betaFeedback

WorksPhysics and Astronomy

How Elasticity and Chiral Damping Shape Domain-Wall Creep

Abstract

Asymmetric domain-wall expansion in the creep regime provides an experimentally accessible probe of the interfacial Dzyaloshinskii-Moriya interaction (iDMI) and, in principle, chiral damping. However, severe discrepancies with complementary measurement techniques and the lack of a model describing the full experimental phenomenology of the creep of chiral domain walls, limits the application of these measurements. Here, using a symmetry decomposition of the velocity curve, we derive an analytical model describing the coupled influence of domain-wall stiffness and chiral damping in the creep regime. Their interplay generates the antisymmetric distortion that shifts the velocity minimum away from the DMI field. Applied to Co-based magnetic multilayers, the model yields iDMI values consistent with BLS and flow-regime benchmarks and identifies the reasons for the longstanding discrepancy. The model provides insights for possible strategies to measure chiral damping via creep dynamics of domain walls.

The interfacial Dzyaloshinskii-Moriya interaction (iDMI) [1, 2, 3, 4] is a key ingredient in the stabilization and dynamics of chiral magnetic textures in heavy-metal/ferromagnet multilayers with perpendicular magnetic anisotropy, including homochiral domain walls [5, 6, 7, 8, 9] and skyrmions [10, 11, 12]. More recently, chiral damping has emerged as a distinct contribution to domain-wall motion, with possible implications for the mobility of magnetic textures [13, 14, 15]. The quantitative determination of both effects is therefore relevant to the fundamental understanding of chiral magnetism and to the design of future spintronic devices [16, 17, 18] [19].

Field-driven domain-wall expansion in the creep regime, in which an out-of-plane field drives the wall in the presence of an in-plane field [14] [20, 21, 22] holds promise because it couples experimental accessibility and, in principle, sensitivity to both effects [13] [23]. To date, however, creep measurements have mainly been used to determine the iDMI. In the conventional analysis, the minimum of the domain-wall velocity curve is identified with the DMI effective field, Hxmin⁡=−HDMIH_x^{\min} = -H_{\mathrm{DMI}}, and the DMI constant is obtained as D=μ0MsHDMIΔD = \mu_0 M_s H_{\mathrm{DMI}}\Delta, where Δ\Delta is the domain-wall width parameter [20, 24–26].

Cross-laboratory comparisons have however highlighted an important discrepancy [19, 27]: Brillouin light scattering (BLS) determines iDMI through spin-wave nonreciprocity [28, 29], and its estimates agree quantitatively with those obtained from DW flow-regime measurements [30, 31]; values obtained from conventional DW creep analysis, instead, often show substantial over- or underestimates [27, 32] (see Fig. 1). This ambiguity limits both DMI extraction and the use of creep dynamics to probe chiral damping, whose quantitative experimental determination remains challenging. Domain-wall elasticity [33–35], and chiral damping [13, 14] have all been investigated as possible corrections to the conventional creep interpretation, including through numerical treatments [23, 36]. Nevertheless, an analytical description of the coupled influence of domain-wall elasticity and chiral damping is still missing.

Here, following a symmetry decomposition of the DW velocity curve [37], we derive a unified model describing the coupled action of domain-wall stiffness and chiral damping. We show why this discrepancy is specific to the creep regime and demonstrate that these contributions shift the creep-velocity minimum away from the DMI field, shedding light on the controversies associated with the conventional HDMIH_{\mathrm{DMI}} extraction method. [20]

We test this description on five magnetron-sputtered Pt/Co/XX multilayers with different capping layers -

Figure 1. iDMI values extracted from BLS and conventional DW creep analysis for the investigated Pt/Co multilayers. The experimental data are from Ref. [27]; the creep values were obtained using the conventional model of Ref. [20]. Lines are guides to the eye.

Figure 2. Geometry of asymmetric bubble expansion, defining HxH_x, HzH_z, the wall-normal angle θ\theta, and the internal wall angle ϕ\phi.

Pt/Co/Pt(3), Pt/Co/Pt(1), Pt/Co/Ir(3), Pt/Co/Ir(1), and Pt/Co/Ta — and on a structurally distinct Si/SiO2_2/TiN(10)/Hf0.5_{0.5}Zr0.5_{0.5}O2_2(10)/Co(0.7)/Pt(6) sample, denoted HZO/Co/Pt. Growth details, structural and magnetic characterization of all samples are reported in the Suppl. Mater. [49]. Across the investigated samples, the symmetry-decomposed DW creep analysis yields iDMI values consistent with the BLS benchmarks and, where available, the DW flow benchmarks. The resulting physical picture improves DMI determination from creep measurements and provides insight to extend creep regime measurements to probe chiral damping.

Domain wall creep in the presence of domain wall stiffness and chiral damping —

We now develop an analytical framework for the combined effects of chiral damping and domain-wall elasticity on creep-regime motion. At low driving fields, the domain-wall velocity is governed by an Arrhenius law [38–40], v=v0exp⁡[−χ(Hx)Hz−1/4]v=v_0\exp[-\chi(H_x)H_z^{-1/4}]. Here, the dimensionless exponent χ(Hx)Hz−1/4\chi(H_x)H_z^{-1/4} represents the field-dependent activation energy in units of kBTk_{\mathrm{B}}T. In the following, we adopt a semi-phenomenological approximation in which the activation-energy coefficient χ(Hx)\chi(H_x) is controlled by the domain-wall stiffness σ~\tilde{\sigma}, whereas the prefactor v0(Hx)v_0(H_x) depends on the equilibrium wall energy σeq=min⁡ϕσ(θ,ϕ)\sigma_{\mathrm{eq}}=\min_{\phi}\sigma(\theta,\phi). Since the stiffness introduces an angular-curvature correction to the equilibrium wall energy, its influence is amplified in the creep exponent, while its contribution to the algebraic prefactor is expected to be comparatively smaller. We therefore retain this correction in the activation barrier while neglecting it in the prefactor [22]. The domain-wall energy density σ(θ,ϕ)\sigma(\theta,\phi) is

σ(θ,ϕ)=σ0−πμ0MsΔ[Hxcos⁡ϕ+HDMIcos⁡(ϕ−θ)]+ln⁡(2)πtμ0Ms2cos⁡2(ϕ−θ).(1)\sigma(\theta,\phi)=\sigma_0-\pi\mu_0M_s\Delta\left[H_x\cos\phi+H_{\mathrm{DMI}}\cos(\phi-\theta)\right]+\frac{\ln(2)}{\pi}t\mu_0M_s^2\cos^2(\phi-\theta). \tag*{(1)}

where θ\theta and ϕ\phi represent the azimuthal angles of the DW normal and magnetization (see Fig. 2). σ0=4AKeff\sigma_0=4\sqrt{AK_{\mathrm{eff}}} is the Bloch wall energy, Δ=A/Keff\Delta=\sqrt{A/K_{\mathrm{eff}}} is the domain-wall width parameter, AA is the exchange stiffness, Keff=Ku−μ0Ms22K_{\mathrm{eff}}=K_u-\frac{\mu_0M_s^2}{2} is the effective perpendicular magnetic anisotropy, MsM_s is the saturation magnetization, HDMIH_{\mathrm{DMI}} is the DMI field, and tt is the effective magnetic thickness. KuK_u represents the uniaxial magnetocrystalline anisotropy. In this approximation, we write

χ(Hx)=χ0(σ~(Hx)σ~(0))1/4,v0∝σeqβαeff−1.(2)\chi(H_x)=\chi_0\left(\frac{\tilde{\sigma}(H_x)}{\tilde{\sigma}(0)}\right)^{1/4},\qquad v_0\propto\sigma_{\mathrm{eq}}^\beta\alpha_{\mathrm{eff}}^{-1}. \tag*{(2)}

where χ0\chi_0 is the creep scaling coefficient at Hx=0H_x=0, and β\beta describes the scaling of the velocity prefactor with the equilibrium domain-wall energy density [41]. We model the DW stiffness according to the description of Refs. [23, 33],

σ~(θ)=σ(θ,ϕeq)+∂2σ(θ,ϕ)∂θ2∣ϕ=ϕeq.(3)\tilde{\sigma}(\theta)=\sigma(\theta,\phi_{\mathrm{eq}})+\left.\frac{\partial^2\sigma(\theta,\phi)}{\partial\theta^2}\right|_{\phi=\phi_{\mathrm{eq}}}. \tag*{(3)}

Here, ϕeq=arg⁡min⁡ϕσ(θ,ϕ)\phi_{\mathrm{eq}}=\arg\min_{\phi}\sigma(\theta,\phi). We couple this expression with the chiral damping parametrization

αeff(ϕ)=α0[1+sηcos⁡(ϕ−θ)],(4)\alpha_{\mathrm{eff}}(\phi)=\alpha_0\left[1+s\eta\cos(\phi-\theta)\right], \tag*{(4)}

where s=±1s=\pm1 labels the two wall chiralities, η\eta is the dimensionless relative chiral-damping strength, and α0\alpha_0 is ordinary Gilbert damping [41, 42]. Assuming θ=0\theta=0, the logarithmic velocity, up to irrelevant constants, is

ln⁡v(Hx,Hz)=βln⁡(σeq(Hx)σeq(0))−ln⁡αeff(Hx)−χ0(σ~(Hx)σ~(0))1/4Hz−1/4.(5)\ln v(H_x,H_z)=\beta\ln\left(\frac{\sigma_{\mathrm{eq}}(H_x)}{\sigma_{\mathrm{eq}}(0)}\right)-\ln\alpha_{\mathrm{eff}}(H_x)-\chi_0\left(\frac{\tilde{\sigma}(H_x)}{\tilde{\sigma}(0)}\right)^{1/4}H_z^{-1/4}. \tag*{(5)}

To translate this coupled expression into a functional tool, we decompose the velocity curve about the field −HDMI-H_{\mathrm{DMI}}, at which the applied in-plane field compensates the DMI contribution to the wall energy [37]. The symmetric component is even and the antisymmetric component is odd about this decomposition center. Under the approximations of small in-plane-field-induced variations of the equilibrium domain-wall energy and stiffness, and retaining chiral damping up to second order O(η2)\mathcal{O}(\eta^2),

Eq. (5) reduces to a minimal semi-phenomenological model (see [49] for the full derivation). Combining all variables not directly related to HDMI+HxH_{\mathrm{DMI}}+H_x and defining HKDW=2ln⁡(2)tMs(π2Δ)−1H_K^{\mathrm{DW}}=2\ln(2)tM_s(\pi^2\Delta)^{-1} as the DW anisotropy field, we identify ∣Hx+HDMI∣≤HKDW\lvert H_x+H_{\mathrm{DMI}}\rvert\leq H_K^{\mathrm{DW}} as the mixed Bloch-Néel region, in which the equilibrium wall magnetization rotates continuously between the two limiting configurations. The symmetric component of the velocity is governed by two parameters, ν1\nu_1 and ν3\nu_3:

(ln⁡v)S={ν1(HKDW)2−ν3(Hx+HDMI+HKDW),Hx+HDMI<−HKDW,ν1(Hx+HDMI)2,∣Hx+HDMI∣≤HKDW,ν1(HKDW)2+ν3(Hx+HDMI−HKDW),Hx+HDMI>HKDW.(6)(\ln v)_S = \begin{cases} \nu_1(H_K^{\mathrm{DW}})^2-\nu_3(H_x+H_{\mathrm{DMI}}+H_K^{\mathrm{DW}}), & H_x+H_{\mathrm{DMI}}<-H_K^{\mathrm{DW}},\\ \nu_1(H_x+H_{\mathrm{DMI}})^2, & \lvert H_x+H_{\mathrm{DMI}}\rvert\leq H_K^{\mathrm{DW}},\\ \nu_1(H_K^{\mathrm{DW}})^2+\nu_3(H_x+H_{\mathrm{DMI}}-H_K^{\mathrm{DW}}), & H_x+H_{\mathrm{DMI}}>H_K^{\mathrm{DW}}. \end{cases} \tag*{(6)}

The antisymmetric component, under these approximations, is governed by a single parameter ν2\nu_2:

(ln⁡v)A={−ν2,Hx+HDMI<−HKDW,ν2Hx+HDMIHKDW,∣Hx+HDMI∣≤HKDW,ν2,Hx+HDMI>HKDW.(7)(\ln v)_A = \begin{cases} -\nu_2, & H_x+H_{\mathrm{DMI}}<-H_K^{\mathrm{DW}},\\ \nu_2\frac{H_x+H_{\mathrm{DMI}}}{H_K^{\mathrm{DW}}}, & \lvert H_x+H_{\mathrm{DMI}}\rvert\leq H_K^{\mathrm{DW}},\\ \nu_2, & H_x+H_{\mathrm{DMI}}>H_K^{\mathrm{DW}}. \end{cases} \tag*{(7)}

Introducing the reduced variables

ζ=πμ0MsΔHDMIσ0,βx=πμ0MsΔHxσ0,(8)\zeta=\frac{\pi\mu_0M_s\Delta H_{\mathrm{DMI}}}{\sigma_0},\qquad\beta_x=\frac{\pi\mu_0M_s\Delta H_x}{\sigma_0}, \tag*{(8)}
κ=πμ0MsΔHKDW2σ0,(9)\kappa=\frac{\pi\mu_0M_s\Delta H_K^{\mathrm{DW}}}{2\sigma_0}, \tag*{(9)}

and defining Cz=(χ0Hz−1/4)/4C_z=(\chi_0H_z^{-1/4})/4 for compactness, where CzC_z carries the HzH_z dependence of the elastic creep contribution, we obtain

ν1=1(HKDW)2[−βκ1+κ−∣ζ∣+5Czκ1−κ+η22],(10)\nu_1=\frac{1}{(H_K^{\mathrm{DW}})^2}\left[-\frac{\beta\kappa}{1+\kappa-\lvert\zeta\rvert}+\frac{5C_z\kappa}{1-\kappa}+\frac{\eta^2}{2}\right], \tag*{(10)}
ν2=sη−ζCz1−κ,(11)\nu_2=s\eta-\frac{\zeta C_z}{1-\kappa}, \tag*{(11)}
ν3=2κHKDW[Cz1−κ−β1+κ−∣ζ∣].(12)\nu_3=\frac{2\kappa}{H_K^{\mathrm{DW}}}\left[\frac{C_z}{1-\kappa}-\frac{\beta}{1+\kappa-\lvert\zeta\rvert}\right]. \tag*{(12)}

We can now analytically express the shift of the velocity minimum. In the mixed Bloch-Néel region, the velocity minimum Hx⋆H_x^\star is located at

(Hx⋆)creep=−HDMI−ν22ν1HKDW.(13)(H_x^\star)_{\mathrm{creep}}=-H_{\mathrm{DMI}}-\frac{\nu_2}{2\nu_1H_K^{\mathrm{DW}}}. \tag*{(13)}

While derived under specific approximations, this expression captures the phenomenology observed experimentally. First, it explicitly shows how the minimum is shifted from the compensation field −HDMI-H_{\mathrm{DMI}} whenever ν2≠0\nu_2\neq0. The shift is controlled by the antisymmetric amplitude ν2\nu_2, which contains two contributions: the chiral-damping term sηs\eta, and the elastic term −ζCz/(1−κ)-\zeta C_z/(1-\kappa) generated by the DMI-dependent DW stiffness in the activation exponent Eq. (11). The sign of the chiral-damping term ∝sη\propto s\eta is sample dependent and not necessarily related to the sign of DMI [15] [42]. We also highlight the HzH_z field dependence of the velocity minimum [40], which in this formalism can be clearly attributed to the elastic contribution ∼ζCz\sim\zeta C_z in ν2\nu_2.

Chiral damping and domain-wall elasticity in BLS and DW flow measurements

As shown above, creep domain-wall dynamics are particularly sensitive to the coupled effects of chiral damping and DW stiffness [41]. It is therefore important to assess whether the same effects bias BLS and DW flow measurements. DW stiffness is not relevant in the interpretation of BLS spectra, as the technique probes propagating spin waves rather than domain-wall motion [43, 44]. The Stokes/anti-Stokes peak positions and linewidths nevertheless contain signatures of both the iDMI and chiral damping. As discussed by Kim et al. [45], the iDMI produces a nonreciprocal frequency shift and, through the frequency dependence of ordinary Gilbert damping, also contributes to the linewidth asymmetry. Chiral damping, instead, introduces an additional odd-in-wave-vector contribution to the linewidth asymmetry. Defining Δf=∣fS∣−∣fAS∣\Delta f=\lvert f_S\rvert-\lvert f_{AS}\rvert where fSf_S and fASf_{AS} represent the Stokes and Anti-Stokes peak frequencies [Fig. 3(i)], the DMI constant is obtained as D=πMsΔf/(2γk)D=\pi M_s\Delta f/(2\gamma k), where kk is the spin-wave wave vector and γ\gamma is the gyromagnetic ratio in frequency units, for the sign convention used here. Defining the measured linewidth sum and difference Γ±=ΓS±ΓAS\Gamma_{\pm}=\Gamma_S\pm\Gamma_{AS} [Fig. 3(i)], we can isolate the chiral-damping contribution via

Figure 3. i) Representative BLS spectrum (sample Pt/Co/Ir(1)) showing both the Stokes/anti-Stokes frequencies fS,fASf_{\mathrm{S}}, f_{\mathrm{AS}}, used to extract DD, and the peak-widths ΓS,ΓAS\Gamma_{\mathrm{S}}, \Gamma_{\mathrm{AS}}. ii) relative chiral damping αc/α0\alpha_{\mathrm{c}}/\alpha_{0} contributions obtained from ΓS,ΓAS\Gamma_{\mathrm{S}}, \Gamma_{\mathrm{AS}} (left axis) and DMI values measured from BLS spectra (right axis).

αcα0=Γ−−rfΓ+Γ+−rfΓ−,(14)\frac{\alpha_{\mathrm{c}}}{\alpha_{0}}=\frac{\Gamma_{-}-r_{\mathrm{f}}\Gamma_{+}}{\Gamma_{+}-r_{\mathrm{f}}\Gamma_{-}}, \tag*{(14)}

where rf=(∣fS∣−∣fAS∣)/(∣fS∣+∣fAS∣)r_{\mathrm{f}}=(|f_{\mathrm{S}}|-|f_{\mathrm{AS}}|)/(|f_{\mathrm{S}}|+|f_{\mathrm{AS}}|) [49]. Here, αc\alpha_{\mathrm{c}} denotes the chiral contribution to the damping. The resulting chiral-damping contributions for our sample series are shown in Fig. 3(ii). Within the Pt/Co/X series, the sign and magnitude of the linewidth asymmetry broadly track those of DD. This suggests a sample-dependent correlation rather than a general proportionality: DMI and chiral damping correspond to distinct energetic and dissipative responses and can vary independently [45], as illustrated here by the weak linewidth asymmetry of sample HZO/Co/Pt despite its sizable DMI.

DW flow measurements – The domain-wall stiffness correction governing the creep activation barrier does not enter the flow regime, where the motion is no longer controlled by thermally activated depinning [46, 47]. In the strong-DMI limit, the Walker-velocity plateau provides a direct estimate of the DMI strength [48],

vW≃π2DMs,(15)v_{\mathrm{W}}\simeq\frac{\pi}{2}\frac{D}{M_{\mathrm{s}}}, \tag*{(15)}

Since Eq. (15) contains no damping parameter at leading order, this estimate is insensitive to chiral damping within the present approximation [41, 42]. A second DW flow estimate of HDMIH_{\mathrm{DMI}} is obtained from the minimum of the precessional-flow velocity as a function of HxH_{x} [30]. Chiral damping weakly breaks the symmetry of this velocity curve. Expanding the average drift velocity [30, 31] to first order in η\eta yields the following (Hx⋆)flow(H_{x}^{\star})_{\mathrm{flow}} field corresponding to the velocity minimum

(Hx⋆)flow≃−HDMI+sηα0Hzπ[1+58(HKDWHz)2].(16)(H_{x}^{\star})_{\mathrm{flow}}\simeq-H_{\mathrm{DMI}}+\frac{s\eta\alpha_{0}H_{z}}{\pi}\left[1+\frac{5}{8}\left(\frac{H_{K}^{\mathrm{DW}}}{H_{z}}\right)^{2}\right]. \tag*{(16)}

The derivation and comparison with the full numerical minimum are given in the Supplementary material [49]. Equation (16) shows that the displacement is suppressed by ηα0\eta\alpha_{0}. For the parameters of the investigated samples, the calculated shift is of the order of 1 mT and is comparable to the experimental resolution of the in-plane-field coils. To summarize, domain-wall elasticity does not enter the flow physics, the Walker velocity is damping independent to leading order, and the chiral-damping-induced displacement of the drift-velocity minimum is experimentally negligible. We conclude that DW flow measurements are expected to agree with BLS, providing a reference for the present analysis of iDMI strength [48].

Experimental interpretation – We now apply the symmetry-decomposed DW creep model of Eqs. (6)–(7) to the samples introduced above. We test whether the model yields iDMI values consistent with BLS and DW flow measurements and whether the fitted antisymmetric response helps identify the physical origin of the discrepancies found with conventional DW creep analysis.

The symmetry-decomposed estimates are obtained by fitting each measured (ln⁡[v(Hx)/v(0)])(\ln[v(H_{x})/v(0)]) as

ln⁡v(Hx)=(ln⁡v)S+(ln⁡v)A,(17)\ln v(H_{x})=(\ln v)_{\mathrm{S}}+(\ln v)_{\mathrm{A}}, \tag*{(17)}

where (ln⁡v)S,(ln⁡v)A(\ln v)_{\mathrm{S}},(\ln v)_{\mathrm{A}} come from Eqs. (6)–(7). We see how Hx=−HDMIH_{x}=-H_{\mathrm{DMI}} is the symmetry axis of (ln⁡v)S(\ln v)_{\mathrm{S}}, while (ln⁡v)A(\ln v)_{\mathrm{A}} displaces the velocity minimum from this center. In Fig. 4(i), The blue circles represent BLS measurements, the green squares are the DW flow estimate extracted from the Walker-velocity plateau using Eq. (15), the red triangles are the conventional DW creep estimate [20], and the black stars are the symmetry-decomposed DW creep estimate [Eq. (17)]. We immediately notice how the symmetry-decomposed DMI estimates agree with independent BLS measurements and, where available, DW flow estimates. The conventional DW creep estimates, on the other hand, show substantial discrepancies. An in depth discussion of the error estimation of D=μ0MsΔHDMID=\mu_{0}M_{\mathrm{s}}\Delta H_{\mathrm{DMI}} for each sample as well as the fitted values of ν1,2,3\nu_{1,2,3} and HKDW,HDMIH_{K}^{\mathrm{DW}}, H_{\mathrm{DMI}} are reported in [49]. We can now discuss some representative cases: sample Pt/Co/Pt(3) [Fig. 4(iii)] has a small antisymmetric component and a symmetric part centered close to zero, with a weakly biased conventional DW creep estimate. Sample Pt/Co/Ta [Fig. 4(iv)] has a large antisymmetric component and a symmetric part centered at large HxH_x, with a large displacement of the minimum from the symmetry axis. Sample HZO/Co/Pt [Fig. 4(v)] combines a small antisymmetric component with a symmetric part centered at large HxH_x, yet retains a finite discrepancy between the conventional DW creep and BLS estimates [inset of Fig. 4(v)]. The inverted sign of HDMIH_{\mathrm{DMI}} in this sample follows from the inverted layer stacking: Pt is on top of Co, in contrast to the Pt/Co/XX series.

Figure 4. i) iDMI constant values measured with different techniques. No Flow comparison was possible for sample HZO/Co/Pt. ii) Normalized log of the velocity ln⁡(v(Hx)/v(0))\ln(v(H_x)/v(0)) in the asymmetric bubble expansion on sample Pt/Co/Ta performed at different OOP fields ranging from μ0Hz=13\mu_0 H_z = 13 mT to 50 mT. A vertical offset is applied to improve clarity. The inset on the top shows the location of the minimum μ0Hxmin⁡\mu_0 H_x^{\min} as a function of the applied OOP field μ0Hz\mu_0 H_z. Representative velocity decompositions according to Eqs. (6)–(7) for samples iii) Pt/Co/Pt(3), iv) Pt/Co/Ta, and v) HZO/Co/Pt. The arrows identify the magnetization directions of the two domains and the intervening DW. The blue dashed line represents the symmetric part (6) of the curve, while the orange dashed line represents its antisymmetric component (7). The shaded area represents the mixed Bloch-Néel regime.

Comparing these decompositions with the BLS spectra provides further insights. As shown in Fig. 3(i), the spectra provide both the Stokes/anti-Stokes frequency shift used to extract DMI and the relative linewidth asymmetry Γ−\Gamma_{-}, which we associate with chiral damping [45, 49] via Eq. (14). Across the Pt/Co/XX series [Fig. 3(ii)], the largest linewidth asymmetries occur in the samples where the discrepancy between conventional DW creep analysis and BLS is largest. However, sample HZO/Co/Pt in Fig. 4(v) shows that this correspondence is not one-to-one: a sizable minimum displacement persists at large

DMI even when the BLS chiral damping signature is weak [Fig. 3(ii)], consistent with an elastic contribution to the asymmetry ν2\nu_2. Its discrepancy is nevertheless smaller than that of Pt/Co/Ta, but accounting for the asymmetric part yields a DMI value closer to the BLS benchmark. Combining these observations, we conclude that elastic and chiral damping contributions mix with sample-dependent weights, obscuring the origin of the discrepancy in the full creep velocity curve.

We emphasize, in fact, how the magnitude of the antisymmetric component alone cannot establish the relative roles of DW elasticity and chiral damping. At a single HzH_z, the fit determines the net ν2\nu_2 of Eq. (11), but does not separate its elastic and dissipative contributions. Such a separation would require extracting ν2\nu_2 from velocity curves measured over a range of HzH_z, since the elastic term varies through Cz∝Hz−1/4C_z \propto H_z^{-1/4}.

The different HzH_z dependencies of elasticity and chiral damping in ν2\nu_2 (see Eq. (11)) provide a way to test this interpretation. Fig. 4(ii) shows that the conventional DW creep minimum is not a fixed material observable. In sample Pt/Co/Ta, the normalized DW creep curves ln⁡[v(Hx)/v(0)]\ln[v(H_x)/v(0)], measured over OOP fields spanning μ0Hz=13\mu_0 H_z = 13–5050 mT (all values below the depinning field for sample Pt/Co/Ta, μ0Hd=55.3±0.3\mu_0 H_d = 55.3 \pm0.3 mT [49]), display a systematic displacement of μ0Hxmin⁡\mu_0 H_x^{\min} with the out-of-plane driving field. This trend, summarized in the inset, is consistent with Eq. (13): changing HzH_z modifies the activation-barrier contribution through CzC_z, and therefore changes the relative weight of the elastic term in the shift of the velocity minimum.

Conclusion - In this work, we analyzed how domain-wall elasticity and chiral damping collectively affect the creep dynamics of chiral domain walls. The derived model shows that the antisymmetric component responsible for shifting the domain-wall creep-velocity minimum away from the DMI field arises from a nontrivial interplay of these contributions. Moreover, because the domain-wall stiffness contribution depends on the out-of-plane driving field, the position of this minimum can vary with the measurement conditions rather than representing an intrinsic material parameter. Fits to velocity curves measured in a series of Pt/Co samples with different overlayers and growth procedures yield DMI values consistent with BLS and the available DW flow benchmarks, whose iDMI signatures are not affected by the same interplay between elasticity and chiral damping. With additional measurement of the HzH_z dependence of the ν2\nu_2 component of the velocity curves (see Eq. (11)), creep measurements could also serve as a probe of chiral damping in magnetic thin films, providing a basis for future quantitative comparisons with BLS.

Acknowledgments - We acknowledge R. Lavrijsen for the fruitful discussions. This project is supported by the European Union (EU) and the Italian ministry of University and Research (MUR) under the grant PRIN 2022 “Metrology for spintronics: A machine learning approach for the reliable determination of the Dzyaloshinskii-Moriya interaction (MetroSpin)”. Project code: 2022SA-YARY. Yasser Shokr gratefully acknowledges financial support from TÜBİTAK under grant number 124C509, which supported his work. This work was supported by the France 2030 government plan managed by the Agence Nationale de la Recherche (project PEPR SPIN CHIREX ANR-22-EXSP-0002). This work was supported in part by the European Metrology Programme for Innovation and Research (EMPIR) Programme co-financed by the Participating States under Project 17FUN08-TOPS.

* [email protected]

Supplementary Material: How Elasticity and Chiral Damping Shape Domain-Wall Creep

A. Di Pietro; A. Magni; S. Pizzini; G. Gandini; F. Dörr; Y. Shokr; M. Erkovan; G. Durin; S. Tacchi; M. Madami; G. Carlotti; E. Darwin; A. J. Huxtable; C. Rinaldi; C. H. Marrows; B. J. Hickey; V. Puliafito; M. Kuepferling

Supplementary.pdf

Sample Preparation and Magnetic Characterization

Previously characterized Co-based samples

Samples a1-a5 are ultrathin Co-based heterostructures previously prepared and characterized in Ref. [1]. All samples contain a nominally 0.80.8-nm-thick Co layer deposited on a common Ta(5)/Pt(3)\mathrm{Ta}(5)/\mathrm{Pt}(3) underlayer, while the material and thickness of the upper interface were varied. The nominal layer sequences, listed from bottom to top and with all thicknesses expressed in nanometers, are reported in Table I. The complete fabrication and initial characterization were reported previously in Ref. [1]; here we summarize only the information required for the analysis presented in this work.

Sample IDLayer structure
a1Ta(5)/Pt(3)/Co(0.8)/Pt(3)/Ta(3)
a2Ta(5)/Pt(3)/Co(0.8)/Pt(1)/Ta(3)
a3Ta(5)/Pt(3)/Co(0.8)/Ir(3)/Ta(3)
a4Ta(5)/Pt(3)/Co(0.8)/Ir(1)/Ta(3)
a5Ta(5)/Pt(3)/Co(0.8)/Ta(3)
a6Hf0.5_{0.5}Zr0.5_{0.5}O2_2(10)/Co(0.7)/Pt(6.0)

Table I. Layer structures of the investigated samples. Layer thicknesses are given in nanometers.

HZO sample

The a6 sample consists of an SiO/TiN/Hf0.5Zr0.5O2(10)/Co(0.7)/Pt(6.0)\mathrm{SiO}/\mathrm{TiN}/\mathrm{Hf}_{0.5}\mathrm{Zr}_{0.5}\mathrm{O}_{2}(10)/\mathrm{Co}(0.7)/\mathrm{Pt}(6.0) stack fabricated at the Fraunhofer Institute Dresden and characterized at the Politecnico di Milano. Atomic Layer Deposition (ALD) was used for TiN and Hf0.5Zr0.5O2\mathrm{Hf}_{0.5}\mathrm{Zr}_{0.5}\mathrm{O}_{2} (HZO), while in-situ Physical Vapour Deposition (PVD) was used for the metallic layers (Co/Pt).

Brillouin Light Scattering Measurements

DMI evaluation

Independent measurements of the interfacial Dzyaloshinskii-Moriya interaction were performed by Brillouin light scattering on samples taken from the same wafers as those used for the domain-wall measurements.

In a typical BLS experiment, a monochromatic laser beam is focused onto the sample surface, and the backscattered light is collected by the same objective and frequency analyzed by a high-resolution spectrometer. BLS measurements were performed in the Damon–Eshbach (DE) geometry by applying an external magnetic field HH, sufficiently large to magnetize the ferromagnetic film in-plane, and sweeping the spin-wave wave vector kk perpendicular to the magnetization direction (Fig. 1). During the BLS scattering process, magnons can be either created (Stokes process) or annihilated (anti-Stokes process), resulting in a redshift or blueshift of the scattered light relative to the incident beam. In wave-vector space, Stokes and anti-Stokes processes correspond to magnons with opposite wave vectors. In the presence of DMI, the frequency difference between DE modes propagating in opposite direction can be written as:

Δf(k)=fS−fAS=γ0DπMsk,(18)\Delta f(k) = f_{\mathrm{S}} - f_{\mathrm{AS}} = \frac{\gamma_0 D}{\pi M_{\mathrm{s}}} k, \tag*{(18)}

where kk is the in-plane spin-wave wave vector and γ0\gamma_0 is the gyromagnetic ratio. The sign of DD was assigned using the field, wave-vector, and Stokes/anti-Stokes conventions described below. The BLS measurements probe an area with a lateral dimension of approximately 10-40 μm\mu\mathrm{m} and spin waves with wavelengths of the order of 0.5 μm\mu\mathrm{m}. The technique therefore provides a spatially averaged measurement that is complementary to domain-wall creep, which is particularly sensitive to disorder and pinning on nanometer length scales.

Extraction of chiral damping from the BLS linewidths

The Stokes and anti-Stokes components of each Brillouin-light-scattering spectrum were fitted independently with Lorentzian line shapes. Their linewidths, denoted by ΓS\Gamma_{\mathrm{S}} and ΓAS\Gamma_{\mathrm{AS}}, characterize the dissipation of counterpropagating spin waves. Following Kim et al. [5], the linewidths for the field, magnetization, and wave-vector convention adopted here can be

Sample IDMsM_s (MA/m)AA (pJ/m)Δ\Delta (nm)KeffK_{\mathrm{eff}} (MJ/m3^3)
a11.58±0.031.58 \pm0.03b13±213 \pm2b4.8±0.44.8 \pm0.4b0.490±0.020.490 \pm0.02b
a21.74±0.041.74 \pm0.04b14±214 \pm2b5.5±0.85.5 \pm0.8b0.420±0.030.420 \pm0.03b
a31.14±0.051.14 \pm0.05b13±413 \pm4a5.6±0.85.6 \pm0.8a0.414±0.0010.414 \pm0.001b
a41.20±0.091.20 \pm0.09b12±412 \pm4a4.8±0.84.8 \pm0.8a0.513±0.0020.513 \pm0.002b
a50.90±0.020.90 \pm0.02b23±623 \pm6a6.3±0.96.3 \pm0.9a0.563±0.0050.563 \pm0.005b
a61.42±0.02∣1.42 \pm0.02 |20 \pm 5sup‘c‘∣{sup}`c` |5.4 \pm 0.9$0.69 \pm 0.07

Table II. Saturation magnetization MsM_s, Exchange stiffness AA, Domain wall (DW) width parameter Δ\Delta, effective anisotropy KeffK_{\mathrm{eff}} for the full sample series of Table I.

written as

ΓS=C(α0+αc)(1+rf),ΓAS=C(α0−αc)(1−rf),(19)\Gamma_{\mathrm{S}} = C(\alpha_0 + \alpha_{\mathrm{c}})(1 + r_f), \qquad\Gamma_{\mathrm{AS}} = C(\alpha_0 - \alpha_{\mathrm{c}})(1 - r_f), \tag*{(19)}

where CC is the common wave-vector-even linewidth prefactor, α0\alpha_0 is the ordinary Gilbert-damping coefficient, and αc\alpha_{\mathrm{c}} is its wave-vector-odd contribution. The DMI frequency contribution is determined from

rf=∣fS∣−∣fAS∣∣fS∣+∣fAS∣=ωDMIω0.(20)r_f = \frac{\lvert f_{\mathrm{S}}\rvert- \lvert f_{\mathrm{AS}}\rvert}{\lvert f_{\mathrm{S}}\rvert+ \lvert f_{\mathrm{AS}}\rvert} = \frac{\omega_{\mathrm{DMI}}}{\omega_0}. \tag*{(20)}

We define the sum and difference of the measured linewidths as

Γ+=ΓS+ΓAS,Γ−=ΓS−ΓAS.(21)\Gamma_{+} = \Gamma_{\mathrm{S}} + \Gamma_{\mathrm{AS}}, \qquad\Gamma_{-} = \Gamma_{\mathrm{S}} - \Gamma_{\mathrm{AS}}. \tag*{(21)}

Equation (2) then gives

Schematic diagram of the BLS interaction setup

Figure 1 (PDF p. 13). Schematic diagram of the BLS interaction setup. The bottom panel shows a typical BLS spectrum in the absence of iDMI (dashed line) and with iDMI (solid line) for negative (left) and positive (right) values of DD, respectively. Here kilightk_i^{\mathrm{light}} denotes the incident light wave numbers, while HH and MM are the applied in-plane field and magnetization, respectively.

Γ+=2C(α0+rfαc),Γ−=2C(αc+rfα0).(22)\Gamma_{+}=2\mathcal{C}(\alpha_{0}+r_f\alpha_{c}), \qquad\Gamma_{-}=2\mathcal{C}(\alpha_{c}+r_f\alpha_{0}). \tag*{(22)}

Eliminating the common prefactor C\mathcal{C} yields the relative chiral-damping contribution directly:

αcα0=Γ−−rfΓ+Γ+−rfΓ−.(23)\frac{\alpha_{c}}{\alpha_{0}}=\frac{\Gamma_{-}-r_f\Gamma_{+}}{\Gamma_{+}-r_f\Gamma_{-}}. \tag*{(23)}

This expression removes the DMI-induced contribution.

Flow-Regime Characterization

At sufficiently large driving fields, domain-wall motion is no longer governed by the thermally activated energy barriers characteristic of the creep regime [6]. The dynamics of chiral domain walls can instead be described within the two-dimensional framework developed for chiral domain walls [2]. Flow-regime measurements therefore provide a reference determination of the DMI that is substantially less sensitive to the pinning and elasticity effects that complicate the interpretation of asymmetric bubble expansion in the creep regime (see Main text).

SampleΓS\Gamma_{\mathrm{S}}ΓAS\Gamma_{\mathrm{AS}}ΔfDMI\Delta f_{\mathrm{DMI}}rfr_fαc/α0\alpha_c/\alpha_0
(GHz)(GHz)(GHz)(-)(%)
a13.0±0.13.0 \pm0.13.0±0.23.0 \pm0.20.1±0.30.1 \pm0.30.4±0.90.4 \pm0.90±40 \pm4
a23.2±0.13.2 \pm0.13.0±0.23.0 \pm0.20.2±0.20.2 \pm0.20.4±0.40.4 \pm0.43±43 \pm4
a34.3±0.34.3 \pm0.35.5±0.95.5 \pm0.9−0.7±0.5-0.7 \pm0.5−2±1-2 \pm1−11±9-11 \pm9
a45.0±0.25.0 \pm0.27.0±0.57.0 \pm0.5−0.9±0.3-0.9 \pm0.3−1.9±0.6-1.9 \pm0.6−15±4-15 \pm4
a54.6±0.24.6 \pm0.27.3±0.77.3 \pm0.7−1.9±0.2-1.9 \pm0.2−3.8±0.5-3.8 \pm0.5−20±5-20 \pm5
a612.1±0.512.1 \pm0.513.7±0.713.7 \pm0.71.1±0.11.1 \pm0.12.8±0.32.8 \pm0.3−3±3-3 \pm3

Table III. Stokes (ΓS\Gamma_{\mathrm{S}}) and anti-Stokes linewidths (ΓAS\Gamma_{\mathrm{AS}}), fitted DMI frequency splitting ΔfDMI\Delta f_{\mathrm{DMI}}, normalized frequency asymmetry rfr_f and extracted relative chiral damping αc/α0\alpha_c/\alpha_0.

Figure 2 (PDF p. 14). (i) Stokes and anti-Stokes BLS linewidths, ΓS\Gamma_{\mathrm{S}} and ΓAS\Gamma_{\mathrm{AS}}, for samples a1-a5 and a6. (ii) Relative chiral-damping contribution αc/α0\alpha_c/\alpha_0, obtained after removing the DMI-induced frequency contribution from the measured linewidth asymmetry according to Eq. (6).

Flow-regime measurements were performed on all Co-base samples described in § 3 A (i.e. samples a1-a5). For each sample, two complementary measurements were carried out. First, the DW velocity was measured as a function of the out-of-plane field HzH_z in order to identify the characteristic Walker or high-velocity plateau regime. From the full velocity measurements (see Fig. 3), the depinning fields were extracted according to the procedure detailed in Ref.[7]. Second, the domain-wall velocity was measured as a function of the in-plane field HxH_x at a fixed out-of-plane driving field HzH_z within the flow regime.

DMI constant from the Walker/plateau velocity

A first and independent estimate of the DMI strength was obtained from the OOP field dependence of the domain-wall velocity plateaus. In systems with sufficiently strong DMI, the conventional velocity drop expected immediately above the Walker field is replaced by an extended high-velocity saturation field region. The characteristic Walker or plateau velocity is related to the DMI constant through [2, 8]

vW=π2γDMs,(24)v_W = \frac{\pi}{2}\gamma\frac{D}{M_s}, \tag*{(24)}

where vWv_W is the characteristic Walker velocity, γ\gamma is the gyromagnetic ratio, DD is the interfacial DMI energy density, and MsM_s is the saturation magnetization.

The domain-wall velocity was measured as a function of μ0Hz\mu_0 H_z for every sample in the a1-a5 series, allowing the corresponding characteristic velocity vWv_W to be determined. The complete set of measurements is shown in Fig. 3. The DMI constants obtained using Eq. (7) are listed in Table IV.

Sampleμ0Hz\mu_0 H_zμ0HDMIflow\mu_0 H^{\mathrm{flow}}_{\mathrm{DMI}}μ0Hd\mu_0 H_dDWalkerD_{\mathrm{Walker}}AA
(mT)(mT)(mT)(mJ/m2^2)(pJ/m)
a1a140≈0\approx036.7±0.536.7 \pm0.5n.o.n.o.
a2a240≈0\approx030.22±0.0430.22 \pm0.04n.o.n.o.
a3a350−70±7-70 \pm731.9±0.531.9 \pm0.5−0.45±0.05-0.45 \pm0.0513±413 \pm4
a4a450−88±9-88 \pm938.1±0.638.1 \pm0.6−0.51±0.05-0.51 \pm0.0512±412 \pm4
a5a5100−160±16-160 \pm1655.3±0.355.3 \pm0.3−0.91±0.09-0.91 \pm0.0923±623 \pm6
a6a6n.m.n.m.n.m.n.m.n.m.

Table IV. Applied OOP field μ0Hz\mu_0H_z, DMI field μ0HDMIflow\mu_0H^{\mathrm{flow}}_{\mathrm{DMI}}, depinning field μ0Hd\mu_0H_d, DMI constant DWalkerD_{\mathrm{Walker}}, and exchange stiffness obtained AA from flow-regime domain-wall velocity measurements. The DMI field is determined from the compensation minimum in the in-plane-field-dependent velocity. When a Walker or high-velocity plateau is observed, the DMI constant is obtained from Eq. (7), and the exchange stiffness is subsequently calculated using Eq. (12). A relative uncertainty of 10% is assigned to the graphically extracted values of μ0HDMIflow\mu_0H^{\mathrm{flow}}_{\mathrm{DMI}} and DWalkerD_{\mathrm{Walker}}. The uncertainty in AA additionally includes the propagated uncertainties in MsM_s and KeffK_{\mathrm{eff}} (see Table I). No Walker plateau was observed for samples a1a1 and a2a2, preventing the extraction of DWalkerD_{\mathrm{Walker}} and AA by this procedure. No flow-regime measurements were performed on the a6a6 sample. Here, n.o. denotes “not observed” and n.m. denotes “not measured”.

The DMI constants obtained from the flow-regime velocity measurements are reported in Table IV and compared with the independently determined BLS values in Table VI. The general agreement between the two techniques supports the DMI values obtained in the creep-regime analysis presented in the main text.

SampleDBLSD_{\mathrm{BLS}} [1]DWalkerD_{\mathrm{Walker}}DcreepoldD_{\mathrm{creep}}^{\mathrm{old}}DcreepnewD_{\mathrm{creep}}^{\mathrm{new}}
(mJ/m2^2)(mJ/m2^2)(mJ/m2^2)(mJ/m2^2)
a10.084 ±\pm 0.005n.o.0.043 ±\pm 0.0050.060 ±\pm 0.008
a20.170 ±\pm 0.006n.o.0.022 ±\pm 0.0040.05 ±\pm 0.02
a3−0.390±0.006-0.390 \pm0.006−0.45±0.05-0.45 \pm0.05−0.24±0.06-0.24 \pm0.06−0.40±0.06-0.40 \pm0.06
a4−0.560±0.008-0.560 \pm0.008−0.51±0.05-0.51 \pm0.05−0.3±0.1-0.3 \pm0.1−0.40±0.08-0.40 \pm0.08
a5−0.810±0.005-0.810 \pm0.005−0.91±0.09-0.91 \pm0.09−0.45±0.06-0.45 \pm0.06−0.8±0.1-0.8 \pm0.1
a60.8 ±\pm 0.1n.m.0.5 ±\pm 0.10.7 ±\pm 0.2

Table VI. DMI values obtained from BLS (DBLSD_{\mathrm{BLS}}), Walker-field (DWalkerD_{\mathrm{Walker}}), conventional-creep [18] (DcreepoldD_{\mathrm{creep}}^{\mathrm{old}}) and new-creep analyses (DcreepnewD_{\mathrm{creep}}^{\mathrm{new}}).

Creep characterization of the a6 sample

We were not able to measure the a6 sample up to the depinning regime as the same setup used for samples a1-a5 was unavailable. To check the validity of the creep regime, we performed low field bubble expansion measurements to confirm the linear scaling of ln⁡(v)\ln(v) [9]

ln⁡(v)=v0+χ0(μ0Hz)−1/4(25)\ln(v) = v_0 + \chi_0(\mu_0 H_z)^{-1/4} \tag*{(25)}

The results are shown in Fig. 4.

Figure 3 (PDF p. 16). Out-of-plane-field-dependent domain-wall dynamics the a1−a5a1-a5 samples investigated in this work. (a) Domain-wall velocity as a function of μ0Hz\mu_0 H_z. The characteristic Walker or high-velocity plateau value vWv_W is used to determine the DMI constant through Eq. (7). (b) Corresponding creep-law representation, ln⁡v\ln v versus (μ0Hz)−1/4(\mu_0 H_z)^{-1/4}. (c) Depinning fields μ0Hd\mu_0 H_d extracted from the full creep-to-depinning [7] fitting procedure for each sample.

DMI field from in-plane-field compensation

For each sample, asymmetric bubble expansion was measured as a function of the applied in-plane field μ0Hx\mu_0 H_x, while maintaining the out-of-plane driving field within the flow regime (see Table IV). The velocities of the up-down and down-up portions of the expanding bubble were evaluated independently. The velocity curves exhibit a minimum near the compensation condition at which the applied in-plane field balances the DMI effective field [2, 8, 10],

Figure 4 (PDF p. 17). Out-of-plane-field-dependent domain-wall dynamics the a6 sample. The dashed line represents the linear fit of Eq. (8)

Hx=−HDMI.(26)H_x = -H_{\mathrm{DMI}}. \tag*{(26)}

The flow-regime asymmetric bubble-expansion measurements for all samples are shown in Figure 5. The recorded depinning fields μ0Hd\mu_0 H_d for the a1-a5 samples and are reported in Table IV alongside the recorded values of the effective DMI field μ0HDMI\mu_0 H_{\mathrm{DMI}}.

Figure 5. Domain-wall velocity measured as a function of the in-plane field μ0Hx\mu_0 H_x in the flow regime for samples a1a_1–a5a_5. For each sample, the velocities of the up-down (u/d) and down-up (d/u) portions of the expanding bubble domain are shown separately. The position of the velocity minimum identifies the compensation condition μ0Hx=−μ0HDMI≡−μ0HDMIflow\mu_0 H_x=-\mu_0 H_{\mathrm{DMI}}\equiv-\mu_0 H_{\mathrm{DMI}}^{\mathrm{flow}} [10].

Exchange-stiffness consistency check

The two flow-regime measurements described above provide complementary information. The in-plane-field-dependent asymmetric bubble expansion determines the DMI effective field HDMIH_{\mathrm{DMI}}, whereas the out-of-plane-field-dependent velocity provides an independent estimate of the DMI energy density DD. With the sign convention adopted in this work, the DMI constant and the corresponding effective field are related by

D=μ0MsHDMIΔ,(27)D = \mu_0 M_s H_{\mathrm{DMI}}\Delta, \tag*{(27)}

where

Δ=AKeff(28)\Delta= \sqrt{\frac{A}{K_{\mathrm{eff}}}} \tag*{(28)}

is the domain-wall-width parameter. Combining Eqs. (10) and (11) gives

A=Keff(Dμ0MsHDMI)2.(29)A=K_{\mathrm{eff}}\left(\frac{D}{\mu_0 M_{\mathrm{s}}H_{\mathrm{DMI}}}\right)^2. \tag*{(29)}

The extraction of AA is not the primary purpose of the flow measurements. Nevertheless, Eq. (12) provides an independent consistency check on the material parameters entering the creep-regime calculations. This is particularly useful in ultrathin multilayers, where direct measurements of the exchange stiffness are experimentally challenging. Conventional approaches determine the exchange stiffness AA from the temperature dependence of the saturation magnetization, using Bloch-law or related spin-wave analyses of Ms(T)M_{\mathrm{s}}(T) [11, 12], or from the exchange contribution to spin-wave spectra measured by ferromagnetic resonance, perpendicular standing spin-wave spectroscopy, or Brillouin light scattering [4, 13]. In ultrathin multilayers, however, these methods can be affected by interface anisotropy, magnetic dead layers, proximity-induced magnetic moments, finite-size effects, and uncertain exchange boundary conditions. Applying Eq. (12) to the samples of the a1-a5 series provides the exchange-stiffness estimates reported in Table IV.

Effect of chiral damping in the precessional regime

We consider the one-dimensional precessional dynamics with an angle-dependent damping

α(ϕ)=α0(1+ηcos⁡ϕ),∣αcα0∣≡∣η∣<1.(30)\alpha(\phi) = \alpha_0 (1 + \eta\cos\phi), \qquad\left|\frac{\alpha_c}{\alpha_0}\right| \equiv|\eta| < 1. \tag*{(30)}

The deviation from the DMI compensation field is defined as

X=Hx+HDMI,(31)X = H_x + H_{\mathrm{DMI}}, \tag*{(31)}

so that the compensation point corresponds to X=0X = 0. The internal torque entering the qq-ϕ\phi equations is written as

G(ϕ,X)=HKDW2sin⁡2ϕ+πX2cos⁡ϕ.(32)G(\phi, X) = \frac{H_K^{DW}}{2}\sin2\phi+ \frac{\pi X}{2}\cos\phi. \tag*{(32)}

The cycle-averaged drift velocity is

⟨v⟩=γΔJI,(33)\langle v\rangle= \gamma\Delta\frac{J}{I}, \tag*{(33)}

where

I=∫02π1+α(ϕ)2Hz−α(ϕ)G(ϕ,X) dϕ,(34)I = \int_{0}^{2\pi} \frac{1+\alpha(\phi)^2}{H_z-\alpha(\phi)G(\phi,X)}\,\mathrm{d}\phi, \tag*{(34)}

and

J=∫02πα(ϕ)Hz+G(ϕ,X)Hz−α(ϕ)G(ϕ,X) dϕ.(35)J = \int_{0}^{2\pi} \frac{\alpha(\phi)H_z+G(\phi,X)}{H_z-\alpha(\phi)G(\phi,X)}\,\mathrm{d}\phi. \tag*{(35)}

To perform the asymptotic expansion, we introduce the dimensionless parameters

ξ=α0HKDW2Hz,ρ=πX2Hz,(36)\xi= \frac{\alpha_0 H_K^{DW}}{2H_z}, \qquad\rho= \frac{\pi X}{2H_z}, \tag*{(36)}

so that

G(ϕ,X)Hz=ξα0sin⁡2ϕ+ρcos⁡ϕ.(37)\frac{G(\phi,X)}{H_z} = \frac{\xi}{\alpha_0}\sin2\phi+\rho\cos\phi. \tag*{(37)}

Multiplying the denominator integral by HzH_z, we define Iˉ=HzI\bar{I}=H_zI, giving

⟨v⟩=γΔHzJIˉ.(38)\langle v\rangle= \gamma\Delta H_z\frac{J}{\bar{I}}. \tag*{(38)}

The integrands are expanded simultaneously in the small parameters ρ\rho, η\eta and ξ\xi. After integration over one full precessional cycle, all terms linear in ξ\xi vanish by symmetry, yielding

Iˉ=2π(1+α02)[1+ξ22+α0(1+3α02)2(1+α02)ηρ+α022ρ2+3α0(3+5α02)4(1+α02)ηρξ2+3α022ρ2ξ2+O(η2,ηρ2,ρ3,ξ4)],(39)\bar{I}=2\pi(1+\alpha_0^2)\left[1+\frac{\xi^2}{2}+\frac{\alpha_0(1+3\alpha_0^2)}{2(1+\alpha_0^2)}\eta\rho+\frac{\alpha_0^2}{2}\rho^2+\frac{3\alpha_0(3+5\alpha_0^2)}{4(1+\alpha_0^2)}\eta\rho\xi^2+\frac{3\alpha_0^2}{2}\rho^2\xi^2+\mathcal{O}\left(\eta^2,\eta\rho^2,\rho^3,\xi^4\right)\right], \tag*{(39)}

and

J=2πα0[1+1+α022α02ξ2+α0ηρ+1+α022ρ2+3(1+2α02)2α0ηρξ2+3(1+α02)2ρ2ξ2+O(η2,ηρ2,ρ3,ξ4)].(40)J=2\pi\alpha_0\left[1+\frac{1+\alpha_0^2}{2\alpha_0^2}\xi^2+\alpha_0\eta\rho+\frac{1+\alpha_0^2}{2}\rho^2+\frac{3(1+2\alpha_0^2)}{2\alpha_0}\eta\rho\xi^2+\frac{3(1+\alpha_0^2)}{2}\rho^2\xi^2+\mathcal{O}\left(\eta^2,\eta\rho^2,\rho^3,\xi^4\right)\right]. \tag*{(40)}

Series division then gives

⟨v⟩=γΔHzα01+α02[1+ξ22α02+ρ22+ρ2ξ2+α0(1−α02)2(1+α02)ηρ(41)\langle v\rangle=\gamma\Delta H_z\frac{\alpha_0}{1+\alpha_0^2}[1+\frac{\xi^2}{2\alpha_0^2}+\frac{\rho^2}{2}+\rho^2\xi^2+\frac{\alpha_0(1-\alpha_0^2)}{2(1+\alpha_0^2)}\eta\rho \tag*{(41)}
+5+5α02−2α044α0(1+α02)ηρξ2+O(η2,ηρ2,ρ3,ξ4)].(42)+\frac{5+5\alpha_0^2-2\alpha_0^4}{4\alpha_0(1+\alpha_0^2)}\eta\rho\xi^2+\mathcal{O}(\eta^2,\eta\rho^2,\rho^3,\xi^4)]. \tag*{(42)}

The position of the extremum is obtained from

∂⟨v⟩∂ρ=0.(43)\frac{\partial\langle v\rangle}{\partial\rho}=0. \tag*{(43)}

Differentiating the expanded velocity gives

(1+2ξ2)ρmin⁡+η[α0(1−α02)2(1+α02)+5+5α02−2α044α0(1+α02)ξ2]=0.(44)(1+2\xi^2)\rho_{\min}+\eta\left[\frac{\alpha_0(1-\alpha_0^2)}{2(1+\alpha_0^2)}+\frac{5+5\alpha_0^2-2\alpha_0^4}{4\alpha_0(1+\alpha_0^2)}\xi^2\right]=0. \tag*{(44)}

which gives

ρmin⁡=−ηα0(1−α02)2(1+α02)+5+5α02−2α044α0(1+α02)ξ21+2ξ2.(45)\rho_{\min}=-\eta\frac{\dfrac{\alpha_0(1-\alpha_0^2)}{2(1+\alpha_0^2)}+\dfrac{5+5\alpha_0^2-2\alpha_0^4}{4\alpha_0(1+\alpha_0^2)}\xi^2}{1+2\xi^2}. \tag*{(45)}

Since the numerator and denominator have been calculated only up to second order in ξ\xi, their ratio must be consistently re-expanded to the same order. This yields

ρmin⁡−η[α0(1−α02)2(1+α02)+5+α02+2α044α0(1+α02)ξ2]+O(η2,ηξ4).(46)\rho_{\min}-\eta\left[\frac{\alpha_0(1-\alpha_0^2)}{2(1+\alpha_0^2)}+\frac{5+\alpha_0^2+2\alpha_0^4}{4\alpha_0(1+\alpha_0^2)}\xi^2\right]+\mathcal{O}\left(\eta^2,\eta\xi^4\right). \tag*{(46)}

Finally, using

ρ=πX2Hz,ξ=α0HKDW2Hz,(47)\rho=\frac{\pi X}{2H_z},\qquad\xi=\frac{\alpha_0H_K^{DW}}{2H_z}, \tag*{(47)}

the displacement of the minimum from the DMI compensation field is

(Hx⋆)flow=−HDMI+α0Hzπη1+α02[1−α02+5+α02+2α048(HKDWHz)2]+O.(48)(H_x^\star)_{flow}=-H_{\mathrm{DMI}}+\frac{\alpha_0H_z}{\pi}\frac{\eta}{1+\alpha_0^2}\left[1-\alpha_0^2+\frac{5+\alpha_0^2+2\alpha_0^4}{8}\left(\frac{H_K^{DW}}{H_z}\right)^2\right]+\mathcal{O}. \tag*{(48)}

For α0≪1\alpha_0\ll1, this simplifies to the expressions used in the main text (Eq. (17))

(Hx⋆)flow≃−HDMI+ηα0Hzπ[1+58(HKDWHz)2].(49)(H_x^\star)_{flow}\simeq-H_{\mathrm{DMI}}+\frac{\eta\alpha_0H_z}{\pi}\left[1+\frac{5}{8}\left(\frac{H_K^{DW}}{H_z}\right)^2\right]. \tag*{(49)}

In Figure 6 we report both the minimum obtained numerically from Eq. (16) and the analytical estimate of Eq. (49).

Figure 6. Difference between the minimum of the domain-wall velocity in the flow regime and the DMI effective field μ0δHxmin⁡=μ0[(Hx⋆)flow+HDMI]\mu_0\delta H_x^{\min} = \mu_0[(H_x^\star)_{\mathrm{flow}} + H_{\mathrm{DMI}}] as computed via the full numerical minimum of Eq. (16) and the analytical estimate Eq. (49). The magnetic parameters used in the calculation are from the Pt/Co/X series. We set Gilbert damping α0=0.05\alpha_0 = 0.05 and μ0Hz=120\mu_0 H_z = 120 mT for all samples. The selected HzH_z field ensures correspondence with the precessional flow regime in all samples. See Table IV for the depinning field analysis of the Pt/Co/X series

Derivation of the Three-Amplitude Model

In this section, we derive the symmetry-decomposed model used in the main text. We start from a creep-velocity Ansatz of the form

v(Hx,Hz)=v⋆σeq(Hx)βαeff(Hx)exp⁡[−χ0(σ~(Hx)σ~(0))1/4Hz−1/4].(50)v(H_x,H_z) = v_\star\frac{\sigma_{\mathrm{eq}}(H_x)^\beta}{\alpha_{\mathrm{eff}}(H_x)} \exp\left[-\chi_0 \left(\frac{\widetilde{\sigma}(H_x)}{\widetilde{\sigma}(0)}\right)^{1/4} H_z^{-1/4}\right]. \tag*{(50)}

Here, v⋆v_\star is a field-independent attempt-frequency prefactor, σeq(Hx)\sigma_{\mathrm{eq}}(H_x) is the equilibrium domain-wall energy density, and σ~(Hx)\widetilde{\sigma}(H_x) is the domain-wall stiffness in the long-wavelength limit [14, 15].

σ~(θ)=σeq(θ)+∂2σeq∂θ2(θ),σeq(θ)=min⁡ϕσ(θ,ϕ).(51)\widetilde{\sigma}(\theta) = \sigma_{\mathrm{eq}}(\theta) + \frac{\partial^2 \sigma_{\mathrm{eq}}}{\partial\theta^2}(\theta), \qquad\sigma_{\mathrm{eq}}(\theta) = \min_{\phi} \sigma(\theta,\phi). \tag*{(51)}

The quantity χ0\chi_0 is the creep scaling parameter at Hx=0H_x = 0, whereas β\beta is an effective exponent describing the dependence of the velocity prefactor on the equilibrium wall energy [15]. It should not be confused with the reduced in-plane field βx\beta_x introduced below. Taking the natural logarithm of the velocity normalized at Hx=0H_x = 0 gives

ln⁡[v(Hx,Hz)v(0,Hz)]=βln⁡[σeq(Hx)σeq(0)]+ln⁡[αeff(0)αeff(Hx)]−χ0Hz−1/4[(σ~(Hx)σ~(0))1/4−1].(52)\ln\left[\frac{v(H_x,H_z)}{v(0,H_z)}\right] = \beta\ln\left[\frac{\sigma_{\mathrm{eq}}(H_x)}{\sigma_{\mathrm{eq}}(0)}\right] + \ln\left[\frac{\alpha_{\mathrm{eff}}(0)}{\alpha_{\mathrm{eff}}(H_x)}\right] - \chi_0 H_z^{-1/4}\left[\left(\frac{\widetilde{\sigma}(H_x)}{\widetilde{\sigma}(0)}\right)^{1/4}-1\right]. \tag*{(52)}

For compactness, we define the relative variations

δσeq(Hx)≡σeq(Hx)−σeq(0)σeq(0),δσ~(Hx)≡σ~(Hx)−σ~(0)σ~(0).(53)\delta\sigma_{\mathrm{eq}}(H_x) \equiv\frac{\sigma_{\mathrm{eq}}(H_x)-\sigma_{\mathrm{eq}}(0)}{\sigma_{\mathrm{eq}}(0)}, \qquad\delta\widetilde{\sigma}(H_x) \equiv\frac{\widetilde{\sigma}(H_x)-\widetilde{\sigma}(0)}{\widetilde{\sigma}(0)}. \tag*{(53)}

Equation (52) can then be expressed as

ln⁡[v(Hx,Hz)v(0,Hz)]=βln⁡[1+δσeq(Hx)]+ln⁡[αeff(0)αeff(Hx)]−χ0[(1+δσ~(Hx))1/4−1]Hz−1/4.(54)\ln\left[\frac{v(H_x,H_z)}{v(0,H_z)}\right] = \beta\ln\left[1+\delta\sigma_{\mathrm{eq}}(H_x)\right] + \ln\left[\frac{\alpha_{\mathrm{eff}}(0)}{\alpha_{\mathrm{eff}}(H_x)}\right] - \chi_0\left[\left(1+\delta\widetilde{\sigma}(H_x)\right)^{1/4}-1\right]H_z^{-1/4}. \tag*{(54)}

Considering the experimentally accessible IP field ranges is asymmetric bubble domain expansions (∣μ0Hx∣≤200|\mu_0H_x| \le200 mT) we assume small in-plane-field-induced variations,

∣δσeq(Hx)∣≪1,∣δσ~(Hx)∣≪1,(55)|\delta\sigma_{\mathrm{eq}}(H_x)| \ll1,\qquad|\delta\widetilde{\sigma}(H_x)| \ll1, \tag*{(55)}

Since direct evaluation of σ~(Hx)/σ~(0)\widetilde{\sigma}(H_x)/\widetilde{\sigma}(0), accounting for both mixed and saturated zero-field configurations, shows that the first-order fourth-root expansion is accurate to better than 0.3% for samples a2–a5 over their respective experimental field ranges. Sample a1 provides the most restrictive case, with a maximum error of approximately 3.6 %. With the mentioned assumption, we obtain

ln⁡[v(Hx,Hz)v(0,Hz)]≃β[σeq(Hx)σeq(0)−1]+ln⁡[αeff(0)αeff(Hx)]−Cz[σ~(Hx)σ~(0)−1],(56)\ln\left[\frac{v(H_x,H_z)}{v(0,H_z)}\right] \simeq\beta\left[\frac{\sigma_{\mathrm{eq}}(H_x)}{\sigma_{\mathrm{eq}}(0)}-1\right] + \ln\left[\frac{\alpha_{\mathrm{eff}}(0)}{\alpha_{\mathrm{eff}}(H_x)}\right] - C_z\left[\frac{\widetilde{\sigma}(H_x)}{\widetilde{\sigma}(0)}-1\right], \tag*{(56)}

where

Cz=χ0Hz−1/44.(57)C_z = \frac{\chi_0 H_z^{-1/4}}{4}. \tag*{(57)}

To retain the correct normalizations of the two energy-dependent terms, we define

S0≡σeq(0)σ0,T0≡σ~(0)σ0,(58)S_0 \equiv\frac{\sigma_{\mathrm{eq}}(0)}{\sigma_0},\qquad T_0 \equiv\frac{\widetilde{\sigma}(0)}{\sigma_0}, \tag*{(58)}

where

σ0=4AKeff(59)\sigma_0 = 4\sqrt{AK_{\mathrm{eff}}} \tag*{(59)}

is the Bloch-wall energy density. Equation (56) then becomes

ln⁡[v(Hx,Hz)v(0,Hz)]≃β[σeq(Hx)/σ0S0−1]+ln⁡[αeff(0)αeff(Hx)]−Cz[σ~(Hx)/σ0T0−1].(60)\ln\left[\frac{v(H_x,H_z)}{v(0,H_z)}\right] \simeq\beta\left[\frac{\sigma_{\mathrm{eq}}(H_x)/\sigma_0}{S_0}-1\right]+\ln\left[\frac{\alpha_{\mathrm{eff}}(0)}{\alpha_{\mathrm{eff}}(H_x)}\right]-C_z\left[\frac{\widetilde{\sigma}(H_x)/\sigma_0}{T_0}-1\right]. \tag*{(60)}

The two reference values S0S_0 and T0T_0 are generally different: the prefactor contribution is normalized by the equilibrium wall energy at Hx=0H_x=0, whereas the activation-barrier contribution is normalized by the stiffness at Hx=0H_x=0. We introduce the dimensionless quantities

ζ=πμ0MsΔHDMIσ0,βx=πμ0MsΔHxσ0,κ=πμ0MsΔHKDW2σ0.(61)\zeta=\frac{\pi\mu_0M_s\Delta H_{\mathrm{DMI}}}{\sigma_0},\qquad\beta_x=\frac{\pi\mu_0M_s\Delta H_x}{\sigma_0},\qquad\kappa=\frac{\pi\mu_0M_s\Delta H_K^{\mathrm{DW}}}{2\sigma_0}. \tag*{(61)}

In terms of these quantities, the domain-wall energy density is

σ(θ,ϕ)=σ0[1−ζcos⁡(ϕ−θ)−βxcos⁡ϕ+κcos⁡2(ϕ−θ)].(62)\sigma(\theta,\phi)=\sigma_0\left[1-\zeta\cos(\phi-\theta)-\beta_x\cos\phi+\kappa\cos^2(\phi-\theta)\right]. \tag*{(62)}

For notational compactness, we introduce

X=Hx+HDMI.(63)X=H_x+H_{\mathrm{DMI}}. \tag*{(63)}

The reference axis for all the subsequent symmetry decompositions of the velocity curve is therefore X=0X=0, while the mixed Bloch-Néel region corresponds to

∣X∣≤HKDW.|X|\le H_K^{\mathrm{DW}}.

For the right-hand side of the expanding bubble, we set θ=0\theta=0. Minimizing Eq. (62) with respect to ϕ\phi gives

cos⁡ϕeq={−1,X<−HKDW,XHKDW,∣X∣≤HKDW,1,X>HKDW.(64)\cos\phi_{\mathrm{eq}}= \begin{cases} -1, & X<-H_K^{\mathrm{DW}},\\ \dfrac{X}{H_K^{\mathrm{DW}}}, & |X|\le H_K^{\mathrm{DW}},\\ 1, & X>H_K^{\mathrm{DW}}. \end{cases} \tag*{(64)}

The corresponding equilibrium wall energy is

σeq(X)σ0={1+κ−2κ∣X∣HKDW,∣X∣>HKDW,1−κ(XHKDW)2,∣X∣≤HKDW.(65)\frac{\sigma_{\mathrm{eq}}(X)}{\sigma_0}= \begin{cases} 1+\kappa-2\kappa\dfrac{|X|}{H_K^{\mathrm{DW}}}, & |X|>H_K^{\mathrm{DW}},\\ 1-\kappa\left(\dfrac{X}{H_K^{\mathrm{DW}}}\right)^2, & |X|\le H_K^{\mathrm{DW}}. \end{cases} \tag*{(65)}

It is therefore symmetric around the compensation axis:

σeq(X)=σeq(−X).(66)\sigma_{\mathrm{eq}}(X)=\sigma_{\mathrm{eq}}(-X). \tag*{(66)}

Consequently, the equilibrium-energy contribution to the velocity prefactor contributes only to the symmetric part of the creep curve. Using Eq. (51), the symmetric part of the stiffness is

σ~S(X)σ0={1−κ−2κ∣X∣HKDW,∣X∣>HKDW,1+2κ−5κ(XHKDW)2,∣X∣≤HKDW.(67)\frac{\widetilde{\sigma}_{S}(X)}{\sigma_{0}} = \begin{cases} 1-\kappa-2\kappa\frac{|X|}{H_{K}^{\mathrm{DW}}}, & |X|>H_{K}^{\mathrm{DW}},\\ 1+2\kappa-5\kappa\left(\frac{X}{H_{K}^{\mathrm{DW}}}\right)^{2}, & |X|\le H_{K}^{\mathrm{DW}}. \end{cases} \tag*{(67)}

whereas its antisymmetric part is

σ~A(X)σ0={ζsgn⁡(X),∣X∣>HKDW,ζXHKDW,∣X∣≤HKDW.(68)\frac{\widetilde{\sigma}_{A}(X)}{\sigma_{0}} = \begin{cases} \zeta\operatorname{sgn}(X), & |X|>H_{K}^{\mathrm{DW}},\\ \zeta\frac{X}{H_{K}^{\mathrm{DW}}}, & |X|\le H_{K}^{\mathrm{DW}}. \end{cases} \tag*{(68)}

The normalization point Hx=0H_{x}=0 corresponds to X=HDMIX=H_{\mathrm{DMI}}. The normalization constants are therefore obtained by evaluating the full equilibrium energy and stiffness at this point:

S0=σeq(X)σ0∣X=HDMI,T0=σ~(X)σ0∣X=HDMI.(69)S_{0}=\left.\frac{\sigma_{\mathrm{eq}}(X)}{\sigma_{0}}\right|_{X=H_{\mathrm{DMI}}}, \qquad T_{0}=\left.\frac{\widetilde{\sigma}(X)}{\sigma_{0}}\right|_{X=H_{\mathrm{DMI}}}. \tag*{(69)}

Explicitly,

S0={1+κ−2κ∣HDMI∣HKDW,∣HDMI∣>HKDW,1−κ(HDMIHKDW)2,∣HDMI∣≤HKDW.(70)S_{0}= \begin{cases} 1+\kappa-2\kappa\frac{|H_{\mathrm{DMI}}|}{H_{K}^{\mathrm{DW}}}, & |H_{\mathrm{DMI}}|>H_{K}^{\mathrm{DW}},\\ 1-\kappa\left(\frac{H_{\mathrm{DMI}}}{H_{K}^{\mathrm{DW}}}\right)^{2}, & |H_{\mathrm{DMI}}|\le H_{K}^{\mathrm{DW}}. \end{cases} \tag*{(70)}

and

T0={1−κ−2κ∣HDMI∣HKDW+ζsgn⁡(HDMI),∣HDMI∣>HKDW,1+2κ−5κ(HDMIHKDW)2+ζHDMIHKDW,∣HDMI∣≤HKDW.(71)T_{0}= \begin{cases} 1-\kappa-2\kappa\frac{|H_{\mathrm{DMI}}|}{H_{K}^{\mathrm{DW}}}+\zeta\operatorname{sgn}(H_{\mathrm{DMI}}), & |H_{\mathrm{DMI}}|>H_{K}^{\mathrm{DW}},\\ 1+2\kappa-5\kappa\left(\frac{H_{\mathrm{DMI}}}{H_{K}^{\mathrm{DW}}}\right)^{2}+\zeta\frac{H_{\mathrm{DMI}}}{H_{K}^{\mathrm{DW}}}, & |H_{\mathrm{DMI}}|\le H_{K}^{\mathrm{DW}}. \end{cases} \tag*{(71)}

The effective damping is written as

αeff(ϕ)=α0[1+sηcos⁡ϕ],(72)\alpha_{\mathrm{eff}}(\phi)=\alpha_{0}\left[1+s\eta\cos\phi\right], \tag*{(72)}

where s=±1s=\pm1 labels the two domain-wall chiralities and η\eta is the relative chiral-damping strength [16, 17]. Expanding the logarithmic damping contribution to second order in η\eta gives

ln⁡[αeff(0)αeff(X)]=const.⁡−sηcos⁡ϕeq+η22cos⁡2ϕeq+O(η3).(73)\ln\left[\frac{\alpha_{\mathrm{eff}}(0)}{\alpha_{\mathrm{eff}}(X)}\right]=\operatorname{const.}-s\eta\cos\phi_{\mathrm{eq}}+\frac{\eta^{2}}{2}\cos^{2}\phi_{\mathrm{eq}}+O(\eta^{3}). \tag*{(73)}

The first-order term is antisymmetric in XX, whereas the second-order term is symmetric. Since s=±1s=\pm1 is a convention-dependent chirality label, we absorb the minus sign in the linear term into the definition of ss from this point onward. With the convention used in the main text, the linear contribution to the antisymmetric amplitude is therefore written as sηs\eta. Collecting all symmetric contributions, and removing one common field-independent offset, symmetric part of the function can be written as

(ln⁡ν)S(X)={ν1(HKDW)2−ν3(X+HKDW),X<−HKDW,ν1X2,∣X∣≤HKDW,ν1(HKDW)2+ν3(X−HKDW),X>HKDW.(74)(\ln\nu)_{S}(X)= \begin{cases} \nu_{1}(H_{K}^{\mathrm{DW}})^{2}-\nu_{3}\left(X+H_{K}^{\mathrm{DW}}\right), & X<-H_{K}^{\mathrm{DW}},\\ \nu_{1}X^{2}, & \lvert X\rvert\le H_{K}^{\mathrm{DW}},\\ \nu_{1}(H_{K}^{\mathrm{DW}})^{2}+\nu_{3}\left(X-H_{K}^{\mathrm{DW}}\right), & X>H_{K}^{\mathrm{DW}}. \end{cases} \tag*{(74)}

The antisymmetric part is

(ln⁡ν)A(X)={−ν2,X<−HKDW,ν2XHKDW,∣X∣≤HKDW,ν2,X>HKDW.(75)(\ln\nu)_{A}(X)= \begin{cases} -\nu_{2}, & X<-H_{K}^{\mathrm{DW}},\\ \nu_{2}\dfrac{X}{H_{K}^{\mathrm{DW}}}, & \lvert X\rvert\le H_{K}^{\mathrm{DW}},\\ \nu_{2}, & X>H_{K}^{\mathrm{DW}}. \end{cases} \tag*{(75)}

The three amplitudes appearing in Eqs. (74) and (75) are

ν1=1(HKDW)2[−βκS0+5CzκT0+η22].(76)\nu_{1}=\frac{1}{(H_{K}^{\mathrm{DW}})^{2}}\left[-\frac{\beta\kappa}{S_{0}}+\frac{5C_{z}\kappa}{T_{0}}+\frac{\eta^{2}}{2}\right]. \tag*{(76)}
ν2=sη−ζCzT0,(77)\nu_{2}=s\eta-\frac{\zeta C_{z}}{T_{0}}, \tag*{(77)}
ν3=2κHKDW[CzT0−βS0].(78)\nu_{3}=\frac{2\kappa}{H_{K}^{\mathrm{DW}}}\left[\frac{C_{z}}{T_{0}}-\frac{\beta}{S_{0}}\right]. \tag*{(78)}

Thus, ν1\nu_{1} has units of T−2\mathrm{T}^{-2}, ν2\nu_{2} is dimensionless, and ν3\nu_{3} has units of T−1\mathrm{T}^{-1}. Dimensionless versions of the amplitudes can be derived by multiplying the values obtained from the fit by the DW anisotropy field νˉ1=(HKDW)2ν1\bar{\nu}_{1}=(H_{K}^{\mathrm{DW}})^{2}\nu_{1}, νˉ2=ν2\bar{\nu}_{2}=\nu_{2}, νˉ3=HKDWν3\bar{\nu}_{3}=H_{K}^{\mathrm{DW}}\nu_{3}. A complete set of the dimensionless fitting amplitudes νˉ1…3\bar{\nu}_{1\ldots3} is reported in Table V. Equations (74) and (75) define the symmetric and antisymmetric shape functions about the compensation axis X=0X=0, with their vertical origin chosen such that both shape functions vanish at X=0X=0. No independent vertical-offset parameter is introduced in the fit.

Sampleμ0Hz\mu_0 H_zνˉ1\bar{\nu}_1νˉ2\bar{\nu}_2νˉ3\bar{\nu}_3μ0HK,fitDW\mu_0 H^{\mathrm{DW}}_{\mathrm{K,fit}}μ0HDMInew\mu_0 H^{\mathrm{new}}_{\mathrm{DMI}}μ0HDMIflow\mu_0 H^{\mathrm{flow}}_{\mathrm{DMI}}
(mT)(−)(−)(−)(−)(mT)(mT)(mT)
a1a112.61.0±0.21.0 \pm0.20.94±0.050.94 \pm0.050.78±0.080.78 \pm0.0815±115 \pm1−2±1-2 \pm1≈0\approx0
a2a211.27±27 \pm20.14±0.090.14 \pm0.094.1±0.54.1 \pm0.540±440 \pm44±24 \pm2≈0\approx0
a3a310.50.7±0.10.7 \pm0.10.69±0.030.69 \pm0.030.83±0.060.83 \pm0.0655±455 \pm4−62±2-62 \pm2−70±7-70 \pm7
a4a412.40.37±0.070.37 \pm0.070.36±0.030.36 \pm0.030.37±0.030.37 \pm0.0362±562 \pm5−69±5-69 \pm5−88±9-88 \pm9
a5a5351.3±0.21.3 \pm0.22.36±0.082.36 \pm0.080.71±0.040.71 \pm0.0470±370 \pm3−144±4-144 \pm4−160±16-160 \pm16
a6a6200.7±0.20.7 \pm0.20.22±0.080.22 \pm0.081.5±0.21.5 \pm0.2113±13113 \pm1388±788 \pm7n.m.

Table V. Fitting parameters obtained from the three-amplitude creep model.

The experimental and calculated velocity curves are instead normalized at Hx=0H_{x}=0. Defining the total symmetry-centered shape as

F(X)=(ln⁡ν)S(X)+(ln⁡ν)A(X),(79)F(X)=(\ln\nu)_{S}(X)+(\ln\nu)_{A}(X), \tag*{(79)}

the function compared with the normalized experimental velocity is

ln⁡[v(Hx,Hz)v(0,Hz)]model=F(Hx+HDMI)−F(HDMI).(80)\ln\left[\frac{v(H_x,H_z)}{v(0,H_z)}\right]_{\mathrm{model}} = F(H_x+H_{\mathrm{DMI}})-F(H_{\mathrm{DMI}}). \tag*{(80)}

Indeed, at the experimental normalization point Hx=0H_x=0, the shifted field is X=HDMIX=H_{\mathrm{DMI}}, and therefore

ln⁡[v(Hx,Hz)v(0,Hz)]model∣Hx=0=F(HDMI)−F(HDMI)=0.(81)\left.\ln\left[\frac{v(H_x,H_z)}{v(0,H_z)}\right]_{\mathrm{model}}\right|_{H_x=0}=F(H_{\mathrm{DMI}})-F(H_{\mathrm{DMI}})=0. \tag*{(81)}

The term F(HDMI)F(H_{\mathrm{DMI}}) is a single parameter-dependent constant evaluated at the normalization point. It is recalculated for each trial set of fitting parameters, but it does not constitute an additional fit parameter. In particular, it does not modify the amplitudes ν1\nu_1, ν2\nu_2, or ν3\nu_3: it only fixes the vertical origin of the complete model. Since a field-independent constant is symmetric under the transformation X→−XX\to-X, the strict symmetry decomposition of the normalized curve may equivalently be written as

(ln⁡vv(0)) ⁣S=(ln⁡v)S(X)−F(HDMI),(82)\left(\ln\frac{v}{v(0)}\right)_{\!S}=(\ln v)_{S}(X)-F(H_{\mathrm{DMI}}), \tag*{(82)}
(ln⁡vv(0)) ⁣A=(ln⁡v)A(X).(83)\left(\ln\frac{v}{v(0)}\right)_{\!A}=(\ln v)_{A}(X). \tag*{(83)}

Thus, the normalization affects only the constant level of the symmetric component. It does not alter the symmetric curvature ν1\nu_1, the high-field slope ν3\nu_3, the antisymmetric amplitude ν2\nu_2, or the position of the velocity minimum. In the mixed Bloch-Néel region, the field-dependent part of the velocity is

ln⁡[v(Hx,Hz)v(0,Hz)]=const.+ν1X2+ν2XHKDW,X=Hx+HDMI.(84)\ln\left[\frac{v(H_x,H_z)}{v(0,H_z)}\right]=\mathrm{const.}+\nu_1X^2+\nu_2\frac{X}{H_K^{\mathrm{DW}}},\qquad X=H_x+H_{\mathrm{DMI}}. \tag*{(84)}

The apparent minimum is therefore located at

(Hx+)creep=−HDMI−ν22ν1HKDW.(85)(H_x^+)_{\mathrm{creep}}=-H_{\mathrm{DMI}}-\frac{\nu_2}{2\nu_1H_K^{\mathrm{DW}}}. \tag*{(85)}

Equation (85) applies provided that the stationary point remains inside the mixed Bloch-Néel region and is a minimum:

∣ν22ν1HKDW∣≤HKDW,ν1>0.(86)\left|\frac{\nu_2}{2\nu_1H_K^{\mathrm{DW}}}\right|\le H_K^{\mathrm{DW}},\qquad\nu_1>0. \tag*{(86)}

Normalization regimes

When the normalization point Hx=0H_x=0 lies in either saturated Néel branch, ∣ζ∣>2κ|\zeta|>2\kappa, one has

S0=1+κ−∣ζ∣,T0=1−κ.(87)S_0=1+\kappa-|\zeta|,\qquad T_0=1-\kappa. \tag*{(87)}

For the positive saturated branch, ζ>2κ\zeta> 2\kappa, this reduces to S0=1−ζ+κS_0 = 1 - \zeta+ \kappa. The corresponding coefficients are

ν1(HKDW)2=−βκ1+κ−∣ζ∣+5Czκ1−κ+η22,(88)\nu_1\left(H_K^{\mathrm{DW}}\right)^2 = -\frac{\beta\kappa}{1+\kappa-\lvert\zeta\rvert}+\frac{5C_z\kappa}{1-\kappa}+\frac{\eta^2}{2}, \tag*{(88)}
ν2=sη−ζCz1−κ,(89)\nu_2 = s\eta-\frac{\zeta C_z}{1-\kappa}, \tag*{(89)}
ν3HKDW=2κ[Cz1−κ−β1+κ−∣ζ∣].(90)\nu_3H_K^{\mathrm{DW}} = 2\kappa\left[\frac{C_z}{1-\kappa}-\frac{\beta}{1+\kappa-\lvert\zeta\rvert}\right]. \tag*{(90)}

If instead the normalization point lies inside the mixed Bloch-Néel region, ∣ζ∣≤2κ\lvert\zeta\rvert\le2\kappa, then

S0=1−ζ24κ,T0=1+2κ−3ζ24κ.(91)S_0 = 1-\frac{\zeta^2}{4\kappa}, \qquad T_0 = 1+2\kappa-\frac{3\zeta^2}{4\kappa}. \tag*{(91)}

The coefficients become

ν1(HKDW)2=−βκ1−ζ24κ+5Czκ1+2κ−3ζ24κ+η22,(92)\nu_1\left(H_K^{\mathrm{DW}}\right)^2 = -\frac{\beta\kappa}{1-\frac{\zeta^2}{4\kappa}}+\frac{5C_z\kappa}{1+2\kappa-\frac{3\zeta^2}{4\kappa}}+\frac{\eta^2}{2}, \tag*{(92)}
ν2=sη−ζCz1+2κ−3ζ24κ,(93)\nu_2 = s\eta-\frac{\zeta C_z}{1+2\kappa-\frac{3\zeta^2}{4\kappa}}, \tag*{(93)}
ν3HKDW=2κ[Cz1+2κ−3ζ24κ−β1−ζ24κ].(94)\nu_3H_K^{\mathrm{DW}} = 2\kappa\left[\frac{C_z}{1+2\kappa-\frac{3\zeta^2}{4\kappa}}-\frac{\beta}{1-\frac{\zeta^2}{4\kappa}}\right]. \tag*{(94)}

Additional fits and full dataset

The complete three-amplitude fits for all investigated samples are shown in Figure 7, while the corresponding fitting parameters are summarized in Table V. The parameter HKDWH_K^{\mathrm{DW}} is treated as an effective crossover field delimiting the central mixed Bloch-Néel region, rather than as a direct measurement of the microscopic domain-wall anisotropy field. Its extraction is most reliable when the experimental range samples both crossover boundaries, Hx=−HDMI±HKDWH_x=-H_{\mathrm{DMI}}\pm H_K^{\mathrm{DW}}, together with sufficient data in the outer branches. When one or both boundaries lie outside the accessible field range, HKDWH_K^{\mathrm{DW}} becomes correlated with HDMIH_{\mathrm{DMI}} and the fitted amplitudes and is consequently more sensitive to the imposed bounds. In these cases, independently calculated magnetic parameters provide a useful physical constraint, but the resulting DMI field should be interpreted in light of the limited transition coverage.

Figure 7. Complete sample series fits with the three amplitude model of Eqs. (6-7) performed on the normalized velocity ln⁡(v(Hx)/v(0))\ln(v(H_x)/v(0)). Panels i)-v) show the fits for samples a1-a5 while panel vi) shows the fit of sample a6. The red line represent the full fitted curve, while the blue and orange dashed lines represent the symmetric and antisymmetric components respectively. The vertical dashed line represents the fitted μ0HDMInew\mu_0 H_{\mathrm{DMI}}^{\mathrm{new}} field reported in Table. Table V. The shaded region displays the mixed Bloch-Néel transition region.

A useful check for the model validity is provided by comparing the DMI values discrepancy and the predicted shift of Eq. (11) in the main text. As can be seen in Figure 8, there is a clear correlation between the discrepancy of the DMI values ∣ΔD∣|\Delta D| (see Table. Table VI) and the factor ν2/(2ν1HKDW)\nu_2/(2\nu_1 H_{\mathrm{K}}^{\mathrm{DW}}) which quantifies the displacement of the DMI field from the velocity curve minimum.

Figure 8. Comparison of the DMI discrepancy between BLS and conventional creep-regime measurements (left axis) with the fitted shift scale ν2/(2ν1HKDW)\nu_2/(2\nu_1 H^{\mathrm{DW}}_{\mathrm{K}}) (right axis). Here, HKDWH^{\mathrm{DW}}_{\mathrm{K}} denotes the crossover field between mixed Bloch–Néel and saturated Néel domain-wall configurations. Error bars represent propagated uncertainties; lines are guides to the eye.

Smoothing procedure and bootstrap uncertainty

The fitting procedure used to extract HDMIH_{\mathrm{DMI}} from the creep-regime velocity curves is based on the three-amplitude model derived above. In this model the logarithmic velocity is written as the sum of a symmetric and an antisymmetric contribution with respect to the shifted

TABLE V: Fitting parameters obtained from the three-amplitude creep model. We report the applied bubble expansion fields μ0Hz\mu_0 H_z, the dimensionless amplitudes νˉ1…3\bar{\nu}_{1\ldots3}, the fitted crossover field μ0HK,fitDW\mu_0 H^{\mathrm{DW}}_{\mathrm{K,fit}}, and the fitted DMI effective field μ0HDMInew\mu_0 H^{\mathrm{new}}_{\mathrm{DMI}}. The uncertainties in νˉ1\bar{\nu}_1 and νˉ3\bar{\nu}_3 include the propagated uncertainty in μ0HK,fitDW\mu_0 H^{\mathrm{DW}}_{\mathrm{K,fit}}, assuming the fitted quantities are statistically independent. We also include the effective DMI field measured in the flow regime μ0HDMIflow\mu_0 H^{\mathrm{flow}}_{\mathrm{DMI}} (see III B) for comparison.

field coordinate

X=Hx+HDMI.(95)X = H_x + H_{\mathrm{DMI}}. \tag*{(95)}

The ideal analytical form contains piecewise branches separated by the boundaries

X=±HKDW,(96)X = \pm H_K^{DW}, \tag*{(96)}

which delimit the mixed Bloch-Néel region. While this piecewise form is useful for identifying the physical origin of the different contributions, the sharp transition at the boundaries can make numerical error estimation unstable. In particular, small changes of the fitting parameters can move experimental points across the boundary between branches, producing a non-smooth local optimization landscape. To avoid this purely numerical instability, the antisymmetric contribution was regularized by replacing the sharp transition between the central linear branch and the saturated outer branches with a smooth interpolation. The unsmoothed form is

(ln⁡v)A={−ν2,X<−HKDW,ν2X/HKDW,∣X∣≤HKDW,ν2,X>HKDW.(97)(\ln v)_A = \begin{cases} -\nu_2, & X < -H_K^{DW}, \\ \nu_2 X/H_K^{DW}, & |X| \le H_K^{DW}, \\ \nu_2, & X > H_K^{DW}. \end{cases} \tag*{(97)}

In the numerical implementation, the three branches were blended using smooth switching functions of the form

S±(X)=12[1+tanh⁡(±X−HKDWw)],(98)S_{\pm}(X) = \frac{1}{2}\left[1+\tanh\left(\frac{\pm X-H_K^{DW}}{w}\right)\right], \tag*{(98)}

where ww is a fixed smoothing width. The smoothed antisymmetric contribution can then be written schematically as

(ln⁡v)Asmooth=S−(X)(−ν2)+[1−S−(X)−S+(X)]ν2XHKDW+S+(X)ν2.(99)(\ln v)_A^{\mathrm{smooth}} = S_-(X)(-\nu_2) + [1-S_-(X)-S_+(X)]\nu_2\frac{X}{H_K^{DW}} + S_+(X)\nu_2. \tag*{(99)}

The smoothing width was kept small compared with the field scale over which the velocity varies and was not treated as a free fitting parameter. Its role is only to regularize the branch boundaries for numerical minimization and uncertainty estimation.

Parameter uncertainties were estimated using a residual bootstrap. First, the model was fitted to the experimental data, yielding the best-fit parameter vector

θ^=(HDMI,ν1,ν2,ν3,HKDW).(100)\hat{\theta} = \left(H_{\mathrm{DMI}}, \nu_1, \nu_2, \nu_3, H_K^{DW}\right). \tag*{(100)}

The fitted values are

y^i=f(Hx,i;θ^),(101)\hat{y}_i = f(H_{x,i}; \hat{\theta}), \tag*{(101)}

where yi=ln⁡viy_i = \ln v_i, and the residuals are

ri=yi−y^i.(102)r_i = y_i - \hat{y}_i. \tag*{(102)}

A synthetic bootstrap dataset was generated by sampling the residuals with replacement,

yi∗=y^i+ri∗,(103)y_i^* = \hat{y}_i + r_i^*, \tag*{(103)}

where each ri∗r_i^* is randomly drawn from the original residual set. The model was then refitted to each synthetic dataset, producing a bootstrap parameter vector θ∗\theta^*. Repeating this procedure NbootN_{\mathrm{boot}} times gives an ensemble

{θ1∗,θ2∗,…,θNboot∗}.(104)\left\{\theta_1^*, \theta_2^*, \ldots, \theta_{N_{\mathrm{boot}}}^*\right\}. \tag*{(104)}

The statistical uncertainty of each fitted parameter was estimated from the spread of this ensemble. For the DMI field, we report

u(HDMI)=std⁡(HDMI,1∗,HDMI,2∗,…,HDMI,Nboot∗).(105)u(H_{\mathrm{DMI}}) = \operatorname{std}\left(H_{\mathrm{DMI},1}^*, H_{\mathrm{DMI},2}^*, \ldots, H_{\mathrm{DMI},N_{\mathrm{boot}}}^*\right). \tag*{(105)}

Since the bootstrap distribution was approximately symmetric, this standard deviation was used as the one-sigma statistical uncertainty. The quoted uncertainty should therefore be interpreted as the statistical fitting uncertainty within the adopted smoothed three-amplitude model.

Uncertainty estimation of D

The uncertainty in the DD value is then obtained through standard uncertainty propagation of the relevant physical quantities through the formula D=μ0MsΔHDMID = \mu_0 M_s \Delta H_{\mathrm{DMI}}, i.e.

u(D)=∣D∣[u(Ms)Ms]2+[u(Δ)Δ]2+[u(HDMI)HDMI]2.(106)u(D) = |D|\sqrt{\left[\frac{u(M_s)}{M_s}\right]^2 + \left[\frac{u(\Delta)}{\Delta}\right]^2 + \left[\frac{u(H_{\mathrm{DMI}})}{H_{\mathrm{DMI}}}\right]^2}. \tag*{(106)}

Given the rather complex ensemble of experimental data, we summarize here the source of all the relevant quantities for each sample. For a1 and a2, MsM_s, u(Ms)u(M_s) and Δ\Delta, u(Δ)u(\Delta) are taken from Ref.[1]. For a3, a4 and a5, MsM_s, u(Ms)u(M_s) is taken from Ref.[1], while Δ\Delta and u(Δ)u(\Delta) are taken from the comparison of flow measurements as discussed in Sec. III. For sample a6, MSM_{\mathrm{S}}, u(MS)u(M_{\mathrm{S}}) and KeffK_{\mathrm{eff}}, u(Keff)u(K_{\mathrm{eff}}) are measured while the exchange stiffness AA, u(A)u(A) is taken from a literature survey on similar stacks [3, 4]. The DW width Δ\Delta and its standard uncertainty u(Δ)u(\Delta) are obtained by propagating the uncertainty in AA and KeffK_{\mathrm{eff}} in Δ=A/Keff\Delta= \sqrt{A/K_{\mathrm{eff}}}.

u(Δ)=Δ2[u(A)A]2+[u(Keff)Keff]2(107)u(\Delta) = \frac{\Delta}{2}\sqrt{\left[\frac{u(A)}{A}\right]^2 + \left[\frac{u(K_{\mathrm{eff}})}{K_{\mathrm{eff}}}\right]^2} \tag*{(107)}

References

  1. [1]A. Magni, G. Carlotti, A. Casiraghi, E. Darwin, G. Durin, L. H. Diez, B. J. Hickey, A. Huxtable, C. Y. Hwang, G. Jakob, C. Kim, M. Klaui, J. Langer, C. H. Marrows, H. T. Nembach, D. Ravelosona, G. A. Riley, J. M. Shaw, V. Sokalski, S. Tacchi, and M. Kuepferling, Key Points in the Determination of the Interfacial Dzyaloshinskii–Moriya Interaction From Asymmetric Bubble Domain Expansion, IEEE Transactions on Magnetics 58, 1 (2022).DOI
  2. [2]T. Ha Pham, J. Vogel, J. Sampaio, M. Vañatka, J.-C. Rojas-Sánchez, M. Bonfim, D. S. Chaves, F. Choueikani, P. Ohresser, E. Otero, A. Thiaville, and S. Pizzini, Very large domain wall velocities in Pt/Co/GdOx and Pt/Co/Gd trilayers with Dzyaloshinskii-Moriya interaction, EPL (Europhysics Letters) 113, 67001 (2016).
  3. [3]B. Zimmermann, G. Bihlmayer, M. Böttcher, M. Bouhassoune, S. Lounis, J. Sinova, S. Heinze, S. Blügel, and B. Dupe, Comparison of first-principles methods to extract magnetic parameters in ultrathin films: Co/Pt(111), Physical Review B 99, 10.1103/PhysRevB.99.214426 (2019), arXiv: 1904.06954.
  4. [4]T. Devolder, J.-V. Kim, L. Nistor, R. Sousa, B. Rodmacq, and B. Diény, Exchange stiffness in ultrathin perpendicularly magnetized CoFeB layers determined using the spectroscopy of electrically excited spin waves, Journal of Applied Physics 120, 183902 (2016).DOI
  5. [5]D.-Y. Kim, I. Berrai, T. Suraj, Y. Roussigné, S. Yang, M. Belmeguenai, F. Hu, G. Shi, H. R. Tan, J. Huang, A. Soumyanarayanan, K.-W. Kim, S. Mourad Cherif, and H. Yang, Chiral Damping of Magnons, Physical Review Letters 133, 156704 (2024).arxiv.org/abs/2409.04713
  6. [6]P. J. Metaxas, J. P. Jamet, A. Mougin, M. Cormier, J. Ferré, V. Baltz, B. Rodmacq, B. Dieny, and R. L. Stamps, Creep and Flow Regimes of Magnetic Domain-Wall Motion in Ultrathin Pt/Co/Pt Films with Perpendicular Anisotropy, Physical Review Letters 99, 217208 (2007), publisher: American Physical Society.DOI
  7. [7]V. Jeudy, A. Mougin, S. Bustingorry, W. Savero Torres, J. Gorchon, A. B. Kolton, A. Lemaître, and J. P. Jamet, Universal Pinning Energy Barrier for Driven Domain Walls in Thin Ferromagnetic Films, Physical Review Letters 117, 1 (2016), arXiv: 1603.01674.DOI
  8. [8]J. P. Garcia, A. Fassatoui, M. Bonfim, J. Vogel, A. Thiaville, and S. Pizzini, Magnetic domain wall dynamics in the precessional regime: Influence of the Dzyaloshinskii-Moriya interaction, Physical Review B 104, 014405 (2021).DOI
  9. [9]S. Lemerle, J. Ferré, C. Chappert, V. Mathet, T. Giamarchi, and P. Le Doussal, Domain wall creep in an Ising ultrathin magnetic film, Physical Review Letters 80, 849 (1998).DOI
  10. [10]V. Krizakova, J. P. Garcia, J. Vogel, N. Rougemaille, D. D. S. Chaves, S. Pizzini, and A. Thiaville, Study of the velocity plateau of Dzyaloshinskii domain walls, Physical Review B 100, 214404 (2019).DOI
  11. [11]I. A. Yastremsky, O. M. Volkov, M. Kopte, T. Kosub, S. Stienen, K. Lenz, J. Lindner, J. Fassbender, B. A. Ivanov, and D. Makarov, Thermodynamics and Exchange Stiffness of Asymmetrically Sandwiched Ultrathin Ferromagnetic Films with Perpendicular Anisotropy, Physical Review Applied 12, 064038 (2019), publisher: American Physical Society.DOI
  12. [12]M. D. Kuz’min, K. P. Skokov, L. V. B. Diop, I. A. Radulov, and O. Gutfleisch, Exchange stiffness of ferromagnets, The European Physical Journal Plus 135, 301 (2020).DOI
  13. [13]H. J. Waring, Y. Li, N. A. B. Johansson, C. Moutafis, I. J. Vera-Marun, and T. Thomson, Exchange stiffness constant determination using multiple-mode FMR perpendicular standing spin waves, Journal of Applied Physics 133, 063901 (2023).
  14. [14]J. P. Pellegren, D. Lau, and V. Sokalski, Dispersive Stiffness of Dzyaloshinskii Domain Walls 10.1103/PhysRevLett.119.027203 (2017).DOI
  15. [15]D. Lau, J. P. Pellegren, H. T. Nembach, J. M. Shaw, and V. Sokalski, Disentangling factors governing Dzyaloshinskii domain-wall creep in Co/Ni thin films using PtₓIr₁₋ₓ seed layers, Physical Review B 98, 184410 (2018).DOI
  16. [16]C. A. Akosa, I. M. Miron, G. Gaudin, and A. Manchon, Phenomenology of chiral damping in noncentrosymmetric magnets, Physical Review B 93, 214429 (2016).DOI
  17. [17]C. A. Akosa, G. Tatara, A. Manchon, and M. Mochizuki, Decoupling the influences of chiral damping and Dzyaloshinskii-Moriya interaction in chiral magnetic domain walls, Physical Review B 110, 024420 (2024).DOI
  18. [18]S. G. Je, D. H. Kim, S. C. Yoo, B. C. Min, K. J. Lee, and S. B. Choe, Asymmetric magnetic domain-wall motion by the Dzyaloshinskii-Moriya interaction, Physical Review B - Condensed Matter and Materials Physics 88, 1 (2013).DOI

Paper details

vlnc:W_veag7e8z
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents