0027 - NEUTRON: 神经粒子群优化实现材料感知的结构色逆设计

一、文献核心概览 (Literature Core Overview)

1.1 基本信息 (Basic Information)

1.2 核心结论 (Core Conclusions)

1. 方法创新: 提出NEUTRON方法，结合材料感知多任务混合密度网络(M3DN)与粒子群优化(PSO)，实现材料选择和厚度优化的协同设计。

2. 分阶段策略: 将材料和厚度设计分为两个阶段，先通过M3DN预测最适合的材料组合和初始厚度分布，再通过PSO精细优化，将搜索空间从6.6×10¹¹降低到约10⁴量级。

3. 多任务学习: 材料分类任务与厚度预测任务共享表示，产生正向知识迁移，提高逆设计精度。

4. 数据增强: 通过混合两种材料形成合成复合材料，增加折射率数据变化，显著提升模型泛化能力。

5. 实际应用: 成功应用于两个实际任务——环保型镀铬替代涂层设计和多图层光学薄膜图像重建。

1.3 核心价值 (Core Value)

1.4 研究方法 (Research Methods)

NEUTRON核心流程（4步）:

1. 数据收集: 使用PSO生成40万数据点，覆盖sRGB色域
2. M3DN训练: 训练材料感知多任务混合密度网络
3. 初始设计生成: M3DN输出多组候选材料和厚度分布
4. PSO精调: 粒子群优化精细调整厚度设计

网络架构:

• 输入: 五层折射率数据(62维向量) + 目标颜色(Lab坐标)
• 任务1: 材料可行性分类(e=0/1)
• 任务2: 厚度分布预测(混合高斯模型，51,200个组件)
• 输出: 可行材料组合 + 厚度概率分布

二、全文双语对照 (Full Bilingual Text)

Abstract 摘要

English: Designing optical structures for generating structural colors is challenging because of the complex relationship between the optical structures and the color perceived by human eyes. Machine learning-based approaches have been developed to expedite this design process. However, existing methods solely focus on structural parameters of the optical design, which could lead to suboptimal color generation because of the inability to optimize the selection of materials. To address this issue, an approach known as Neural Particle Swarm Optimization is proposed in this paper. The proposed method achieves high design accuracy and efficiency on two structural color design tasks; the first task is designing environment-friendly alternatives to chrome coatings, and the second task concerns reconstructing pictures with multilayer optical thin films. Several designs that could replace chrome coatings have been discovered; pictures with more than 200,000 pixels and thousands of unique colors can be accurately reconstructed in a few hours.

中文: 为产生结构色而设计光学结构具有挑战性，因为光学结构与人类眼睛感知的颜色之间存在复杂关系。基于机器学习的方法已被开发用于加速这一设计过程。然而，现有方法仅关注光学设计的结构参数，这可能导致次优的颜色生成，因为无法优化材料的选择。为解决这一问题，本文提出了一种称为神经粒子群优化(NEUTRON)的方法。该方法在两个结构色设计任务上实现了高精度和高效率；第一个任务是设计环保的镀铬涂层替代方案，第二个任务是使用多层光学薄膜重建图片。已发现多种可替代镀铬涂层的设计；包含超过20万像素和数千种独特色彩的图片可在数小时内准确重建。

Introduction 引言

Paragraph 1:

Structural color refers to the color generated through the light interaction with patterned or layered optical structures.

结构色是指光与图案化或层状光学结构相互作用产生的颜色。

It is more stable than colors produced from chemical pigments and serves as an environment-friendly alternative.

它比化学颜料产生的颜色更稳定，是环保的替代方案。

However, designing the structures for producing desired colors is challenging because of the complex relationship between the optical structures and their spectral properties.

然而，设计产生期望颜色的结构具有挑战性，因为光学结构与其光谱特性之间存在复杂关系。

In addition, color metamerism, i.e., different spectra may correspond to the same color perceived by human eyes, makes the relationship between the structures and the perceived color more complex because multiple different structures could have the same color appearance.

此外，颜色同色异谱现象，即不同的光谱可能对应于人眼感知的相同颜色，使得结构与感知颜色之间的关系更加复杂，因为多种不同的结构可能具有相同的外观颜色。

Human experts often design optical structures based on the understanding of the physical properties of structures, including multilayer thin films, metasurfaces, and self-assembled colloidal particles, to name a few.

人类专家通常基于对结构物理性质的理解来设计光学结构，包括多层薄膜、超表面和自组装胶体颗粒等。

Owing to the complex relationship between the structures and the generated color, the manual design process is often slow and could lead to suboptimal color production.

由于结构与产生的颜色之间的复杂关系，手动设计过程通常缓慢，可能导致次优的颜色生成。

Paragraph 2:

Recently, machine learning-based optical inverse design approaches have been developed to predict optical structures that can achieve user-specified color properties.

最近，基于机器学习的光学逆设计方法已被开发用于预测可实现用户指定颜色特性的光学结构。

These inverse design methods often involve training a machine learning model such as deep neural networks or support vector machines on a curated dataset that contains a large number of data points mapping structural parameters to the corresponding color, e.g., represented by coordinates in CIE xyY or LAB color space.

这些逆设计方法通常涉及在精心整理的数据集上训练机器学习模型（如深度神经网络或支持向量机），该数据集包含大量将结构参数映射到相应颜色的数据点，例如用CIE xyY或LAB颜色空间中的坐标表示。

Though previous methods have been demonstrated to be efficient in designing a wide range of colors, they often require materials constituting the optical structures to be fixed.

虽然先前方法已被证明在设计广泛颜色范围方面高效，但它们通常要求构成光学结构的材料是固定的。

Because the refractive index of materials affects their reflection and absorption properties, it could be challenging or even impossible to produce specific colors when the materials are not appropriately selected.

由于材料的折射率影响其反射和吸收特性，当材料未适当选择时，产生特定颜色可能具有挑战性甚至不可能。

Thus, the first step of screening appropriate materials for latter inverse design with machine learning models still requires extensive effort from human experts and is difficult for people without sufficient prior experience, which should be adequately addressed for the wide adoption of the developed machine learning models.

因此，为后续机器学习模型逆设计筛选适当材料的第一步仍然需要人类专家的广泛努力，对于缺乏足够先前经验的人来说很困难，这应该得到充分解决以实现所开发机器学习模型的广泛采用。

Paragraph 3:

For optical multilayer thin-film design, the recently reported reinforcement learning approach addresses the material selection challenge by searching the material and thickness design space simultaneously.

对于光学多层薄膜设计，最近报道的强化学习方法通过同时搜索材料和厚度设计空间来解决材料选择挑战。

However, this method only designs a single color at a time because the reward function for training the reinforcement learning algorithm has to be defined for a specific color.

然而，这种方法一次只能设计一种颜色，因为训练强化学习算法的奖励函数必须为特定颜色定义。

It could be impractical when many colors need to be designed, e.g., reproducing all the colors found in a paint shop or designing a large array of reflective color pixels to reconstruct a colored picture, which would take an extremely long time using the reinforcement learning method reported in our previous work.

当需要设计许多颜色时，例如重现油漆店中找到的所有颜色或设计大阵列反射彩色像素以重建彩色图片，这可能不切实际，使用我们先前工作中报道的强化学习方法将需要极长的时间。

To address these issues, we propose an inverse structural color design method that can efficiently predict both the materials and structural parameters in a synergistic manner.

为解决这些问题，我们提出了一种逆结构色设计方法，可以以协同方式高效预测材料和结构参数。

Paragraph 4:

The proposed method termed Neural ParTicle SwaRm OptimizatioN (NEUTRON) is a hybrid approach that combines a Material-aware Multitask Mixture Density Network (M3DN) and Particle Swarm Optimization (PSO).

所提出的方法称为神经粒子群优化(NEUTRON)，是一种混合方法，结合了材料感知多任务混合密度网络(M3DN)和粒子群优化(PSO)。

Instead of searching the material and thickness space simultaneously, NEUTRON first predicts the most suitable materials and provides a diverse set of initial guesses of the thicknesses in the form of probability distributions that could fulfill the target color and then applies particle swarm optimization to fine-tune the initial thickness designs.

NEUTRON不是同时搜索材料和厚度空间，而是首先预测最合适的材料，并以概率分布形式提供可满足目标颜色的厚度初始猜测的多样化集合，然后应用粒子群优化来精细调整初始厚度设计。

We demonstrate the effectiveness of the proposed approach on two optical multilayer thin film design tasks.

我们在两个光学多层薄膜设计任务上展示了所提出方法的有效性。

The results show that our approach can lead to accurate color inverse designs efficiently.

结果表明，我们的方法可以高效地实现精确的颜色逆设计。

We believe that the proposed approach can be readily extended to many other optical design tasks where material selection and structural designs are both important.

我们相信，所提出的方法可以 readily 扩展到许多其他材料选择和结构设计都很重要的光学设计任务。

Background 背景

Paragraph 5:

Both numerical optimization and machine learning have been applied to the task of structural color design.

数值优化和机器学习都已应用于结构色设计任务。

Compared to machine learning, optimization-based methods for structural color design are often more accurate but suffer from low computational efficiency.

与机器学习相比，基于优化的结构色设计方法通常更准确，但计算效率低。

Various optimization methods including PSO and genetic algorithms (GAs) have been applied to design optical structures for structural color applications.

包括PSO和遗传算法(GA)在内的各种优化方法已被应用于设计结构色应用的光学结构。

While applying optimization methods to design optical structures for structural color, the designs are iteratively updated through the feedback from the electromagnetic simulations.

在应用优化方法设计结构色光学结构时，设计通过电磁仿真的反馈迭代更新。

Owing to the active feedback loop, optimization methods can often lead to highly accurate results but may suffer from low computational efficiency because of the large number of EM simulations.

由于主动反馈循环，优化方法通常可以产生高精度的结果，但由于大量的电磁仿真，可能遭受低计算效率的困扰。

Paragraph 6:

Machine learning methods, especially deep learning, have been recently applied to the task of structural color designs.

机器学习方法，特别是深度学习，最近已被应用于结构色设计任务。

Unlike numerical optimizations, machine learning methods learn models from the data points that can later be used to efficiently predict designs corresponding to user-specified color targets.

与数值优化不同，机器学习方法从数据点学习模型，以后可用于高效预测对应于用户指定颜色目标的设计。

However, machine learning could lead to lower design performance than optimizations because of the lack of active evaluations with EM simulations.

然而，由于缺乏电磁仿真的主动评估，机器学习可能导致比优化更低的设计性能。

Tandem networks combine a forward network and an inverse network to directly predict the designs corresponding to a color target.

串联网络结合前向网络和逆网络直接预测对应于颜色目标的设计。

The forward network predicted optical properties given the design, whereas the inverse network performs the opposite.

前向网络预测给定设计的光学特性，而逆网络执行相反操作。

However, both works can only output a single design given a user-specified color input, whereas our proposed method can output a set of potential solutions.

然而，这两项工作只能输出给定用户指定颜色输入的单一设计，而我们提出的方法可以输出一组潜在解决方案。

The capability of outputting a set of solutions allows researchers to select designs that are amenable to the fabrication process.

输出一组解决方案的能力允许研究人员选择适合制造工艺的设计。

Results 结果

Paragraph 7 - Overview:

The proposed method consists of four steps for structural color designs.

所提出的方法包括结构色设计的四个步骤。

In the first step, we collect training data with a wide range of colors.

在第一步中，我们收集具有广泛颜色的训练数据。

Then, a novel neural network model called material-aware multitask mixture density network (M3DN) is trained on the collected dataset.

然后，在收集的数据集上训练称为材料感知多任务混合密度网络(M3DN)的新型神经网络模型。

Next, for a given color design target, M3DN outputs a set of potential designs with different material combinations and layer thicknesses.

接下来，对于给定的颜色设计目标，M3DN输出具有不同材料组合和层厚度的一组潜在设计。

Finally, the set of initial designs are fine-tuned with PSO to obtain the final designs with optimized material selections and layer thicknesses.

最后，用PSO精细调整初始设计集合，以获得具有优化材料选择和层厚度的最终设计。

Paragraph 8 - Dataset Generation:

Previous research shows that five-layer optical thin films with two absorbing layers sandwiched by two dielectric layers and a bottom metal reflecting layer can achieve high color purity and brightness, where the layers can be easily deposited by physical vapor evaporation.

先前研究表明，具有两个吸收层被两个介电层和底部金属反射层夹在中间的五层光学薄膜可以实现高颜色纯度和亮度，其中各层可以通过物理气相蒸发轻松沉积。

Owing to the high performance and feasibility for large-scale fabrications of such structures, we synthesize a dataset with diverse designs based on the same five-layer structural template by varying both the material and thickness of each layer.

由于这种结构的高性能和大规模制造可行性，我们通过改变每层的材料和厚度，基于相同的五层结构模板合成具有多样化设计的数据集。

All designs are based on randomly sampled materials from ten candidate metal materials Au, Ag, Al, Cu, Cr, Ge, Ni, Ti, W, Zn, and ten dielectric materials Al₂O₃, Fe₂O₃, HfO₂, MgF₂, SiO₂, Ta₂O₅, TiO₂, ZnO, ZnS, and ZnSe.

所有设计基于从十种候选金属材料(Au, Ag, Al, Cu, Cr, Ge, Ni, Ti, W, Zn)和十种介电材料(Al₂O₃, Fe₂O₃, HfO₂, MgF₂, SiO₂, Ta₂O₅, TiO₂, ZnO, ZnS, ZnSe)中随机采样的材料。

Both absorber layers and the bottom reflective layers are composed of metals, whereas the other two layers are based on dielectric materials.

两个吸收层和底部反射层由金属组成，而其他两层基于介电材料。

Including a wide range of candidate materials with different refractive indices makes it possible to search for the most suitable materials combinations for specific color targets.

包括具有不同折射率的广泛候选材料使得可以搜索特定颜色目标的最合适材料组合。

Discussion 讨论

Paragraph 9:

In both the chrome color design and the picture reconstruction tasks, NEUTRON achieves exceptional design accuracy efficiently through combining a material classification network, a mixture density network for thickness predictions and the final PSO fine-tuning.

在镀铬颜色设计和图像重建任务中，NEUTRON通过结合材料分类网络、用于厚度预测的混合密度网络和最终PSO精细调整，高效地实现了卓越的设计精度。

Unlike previous methods that either assume fixed materials or search the materials and thickness designs simultaneously, our method split the materials and thickness design into two stages to enable an approach that can search for the best materials and design the thickness accordingly in an efficient manner.

与先前假设固定材料或同时搜索材料和厚度设计的方法不同，我们的方法将材料和厚度设计分为两个阶段，以实现可以高效搜索最佳材料并相应设计厚度的方法。

The reason why splitting the material and thickness design is helpful can be understood by the fact that the entire design space is much larger than the material design space or the thickness design space alone.

将材料和厚度设计分开有帮助的原因可以理解为，整个设计空间比单独的材料设计空间或厚度设计空间大得多。

In our five-layer optical thin film design task, the material design space size is 9 × 10⁴, and the thickness design space size is 7.3 × 10⁶, which lead to a huge full design space with a size of (9 × 7.3) × 10¹⁰ = 6.6 × 10¹¹ when considering the material design space and the thickness design space simultaneously.

在我们的五层光学薄膜设计任务中，材料设计空间大小为9×10⁴，厚度设计空间大小为7.3×10⁶，当同时考虑材料设计空间和厚度设计空间时，导致大小为(9×7.3)×10¹⁰=6.6×10¹¹的巨大完整设计空间。

With the two-step process, we first narrow down the promising material combinations with the material classification model and only optimize the thickness designs for the selected small set of materials.

通过两步过程，我们首先用材料分类模型缩小有前景的材料组合，然后仅为选定的材料小集合优化厚度设计。

Thus, we only need to consider a design space that is one or multiple times of the thickness design space, which is a ≈ 10,000× complexity reduction compared to searching both materials and thickness in one step.

因此，我们只需要考虑为厚度设计空间一倍或多倍的设计空间，与一步搜索材料和厚度相比，复杂度降低了约10,000倍。

Paragraph 10:

We combine M3DN and PSO by initializing the particle positions with designs sampled from the M3DN.

我们通过用从M3DN采样的设计初始化粒子位置来结合M3DN和PSO。

This process is easy to implement and highly effective in obtaining optimal designs because of the probabilistic nature of the mixture density network.

这个过程易于实现，在获得最优设计方面非常有效，因为混合密度网络具有概率性质。

On both design tasks, we show that PSO significantly improved the initial solutions by the M3DN.

在两个设计任务上，我们展示了PSO显著改进了M3DN的初始解决方案。

This result is not surprising, because PSO involves iterative updates of the design parameters based on the feedback from optical simulations.

这个结果并不令人惊讶，因为PSO涉及基于光学仿真反馈的设计参数迭代更新。

In addition, the exceptional design accuracy after fine-tuning the initial M3DN designs with PSO indicates that probabilistic machine learning models that can output diverse predictions are highly compatible with PSO, which requires an initial population of designs to begin with.

此外，用PSO精细调整初始M3DN设计后的卓越设计精度表明，能够输出多样化预测的概率机器学习模型与PSO高度兼容，PSO需要初始设计群体来开始。

Paragraph 11 - Limitations:

The training dataset for NEUTRON consists of PSO-optimized designs.

NEUTRON的训练数据集由PSO优化设计组成。

To achieve accurate design predictions, we collected 400K designs as the entire dataset.

为了实现准确的设计预测，我们收集了40万个设计作为整个数据集。

However, when applying NEUTRON to optical structures that require a longer simulation time, collecting 400K PSO-optimized designs might be infeasible.

然而，当将NEUTRON应用于需要更长仿真时间的光学结构时，收集40万个PSO优化设计可能不可行。

Thus, future study on improving the sample efficiency of NEUTRON is required.

因此，需要未来研究来提高NEUTRON的样本效率。

三、语言学习 (Language Learning)

3.1 雅思词汇 (IELTS Vocabulary)

3.2 科研术语 (Technical Terms)

3.3 学术表达 (Academic Expressions)

3.3.1 研究背景与动机

3.3.2 方法描述

3.3.3 结果与讨论

四、关键图表说明 (Key Figures)

Figure 1: NEUTRON Pipeline

• M3DN训练: 使用TMM生成的数据集训练
• 设计输出: 基于不同候选材料输出设计集合
• PSO精调: 通过粒子群优化进一步精细调整厚度

Figure 2: 五层光学结构与数据分布

• (A) 五层光学薄膜结构示意图
• (B) 数据集生成流程: 40万数据点通过PSO获得
• (C) 验证集颜色分布在CIE 1931 xy空间的覆盖
• (D) 随机采样RGB目标与PSO获得颜色的对比

Figure 5: 图像重建结果

• The White Orchard (梵高): CIEDE2000 = 3.76
• The Tulip Field (梵高): CIEDE2000 = 3.48
• The Great Wall: CIEDE2000 = 5.69

五、延伸阅读 (Further Reading)

基础方法

1. Bishop, C.M. (1994). “Mixture Density Networks.” (MDN奠基之作)
2. Kennedy, J. & Eberhart, R. (1995). “Particle Swarm Optimization.” (PSO奠基之作)

相关应用

3. Gao, H., et al. (2019). “Optimization of reflective color from silicon metasurface.” (Tandem网络)
4. Dai, Y., et al. (2021). “Design of three-layer Fabry-Perot color filters.” (多层薄膜设计)
5. Ma, T., et al. (2024). “OptoGPT: A foundation model for inverse design.” (后续基础模型)

Published: 2022 | Journal: iScience | DOI: 10.1016/j.isci.2022.104339