无序系统与神经网络
查看 最近的 文章
显示 2025年07月23日， 星期三 新的列表
- [1] arXiv:2507.16016 [中文pdf, pdf, 其他]
-
标题： 在a-InOx和颗粒铝中的电子玻璃动力学异常热激活标题： Anomalous thermal activation of the electron glass dynamics in a-InOx and granular aluminum评论： 将发表在《Journal of Physics Condensed Matter》上。当前版本是最初提交给期刊的版本。接受的手稿：https://doi.org/10.1088/1361-648X/adeef7主题： 无序系统与神经网络 (cond-mat.dis-nn)
在本文中，我们探讨了绝缘非晶氧化铟（a-InOx）和颗粒状铝薄膜中电玻璃动力学的温度依赖性。 我们使用非等温栅极电压协议，当对数弛豫没有特征时间时，可以揭示温度引起的动力学变化。 我们证明，与过去近20年的相反观点相反，液氦温度范围内非晶氧化铟薄膜的动力学是热激活的，即它在冷却时变慢，在加热时加速。 因此，非晶氧化铟加入了我们已经证明热激活的玻璃无序系统的列表，其中包括颗粒状铝和非晶NbxSi1-x薄膜。 此外，在a-InOx和颗粒状铝薄膜中进行的高达40 K的测量显示了这两个系统的密切相似性，以及热激活的非常异常特性，其有效活化能随T呈T^2增加。 到目前为止，我们还没有对此的解释。进一步的研究和理解可能对电子玻璃的物理具有重要意义。
In this article, we explore the temperature dependence of the electrical glassy dynamics in insulating amorphous indium oxide (a-InOx) and granular Al films. We use non-isothermal gate voltage protocols, which can reveal changes in the dynamics induced by the temperature, when logarithmic relaxations devoid of characteristic times are at work. We demonstrate that, contrary to almost 20 years of opposite belief, the dynamics of amorphous indium oxide films in the liquid helium temperature range is thermally activated, i.e. it slows down under cooling and accelerates upon heating. Amorphous indium oxide thus adds to the list of glassy disordered systems in which we already demonstrated thermal activation, which includes granular Al and amorphous NbxSi1-x films. Moreover, measurements up to 40 K in a-InOx and granular Al films reveal the close similarity between the two systems and a very anomalous character of the thermal activation, with an effective activation energy increasing with T as T^2 . We so far have no explanation for it. Its further study and understanding may be important for the physics of electron glasses.
- [2] arXiv:2507.16567 [中文pdf, pdf, html, 其他]
-
标题： 无序量子多体系统中非遍历行为的虚假特征标题： False signatures of non-ergodic behavior in disordered quantum many-body systems评论： 12页的精彩文字内容，包括参考文献，11幅图，欢迎提出评论主题： 无序系统与神经网络 (cond-mat.dis-nn) ; 统计力学 (cond-mat.stat-mech) ; 量子物理 (quant-ph)
Ergodic isolated quantum many-body systems satisfy the eigenstate thermalization hypothesis (ETH), i.e., the expectation values of local observables in the system's eigenstates approach the predictions of the microcanonical ensemble. However, the ETH does not specify what happens to expectation values of local observables within an energy window when the average over disorder realizations is taken. As a result, the expectation values of local observables can be distributed over a relatively wide interval and may exhibit nontrivial structure, as shown in [Phys. Rev. B \textbf{104}, 214201 (2021)] for a quasiperiodic disordered system for site-resolved magnetization. We argue that the non-Gaussian form of this distribution may \textit{错误地} suggest non-ergodicity and a breakdown of ETH. By considering various types of disorder, we find that the functional forms of the distributions of matrix elements of the site-resolved magnetization operator mirror the distribution of the onsite disorder. We argue that this distribution is a direct consequence of the local observable having a finite overlap with moments of the Hamiltonian. We then demonstrate how to adjust the energy window when analyzing expectation values of local observables in disordered quantum many-body systems to correctly assess the system's adherence to ETH, and provide a link between the distribution of expectation values in eigenstates and the outcomes of quench experiments.
Ergodic isolated quantum many-body systems satisfy the eigenstate thermalization hypothesis (ETH), i.e., the expectation values of local observables in the system's eigenstates approach the predictions of the microcanonical ensemble. However, the ETH does not specify what happens to expectation values of local observables within an energy window when the average over disorder realizations is taken. As a result, the expectation values of local observables can be distributed over a relatively wide interval and may exhibit nontrivial structure, as shown in [Phys. Rev. B \textbf{104}, 214201 (2021)] for a quasiperiodic disordered system for site-resolved magnetization. We argue that the non-Gaussian form of this distribution may \textit{falsely} suggest non-ergodicity and a breakdown of ETH. By considering various types of disorder, we find that the functional forms of the distributions of matrix elements of the site-resolved magnetization operator mirror the distribution of the onsite disorder. We argue that this distribution is a direct consequence of the local observable having a finite overlap with moments of the Hamiltonian. We then demonstrate how to adjust the energy window when analyzing expectation values of local observables in disordered quantum many-body systems to correctly assess the system's adherence to ETH, and provide a link between the distribution of expectation values in eigenstates and the outcomes of quench experiments.
- [3] arXiv:2507.16654 [中文pdf, pdf, html, 其他]
-
标题： 构建动力学平均场理论的直觉：一个简单模型和腔体方法标题： Building Intuition for Dynamical Mean-Field Theory: A Simple Model and the Cavity Method评论： 33页，2张图，5张边距图，未发表的教程主题： 无序系统与神经网络 (cond-mat.dis-nn) ; 生物物理 (physics.bio-ph)
动态平均场理论（DMFT）是一种用于分析具有许多相互作用自由度系统的强大理论框架。 本教程提供了一个易于理解的DMFT介绍。 我们从一个线性模型开始，其中DMFT方程可以精确推导，使读者能够建立对基本原理的清晰直觉。 然后我们引入腔方法，这是一种用于推导非线性系统DMFT方程的通用方法。 本教程最后以广义Lotka-Volterra相互作用物种模型的应用结束，展示了如何将多物种群落的复杂动力学简化为可处理的单物种随机过程。 关键见解包括理解淬火无序如何使多体动力学减少到有效单粒子动力学，认识到自平均在简化复杂系统中的作用，以及看到集体相互作用如何产生非马尔可夫反馈效应。
Dynamical Mean-Field Theory (DMFT) is a powerful theoretical framework for analyzing systems with many interacting degrees of freedom. This tutorial provides an accessible introduction to DMFT. We begin with a linear model where the DMFT equations can be derived exactly, allowing readers to develop clear intuition for the underlying principles. We then introduce the cavity method, a versatile approach for deriving DMFT equations for non-linear systems. The tutorial concludes with an application to the generalized Lotka--Volterra model of interacting species, demonstrating how DMFT reduces the complex dynamics of many-species communities to a tractable single-species stochastic process. Key insights include understanding how quenched disorder enables the reduction from many-body to effective single-particle dynamics, recognizing the role of self-averaging in simplifying complex systems, and seeing how collective interactions give rise to non-Markovian feedback effects.
新提交 (展示 3 之 3 条目 )
- [4] arXiv:2507.16042 (交叉列表自 cond-mat.quant-gas) [中文pdf, pdf, html, 其他]
-
标题： 量子自旋晶格中分散掺杂动力学之外：稳健局域化和亚扩散标题： Beyond fragmented dopant dynamics in quantum spin lattices: Robust localization and sub-diffusion评论： 19页，12图主题： 量子气体 (cond-mat.quant-gas) ; 无序系统与神经网络 (cond-mat.dis-nn) ; 强关联电子 (cond-mat.str-el) ; 超导性 (cond-mat.supr-con)
掺杂剂在磁自旋晶格中的运动在过去至少四十年中受到了极大的关注，因为它与高温超导性有关。 尽管有这些努力，我们仍然缺乏对其行为的完整理解，尤其是在非平衡状态和非零温度下。 在本文中，我们基于最先进的矩阵乘积态计算，朝着更深入的理解迈出了重要的一步。 特别是，我们研究了两腿$t$--$J$梯子中掺杂剂的非平衡动力学，其中具有反铁磁性的XXZ自旋相互作用。 在伊辛极限下，由于出现的无序势，所有研究的\emph{非零}温度下掺杂剂都是\emph{局部的}，其局域化长度由自旋晶格的基本关联长度控制，因此它仅在零温度极限下渐近地离域化。 这极大地推广了最近在希尔伯特空间碎片化模型中发现的局域化效应。 在存在自旋交换过程的情况下，掺杂剂根据幂律行为离域化，在弱自旋交换下表现出强烈的亚扩散，但当交换足够强时最终变为扩散。 此外，我们表明，在无限温度下的自旋动力学行为定性相同，尽管存在重要的定量差异。 我们通过显示动力学表现出自相似的标度行为来证实这些发现，这种行为与常规扩散的高斯行为强烈偏离。 最后，我们展示了在高温下扩散系数遵循阿伦尼乌斯关系，随着温度的降低，它呈指数下降。
The motion of dopants in magnetic spin lattices has received tremendous attention for at least four decades due to its connection to high-temperature superconductivity. Despite these efforts, we lack a complete understanding of their behavior, especially out-of-equilibrium and at nonzero temperatures. In this Article, we take a significant step towards a much deeper understanding based on state-of-the-art matrix-product-state calculations. In particular, we investigate the non-equilibrium dynamics of a dopant in two-leg $t$--$J$ ladders with antiferromagnetic XXZ spin interactions. In the Ising limit, we find that the dopant is \emph{localized} for all investigated \emph{nonzero} temperatures due to an emergent disordered potential, with a localization length controlled by the underlying correlation length of the spin lattice, whereby it only delocalizes asymptotically in the zero temperature limit. This greatly generalizes the localization effect discovered recently in Hilbert space fragmented models. In the presence of spin-exchange processes, the dopant delocalizes according to a power-law behavior, which is strongly sub-diffusive for weak spin-exchange but which eventually becomes diffusive for strong enough exchange. Moreover, we show that the underlying spin dynamics at infinite temperature behaves qualitatively the same, albeit with important quantitative differences. We substantiate these findings by showing that the dynamics shows self-similar scaling behavior, which strongly deviates from the Gaussian behavior of regular diffusion. Finally, we show that the diffusion coefficient follows an Arrhenius relation at high temperatures, whereby it exponentially decreases for decreasing temperatures.
- [5] arXiv:2507.16336 (交叉列表自 cond-mat.mtrl-sci) [中文pdf, pdf, html, 其他]
-
标题： 构建智能非晶合金设计的材料网络表示标题： Constructing material network representations for intelligent amorphous alloys design评论： 5个数字主题： 材料科学 (cond-mat.mtrl-sci) ; 无序系统与神经网络 (cond-mat.dis-nn) ; 计算复杂性 (cs.CC) ; 机器学习 (cs.LG)
设计高性能的非晶合金在各种应用中具有挑战性。 但这一过程高度依赖经验法则和无限制的尝试。 传统策略的高成本和低效率特性阻碍了在庞大的材料空间中进行有效的采样。 在此，我们提出材料网络以加速二元和三元非晶合金的发现。 网络拓扑结构揭示了被传统表格数据表示所掩盖的隐藏材料候选。 通过仔细分析不同年份合成的非晶合金，我们构建了动态材料网络来追踪合金发现的历史。 我们发现过去设计的一些创新材料被编码在网络中，证明了它们在指导新合金设计方面的预测能力。 这些材料网络与我们日常生活中的一些现实世界网络表现出物理相似性。 我们的研究为智能材料设计开辟了一条新途径，特别是对于复杂合金而言。
Designing high-performance amorphous alloys is demanding for various applications. But this process intensively relies on empirical laws and unlimited attempts. The high-cost and low-efficiency nature of the traditional strategies prevents effective sampling in the enormous material space. Here, we propose material networks to accelerate the discovery of binary and ternary amorphous alloys. The network topologies reveal hidden material candidates that were obscured by traditional tabular data representations. By scrutinizing the amorphous alloys synthesized in different years, we construct dynamical material networks to track the history of the alloy discovery. We find that some innovative materials designed in the past were encoded in the networks, demonstrating their predictive power in guiding new alloy design. These material networks show physical similarities with several real-world networks in our daily lives. Our findings pave a new way for intelligent materials design, especially for complex alloys.
- [6] arXiv:2507.16607 (交叉列表自 physics.optics) [中文pdf, pdf, 其他]
-
标题： 粗糙的法布里-珀罗腔：一个高度多尺度的数值问题标题： Rough Fabry-Perot cavity: a vastly multi-scale numerical problemTetiana Slipchenko, Jaime Abad-arredondo, Antonio Consoli, Francisco J García Vidal, Antonio I Fernández-domínguez, Pedro David García, Cefe López主题： 光学 (physics.optics) ; 无序系统与神经网络 (cond-mat.dis-nn)
一种商用法布里-珀罗激光二极管具有高度不成比例的尺寸，这即使对于最先进的工具来说也是一个重大的数值挑战。 当腔体镜面之一变得粗糙时，这种挑战会更加严重，这在制造随机激光二极管时是常见的情况。 这样的系统涉及从几百微米（长度）到几纳米（粗糙度）的尺度，所有这些在研究可见光范围内的光学特性时都是相关的。 尽管涉及极端的尺寸范围，但这些腔体不能通过统计方法进行处理，例如那些用于具有已知良好研究特性的自相似分形结构的方法。 在这里，我们采用数值方法来计算腔体模式，并展示法布里-珀罗腔壁的随机起伏如何影响其光谱特征的统计特性。 我们的研究是开发用于光子计算和高效无斑点照明的技术关键设备的必要第一步。
A commercial Fabry-Perot laser diode is characterized by highly disproportionate dimensions, which poses a significant numerical challenge, even for state-of-the-art tools. This challenge is exacerbated when one of the cavity mirrors is rough-ened, as is the case when fabricating random laser diodes. Such a system involves length scales from several hundred mi-crometres (length) to a few nanometres (roughness) all of which are relevant when studying optical properties in the visi-ble. While involving an extreme range of dimensions, these cavities cannot be treated through statistical approaches such as those used with self-similar fractal structures known to show well-studied properties. Here we deploy numerical meth-ods to compute cavity modes and show how random corrugations of the Fabry-Perot cavity wall affect statistical proper-ties of their spectral features. Our study constitutes a necessary first step in developing technologically essential devices for photonic computation and efficient speckle-free illumination.
交叉提交 (展示 3 之 3 条目 )
- [7] arXiv:2502.10918 (替换) [中文pdf, pdf, html, 其他]
-
标题： 玻璃动力学在动态模式分解中的特征标题： Signature of glassy dynamics in dynamic modes decompositions评论： 4页，4图，加上附录主题： 无序系统与神经网络 (cond-mat.dis-nn) ; 软凝聚态物理 (cond-mat.soft) ; 统计力学 (cond-mat.stat-mech) ; 模式形成与孤子 (nlin.PS)
玻璃传统上以其无序的低能态的崎岖景观以及其缓慢趋向热力学平衡的特性来描述。 远离平衡时，也观察到了具有异常代数弛豫的动力学玻璃行为形式，例如在耦合振子网络中。 由于这些系统的无序性和高维特性，它们在理论上一直难以研究，但数据驱动的方法正在成为一种有前景的替代方法，可能有助于它们的分析。 在此，我们使用动态模态分解来表征玻璃动力学，这是一种数据驱动的谱计算方法，用于近似Koopman谱。 我们展示了在表现出代数弛豫的系统中，Koopman谱中的振荡模式和衰减模式之间的间隙消失，因此，我们提出了一种与模型无关的特征，用于稳健地检测和分析玻璃动力学。 我们通过一个一维常微分方程的最小示例和一个耦合振子的高维示例来展示我们方法的实用性。
Glasses are traditionally characterized by their rugged landscape of disordered low-energy states and their slow relaxation towards thermodynamic equilibrium. Far from equilibrium, dynamical forms of glassy behavior with anomalous algebraic relaxation have also been noted, for example, in networks of coupled oscillators. Due to their disordered and high-dimensional nature, such systems have been difficult to study theoretically, but data-driven methods are emerging as a promising alternative that may aid in their analysis. Here, we characterize glassy dynamics using the dynamic mode decomposition, a data-driven spectral computation that approximates the Koopman spectrum. We show that the gap between oscillatory and decaying modes in the Koopman spectrum vanishes in systems exhibiting algebraic relaxation, and thus, we propose a model-agnostic signature for robustly detecting and analyzing glassy dynamics. We demonstrate the utility of our approach through both a minimal example of a one-dimensional ODE and a high-dimensional example of coupled oscillators.
- [8] arXiv:2504.10310 (替换) [中文pdf, pdf, html, 其他]
-
标题： 非平衡玻璃在简并隐秘超均匀基态流形中的存在标题： Existence of Nonequilibrium Glasses in the Degenerate Stealthy Hyperuniform Ground-State Manifold评论： 10页，7图期刊参考： 软物质，21，4898 (2025)主题： 无序系统与神经网络 (cond-mat.dis-nn) ; 统计力学 (cond-mat.stat-mech) ; 计算物理 (physics.comp-ph)
隐秘相互作用是一种新兴的非平凡、有界长程振荡对势类，其经典基态可以是无序的、超均匀的和无限简并的。 它们的晶体-液体混合特性赋予了它们优于其晶体对应物的新物理性质。 在这里，我们展示了在隐秘参数$\chi$趋近于零时，这种不寻常的基态流形中存在非平衡硬球玻璃，这些玻璃在构型上非常接近超均匀的三维最大随机紧密堆积(MRJ)球体填充。 后者是典型的玻璃，因为它们是最高度无序的、完全刚性的和完全不可遍历的。 我们的优化过程利用了无限基态集的最大基数，不仅保证了我们的填充具有与MRJ状态相同的结构因子标度指数的超均匀性，而且它们还具有其他显著的结构特征，包括填充分数为$0.638$，每个粒子的平均接触数为6，间隙指数为$0.44(1)$，以及对于所有$r$和$k$分别几乎彼此相同的对相关函数$g_2(r)$和结构因子$S(k)$。 此外，我们证明在无序区域（$0 < \chi <1/2$）内可以创建隐身超均匀包装，其最大包装分数达到了前所未有的水平。 当$\chi$从零开始增加时，它们总是形成粒子间接触，尽管随着$\chi$从零开始增加，接触网络变得更加稀疏，从而导致接触粒子的线性聚合物状链，且链长逐渐变短。 对于所有$\chi$，生成超密集隐身超均匀包装的能力为光学和声学领域开辟了新的材料应用。
Stealthy interactions are an emerging class of nontrivial, bounded long-ranged oscillatory pair potentials with classical ground states that can be disordered, hyperuniform, and infinitely degenerate. Their hybrid crystal-liquid nature endows them with novel physical properties with advantages over their crystalline counterparts. Here, we show the existence of nonequilibrium hard-sphere glasses within this unusual ground-state manifold as the stealthiness parameter $\chi$ tends to zero that are remarkably configurationally extremely close to hyperuniform 3D maximally random jammed (MRJ) sphere packings. The latter are prototypical glasses since they are maximally disordered, perfectly rigid, and perfectly nonergodic. Our optimization procedure, which leverages the maximum cardinality of the infinite ground-state set, not only guarantees that our packings are hyperuniform with the same structure-factor scaling exponent as the MRJ state, but they share other salient structural attributes, including a packing fraction of $0.638$, a mean contact number per particle of 6, gap exponent of $0.44(1)$, and pair correlation functions $g_2(r)$ and structures factors $S(k)$ that are virtually identical to one another for all $r$ and $k$, respectively. Moreover, we demonstrate that stealthy hyperuniform packings can be created within the disordered regime ($0 < \chi <1/2$) with heretofore unattained maximal packing fractions. As $\chi$ increases from zero, they always form interparticle contacts, albeit with sparser contact networks as $\chi$ increases from zero, resulting in linear polymer-like chains of contacting particles with increasingly shorter chain lengths. The capacity to generate ultradense stealthy hyperuniform packings for all $\chi$ opens up new materials applications in optics and acoustics.
- [9] arXiv:2505.16020 (替换) [中文pdf, pdf, html, 其他]
-
标题： 单点模型中的霍尔斯坦机制与单位演化标题： Holstein mechanism in single-site model with unitary evolution主题： 无序系统与神经网络 (cond-mat.dis-nn) ; 量子物理 (quant-ph)
我们研究单电子（单点）系统中的Holstein机制，在非绝热条件下，单位演化本质上涉及费米子和玻色子算符。 产生的单位动力学和玻色子频率依赖性揭示了一种量子相变，这由短时间（幂律衰减）和长时间（指数衰减）行为的不同表现所证实，这体现在极化子位移、玻色能量和约化密度矩阵的动力学中。 这一观察结果与非马尔可夫到马尔可夫的转变一致。
We investigate the Holstein mechanism in a single-electron (one-site) system, where unitary evolution intrinsically involves both fermion and boson operators under nonadiabatic conditions. The resulting unitary dynamics and boson-frequency dependence reveal a quantum phase transition, evidenced by distinct short-time (power-law decay) and long-time (exponential decay) behaviors, which are manifested in the polaronic shift, bosonic energy, and dynamics of reduced density matrix. This observation is consistent with a non-Markovian to Markovian transition.
- [10] arXiv:2408.04607 (替换) [中文pdf, pdf, html, 其他]
-
标题： 岭回归中的风险与相关样本的交叉验证标题： Risk and cross validation in ridge regression with correlated samples评论： 44页，19图。v4：ICML 2025最终稿。v5：修正定理5陈述中的拼写错误。主题： 机器学习 (stat.ML) ; 无序系统与神经网络 (cond-mat.dis-nn) ; 机器学习 (cs.LG)
近年来，我们在高维岭回归的理解方面取得了显著进展，但现有的理论假设训练样本是独立的。 通过利用随机矩阵理论和自由概率的技术，我们提供了当数据点具有任意相关性时，岭回归的内部和外部样本风险的精确渐近分析。 我们证明在这一设定下，广义交叉验证估计量（GCV）无法正确预测外部样本风险。 然而，在噪声残差与数据点具有相同相关性的条件下，可以对GCV进行修改，得到一个高效计算的无偏估计量，在高维极限下集中，我们将其称为CorrGCV。 我们进一步将渐近分析扩展到测试点与训练集具有非平凡相关性的场景，这种场景在时间序列预测中经常遇到。 假设已知时间序列的相关结构，这再次产生了一个GCV估计量的扩展，并精确描述了此类测试点在多大程度上会导致长期风险的过于乐观的预测。 我们在各种高维数据中验证了我们理论的预测。
Recent years have seen substantial advances in our understanding of high-dimensional ridge regression, but existing theories assume that training examples are independent. By leveraging techniques from random matrix theory and free probability, we provide sharp asymptotics for the in- and out-of-sample risks of ridge regression when the data points have arbitrary correlations. We demonstrate that in this setting, the generalized cross validation estimator (GCV) fails to correctly predict the out-of-sample risk. However, in the case where the noise residuals have the same correlations as the data points, one can modify the GCV to yield an efficiently-computable unbiased estimator that concentrates in the high-dimensional limit, which we dub CorrGCV. We further extend our asymptotic analysis to the case where the test point has nontrivial correlations with the training set, a setting often encountered in time series forecasting. Assuming knowledge of the correlation structure of the time series, this again yields an extension of the GCV estimator, and sharply characterizes the degree to which such test points yield an overly optimistic prediction of long-time risk. We validate the predictions of our theory across a variety of high dimensional data.
- [11] arXiv:2501.00589 (替换) [中文pdf, pdf, html, 其他]
-
标题： 金属玻璃的机器学习势函数的高效训练：CuZrAl验证标题： Efficient training of machine learning potentials for metallic glasses: CuZrAl validationAntoni Wadowski, Anshul D.S. Parmar, Filip Kaśkosz, Jesper Byggmästar, Jan S. Wróbel, Mikko J. Alava, Silvia Bonfanti评论： 10页，6图，补充信息主题： 材料科学 (cond-mat.mtrl-sci) ; 无序系统与神经网络 (cond-mat.dis-nn)
原子间势能是揭示微观结构-性能关系的关键，对于多尺度模拟和高通量实验至关重要。 对于金属玻璃而言，其无序的原子结构使得势能的发展尤其具有挑战性，导致这一重要材料类别的化学特异性参数化方法稀缺。 我们通过引入一种高效的方法来设计机器学习原子间势能（MLIPs），并在CuZrAl系统上进行了基准测试。 使用Lennard-Jones近似模型、交换蒙特卡罗采样和单点密度泛函理论（DFT）修正，我们捕捉到了跨越14个数量级过冷度的非晶结构。 这些具有代表性的构型与实验时间尺度竞争，能够在多种状态下实现稳健的模型训练，同时减少对大量DFT数据集的需求。 所产生的MLIP在结构、动力学、能量和力学性质方面与实验数据以及经典嵌入原子方法（EAM）的预测相匹配。 这种方法为开发复杂金属玻璃的准确MLIP提供了一条可扩展的路径，包括新兴的多组分和高熵系统。
Interatomic potentials are key to uncovering microscopic structure-property relationships, essential for multiscale simulations and high-throughput experiments. For metallic glasses, their disordered atomic structure makes the development of potentials particularly challenging, resulting in the scarcity of chemistry-specific parametrizations for this important class of materials. We address this gap by introducing an efficient methodology to design machine learning interatomic potentials (MLIPs), benchmarked on the CuZrAl system. Using a Lennard-Jones surrogate model, swap-Monte Carlo sampling, and single-point Density Functional Theory (DFT) corrections, we capture amorphous structures spanning 14 decades of supercooling. These representative configurations, competing with the experimental time scale, enable robust model training across diverse states, while minimizing the need for extensive DFT datasets. The resulting MLIP matches the experimental data and predictions of the classical embedded atom method (EAM) for structural, dynamical, energetic, and mechanical properties. This approach offers a scalable path to develop accurate MLIPs for complex metallic glasses, including emerging multi-component and high-entropy systems.