SYSTEM ONLINE · 2026 香港能带间隙量子科技有限公司Hong Kong Bandgap Quantum Technology

在能带之间， 重新定义价值。 Engineering the gap.

每一段能带间隙之间，都藏着一个可设计的光谱身份——以及一个从未被打开的市场。 Between every bandgap lies a designable spectral identity —— and a market that has never been opened.

从物理本质开始Start with the physics→ 阅读博士论文Read the thesis↗
1.6亿
吨退役光伏tons decommissioned PV
2050 全球累计 · IRENACUMULATIVE BY 2050 · IRENA
92.3%
电效率保留率 · 95% CI [89.5–95.1%]power retention · 95% CI [89.5–95.1%]
ANOVA p<0.001 · GUM k=2 · 商用 78%ANOVA p<0.001 · GUM k=2 · BIPV 78%
20×
单价跃迁 (鲁棒性已验证)unit-price uplift (robust)
LCOA 7 参数 ±20% 敏感性测算LCOA SENSITIVITY VERIFIED
96.4%
BgClassNet 分类准确率BgClassNet accuracy
12,378 模块 · 5 模态 · 5 失效模式12,378 MODULES · 5 MODALITIES
向下滚动开始SCROLL TO BEGIN
立 论 · MANIFESTOMANIFESTO / 立论

硅的能带间隙是 1.12 eV。这一个数字解释了所有问题——为什么每一块退役光伏看起来都一样，为什么世界上最大的工业固废流（2050 年累计 1.6 亿吨）至今没有出路。

我们把退役视为一个能带间隙工程问题：在保留 90% 以上光电效率的前提下，调制每一块退役组件的光谱身份，让硅从 0.3 元/W 的工业商品跃迁为 2,500 元/m² 的建筑材料。每一块面板都拥有可量化的光学指纹，于是我们把这枚指纹搬上链。 Silicon's bandgap is 1.12 eV. That single number explains everything from why every retired PV panel looks the same, to why the world's largest industrial waste stream — 160 million tons by 2050 — has nowhere to go.

We treat decommissioning as a bandgap engineering problem: by modulating the spectral identity of every retired module while preserving 90%+ of its photoelectric efficiency, we move silicon from 0.3 RMB / W commodity to 2,500 RMB / m² architectural material.

RESEARCH-DRIVEN MATERIALS COMPANY · FOUNDED 2026A RESEARCH-DRIVEN MATERIALS COMPANY · FOUNDED 2026
// 问题THE PROBLEM
160
百万吨 (1.6 亿吨)million metric tons
2050 全球累计退役光伏组件
来源：IRENA 2024 + 国家能源局测算 CUMULATIVE GLOBAL DECOMMISSIONED PV — 2050
SOURCE: IRENA 2024 + CHINA NEA

从 2026 年开始，仅中国每年就会产生 50 万吨退役光伏组件。到 2040 年这个数字将攀升至 200 万吨/年。

当前两条主流处置路径——物理拆解回收毛利不到 5%、低价梯次利用市场接受度极低——让整个行业默认这些资产是负资产。

我们不这样认为。 Starting 2026, China alone produces 0.5 million tons of decommissioned PV per year. By 2040 that number reaches 2 million tons annually.

Today's two disposal pathways — mechanical recovery at <5% gross margin, and cascade reuse with negligible market acceptance — leave the entire industry treating these assets as liabilities.

We disagree.

60 秒看懂IN 60 SECONDS

如果只读这一节——你需要知道四件事。 If you read only this — four things to know.

01
问题：1.6 亿吨退役潮已经开始。The problem: a 160-million-ton wave has started.
2026 年起，中国每年退役 50 万吨光伏组件；到 2050 年全球累计 1.6 亿吨。传统处置只有两条死路：低价回收（毛利 <5%）或低值再用（无人接盘）。From 2026, China alone retires 0.5M tons of PV modules per year; by 2050 global cumulative exceeds 160M tons. Today only two dead-ends exist: scrap recycling (<5% margin) or low-grade reuse (no demand).
02
方案：用「干涉」给它上色，发电不损耗。The solution: color it with interference — without losing power.
业界用染料涂色，必然损耗（天花板 ~78%）。我们用薄膜干涉造色，物理上无损（92.3% ± 1.2%，95% CI、ANOVA p<0.001、GUM k=2）。差 14.3 个百分点——是两个不同的行业。The industry uses dyes that absorb light (ceiling ~78%). We use thin-film interference that reflects it (92.3% ± 1.2%, with full statistical rigor). The 14.3-point gap separates two distinct industries.
03
阶段：M0 天使 · 物理验证中。Stage: M0 angel · proving the materials.
天使关账后 18 个月内完成宜兴中试线 + 3 个灯塔项目（一线商业 + 高端文旅 + 国际地标），同步推进 IEC/GB 国标主导，让「无损彩色」从公司主张升级为行业规范。After angel close, within 18 months: Yixing pilot line + 3 lighthouse projects (premium commercial + cultural tourism + international landmark); parallel push for IEC/GB standards leadership — turning "lossless color" from claim to regulation.
04
我们要找：耐心 LP · 战略产业方 · 联合创始人。We're looking for: patient LPs · strategic CVCs · co-founders.
不接快速套利型资本；不做横向平台；不做 Web3 优先叙事。如果你来自材料、能源、地产或建筑业，并能在三年的时间窗里耐心同行——往下读，或直接联系戴博士。No fast-arbitrage capital; no horizontal platforms; no Web3-first narrative. If you come from materials, energy, real-estate, or architecture and can walk a 3-year horizon — keep reading, or reach Dr. DAI directly.
→ 合作通道→ Partnership → 关键差异 14.3→ The Gap · 14.3
01 / 物理本质THE PHYSICS

颜色与价值，
共享同一个物理源头。 Color and value share
the same physical source.

半导体的能带间隙决定它吸收哪些光子、反射哪些、透射哪些——也就是同时决定了它的颜色、它的开路电压，以及它的市场定位。 A semiconductor's bandgap determines which photons it absorbs, which it reflects, and which it transmits — simultaneously its color, its open-circuit voltage, and its market position.

硅在 300 K 下的能带间隙为 Eg = 1.12 eV，对应吸收边 1107 nm。这迫使每一块硅基面板都吸收整个可见光谱，呈现出统治当今光伏界的工业深蓝。

通过沉积薄膜干涉堆栈、封装钙钛矿量子点、激光局部调制表面，我们在每块退役组件上工程化一个可调的光谱身份——同时不牺牲硅在敏感波段的光电响应。 Silicon at 300 K has a bandgap of Eg = 1.12 eV, corresponding to an absorption edge at 1107 nm. This forces every silicon panel to absorb across the entire visible spectrum, producing the industrial deep-blue that defines today's solar landscape.

By depositing thin-film interference stacks, encapsulating perovskite quantum dots, and laser-modulating the surface, we engineer a tunable spectral identity on every retired module — without sacrificing the silicon's photoelectric response in its sensitive bands.

"光伏的『工业感』不是审美选择。它是能带间隙未被调制的物理后果。" "The 'industrial' look of solar isn't an aesthetic choice. It's the physical consequence of an unmodulated bandgap."
摘自博士论文 第一章FROM PH.D. DISSERTATION, CHAPTER 1
第一性原理 · 理论FIRST PRINCIPLES · THEORY

不要看颜色，
要看波。 Don't look at the color.
Look at the wave.

一块组件为什么是工业深蓝，与它能卖多少钱，背后是同一个物理对象——电子在周期性势场中的波。我们的全部工作，都建立在看穿这一层之上。 Why a module is industrial blue, and what it can sell for, trace back to one physical object — a wave in a periodic potential. Everything we do is built on seeing through to that layer.

「废料」与「资产」之间，并没有一道材料的鸿沟。
它们隔着的，只是一次相干干涉——一个可以被计算、被设计、被工程化的物理过程。 能带间隙的物理灵魂 / THE SOUL BEHIND THE BANDGAP Between "waste" and "asset" there is no gulf of materials.
What separates them is a single coherent interference — a physical process that can be computed, designed, and engineered. THE SOUL BEHIND THE BANDGAP / 能带间隙的物理灵魂

I
理论介绍The Principle
能带从何而来Where bandgaps come from

能带间隙不是材料的属性，
而是波的命运。A bandgap is not a property of matter.
It is the fate of a wave.

1928 年，年仅 23 岁的费利克斯·布洛赫（Felix Bloch）在海森堡门下写下这一切的起点：当任何一种波穿过一个周期性结构，干涉会禁止某些能量存在。对硅晶体里的电子来说，这个周期就是原子点阵（约 0.5 纳米），于是出现一条由原子锁死的电子能带间隙 Eg = 1.12 eV。不换晶体，就动不了它——所以我们故意不去碰它，让硅继续安静发电。 In 1928, a 23-year-old Felix Bloch, working under Heisenberg, wrote down the starting point for all of this: when any wave travels through a periodic structure, interference forbids certain energies from existing. For electrons in a silicon crystal, that period is the atomic lattice (~0.5 nm), producing an electronic bandgap Eg = 1.12 eV, locked by the atoms. You cannot move it without changing the crystal — so we deliberately leave it untouched, and let the silicon keep generating power.

第一性原理First principle
周期性 → 禁带。同一个布洛赫定理，既支配晶格中的电子，也支配介质堆栈中的光子。Periodicity forbids. The same Bloch theorem governs electrons in a lattice and photons in a dielectric stack.

但波不在乎自己是电子还是光子。把周期性换一个尺度——在光波长的尺度（数百纳米）上交替堆叠折射率不同的薄膜——同一套数学就会产生一条光子能带间隙：一段被结构禁止通过、因而被反射的波长。它不由原子决定，而由几何决定；而几何，正是我们可以设计的东西。 But a wave does not care whether it is an electron or a photon. Move the periodicity to a different scale — alternating films of differing refractive index at the scale of light's wavelength (hundreds of nm) — and the same mathematics yields a photonic bandgap: a band of wavelengths the structure forbids, and therefore reflects. It is set not by atoms but by geometry — and geometry is precisely what we can design.

同一条定律，两个尺度：电子能带由原子锁定，光子能带由几何可调One law, two scales — the electronic gap is locked by atoms, the photonic gap is tunable by geometry
ELECTRONIC · 电子能带
Eg = 1.12 eV
由硅点阵决定，决定发电能力。不可调——我们刻意保留它。Set by the silicon lattice; governs power generation. Not tunable — we keep it intact by design.
PHOTONIC · 光子能带
Δω 可调Δω · tunable
由多层膜几何决定，决定反射出的颜色。可经厚度调谐——这就是杠杆。Set by multilayer geometry; governs the reflected color. Tuned by layer thickness — this is the lever.
CONFINEMENT · 量子限域
E ∝ 1/R²
由量子点尺寸决定（Brus），决定再发射的辉光。由粒径调谐。Set by quantum-dot size (Brus); governs the re-emitted glow. Tuned by particle radius.

这三种能带，来自三条原本独立的谱系——而它们在一块退役组件上第一次交汇。我们站在巨人的肩上：These three bandgaps descend from three originally separate lineages — and they converge, for the first time, on a single retired module. We stand on the shoulders of giants:

固体能带理论Solid-State Band Theory
1928
Felix Bloch
布洛赫定理：周期势场中的电子波——能带理论的基石 ★Bloch's theorem: electron waves in a periodic potential — the cornerstone of band theory ★
1929
Rudolf Peierls
禁带与「空穴」概念的起源The origin of the forbidden band and the concept of "holes"
1930
Léon Brillouin
布里渊区——倒空间中允带与禁带的边界Brillouin zones — the boundary of allowed and forbidden states in reciprocal space
1931
Alan H. Wilson
以能带填充区分金属 / 半导体 / 绝缘体Distinguished metals, semiconductors and insulators by band filling
光子能带间隙Photonic Bandgap
1887
Lord Rayleigh
一维周期介质堆栈的反射禁带——光子能带的最早先声The reflection stop-band of 1-D periodic dielectric stacks — the earliest photonic bandgap
1913
W.H. & W.L. Bragg
布拉格反射定律——多层膜干涉的工程基础Bragg's law of reflection — the engineering basis of multilayer interference
1987
Yablonovitch & John
光子晶体与完整光子能带间隙——「光子能带间隙」一词由此而来 ★Photonic crystals and the full photonic bandgap — the term itself originates here ★
量子限域Quantum Confinement
1981
Alexei Ekimov
在玻璃中首次观测到尺寸依赖的量子效应First observed size-dependent quantum effects, in glass
1983
Louis Brus
自由胶体量子点与尺寸–能带关系（Brus 公式）Free-standing colloidal quantum dots and the size–bandgap relation (the Brus equation)
2023
Nobel · Ekimov · Brus · Bawendi
诺贝尔化学奖：量子点的发现与合成 ★Nobel Prize in Chemistry: the discovery and synthesis of quantum dots ★
II
理论应用The Application
于是颜色可以无损Color without loss

于是，颜色
可以是无损的。Therefore, color
can be lossless.

行业给组件上色，靠的是染料和颜料——也就是分子吸收。吸收是做减法：每一个让面板「看起来有颜色」的光子，都是一个从发电里偷走的光子。这正是商用彩色 BIPV 普遍只能保留约 78% 功率的原因。那不是工艺缺陷，是热力学的征税。 The industry colors modules with dyes and pigments — that is, molecular absorption. Absorption is subtractive: every photon that makes a panel "look colored" is a photon stolen from power. This is exactly why commercial colored BIPV typically retains only ~78% of its power. It is not a flaw of craft; it is a tax levied by thermodynamics.

关键分野The decisive distinction
染料用「吸收」造色——做减法，必然损耗；涂层用「干涉」造色——做加法，原理上无损。Dye makes color by absorption — subtractive, inherently lossy. A coating makes color by interference — additive, in principle lossless.

因为光子能带是把选定的波段反射出去、而不是吸收掉，我们就能把反射峰放在硅最不敏感的波段，同时把透射窗口留在它最需要的地方。四要素能带特征向量（FBFV）把这件事写成一个可优化的目标；传输矩阵法（TMM）不过是布洛赫定理在有限层堆栈上的写法，差分进化把反射峰对准到 ±8 nm。最终 92.3% 的功率保留率不是技巧，而是——当你不再把「颜色」误当成「吸收」——物理本就允许的结果。 Because a photonic gap reflects the chosen band rather than absorbing it, we place the reflection peak where the silicon is least sensitive while keeping the transmission window open where it matters most. The Four-Component Bandgap Feature Vector (FBFV) writes this as one optimizable target; the Transfer Matrix Method is simply Bloch's theorem for a finite stack, and differential evolution aligns the peak to ±8 nm. The resulting 92.3% power retention is not a trick — it is what the physics permits once you stop mistaking "color" for "absorption."

F = [ Egeff , R(λ) , ηret , IDchain ]T四要素能带特征向量 · 把彩色化、低反射、保功率、可标识统一为一个可优化对象Four-component bandgap feature vector · unifies colorization, anti-reflection, power retention, and identity into one optimizable object
INTERFERENCE · 干涉
加法 · 结构色Additive · structural color
多层膜反射选定波段。原理上无损，颜色来自结构而非色素。我们的主路径。Multilayer films reflect a chosen band. Lossless in principle; color from structure, not pigment. Our primary path.
RE-EMISSION · 再发射
回收 · 量子点辉光Recovery · quantum-dot glow
量子点把损失的紫外下转换为可见辉光，能把能量「加回来」。增益路径。Quantum dots down-convert lost UV into a visible glow — they can add energy back. A gain path.
ABSORPTION · 吸收
减法 · 染料（我们规避）Subtractive · dye (we avoid)
分子吸收造色，必然偷走光子。行业的旧办法，也是 78% 天花板的来源。Molecular absorption makes color but inevitably steals photons. The incumbent approach — and the source of the 78% ceiling.
III
理论未来The Frontier
把光谱变成可编程资源A programmable spectrum

把光谱
当作可编程的资源。Treating the spectrum
as a programmable resource.

如果能带就是几何，那么「设计颜色」就变成一个逆问题：先指定你想要的光谱，再让计算去寻找能产生它的结构。正向设计（猜一个结构、算它的光谱）让位于逆向设计——由机器学习生成，最终走向比简单布拉格堆栈更丰富的非周期、准周期结构。 If a bandgap is geometry, then "designing color" becomes an inverse problem: specify the spectrum you want, then let computation find the structure that produces it. Forward design — guess a structure, compute its spectrum — gives way to inverse design, generative and ML-driven, eventually reaching aperiodic and quasi-periodic structures richer than a simple Bragg stack.

再往前，今天彼此分开的电子能带与光子能带，将开始耦合：光子再循环与荧光聚光，把再发射的光重新喂回硅，把功率保留率推过当前的天花板。「为了好看而管理光」和「为了发电而管理光」之间的界线，会逐渐消失。 Further out, the electronic and photonic gaps — kept separate today — begin to couple: photon recycling and luminescent concentration feed re-emitted light back into the silicon, pushing power retention past today's ceiling. The line between "managing light for looks" and "managing light for power" dissolves.

物理推论，而非营销A corollary, not a campaign
每块组件的光谱响应，都是由其物理结构唯一决定的指纹——这是物理的直接推论，可被独立测量与复核。Every module's spectral response is a fingerprint uniquely determined by its physical structure — a direct corollary of the physics, independently measurable and verifiable.
INVERSE DESIGN · 逆向设计
从结构→光谱，到光谱→结构Spectrum → structure
指定目标光谱，让生成式模型反解出薄膜结构。Specify the target spectrum; let a generative model solve back for the film structure.
PHOTON RECYCLING · 光子再循环
耦合两种能带Couple both gaps
把再发射光喂回硅，让保留率突破现有上限。Feed re-emitted light back into the silicon, breaking the present retention ceiling.
SPECTRAL IDENTITY · 光谱指纹
光谱即身份The spectrum is the ID
每块面板由结构唯一决定的反射光谱，本身就是不可伪造的物理指纹。A reflection spectrum uniquely determined by structure is itself an unforgeable physical fingerprint.

而最深的一次重构，是热力学的。传统回收用能量对抗熵——把材料熔回原料，付出巨大的能耗与碳成本，只换回很少的价值。高值再制造反其道而行：它注入的是信息——波长尺度上的结构——而不是能量。价值由「组织物质」创造，而非「摧毁物质」。这就是物理背后的灵魂，也是一块退役组件在我们手里能值二十倍的根本原因。 And the deepest reframing is thermodynamic. Conventional recycling fights entropy with energy — melting matter back to raw material at great energetic and carbon cost to recover little value. High-value remanufacturing does the opposite: it injects information — structure at the scale of a wavelength — rather than energy. Value is created by organizing matter, not by destroying it. That is the soul behind the physics, and the reason a retired module can be worth twenty times more in our hands than in a smelter's.

看任何问题，都不要停在表面。
颜色之下是波，废料之下是结构，价值之下是被重新组织的信息。 Never stop at the surface of any problem.
Beneath color is a wave; beneath waste is structure; beneath value is information, reorganized.

关键差异THE GAP · 14.3 POINTS

同一块退役光伏，
两条物理路径，两个结果。 Same module.
Two physics. Two outcomes.

业界普遍把「彩色光伏」当作一类产品——但物理上，染料造色与干涉造色是两个行业。一个吸收、必然损耗；一个反射、原理上无损。差距落在 14.3 个百分点上。 The industry treats "colored PV" as one category. Physically, it is two — absorption (dye) inevitably loses power; interference (coating) does not. The gap is 14.3 percentage points.

PATH A · INTERFERENCE
薄膜干涉 · 反射造色Thin-film interference · structural color
Reflect a chosen band · transmit the rest · silicon untouched
92.3%
95% CI [89.5 – 95.1%]
ANOVA p < 0.001 · GUM k = 2
能带间隙 · 无损彩色BANDGAP · LOSSLESS COLOR
PATH B · ABSORPTION
分子吸收 · 染料造色Molecular absorption · pigment color
Absorb the chosen band · convert to heat · cover the silicon
~78%
业界染料 BIPV 公认天花板industry ceiling for dye / pigment BIPV
无公开统计严苛性no published CI · no statistical rigor
染料 BIPV · 物理上必然损耗DYE BIPV · LOSSY BY PHYSICS
关键结论The decisive insight
两种工艺看起来都是「彩色光伏」——物理上，它们是两个行业。能带间隙争取把这条 14.3 个百分点的物理鸿沟，写进 IEC TC82 与 GB 国标，让「无损彩色」与「损耗型彩色」从此在合规层面分开。Two crafts that look like "colored PV" are physically two industries. We aim to write the 14.3-point gap into IEC TC82 and Chinese GB standards — so that "lossless color" and "lossy color" are compliance-distinct from now on.
02 / 四类机制FOUR MECHANISMS

四种方法，
调制一个能带间隙。 Four ways
to engineer a bandgap.

单一机制无法跨越 5% 毛利的护城河。把多层薄膜干涉、量子限制、钙钛矿组分调控、飞秒激光图案化四种机制叠加协同，才能在保留光电效率的前提下产生 20× 的价值跳升。 No single mechanism crosses the 5% margin moat. Stacking four of them — multilayer interference, quantum confinement, perovskite tuning, and femtosecond patterning — produces a 20× value lift while preserving photoelectric efficiency.

M1 / 01

薄膜干涉Thin-Film Interference

Thin-Film Interference薄膜干涉

基于传输矩阵法（TMM）设计多层介质堆栈，磁控溅射沉积。五系列彩色（深蓝 → 深空黑），反射峰位精度 ±8 nm。 Multilayer dielectric stacks designed via Transfer Matrix Method, deposited by magnetron sputtering. Five-color series with peak accuracy ±8 nm.

→ ΔEab* ≤ 1.6 / ηret 92.3%
M2 / 02

量子限制Quantum Confinement

Quantum Confinement量子限制

钙钛矿量子点（CsPbBr3 / CsPbBrxI3-x）通过尺寸与卤素组分调谐，封装入梯度浓度 EVA 复合膜。 Perovskite quantum dots (CsPbBr3 / CsPbBrxI3-x) tuned by dot size and halide composition, encapsulated into gradient-concentration EVA composite films.

→ PLQY 78.5% / IEC 61215 ×2 passed
M3 / 03

钙钛矿组分Perovskite Composition

Perovskite Composition钙钛矿组分

通过卤素阴离子替换实现带隙连续调谐——从 CsPbI3（1.73 eV）到 CsPbBr3（2.30 eV），覆盖整个可见光发射谱。 Continuous bandgap tuning from 1.73 eV (CsPbI3) to 2.30 eV (CsPbBr3) via halide anion substitution. Enables the full visible emission palette.

→ 1.73 — 2.30 eV continuous
M4 / 04

飞秒激光Femtosecond Patterning

Femtosecond Patterning飞秒激光

超短脉冲激光刻写亚波长光栅，实现空间光谱调制、渐变色彩与表面图案化身份。 Ultrashort-pulse laser inscription of subwavelength gratings, enabling spatial color modulation, gradient effects, and patterned identity on the coating surface.

→ 空间光谱图案化Spatial spectral patterning
03 / 性能突围THE PERFORMANCE

电效率保留率
重新定义市场。 Power retention
that reorders the market.

现有彩色光伏产品都在用电效率换审美。把四种机制放进同一个能带间隙工程框架，我们在江苏宜兴中试线达到了 92.3% 的电效率保留率——本品类首个突破 90% 阈值的产品。 Every existing colored-PV product trades efficiency for aesthetics. By placing all four mechanisms inside a single bandgap-engineering framework, our pilot line at Yixing reached 92.3% efficiency retention — the first product in this category to break the 90% threshold.

国内染料 BIPVDomestic dye BIPV
60.0%
Solaxess (CH)
70.0%
Solar Decor (KR)
75.0%
Kromatix (CH)
78.0%
MorphoColor (DE)
82.0%
能带间隙量子（本工作）Bandgap Quantum
95% CI [89.5–95.1%] · ANOVA p<0.001
92.3%
05 / 学术研究THE RESEARCH

五年时间凝结的
一篇博士论文。 A doctoral thesis
five years in the making.

支撑这家公司的物理学并非为融资而发明。它是五年博士研究的成果——9 万字、22 张正式实验三线表、237 条同行评审引用，以及一条自 2024 年 10 月起运行的中试线。 The physics underwriting our company is not invented for fundraising. It is the result of a five-year doctoral investigation that produced 90,000 words, 22 formal experimental tables, 237 peer references, and a pilot line operational since October 2024.

面向退役晶硅光伏组件高值再制造的能带间隙工程理论与实验研究Bandgap Engineering Theory and Experimental Investigation for High-Value Remanufacturing of Decommissioned c-Si PV Modules

Bandgap Engineering Theory and Experimental Investigation for High-Value Remanufacturing of Decommissioned Crystalline Silicon Photovoltaic Modules面向退役晶硅光伏组件高值再制造的能带间隙工程理论与实验研究

PH.D. DISSERTATION DEFENSE · 2026

面向退役晶硅光伏组件高值再制造的能带间隙工程理论与实验研究

Bandgap Engineering for High-Value Remanufacturing of Decommissioned c-Si PV Modules

DAI JITAO · 戴 继 涛 Bandgap engineering · Decommissioned PV · High-value remanufacturing May 2026
8CH.
章节Chapters
90,000 字90,000 WORDS
237
参考文献References
同行评审PEER-REVIEWED
12,378
实验组件 BgDB v1.0Modules in BgDB v1.0
五模态数据5-MODALITY DATA
96.4%
BgClassNet 准确率BgClassNet accuracy
RESNET-50 + X-注意力RESNET-50 + X-ATTN
12 / VIDEO · ON REQUEST

答辩视频精讲Defense walkthrough

06'39" · 18 学术帧 · 双语字幕。
致信索取，我们将通过加密链接发送。06'39" · 18 academic frames · bilingual subtitles.
Email to request — we'll send via encrypted link.

致信索取Email to request ↗
hi@bandgap.hk 12 / V
11

论文全文 · 完整版Dissertation · Full Text

9 万字 · 8 章 · 237 引用 · 22 张三线表90,000 words · 8 chapters · 22 three-line tables · 237 references

致信索取 →Request by email →
12

答辩 PPT · 23 页Defense Slides · 23 pages

学术风格 · 八章核心 · Q&A 附录Academic format · 8-chapter compression · Q&A appendix

致信索取 →Request by email →
06 / 论文解析CHAPTER ANALYSIS

把 10.7 万字博士论文，
拆给所有人看。 Unpacking 107,000 words
for everyone.

论文不应该只属于答辩委员会。我们把八章核心成果逐章拆解，每章用「研究问题 → 解决方法 → 关键发现」三层结构呈现，配以原创数据可视化与原文引用——让投资人、社区、技术合作伙伴都能在 10 分钟内吃透 5 年研究的精髓。 Research shouldn't only belong to defense committees. We break down all eight chapters using a Question → Method → Finding three-layer structure, with original data visualizations and source citations — letting investors, the community, and technical partners absorb five years of research in ten minutes.

// 最新版本 v3 (v2026.5-rev) · 178 页 · 严苛性修订完成：A1-A6 关键漏洞 + B1-B7 中等问题 + D 研究方法学声明全部修补。新增 §1.6 学位论证范畴声明 / §3.4.1 数据集分布偏倚 / §4.6.3 统计严密性（ANOVA + GUM 不确定度）/ §4.6.4 批次重复性 / §6.4.3 LCOA 敏感性分析 / §7.1 综述法方法学声明 / Brus 介电屏蔽修正 / SiNx 减反层论证 / PVQAT-TG3 标准化引用。 // LATEST v3 (v2026.5-rev) · 178 pages · Rigorous revision complete: 6 critical + 7 moderate + structural fixes. New: §1.6 thesis scope declaration / §3.4.1 distribution shift disclosure / §4.6.3 statistical rigor (ANOVA + GUM uncertainty) / §4.6.4 batch reproducibility / §6.4.3 LCOA sensitivity / §7.1 survey methodology / Brus dielectric screening correction / SiNx ARC argument / PVQAT-TG3 standard.

01
CHAPTER ONE · 绪 论INTRODUCTION

为什么是能带间隙工程Why Bandgap Engineering

A trillion-yuan industrial waste stream meets a Nobel-prize-grade physics framework

KEY INSIGHT
退役光伏不是工艺问题，是物理问题——硅 1.12 eV 的天然能带间隙未被人为调制，导致所有面板光谱身份完全一致。 Decommissioned PV is not a process problem — it is a physics problem. Silicon's native 1.12 eV bandgap remains unmodulated, leaving every panel with an identical spectral signature.
QUESTION / 问 题Q

为什么 50 万吨/年退役光伏的处置工业，毛利率始终被锁死在 5% 以下？ Why is the gross margin of the 0.5M-tons-per-year decommissioning industry structurally capped below 5%?

METHOD / 方 法M

从经济学诊断转向物理学诊断——把"工业感"还原为可调制的能带物理量。 Pivot from an economic diagnosis to a physical one — recasting the "industrial look" as a tunable bandgap parameter.

FINDING / 发 现F

问题的物理根源被定位：所有退役面板的光谱身份完全一致，因此被天然排除在高价值视觉空间之外。 Root cause located: every retired panel shares the same spectral identity, structurally excluding them from high-value visual markets.

2026 年起，仅中国每年产生 50 万吨退役光伏组件；2040 年达到 200 万吨；2050 年全球累计 1.6 亿吨。本章首先用三组定量数据证明这是一个不可避免的工业级冲击。 From 2026, China alone generates 0.5M tons of retired PV panels per year, rising to 2M by 2040 and 160M cumulative globally by 2050. The chapter establishes this as an unavoidable industrial shock through three quantitative datasets.

现有处置只有两条路：拆解卖材料（毛利 < 5%、能耗高、化学品污染）或低价梯次利用（接受度极低、电站运营商拒绝采购）。本章对这两条路径分别建立失效模型，定量估计其经济边界。 Existing pathways — material recovery (margin <5%, high-energy, chemical-heavy) and cascade reuse (rejected by operators) — are formally modeled with quantitative failure boundaries.

本章的破局起点：跳出"工艺成本博弈"，把退役光伏的高值化重新定义为能带间隙工程问题。这一物理学起点贯穿后续所有章节，是整个研究范式的逻辑锚点。 The breakthrough framing: step outside the "process-cost game" and reframe high-value remanufacturing as a bandgap engineering problem. This physics-first starting point anchors every subsequent chapter.

CHINA · ANNUAL DECOMMISSIONED PV (MILLION TONS) 2.0M 1.5M 0.5M 20260.5M 20301.0M 20351.5M 20402.0M
图 1.1 · 中国 2026-2040 年退役光伏组件年产生量预测（数据：IRENA + 国家能源局）Fig. 1.1 · China annual decommissioned PV stream, 2026-2040 (Source: IRENA + China NEA)
"现有处置工艺均处于低毛利-低单位价值区间，整个行业默认这些资产是负资产。" "All existing disposal pathways trap end-of-life modules in low-margin, low-value commodity territory."
— DISSERTATION CH.1, p.6
02
CHAPTER TWO · 理 论 基 础FOUNDATIONS

FBFV 四要素能带特征向量FBFV: Four-Component Bandgap Feature Vector

A unified mathematical framework that collapses four engineering goals into one physical equation

KEY INSIGHT
把彩色化、低反射、保功率、可标识四个原本割裂的工程目标，归一到一个四维向量 F = [E_g_eff, R(λ), η_ret, ID_chain]ᵀ。 Four previously disjoint engineering goals — colorization, anti-reflection, power preservation, traceable identity — collapse into a single four-vector F = [E_g_eff, R(λ), η_ret, ID_chain]ᵀ.
QUESTION / Q

如何在统一的物理框架下，描述退役光伏高值再造的全部工程目标？ How can all engineering goals of high-value remanufacturing be unified under a single physical framework?

METHOD / M

梳理 Bloch 定理、量子限制、Shockley-Queisser 极限等核心物理，构造四要素向量。 Drawing on Bloch's theorem, quantum confinement, and the Shockley-Queisser limit to construct a four-component vector.

FINDING / F

FBFV 把"商业目标"转化为"物理参数"，让设计、检测、标准化都有了同一种语言。 FBFV translates business goals into physical parameters, providing a shared language for design, inspection, and standardization.

本章从 Bloch 定理出发，回顾晶体能带结构的核心结论，详细推导硅在 X 方向 k ≈ 0.85·(2π/a) 处的 1.12 eV 间接带隙的物理来源；继而回顾 Shockley-Queisser 极限与外量子效率公式，说明 EQE 主要由 (1 - R(λ)) 决定。 Beginning with Bloch's theorem, the chapter derives the physical origin of silicon's 1.12 eV indirect gap at X-point k ≈ 0.85·(2π/a). It then revisits the Shockley-Queisser limit, showing that EQE is dominated by (1 - R(λ)).

系统比较四类能带调制机制——薄膜光学干涉（结构色）、量子限制（尺寸调控）、钙钛矿组分调控、飞秒激光局部调制——并把每一类对应到 FBFV 中独立的"调制方向"。 A systematic comparison of four bandgap modulation mechanisms — thin-film interference (structural color), quantum confinement (size tuning), perovskite composition tuning, and femtosecond patterning — maps each mechanism to an independent modulation direction in FBFV space.

FBFV 的几何意义：每一块经过能带工程化处理的退役组件，对应于 FBFV 空间中的一个唯一向量；商业上有价值的产品集合是一个紧致子集；本研究的工程目标是找到能稳定到达这一子集的工艺参数空间。 Geometrically: every bandgap-engineered module is a unique vector in FBFV space; commercially valuable products form a compact subset; engineering goal = find process-parameter space that reliably reaches this subset.

F = [E_g_eff, R(λ), η_ret, ID_chain]ᵀ E_g_eff 等效带隙 eV R(λ) 反射光谱 CIE color η_ret 效率保留 ≥ 0.90 ID_chain 光谱指纹 SPECTRAL ID FOUR-COMPONENT BANDGAP FEATURE VECTOR
图 2.9 · FBFV 四要素能带特征向量的几何表示Fig. 2.9 · Geometric representation of the four-component bandgap feature vector
"FBFV 框架是后续第三至第七章工作的理论锚点。" "The FBFV framework is the theoretical anchor for all of Chapters 3-7."
— DISSERTATION CH.2, p.71
03
CHAPTER THREE · 退 役 检 测INSPECTION

BgDB v1.0 + BgClassNetBgDB v1.0 + BgClassNet

A 12,378-module multimodal dataset and a 96.4%-accurate fusion classifier — the industry's first "raw-material fingerprint."

KEY INSIGHT
每一块退役组件入厂前必须先回答："你属于哪类？" — 通过 5 模态 220 GB 数据 + Cross-Attention 融合网络，准确率 96.4%。 Every retired panel must first answer: "Which failure class are you?" — solved with 5-modality 220 GB data and a Cross-Attention fusion net at 96.4% accuracy.
QUESTION / Q

退役组件相对原始组件，能带特征发生了哪些可量化变化？ How does the bandgap signature of a retired panel quantitatively differ from a fresh panel?

METHOD / M

PL/EL/反射光谱/椭偏/IR 五模态采集 + ResNet-50 + Cross-Attention 融合网络。 5-modality acquisition (PL/EL/reflectance/ellipsometry/IR) + ResNet-50 + Cross-Attention fusion network.

FINDING / F

12,378 块组件、6 类标签、96.4% 准确率，国内首个产业级 AI 检测系统。 12,378 panels · 6 classes · 96.4% — the first industrial-grade AI inspection system for retired PV in China.

本章首先建立失效模式与能带特征演化的对应关系：PID 导致少子寿命下降 60%、EVA 黄化使 380-500 nm 反射率下降 4-12%、热斑/断栅/微裂纹局部影响 EQE。 The chapter first maps failure modes to bandgap signature evolution: PID drops minority-carrier lifetime by 60%, EVA yellowing reduces 380-500 nm reflectance by 4-12%, hotspots/broken fingers/microcracks locally compromise EQE.

基于此构建 BgDB v1.0：4 省 11 家电站 12,378 块退役组件，每块包含 5 模态原始数据（PL、EL、反射光谱、椭偏、IR），共 220 GB。这是国内外第一个公开的退役晶硅组件能带特征数据集。 From this we constructed BgDB v1.0: 12,378 panels from 11 stations across 4 provinces, each containing 5 modalities of raw data (PL, EL, reflectance, ellipsometry, IR), totaling 220 GB — the first publicly available bandgap-feature dataset for retired c-Si modules.

BgClassNet 在测试集上达到 96.4% 分类准确率，相比 ViT-B/16 多模态基线提升 2.9 个百分点，已部署在江苏宜兴中试基地。 BgClassNet reaches 96.4% on held-out tests, +2.9pp over the ViT-B/16 multimodal baseline, deployed at the Yixing pilot site since October 2024.

CLASSIFICATION ACCURACY · TEST SET ResNet-50 (PL only)83.7% ResNet-50 multimodal88.2% MMTM92.1% ViT-B/16 multimodal93.5% BgClassNet (本工作)96.4%
表 3.3 · BgClassNet 与基线模型分类性能对比Table 3.3 · BgClassNet vs. baseline models
"为后续高值再制造提供了精准的『原料指纹』。" "Provides a reliable 'raw-material fingerprint' for downstream remanufacturing."
— DISSERTATION CH.3, p.96
04
CHAPTER FOUR · 薄 膜 涂 层COATINGS

涂膜能带间隙工程的统一理论与五系列彩色涂层Unified Theory of Coating Bandgap Engineering & Five-Series Color Coatings

From Bloch theorem extension to photonic bandgap design — why coating beats dyeing by physics, not by craft.

KEY INSIGHT
涂膜从未改变硅的电子带隙——它在硅之上人造一个光子带隙。这是涂膜（保留 92.3% 效率）vs 染料（仅 60-75%）差异的物理根源——不是工艺差距，是机制差距。 Coatings never alter silicon's electronic bandgap — they engineer a photonic bandgap on top of it. This is the physics-level reason coatings retain 92.3% efficiency while dyes peak at 60-75% — not a craft gap, but a mechanism gap.
QUESTION / Q

"能带间隙工程"在涂膜语境下到底改变了什么？为什么涂膜可以保留 90%+ 电效率而染料不能？What does "bandgap engineering" actually change in coating? Why do coatings preserve 90%+ efficiency while dyes cannot?

METHOD / M

把 Bloch 定理推广到光学周期介质，证明涂膜本质是构造一维光子带隙；TMM 主方程 + 差分进化求解。Extend Bloch theorem to optical periodic media, proving coating = engineering a 1D photonic bandgap; TMM + Differential Evolution.

FINDING / F

三类能带间隙严格区分；涂膜（加法着色）vs 染料（减法着色）= 机制级差异；五色 92.3% η_ret 由理论必然推出。Three bandgap types rigorously distinguished; coating (additive) vs dye (subtractive) = mechanism-level difference; 92.3% η_ret follows necessarily from theory.

// THREE BANDGAPS, ONE TERM — A SOURCE OF SYSTEMIC CONFUSION
ELECTRONIC E_g
电子能带间隙Electronic bandgap
硅天然 1.12 eV，由材料决定，涂膜不改变Si: 1.12 eV, intrinsic to material, coating cannot alter
PHOTONIC Δω_PBG
光子能带间隙Photonic bandgap
由结构决定，涂膜的主战场，可任意工程化Determined by structure, where coating engineers act, fully tunable
EFFECTIVE E_g_opt
等效光学带隙Effective optical gap
系统对外的吸收边，由 Tauc plot 提取，视觉意义上的颜色External absorption edge via Tauc plot — "color" as we see it

本章把 Bloch 定理从晶体周期势 V(r) 推广到光学周期介质 ε(z)。证明：在折射率周期排布的多层薄膜（H/L/H/L/...）中存在频率区间，电磁波无法以传播模式存在——这就是光子能带间隙（Photonic Bandgap, PBG）。涂膜彩色化的全部物理学，都还原到这一个机制：把 PBG 中心频率调到对应的可见光波长。 This chapter extends Bloch's theorem from crystal periodic potential V(r) to optical periodic media ε(z). The proof: in periodic dielectric stacks (H/L/H/L/...), certain frequency intervals admit no propagating EM modes — the photonic bandgap (PBG). All coating-based colorization reduces to this single mechanism: tuning the PBG center frequency to the target visible wavelength.

由此推出 Bragg 反射镜的两个工程公式：中心频率 ω₀ = πc/(n_H·d_H + n_L·d_L)，相对带宽 Δω/ω₀ = (4/π)·arcsin[(n_H−n_L)/(n_H+n_L)]。这是涂膜设计最干净的"配方"——对于 545 nm 翠绿色，TiO₂ (2.45) + SiO₂ (1.46)，单层 55.6 / 93.3 nm，7 层堆栈即可获得 R>47% 的反射峰，FWHM 80 nm，NIR 反射率天然<6%。 From this follow two engineering formulas for Bragg mirrors: center frequency ω₀ = πc/(n_H·d_H + n_L·d_L); relative bandwidth Δω/ω₀ = (4/π)·arcsin[(n_H−n_L)/(n_H+n_L)]. The cleanest "recipe" for coating: for 545 nm emerald, TiO₂ (2.45) + SiO₂ (1.46), thicknesses 55.6 / 93.3 nm, a 7-layer stack delivers R>47% peak with FWHM 80 nm and intrinsic NIR reflectance <6%.

实际工程化设计无解析解，由 TMM 主方程（特征矩阵 M_j → 总矩阵 M → 反射率 R(λ)）+ 差分进化算法（种群 200 × 1000 代）数值求解。多目标损失 L = w₁·‖R−R*‖² + w₂·(1−η_ret) + w₃·ΔE，权重 (0.5, 1.5, 1.0)——w₂ 取最大不是巧合：把"保发电"放到最高优先级，是 92.3% 超越行业 78% 的核心工程决策。 Real engineering has no closed form — solved numerically via the TMM master equation (per-layer M_j → total M → reflectance R(λ)) + Differential Evolution (population 200 × 1000 generations). Multi-objective loss L = w₁·‖R−R*‖² + w₂·(1−η_ret) + w₃·ΔE with weights (0.5, 1.5, 1.0) — the maximal w₂ is no accident: prioritizing power preservation is the core engineering decision behind 92.3% beating the 78% industry mean.

涂膜 vs 染料 — 机制级差异，不是工艺差距：染料是分子能级吸收（减法着色），分子振动展宽不可避免地把硅敏感的 NIR 也吃掉，电效率损失 25-40%；涂膜是相干干涉反射（加法着色），可显式约束 NIR 反射率<6%，电效率损失 5-12%。把染料工艺优化到极致仍突破不了 75% 上限——这是物理决定的天花板，不是工艺水平问题。 Coating vs dye — mechanism-level, not craft-level: dyes work by molecular electronic absorption (subtractive), with vibrational broadening inevitably eating into silicon's NIR-sensitive band — 25-40% efficiency loss. Coatings work by coherent interference reflection (additive), with explicit constraint of NIR reflectance below 6% — 5-12% efficiency loss. Optimizing dye craft to perfection cannot break the 75% ceiling — that's a physics-imposed wall, not a craftsmanship issue.

论文进一步在 4.1.6 节梳理了三种嵌入纳米结构的扩展机制：(1) 嵌入半导体量子点 → Brus 公式给出尺寸调谐带隙；(2) 嵌入金属纳米颗粒 → LSPR 共振 ω_LSPR ≈ ω_p/√(1+2ε_m)；(3) 梯度折射率 GRIN → Bruggeman 等效介质理论。这三种机制可与 Bragg 基底叠加，构成"基础堆栈 + 嵌入 QD + 嵌入 Au + GRIN 渐变"的复合涂膜——这是论文 8.4 节"未来研究方向 1：多机制协同"的物理基础。 Section 4.1.6 further develops three embedded nanostructure mechanisms: (1) semiconductor QD embedding → Brus formula for size-tuned gap; (2) metallic NP embedding → LSPR resonance ω_LSPR ≈ ω_p/√(1+2ε_m); (3) graded-index GRIN → Bruggeman effective medium theory. All three layer onto a Bragg base, forming "stack + QD + Au NP + GRIN" composite coatings — the physics underpinning Section 8.4 future direction 1: multi-mechanism synergy.

BRAGG REFLECTOR · 1D PHOTONIC BANDGAP CONSTRUCTION 7-layer Bragg stack (TiO₂/SiO₂) H L H L H L H d_H = 55.6 nm d_L = 93.3 nm CENTER FREQUENCY ω₀ = πc / (n_H·d_H + n_L·d_L) → λ₀ = 545 nm (emerald green) RELATIVE BANDWIDTH Δω/ω₀ = (4/π) · arcsin[(n_H − n_L)/(n_H + n_L)] → TiO₂/SiO₂: Δω/ω₀ ≈ 0.31, FWHM ≈ 80 nm EFFICIENCY LOSS COMPARISON DYE (subtractive): 25-40% lost COATING (additive): 5-12% lost
图 4.1 · Bragg 反射镜的物理构造：周期性 H/L 介电堆栈 + 中心频率/带宽公式 + 涂膜与染料的电效率损失对比Fig. 4.1 · Physical construction of a Bragg reflector: periodic H/L dielectric stack + center frequency/bandwidth formulas + coating vs. dye efficiency loss comparison
REFLECTANCE SPECTRA · R(λ) — FIVE COATING SERIES 60% 30% 380nm 1100nm 465 545 590 650 Black: broadband < 6%
图 4.2 · 五系列涂层反射光谱（峰位 465 / 545 / 590 / 650 nm + 深空黑全波段）Fig. 4.2 · Reflectance spectra of five coating series
REFLECTANCE SPECTRA · R(λ) — FIVE COATING SERIES 60% 30% 380nm 1100nm 465 545 590 650 Black: broadband < 6%
图 4.1 · 五系列涂层反射光谱（峰位 465 / 545 / 590 / 650 nm + 深空黑全波段）Fig. 4.1 · Reflectance spectra of five coating series (peaks at 465 / 545 / 590 / 650 nm + broadband black)
"涂膜从未改变硅的电子能带间隙；被工程化的是光子能带间隙——这是涂膜彩色化的根本物理基础。" "Coatings never alter silicon's electronic bandgap; what's engineered is the photonic bandgap — the fundamental physical basis of coating-based colorization."
— DISSERTATION CH.4.1, p.99
05
CHAPTER FIVE · 量 子 点 封 装QUANTUM DOTS

钙钛矿量子点光致发光封装Perovskite QD Photoluminescent Encapsulation

Gradient-concentration EVA composite films deliver 78.5% PLQY with 25-year outdoor stability.

KEY INSIGHT
不只是给面板"染色"——通过 UV 下转换，挽回硅响应较弱的紫外光，实现"白天发电、夜间发光"双功能。 More than aesthetic colorization — UV down-conversion recovers UV photons that silicon barely responds to, enabling a panel that generates by day, glows by night.
QUESTION / Q

如何让退役光伏组件同时具备发电与可见光发射两种能力？How to give a retired panel both power generation and visible-light emission?

METHOD / M

CsPbBr₃ 系量子点 + 双配体钝化 + 梯度浓度 EVA 复合膜（5/3/1 wt%）。CsPbBr₃-family QDs + dual-ligand passivation + gradient EVA encapsulation (5/3/1 wt%).

FINDING / F

PLQY 78.5%，IEC 61215 双倍加速 PLQY 衰减 < 17.2%，黄/橙/红三色组件。PLQY 78.5%, <17.2% drop after doubled IEC 61215, yellow/orange/red modules.

本章合成 CsPbBr₃（黄）、CsPbBr₂I（橙）、CsPbBr₁.₅I₁.₅（红）三系列全无机钙钛矿量子点，TEM 实测尺寸 8.7-10.2 nm，原始 PLQY 67.5-88.2%。 Three all-inorganic perovskite QD series synthesized: CsPbBr₃ (yellow), CsPbBr₂I (orange), CsPbBr₁.₅I₁.₅ (red), TEM-confirmed sizes 8.7-10.2 nm with raw PLQY 67.5-88.2%.

引入双配体钝化（C20-酸 + 三辛胺 TOA），让 QD 在 65% RH 环境 30 天后 PLQY 保留率从 31% 提升到 86%；进一步开发梯度浓度 EVA 复合膜（上 5%/中 3%/下 1%）平衡下转换效率与吸收损失。 A dual-ligand passivation (C20-acid + TOA) lifts 30-day PLQY retention at 65% RH from 31% to 86%. A gradient EVA composite film (5%/3%/1% top-to-bottom) balances down-conversion gain against absorption loss.

最终三色 PL-PV 组件电效率保留率 92.7-97.4%（黄色组件最佳），EQE 在 365-450 nm 波段提升 12-18%——这意味着同样硅基底 + 同样阳光下，年发电增量 1-3%。 Final tri-color PL-PV modules retain 92.7-97.4% efficiency (yellow leads), with EQE gain of 12-18% in 365-450 nm — equating to 1-3% extra annual yield on the same silicon under the same sun.

PEROVSKITE QUANTUM DOTS · EMISSION SPECTRA CsPbBr₃ 518 nm PLQY 88% CsPbBr₂I 586 nm PLQY 75% CsPbBr₁.₅I₁.₅ 658 nm PLQY 67% +I⁻ +I⁻
图 5.4 · 通过卤素阴离子替换实现量子点带隙连续调谐（黄 → 橙 → 红）Fig. 5.4 · Continuous QD bandgap tuning via halide substitution (yellow → orange → red)
"突破了量子点光伏封装的稳定性瓶颈。" "Breaks the long-standing stability bottleneck of QD-PV encapsulation."
— DISSERTATION CH.5, p.158
06
CHAPTER SIX · 产 业 化INDUSTRIALIZATION

三层一体产业化系统Three-Layer Industrialization System

Pilot line + digital twin + spectral identity registry — the LCOA Pareto frontier proves business viability.

KEY INSIGHT
从实验室走向产业不是"放大问题"，而是"重新定价问题"——本章提出 LCOA（美学资产度量化成本）模型，论证 BEHVR 在 LCOA / 美学耐久 / 碳成本三维同时帕累托最优。 Going from lab to industry is not a "scale problem" but a "pricing problem." We introduce LCOA (Levelized Cost of Aesthetic Asset), proving BEHVR's Pareto-optimality across LCOA / aesthetic durability / carbon cost.
QUESTION / Q

实验室成果如何变成可规模化、可经济性站得住脚的产业？How does a lab result become a scalable, economically defensible industry?

METHOD / M

中试产线 + 工艺数字孪生 + 光谱指纹注册库 + LCA + LCOA。Pilot line + process digital twin + spectral identity registry + LCA + LCOA.

FINDING / F

江苏宜兴 1.2 万片/月产能；LCOA 835-1080 元/m²，三维度帕累托最优。Yixing pilot at 12,000 panels/month; LCOA 835-1080 RMB/m², Pareto-optimal across three axes.

本章设计了 11 工位 U 型布局的自动化能带特征检测产线，节拍 80 秒/片，BgClassNet 推理工位的产线实测准确率 95.7%（实验室 96.4%）。 An 11-station U-shaped pilot line processes one panel every 80 seconds; BgClassNet inference reaches 95.7% accuracy in-line (vs. 96.4% in-lab).

基于 Unity3D + Python + AWS 开发 BgTwin v1.0 工艺数字孪生平台，包含 TMM 光学仿真、磁控溅射沉积仿真、QD-EVA 共熔仿真、IV 特性仿真，支持 60 秒级单组件全工艺仿真，已申请 8 项软著。 BgTwin v1.0 (Unity3D + Python + AWS) integrates TMM optical, sputtering deposition, QD-EVA blending, and IV simulators, with 60-second single-panel runs and 8 registered software copyrights.

基于 ISO 14040 LCA 方法测得 BEHVR 产品全生命周期碳排 9.10 kg CO₂eq/m²，相比铝单板 28.4、玻璃幕墙 35.7 显著降低。LCOA 模型证明 835-1080 元/m² 的成本结构在三维度同时占据帕累托最优。 ISO-14040 LCA gives BEHVR a lifecycle footprint of 9.10 kg CO₂eq/m², far below aluminum panels (28.4) and glass curtain walls (35.7). The LCOA model places BEHVR's 835-1080 RMB/m² cost on the Pareto frontier across all three axes.

"BEHVR 在 LCOA / 美学耐久 / 碳成本三维度同时占据帕累托最优区域。" "BEHVR is Pareto-optimal across LCOA, aesthetic durability, and carbon cost."
— DISSERTATION CH.6, p.172
08
CHAPTER EIGHT · 结 论 与 展 望CONCLUSIONS

六大创新与五个未来方向Six Innovations & Five Future Directions

Theory → data → algorithm → process → material → industrialization — and where the next wave of research opens.

KEY INSIGHT
六大创新覆盖理论、数据、算法、工艺、材料、产业化六个层面——这种「全栈创新」在博士论文中极为罕见，也是公司技术护城河的真正来源。 Six innovations span theory, data, algorithm, process, materials, and industrialization — a "full-stack innovation" rare in dissertations, and the genuine source of the company's technical moat.
CONTRIBUTION / 贡 献C

完整的「理论 → 实验 → 工艺 → 产业 → 量产」五层级递进体系。A full theory → experiment → process → industry → mass-production cascade.

LIMITS / 局 限L

数据集气候带覆盖、透射型设计、量子点真实 25 年稳定性、产线良率提升。Limited climate coverage, transmissive coatings, real-world QD durability, line yield improvement.

FUTURE / 未 来F

多机制协同 / AI 工艺生成 / 无铅替代 / 全光谱量产 / 社会经济影响。Multi-mechanism synergy / AI process generation / Pb-free alternatives / full-spectrum mass production / social-economic impact.

本章总结六大研究结论：(1) 揭示工业感的物理本质；(2) 建立 FBFV 统一框架；(3) 构建 BgDB + BgClassNet 96.4% 准确率；(4) 五系列涂层 92.3% 保留率；(5) 梯度 PL-EVA 通过 PVQAT-TG3 双倍加速衰减；(6) 三层产业化系统帕累托最优。 Six research conclusions: (1) physics origin of "industrial look" identified; (2) FBFV unified framework established; (3) BgDB + BgClassNet at 96.4%; (4) 5-color coatings retain 92.3%; (5) gradient PL-EVA passes PVQAT-TG3 doubled-stress; (6) three-layer industrialization Pareto-optimal.

研究局限性诚实披露：BgDB 数据集主要覆盖国内 4 省，海洋性 / 热带湿热气候数据不足；薄膜涂层主要针对反射型，透射型 BIPV 设计未覆盖；钙钛矿真实户外 25 年稳定性需实地长期监测；产线良率提升与跨气候带应用扩展仍待实践检验。 Honest limitations: BgDB covers 4 provinces, missing maritime and tropical climates; coatings target reflective only, not transmissive BIPV; real-world 25-year QD durability still needs field validation; mainnet, liquidity bootstrap, and DAO transition remain unproven in practice.

四个未来研究方向：(1) 多机制协同——钙钛矿叠层 + QD + 薄膜协同；(2) 智能化——基于 BgDB 的扩散模型直接生成工艺参数；(3) 材料——CsSnI₃ 等无铅替代方案；(4) 社会经济——量化对中国千万吨级退役潮的政策协同效应。 Four future directions: (1) multi-mechanism synergy — perovskite tandem + QD + coatings; (2) AI-driven process generation via diffusion models trained on BgDB; (3) Pb-free materials (CsSnI₃ etc.); (4) quantified social-economic impact at China's tens-of-millions-of-tons scale.

「从物理基础、到工艺与产业化的完整理论与技术体系。」 "A complete pathway from fundamental physics to industrial deployment."
— DISSERTATION CH.8, p.207
// v3 RIGOR SYNCHRONIZATION

严苛性修订徽章 · 13 项验证全部通过Rigor Audit · 13/13 Verifications Passed

论文 v3 (v2026.5-rev) 完成全套严苛性修订。所有数据声明背后都有统计学严密性、物理学严密性、文献严密性、方法学严密性作为支撑。13 项修订要点已同步至本项目所有对外资产——白皮书、IPO 计划书、答辩 PPT、研究报告、官网。Thesis v3 has completed all rigor revisions. Every data claim is backed by statistical, physical, literature, and methodological rigor. All 13 revision items have been synchronized across every external asset — whitepaper, IPO plan, defense slides, research report, this website.

A1-A6 · 6 critical
关键漏洞Critical fixes
范畴 / SiNx / 统计 / 偏倚 / Brus / PVQATScope / SiNx / Stats / Shift / Brus / PVQAT
B1-B7 · 7 moderate
中等问题Moderate fixes
综述法 / TMM / PLQY / 批次 / 文献 / 敏感性 / JoannopoulosSurvey / TMM / PLQY / Batch / Refs / Sensitivity / Joannopoulos
D · structural
结构性根治Structural fix
§1.6 学位论证范畴声明§1.6 Thesis scope declaration
致信索取 9 万字论文全文Request the full 90,000-word thesis ↗
盈利模型THE REVENUE ENGINE

毛利，
来自那道间隙。 Margin lives
in the gap.

收入沿价值阶梯逐层上移——从制造驱动，到项目驱动，再到设计与运维驱动。营收规模增长的同时，高毛利的设计与运维订阅层占比上升，估值倍数随之跃升。 Revenue climbs the value ladder — from manufacturing-led, to project-led, to design-and-operations-led. As scale grows, the high-margin design and SaaS subscription layers take a larger share, and the valuation multiple jumps with them.

盈利模型 · Revenue engine Three stacked bars (manufacturing-led, project-led, platform-led) showing revenue growing and shifting from product and project layers toward SaaS and platform layers, with the estimated valuation multiple rising from 8x to over 30x. VALUATION MULTIPLE · 估值倍数 8× 14× 18×+ STAGE 01 Manufacturing-led 制造驱动 M0–18 STAGE 02 Project-led 项目驱动 M18–36 STAGE 03 Platform-led 平台驱动 M36–60
产品Products再制造彩色组件Remanufactured colored modules
项目Projects场景交钥匙工程Turnkey scenario builds
设计Design方案与集成Solutions & integration
运维 SaaSOps SaaS数字孪生订阅Digital-twin subscription
盈利的灵魂The soul of the margin
毛利来自废料价与美学资产价之间的那道间隙——与 logo 同一个隐喻。价值由「在波长尺度上重新组织物质、注入信息」创造，而非靠熔回原料。Margin lives in the gap between scrap price and aesthetic-asset price — the same metaphor as the logo. Value is created by reorganizing matter and injecting information at the wavelength scale, not by melting it back to raw material.
06 / 路 径THE PATH

从中试到上市，
60 个月的路径。 60 months from pilot
to public listing.

M0 — M6 / FOUNDATION

奠基期Foundation

首面能带间隙样板墙公开揭幕。RMB 1000 万天使轮，5000 万 pre-money。6 人核心团队搭建。First bandgap wall publicly unveiled. RMB 10M angel close at 50M pre-money. 6-person core team operational.

M7 — M24 / VALIDATION

验证期Validation

3-5 个商业项目落地。Pre-A 轮 2 亿估值。8 篇 SCI、12 项专利进入实质审查。3–5 commercial projects on the ground. Pre-A round closed at 200M valuation. 8 SCI papers, 12 patents in active examination.

M25 — M48 / REPLICATION

复制期Replication

材料 / EPC / 平台三主体并行。A+B 轮共 3-5 亿。M48 触发 30× 估值倍数。Material / EPC / Platform companies established. A+B rounds total 300–500M. 30× valuation multiple unlocked at M48.

M49 — M60 / CAPITALIZATION

资本化期Capitalization

提交港交所主板申报。Pre-IPO 5-10 亿。上市估值 30-80 亿。天使回报 60-160×。Hong Kong main board IPO submission. Pre-IPO 500M – 1B. Listing valuation 3-8B RMB. Angel return 60–160×.

戴

问题不是我们应该把面板染成什么颜色。问题是为什么硅的能带间隙一定是 1.12 eV。 The question is not what color we should paint these panels. The question is why silicon's bandgap has to be 1.12 eV.

创始人 兼 首席科学家FOUNDER & CHIEF SCIENTIST

戴继涛 博士

香港能带间隙量子科技有限公司Hong Kong Bandgap Quantum Technology Co., Ltd.

5y
五年退役光伏能带间隙工程博士研究Doctoral research on bandgap engineering for retired PV
30+
商业计划 30+ 章 · 5 份执行规划Business plan chapters across 5 execution playbooks
9
18 份对外文档 · 1 部视频 · 1 个交互式知识图谱 · 3 战略视觉18 documents · 1 video · 1 knowledge graph · 3 strategic visuals
合作共赢TOGETHER WE BUILD

能带间隙
不是一个人的实验室。 Not a one-scientist
laboratory.

我们在寻找愿意把这件事做成行业标准的人——既包括加入团队的早期合伙人，也包括从上游硅片、下游建筑文旅、退役长约、标准制定到学术发表的外部战略合作方。下面是我们公开的「合作类型清单」，按战略章程「四不做」校验。 We are looking for people who want to turn this into an industry standard — both founding partners joining the team, and strategic external collaborators across upstream silicon, downstream architecture, end-of-life supply, standards bodies and academic publication. Below is our public partnership menu, screened against the Four Prohibitions in our Strategy Charter.

I
内部合伙人 · 加入团队INNER · Join the Founding Team
联合创始人、核心团队、战略天使与导师网络Co-founders, early team, strategic angels & advisors
天使阶段 4 人 → Pre-A 12 人 → A 轮 30+ 人。下面四类位置目前开放或近期开放，请直接与戴博士对话。From a 4-person seed team to ~12 at Pre-A to 30+ at Series A. The four positions below are open now or opening soon — talk directly to Dr. DAI.
CO-FOUNDER · CTO
技术联合创始人（材料 / 工艺方向）Technical Co-founder (Materials / Process)
半导体材料 / 薄膜涂层 / 量子点 / TMM 建模背景；带宜兴中试线、领导工艺与质控；与戴博士共建技术决策权。M3–M6 入职。Semiconductor materials / thin-film coatings / quantum dots / TMM modelling background; lead the Yixing pilot line and process & QC; share technical decision rights with Dr. DAI. Joins M3–M6.
画像：博士级 + 头部 PV / 显示 / 涂层企业 5 年以上经验Profile: doctoral degree + 5+ yrs at a top PV / display / coatings company
CO-FOUNDER · BD
商业拓展联合创始人Co-founder of Business Development
高端商业地产 / 文旅 / 政府客户网络；带 3 个灯塔签约；M0–M3 入职。Deep network in premium commercial real estate / cultural tourism / government clients; brings 3 lighthouse signings; joins M0–M3.
画像：头部地产 / 文旅集团高级总监 8 年以上经验，或大型央企政府事务背景Profile: 8+ yrs senior director at top developer / cultural-tourism group
EARLY TEAM
核心团队（CFO / 数据 / 产线 / 国际市场）Early Team (CFO / Data / Line / International)
数据科学家（BgDB + BgClassNet 升级、JMCA 发表）、产线工程师（薄膜 + 检测）、CFO（双运行 IPO 准备）、国际市场负责人（HK / 中东 / 欧盟）。Pre-A 至 A 轮分批入职。Data scientist (BgDB + BgClassNet, JMCA paper), line engineers (coatings + inspection), CFO (dual-IPO prep), international BD (HK / Middle East / EU). Joining Pre-A through Series A.
期权与现金组合 · 与阶段、岗位、贡献挂钩Equity + cash · staged by phase, role & contribution
STRATEGIC ANGELS
战略天使个人 / 战略产业方Strategic Angels & Industry Partners
能带来技术、客户、政策、海外网络其中之一的有耐心战略个人；或材料 / 能源 / 地产 / 建筑产业方 CVC。明确不接：纯财务套利、需要快速退出、要求品牌联动的资本。Patient strategic individuals bringing tech, clients, policy, or overseas networks; or CVCs from materials / energy / real-estate / architecture. Not accepted: pure financial arbitrage, fast-exit capital, brand-tying conditions.
天使 50M post · Pre-A 200–400M · A 轮 1–3 亿Seed @ 50M post · Pre-A 200–400M · A 1–3 亿
SCIENTIFIC ADVISORS
学术导师网络（中科院体系 + 港校）Scientific Advisors (CAS & HK universities)
中科院深圳先进院光伏 / 量子点 / 钙钛矿研究员；港大 / 港科大 / 港中文材料与建筑环境教授；联合 SCI 发表、共建 BgDB 数据扩展、共申标准。Researchers from SIAT-CAS in PV / quantum-dot / perovskite; HKU / HKUST / CUHK professors in materials and built-environment; joint SCI publications, BgDB co-development, joint standards drafting.
权益：联合署名 · 研究经费支持 · 期权激励（按贡献）Co-authorship · research-grant support · contribution-based equity
INDUSTRY MENTORS
产业导师团（PV / BIPV / 建筑业）Industry Mentors (PV / BIPV / Architecture)
头部 PV 厂前高管、知名建筑师事务所合伙人、BIPV 工程顾问；以 Mentor 身份提供产业 know-how、客户引荐、行业人脉。Former senior execs of top PV manufacturers, partners at renowned architecture firms, BIPV engineering consultants — providing industry know-how, client introductions and ecosystem reach.
每季度 1–2 次圆桌 · 关键事件咨询 · 期权激励1–2 round-tables per quarter · critical-event consults · equity
II
外部战略合作OUTER · Strategic Partnerships
从上游硅片到下游建筑、从退役长约到国标制定、从国内到国际From upstream silicon to downstream architecture, from end-of-life supply to standards, from domestic to international
下方五类合作方均来自 18 号战略发展规划中明确点名的目标对象，公开发布以表达诚意——若你或你认识的人来自这些类别，请直接联系戴博士。All five categories below are named in our Strategic Plan v1.0. Publishing them openly signals our intent — if you or someone you know is from one of these categories, please reach Dr. DAI directly.
01 · UPSTREAM PV
上游 c-Si 头部锚定方Upstream c-Si Anchor Partner
提供硅片产能 + 全球渠道；我们带「无损彩色」差异化产品。形式：战略股东 + 联合产品 + 共申 BIPV 国际标准。Provides silicon-wafer capacity + global channels; we bring the differentiated "lossless color" product line. Form: strategic shareholder + co-product + joint BIPV international standard.
画像：全球出货量前 5 的硅基组件厂商 · 钙钛矿-硅叠层创新者Profile: top-5 global silicon module makers · perovskite-Si tandem innovators
02 · DOWNSTREAM ARCHITECTURE
下游建筑 / 文旅地产 / 建筑师圈层Downstream Real Estate / Cultural Tourism / Architects
提供项目场景与建筑师指定权；我们提供 BIPV「无损彩色」材料系统。形式：项目级合作 + 灯塔案例 + 联合标准。Provides project sites and architect specification; we provide the BIPV "lossless color" material system. Form: project-level collaboration + lighthouse cases + joint standards.
画像：国内 Top 10 商业地产开发商 · Top 5 文旅集团 · 国际知名建筑师事务所Profile: top-10 domestic developers · top-5 cultural-tourism groups · world-class architecture studios
03 · EoL c-Si SUPPLY
退役 c-Si 长约供应方End-of-Life c-Si Long-Term Suppliers
提供 10 年期定向退役 c-Si 组件优先采购权；我们以战略股东 / 优先合作方身份分享再制造产业增值。Provide 10-year-term priority offtake of decommissioned c-Si modules; we offer strategic-shareholder / priority-partner status in the remanufacturing value chain.
画像：装机量 Top 10 的央/国企新能源运营商Profile: top-10 SOE renewable-energy operators by installed capacity
04 · STANDARDS BODIES
标准制定机构Standards Bodies
19 号《光伏组件干涉型彩色化技术规程》起草大纲已就位，等待与标准机构正式接触。Document #19 — the draft outline for GB/T "Interference-Type Colorization of PV Modules" — is ready for formal engagement with standards bodies.
合作机构：CESI · 中建科 · SAC/TC562 · IEC TC82 · ISO TC163（公开标准机构，可直接接触）Bodies: CESI · CABR · SAC/TC562 · IEC TC82 · ISO TC163 (public standards bodies)
05 · ACADEMIC & JOURNALS
学术与期刊合作Academic Institutions & Journals
联合实验室、共建论文、共申标准。戴博士 BgDB v2 + BgClassNet 升级版准备 M12 前后投稿。Joint laboratories, co-authored papers, joint standards. Dr. DAI's BgDB v2 + BgClassNet upgrade is targeted for submission around M12.
画像：中科院体系研究员 · 港校教授 · 高水平期刊（JMCA / ACS EL / Nature 系）Profile: CAS researchers · HK university professors · high-impact journals
06 · INTERNATIONAL (Phase 2+)
国际项目方（Phase 2 启动）International Project Owners (Phase 2)
HK 主体的天然窗口——中东、欧盟、东南亚的高端 BIPV 与历史建筑改造项目方。Natural channels for our HK entity — premium BIPV and historic-building retrofit project owners in the Middle East, EU, and Southeast Asia.
画像：中东 / 欧盟 / 东南亚高端 BIPV 业主 · 建筑遗产改造项目方Profile: premium BIPV owners in ME / EU / SE Asia · heritage-retrofit project owners
联系方式How to reach us
所有合作类型——无论你是想加入团队的同行者，还是想战略入股的产业方，或是想共建标准的学术机构——请直接对话戴博士，由他亲自回复。For all partnership types — joining the team, strategic investment, or co-developing standards — please reach Dr. DAI directly. He replies personally.
FOUNDER · CEO戴继涛 博士 · Dr. DAI Jitao
EMAILdaiyile396@gmail.com · founder@bandgap.hk (M3 启用)
HKHong Kong Bandgap Quantum Technology Co., Ltd.
WECHAT扫码或搜索「DAI_Jitao_Dr」QR / ID: DAI_Jitao_Dr
内部合伙人请致信 For internal partner candidates, please email hi@bandgap.hk 索取「内部合伙人申请表」（约 30–60 分钟填写，含与决策五问的对齐自答），完成后回传。 to request the Internal Partner Application Form (30–60 min, includes alignment with the Five Questions), then send back to the same address.
窗口只开一次THE WINDOW OPENS ONCE

退役潮始于 2026。
与我们一起构建。 The wave starts in 2026.
Build with us.

资本、人才、产业——无论你带来哪一边，能带间隙这件事正在被一块一块地工程化。 Capital, talent, industry — whichever side you bring, the bandgap is being engineered one module at a time.

致信约谈Email to Meet→ 学术研究Academic Research↗
标识含义THE MARK
能带 × 间隙BAND × GAP
一条能带闭合成环，顶部留下一道间隙——这正是公司的名字「能带间隙」。环身自顶部的青（导带）渐变到底部的金（价带）。An energy band closed into a ring, with a single gap at the top — literally the company's name, the bandgap. The ring fades from cyan (conduction) at the top to gold (valence) at the bottom.
量子跃迁THE QUANTUM LEAP
间隙处那颗发光的电子正在跃迁——光与物质相互作用的本征瞬间，也是价值被创造的起点：一块退役组件重新被「点亮」。The glowing electron at the gap is mid-leap — the fundamental instant of light–matter interaction, and where value begins: a retired module brought back to light.
闭环再造THE CLOSED LOOP
环亦是循环经济：退役光伏经能带工程重获价值，重新进入建筑、能源与公共空间，形成不断回流的价值闭环。The ring is also the circular economy: decommissioned PV regains value through bandgap engineering and returns to buildings, energy systems, and public spaces — a value loop that keeps returning.
纪律THE DISCIPLINE

纪律，
就是护城河。 Discipline is
the moat.

能带间隙最终的核心护城河，不是某一项专利、某一份数据集、某一段代码，而是把同一件事在波长尺度上做透的纪律本身。下面是我们公开的部分——也是检验我们是否对得起它的方式。 Our deepest moat is not a single patent, dataset, or codebase. It is the discipline of doing the same thing — engineering value at the wavelength scale — over and over. Below is the public part of that discipline, and the way to check whether we are living up to it.

「在能带之间，把废料变成可投资的光谱。」BRAND AXIS · 品牌主轴句 "Between the bands, we turn waste into investable spectrum."BRAND AXIS · 品牌主轴句

四不做Four Prohibitions
01
不做横向平台 / 孵化器No horizontal platform or incubator
跨行业平台业务交独立主体承载，不进能带间隙品牌。Cross-industry platform work belongs in a separate entity — not under this brand.
02
不做硬科技之外的副业No side-businesses outside deep tech
任何延伸都必须先服务于核心材料与工艺价值链。Any extension must first serve the core materials & process value chain.
03
不做泛能源 / 泛 ESG 包装No generic energy or ESG packaging
护城河是材料 + 工艺 + 数据三锚，不是热词。The moat is the material × process × data triple-anchor, not buzzwords.
04
不做「通用企业家」创始人故事No generic-entrepreneur founder story
戴继涛 博士的科学家身份是品牌不可让渡的核心资产。The scientist-CEO identity of Dr. DAI Jitao is non-transferable.
决策五问The Five Questions
1
是否在波长尺度上创造价值？Does it create value at the wavelength scale?
2
是否会稀释「科学家-CEO」的创始人品牌？Will it dilute the scientist-CEO founder brand?
3
是否要求放弃「四不」中任何一条？Does it require breaking any of the Four Prohibitions?
4
能否在不抢主问题资源的前提下推进？Can it proceed without taking resources from the current main question?
5
它的失败是否会污染公司核心叙事？Would its failure contaminate the company's core narrative?
使用方式How to use this
任何提案——新业务、新合作、新场景、新创始人故事——落笔前先比对「四不」与「决策五问」。任一不过，原则上不应推进。Any proposal — new business, new partnership, new scenario, new founder story — must be checked against the Four Prohibitions and the Five Questions before being pursued. Failing any one is, by default, a veto.