--- title: 'Zhang–Li Term: Spintronics, Geometry & Kähler Flow' url: https://www.emergentmind.com/topics/zhang-li-term type: topic --- # Zhang–Li Term: Spintronics, Geometry & Kähler Flow Searching arXiv for recent and relevant papers on the different uses of “Zhang–Li term”. “Zhang–Li term” is a context-dependent designation rather than a single universally fixed object. In contemporary arXiv usage represented by the literature considered here, it denotes three different constructions: an additional spin-transfer torque in magnetization dynamics for non-uniform magnetic textures; a zeroth-order curvature contribution in noncompact Llarull-type twisted Dirac estimates; and the explicit coupling between an evolving Kähler metric and a closed \((1,1)\)-form in the Li–Yuan–Zhang flow. This distribution of meanings suggests that the phrase is best understood relationally: in each setting it names the extra term introduced when a baseline evolution or Bochner identity is coupled to an additional geometric or transport structure [2005.13364], [2601.12803], [1808.06968]. ## 1. Distinct uses of the designation The three uses represented in the cited arXiv literature can be organized as follows. | Context | Object called the Zhang–Li term | Structural role | |---|---|---| | Spintronics and STNO micromagnetics | The Zhang–Li torque \(\boldsymbol{\tau}_{\text{ZL}}\) | Extra spin-transfer torque driven by \((\mathbf{j}\cdot\nabla)\mathbf{m}\) | | Noncompact Llarull-type scalar curvature rigidity | A zeroth-order curvature term in a twisted Dirac Weitzenböck formula | Nonnegative curvature-defect term localized on \(\operatorname{supp}(d\Phi)\) | | Li–Yuan–Zhang coupled Kähler flow | The coupling \(+\alpha_t\) in \(\partial_t\omega_t\), equivalently \(-F\) in \(\partial_t\varphi\) | Dynamical coupling between \(\omega_t\) and \(\alpha_t\) | In the spintronics paper, the terminology is explicit and formulaic: the Zhang–Li term is an actual torque added to the Landau–Lifshitz-type equation for \(\mathbf{m}(\mathbf{r},t)\) [2005.13364]. In the noncompact Llarull paper, by contrast, the phrase “Zhang–Li term” does not appear explicitly; the paper imports the underlying analytic machinery from Zhang and Li–Su–Wang–Zhang as a black box and only allows a conceptual identification of the relevant curvature contribution [2601.12803]. In the \(\kappa\)-LYZ flow paper, the term is again contextual: it refers to the explicit coupling between the metric side and the \((1,1)\)-form side, either as \(+\alpha_t\) in the form equation or as \(-F\) in the scalar potential formulation [1808.06968]. A common misconception is to treat “Zhang–Li term” as having a single cross-disciplinary definition. The available evidence instead indicates that the phrase is overloaded. Any rigorous use therefore requires immediate specification of the ambient framework: micromagnetic dynamics, Dirac-operator scalar curvature rigidity, or coupled Kähler flow. ## 2. Spin-transfer form: the Zhang–Li torque in magnetization dynamics In nanocontact spin-torque nano-oscillators, the reduced magnetization \(\mathbf{m}(\mathbf{r},t)=\mathbf{M}/M_{\text{sat}}\) is evolved by a Landau–Lifshitz-type equation with three torques, \[ \dot{\mathbf{m}}(\mathbf{r},t)=\boldsymbol{\tau}_{\text{LL}}+\boldsymbol{\tau}_{\text{SL}}+\boldsymbol{\tau}_{\text{ZL}}, \] where \(\boldsymbol{\tau}_{\text{LL}}\) is the precession-plus-damping part, \(\boldsymbol{\tau}_{\text{SL}}\) is the Slonczewski spin-transfer torque, and \(\boldsymbol{\tau}_{\text{ZL}}\) is the Zhang–Li torque [2005.13364]. In the first-order adiabatic approximation used there, the Zhang–Li torque is \[ \boldsymbol{\tau}_{\text{ZL}} = -\gamma \,\frac{\hbar}{2 e M_\text{sat}} \frac{1}{1+\alpha^2} \Big[ \mathbf{m} \times \big(\mathbf{m}\times (\mathbf{j}\cdot\nabla)\mathbf{m}\big) - \alpha \,\big(\mathbf{m}\times (\mathbf{j}\cdot\nabla)\mathbf{m}\big) \Big]. \] The quantities entering this expression are defined internally: \(\gamma\) is the gyromagnetic ratio of the electron, \(\hbar\) the reduced Planck constant, \(e\) the electron charge, \(M_{\text{sat}}\) the saturation magnetization of the free layer, \(\alpha\) the Gilbert damping parameter, and \(\mathbf{j}\) the charge current density. The operator \((\mathbf{j}\cdot\nabla)\mathbf{m}\) is the directional derivative of \(\mathbf{m}\) along the current flow, so the torque is intrinsically sensitive to magnetization gradients rather than merely to local magnetization orientation [2005.13364]. The same paper emphasizes that the written form contains both the so-called adiabatic and non-adiabatic components, with the non-adiabaticity effectively fixed by \(\beta=\alpha\). In the representation adopted there, the term \[ \mathbf{m}\times\big(\mathbf{m}\times(\mathbf{j}\cdot\nabla)\mathbf{m}\big) \] is the adiabatic or transport-like contribution, while \[ -\alpha\,\mathbf{m}\times(\mathbf{j}\cdot\nabla)\mathbf{m} \] acts as the non-adiabatic part. The authors do not introduce an independent \(\beta\); instead they use MuMax\(^3\)’s built-in Zhang–Li implementation with \(\beta=\alpha\) [2005.13364]. The physical interpretation given there is transport-theoretic. Conduction electrons traversing a non-uniform magnetization experience a spatially varying exchange field. As their spins attempt to follow the local magnetization but cannot do so perfectly because of finite spin relaxation, they develop a spin accumulation misaligned with \(\mathbf{m}\). Angular-momentum transfer then produces a torque that tends to translate spin textures along the current direction and adds an effective field-like or damping-like component. In the paper’s own characterization, “the ZL torque is related to magnetic gradients and can be seen as a modulation of the generated excitations,” and it is expected to “have an impact on magnetic domain walls; with opposite effects for opposite current directions” [2005.13364]. For nanocontact geometries with cylindrical symmetry and negligible \(z\)-dependence of \(\mathbf{m}\), the dominant magnitude scales as \[ \left|\boldsymbol{\tau}_{\text{ZL}_1}\right| \propto j_\rho \left|\frac{\partial \mathbf{m}}{\partial \rho}\right|. \] This identifies the effective control parameters of the torque: the radial in-plane current density \(j_\rho(\rho,z)\) and the radial magnetization gradient. If the in-plane current is artificially suppressed so that \(j_\rho=0\), the Zhang–Li term disappears in the thin-layer approximation [2005.13364]. ## 3. Nanocontact STNOs: thresholds, mode size, and current geometry The nanocontact STNO study treats the Zhang–Li torque as a quantitatively relevant ingredient of microwave auto-oscillator dynamics rather than as a small correction [2005.13364]. The simulated geometry is a circular metallic nanocontact of radius \(a=50\) nm on a multilayer stack approximated electrically as a single ohmic layer of thickness \(h\), with a typical case \(h=54\) nm. The free layer thickness is \(4\) nm, the applied field is either in-plane for bullets or out-of-plane for propagating spin waves and droplet solitons, and the magnetization dynamics of the free layer alone is simulated with MuMax\(^3\) [2005.13364]. A critical step is the current model. Instead of assuming a uniform vertical current, the paper solves Laplace’s equation \[ \nabla^2 V=0 \] for the electrostatic potential in an infinite cylindrical conductor of thickness \(h\), subject to boundary conditions that impose a uniform vertical current through the contact and no vertical current at the bottom. With Ohm’s law \[ \mathbf{j}=-\frac{1}{\rho_{\text{el}}}\nabla V, \] the resulting current distribution has substantial radial components \(j_\rho\), especially near the contact edge and for thin electrodes [2005.13364]. This point is decisive because the Zhang–Li torque vanishes in the thin-layer approximation if one imposes purely vertical current. Three simulation protocols are compared: no-ZL, ZL+, and ZL−. In the latter two, the sign of the charge current is reversed while the Slonczewski torque is held effectively unchanged by simultaneously flipping the projection of the polarizer, so only the Zhang–Li contribution changes sign. This isolates the effect of \(\boldsymbol{\tau}_{\text{ZL}}\) from the rest of the dynamics [2005.13364]. Across propagating spin waves, localized bullets, and magnetic droplet solitons, the threshold ordering is the same: \[ I_\text{th}^{\,+} < I_\text{th}^\text{no-ZL} < I_\text{th}^{\,-}. \] Thus a positive Zhang–Li torque lowers the threshold current, whereas a negative Zhang–Li torque raises it [2005.13364]. For propagating spin waves, the dispersion relation \(f(k)\) is reported to be largely unaffected, but the frequency-current relation and the onset current shift with the Zhang–Li contribution. The interpretation offered is that ZL+ acts like an additional effective excitation drive under the contact, producing larger amplitude spin waves and hence higher frequency at fixed current [2005.13364]. For bullets and droplets, the effect is stronger on spatial extent than on dispersion. At fixed current above threshold, ZL+ produces larger effective bullet or droplet area, the no-ZL case is intermediate, and ZL− produces smaller localized modes [2005.13364]. This is consistent with the schematic interpretation that the torque can expand or shrink localized reversed regions depending on current direction and surrounding magnetization orientation. The paper also studies geometric control through the ratio \(h/a\). When \(h\) is small compared to \(a\), current continuity forces stronger lateral spreading and hence larger \(j_\rho\); when \(h\gg a\), current lines are nearly vertical near the free layer and \(j_\rho\) becomes small [2005.13364]. For a \(50\) nm-radius contact, varying thickness from very thin up to about \(50\)–\(75\) nm can change thresholds and effective sizes of localized excitations by up to approximately \(20\%\) due to Zhang–Li torques, while beyond about \(75\)–\(100\) nm the behavior approaches the \(h\to\infty\) limit [2005.13364]. The practical implication drawn there is that omission of the Zhang–Li term, or inadvertent suppression of it through a vertical-current approximation, can yield incorrect thresholds and incorrect mode sizes in thin-film nanocontact oscillators. ## 4. Noncompact Llarull theory: the Zhang–Li term as a Dirac–Weitzenböck curvature contribution In noncompact scalar-curvature rigidity, the phrase refers to something entirely different. The paper on Llarull’s theorem for noncompact manifolds with boundary states explicitly that the phrase “Zhang–Li term” does not occur there and that no detailed formulas from Zhang or Li–Su–Wang–Zhang are reproduced; nonetheless, it identifies the relevant object conceptually from the known structure of noncompact Llarull-type arguments [2601.12803]. The underlying noncompact theorem used there is the following: if \((M,g_M)\) is a complete noncompact Riemannian spin manifold, \(\Phi:M\to S\) is a smooth area decreasing map which is locally constant at infinity and of nonzero degree, and \[ \mathrm{scal}_{g_M}\ge m(m-1)\quad\text{on }\operatorname{supp}(d\Phi), \] then \[ \inf \mathrm{scal}_{g_M}<0. \] The key novelty relative to the classical closed case is that the scalar-curvature comparison is imposed only on \(\operatorname{supp}(d\Phi)\), not globally [2601.12803]. Within the imported Dirac framework, one considers a twisted Dirac operator \[ D^\Phi:\Gamma(\mathbb{S}_M\otimes E_\Phi)\to \Gamma(\mathbb{S}_M\otimes E_\Phi), \] where \(E_\Phi\) is typically \(\Phi^*TS\) or a related bundle. Its square satisfies a Weitzenböck formula of the form \[ (D^\Phi)^2=\nabla^*\nabla+\frac{1}{4}\mathrm{scal}_{g_M}+\pazocal{R}(\Phi), \] with \(\pazocal{R}(\Phi)\) a zeroth-order endomorphism built from the curvature of \(M\), the curvature of the sphere, and the differential \(d\Phi\) [2601.12803]. The conceptual identification made there is that the Zhang–Li term is the nonnegative or controlled remainder term that appears after one optimally separates the model scalar-curvature contribution \(m(m-1)\) from the rest of the curvature operator. Schematically, the paper describes an inequality of the form \[ (D^\Phi)^2 \ge \frac{1}{4}\bigl(\mathrm{scal}_{g_M}-m(m-1)\bigr)+\pazocal{Z}_{\mathrm{ZL}(\Phi)}, \] where \(\pazocal{Z}_{\mathrm{ZL}(\Phi)}\) is pointwise nonnegative under the area-decreasing assumption [2601.12803]. In that description, the Zhang–Li term is a zeroth-order operator whose quadratic form measures the defect of \(\Phi\) from infinitesimal isometry on \(2\)-vectors. It vanishes when \(d\Phi\) behaves isometrically on \(2\)-planes and captures the positivity gained from mapping into the positively curved target sphere by an area-decreasing map. The same paper presents the analytic origin more explicitly through the twisted-curvature term \[ \frac{1}{2}\sum_{i0\ \text{near }\partial M, \] and mean-convexity \[ H_{\partial M}\ge 0 \] with respect to the interior unit normal. The conclusion is again \[ \inf \mathrm{scal}_{g_M}<0 \] [2601.12803]. The proof proceeds by doubling. After a deformation that strengthens \(H_{\partial M}\ge0\) to \(H_{\partial M}>0\), one removes a thin collar \(C\) of the boundary, keeps a larger collar \(C'\) with positive scalar curvature and disjoint from \(\operatorname{supp}(d\Phi)\), forms \(M\times[0,1]\), and defines an \(\varepsilon\)-tube \[ D(M):=\{p\in M\times[0,1]: \operatorname{dist}(p,M_1\times\{1/2\})=\varepsilon\}, \] where \(M_1=M\setminus C\). Gromov–Lawson smoothing then yields a complete noncompact spin manifold \(D(M)\) without boundary, with scalar curvature positive on the bending region and unchanged on the parallel part containing the support of the differential [2601.12803]. A map \(\Theta:D(M)\to S\) is constructed so that \[ \deg(\Theta)=\deg(\Phi)\neq0,\qquad \operatorname{supp}(d\Theta)=\operatorname{supp}(d\Phi), \] \(\Theta\) is locally constant at infinity, and the metric on the relevant support region agrees with the original one. Consequently, \[ \mathrm{scal}_{g_{D(M)}}\ge m(m-1)\quad\text{on }\operatorname{supp}(d\Theta). \] At that point the boundaryless noncompact theorem is invoked directly on \(D(M)\) [2601.12803]. The significance for the Zhang–Li term is explicit: no new boundary version of the term is introduced. The analytic machinery remains entirely boundaryless because the proof never imposes APS, chiral, or other explicit boundary conditions on spinors. The Zhang–Li term sits inside the imported proof on the doubled manifold \(D(M)\), exactly as in the original noncompact boundaryless setting [2601.12803]. ## 6. Coupled Kähler geometry: the Zhang–Li term in the Li–Yuan–Zhang flow In the \(\kappa\)-LYZ flow, the relevant object is the coupling between a Kähler metric \(\omega_t\) and a closed real \((1,1)\)-form \(\alpha_t\). The flow is \[ \partial_t \omega_t = - \operatorname{Ric}(\omega_t) + \lambda\,\omega_t + \alpha_t, \] \[ \partial_t \alpha_t = - \kappa\,\Delta_{\omega_t}\alpha_t. \] When \(\kappa=1\) and \(\lambda=-1\), this is exactly the flow introduced by Li, Yuan, and Zhang [1808.06968]. In this context, the “Zhang–Li term” is the explicit coupling: on the metric side it is the extra \(+\alpha_t\) added to the Kähler–Ricci evolution, and in the scalar potential formulation it appears as \(-F\) [1808.06968]. Indeed, with \[ \omega_t=\omega_0+i\partial\bar\partial\varphi_t,\qquad \alpha_t=\alpha_0-i\partial\bar\partial F_t, \] the flow becomes \[ \partial_t\varphi=\log\frac{\omega_t^n}{\omega_0^n}+H_0+\lambda\varphi-F, \] \[ \partial_t F=\kappa\,\Delta_{\omega_t}F-\kappa\,\operatorname{tr}_{\omega_t}\alpha_0+\kappa b. \] The paper states explicitly that “the coupling term in \(\partial_t\varphi\) is \(-F\)” [1808.06968]. Thus, unlike the spintronics usage, the Zhang–Li term here is not an extra transport torque; it is the dynamical twisting that makes the metric evolution depend on the evolving \((1,1)\)-form and vice versa. The parameter \(\kappa\) enters not in the metric coupling strength but in the diffusion of \(\alpha_t\). The central analytic observation is that for \(\kappa\neq1\), the combination \[ Q = R + \frac{1}{\kappa-1}\operatorname{tr}_{\omega}\alpha \] satisfies a favorable parabolic inequality, which yields a lower bound for the scalar curvature from \(C^0\) control [1808.06968]. This is the technical reason the paper can prove that, assuming \[ K^{-1}\omega_0\le \omega_t\le K\omega_0,\qquad -K\omega_t\le \alpha_t\le K\omega_t, \] all higher derivatives of \(\omega_t\) and \(\alpha_t\) are uniformly bounded for \(0<\kappa\neq1\) [1808.06968]. The coupling term is therefore not peripheral; it is precisely what forces the use of mixed scalar combinations and makes the \(\kappa\neq1\) condition analytically decisive. The paper also introduces monotone functionals tailored to this coupling. On Riemann surfaces it defines \[ E(\phi,\tau)=\int_X\left(|\nabla\phi|_{\omega_0}^2+R_0\phi-e^\phi\right)\omega_0 +\int_X \tau\log\tau\,e^\phi\omega_0, \] whose time derivative is \[ \frac{d}{dt}E(\phi,\tau) = -\int_X(\partial_t\phi)^2\,\omega_t -\kappa\int_X \frac{|\nabla\tau|_{\omega_t}^2}{\tau}\,\omega_t \le 0 \] [1808.06968]. In higher dimensions it defines the coupled Mabuchi-type functional \[ M_{\omega_0}(\varphi,F)=N_{\omega_0}(\varphi)+\int_X F\,\omega_\varphi^n, \] and proves \[ \frac{d}{dt}M_{\omega_0}(\varphi,F) =-\frac{1}{V}\int_X |\dot\varphi|^2\,\omega_t^n\le0. \] These formulas show that the additional field \(F\), equivalently the Zhang–Li coupling, is built directly into the Lyapunov structure of the flow [1808.06968]. ## 7. Comparative interpretation and terminological cautions The three meanings share a formal family resemblance. In each case, the Zhang–Li term is the additional contribution that couples a baseline evolution to an auxiliary structure: a current-driven gradient in micromagnetics, a sphere-valued differential in twisted Dirac analysis, or an evolving closed \((1,1)\)-form in Kähler geometry [2005.13364], [2601.12803], [1808.06968]. This suggests a useful comparative description: the term always marks the place where nontrivial interaction enters. The differences, however, are structurally stronger than the similarity. In spintronics, the term is a dynamical torque depending on \((\mathbf{j}\cdot\nabla)\mathbf{m}\) and changes sign with current polarity [2005.13364]. In noncompact Llarull theory, it is a zeroth-order endomorphism in a Bochner–Weitzenböck inequality, nonnegative under the area-decreasing hypothesis and tied to the support of \(d\Phi\) [2601.12803]. In the Li–Yuan–Zhang flow, it is the explicit metric-form coupling \(+\alpha_t\), or equivalently \(-F\), whose interaction with the parameter \(\kappa\) determines whether favorable scalar combinations and full a priori estimates are available [1808.06968]. Several misconceptions therefore require qualification. It is inaccurate to identify the Zhang–Li term exclusively with the torque from spintronics, because current arXiv usage includes at least the geometric-analysis and Kähler-flow meanings represented here. It is equally inaccurate to assume that the term is always explicitly displayed in the source paper: in the boundary extension of noncompact Llarull’s theorem, the terminology is absent from the text and only recoverable conceptually from the imported Zhang and Li–Su–Wang–Zhang framework [2601.12803]. Finally, it is misleading to think of the term as invariably perturbative. In all three settings, the cited papers treat it as structurally decisive: it changes threshold currents and mode sizes in STNOs, enables localization in noncompact scalar-curvature rigidity, and governs the analytic coupling and convergence theory of the \(\kappa\)-LYZ flow [2005.13364], [2601.12803], [1808.06968]. Taken together, these usages show that “Zhang–Li term” functions as a domain-specific label whose meaning is fixed only by the surrounding formalism. Precise scholarship therefore requires citing the operative framework, the governing equation or Bochner identity, and the exact role of the term in the argument or dynamics under discussion. Source: https://www.emergentmind.com/topics/zhang-li-term