available online at stacks.iop.org/JPMATER/4/035003/mmedia). In this study, the ferrihydrite NPs were formed inside the ferritin under anaerobic (An1200Fe AfFtn-AA) which is more magnetic than that prepared under aerobic (A1200Fe AfFtn-AA) conditions. The loading of iron ions per ferritin cage is indicated by 1200Fe (i.e. we loaded 1200 iron ions per cage, determined with inductively coupled plasma atomic emission spectroscopy as reported before) [39]. The fabrication of the Ni^TS and the GaO_x /EGaIn top electrodes (EGaIn stands for eutectic gallium–indium; 75.5% Ga and 24.5% In by weight) have been reported elsewhere (see Section 4 for details) [4]. We note that the ∼0.7 nm thin layer of GaO_x is highly conductive (resistivity ∼2.9 × 10⁻² Ω cm²) [40], and has a metal-like behavior [41], which prevents the bulk EGaIn from alloying with the bottom electrode. The TMR response of the junction was measured as a function of an in-plane magnetic field to control the magnetic polarization of both magnetic ferrihydrate NPs inside the ferritin and the Ni^TS electrode. Figure 1 displays the typical energy landscape in a TMJ resulting in TMR, where the two magnetic layers align parallel (low R) or antiparallel (high R) to each other. Before describing the TMR measurements of the junctions, it is instructive to discuss the characterization of the junctions in detail.

First, we characterized the magnetic properties of An1200Fe AfFtn-AA and A1200Fe AfFtn-AA in freeze-dried powder form by using a superconducting quantum interference device (SQUID). Figure 2(A) shows the room temperature magnetic hysteresis loops, which saturate at an applied magnetic field of around 0.89 and 1.0 T for AnFe1200 AfFtn-AA and AFe1200 AfFtn-AA, respectively (table 1). Figure S4 shows the room temperature magnetic hysteresis loops for apo-AfFtn-AA, which exhibit diamagnetic characteristics. The corresponding magnetic moment from the saturation of the magnetic moment is estimated to be ∼2500 µ_B and ∼481 µ_B per ferritin cage for An1200Fe AfFtn-AA and A1200Fe AfFtn-AA, respectively. These values are larger than those of ferrihydrite NPs inside ferritin (260–500 µ_B) reported elsewhere [37, 43–46], but smaller than those of NPs of pure magnetite or maghemite with diameters of 5–11 nm, which reach up to 21 600 µ_B [47]. Thus, the measured magnetic moments in our experiments are within the range of previously reported values, whereas ferritin loaded under anaerobic conditions has a value of µ_B of about one order of magnitude higher than that of ferritin loaded in aerobic conditions.

Zoom In Zoom Out Reset image size

It is important to observe that ferritins in powder form do not exhibit coercivity at 300 K (figure 2(A)), indicating superparamagnetic behavior. Such a superparamagnetic behavior is observed down to ∼10 K (see below and figure S5). For lower temperatures, the magnetic moment of the particles is pinned to the anisotropy axis resulting in small coercivity and hysteretic magnetization dynamics (see open hysteresis loops obtained at T < 10 K in figure S5). The observed slow saturation of the magnetization curve in the magnetic hysteresis loops of the An1200Fe sample suggest that this sample contains a larger variation of the Fe ion content inside the ferritins cages as compared to that of the A1200Fe which shows a faster and smoother approach to saturation. This slow saturation is usually related to the nanometer size of the ferrihydrite NPs inside the ferritin cage (which are a mixture of magnetite and maghemite, see figure 3 below). At 2 K, we derive the saturated magnetization values of A1200Fe AfFtn-AA of ∼33 emu g⁻¹ and ∼44 emu g⁻¹ for An1200Fe AfFtn-AA, which are lower than the value of 71 emu g⁻¹ corresponding to pure FM Fe₃O₄ NPs of ∼6.0 nm [48]. Thus, it is likely that the mineralization of iron oxide inside the ferritin at anaerobic conditions induces the formation of FM NPs, as previously observed in other types of ferritin (e.g. human, horse spleen, or bacteria) [49, 50]. The ferritin in this study is derived from the extremothermophilic archaeon Archaeoglobus fulgidus (AfFtn) and exists in the form of dimers which self-assemble in the presence of divalent metal ions (Fe^II ions) to form a 24-mer cage structure. The iron core radius of AfFtn (4.26 nm) is larger than that of bacterial ferritin (3.96 nm) and human ferritin (3.95 nm) [51]. An engineered version (K150 A, L151 A) of the archeal ferritin (AfFtn-AA) [10] with a closed pore has been used in this study (note, the Fe ions are mineralized during the self-assembly process). In addition, we carried-out the iron oxide mineralization under a controlled environment at elevated temperature (65 °C) under anaerobic conditions which results in a more magnetic form of the iron oxide mineral (magnetite) than under aerobic conditions at room temperature [52]. This difference in preparation methods results in the higher magnetic moment as observed in this study than those for commonly used ferritin. The presence of FM NPs in the An1200Fe AfFtn-AA is confirmed by detailed SQUID and x-ray magnetic circular dichroism (XMCD) characterization measurements (figures 2 and 3).

Zoom In Zoom Out Reset image size

Figure 2(B) shows the temperature dependence of the magnetization measured upon FC and ZFC under an applied field of 1000 Oe. The curves show typical superparamagnetic behavior with a well-defined blocking temperature (T_B) for An1200Fe AfFtn-AA and A1200Fe AfFtn-AA of ∼10 K. Upon further cooling the sample below the T_B in the FC configuration, the magnetization curve increases considerably for both instead of the usual leveling off [53] which further confirms that the ferritin core consists of superferromagnetic NPs. Here, long-range magnetic interactions occuring between the NPs inside the ferritin core can explain our observations, similar to that of Curie-Weiss law (M_FC ∝ (T-T_B)⁻¹) behavior [54] reported previously for an ultra-small iron oxide NP system with a diameter of less than 3 nm and a dense frozen ferrofluid containing Fe₅₅Co₄₅ particles of 4.6 nm. The data show a small bifurcation between the ZFC and FC curves for An1200Fe AfFtn-AA (figure S6) between 50 K and 250 K, which indicates a deviation from superparamagnetic behavior, most likely caused by the presence of a FM component (i.e. the magnetite phase in the ferrihydrite NPs as discussed below). The observed values of T_B are similar to previously reported values for ferritins which fall in the range of 5 K to 25 K [43, 46]. Apo-AfFtn-AA shows a deviation from the diamagnetic behavior down to 10 K. Despite the absence of additional Fe ions, trace amount of Apo-AfFtn-AA is present in the preparation. However, the trace amount is not expected to contribute to the observed deviation from diamagnetic behavior, indicating that the superparamagnetic behavior of these proteins is exclusively caused by the ferrihydrite NPs inside the ferritin core. From these observations, we conclude that the ferrihydrite NP inside one ferritin does not magnetically interact with the neighboring ferritins.

To characterize the magnetic properties of the SAMs of ferritin, we analyzed the magnetic behavior of A1200Fe AfFtn-AA SAM on template-stripped gold electrodes (Au^TS) with magneto-optical Kerr effect (MOKE) and that of An1200Fe AfFtn-AA with SQUID (see Section 4 for details). Figure 2(C) shows the full hysteresis curves which were recorded with an in-plane magnetic field up to 1.50 T. The curves saturate around 0.89 T. These results suggest that the magnetic behavior of the ferritin particles remain unaltered when self-assembled on the surface. The in-plane magnetic hysteresis loop of the Ni^TS substrate measured with SQUID is also shown in figure 2(C), displaying a much sharper behavior with field, reaching room temperature magnetic saturation at H = 0.1 T. To determine the magnetic properties of the ferritin adsorbed on Au in more detail, we measured the ZFC/FC curves of An1200Fe AfFtn-AA monolayers (the A1200Fe AfFtn-AA monolayers gave too weak signals to be measured with this method). Figure 2(D) shows a similar blocking behavior with a departure between the ZFC and FC curves below 10 K. The increase of both ZFC and FC curve for higher temperatures, all the way into room temperature may indicate that the magnetic NPs may indeed be weakly interacting and depart from a pure superparamagnetic behavior. Gandhi et al [55] reported similar observations for weakly interacting Fe₃O₄ NPs and Cr₂O₃ nanorods, where the shape of the NPs resulted in large coercive fields. In our case, we could not observe the high coercive field from the magnetization measurements due to the minimal contribution of the FM core of the ferritin relative to those from the other diamagnetic contributions from the ferritin shell and substrate. We also tried to measure the magnetization hysteresis loop of the complete device as a function of the applied magnetic field, unfortunately, the measured total magnetization of the complete device is completely dominated by the Ni FM bottom electrode because the amount of magnetic ferritin core (weight) in the MTJs is much smaller than that of the 200 nm Ni FM bottom electrode. It is possible that this behavior results from some particles that have changed shape after deposition (distortion), inducing some hysteretic behavior. From prior study [13] we know that ferritins adsorbed on Au with loading of Fe1200 has only a height of 5.2 ± 0.2 nm determined from atomic force microscopy (AFM) which is considerably smaller than that expected for fully extended, spherical ferritins which have an outer diameter of 12 nm. Pen et al [56] reported that the there are multiple nucleation sites for the Fe ions inside the cavity resulting in small NPs of 1.8–2.7 nm in diameter, and Jian et al [57] reported that morphology of the NPs depend on the Fe ion loading inside the equine spleen ferritin. For very high Fe ion loadings, these small NPs can merge into larger NPs with diameters of 2–6 nm [50]. We used relatively low Fe ion loadings in this study, therefore it is reasonable to assume that in our case the ferritins consist of several small NPs distributed inside the cavity.

We performed x-ray absorption spectroscopy (XAS) of An1200Fe AfFtn-AA and A1200Fe AfFtn-AA SAMs on Ni^TS to characterize the structure of the ferrihydrite NPs in more detail. Figures 3(A) and (B) show that the XAS spectra are dominated by two peaks at 708.2 and 709.6 eV along with a shoulder within the L₃ edge and three components within the L₂ edge at 719.4, 721.2 and 722.8 eV. We focus our attention on the L₃ region. The integrated intensity of each peak at the L₃ edge was used to determine the Fe²⁺/Fe³⁺ ratio from which the composition of the ferrihydrite NPs inside the ferritin can be determined as described in [48] (see Section 4 for the details). This ratio is 0.59 for An1200Fe AfFtn-AA monolayers which indicates that the magnetite component dominates. In contrast, for A1200Fe AfFtn-AA monolayer, the ratio is 0.68 which indicates that there is a mixture of γ-Fe₂O₃ (maghemite) and Fe₃O₄ (magnetite) phases present in the ferrihydrite NPs in agreement with previous reports [58]. Thus, we conclude that the ferritin core formed under aerobic conditions forms a mixed iron oxide phase, whereas the ferritin core formed under anaerobic conditions is dominated by the magnetite phase which has better magnetic properties than maghemite in particular, magnetiocrystalline anisotropy and saturation magnetization. These observations are consistent with the observed 5.9 times higher magnetic moment per ferritin cage for An1200Fe AfFtn-AA than for A1200Fe AfFtn-AA (table 1).

To study the magnetic interaction between the Ni^TS and An1200Fe AfFtn-AA, we performed XMCD at room temperature on the monolayers on Ni^TS. We note that usually monolayers give weak signals and usually such measurements are performed in high magnetic fields (>10 T) and at low temperatures (<4 K), but we aimed to characterize the monolayers in close resemblance to the experimental conditions at which the junctions were electrically characterized (unfortunately, the observed XMCD signal at low magnetic fields (⩽1.5 T) at room temperature (300 K) is very weak so we could not obtain magnetic hysteresis loops from the XMCD characterization). Figures 3(C) and (D) show the XAS and XMCD spectra of Fe L_2,3 and Ni L_2,3 edges in the presence of a ±1.5 T field (the results obtained for A1200Fe AfFtn-AA on Ni^TS are given in figure S7). The observed XMCD signal indicates the presence of uncompensated iron spin [59, 60]. The sign of the XMCD signal for Fe L_2,3 and Ni L_2,3 are the same, implying that there is no substantial antiFM coupling between ferritin core and Ni^TS substrate. To further confirm the magnetic interaction between the Ni^TS surface and the monolayer of ferritins, we performed XMCD of the ferritin monolayer on Ni^TS surface without external magnetic field (0 T) at room temperature. Figure S8 shows a negative XMCD signal at 709.43 eV which indicates the presence of uncompensated spins of Fe³⁺ ion, moreover the sign of the XMCD signal is also similar to that of Ni L_2,3 edge (figure S7). The XMCD signal for A1200Fe AfFtn-AA on Ni^TS was weaker than that of An1200Fe AfFtn-AA due to the lower magnetic moment as discussed above. We also carried out XMCD on An1200Fe ferritin adsorbed on a non-magnetic metal surface (Au^TS) to further confirm there are no long-range magnetic interactions between the ferritin and FM surface. Indeed, figure S9 shows the same sign of the XMCD signal as that of ferritin on a Ni surface (figure S8). From these experiments, we conclude that the ferrihydrite NPs inside the ferritins are decoupled from the Ni^TS substrate by the protein shell and that they are spin-polarized (FM). These results confirm the observation mentioned earlier and that the superparamagnetic properties of the ferritin in powder form changes to FM behavior for ferritins adsorbed on the surface.

We studied the electrical characteristics of the Ni^TSlinker-ferritin//GaO_x /EGaIn junctions with monolayers of apo-AfFtn-AA, A1200Fe AfFtn-AA, and An1200Fe AfFtn-AA. We used EGaIn stabilized in a through-hole in a microfluidic network in PDMS as the top-electrode and formed five to six junctions on four to five different samples of ferritin monolayers on Ni^TS. From each junction, 20 different current-density traces were recorded. All the data were analyzed as reported [11, 61] (see Section 4 for a brief description).

Figure 4(A) shows the Gaussian log-average J(V) (<lnJ > _G) curves of all the junctions. We used the Z-test to calculate the 95% confidence intervals as reported before [61]. The observed J(V) behavior for A1200Fe AfFtn-AA and apo-AfFtn-AA is similar to that of the junction Au^TS-linker-A1200Fe AfFtn-AA//GaO_x /EGaIn and Au^TS-linker-apo-AfFtn-AA//GaO_x /EGaIn reported previously [13]. To clarify the possible charge transport mechanism across the junctions, we measured J(V) curves over the range of temperatures of 160 K–300 K. Figure 4(B) shows the Arrhenius plots and that only in junctions with apo-AfFtn-AA manifest a small thermally activated component with an activation energy E_a of 98 ± 4 meV, obtained from the fit to equation (2) (where k_B is the Boltzmann constant and J₀ is a pre-exponential factor) which indicates the charge transfer mechanism driven by a hopping mechanism:

$$\begin{equation}J = {J_0}{e^{ - {E_{\text{a}}}/{k_{\text{B}}}T}}.\end{equation} \tag{ 2 }$$

Zoom In Zoom Out Reset image size

In contrast, junctions with A1200Fe AfFtn-AA and An1200Fe AfFtn-AA display electrical characteristics that are independent of the sample temperature. This temperature independence suggests prevailing 'long range' coherent tunneling mechanism across these junctions. Such type of activation-less long range charge transfer has been already observed in similar junctions but with Au electrodes [13], or for different types of biomolecular junctions [62] and organic thin film [63] based junctions.

We further measured the magnetoresistance (equation (1)) of the ferritin-based junctions as a function of magnetic field (±1.5 T) at constant bias of −1.0 V. Figure 4(C) shows the results obtained at T = 270 K (figures S10–12 show the results for another three junctions to demonstrate reproducibility), with a clear positive TMR effect of ∼10% for junctions with A1200Fe AfFtn-AA and ∼40% for junctions with An1200Fe AfFtn-AA at 270 K. The results are consistent with the higher magnetization of ∼2500 µ_B of An1200Fe AfFtn-AA than that of ∼418 μ_B of A1200Fe AfFtn-AA (table 1). In the junctions with apo-AfFtn-AA and only linker molecules, the resistance is independent of the applied magnetic field. Figure 4(D) shows the TMR response of An1200Fe AfFtn-AA junctions at three different temperatures. As a control, figure S13 shows the J(V) curves recorded from an An1200Fe AfFtn-AA junction with parallel and anti-parallel field orientations (±0.5 T).

To understand the effect of T and V on TMR response of the ferritin-based junctions, TMR of the junctions were measured as a function of T and V. Figure S9 shows bias dependent TMR for a junction with An1200Fe AfFtn-AA at 260 K. TMR increases with V for V > 0.5 V (at lower V values we did not observe TMR) following a power law dependency (see figure S14). This observation is in contrast with conventional MTJs, where the maximum TMR value is close to zero bias and decreases with increasing V because of the presence of defects [64]. Figure 4(D) shows that well-defined resistance states are observed in both forward and backward field sweeps, with TMR increasing with decreasing T, achieving a maximum value of 60 ± 5% at 200 K. The improvement of TMR with decreasing temperature likely stems from the increase of the electrode spin polarizations with lowering temperature [65, 66]. As previously remarked, achieving reliable room temperature operation of biomolecule-based MTJs is truly challenging owing to the difficulties to incorporate molecules inside MTJs without disturbance or suppression of their magnetic properties. Accordingly, the measured TMR performance close to room temperature is remarkably large considering the small size of the ferrihydrite NPs.

2.1. Proposed mechanisms for MR

We have demonstrated that the TMR characteristics of magnetic tunneling junctions based on ferritins are induced by the magnetic iron oxide core inside the ferritin. The 2 nm thick protein cage: (a) serves as a tunneling barrier, (b) magnetically decouples the iron oxide core from the Ni^TS bottom-electrode, and (c) provides stability. These properties of ferritin allowed us to fabricate a highly efficient biomolecular TMR device with a relatively simple structure with only one FM electrode. The TMR was 60 ± 10% at 200 K and 25 ± 5% at room temperature. The TMR depends on how the magnetic core of ferritin was prepared. The magnetic moment µ_B per ferritin of the core was 5.9 times (table 1) higher for ferritin loaded with iron oxide under anaerobic than aerobic conditions resulting in a factor of four increase of the TMR. The mechanism of charge transport across iron oxide loaded ferritin junction studied here is independent of the temperature, which suggests that a long-range tunneling mechanism of charge transport predominates. To explain the observed TMR involving large coercive fields present in the junctions, but absent in apo ferritin, we propose that 'long-range tunneling' is assisted by magnon excitations of ferritins. Our findings showcase that ferritins form a highly interesting class of biomolecules for fundamental studies of new tunneling phenomena. They not only link long-range tunneling and spin dependent transport at the molecular level [77] but also enable large TMR performances and room-temperature operation of spintronic devices, important requirements for practical applications. Currently, we are investigating the mechanism of spin dependent transport and role of magnons in more detail.

4.1. Materials

4.2. Template-stripped nickel (Ni^TS)

4.3. Ferritin monolayer on Ni^TS

4.4. Device fabrication

4.5. Ferritin production and purification

4.6. Iron loading in ferritin

4.7. Freeze dried ferritin

4.8. Superconducting quantum interference device (SQUID)

4.9. MOKE measurements

4.10. X-ray photoelectron spectroscopy (XPS)

4.11. X-ray absorption spectroscopy (XAS)

4.12. X-ray magnetic circular dichroism (XMCD)

4.13. Atomic force microscopy (AFM)

4.14. Data collection of DC measurements

4.15. Temperature-dependent measurements

4.16. Magnetic field dependent measurements

The data generated and/or analyzed during the current study are not publicly available for legal/ethical reasons but are available from the corresponding author on reasonable request.

We acknowledge the Ministry of Education (MOE) for supporting this research under award No. MOE2019-T2-1-137. Prime Minister's Office, Singapore under its Medium sized center program is also acknowledged for supporting this research. We kindly acknowledge the Singapore Synchrotron Light Source (SSLS) supporting our experiments at the SINS beam line under NUS core support C-380-003-003-001. We would like to acknowledge beamline scientists Dr Bruce Cowie and Dr Anton Tadich (soft x-ray (SXR) Beamline at the Australian Synchrotron) for their immense support. E d B acknowledges support from the US National Science Foundation (Grant No.: ECCS#1916874). S R acknowledges funding from CA2DM-NUS during his stay in Singapore. ICN2 is funded by the CERCA programme/generalitat de Catalunya, and the Severo Ochoa program from Spanish MINECO (Grant No. SEV-2017-0706).