Robust chirality-induced spin selectivity in topologically chiral molecular knots Authors: Xi Sun, Kai-Yuan Zhang, Shu-Zheng Zhou, Hua-Hua Fu Published: 2025 ABSTRACT Molecular knots with topological chirality demonstrate superior spin selectivity compared to conventional chiral materials, exhibiting ultra-high spin polarization of nearly 90%, conductivity increased by two orders of magnitude and high-temperature stability (up to 350°C). However, the physical origin of this phenomenon remains elusive. We establish a fundamental theoretical framework for the chirality-induced spin selectivity in topologically chiral molecular knots and identify its essential conditions. Our calculations show that a single trefoil knot molecule exhibits spin polarization exceeding 60% along with significant conductivity, robustness unaffected by lattice reduction or structural strain. Importantly, when the topological knot degenerates into a trivial structure, accompanied by the transition from topological chirality to structural chirality, the spin polarization sharply declines, demonstrating a strong correlation between the ultrahigh spin polarization and the knot topology. This work elucidates the mechanism of spin selectivity in knot molecules and provides new design principles for future spintronic devices. Introduction The interplay between electron spin and structural chirality leads to a spin-dependent transport phenomenon known as chirality-induced spin selectivity (CISS)1–6. This effect was first discovered in photoelectron spectroscopy measurements on DNA molecules7 and has since been reported in various chiral systems such as helicenes8, 9, chiral crystals10,11, donor-acceptor molecules12, and supramolecular assemblies13–16. A key signature of CISS is the generation of high spin polarization (up to 80%) in charge transport even without any external magnetic field17,18, and the spin-polarized state that can persist even at room temperature19. This capability for manipulating spin states and generating spin currents in non-magnetic materials at high temperatures makes CISS promising for spintronic device design20–22, as evidenced by recent demonstrations of room-temperature spin-emitting LEDs23,24 and high-efficiency chiral perovskite photovoltaics25–29. Beyond spin-dependent electronics, CISS also influences enantiospecific chemistry30, catalytic activity31–34, and biological functions35,36. Although multiple theoretical models have been proposed to understand this effect37–46, they still exhibit systematic contradictions with or explanatory limitations when confronted by experimental observations2, and a universal theory capable of fully explaining the CISS effect in all chiral materials remains elusive47. The traditional CISS effect is typically related to structural chirality, such as helical chains that break the inversion symmetry to produce left- or right-handed chirality (Fig. 1a). However, recent experiments point to a different kind of system where strong CISS arises without such conventional stereogenic units, namely, in molecular trefoil knots19 (Fig. 1b). These molecules derive their chirality not from a helical shape but from the topological configuration of their entanglement, which is determined by the sequence and orientation of the strand crossings (Fig. 1c, d). This form of topological chirality constitutes a distinct level of spatial organization, one that appears to confer exceptional stability on the resulting spin selectivity. Experimentally, molecules that display this knot-based chirality achieve spin polarization as high as 90%, accompanied by an increase in conductivity of roughly two orders of magnitude compared to traditional chiral molecules, and retain performance up to temperatures near 350∘C19, exhibiting various advantages over traditional CISS1–6. These experimental observations establish topologically chiral knots as a distinctive platform for achieving enhanced and robust CISS. Although pioneering experimental work has presented a theoretical model to explain the highly spin-polarized current in these systems, the model itself remains grounded in a simplified picture, that of a uniform helical chain with structural chirality19. This representation, although valuable, does not fully capture the spatial variations and specific geometrical arrangement inherent in a knot topology. As a result, it offers limited insight into the distinctive physical processes that govern CISS in these entangled structures. A dedicated theoretical framework that directly addresses topological chirality is still needed to clarify the underlying microscopic mechanisms.Fig. 1Schematic descriptions of two different chiral materials to exhibit spin selectivity.a The conventional CISS device structured on a single helix chain with structural chirality. b The new-developed CISS device based on the trefoil knot molecules characterized by topological chirality. c Paired enantiomers of real trefoil knot molecules possessing opposite chirality exampled by C193H237N15O20S8. d Trefoil knots with opposite topological chirality depending on the spatial arrangements of crossings. Schematic descriptions of two different chiral materials to exhibit spin selectivity. a The conventional CISS device structured on a single helix chain with structural chirality. b The new-developed CISS device based on the trefoil knot molecules characterized by topological chirality. c Paired enantiomers of real trefoil knot molecules possessing opposite chirality exampled by C193H237N15O20S8. d Trefoil knots with opposite topological chirality depending on the spatial arrangements of crossings. In this work, we construct a series of spin-dependent transport devices based on trefoil knot molecules to establish a theoretical framework for the CISS effect in topologically complex systems. A central challenge is to clarify the origin of spin polarization in such structures. Although established mechanisms, including intramolecular spin-orbit coupling (SOC)37,38,48,49, chiral-induced orbital polarization43, spin-interface interactions50–52, and non-Hermitian skin effects under open boundaries46, have been invoked to explain CISS in other chiral molecules, they do not directly account for the role of global molecular topology. Recent studies highlight that SOC in chiral molecules is atomic in nature but also a component related to molecular geometry and chirality53–55. In trefoil knots19,56, the three-dimensionally curved and closed transport pathways further enrich the SOC mechanism. Here, SOC emerges from the non-relativistic limit of the Dirac equation in a curved space, intrinsically linked to the geometry of the twisted electron transport paths. This form of SOC, termed geometric SOC57,58, is different from conventional Rashba or atomic SOC in its global geometric origin and strong spatial variation along the knotted contour. Moreover, its magnitude can significantly exceed that of conventional SOC57, providing a physical basis for ultra-high spin polarization (up to 90%), observed experimentally in trefoil-knot molecules19. Results Theoretical model with geometric SOC From a theoretical modeling perspective, the trefoil knot molecule can be simplified as a closed helical configuration with continuously varying curvature, forming its characteristic topological knot structure (Fig. 1d). The three-dimensional twisted pathways in this configuration increase the spatial degrees of freedom available for electron transport, giving rise to complex SOC features. Crucially, conventional SOC models for CISS, often derived from the Dirac equation in a flat spacetime electromagnetic field37,38,48, are fundamentally inapplicable to these continuously deforming trefoil structures. Here, we demonstrate that the trefoil-knot arrangement of monomers generates geometric SOC. Specifically, within the helical structure of trefoil knot molecules, quantum interference fundamentally occurs between direct (i-j) and indirect (i-k-j) paths for electron transmission, introducing an extra spin-dependent Berry phase during electron motion (see “Method”). This phenomenon constitutes the microscopic origin of geometric SOC, whose form is determined by molecular chirality. Like a conventional SOC, this geometric SOC can induce spin flipping during electron transport. To quantitatively characterize this mechanism, we construct a closed lattice-knot model in which the molecule is discretized into N lattice sites spaced at equal arc length (positions of sites 1 and N are shown in Fig. 2a; see also Supplementary Movie 1). The spatial coordinates of the ith site are given by 1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\bf{R}}}}_{i}={{{\bf{R}}}}_{i}({\theta }_{i})={[r\sin {\theta }_{i}+R\sin (2{\theta }_{i}),\zeta r\cos {\theta }_{i}-\zeta R\cos (2{\theta }_{i}),h\sin (3{\theta }_{i})]}^{T},$$\end{document}Ri=Ri(θi)=[rsinθi+Rsin(2θi),ζrcosθi−ζRcos(2θi),hsin(3θi)]T, where R+r and R−r denote the maximum and minimum radii, respectively, in the top view projection of the trefoil knot; ζ symbolizes the chirality of the molecular (left handed when ζ=1 and right handed when ζ=–1); h represents the molecular height and θi is the angle coordinate of the ith lattice. Ti and Ni are defined as the unit tangent and normal vectors of the twisted path (Supplementary Section A). Electrons traversing this trefoil knot experience dynamics analogous to a noninertial system with finite acceleration ai = − κc2Ni (where c is the speed of light), inducing the geometric SOC term Hgeo = ℏ(a × p) ⋅ σ/(4mc2) = − Σiℏκi(Ni × pi) ⋅ σ/(4m) (where pi is the electron momentum, m is electron mass, and σ is the Pauli vector) in the equation. Further calculation demonstrates that this geometric SOC is explicitly governed by lattice-dependent curvature κi, which can be calculated by \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\kappa }_{i}=| {{{\bf{R}}}}_{i}^{{\prime} }\times {{{\bf{R}}}}_{i}^{{\prime\prime} }| /| {{{\bf{R}}}}_{i}^{{\prime} }{| }^{3}$$\end{document}κi=∣Ri′×Ri″∣/∣Ri′∣3. To validate this, we compute the real-space distribution of κi for a trefoil knot with R: h = 2: 1 (Fig. 2b). Furthermore, κi exhibits strict 2π/3-periodicity versus θi in all models (Fig. 2c, d), confirming that the magnitude of κi, directly reflecting the twisted geometry in 3D and the rotational symmetry of the knot with C3, dictates the modulation of the SOC strength through geometric scaling.Fig. 2Geometrical model and spatial curvature of trefoil knots.a A trefoil knot model composed by N lattices (top view), where the first and the last lattice sites are denoted, and the tangent (Ti), normal (Ni), and binormal vector (Bi) are represented by red, black, magenta arrows, respectively. b the lattice-dependent weigh distribution of spatial curvature κ, where four structural parameters R, r, θ, and h are denoted accordingly. c, d The spatial curvature κ versus the helix angle θ in the six representative trefoil knot models, with the specific structures drawn in Supplementary Fig. 1. Geometrical model and spatial curvature of trefoil knots. a A trefoil knot model composed by N lattices (top view), where the first and the last lattice sites are denoted, and the tangent (Ti), normal (Ni), and binormal vector (Bi) are represented by red, black, magenta arrows, respectively. b the lattice-dependent weigh distribution of spatial curvature κ, where four structural parameters R, r, θ, and h are denoted accordingly. c, d The spatial curvature κ versus the helix angle θ in the six representative trefoil knot models, with the specific structures drawn in Supplementary Fig. 1. For this trefoil knot model, given that \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\mathcal{H}}}}_{\rm geo}$$\end{document}Hgeo is determined, the overall tight-binding Hamiltonian can be formulated as \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{\mathcal{H}}}={{{\mathcal{H}}}}_{\rm mol}+{{{\mathcal{H}}}}_{\rm geo}+{{{\mathcal{H}}}}_{\rm el}+{{{\mathcal{H}}}}_{\rm d}$$\end{document}H=Hmol+Hgeo+Hel+Hd, where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\mathcal{H}}}}_{\rm mol}=\widehat{T}+\widehat{V}=\frac{{{{\bf{p}}}}^{2}}{2m}+V(x,y,z)$$\end{document}Hmol=T^+V^=p22m+V(x,y,z). Then, we discretize the Hamiltonian \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\mathcal{H}}}}_{\rm mol}$$\end{document}Hmol and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\mathcal{H}}}}_{\rm geo}$$\end{document}Hgeo within the Wannier representation, and the discretization form of the former is given by 2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\mathcal{H}}}}_{\rm mol}={\sum }_{n=1}^{N}{\varepsilon }_{n}{c}_{n}^{{\dagger} }{c}_{n}+{\sum }_{n=1}^{N}({t}_{n}{c}_{n}^{{\dagger} }{c}_{n+1}+{\rm H.c.})$$\end{document}Hmol=∑n=1Nεncn†cn+∑n=1N(tncn†cn+1+H.c.) where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${c}_{n}^{{\dagger} }=({c}_{n,\uparrow }^{{\dagger} },{c}_{n,\downarrow }^{{\dagger} })$$\end{document}cn†=(cn,↑†,cn,↓†) and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\varepsilon }_{n}=\langle {\Psi }_{n}| {{{\mathcal{H}}}}_{\rm mol}| {\Psi }_{n}\rangle $$\end{document}εn=⟨Ψn∣Hmol∣Ψn⟩ represent the creation operator and the onsite energy located at the nth lattice site, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${t}_{n}=\langle {\Psi }_{n}| {{{\mathcal{H}}}}_{\rm mol}| {\Psi }_{n+1}\rangle $$\end{document}tn=⟨Ψn∣Hmol∣Ψn+1⟩ signifies the nearest-neighbor hopping integral between the nth and (n+1)th lattice sites and can be regarded as a constant. Here, Ψn is the electron wave function at the nth lattice site. Similarly, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\mathcal{H}}}}_{\rm geo}$$\end{document}Hgeo can be discretized as 3\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\mathcal{H}}}}_{\rm geo}=i\frac{{\lambda }_{soc}}{2}{\sum }_{n=1}^{N}{c}_{n}^{{\dagger} }({\kappa }_{n}+{\kappa }_{n+1})({\sigma }_{n}+{\sigma }_{n+1}){c}_{n+1}+{\rm H.c.}$$\end{document}Hgeo=iλsoc2∑n=1Ncn†(κn+κn+1)(σn+σn+1)cn+1+H.c. where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\lambda }_{\rm soc}=\frac{{\hslash }^{2}}{16mS}$$\end{document}λsoc=ℏ216mS with S being the arc length between two adjacent lattices (Supplementary Section A). Within the framework of (Ti, Ni, Bi) and for each lattice, Bi is given by Bi = Ti × Ni. This unit vector has three components, namely Bi = (Bix, Biy, Biz), and then the Pauli matrix has the relation σi = σxBix + σyBiy + σzBiz, where (σx, σy, σz) denotes the three dependent Pauli matrix. Unless otherwise specified, we adopt λsoc=0.737t ⋅ nm to perform all calculations. For example, t = 0.3 eV, resulting in λsoc ≈ 0.221 (eV ⋅ nm), satisfies well the basic properties of electronic hopping and the geometric SOC established here in trefoil knot molecules. To simulate the hopping of electrons between trefoil knot molecules, STM probes, and metal substrates, we adopt the coupling parameters Γγ, with γ indicating the number of semi-infinite metal electrodes, to design a multi-terminal device model. Then, the Hamiltonian to describe the electron tunneling between the electrodes and the molecules can be described by \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\mathcal{H}}}}_{\rm el}={\sum }_{n,\gamma }{t}_{n}^{{\prime} }{c}_{n,\gamma }^{{\dagger} }{c}_{n+1,\gamma }+{\sum }_{i,\gamma }{\Gamma }_{\gamma }{c}_{0,\gamma }^{{\dagger} }{c}_{i}+{\rm H.c.}$$\end{document}Hel=∑n,γtn′cn,γ†cn+1,γ+∑i,γΓγc0,γ†ci+H.c. (n≥0; i: lattice site in knot). \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\mathcal{H}}}}_{\rm d}$$\end{document}Hd represents the naturally occurring dephasing Hamiltonian in experiments. The dephasing process can arise from electron-phonon and electron-electron interactions. This inelastic scattering results in the loss of electron phase memory and can be simulated by connecting each position of the molecule to Büttiker’s virtual electrodes. Subsequently, under the boundary condition of zero net current on each virtual electrode, the spin-up conductance G↑ and spin-down conductance G↓ can be calculated by combining the Landauer-Büttiker formula and non-equilibrium Green’s functions, and then the spin polarization is defined as \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${P}_{s}=\frac{{G}_{\uparrow }-{G}_{\downarrow }}{{G}_{\uparrow }+{G}_{\downarrow }}$$\end{document}Ps=G↑−G↓G↑+G↓ (see Methods). Fundamental conditions for CISS in knot molecules Using the Hamiltonian model described above, we identify the key conditions for spin-polarized electron transport through trefoil knot molecules. We begin by analyzing a two-terminal setup (γ = 2) in which a trefoil knot molecule (N = 60 lattice sites) bridges two non-magnetic metal electrodes (Fig. 3a, top). Theoretical calculations reveal that when the electrodes are coupled to the lattice sites i = 1 and j = 31, dividing the molecule into two perfectly symmetric halves, the system exhibits high conductivity but no spin polarization in the transmitted charge current (Fig. 3a, bottom). This behavior comes from the presence of two symmetric tunneling channels. Electrons passing through these channels in the curved space acquire Berry phases \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${e}^{-i{\Phi }_{L}}$$\end{document}e−iΦL and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${e}^{-i{\Phi }_{R}}$$\end{document}e−iΦR (L, R denote left and right pathways), which are equal in magnitude but opposite in sign. The resulting phase cancellation suppresses quantum interference, and consequently spin polarization, a mechanism previously reported in closed helical molecular rings59–61, although the origins of SOC differ. We further test an open trefoil knot configuration, where the knot is cut at one point, and electrodes are placed at the resulting terminals (Fig. 3a, right). Numerical simulations show that although conductivity remains high (Supplementary Fig. 2), spin polarization remains negligible. These results confirm that a single transport channel through the trefoil knot is insufficient to generate spin polarization, consistent with an earlier theoretical conclusion drawn for single-channel helical systems38.Fig. 3Spin-dependent conductance (G↑ and G↓) and spin polarization Ps in knot molecule-based devices.a Two two-terminal device configurations (top view): left device (1,31)60 exhibits minor symmetry, while the right device features an open trefoil knot with ends coupled to electrodes; low panels show G↑, G↓ and Ps. b Asymmetric two-terminal devices (1, 31)60 and (1, 47)60 with G↑, G↓ and Ps plotted below. c Three-terminal devices (7;14,40)60 with opposite chirality and corresponding G↑, G↓ below. d–f Three-terminal devices (7;15,36)60, (7;36,56)60 and (7;15,56)60 with corresponding spin polarization. Spin-dependent conductance (G↑ and G↓) and spin polarization Ps in knot molecule-based devices. a Two two-terminal device configurations (top view): left device (1,31)60 exhibits minor symmetry, while the right device features an open trefoil knot with ends coupled to electrodes; low panels show G↑, G↓ and Ps. b Asymmetric two-terminal devices (1, 31)60 and (1, 47)60 with G↑, G↓ and Ps plotted below. c Three-terminal devices (7;14,40)60 with opposite chirality and corresponding G↑, G↓ below. d–f Three-terminal devices (7;15,36)60, (7;36,56)60 and (7;15,56)60 with corresponding spin polarization. When the two electrodes are attached to lattice sites that break the local left-right symmetry of the trefoil knot, the Berry phases induced by spatial curvature no longer cancel. This leads to the emergence of spin polarization, as illustrated in Fig. 3b. Specifically, in the asymmetric two-terminal configuration (1,15)60, the system maintains robust conductivity (Supplementary Fig. 2) and exhibits spin polarization of conducting electrons up to 20%. In particular, the mirror-symmetric configuration (1,47)60 shows spin polarization of the same magnitude but opposite sign (Fig. 3b, lower panel). This inversion of the spin direction upon spatial mirroring reflects the intrinsic C3 symmetry of the trefoil knot and strongly confirms that the geometric SOC term correctly captures the relationship between molecular geometry and spin-dependent transport. These results establish that asymmetric multichannel transport is necessary for spin selectivity in trefoil knots, with geometric SOC providing a key mechanism for spin polarization in these topologically chiral knot molecules. Given the distinctive geometry of the trefoil knot molecule and the experimental setup typically used for current measurements (Fig. 1b), this molecular system presents three symmetrically positioned lower lattice sites that can be coupled to a metal substrate, while any one of the three upper sites may connect to an STM probe. To simulate this arrangement, we construct a three-terminal device (γ = 3) based on a trefoil knot with N = 60 lattice sites: two of its lower sites are connected to the metal substrate, and one upper site (i = 7) is coupled to the STM probe. We examine three different coupling configurations for the lower electrode pairs, corresponding to the structures shown in Fig. 3d (j = 15, 36), Fig. 3e (j = 36, 56), and Fig. 3f (j = 15, 56). Across these three configurations, the spatial symmetry between the left and right transport channels formed by the lower electrodes relative to the top electrode progressively decreases. The calculation results show a corresponding sequential enhancement in spin polarization, from approximately 20 to 75%, providing direct evidence that reduced symmetry in transport channels enhances spin selectivity. Robustness of high-spin polarization in knot molecules To further elucidate the origin of high spin polarization in trefoil knot molecules, we model a four-terminal transport device (γ = 4) that mirrors experimental setups for charge current measurement19. In this configuration, the three lowest lattice sites are connected to a metal substrate, while the uppermost site is coupled to an STM probe (Fig. 4a). We first consider a trefoil knot with N = 20 lattice sites, coupled at positions i = 3, j = 6, 14, 20. The calculated conductances G↑ and G↓ reveal a spin polarization of up to 60% (Fig. 4b, top). Reversing the molecular chirality from left- to right-handed also reverses the spin polarization direction, confirming that the effect is intrinsically tied to the topological chirality of the knot. Moreover, the four-terminal setup exhibits high electrical conductivity, especially near the Fermi level, where finite conductance indicates metallic transport behavior (Supplementary Fig. 3). This contrasts with conventional chiral materials such as DNA7 and helicenes8, which typically show low conductance due to significant band gaps. The observed metallic conduction is consistent with experimental reports of nearly two orders of magnitude conductivity enhancement relative to traditional chiral systems. Additional simulations for a knot with N = 60 lattice sites (at constant molecular length) show that the maximum spin polarization remains around 60% (Fig. 4b, bottom). In particular, this value is slightly lower than the ~75% polarization obtained in the analogous three-terminal model with the same number of lattice sites. This decrease can be attributed to the additional electrode improving the overall symmetry of the transport pathways, consistent with the C3 rotational symmetry of the trefoil knot, which partially counteracts the spin-dependent Berry phase generated by spatial curvature and thus moderates the resulting spin polarization.Fig. 4Spin polarization of trefoil knot molecules versus the lattice density.a A typical four-terminal device to simulate experiment measurements. b Spin polarization of two typical trefoil knot models composed by N = 20, and 60 lattices, where the curved length is fixed. c Spin polarization of trefoil knot molecules versus the total lattice number N. Spin polarization of trefoil knot molecules versus the lattice density. a A typical four-terminal device to simulate experiment measurements. b Spin polarization of two typical trefoil knot models composed by N = 20, and 60 lattices, where the curved length is fixed. c Spin polarization of trefoil knot molecules versus the total lattice number N. In particular, in the four-terminal device based on the trefoil knot, spin polarization remains stable at approximately 60% as the total lattice density increases from N = 20 to 100, while the overall molecular length remains constant (Fig. 4c; Supplementary Movie 1). These results indicate that increasing lattice resolution does not significantly alter the spin polarization. Rather, the polarization is governed primarily by the intrinsic geometric configuration of the molecule and the coupling scheme, specifically the number and placement of electrodes, rather than by the total number of discrete lattice sites. This behavior stands in sharp contrast to conventional CISS effects observed in helical molecular chains, where spin polarization often varies considerably, typically increasing with either the number of lattice points per unit length or the total chain length. Together, these findings suggest that the robust spin selectivity in trefoil knots stems from the molecule’s intrinsic topological character: the stable knotted geometry enforces a form of topological invariance, which in turn ensures consistent spin polarization across different lattice discretizations. To assess the stability of high spin polarization observed in trefoil knot molecules, we examine a four-terminal device (γ = 4) based on a N = 60 trefoil knot with initial parameters l1=l2 = 1.0 and h = 1.0 (central structures in Fig. 5). Two different strain protocols are applied to the 3D knot structure: (i) lateral strain, in which the height is fixed (h = 1.0) while one side is asymmetrically compressed (l1 increases from 1.0 to 2.0, l2 remains 1.0; Fig. 5a, left) and (ii) vertical strain, in which the lateral dimensions are maintained while the height is stretched (h increases from 1.0 to 2.0; Fig. 5a, right). The numerical results show that lateral deformation does not cause significant changes in conductivity (Supplementary Fig. 4) or spin polarization, which remains close to ~ 60%. Similarly, under vertical strain, spin polarization remains largely unchanged compared to the unstrained knot (Fig. 5a, right curves). The preservation of high spin selectivity under both deformation modes underscores the structural robustness conferred by the knot topology. These findings offer a theoretical basis for experimental reports of high-temperature spin polarization stability in trefoil knots: lattice vibrations, which induce dynamic structural distortions, do not disrupt the spin-selectivity transport pathways protected by the underlying topological configuration. This resilience to structural variation distinguishes trefoil knots from conventional chiral materials in the context of CISS, highlighting their potential for spintronic applications where size-independent robustness is essential.Fig. 5Structural regulation of spin polarization in trefoil knot molecules.a Strain effects: Left - Lateral strain varies l1 (1.0→2.0, top view) while l2 and h is fixed; Right - Vertical strain increases height h (1.0→2.0, side view) with fixed l1 and l2. b Chirality phase transition: Left - reducing structural parameter R (2.0→1.0) collapses the hollow triangle to a crossing point and reopens as a trivial ring with further reducing R (1.0→0.0, top view); Right - further reducing the ring height h (1.0→0.0). All systems share an identical lattice number (N = 60 in (a) and N = 50 in (b). Structural regulation of spin polarization in trefoil knot molecules. a Strain effects: Left - Lateral strain varies l1 (1.0→2.0, top view) while l2 and h is fixed; Right - Vertical strain increases height h (1.0→2.0, side view) with fixed l1 and l2. b Chirality phase transition: Left - reducing structural parameter R (2.0→1.0) collapses the hollow triangle to a crossing point and reopens as a trivial ring with further reducing R (1.0→0.0, top view); Right - further reducing the ring height h (1.0→0.0). All systems share an identical lattice number (N = 60 in (a) and N = 50 in (b). Topological phase transition of CISS in knot molecules To clarify the role of knot topology in stabilizing high spin polarization, we systematically vary the structural parameter R in a trefoil knot with N = 50 lattice sites. Changing R progressively suppresses the knot geometry and ultimately removes its topological character, allowing us to track the corresponding evolution of spin polarization. As R decreases from 2.0 to 1.1, the central triangular region contracts while the knot topology remains intact (Structure 1, Fig. 5b; Supplementary Movie 2). In this regime, spin polarization stays stable near 60%. At the critical value R = 1.0, the central region collapses to a crossing point (Structure 2, Fig. 5b) and the spin polarization drops markedly, although it remains finite. When R is further reduced to 0.9, the topological knot is eliminated: the helical ring opens into a trivial conformation that can be continuously deformed into a simple annulus, while maintaining a fan-like structural chirality (Structures 3–4, Fig. 5b and Supplementary Movie 2). This structural change corresponds to a topological transition in which topological chirality gives way to structural chirality. Upon completing this transition, spin polarization falls to approximately 20%, and conductivity is substantially reduced (Supplementary Fig. 5). Sequential suppression of spin polarization as the knot topology is removed provides clear evidence that the previously observed high polarization (>60%) is protected by the nontrivial topology of the trefoil knot. As the parameter R gradually decreases, the trefoil knot deforms into a fan-shaped geometry. At R = 0, the structure opens completely into a vertically oriented circular ring (Structures 5 and 6, top view, Fig. 5b and Supplementary Movie 2). Although the knot topology is lost in this limit, the persistence of asymmetric transport paths with non-zero curvature maintains a finite degree of spin polarization in the transmitted electrons. In contrast, reducing the height parameter h flattens the molecular structure (Structures 5-6, side view, Fig. 5b). When h reaches zero, the system degenerates into a circular planar ring with uniform curvature, and the spin polarization decays asymptotically to zero. This marked difference in behavior reveals that spin selectivity in structurally chiral systems is highly sensitive to geometric deformation, a property that fundamentally distinguishes it from the robust spin polarization observed in topologically chiral knots. Importantly, the geometric SOC introduced in this work offers a unified physical framework that accounts for both conventional CISS in helical chains and robust CISS in trefoil knots. Discussion This study presents a theoretical framework to explain the microscopic origin of ultra-high spin polarization in trefoil knot molecules. The underlying mechanism involves two key elements: asymmetric multiple transport channels and geometric SOC. The latter stems from the accumulation of a quantum geometric phase in a curved configuration space, which distinguishes it from conventional SOC typically used to explain spin selectivity in chiral chains. We develop multi-terminal device models based on a tight-binding Hamiltonian that incorporates both electronic hopping and the geometric SOC term. Theoretical results show that when electrode coupling breaks spatial symmetry while maintaining the connectivity of the molecular knot, substantial spin polarization emerges. Simulations performed at the single-particle level, without additional electron correlation effects, reproduce spin polarization exceeding 60% with a direction governed solely by the topological chirality of the knot, alongside consistently high electrical conductivity. These theoretical predictions align with and help explain recent experimental reports19, establishing a coherent microscopic picture of spin selectivity in topologically chiral materials. Theoretical results further show that high spin polarization (>60%) and high conductivity remain stable under changes in lattice site density when the overall molecular length is fixed. These properties also persist under considerable structural deformation, with lateral and vertical strains of up to 200% causing no significant degradation. In contrast, suppressing the topological knot structure, leading to its degeneration into a trivial knot along with a transition from topological to structural chirality, results in a pronounced decrease in spin polarization. This clear dependence on topological integrity confirms that the observed spin selectivity in trefoil knots is protected by the knot topology, endowing it with notable robustness. This inherent stability provides a natural explanation for experimental observations of sustained high spin polarization at elevated temperatures. Our theoretical model predicts a maximum spin polarization of approximately 60% in a single trefoil knot molecule, which is significantly lower than the 90% polarization reported in recent experiments19. This difference can be attributed to several important factors that are not included in our idealized model. First, in the simulated single-molecule device, only one upper lattice site is connected to the STM tip, whereas experimentally, the tip may interact with multiple sites or even several molecules simultaneously, enhancing the overall spin-selective signal. More significantly, experimental samples typically consist of molecular thin films with thicknesses ranging from 2 to 20 nm, in which spin-polarized electrons traverse multiple molecular layers. Such a multilayer arrangement can substantially amplify spin filtering through sequential polarization steps. Additional physical mechanisms that are not captured in our current single-particle treatment, such as electron-electron interactions that may enhance spin correlations41 and electron-phonon coupling that could influence spin relaxation dynamics42, are also likely to contribute to higher polarization in real materials. Incorporating these structural and many-body effects into future models is expected to bring theoretical predictions closer to the 90% polarization observed experimentally, thus reconciling the current discrepancy while affirming the physical consistency of the proposed geometric SOC mechanism. The theoretical framework presented here offers a consistent explanation for variations in spin polarization across systems with different types of chirality, including the transition from topological to structural chirality. Although developed for trefoil knots, the model can be naturally extended to other topologically chiral molecules, such as the cinquefoil knot (51)62,63, suggesting that molecular knots may form a versatile platform for exploring spin-dependent transport. More broadly, the approach introduced in this work, centered on geometric SOC and multi-terminal transport, provides a conceptual connection between different classes of chiral materials. By identifying common physical elements underlying spin selectivity in both conventional helical and topologically knotted systems, this study contributes to the ongoing effort toward a unified understanding of CISS phenomena. Methods The derivation of geometric SOC in curved pathways The geometric SOC in a curved molecular structure, such as a trefoil knot, can be derived starting from a continuum Hamiltonian of the form 4\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\mathcal{H}}}}_{\rm geo}=\frac{\hslash }{4m{c}^{2}}({{\bf{a}}}\times {{\bf{p}}})\cdot {{\boldsymbol{\sigma }}},$$\end{document}Hgeo=ℏ4mc2(a×p)⋅σ, where a= −κc2N is an effective gauge field related to curvature, p is the momentum operator, and σ denotes the Pauli matrices. In the Frenet-Serret frame, the momentum is decomposed into tangential and normal components such as p = p∥T + p⊥N. Substituting this expression into the Hamiltonian leads to 5\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\mathcal{H}}}}_{\rm geo}={\sum }_{i}\frac{\hslash }{4m}{\kappa }_{i}{p}_{\parallel }{{{\bf{B}}}}_{i}\cdot {{{\boldsymbol{\sigma }}}}_{i},$$\end{document}Hgeo=∑iℏ4mκip∥Bi⋅σi, where Bi is a local axial vector depending on the Frenet frame, and σi = σxBix + σyBiy + σzBiz. To ensure Hermiticity, the product p∥σi is symmetrized, giving the following form 6\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\mathcal{H}}}}_{\rm geo}={\sum }_{i}\frac{\hslash }{8m}{\kappa }_{i}({p}_{\parallel }{\sigma }_{i}+{\sigma }_{i}{p}_{\parallel }).$$\end{document}Hgeo=∑iℏ8mκi(p∥σi+σip∥). We then proceed to the second quantization. Using the tangential momentum operator p∥=\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-i\hslash \frac{\partial }{\partial T}$$\end{document}−iℏ∂∂T and applying a discretization of finite difference on a discrete chain of sites, we approximate the action of p∥ on a lattice wave function Fn as 7\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\widehat{p}}_{\parallel }{F}_{n}=-i\hslash \frac{{F}_{n+1}-{F}_{n-1}}{2S},$$\end{document}p^∥Fn=−iℏFn+1−Fn−12S, where S is the length of the arc between two neighboring sites. After algebraic manipulation, the second-quantized geometric SOC Hamiltonian takes the form 8\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\mathcal{H}}}}_{\rm geo}=i\frac{{\lambda }_{soc}}{2}{\sum }_{n=1}^{N}{c}_{n}^{{\dagger} }({\kappa }_{n}+{\kappa }_{n+1})({\sigma }_{n}+{\sigma }_{n+1}){c}_{n+1}+{\rm H.c.}$$\end{document}Hgeo=iλsoc2∑n=1Ncn†(κn+κn+1)(σn+σn+1)cn+1+H.c. Here, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${c}_{n}^{{\dagger} }$$\end{document}cn† and cn are the creation and annihilation operators at the site n, κn is the local curvature, σn denotes the Pauli vector and λsoc is the effective coupling strength of geometric SOC. Further details of the derivation are provided in Supplementary Section A. Spin-polarized conductance and spin polarization We describe the theoretical approach for calculating spin-resolved conductance using a two-terminal device (γ = 2) as an example. The device model includes a trefoil knot molecule coupled to left (L) and right (R) metal electrodes, with additional semi-infinite virtual probes attached to each lattice site n to model phase-breaking processes and ensure current conservation. Within the Landauer-Büttiker framework, the charge current Iν (ν = L, R) flowing in the physical electrode and the current In in the virtual electrode n can be computed below 9\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${I}_{\nu }=\frac{e}{\hslash }{\sum }_{\alpha,\sigma }{\int }_{-\infty }^{\infty }{T}_{\nu,\alpha }^{\sigma }[\;{f}^{\nu }(E)-{f}^{\alpha }(E)]dE,\\ {I}_{n}=\frac{e}{\hslash }{\sum }_{\alpha,\sigma }{\int }_{-\infty }^{\infty }{T}_{n,\alpha }^{\sigma }[\;{f}^{n}(E)-{f}^{\alpha }(E)]dE,$$\end{document}Iν=eℏ∑α,σ∫−∞∞Tν,ασ[fν(E)−fα(E)]dE,In=eℏ∑α,σ∫−∞∞Tn,ασ[fn(E)−fα(E)]dE, where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${T}_{\nu,\alpha }^{\sigma }(E)$$\end{document}Tν,ασ(E) denotes the spin-dependent transmission probability from electrode α to ν, and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${f}^{\mu }(E)={[{e}^{{\beta }_{\mu }(E-{\mu }_{\mu })}+1]}^{-1}$$\end{document}fμ(E)=[eβμ(E−μμ)+1]−1 is the Fermi-Dirac distribution in the electrode μ, with \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\beta }_{\mu }={({k}_{B}{T}_{\mu })}^{-1}$$\end{document}βμ=(kBTμ)−1 and chemical potential μμ. To determine the local chemical potentials μn in the virtual probes, we impose the condition In = 0 for all n, leading to a system of nonlinear equations 10\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\sum }_{\alpha,\sigma }{\int }_{-\infty }^{\infty }{T}_{n,\alpha }^{\sigma }{f}^{n}(E)dE={\sum }_{\alpha,\sigma }{\int }_{-\infty }^{\infty }{T}_{n,{n}^{{\prime} }}^{\sigma }{f}^{{n}^{{\prime} }}(E)dE+{\sum }_{\alpha,\sigma }{\int }_{-\infty }^{\infty }{T}_{n,\nu }^{\sigma }{f}^{\nu }(E)dE$$\end{document}∑α,σ∫−∞∞Tn,ασfn(E)dE=∑α,σ∫−∞∞Tn,n′σfn′(E)dE+∑α,σ∫−∞∞Tn,νσfν(E)dE To simplify current calculations, we restrict the calculation of local current In within the scope of the linear response for this matrix equation. Furthermore, under low-voltage conditions, we can employ the Taylor series to expand the Fermi-Dirac distribution function to the first nontrivial order, thereby facilitating the approximation of electronic transport characteristics. The Fermi-Dirac distribution function can be approximately expanded as 11\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${f}^{\alpha }(E,{\mu }_{\alpha })={f}^{eq}(E,{F}_{F})-\frac{\partial {f}^{eq}(E,{E}_{F})}{\partial E}({\mu }_{\alpha }-{E}_{F}),({\mu }_{eq}={E}_{F}),$$\end{document}fα(E,μα)=feq(E,FF)−∂feq(E,EF)∂E(μα−EF),(μeq=EF), where EF is the Fermi energy. Setting EF = 0 and substituting it into the current conservation equations yields a linear system 12\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\mu }_{n}{\sum }_{\alpha,\sigma }{\int }_{-\infty }^{\infty }{T}_{n,\alpha }^{\sigma }(-\frac{\partial {f}^{eq}(E,{E}_{F})}{\partial E})-{\sum }_{{n}^{{\prime} },\sigma }^{N}{\int }_{-\infty }^{\infty }{T}_{n,{n}^{{\prime} }}^{\sigma }(-\frac{\partial {f}^{eq}(E,{E}_{F})}{\partial E})\\={\int }_{-\infty }^{\infty }(-\frac{\partial {f}^{eq}(E,{E}_{F})}{\partial E})[{T}_{n,L}^{\sigma }(E){\mu }_{L}+{T}_{n,R}^{\sigma }(E){\mu }_{R}].$$\end{document}μn∑α,σ∫−∞∞Tn,ασ(−∂feq(E,EF)∂E)−∑n′,σN∫−∞∞Tn,n′σ(−∂feq(E,EF)∂E)=∫−∞∞(−∂feq(E,EF)∂E)[Tn,Lσ(E)μL+Tn,Rσ(E)μR]. In this way, the matrix equation comprising n equations can be reformulated as an equation \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\widehat{M}\cdot \mu=\nu $$\end{document}M^⋅μ=ν, and the solutions to this matrix equation can be obtained by inversion of the matrix. Furthermore, the solution of this equation will allow us to determine the specific form of the chemical potential associated with the probe coupled to each lattice site. The matrix \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\widehat{M}$$\end{document}M^, in this context, can be expressed as 13\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{\bf{M}}}=\left[\begin{array}{ccc}{\sum }_{\alpha \ne 1,\nu,\sigma }{\int }_{-\infty }^{\infty }dE{T}_{1,\alpha }^{\sigma }(E)(-\frac{\partial {f}^{\alpha }(E,{E}_{F})}{\partial E}) & \ldots & -{\sum }_{\sigma }{\int }_{-\infty }^{\infty }dE{T}_{1,n}^{\sigma }(E)(-\frac{\partial {f}^{\alpha }(E,{E}_{F})}{\partial E})\\ -{\sum }_{\sigma }{\int }_{-\infty }^{\infty }dE{T}_{2,1}^{\sigma }(E)(-\frac{\partial {f}^{\alpha }(E,{E}_{F})}{\partial E}) & \ddots & -{\sum }_{\sigma }{\int }_{-\infty }^{\infty }dE{T}_{2,n}^{\sigma }(E)(-\frac{\partial {f}^{\alpha }(E,{E}_{F})}{\partial E})\\ \ldots & \ldots & {\sum }_{\alpha \ne n,\nu,\sigma }{\int }_{-\infty }^{\infty }dE{T}_{n,\alpha }^{\sigma }(E)(-\frac{\partial {f}^{\alpha }(E,{E}_{F})}{\partial E})\end{array}\right].$$\end{document}M=∑α≠1,ν,σ∫−∞∞dET1,ασ(E)(−∂fα(E,EF)∂E)…−∑σ∫−∞∞dET1,nσ(E)(−∂fα(E,EF)∂E)−∑σ∫−∞∞dET2,1σ(E)(−∂fα(E,EF)∂E)⋱−∑σ∫−∞∞dET2,nσ(E)(−∂fα(E,EF)∂E)……∑α≠n,ν,σ∫−∞∞dETn,ασ(E)(−∂fα(E,EF)∂E). By solving the matrix equation mentioned above, we can determine the chemical potential of the virtual probe and subsequently calculate the spin-dependent charge currents flowing from the metal electrode L as follows 14\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${I}_{L,\sigma }=\frac{{e}^{2}}{\hslash }{\sum }_{\alpha \ne 1,\nu }{T}_{L,n}^{\sigma }(E)[-\frac{\partial {f}^{\alpha }(E,{E}_{F})}{\partial E}]({\mu }_{L}-{\mu }_{R}).$$\end{document}IL,σ=e2ℏ∑α≠1,νTL,nσ(E)[−∂fα(E,EF)∂E](μL−μR). Within the linear response regime, the conductivity is simply given by \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${G}_{\sigma }=\frac{{I}_{L,\sigma }}{\Delta V}$$\end{document}Gσ=IL,σΔV, where ΔV = μR − μL. Furthermore, under low-temperature conditions, the Fermi distribution function reduces to the Dirac delta function, and the spin conductivity can be represented as follows 15\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${G}_{\sigma }=\frac{{e}^{2}}{\hslash }[{T}_{L,R}^{\sigma }({E}_{F})+{\sum }_{n,\sigma }{T}_{L,n}^{\sigma }({\mu }_{L}-{\mu }_{n})/\Delta \mu ],$$\end{document}Gσ=e2ℏ[TL,Rσ(EF)+∑n,σTL,nσ(μL−μn)/Δμ], Thus, we will derive the specific expression for spin polarization Ps as 16\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${P}_{s}=\frac{{G}_{\uparrow }-{G}_{\downarrow }}{{G}_{\uparrow }+{G}_{\downarrow }}.$$\end{document}Ps=G↑−G↓G↑+G↓. Supplementary information Supplementary Information Description of Additional Supplementary Files Supplementary Movie 1 Supplementary Movie 2 Transparent Peer Review file Supplementary Information Description of Additional Supplementary Files Supplementary Movie 1 Supplementary Movie 2 Transparent Peer Review file Source data Source Data Source Data