从原子,到江河。From atoms, to rivers.

接下来我们按顺序讲:这个人是谁、问题到底是什么、一百多年的接力赛、他们的巧招、熵增的谜底、这项工作的边界在哪里,最后回答一个更"飘"的问题——它对"宇宙本身就是数学"这个猜想,有没有影响。What follows takes these in order: who this person is, what exactly the problem is, the century-long relay race, the clever trick they found, the answer to the entropy puzzle, where the boundaries of this work lie, and finally a more airy question — whether it has any bearing on the conjecture that "the universe itself is mathematics".

主角邓煜,1989 年生,在深圳长大。他的履历几乎是"数学少年"的标准样本:高二在中国数学奥林匹克(CMO)拿到满分金牌、入选国家集训队;2006 年,16 岁的他代表中国参加第 47 届国际数学奥林匹克(IMO),获金牌(35/42 分,并列世界第六)。2007 年保送北京大学数学科学学院——巧的是,与后来同届获菲尔兹奖的王虹正是同一级的同学。2009 年他转学到麻省理工学院(MIT),2010 年拿下美国大学生数学竞赛最高荣誉 Putnam Fellow,2011 年本科毕业后进入普林斯顿大学,师从 Alexandru Ionescu,2015 年博士毕业并获普林斯顿授予研究生的最高荣誉 Jacobus Fellowship。此后:纽约大学 Courant 研究所博士后(2015–2018)→ 南加州大学任教(2018)→ 斯隆研究奖(2021)→ 2024 年加入芝加哥大学任正教授。The protagonist, Yu Deng, was born in 1989 and grew up in Shenzhen. His résumé is almost the textbook sample of a "mathematical prodigy": in his second year of high school he took a perfect-score gold medal at the Chinese Mathematical Olympiad (CMO) and was selected for the national training team; in 2006, aged 16, he represented China at the 47th International Mathematical Olympiad (IMO) and won gold (35/42 points, tied for sixth in the world). In 2007 he was admitted without examination to the School of Mathematical Sciences at Peking University — and, as it happens, was in the same year group as Hong Wang, who would later share the Fields Medal with him. In 2009 he transferred to the Massachusetts Institute of Technology (MIT); in 2010 he took the highest honour in the American collegiate mathematics competition as a Putnam Fellow; after graduating in 2011 he went to Princeton University to study under Alexandru Ionescu, finishing his doctorate in 2015 and receiving the Jacobus Fellowship, Princeton's highest honour for a graduate student. After that: a postdoc at NYU's Courant Institute (2015–2018) → a faculty post at the University of Southern California (2018) → a Sloan Research Fellowship (2021) → joining the University of Chicago as a full professor in 2024.

2026 年 7 月 23 日,费城国际数学家大会开幕式上,邓煜获菲尔兹奖——四年一评、限 40 岁以下、每届至多四人,常被称为"数学界的诺贝尔奖"。他与同届得主王虹是菲尔兹奖九十年历史上首批在中国大陆出生并成长的获奖者(此前的华人得主是 1982 年的丘成桐与 2006 年的陶哲轩)。在此之前,这项工作已连续摘下 ICCM 数学奖金奖(2025)、克雷研究奖与美国数学会 Eisenbud 数学物理奖(2026)。On 23 July 2026, at the opening ceremony of the International Congress of Mathematicians in Philadelphia, Deng was awarded the Fields Medal — given every four years, restricted to those under 40, at most four per congress, and often called "the Nobel Prize of mathematics". He and fellow laureate Hong Wang are the first recipients in the medal's ninety-year history to have been born and raised in mainland China (the earlier ethnically Chinese laureates were Shing-Tung Yau in 1982 and Terence Tao in 2006). Before that, this work had already collected the ICCM Gold Medal of Mathematics (2025), a Clay Research Award, and the American Mathematical Society's Eisenbud Prize for Mathematics and Physics (2026).

但这不是一个人的战役。解决问题的是一个三人组:But this was not one person's battle. The problem was solved by a trio:

据合作者回忆,证明最紧张的阶段,三人几乎每天开视频会议,常常从深夜聊到凌晨——"我们不断卡住",再一点点凿开。His collaborators recall that during the most intense stretch of the proof the three of them held video calls almost every day, often talking from late at night until the small hours — "we kept getting stuck", and then chipped their way through bit by bit.

1900 年,数学家大卫 · 希尔伯特在巴黎提出了著名的 23 个问题,为整个二十世纪的数学定调。其中第六问题与众不同——它不是一道具体的题,而是一个纲领:把物理学建立在严格的数学公理之上。希尔伯特特别点名了一条路线:沿着玻尔兹曼的工作,把"从原子观点到连续介质运动定律"的极限过程严格化。In 1900, the mathematician David Hilbert put forward his famous list of 23 problems in Paris, setting the agenda for mathematics across the whole twentieth century. Among them, the sixth problem stands apart — it is not a specific exercise but a programme: to build physics upon rigorous mathematical axioms. Hilbert named one route in particular: following Boltzmann's work, make rigorous the limiting process leading "from the atomistic view to the laws of motion of continua".

为什么这重要?看下面这张图——同一团气体,三种"看"法:Why does this matter? Look at the diagram below — one and the same body of gas, seen in three ways:

难点在哪?微观有天文数字个变量(1 升空气约有 10²² 个分子,每个 6 个变量),玻尔兹曼方程却只用一个函数 f 概括全部。信息被暴力压缩,凭什么压缩后的方程还是对的?这一"凭什么",数学家答了 125 年。Where is the difficulty? The microscopic picture has an astronomical number of variables (one litre of air holds about 10²² molecules, six variables each), yet the Boltzmann equation sums it all up in a single function f. The information is brutally compressed — so why should the compressed equation still be right? Mathematicians have spent 125 years answering that "why".

敌人:带记忆的重逢The enemy: a reunion that remembers

玻尔兹曼方程有一个隐藏前提,叫分子混沌假设:任意两颗即将相撞的分子,彼此是"陌生人"——速度互不相关。这就是那次暴力信息压缩合法的前提。可牛顿力学偏偏爱制造熟人:A 撞了 B,两者的速度从此纠缠;若它们(或它们各自撞出去的"后代")日后再次相遇,就不是陌生人了。这种事件叫重碰撞。时间越长,重逢的机会越多,"陌生人假设"就越可疑——兰福德的证明正是死在这里:超过约五分之一次碰撞的时间,他就无法排除重逢的干扰。The Boltzmann equation rests on a hidden premise called the molecular chaos assumption (Stosszahlansatz): any two molecules about to collide are "strangers" to each other — their velocities are uncorrelated. That is exactly what makes the brutal compression of information legitimate. But Newtonian mechanics loves manufacturing acquaintances: once A has hit B, their velocities are entangled from then on; and if they (or the "descendants" they each knock away) meet again later, they are no longer strangers. Such an event is a recollision. The longer the time, the more chances there are for reunions, and the more suspect the "stranger assumption" becomes — this is precisely where Lanford's proof died: beyond about one fifth of a collision time, he could no longer rule out the interference of reunions.

兵器一:碰撞家谱图Weapon one: the collision genealogy graph

他们把每颗分子的完整碰撞史画成一张图:点是碰撞事件,线是分子的轨迹——很像物理学家算粒子相互作用时用的费曼图。一段时间内所有可能的历史,对应天文数字张不同形状的图。重碰撞在图里有清晰的长相:它让图出现"环"(回路)。于是"证明重碰撞稀少"变成一个可以下手的组合问题:证明带环的图贡献极小。They drew the complete collision history of every molecule as a graph: the vertices are collision events and the edges are molecular trajectories — much like the Feynman diagrams physicists use to compute particle interactions. All possible histories over a stretch of time correspond to an astronomical number of differently shaped graphs. A recollision has a clear signature in such a graph: it makes a "cycle" (a loop) appear. So "proving that recollisions are rare" turns into a combinatorial problem one can actually attack: proving that graphs containing cycles contribute almost nothing.

兵器二:累积量记账法Weapon two: cumulant bookkeeping

"分子混沌"不可能严格成立——总有一点点没忘干净的相关性。与其假装它不存在,不如专门给它建账本:累积量(cumulant)就是这样一个数学对象,精确记录"偏离完全独立的那一小部分"。三人证明了一个长时间累积量拟设:这本"残余记忆账"随时间推移始终保持极小,可以一段一段地向前传递(把总时长切成许多小段,逐段证明账本不失控)。"Molecular chaos" can never hold exactly — there is always a little correlation that has not quite been forgotten. Rather than pretend it does not exist, they opened a dedicated ledger for it: a cumulant is precisely such a mathematical object, recording exactly "the small part by which things deviate from complete independence". The trio proved a long-time cumulant ansatz: this ledger of residual memory stays extremely small as time goes on, and can be propagated forward stretch by stretch (cutting the total duration into many short intervals and proving interval by interval that the ledger never runs out of control).

兵器三:切割算法Weapon three: the cutting algorithm

剩下的困难是纯技术但极其凶残的:那些巨大的家谱图如何估计?他们设计了一套切割算法——像剪纸一样,把一张庞大的图按精心设计的规则切成小块,每一块都能单独估出大小,而切法保证"环越多(重碰撞越多)的图,总贡献越小"。这是两篇论文中最核心、也最艰苦的部分:找到正确的切法,靠的是"数月的卡壳与深夜视频会"。The remaining difficulty is purely technical and utterly brutal: how do you estimate those enormous genealogy graphs? They designed a cutting algorithm — like paper-cutting, slicing a huge graph into small pieces according to carefully designed rules, so that each piece can be bounded on its own, while the way you cut guarantees that "the more cycles a graph has (the more recollisions), the smaller its total contribution". This is the core and the most gruelling part of the two papers: finding the right way to cut took "months of getting stuck and late-night video calls".

三件兵器合璧,结论是:只要玻尔兹曼方程的解存在,微观硬球系统就在极限下精确地跟随它——不限时长。再叠加此前已发展成熟的"玻尔兹曼 → 流体"极限理论(适当调节参数令碰撞频率 α→∞),整条链就通了。值得一提的是,这套打法不是为粒子量身定做的孤例:它先在波湍流里验证过一遍——同一种数学,统一了"粒子的气体"与"波的气体",这本身就暗示它触到了某种更普遍的结构。With the three weapons combined, the conclusion is: for as long as a solution of the Boltzmann equation exists, the microscopic hard-sphere system follows it exactly in the limit — with no time restriction whatsoever. Layer on top of that the already mature "Boltzmann → fluid" limit theory (tuning the parameters so that the collision frequency α→∞), and the whole chain is complete. It is worth noting that this line of attack was not a one-off tailored to particles: it had first been tested in wave turbulence — the same mathematics unifying "a gas of particles" and "a gas of waves", which in itself hints that it has touched some more universal structure.

佯谬本身The paradox itself

拍一段两颗分子相撞的视频,倒着放,依然完全符合牛顿定律——你根本看不出正反。但拍一段墨水滴入清水的视频,倒放立刻穿帮:散开的墨绝不会自己收回一滴。玻尔兹曼方程站在宏观一侧:它的 H 定理说,熵(可以粗略理解为"混乱程度"或"还能再散开的余地")只增不减。洛施密特 1876 年的质问由此而来:从完全可逆的原料,怎么可能推导出不可逆的结论?要么推导有诈,要么某处偷偷塞进了方向。Film two molecules colliding, play it backwards, and it still obeys Newton's laws perfectly — you cannot tell which way round it runs. But film a drop of ink falling into clear water and the reversed version gives itself away instantly: dispersed ink never gathers itself back into a drop. The Boltzmann equation stands on the macroscopic side: its H-theorem says that entropy (loosely, "the degree of disorder", or "how much room there still is to spread out") only increases and never decreases. Hence Loschmidt's challenge of 1876: how can an irreversible conclusion possibly be derived from perfectly reversible ingredients? Either the derivation is a cheat, or a direction has been smuggled in somewhere.

谜底的三个零件Three parts to the answer

其一,起点是特殊的。定理假设初始时刻分子近似独立地随机撒开——一个"有序、低相关"的开局。就像一副刚按花色理好的扑克:定律(洗牌规则)本身不偏心,但从有序开局出发,几乎必然越洗越乱;"越洗越整齐"并非被定律禁止,只是概率小到宇宙年龄内看不到。First, the starting point is special. The theorem assumes that at the initial moment the molecules are scattered at random, approximately independently — an opening that is "ordered and weakly correlated". It is like a deck of cards freshly sorted by suit: the law (the shuffling rule) plays no favourites, but starting from an ordered opening the deck almost inevitably grows more disordered with every shuffle; "shuffling into ever tidier order" is not forbidden by the law, its probability is merely so small that you would never see it within the age of the universe.

其二,方向是统计的,不是逐一保证的。定理说的是:对几乎所有随机初始配置,系统跟随玻尔兹曼方程(从而熵增)。确实存在精心构造的"倒放"初始条件能让熵暂时下降——但它们在所有可能配置中占的比例,随分子数暴增而指数式消失。Second, the direction is statistical, not guaranteed case by case. What the theorem says is: for almost all random initial configurations, the system follows the Boltzmann equation (and hence entropy increases). Carefully constructed "played-backwards" initial conditions that make entropy dip temporarily really do exist — but their share of all possible configurations vanishes exponentially as the number of molecules explodes.

其三,也是最深的一件:记忆去了哪里?信息并没有被销毁——牛顿力学不销毁信息。它是被流放了:每次碰撞把"谁曾遇见谁"的记录推入越来越高阶、涉及越来越多分子的关联之中。而玻尔兹曼方程只看单分子的分布 f,那些高阶关联对 f 的影响,正是三人的累积量账本所控制的对象——他们证明它始终微小,并且在 N→∞ 的极限下一去不返。不可逆性,就是"信息流向了你永远不再查看的账户"。他们在论文中专辟一节,用这一显式结构解释时间之箭如何在方程中涌现。Third, and deepest of all: where did the memory go? The information was not destroyed — Newtonian mechanics destroys no information. It was exiled: every collision pushes the record of "who once met whom" into correlations of ever higher order, involving ever more molecules. The Boltzmann equation, meanwhile, looks only at the single-molecule distribution f, and the effect of those high-order correlations on f is exactly what the trio's cumulant ledger controls — they proved it stays minute, and that in the N→∞ limit it never comes back. Irreversibility is simply "information flowing into an account you will never look at again". In their paper they devote a dedicated section to explaining, through this explicit structure, how the arrow of time emerges within the equations.

物理得到地基。玻尔兹曼方程与流体方程是航空、气象、半导体、等离子体等无数工程计算的日常工具,一百多年来"好用"从未被怀疑,但它们与更基本定律之间的逻辑链条一直悬空。现在,在稀薄硬球气体这一经典模型内,这条链条闭合了:统计力学的核心方程不再是"经验上极其成功的假设",而是牛顿力学的数学推论。可逆性佯谬也随之在模型内获得闭环解答。Physics gains a foundation. The Boltzmann equation and the fluid equations are everyday tools in countless engineering calculations — aviation, meteorology, semiconductors, plasmas — and for more than a century nobody has doubted that they work; but the logical chain linking them to more fundamental laws was left dangling. Now, within the classical model of a dilute hard-sphere gas, that chain has been closed: the central equations of statistical mechanics are no longer "an empirically wildly successful hypothesis" but a mathematical consequence of Newtonian mechanics. The reversibility paradox thereby receives a closed-loop answer inside the model as well.

数学得到兵器与信心。兵器,是那套"图 + 累积量 + 切割"的通用打法——它已经在两个截然不同的战场(波湍流、粒子气体)各赢一次,数学家有理由期待它能推广到更多"从微观推导宏观方程"的问题上(不同的相互作用势、其他动理学方程等)。信心,则是示范意义:一个悬置五十年、许多人认为本世代无望的问题,被"换一个相邻战场先练兵、再回师主攻"的路线攻克——这对整个领域的士气与方法论都是一针强心剂。顺带一提,2025–2026 年间,三维挂谷猜想(王虹)与本问题接连告破,被不少数学家视为分析学的一个丰收期。Mathematics gains weapons and confidence. The weapons are that general "graphs + cumulants + cutting" line of attack — it has already won on two completely different battlefields (wave turbulence and particle gases), and mathematicians have reason to expect it to extend to many more problems of "deriving macroscopic equations from microscopic ones" (other interaction potentials, other kinetic equations, and so on). The confidence lies in the example it sets: a problem left hanging for fifty years, which many thought hopeless for this generation, was cracked by the strategy of "training first on a neighbouring battlefield, then wheeling back for the main assault" — a shot in the arm for the morale and the methodology of the whole field. Incidentally, in 2025–2026 the three-dimensional Kakeya conjecture (Hong Wang) and this problem fell one after the other, which many mathematicians regard as a bumper harvest for analysis.

公允的总结是:在希尔伯特明确写下的那条路线上,这项工作是终点;在希尔伯特心中更宏大的蓝图上,它是迄今最坚实的一段路基。两种说法都对,取决于你把"第六问题"的边界画在哪里——而这恰是学界当下真实的讨论状态。A fair summary: along the route Hilbert wrote down explicitly, this work is the terminus; on the grander blueprint Hilbert had in mind, it is the most solid stretch of roadbed laid so far. Both statements are correct, depending on where you draw the boundary of "the sixth problem" — and that is precisely the state of the real discussion in the field today.

先厘清概念。物理学家马克斯 · 泰格马克提出的数学宇宙猜想(Mathematical Universe Hypothesis, MUH)主张:物理实在不只是"能用数学描述",而是本身就是一个数学结构——宇宙与某个数学对象之间没有剩余的、非数学的"质料"。它还有一个更激进的推论:所有自洽的数学结构都同等地"物理存在"。First, some conceptual clarity. The Mathematical Universe Hypothesis (MUH), proposed by the physicist Max Tegmark, claims that physical reality is not merely "describable by mathematics" but is itself a mathematical structure — that between the universe and some mathematical object there is no leftover, non-mathematical "stuff". It has a still more radical corollary: all self-consistent mathematical structures "physically exist" on an equal footing.

支持方向:为什么有人会说"概率略升"In favour: why some would say "the probability rises slightly"

论证一:跨尺度的无缝自洽,正是"世界是一个数学结构"该有的样子。牛顿力学、玻尔兹曼方程、流体方程诞生于不同世纪、面向不同现象、形式毫无相似之处。如今被证明:它们是同一个数学结构在不同分辨率下的三张截面,层与层之间由定理而非信仰连接。若世界"本身是数学",这正是预期图景;每打通一层,图景就更完整一分。Argument one: seamless consistency across scales is exactly what "the world is a mathematical structure" ought to look like. Newtonian mechanics, the Boltzmann equation and the fluid equations were born in different centuries, addressed different phenomena, and bear no formal resemblance to one another. It has now been proved that they are three cross-sections of one and the same mathematical structure at different resolutions, with the layers joined by theorems rather than by faith. If the world "is mathematics itself", this is precisely the expected picture; every layer connected makes the picture one degree more complete.

论证二:世界需要的"非数学补丁"又少了一块。熵增/时间之箭长期像一条独立的、需要额外塞进物理学的原理(热力学第二定律)。现在它在此模型内被降级为定理——从公设变成推论。MUH 的一个间接卖点是"最终一切物理事实都还原为数学事实";每当一条看似独立的物理原理被证明是推论,这个卖点就多一个案例。这延续了维格纳"数学不合理的有效性"的惊叹,并给了它一个新的、极硬的注脚。Argument two: the world needs one fewer "non-mathematical patch". Rising entropy, the arrow of time, long looked like an independent principle that had to be stuffed into physics separately (the second law of thermodynamics). Within this model it has now been demoted to a theorem — from postulate to corollary. One indirect selling point of MUH is that "ultimately every physical fact reduces to a mathematical fact"; every time a seemingly independent physical principle is proved to be a corollary, that selling point gains another case. This continues Wigner's astonishment at "the unreasonable effectiveness of mathematics", and gives it a new and exceptionally hard footnote.

保留方向:为什么更冷静的读法是"概率不动"Against: why the cooler reading is "the probability does not move"

论证一:整场推导从未离开数学的内部。硬球本身就是一个数学理想物——现实中不存在完美刚性、零形变的球。定理证明的是"数学模型 A ⇒ 数学模型 B",即两个人类建构的模型彼此自洽;它没有、也无法检验"实在本身是否为数学结构"。地图与地图之间对得上,不能证明大地就是地图。Argument one: the entire derivation never leaves the interior of mathematics. A hard sphere is itself a mathematical idealisation — perfectly rigid, zero-deformation spheres do not exist in reality. What the theorem proves is "mathematical model A ⇒ mathematical model B", that is, that two humanly constructed models are consistent with each other; it does not, and cannot, test whether "reality itself is a mathematical structure". That two maps agree with each other does not prove that the land is a map.

论证二:贝叶斯视角下,似然比约等于一。问自己:"若 MUH 为真,我们会观察到好模型之间可以互推"——会;"若 MUH 为假、数学只是人类最好的描述工具,我们会观察到同样的事吗"——同样会,因为描述同一现象、各自都准确的模型,本就应当在重叠区域一致。一个在两种假设下都必然出现的证据,不构成判别力:后验 ≈ 先验。MUH 真正独特的主张(实在无非数学残余、一切数学结构皆存在)在这项工作中完全未被触及。Argument two: from a Bayesian standpoint, the likelihood ratio is about one. Ask yourself: "if MUH were true, would we observe that good models can be derived from one another?" — yes; "if MUH were false, and mathematics merely humanity's best descriptive tool, would we observe the same thing?" — yes again, because models that describe the same phenomenon and are each accurate ought to agree where they overlap. Evidence that must appear under either hypothesis carries no discriminating power: the posterior ≈ the prior. The genuinely distinctive claims of MUH (that reality has no non-mathematical residue, that all mathematical structures exist) are left entirely untouched by this work.

论证三(甚至可以反向读):极限过程本身是理想化。定理生效于 N→∞、ε→0 的极限;而真实气体 N 巨大却有限。严格说来,玻尔兹曼方程对真实气体永远只是(极好的)近似。有人会由此读出相反的教训:数学结构是实在的渐近影子,而非实在本身。Argument three (which can even be read the other way round): the limiting process is itself an idealisation. The theorem operates in the limit N→∞, ε→0; whereas for a real gas N is enormous but finite. Strictly speaking, the Boltzmann equation will always be only an (excellent) approximation for a real gas. Some will draw the opposite lesson from this: that mathematical structures are the asymptotic shadow of reality rather than reality itself.

把两侧放上天平,较稳妥的判断是:对 MUH 的本体论主张,这项工作近乎中性——它不是那种能区分"世界是数学"与"数学善于描述世界"的证据。但它确凿地加固了一个人人都能受益的弱命题:物理的可压缩性又深了一层——三套定律被压缩成一套加两个极限。如果你原本就因"数学的不合理有效性"而对某种柏拉图式图景抱有好感,这里多了一枚漂亮的砝码;如果你是工具主义者,你同样可以心安理得地说:看,是我们的模型彼此咬合得好。证据慷慨地兼容双方——这正是它无法裁决双方的原因。Putting both sides on the scales, the safer judgement is: with respect to the ontological claim of MUH, this work is close to neutral — it is not the kind of evidence that can distinguish "the world is mathematics" from "mathematics is good at describing the world". But it definitively reinforces a weaker proposition from which everyone benefits: the compressibility of physics has deepened by another layer — three sets of laws compressed into one set plus two limits. If the "unreasonable effectiveness of mathematics" already disposed you towards some Platonic picture, here is one more handsome weight for that pan; if you are an instrumentalist, you can just as comfortably say: look, it is our models that mesh well with one another. The evidence is generously compatible with both sides — which is exactly why it cannot adjudicate between them.

玻尔兹曼方程Boltzmann equationBoltzmann equation玻尔兹曼方程

不追踪每颗分子,只追踪一张统计表 f(x, v, t):"t 时刻,位置 x 附近、速度约为 v 的分子占多大比例",并规定这张表如何因分子对撞而演化。Instead of tracking every molecule, it tracks a single statistical table f(x, v, t): "at time t, what fraction of the molecules are near position x with velocity roughly v", and prescribes how that table evolves as molecules collide.

熵Entropyentropy熵

粗略说:系统"混乱程度"或"可能微观排法的多少"的量度。理好的扑克熵低,洗乱的熵高。热力学第二定律:孤立系统的熵不减。Roughly: a measure of a system's "degree of disorder", or of "how many microscopic arrangements are possible". A sorted deck of cards has low entropy, a shuffled one high. The second law of thermodynamics: the entropy of an isolated system does not decrease.

H 定理H-theoremH-theoremH 定理

玻尔兹曼 1872 年的证明:他的方程的解,熵随时间只增不减——不可逆性第一次成为公式的性质。Boltzmann's 1872 proof that for solutions of his equation entropy only increases with time and never decreases — the first time irreversibility became a property of a formula.

分子混沌假设Molecular chaos (Stosszahlansatz)molecular chaos分子混沌假设

推导玻尔兹曼方程的关键前提:任意两颗即将相撞的分子,统计上互为"陌生人"。本次工作的核心即严格论证此假设可长时间维持。The key premise for deriving the Boltzmann equation: any two molecules about to collide are statistically "strangers" to each other. The core of this work is a rigorous argument that the assumption can be maintained over long times.

重碰撞Recollisionrecollision重碰撞

曾直接或间接相遇过的分子再次相撞。它携带"记忆"、破坏陌生人假设,是卡住领域五十年的元凶;新工作证明其影响在极限中消失。A second collision between molecules that have already met, directly or indirectly. It carries a "memory" and destroys the stranger assumption; it is the culprit that blocked the field for fifty years, and the new work proves that its effect vanishes in the limit.

玻尔兹曼–格拉德极限Boltzmann–Grad limitBoltzmann–Grad limit玻尔兹曼–格拉德极限

正确的取极限方式:分子数 N→∞、直径 ε→0,同时保持每颗分子的"平均自由程"(两次碰撞间的平均飞行距离)不变。The correct way to take the limit: let the number of molecules N→∞ and their diameter ε→0 while keeping each molecule's "mean free path" (the average distance flown between two collisions) fixed.

纳维–斯托克斯方程Navier–Stokes equationsNavier–Stokes纳维–斯托克斯方程

描述黏性流体(水、空气)运动的宏观方程,飞机与天气预报的数学心脏。其解是否永远光滑,是另一个悬赏百万美元的千禧年难题——与本工作互不解决。The macroscopic equations describing the motion of a viscous fluid (water, air), the mathematical heart of aircraft design and weather forecasting. Whether their solutions always stay smooth is another Millennium Problem carrying a million-dollar prize — one that this work neither solves nor is solved by.

累积量Cumulantcumulant累积量

专门度量"偏离完全独立多少"的统计量:若两组量真的独立,相应累积量为零。三人的账本记的就是它。A statistic designed to measure "by how much things depart from complete independence": if two sets of quantities really are independent, the corresponding cumulant is zero. This is what the trio's ledger records.

波湍流Wave turbulencewave turbulence波湍流

大量非线性波(如海面涟漪)相互作用的统计理论,结构与气体动理学惊人平行。邓煜与 Hani 先在此建立严格数学基础,再移师粒子。The statistical theory of large numbers of interacting nonlinear waves (ripples on the sea, for example), structurally parallel to gas kinetics to a striking degree. Deng and Hani first built a rigorous mathematical foundation here, then moved their forces to particles.

数学宇宙猜想Mathematical Universe HypothesisMUH · TegmarkMUH · 数学宇宙猜想

主张物理实在本身就是数学结构(而非仅"可被数学描述"),且一切自洽的数学结构都物理存在。属哲学层面的猜想,目前无公认检验方法。The claim that physical reality is itself a mathematical structure (not merely "describable by mathematics"), and that every self-consistent mathematical structure physically exists. It is a conjecture at the philosophical level, with no accepted method of testing at present.