High‐Efficiency Asymmetric Spin Transport Enabled by Nanocolumn Molecular Semiconductors

ABSTRACT

Owing to weak spin‐orbit coupling, molecular semiconductors are among the few materials supporting room‐temperature spin functionality, yet their low spin‐transport efficiency (η _s, ∼5%) limits applications. Here, we report molecular spintronic devices featuring vertically asymmetric nanocolumn channels formed by phase separation. These channels confine spins and generate built‐in electric fields, boosting room‐temperature η _s to 20%—the highest value reported to date, over five times that of unstructured films. Simultaneously, the nanocolumn channels induce pronounced bias‐dependent asymmetry, with η _s of 20% at +0.2 V versus 1% at −0.2 V, yielding a record asymmetry factor, significantly outperforming other material systems (e.g., metal oxides, 2D materials, conventional molecular/inorganic semiconductors). This dual achievement of record‐high efficiency and strong asymmetry establishes a platform for new spintronic functionalities. As a proof of concept, we demonstrate its potential for information‐secure applications via spin‐signal encryption elements and two‐stage spin true random number generators, integrating structural design with spintronic operation.

Molecular semiconductors are engineered into vertically asymmetric, phase‐separated nanostructures that confine spins and create built‐in electric fields. This design boosts room‐temperature spin‐transport efficiency to 20% and produces strong bias‐dependent asymmetry. The combined effects establish a robust device platform, enabling hardware‐level spin security applications such as spin‐signal encryption and true random number generation.

1. Introduction

Spintronics offers a new paradigm for computing by leveraging electron spin—instead of charge alone—for data storage, transfer, and processing [1, 2, 3]. However, in spintronic devices, spin‐polarized carriers often lose their spin orientation as they traverse nonmagnetic materials, a process known as spin relaxation [4]. This effect becomes more pronounced as the temperature increases from cryogenic levels, posing a major challenge for preserving spin information throughout the spin transport process, particularly at room temperature (RT) [5]. Advancing spintronics toward practical applications therefore requires materials that can sustain spin orientation under RT conditions [6, 7]. In this context, molecular semiconductors (MSCs) are attracting growing interest because their intrinsically weak spin relaxation enables robust RT spin functionality compared to conventional spintronic materials [8, 9, 10, 11, 12], making them promising candidates for advancing next‐generation spintronic technologies [13, 14, 15].

In MSC systems, molecular spin valves (MSVs) serve as a fundamental platform for exploring spin‐related phenomena [5, 16]. A typical MSV features a layer‐by‐layer structure, with the MSC layer sandwiched between two ferromagnetic (FM) electrodes. Spin‐transport efficiency (η _s), quantitatively determined by the ratio of spin‐polarized current (I _{spin‐polarized}) to the total current (I _total) and ranging from 0% to 100%, η _s = (I _{spin‐polarized}/I _total) × 100%, serves as a key metric for evaluating the performance of MSVs and reflects their viability for practical spintronic applications [15, 17]. Despite substantial progress over recent decades in molecular spintronics, η _s at RT remains low, typically only a few percent, primarily because the MSCs currently used for spintronics research possess inherently limited spin‐transport capability [15, 18, 19], underscoring the persistent challenge of achieving high efficiency spin transport under practical ambient conditions. Efforts to improve η _s in MSVs have primarily focused on molecular design and aggregation control previously [8, 20, 21, 22, 23]. However, these strategies have only yielded limited success, and η _s at RT remains around 5%. This persistent inefficiency continues to constrain the development of practical molecular spintronic devices.

Reducing the widths of spin‐transport channels to the sub‐micrometer or nanometer scale offers a promising route to address current limitations in molecular spintronics [24, 25]. Theoretical studies have revealed that reducing channel widths can significantly suppress spin relaxation [26, 27], and experimental efforts have preliminarily demonstrated enhanced η _s in spatially confined geometries at cryogenic temperatures [24, 28, 29]. These advances underscore the potential of solid‐state structural engineering of spin‐transport channels to boost spintronic performance without relying on complex molecular design. However, the enhancement of η _s in such confined systems–based spintronic devices remains modest even at cryogenic temperatures, and robust detection of spin signals at RT has yet to be realized. A key obstacle lies in the reliable fabrication of micro‐ and nanoscale MSC‐based devices, which is hindered by electrode‐induced metallic filament formation during processing and measurement [16, 30, 31]. Consequently, exploiting spin functionalities inherent to confined MSC channels remains largely unexplored. Achieving a substantial enhancement in η _s, thereby unlocking new spintronic functionalities, thus remains a central challenge for the field.

In this study, we engineered vertically asymmetric nanocolumn MSC layers via liquid–liquid (L–L) phase separation and integrated them with a magnetic electrode transfer technique to construct MSVs. The spatially confined and directionally asymmetric transport channels strongly suppress spin relaxation and induce transport disparity, enabling pronounced bias‐dependent spin‐transport asymmetry. As a result, the η _s is enhanced by more than 500% at RT, reaching 20%—a new benchmark in molecular spintronics—while maintaining over 95% of its initial value after 100 days of ambient storage and 5000 operation cycles, demonstrating outstanding environmental and operational stability. Building on the η _s enhancement, the asymmetric nanocolumn channels induce pronounced bias‐dependent spin‐transport asymmetry, with η _s reaching 20% at +0.2 V versus 1% at −0.2 V, corresponding to a record‐high asymmetry factor of 0.861—far larger than those reported for other material systems, including metal oxides, 2D materials, inorganic semiconductors, and conventional MSCs. This structural engineering strategy delivers η _s among the highest reported to date, with strong bias‐dependent asymmetry, establishing a foundation for spin functionalities not accessible in conventional MSC architectures, such as spin‐signal encryption elements (SSEEs) and two‐stage spin true random number generators. These results highlight solid‐state structural engineering of spin‐transport media as a powerful route to high‐efficiency and asymmetric spin transport at RT, thereby opening a pathway toward novel spintronic applications, such as hardware‐level data security.

2. Results and Discussion

2.1. Phase‐Separated Fabrication of Nanocolumn MSC

Phase separation in blended systems is a well‐established strategy for tailoring the active‐layer morphology of MSCs [32]. Spatially confined nanostructured MSC films can be fabricated through a simple solution spin‐coating process, driven by L–L phase separation in fullerene‐derivative/polymer blends [33, 34, 35], where the fullerene derivatives are ideal spin‐transport media owing to their intrinsically weak spin relaxation [36]. As illustrated in Figure 1a, the formation process involves two key steps: L–L phase separation, followed by solvent evaporation. In a ternary solution comprising a fullerene derivative, a polymer, and a solvent, L–L phase separation occurs when repulsive interactions exist between the fullerene and polymer components [33]. To enable nanocolumn formation, the solvent must also exhibit a high vapor pressure to promote rapid drying [33], since slower evaporation leads to uniform, flat morphologies that inhibit nanocolumn growth. Thus, successful fabrication of nanocolumn MSC structures requires a fullerene/polymer blend with mutual repulsion and the use of high‐vapor‐pressure solvents.

FIGURE 1.

Formation and morphology of phase‐separated ICBA–PS films. (a) Schematic illustration of liquid–liquid (L–L) phase separation in a fullerene derivative/polymer/solvent mixed system, leading to the formation of phase‐separated films. (b) Contact angle (CA) measurements of pure indene‐C₆₀ bis‐adduct (ICBA) and polystyrene (PS). The calculated surface free energies (γ _s) of ICBA and PS are 22.9 and 28.6 mJ m⁻², respectively, indicating repulsive interactions. (c) Atomic force microscopy (AFM) images of ICBA–PS films with varying blend ratios (x wt%, where x represents the weight of PS as a percentage of ICBA, x = 0–25). Approximately circular domains emerge when x ≥ 10, while the film exhibits flat and uniform at x = 0 with a root‐mean‐square roughness (R _q) of only 0.3 nm. (d) Frequency distribution of circular domain diameter under different x conditions (10–25 wt%), extracted from AFM images using ImageJ.JS software (see Experimental Section for details). (e) Summary of the statistical results in (d), showing a decrease in domain diameter with increasing x. Data are presented as mean ± SD.

Herein, a ternary solution comprising indene‐C₆₀ bis‐adduct (ICBA), polystyrene (PS), and chloroform (CF) was used to fabricate nanocolumn MSC channels. ICBA—a widely used fullerene derivative composed solely of carbon and hydrogen—is well suited for spin signal transmission owing to its intrinsically weak spin relaxation. PS, with excellent solution processability and dielectric properties, ensures that spin transport occurs exclusively within ICBA‐rich domains. The high vapor pressure of CF enables rapid drying, which is essential for inducing L–L phase separation. The interaction between ICBA and PS was assessed via surface free energy (γ _s), determined through contact angle measurements [32]. As shown in Figure 1b, the pronounced differences in contact angles and γ _s between pure ICBA and PS indicate sufficient repulsive interactions to drive L–L phase separation. To verify whether such repulsive interactions can indeed facilitate the formation of phase separation–driven nanostructured films, a series of ICBA–PS films with varying weight ratios (x wt%, where x denotes the weight of PS relative to ICBA) was prepared. Atomic force microscopy (AFM) (Figure 1c) and scanning electron microscope (SEM) (Figure S1a) images reveal the emergence of discrete, circular‐like domains at x ≥ 10, whereas the film exhibits flat and uniform at x  =  0. In accordance with L–L phase separation principles [33, 34], these circular domains correspond to ICBA nanocolumns dispersed within the PS matrix (as further confirmed by conductive AFM measurements). Statistical analysis of domain diameters (Figure 1d,e) shows a systematic reduction in nanocolumn size with increasing x, and this trend is highly reproducible across multiple samples (Figure S1b).

2.2. Strongly Suppressed Spin Relaxation in Nanocolumn MSCs

To evaluate the influence of phase‐separated structures on charge and spin transport, we investigated molecular packing, charge carrier mobility, and spin relaxation dynamics. Grazing‐incidence wide‐angle X‐ray scattering (GIWAXS) patterns of both pure ICBA and ICBA–PS blend films (Figure 2a) show amorphous scattering rings without preferential orientation, indicating that the presence of PS does not alter ICBA's molecular packing. This observation is further supported by molecular dynamics simulations of the center‐of‐mass radial distribution functions (Figure 2b) [37, 38], which confirm that intermolecular spacing remains unchanged. Therefore, the reduction in nanocolumn size does not lead to enhanced dipolar or exchange interactions [39]. In line with the constant molecular packing, the energy level alignment of ICBA (Figure 2c) likewise exhibits negligible variation across different compositions. Despite the invariant molecular‐level structures, charge carrier transport in nanocolumn ICBA layers exhibits pronounced variations arising from the size‐dependent effects [31]. As shown in Figure 2d, zero‐field electron mobilities (µ) in ICBA–PS films were extracted using impedance spectroscopy [40, 41, 42], which were determined using:

$$\mu =\frac{e{d}^2}{k_{\mathrm{B}}T{\tau }_{\mathrm{d}}}$$ (1)

where e is the elementary charge, k _B is the Boltzmann constant, T is the absolute temperature, d is the film thickness (80 nm for these measurements), and τ _d is the charge carrier diffusion time. τ_d = R ₁ C _μ, transport resistance (R ₁) and chemical capacitance (C _µ) were obtained by fitting experimental data to an equivalent circuit model (Figure S3). Here, the extracted µ values (Figure 2e) exhibit a systematic decrease with increasing x, consistent with previous studies on size‐dependent effects [24, 31].

FIGURE 2.

Electronic and spintronic properties of phase‐separated ICBA–PS films. (a) Grazing‐incidence wide‐angle X‐ray scattering (GIWAXS) patterns of ICBA–PS films. (b) Center‐of‐mass radial distribution functions g (r) of ICBA intermolecular distance (r) in ICBA–PS films, obtained via molecular dynamics simulations. (c) Energy level alignment of the frontier molecular orbitals of ICBA (HOMO: highest occupied molecular orbital; and LUMO: lowest unoccupied molecular orbital) in ICBA–PS blend films with respect to the vacuum level, as determined by ultraviolet photoelectron spectroscopy (UPS) and low‐energy inverse photoelectron spectroscopy (LEIPS) measurements (Figure S2). (d) Impedance spectra of the real (Z′) and imaginary (Z″) components. Grey lines represent fitting results using the equivalent circuit model (Figure S3). (e) Zero‐field electron mobility of ICBA–PS films extracted from the fitting results in (d). (f) Spin relaxation strength (1/T ₁) of ICBA–PS films measured by field‐induced electron paramagnetic resonance (EPR), experimental data and corresponding fitting curves are shown in Figure S4. Error bars in (e) and (f) represent mean ± SD (n = 3).

For investigating the spin dynamics of nanocolumn MSCs, we extracted their spin relaxation strength (1/T ₁) and spin dephasing strength (1/T ₂) by analyzing the power saturation and unsaturated linewidth spectra of field‐induced electron paramagnetic resonance (EPR) spectroscopy (see Experimental Section & Figure S4) [20, 43]. Since 1/T ₁ governs spin relaxation during transport and directly determines spin transport efficiency [9, 21], we primarily focus on the evolution of 1/T ₁. A pronounced reduction in 1/T ₁ was observed in all ICBA–PS blend films with phase‐separated morphology (Figure 2f), with a more than 250% decrease compared to the pure ICBA film. This indicates a substantial suppression of spin relaxation in the nanostructured transport channels. In addition, 1/T ₂ remains largely unchanged between bulk and nanostructured films (Figure S5). Since 1/T ₂ reflects spin dephasing induced by local inhomogeneities [9, 21], this result suggests that the phase‐separated transport channels have a limited influence on spin dephasing processes. To further understand the origin of the reduced 1/T ₁, we consider the expression [39, 44]:

$$\frac{1}{T_1}=\frac{16{\gamma }^2}{{\overline{R}}^2}\times \frac{\mu k_{\mathrm{B}}T}{e}$$ (2)

where γ ² is the spin admixture parameter, and ${\overline{R}}^2$ is the mean hopping distance. Given that the molecular packing and intermolecular distance of ICBA remain unchanged, ${\overline{R}}^2$ is expected to be constant. According to this Equation (2) and the measured 1/T ₁ and µ values, γ ² was found to be closely related to the phase‐separated growth of the ICBA layers [29, 31], exhibiting a clear decrease as the nanocolumn size diminishes (Figure S6), thereby demonstrating a pronounced suppression of spin relaxation in these nanocolumn MSCs. This suppression arises from the nano‐structuring induced modification of the transport geometry in the phase‐separated films, where laterally separated transport channels spatially confine carriers within isolated pathways and reduce pathway crossings, rendering the transport process less stochastic [29, 31]. Notably, this effect does not rely on lateral dimensions comparable to the carrier mean free path, but instead originates from geometrical constraints on carrier trajectories that suppress spin mixing and spin relaxation.

2.3. Significantly Enhanced η _s in Nanocolumn MSVs

In this study, MSVs—the most prototypical molecular spintronic devices—were employed to investigate spin‐transport properties in phase‐separated nanocolumn spacers [16]. In conventional MSV fabrication, top FM electrode is directly deposited onto the MSC surface via evaporation. However, for discontinuous nanocolumn ICBA–PS films, direct evaporation of the top FM electrode can penetrate the nanogaps [30], thereby compromising the integrity of the spacer layer and disrupting the device architecture. As shown in Figure S7, we fabricated control devices in which the top Ni₈₀Fe₂₀ electrode was deposited directly onto the active layer using conventional electron‐beam evaporation. These devices exhibit only very weak spin signals. In particular, for phase‐separated nanostructured films (x ≥ 10 wt%), the spin valve signal is nearly undetectable, indicating that reliable spin injection and detection cannot be achieved under such conditions. This explains why previously reported MSVs with nanostructured channels often failed to exhibit detectable spin signals at RT. To overcome this critical bottleneck, we implemented an electrode transfer technique to fabricate nanocolumn ICBA–based MSVs (Figure 3a) [45]. A high‐modulus polymer film (e.g., PS) was used to assist the transfer, while a self‐assembled octadecyl trichlorosilane (OTS) layer reduced the adhesion between the PS film and the glass substrate. This approach limits the stress experienced by the electrode during transfer to an extremely low level, ensuring damage‐free assembly. The lamination was performed in a vacuum oven at 105°C, where controlled softening of the PS support layer enabled conformal contact between the transferred electrode and the underlying ICBA layer. Importantly, this thermal treatment does not alter the morphology or molecular packing of the ICBA film (Figure S8). Additional fabrication details are provided in the Experimental Section.

FIGURE 3.

Spin‐transport efficiency (η _s) of molecular spin valves (MSVs) based on phase‐separated ICBA films. (a) Schematics illustrating the process of constructing nanocolumn‐ICBA MSVs through the transfer of ferromagnetic (FM) electrodes, forming a Si/SiO₂/Co/AlO_x/ICBA–PS/Ni₈₀Fe₂₀/PS (bottom to top) structure. The constructed MSVs were then measured by the four‐point probe method. B indicates an in‐plane sweeping magnetic field. More details in Experimental Section. (b) The measured η _s of MSVs based on ICBA–PS films with varied PS content. The thickness of the ICBA–PS layer in these MSVs is all 48 nm. The red curves correspond to the magnetic field sweeping from negative to positive, while the blue curves from positive to negative. (c) η _s variation with the PS content (x = 0–25). (d) Environmental stability of MSVs (x = 0, 15), η _s has no significant degradation kept under ambient air at RT for 100 days. (e) η _s values of MSVs (x = 0, 15, 20) with different thicknesses of ICBA layer, the lines were fitted by Equation (3). All the measurements were carried out at 0.01 V bias voltage, RT and air condition. Error bars in (c) and (e) were calculated by three samples under the same condition.

The η _s of the nanocolumn MSVs were measured using a four‐probe method, as illustrated in Figure 3a, within a magneto‐electric probe station. Applying a positive bias voltage between the two FM electrodes (Co‐side grounded), spin‐polarized electrons are injected from the Co/AlO_x terminal into the ICBA nanocolumns for spin transport, and then detected at the Ni₈₀Fe₂₀ terminal. Under the sweeping magnetic field, the output current of the MSVs switches between high and low current states according to the magnetization alignment of the two FM electrodes. The η _s reflects the proportion of spin‐polarized charge carries transported through the device channel, expressed as [15, 16]:

$${\eta }_{\mathrm{s}}=\frac{I_{\mathrm{spin}-\mathrm{polarized}}}{I_{\mathrm{total}}}\times 100\%=\frac{2P_1{P}_2\exp \left(-d/L_{\mathrm{s}}\right)}{1-P_1{P}_2\exp \left(-d/L_{\mathrm{s}}\right)}\times 100\%$$ (3)

where P ₁ P ₂ is the product of the spin polarizations at the two terminals, L _s is the spin diffusion length, d is the thickness of the ICBA–PS layer. As shown in Figure 3b, all the MSVs (x = 0–25, d = 48 nm) display distinct current switching behaviors under magnetic field sweeps, confirming that high‐quality MSVs have been successfully constructed. The η _s values versus x are summarized in Figure 3c. The MSVs based on nanocolumn ICBA (x = 10–25) have obvious higher η _s values than those based on bulk ICBA films (x = 0). The η _s reaches its maximum value (> 15%) at x = 15, which is more than 400% of the value at x = 0. Figure 3d demonstrates the excellent environmental stability of both MSVs, with negligible degradation in spin‐transport performance observed after 100‐days storage under ambient air at RT, which is attributed to the encapsulation effect provided by the top PS layer [45]. Moreover, both devices retain 95% of their initial η _s values after 5000 biasing cycles, demonstrating great operational stability (Figure S9). In addition, measurements under out‐of‐plane magnetic fields and angle‐dependent in‐plane configurations reveal that η _s exhibits nearly identical magnitude (Figure S10), indicating that the spin transport characteristics are largely insensitive to the magnetic field orientation in these vertically structured devices. This further confirms that the observed spin‐valve signals originate from spin transport within the ICBA nanocolumns, rather than from contributions of anisotropic magnetoresistance in the FM electrodes.

To elucidate the origin of the η _s evolution, we investigated its dependence on the thickness (d) of the ICBA–PS layer (Figure 3e). By fitting their relationship using Equation (3), the P ₁ P ₂ and L _s values can be extracted. Specifically, P ₁ P ₂ = 0.096 and L _s = 28.0 nm for x = 0, and P ₁ P ₂ = 0.100 and L _s = 116.2 nm for x = 15. This result indicates that the remarkable enhancement of η _s is attributed to the increase of L _s. At x = 20, L _s decreases to 61.0 nm, leading to a moderate decline in η _s. According to the expression of L _s [44]:

$$L_{\mathrm{s}}=\sqrt{\frac{\mu k_{\mathrm{B}}T}{e}\times T_1}=\frac{\overline{R}}{4\sqrt{{\gamma }^2}}$$ (4)

The normalized L _s was calculated based on the γ ² values (Figure S11). As x increases, L _s reaches a maximum at x = 15, followed by a gradual decline, consistent with the trend observed in Figure 3c. Specifically, as x increases, the diameter of the ICBA nanocolumns gradually decreases, leading to a corresponding decrease of 1/T ₁ (Figure 2f). In contrast, the carrier mobility µ decreases monotonically with increasing x (Figure 2e), likely due to enhanced Coulomb blockade at nanocolumn boundaries and transport hindrance from residual PS components acting as trap states [24, 46]. The combined effect of these two factors results in a non‐monotonic dependence of L _s and η _s, both of which reach their maximum at x = 15.

2.4. Asymmetric Bias Voltage Dependent η _s in Nanocolumn MSVs

We further investigated the bias voltage (V _b) dependence of η _s in phase‐separated nanocolumn (x = 15) and bulk (x = 0) MSVs, as shown in Figure 4a. In this measurement, spin‐polarized electrons are injected from the top Ni₈₀Fe₂₀ at V _b < 0, while from the bottom Co at V _b > 0. In the x = 0 MSV, η _s decreases monotonically with increasing |V _b|, consistent with previous reports [16, 18], which is commonly attributed to enhanced spin relaxation at higher |V _b| [47]. In contrast, the x = 15 MSV exhibits pronounced asymmetry: while η _s still decreases with increasing |V _b| for top injection, it initially increases with |V _b| for bottom injection until reaching a maximum and then decreases (Figure 4a). Notably, the maximum η _s occurs well away from V _b = 0, an exceedingly rare feature in molecular spintronics with potential to unlock unconventional spintronic functionalities. To accurately evaluate this unique behavior, an asymmetry factor (AF) was introduced to quantitatively characterize the V _b‐dependent asymmetry of spin transport, defined as [48]:

$$AF=\left|\frac{{\eta }_{\mathrm{s}}\left(V_{\mathrm{b}}\right)-{\eta }_{\mathrm{s}}\left(-V_{\mathrm{b}}\right)}{{\eta }_{\mathrm{s}}\left(V_{\mathrm{b}}\right)+{\eta }_{\mathrm{s}}\left(-V_{\mathrm{b}}\right)}\right|$$ (5)

where η _s(V _b) and η _s(−V _b) refer to η _s at V _b of the same magnitude but opposite polarity, and η _s(V _b) represents the maximum value of η _s in the η _s–V _b relation. The AF ranges from 0 to 1, with values approaching 0 indicating that the η _s–V _b characteristics are nearly symmetric about V _b = 0, and values approaching 1 indicating pronounced asymmetry. For the x = 0 MSV, the AF is as low as 0.004, reflecting the monotonic decrease of η _s on both sides of V _b = 0. By contrast, the x = 15 MSV displays an exceptionally high AF of 0.861, arising from a distinctive enhancement of η _s under positive V _b. Notably, measurements under out‐of‐plane magnetic fields reveal the same bias dependence of η _s (Figure S12), indicating that this asymmetric behavior is independent of the magnetic field orientation. This pronounced bias‐induced asymmetry, reminiscent of rectifying behavior in molecular diodes [49], could be exploited to enable electrically controlled spintronic functionalities.

FIGURE 4.

Bias voltage dependent spin transport properties. (a) η _s values of MSVs (x = 0, 15 heated at 105°C, and x = 15 heated at 120°C) versus bias voltage (V _b). Electrons are injected from top Ni₈₀Fe₂₀ terminal at negative V _b, and from bottom Co terminal at positive V _b. For x = 0 MSV, the η _s–V _b relation is nearly symmetric about V _b = 0, whereas for x = 15, the η _s increases markedly under positive V _b, and this enhancement attenuated for the MSV heated at 120°C. (b) Height analysis of ICBA–PS film (x = 15) based on AFM imaging. Besides the main nanocolumns exhibiting uniform height, several satellite nanocolumns with lower heights are also observed. (c, d) Schematic illustration of asymmetric η _s at opposite V _b. (c) At negative V _b, electrons injected from the Ni₈₀Fe₂₀ are confined to spin transport through the main nanocolumns, as the satellite nanocolumns lack a conduction path to the top contact. (d) At positive V _b, electrons injected from the bottom Co electrode into the satellite nanocolumns cannot reach the top Ni₈₀Fe₂₀ electrode, leading to charge accumulation within the nanocolumns. This accumulation induces a built‐in electric field (E _i) oriented opposite to the externally applied field (E _a). The resulting unpolarized reverse current selectively cancels the unpolarized component of the injected current in main nanocolumns, thereby enhancing the η _s. (e, f) Conductive AFM images of ICBA–PS film (x = 15) before (e) and after (f) charged by preset voltage. The overall current in (f) is reduced compared with that in (e), which is attributed to charge accumulation within the ICBA nanocolumns after the preset bias. Upon reapplying the bias, the accumulated charges partially counteract the measured current. The image size is 1.5 × 1.5 µm². (g, h) Elevating the lamination temperature to 120°C promotes partial satellite nanocolumns to establish conduction paths with both electrodes. These connected nanocolumns lose the ability to generate the internal field E _i, and thus no longer enhance spin transport under positive V _b. (i) Capacitance of MSVs (x = 0, 15 heated at 105°C, and x = 15 heated at 120°C) at 1 kHz, the junction area is 4 mm². (j) η _s values at 0.2 V of MSVs (x = 15 heated at 105, and 120°C, respectively) with different thicknesses of ICBA–PS layer, the lines were fitted by Equation (3). Error bars in (i) and (j) represent mean ± SD (n = 3).

The pronounced enhancement in AF can be ascribed to the vertically asymmetric solid‐state structure of ICBA nanocolumns, which induces distinct conductive pathways at the top and bottom FM/MSC interfaces in the MSVs. During the growth of phase‐separated films, partial satellite nanocolumns form around the main nanocolumns. These satellite nanocolumns exhibit significantly lower heights compared to the main nanocolumns, resulting in a unique vertically asymmetric solid‐state MSC architecture (Figure 4b). At the standard lamination temperature of 105°C, the satellite nanocolumns are only electrically connected to the bottom Co electrode and remain disconnected from the top Ni₈₀Fe₂₀ electrode, as illustrated in Figure 4c,d and confirmed by the cross‐sectional transmission electron microscopy (TEM) image (Figure S13a). The difference in conductive pathways between these satellite nanocolumns and the Co and Ni₈₀Fe₂₀ electrodes results in markedly different spin polarizations in MSVs under opposite V _b. In detail, at negative V _b (Figure 4c), the Ni₈₀Fe₂₀ electrode cannot inject electrons into the satellite nanocolumns, restricting spin transport to the main nanocolumns. As |V _b| increases, η _s would decrease monotonically due to enhanced spin relaxation. In contrast, under positive V _b (Figure 4d), electrons injected from the Co electrode into the main nanocolumns can reach the Ni₈₀Fe₂₀ electrode, those entering the satellite nanocolumns lack a conduction path to the top contact and consequently accumulate. This charge buildup renders the satellite nanocolumns effective capacitive elements. The resulting electron accumulation gives rise to an electrostatic potential gradient, which in turn generates an internal electric field (E _i) that opposes the externally applied field (E _a). As the induced current by E _i is non‐polarized, it selectively compensates the non‐polarized portion of the injected current in main nanocolumns, thereby enhancing the overall spin polarization [15, 17]. As V _b increases, the induced E _i in the satellite nanocolumns becomes stronger, leading to a more pronounced enhancement in η _s. This effect competes with the suppression of η _s caused by increased spin relaxation at higher V _b. The interplay between these two opposing effects gives rise to a non‐monotonic dependence of η _s on V _b, which first increases and then decreases (as shown in Figure 4a). Conductive AFM also supports this interpretation: nanocolumns in the x = 15 film exhibit substantially reduced conductivity (Figure 4f) upon preset voltage, in contrast to their pristine state (Figure 4e), consistent with the formation of an E _i that suppresses the total current.

To substantiate the above analysis, we increased the lamination temperature to 120°C, enabling more satellite nanocolumns to form electrical contact with both electrodes (sketch in Figure 4g,h and TEM image in Figure S13b). In this case, the spin‐transport asymmetry is expected to drop substantially, as partial satellite nanocolumns lose their ability to enhance spin polarization. This is evidenced by the significant reduction of the AF to 0.355 (see Figure 4a). Moreover, impedance measurements were conducted to further elucidate the role of satellite nanocolumns in modulating spin polarization. As shown in Figure 4i, the x = 15 MSV exhibits a capacitance of 8.07 nF, which is substantially higher than the 1.33 nF of the x = 0 counterpart, indicating enhanced capacitive behavior. Upon lamination at 120°C, this elevated capacitance significantly decreases to 4.09 nF. These observations corroborate the capacitive nature of satellite nanocolumns, which function as local charge reservoirs contributing to spin‐transport modulation.

As outlined above, the E _i induced by the vertically asymmetric ICBA–PS structure plays a crucial role in modulating the spin polarization of charge carriers. Under positive V _b, the accumulation of electrons in satellite nanodomains generates E _i, which selectively compensates the non‐spin‐polarized component of the current flowing through the main nanodomains. This electrostatic filtering effect effectively enhances the P ₁ P ₂. To quantitatively evaluate this effect, we extracted the P ₁ P ₂ values by fitting the η _s–d relation (Figure 4j). At 105°C, the x = 15 MSV exhibits a P ₁ P ₂ value of 0.136, which decreases to 0.091 when the lamination temperature is increased to 120°C. This reduction can be attributed to the increased electrical connectivity of satellite nanodomains, which suppresses charge accumulation and weakens the electrostatic filtering effect. These results clearly demonstrate that the enhanced P ₁ P ₂ and η _s under positive bias originate from the asymmetric nanostructure of the phase‐separated ICBA–PS film and its ability to selectively filter non‐polarized carriers via electrostatic modulation.

2.5. Record‐High η _s and AF With Emerging Spintronic Functionalities

In our phase‐separated ICBA MSVs, η _s is markedly enhanced by suppressed spin relaxation within laterally segregated channels and is further amplified by bias‐induced effects arising from the vertically asymmetric architecture. Together, these mechanisms yield a record‐high RT η _s of 20%—over fivefold higher than non‐engineered films and far surpassing previous reports (Figure 5a). In contrast, recent efforts relying on elaborate molecular design strategies have typically achieved only modest improvements, with η _s remaining near 5% [20, 21, 22]. Our phase‐separation approach thus provides a simple, reproducible, and scalable route to substantially improving η _s, a prerequisite for realizing spintronic devices with high signal‐to‐noise ratios at RT. Beyond this efficiency milestone, the vertical solid‐state asymmetry of the ICBA–PS layer induces a pronounced bias‐dependent spin‐transport asymmetry, yielding an AF of up to 0.861. Such a pronounced asymmetric response is rare in molecular spintronic devices and has previously been reported only at low temperatures (Figure 5b) [50]. Notably, the high AF achieved in this work at RT exceeds the maximum values reported in other material systems, including metal oxides, 2D materials, inorganic semiconductors, and conventional MSCs, by more than a factor of three (Figure 5c). From this perspective, our work establishes solid‐state structural engineering as a systematic route to high‐efficiency asymmetric spin transport, offering a versatile platform for spintronic functionalities beyond conventional device paradigms.

FIGURE 5.

Benchmarking spintronic performance against literature. (a) RT η _s, phase‐separated ICBA MSVs exhibit a two‐step enhancement of RT η _s, setting a new record in molecular spintronics (Table S1). (b) Bias‐dependent asymmetry factor (AF) in molecular spintronics, while pure ICBA MSVs show AF values comparable to most previous reports, typically below 0.1, the phase‐separated ICBA MSVs achieve AF up to 0.861. Such a pronounced asymmetric response has previously been observed only at low temperatures (Table S2). (c) Maximum AF across different materials at RT (Table S3). The AF of 0.861 achieved here is more than three times than other material systems. Considering the limited signal‐to‐noise ratio and stability of the previously reported spintronic devices, asymmetric spin transport can be regarded as practically relevant when AF consistently exceeds 0.3.

Leveraging these advances, we illustrate the platform's capability for information‐secure applications at the hardware level, exemplified by two conceptual spintronic devices: a SSEE and a two‐stage spin true random number generator (TS‐TRNG). Figure 6a presents an electrical schematic for SSEE. In single‐supply spintronic devices, implementing SSEE requires spin signals with a high signal‐to‐noise ratio under one bias polarity, while becoming indistinguishable under the opposite bias of equal magnitude (Figure S14). In our phase‐separated MSVs (x = 15), the combination of enhanced η _s and a large AF produces spin signals with a high signal‐to‐noise ratio (η _s = 20%) at V _b = 0.2 V, while η _s is nearly zero at V _b = −0.2 V. This pronounced bias‐dependent contrast makes the device particularly well suited for SSEE. Through the control of the multiplexer (MUX), the encryption and readout of spin signals can be achieved at the hardware level (Figure 6b). Furthermore, under a higher bias of V _b = 0.5 V, the device exhibits stochastic fluctuations in output base current between two metastable states (I ₁ and I ₂, or I ₃ and I ₄) (Figure 6c–e). This behavior constitutes a TS‐TRNG, in which the amplitude of randomness is magnetically programmable (mechanism as discussed in Note S1). Compared to single‐state TRNGs, TS‐TRNGs offer significantly greater promise for enhanced data protection and more efficient probabilistic computing [51, 52].

FIGURE 6.

Proof‐of‐concept spintronic functionalities enabled by high‐efficiency asymmetric spin transport. (a) Electrical schematic of the spin‐signal encryption element (SSEE), integrating an MSV, inverter (−1×), and multiplexer (MUX). (b) The large η _s and AF allow high signal‐to‐noise spin signal readout at V _b = 0.2 V and concealment at V _b = −0.2 V. By using the MUX to control the V _b input, spin signal readout (S = 0) and encryption (S = 1) are achieved. When the MUX input is set to S = 0, a bias of V _b = 0.2 V is applied to the device, allowing readout of the spin signal (as transitions between “0” and “1” state). Conversely, when S = 1, a reverse bias of −0.2 V is applied, under which the spin transport is strongly suppressed to nearly undetectable levels, effectively concealing the spin signal and enabling device‐level encryption. (c−e) Proof‐of‐concept demonstration of a two‐stage spin true random number generator (TS‐TRNG) based on a phase‐separated MSV (x = 15) operating at V _b = 0.5 V. The implementation of TS‐TRNG also benefits from the high η _s and AF, which ensure a robust current switching between different magnetization alignments, even at V _b up to 0.5 V. The output current fluctuates between two metastable states and can be further modulated by an external magnetic field, enabling a TS‐TRNG with tunable output ranges (I ₁, I ₂) and (I ₃, I ₄). (d, e) Magnified view of the dashed region in (c).

Collectively, these results highlight that the rational construction of nanoscale solid‐state molecular architectures is a highly effective strategy to enhance η _s and AF, thereby advancing MSVs toward emerging spintronic applications such as hardware‐level data security. Hardware‐based cryptographic systems, owing to their inherent physical unpredictability, offer significantly higher levels of security than traditional software‐based encryption schemes, which is increasingly critical in the era of artificial intelligence and emerging quantum computing [52, 53].

3. Conclusion

We have developed and demonstrated an effective strategy to achieve high‐efficiency and bias‐dependent asymmetric spin transport in MSVs through solid‐state structure engineering of MSC transport channels. The enhanced η _s and its pronounced asymmetry stem from two key mechanisms: the substantial suppression of spin relaxation strength within spatially confined transport channels, and the amplification of spin polarization driven by reverse electric fields induced by the asymmetric architecture. Harnessing these effects, we attained a RT η _s of up to 20% and an AF of 0.861—both surpassing previous records. The record‐high η _s, together with the distinctive spin‐transport asymmetry, enables emerging spintronic functionalities such as SSEE and TS‐TRNG. These results establish solid‐state structural engineering as a powerful and versatile strategy to optimize spintronic performance and unlock multifunctional device applications, offering a new paradigm for molecular spintronics.

Given that the optimization of spin transport performance originates from the nano‐structuring of transport channels rather than specific molecular properties, this approach may be extended to a broad range of material systems, such as molecular systems with intrinsic paramagnetic centers [54, 55]. In this context, combining such nano‐structured transport channels with spectroscopic probes such as EPR could provide new opportunities for engineering and understanding spin dynamics [10, 43], as well as enabling emerging spintronic functionalities in molecular platforms.

4. Experimental Section

4.1. Materials

The fullerene derivative ICBA (purity, 99%), insulator polymer PS (average M _w, ∼192 000), and OTS were purchased from Sigma Aldrich. The metals, including Co, Ni₈₀Fe₂₀, Ag (purity, 99.95%), and Al (purity, 99.99%) were purchased from Kurt J. Lesker Company. PDINO and solvents were purchased from J&K Scientific.

4.2. Device Fabrication

The substrates n‐type Si/SiO₂ (300 nm) and glass were ultrasonically cleaned in deionized water, alcohol, acetone, and isopropyl alcohol, successively. The silicon substrates were also treated with piranha solution before use. The fabrication of the phase‐separated MSVs involves three main steps: top‐part preparation, bottom‐part preparation, and electrode transfer. For the top part, 30‐nm‐thick Ni₈₀Fe₂₀ electrodes were patterned using a shadow mask and evaporated onto a glass/OTS/PS substrate. The OTS layer was self‐assembled onto the glass surface by heating in a vacuum oven at 120°C, while the 3‐µm‐thick PS film was deposited onto the glass/OTS surface via spin coating. For the bottom part, 20‐nm‐thick Co was patterned onto the Si/SiO₂ substrate using a shadow mask, followed by the deposition of an all‐around 1.5‐nm‐thick Al layer. The Al layer was then mildly oxidized using air plasma treatment to form an AlO_x interfacial layer. Subsequently, ICBA/PS/CF blend solutions with various conditions were spin coated to form nanostructured films via L–L phase separation. For electrode transfer, the top glass/OTS/PS/Ni₈₀Fe₂₀ stack was aligned with the bottom Si/SiO₂/Co/AlO_x/ICBA–PS layer using a homemade tool. The OTS layer significantly reduced the adhesion between the PS film and the glass substrate, enabling damage‐free peel off the PS film. Then, thermal treatment at 105°C in a vacuum oven softened the PS layer, facilitating intimate contact between the Ni₈₀Fe₂₀ electrodes and the target ICBA layer. Co, Al, and Ni₈₀Fe₂₀ were all deposited via electron‐beam evaporation. The deposition rates were 1 Å s⁻¹ for both Co and Ni₈₀Fe₂₀, and 0.3 Å s⁻¹ for Al. As for the direct deposition of the Ni₈₀Fe₂₀ electrodes, a low‐temperature and rate‐controlled deposition technique was employed [15, 23]. Specifically, the first 2 nm of Ni₈₀Fe₂₀ was deposited at a rate of 0.1 Å s⁻¹, followed by 28 nm at 1 Å s⁻¹. During the deposition process, the substrate was cooled using liquid nitrogen. The deposition rates and film thicknesses during evaporation were calibrated using a quartz crystal monitor (INFICON SQM‐160). The thicknesses of the ICBA layer were determined using a spectroscopic ellipsometer (SE‐VE‐L, Wuhan Eoptics Technology). All fabrication steps of the device were carried out in a nitrogen‐filled glovebox (O₂, and H₂O concentrations < 0.1 ppm). These processes resulted in the final MSVs with the structure: Si/SiO₂/Co/AlO_x/ICBA–PS/Ni₈₀Fe₂₀/PS.

4.3. Characterization and Measurements

4.3.1. Characteristics of γ _s

Water contact angle measurements were performed on a DSA100S analyzer to characterize γ _s of pure ICBA and PS. ICBA and PS (12 mg mL⁻¹ in CF) were spin coated onto Si/SiO₂ at 3000 r.p.m. for 60 s. γ _s was calculated from the contact angles using the Neumann method [33, 56]:

$${\left({\gamma }_{\mathrm{s}}/{\gamma }_{\mathrm{l}}\right)}^{0.5}\exp \left\{-\beta {\left({\gamma }_{\mathrm{l}}-{\gamma }_{\mathrm{s}}\right)}^2\right\}=0.5\left(1+\cos \theta \right)$$ (6)

where γ _l is the surface free energy of the measuring liquid, γ _l = 72.8 mJ m⁻² for water, the coefficient β = 0.0001247, θ is the contact angle.

4.3.2. Characteristics of the Morphology

Samples were prepared by spin coating the ICBA–PS blend solutions (x = 0–25, 12 mg mL⁻¹ in CF) onto Si/SiO₂ at 3000 r.p.m. for 60 s. AFM images were obtained using a Burker Multimode 8HR under tapping mode. The domain diameter of the ICBA nanocolumns was statistically determined using the ImageJ.JS software. AFM images were converted to grayscale, thresholded, and processed to isolate domains. Equivalent circular diameters were obtained via the “Analyze Particles” function based on measured areas, with scale calibration applied. SEM images were performed on a Hitachi SU8200 working at 5 kV. GIWAXS measurements were carried out by an XEUSS SAXS/WAXS equipment to probe the aggregation behavior of ICBA molecules.

4.3.3. Characteristics of the Energy Level Alignment

HOMO and LUMO levels of ICBA were determined using UPS (Kratos AXIS ULTRA DLD) and LEIPS (ULVAC‐PHI LEIPS instrument, with Bremsstrahlung isochromatic mode), respectively. 4 mg mL⁻¹ ICBA–PS blend solutions were spin coated onto Si/Co (20 nm)/AlO_x (1.5 nm) to prepare the samples.

4.3.4. Impedance spectroscopy measurements

µ of ICBA–PS blend films were extracted from impedance spectroscopy measurements, which were conducted using a Keysight E4990A impedance analyzer. The samples have a structure as ITO/ZnO/ICBA–PS (80 nm)/PDINO (3 nm)/Ag (60 nm), where the Ag electrodes coated with PDINO were transferred onto the ICBA. ZnO was spin coated on UVO‐treated ITO glasses. The Cole–Cole plots were obtained under dark conditions with an AC oscillation level of 0.1 V and frequencies between 20 Hz and 10 MHz. The experimental data were fitted using an equivalent circuit model (Figure S3), τ _d and µ were calculated accordingly. The capacitances in Figure 4i were measured with a DC bias of 0.2 V and a frequency of 1 kHz. The measured device features the same structure as the ICBA MSVs, with a junction area of 4 mm².

4.3.5. EPR Measurements

Field‐induced EPR measurements were performed to extract 1/T ₁ and 1/T ₂ of the ICBA–PS blend films. The measurements were carried out using a Bruker Magnettech ESR5000 X‐band continuous‐wave EPR spectrometer with an external magnetic field modulated at 10 kHz. To ensure the reproducibility of the microwave magnetic field at the sample position, all measurements were performed using the same resonant cavity under identical coupling conditions, and the sample position and orientation were carefully fixed throughout the experiments. For all samples, identical instrumental modulation parameters were applied. The samples have a structure of Si/SiO₂/ICBA–PS/Au, where the Au electrode was fabricated via a transfer method. Before the onset of power saturation, spin dephasing strength (1/T ₂) is given by 1/T ₂ = γ_eΔB _1/2, where γ _e is the gyromagnetic ratio of electrons, and ∆B _1/2 is the half width at half maximum of the spectrum [43]. 1/T ₂ can therefore be extracted from the least‐squares fitting of Lorentzian resonance lines using the EasySpin‐5.2.36 in MATLAB [57]. T ₁ T ₂ can be determined by fitting the power saturation curves, according to the relationship between the double integral (DI) over the derivative spectrum and the microwave field (B ₁) [43]:

$$\mathrm{DI}\left(B_1\right)\propto \frac{B_1}{\sqrt{1+{\gamma }_{\mathrm{e}}^2{T}_1{T}_2{B}_1^2}}$$ (7)

Then 1/T ₁ can be calculated by T ₁ T ₂ and 1/T ₂.

4.3.6. Characteristics of η _s

All η _s measurements were conducted using a high‐vacuum magnetic probe station system (CRX‐EM‐HF, Lake Shore), with a Keithley 4200A‐SCS semiconductor parameter analyzer serving as the electrical source and measurement unit. All measurements were performed at RT using the standard four‐point probe method to ensure accuracy and eliminate contact resistance effects.

4.3.7. Cross‐Sectional TEM Characterization

The cross‐sectional TEM images were obtained using Talos F200X operating at 200 kV. The samples were prepared by the focused ion beam (FIB) technique (Helios Nanolab 600i).

4.3.8. Conductive AFM Characterization

Conductive AFM was employed to investigate the local electrical conductivity of the ICBA nanocolumns. The samples have a structure as Si/SiO₂/Co/AlO_x/ICBA–PS (x = 15). The measurements were conducted using a commercial scanning probe microscope (Dimension Icon, Bruker) operated in quantitative imaging mode. The Co electrode was grounded, and a +7 V bias was applied to the Pt/Ir‐coated tip. Charging by preset voltage refers to performing an initial scan under these conditions to charge the ICBA nanocolumns.

4.3.9. Molecular Dynamics Simulations

The coarse‐grained model for ICBA was based on the Martini “F16” fullerene representation and the Martini PCBM model [58]. PS chains were constructed using the polyply tool [59]. CF molecules were modeled using the default parameters provided within the Martini force field. The schematic models were presented in Figure S15. All simulations were performed with a time step of 20 fs at RT. Non‐bonded interactions were truncated at a cutoff of 1.1 nm. Neighbor lists were constructed using the Verlet algorithm. Temperature and pressure were controlled using the velocity‐rescaling thermostat and the Parrinello‐Rahman barostat, respectively. The coupling time constants were set to 1 ps for temperature and 15 ps for pressure. All simulations were carried out under three‐dimensional periodic boundary conditions using GROMACS versions 2019 [60].

Author Contributions

X.S. proposed and supervised this work. X.S., L.G., and S.H. designed the experiments. S.H. completed sample and device fabrication, measurements, and data analysis. T.Y. performed the electron paramagnetic resonance measurements. S.X. performed the molecular dynamics simulations and the morphology analysis. Y.W. helped with the device measurements. R.Z. drew the schematic of the device fabrication process. G.Z. performed the conductive atomic force microscopy characterization. K.W. supervised the impedance spectroscopy measurements. S.H. and X.S. co‐wrote the paper. All the authors participated in multiple‐round discussions and revisions of this article.

Funding

National Key R&D program of China (Grant No. 2024YFA1209600), the National Natural Science Foundation of China (Grant Nos. W2411042, 52473170, 52303225 and 22505048), the Strategic Priority Research Program of the Chinese Academy of Sciences (Grant Nos. XDB0520000, XDB1300000), the Beijing Nova Program (20240484551), the Beijing Natural Science Foundation (2264116), the China Postdoctoral Science Foundation (Grant Nos. 2024T170203, 2024M760684), the Postdoctoral Fellowship Program of CPSF (Grant Nos. GZB20250272), and Distinguished Professor Fund of Taishan Scholars.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Supporting File: adma73826‐sup‐0001‐SuppMat.docx.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.