2021年诺贝尔奖得主乔治·帕里西对统计物理学的贡献

  2021年诺贝尔物理学奖由三位获奖者共享,以表彰他们对复杂现象的研究。真锅淑郎(Syukuro Manabe)和克劳斯·哈塞尔曼(Klaus Hasselmann)奠定了我们对地球气候及人为影响知识的基础。乔治·帕里西(Giorgio Parisi)则因他对无序和随机现象的理论的创造性贡献而获奖。在本文中,乔治·帕里西(Giorgio Parisi)的亲密合作者和朋友罗伯特·本齐(Roberto Benzi)和乌里尔·弗里希(Uriel Frisch)对他们曾分别参与的帕里西的两项主要工作阐述了个人见解。其中,罗伯特·本齐阐明帕里西及其合作者对气候从间冰期向冰河期转变的观念认识,而乌里尔·弗里希则论述尚未解决的一个古老科学问题:涉及湍流动力学及其“分形”的扩散问题。另两位诺贝尔物理奖获奖者的贡献则收录于“气候”主题下的另一篇文章中。

1. 因对复杂系统理解作出突破性贡献获得诺贝尔奖

  虽然IPCC在2007就获得过诺贝尔和平奖,但气候科学直到2021年才首次获得诺贝尔物理学奖的荣誉。这门科学依赖于经典流体力学和热力学过程,似乎与当前物理学的前沿相去甚远。然而,真正的困难在于,不同过程间发生的相互作用极为复杂,而且发生的空间和时间尺度的跨度十分巨大。

  在三位获奖者中,真锅淑郎(Syukoro Manabe)和克劳斯·哈塞尔曼(Klaus Hasselmann)首创了当前气候模型的建模方法(详见《2021年诺贝尔奖得主克劳斯·哈塞尔曼(Klaus Hasselmann)和真锅淑郎(Syukoro Manabe)对气候科学的贡献》)。在计算机能力十分有限的情况下,创建这样的模型来评估主导因素需要有深刻的物理洞见。一个更为根本的限制源于系统的不稳定,比如说,即便用现在最强大的计算机,也很难准确预报天气。然而,这样的混沌行为却帮助我们获得了长时间尺度上定义明确的统计数据,而这正是气候研究所需要的。上述问题就涉及到统计物理学(statistical physics),乔治•帕里西(Giorgio Parisi)在这一研究领域作出了若干重要贡献。

  统计物理学起初是为了从第一原理中推导出物质的特性而发展起来的。路德维希·玻尔兹曼(Ludwig Boltzmann)率先阐明了熵的概念,作出了开创性贡献。之后,阿尔伯特·爱因斯坦(Albert Einstein)和保罗·朗之万(Paul Langevin)量化了随机分子碰撞作用下胶体粒子的布朗运动。克劳斯·哈塞尔曼(Klaus Hasselmann)将类似的方法应用于气候系统,在该系统中,快速天气波动取代了分子碰撞。乔治·帕里西及其同事分析了这种波动如何触发两种稳定状态之间的转换,这也是第2节中讨论的主要内容。

  第3节讨论另外一个不同的主题:湍流的老问题,即不同尺度旋涡之间的复杂相互作用。一个多世纪以来,这个问题一直困扰着物理学界和数学界。为此,乔治·帕里西提出了“多重分形”的概念来解释实验数据。

  值得注意的是,以上两个课题并不是乔治·帕里西获得诺贝尔奖的唯一贡献。他最著名的工作是无序磁系统理论,瑞典皇家学院的官方报告[1]对此作了解释。他的诸多贡献为理解复杂系统的提供了深入的见解,不仅影响了物理学,也影响了数学、生物学、神经科学和机器学习领域,同时使他获得学生与同事的尊敬和青睐。

2. 随机共振和间冰期(作者:罗伯特·本齐)

  科学家分析了格陵兰岛和南极洲的冰层成分,提供了很好的证据,证明在过去一百万年左右的时间里,地球气候经历了数次冰川期和间冰期,周期约为10万年,温差约为10度。难题在于如何解释所观察的气候行为的准周期性交替(详见《气候的天文理论:漫长的历史》)。

环境百科全书-诺贝尔奖-地球能量收支
图1. 地球能量收支示意图。地球接受太阳辐射(Rin),辐射量可以根据地日距离精确计算。大部分太阳辐射被地球表面吸收,部分被冰和云反射回来(Rout)。地球会基于其温度向外发出红外辐射,因而平衡了太阳辐射和反射Rin-Rout。[来源:作者]

  要解决这个问题,可以参考最简单的气候模型,该模型涉及地球表面能量收支(图1)。太阳提供的入射辐射(Rin)中,部分被地球表面(冰、云和其他表面效应)反射(Rout)到太空。净入射辐射Rin-Rout使地球表面升温,发出红外辐射。知道了净入射辐射,就可以利用辐射与温度间的相关性来估算全球范围内的地面温度。在定常状态下,地球的红外辐射刚好平衡净入射辐射Rin-Rout。由此可以很好地估算出平均全球温度。目前,地球处于间冰期,对应的均温约为15℃(详见《地球平均温度》)。

  尽管出射的红外辐射中有一部分被大气中的温室气体吸收,但净发射量是温度的递增函数,接近于T4下的黑体表达式(详见《黑体的热辐射》)。这种关系在图2中被画成了T的线性函数。至于反射辐射,它在很大程度上取决于积雪的覆盖面积,因为雪能够比地面和自由流动的水更有效地反射光线。因此,入射通量Rin-Rout也随着温度的升高而增加。正如布迪科(Budyko)[2]和赛勒斯(Sellers)的研究所示,随着北美和欧亚大陆上广阔冰盖的融化,这种增长是相当急剧的,体现为图2中的黑色曲线。

  在图2中,定常状态对应于两条曲线的三个交点。然而,中间点代表了一个不稳定的状态:离开该驻点的温度小幅上升会导致入射辐射Rin-Rout过量,使温度进一步上升,直到系统到达温暖的稳定点。同样,温度的轻微下降也会被放大,并最终使系统到达代表冰川气候的寒冷稳定点。因此,该模型为两种稳定状态提供了一个简单的概念解释。在最初的布迪科-赛勒斯(Budyko-Sellers)模型中,冰川状态是指冰雪覆盖的地球。对应的是极端低温,在过去一百万年的古气候记录中从未观察到。

环境百科全书-诺贝尔奖-本齐等人的辐射平衡模型
图2. 本齐等人的辐射平衡模型简图,其灵感来自布迪科-赛勒斯模型。黑线表示净入射辐射(Rin-Rout),红线表示红外辐射。两条曲线在三个点上相交:两端的绿色点代表稳定的气候状态,中间的灰色点则代表不稳定的状态。右侧面板详细展现了较温暖状态。其中,虚线表示太阳辐射的米兰科维奇变化:平均气候状态降温0.2℃。[来源:作者]

  20世纪20年代末,塞尔维亚地球物理学家米卢廷·米兰科维奇(Milutin Milanković)对地球围绕太阳的轨道变化进行了精确而繁琐的计算。他发现,由于偏心率的变化,全球的入射辐射正以大约10万年的周期变化。而后,约翰·英博瑞(John Imbrie)等人发现,米兰科维奇计算出的太阳辐射入射振荡与冰川期-间冰期振荡期间的地球古气候温度几乎同相。

  乍看之下,人们可能会认为,米兰科维奇周期能够解释古气候记录中观察到的准周期性行为。然而,有一个困扰了科学家们几十年的大问题:米兰科维奇的太阳入射辐射的波动极其微弱,只有0.1%的数量级。这种微小的入射辐射Rin-Rout变化仅会导致约0.2℃的温度变化,比记录中冰川期和间冰期的温差要小得多。

  80年代初,乔治·帕里西及其同事提出了一种新的理论方法来解决这一难题[3]。该方法依然以地表能量收支为切入点。通过观察气候记录和环流数值模型,他们发现,由于气候的内部及非线性动力学,地球平均温度每年都在波动。关键在于要将这些波动视为“噪声”,而且是嵌入气候系统的“内部噪声”。在此基础上,帕里西和同事发现,如果存在两种温差约为10度的气候状态,那么由于内部气候波动,地球气候会呈现出从一种状态到另一种状态的随机偏移。如果将这两种状态与观测到的冰期和间冰期相对应,那么这种偏移平均每5万年就会随机发生一次。这与典型的5万年转换间隔相吻合,但并未与古气候记录中观测到的真实周期相吻合。

  可将这两种状态视作地表反射与温度的函数,不难想象,随着温度降低,气候系统中冰量和云量都会增加,并且达到平衡气候状态。

环境百科全书-诺贝尔奖-帕里西和同事使用随机共振机制搭建的模型
图3. 帕里西和同事使用随机共振机制搭建的模型,以模拟温度的时间行为。[来源:作者]

  最后一步是考虑米兰科维奇循环对该系统的影响。在这一点上,他们发现了如今的随机共振机制(the mechanism of stochastic resonance):即便很小,米兰科维奇循环也能够改变从当前的间冰期气候转换到冰期状态的概率。这些转换几乎与米兰科维奇循环同相发生。事实上,随着与不稳定中间态间距的增加,转换的概率呈指数递减。因此,米兰科维奇循环仅将这一距离微微缩减,就足以有力推动转换。图3为本齐(Benzi)、帕里西、苏特拉(Sutera)和伍尔皮亚尼(Vulpiani)原文[3]模型反映的温度行为。

  除了两种气候状态(冰期和间冰期)的存在这一重要观点之外,帕里西及其同事提出的理论还基于两个基本概念:长时间尺度上的噪声效应(noise effects)以及噪声-强迫协同效应(noise-forcing cooperation)。

  在气候理论中,这是噪声首次被视为长时间尺度上气候变化的驱动机制。在气候中,我们所说的噪声是指小尺度非线性气候变量对所谓的“粗粒度”变量(如地球全球温度)的影响。如今,尽管“噪声”这个词本身可能产生误导,但大量定性与定量例证已经允许我们就其进行讨论。受噪声影响,任何系统都将围绕某个稳定状态(如果存在)波动。然而,如果系统是非线性的,且存在多个稳定状态,噪声则能够以指数级的小概率将系统从一个状态向另一个状态驱动。尽管概率较小,但在较长的时间尺度上,如气候问题上,人们仍能观测到状态之间的随机(非周期)转换。

  第二个概念,即噪声和外部强迫的协同效应,该效应在气候理论中,乃至物理学中都是全新的:从来没有人考虑过这种可能性。帕里西和同事的论文[3]首次表明,这种协同可能发生在地球气候一类的非线性系统中。后来,这种协同效应被推广到混沌系统[4]中。如今,随机共振机制在物理学、生物学和其他科学研究中已有数以千计的应用。

3. 湍流的多重分形模型(作者:乌里尔·弗里希)

环境百科全书-诺贝尔奖-列奥纳多·达·芬奇的湍流速写
图4. 列奥纳多·达·芬奇的湍流速写,展现了不同尺度涡之间的相互作用[图源:https://www.oist.jp/photo/eddies-turbulent-pool-sketch-15th-century-leonardo-da-vinci]

  在罗马时代,卢克莱修(Lucretius)就已经对空气和水中湍流运动的共性产生了兴趣。如图4所示,16世纪初,列奥纳多·达·芬奇(Leonardo da Vinci)研究了河流中涡的衰变。上述事实表明,湍流是最古老的科学未解难题之一。

  我们首先介绍这一领域最早的成果,该成果与乔治·帕里西的研究直接相关,是A·N·柯尔莫哥洛夫(A.N. Kolmogorov, 1941)[5]的研究,简称K41。该研究以“充分发展的湍流”为研究对象,“充分发展”意味着涡可以在大尺度范围内自由相互作用。这些相互作用将能量传递给受到黏粘性阻尼的较小的涡。这一级串过程会使特定的大尺度流动特征受这种相互作用的重大影响,从而导致“阻力危机“(详见《运动物体受到的阻力》)(详见《运动物体受到的阻力》)。柯尔莫哥洛夫随后发现,若完全忽略黏粘性影响,可以得出一个动力学方程,在该方程中,较小的旋涡在统计意义上与主导的大涡旋相似。当黏粘度趋于0时,相似性程度由有限的能量耗散率决定。这个自相似图像(图5a)取得了惊人的成功。

环境百科全书-诺贝尔奖-级联
图5. 左:柯尔莫哥洛夫(1941)提出的大小涡级串图,空间尺度形成公比为r的几何级数。右:弗里希(Frisch)、苏莱姆(Sulem)、内尔金(Nelkin)(1978)提出的分形模型中的间歇级串。

  尽管柯尔莫哥洛夫主攻数学,而且是20世纪最伟大的数学家之一,但他仍希望使用从莫斯科实验室(Moscow laboratory)及周边收集到的实验数据证实自己的物理理论。这些数据有力地证明,湍流与上述自相似图像相去甚远:小涡比大涡填充的空间少得多,并呈现出间歇性(intermittency)(图5b)。因此,柯尔莫哥洛夫与其合作者在1961年提出了间歇性模型[6]

  G·K·巴切勒(G.K. Batchelor)和A·A·汤森(A.A. Townsend)(1949)[7]也发现了间歇现象,问题在于他们发现的这一现象是否能与K41的自相似图像兼容。为描述柯尔莫哥洛夫的间歇性,B·曼德尔布罗特(B. Mandelbrot,1974)[8]提出使用“分形(fractal)”模型,即通过在越来越小的尺度上迭代插入相似的图案而获得的几何对象。自19世纪末以来,数学家们就提出了这一概念。1950年,湍流和大气动力学研究的另一位巨匠L·F·理查森(L.F. Richardson)在完全不同的背景下使用了分形,研究如何避免国家之间的冲突。国家之间分形错综复杂的分形边界似乎确实比直线的边界更容易引发冲突。

环境百科全书-诺贝尔奖-多重分形信号示例
图6. 多重分形信号示例,分别为湍流(上)和金融市场(下)。对信号的分析基于直方图,其中横轴代表曲线上不同两点的高度差值。对于湍流,差值的尺度可以柯尔莫哥洛夫(Kolmogorov)尺度(即最小涡的尺度)的倍数来表示;而对于金融市场(此处为两种货币之间的汇率),时间间隔则以小时表示。(直方图取自S. Ghashghaie, W. Breymannt, J. Peinke, P. Talkner &Y. Dodgell (1996), Nature 381)

  回到湍流,穿过湍流场的速度探针会产生典型的“噪声”切口,如图6(上)所示。类似的噪声信号在很多复杂系统中都会遇到,图6(下)所示的金融市场就是其中之一。这些信号令数学家为之着迷,因为该曲线处处都是粗糙的(斜率未定义),数学家对此产生极大兴趣。

  要掌握信号的多尺度结构,一种自然方法就是研究相隔给定距离Δr的两点之间的速度差Δv。方差(|Δv|2的平均值)即表示能量在不同尺度间的分布[9]

  然而,这一方法并不全面,没有将稀少且强烈的涡和具有相同全局能量但充满空间的大涡区分开来(相关信息详见图6右侧的概率分布直方图)。这些直方图的“长尾”代表了稀少且强烈的涡旋。随着间距Δr的减小,频率不断增加,说明小涡旋很稀疏。这就是间歇性的本质。在直方图中,如果横轴考虑时间间隔Δt而非距离,则可以对任何时间序列进行类似分析,如图6中关于金融市场的分析。图中尾部表示在给定的时间间隔内大幅下跌或上涨的概率,与风险管理直接相关。

  再回到湍流的问题上,这些尾部可以利用|Δv|p(p阶矩)的平均值来评估。p阶越高,该值对大的波动就越来越敏感,其对间距Δr的依赖称为结构函数,能够体现涡间歇性的特征。在K41描述的真正自相似级串中,p阶结构函数表现为幂律Δrp/3(图7)。实验表明,对指数ζ(p)小于p/3的幂律|Δr|ζ(p)而言:在小间距Δr的极限下,高阶矩相对较强,这在实际上量化了间歇性。

  柯尔莫哥洛夫(Kolmogorov)在1961年提出的模型更接近实验结果,但该模型会导致指数p较大时ζ(p)值减小,在数学上不一致。因此,作为替代,以弗里希(Frisch)、苏莱姆(Sulem)、内尔金(Nelkin)(1978)[10]等为代表的科学家引入了与湍流动力学方程有一定联系的分形模型。该模型下ζ(p)随p线性增加,且斜率小于1/3。斜率取决于单一参数,如分形维数,从而量化了活跃涡尺度愈小愈稀疏的现象。然而实验结果却是图7中所呈现的凸曲线。

环境百科全书-诺贝尔奖-结构函数
图7. 柯尔莫哥洛夫将“结构函数”Sp(Δr)的标度指数ζ(p)定义为在距离Δr(沿Δr方向投影)上速度增量的p阶(统计)矩。这里假定湍流是均匀(平移不变)各向同性(旋转不变)的。结构函数Sp(Δr),大致是一个幂律|Δr|ζ(p)。K41是第一个柯尔莫哥洛夫理论,其中Sp(Δr)与|Δr|p/3成正比。β模型显示了弗里希-苏莱姆-内尔金(Frisch-Sulem-Nelkin)(1978)的结果,它具有单一的分形维数。黑色三角形是安索梅(Anselmet)等人(1984)的实验数据。这些数据现在仍然是多重分形模型的基准。对数正态模型(K61)提供了第一种确定间歇性的方法,但对于大指数p来说,ζ(p)的减少与数学上的情况是不一致的。

  1983年夏天,乔治·帕里西和乌里尔·弗里希参加了意大利瓦伦纳的暑期班,主题是“地球物理流体动力学和气候动力学中的湍流和可预测性”。F. Anselmet, Y. Gagne, E.J. Hopfinger, R.A. Antonia的实验数据[11]几乎不能与单个分形兼容(如图7所示)。乔治·帕里西凭着非凡的物理直觉,提出了一个多重分形模型来对其进行解释。这种一致性依赖于拟合参数,但背后的分析揭示了深刻的数学属性。

  这种具有不止一个、甚至是无限多个分形维数的想法源于一个非常高难度的主题:评估大风险或金融破产。瑞典数学家哈拉尔德·克拉梅尔(H.Cramér)(1938)[12]有着金融相关的背景,在这方面做出了一些开创性的工作。在克拉梅尔的研究之前,普遍观点认为,通过增加n个独立均匀分布的随机变量,就可以(在适当的条件下)得到两个主要性质:平均数将趋向于平均值(大数定律);将离均差除以n1/2,结果将趋向于高斯定律(中心极限定理)。

  如果我们将变量相乘而非相加(或等效地,将变量相加并取其指数),会发生什么情况呢?这正是柯尔莫哥洛夫(Kolmogorov)[13]研究的岩石粉碎问题。给定颗粒尺寸的概率实际上是连续独立破碎事件概率的乘积。因此,获得高斯极限的情况非常罕见。所有这些问题都涉及一个奇怪的函数,称为大偏差函数,表示的是与大数定律的偏差。该函数的名称在不同的领域有所区别:“速率函数”、“克拉梅尔函数”或“熵”。是的,玻尔兹曼熵只是一个特例;在没有任何高级概率理论的情况下理解熵堪称壮举,除非你是玻尔兹曼本人,否则很难做到。

  以下是一些关于多重分形的主要参考文献。首篇涉及多重分形的文献实际上是帕里西和弗里希在1983年瓦伦纳会议论文集上发表的一篇论文的2.5页附录[14]。紧接着,本齐、帕拉丁(Paladin)、帕里西和武尔皮亚尼(Vulpiani)撰写了一篇更详细的论文[15]。该论文涉及从充分发展的湍流引申到混沌动力系统中的奇异吸引子的重要内容。这些八十年代的论文以勒让德变换(与统计热力学中定义熵时出现的变换相同)为关键论点,却没有提到克拉梅尔的大偏差。弗里希(Frisch)(1995)[16]的书籍则侧重于对大偏差的基础介绍,作出了更加详细的阐释。最后,伊夫·梅耶尔(Yves Meyer)(2021)[17]在网上发表了一篇文章,原本是为数学家所撰写,但不具备高等数学知识的人也可以阅读。

环境百科全书-诺贝尔奖-湍流和多重分形
图8. 从乔纳森·斯威夫特(Jonathan Swift)的跳蚤级联到湍流和多重分形。[来源:作者]

  回想一下,路易斯·弗莱·理查德森(L.F. Richardson)早在20世纪20年代引入级串来描述湍流时,就已经意识到分形。L·F·理查德森借用了乔纳森·斯威夫特(Jonathan Swift)的一首诗(图8,附杰雷米·贝克(Jérémie Bec)的插图)。除非左跳蚤和右跳蚤的头间距变得更小,否则跳蚤模型显然是分形的,而且很容易多重分形。正如伯努瓦·曼德尔布罗特(Benoit Mandelbrot)所指出的,分形在自然现象中相当普遍,这正是我们希望看到的。他反对那些把一切都弄得直截了当的人。

  最后,我们要强调一个悖论。毋庸置疑,湍流测量正越来越精确,推动着多重分形分析的发展。关于模型方程中出现的多重分形奇点,有一些切实可靠的定理[18]。然而,我们还没有找到一个已被证明的、适用于描述三维强湍流的Navier-Stokes方程的定理。尽管数学基础并不牢固,但多重分形分析仍是当前常用的工具,用来定量分析环境中观察到的各种复杂的混沌或湍流现象,如气候时间记录、阵风、云的形状、太阳风。

4. 总结

  • 乔治·帕里西(Giorgio Parisi)对统计物理学作出了卓越的贡献,取得重大进展,因而被授予2021年诺贝尔物理学奖。
  • 他所作出的贡献中,有两项与气候和环境流动有关,即“随机共振”和“多重分形”概念;
  • 随机共振是噪声和外部周期性强迫的协同效应,可以实现两个稳态之间近乎周期性的切换。它最初是作为一种概念模型提出的,旨在记录过去百万年间的古气候,但后来在物理、生物学和其他科学领域也有许多应用;
  • 引入多重分形可以描述湍流中观察到的多尺度结构。事实证明,这一概念与其他混沌系统有着深刻的联系,同时促进了随机过程数学分析的进步。

 


参考资料及说明

封面图片:2021年诺贝尔物理学奖的三位获奖者,从左到右分别是真锅淑郎(Syukuro Manabe)、克劳斯·哈塞尔曼(Klauss Hasselmann)和乔治·帕里西(Giorgio Parisi)[来源:©插图 Niklas Elmehed,诺贝尔奖推广部]

[1] Popular science background : https://www.nobelprize.org/uploads/2021/10/popular-physicsprize2021.pdf, higher level version:  https://www.nobelprize.org/uploads/2021/10/sciback_fy_en_21.pdf

[2] Budyko, I. (1969) “The effect of solar radiation variations on the climate of the earth.” Tellus 21, 611-619.

[3] Benzi , Parisi G., Sutera A., Vulpiani A. (1982) “Stochastic resonance in climatic change”, Tellus 34, 10-16.

[4] Benzi , Sutera A., Vulpiani A. (1981) “The mechanism of stochastic resonance”, J. Phys. A: Math. Gen. 14, L453-457.

[5] Kolmogorov, N. (1941a) “The local structure of turbulence in incompressible viscous fluid for very large Reynolds number”. Dokl. Akad. Nauk SSSR 30, 9-13 (translated in Proc. R. Soc. Lond. A 434, 9-13 (1991)).

[6] Kolmogorov, N. (1961) “Précisions sur la structure locale de la turbulence dans une fluide visqueux aux nombres de Reynolds élevés”, in “La Turbulence en Mécanique des Fluides”, 447-451, eds. A. Favre, L.S.G. Kovasznay, R. Dumas, J. Gaviglio & M. Coantic. Gauthiers-Villard, Paris. An English translation can be found in Kolmogorov, A.N. (1962) J. Fluid Mech. 13, 82 – 85.

[7] Batchelor, K and Townsend, A.A. (1949) “The nature of turbulent motion at large wave-numbers”. Proc. R. Soc. Lond. A 199, 238-255.

[8] Mandelbrot, (1974) “Intermittent turbulence in self-similar cascades: divergence of high moments and dimension of the carrier”. J. Fluid Mech. 62, 331-358.

[9] This variance is related to the « energy spectrum » which can be obtained also from the Fourier transform of the

[10] Frisch, , Sulem, P.L. and Nelkin, M (1978) “A simple dynamical model of intermittent fully developed turbulence”. J. Fluid Mech. 736, 5-23.

[11] Anselmet, , Gagne, Y., Hopfinger, E.J. and Antonia, R.A. (1984) “High-order velocity structure function in turbulent shear flow”. J. Fluid Mech. 140, 63-89.

[12] Cramér, (1938) “Sur un nouveau théorème-limite de la théorie des probabilités”, Actualités Scientifiques et Industrielle. 736, 5-23.

[13] Kolmogorov, N. (1941) “On the logarithmic normal law of distribution of the size of particles under pulverization” Dokl. Akad Nauk SSSR 31, 99-101.

[14] Parisi, and Frisch, U. (1985) On the singularity structure of fully developed turbulence, in Turbulence and Predictability in Geophysical Fluid Dynamics. Proceed Intern. School of Physics ‘E. Fermi’. 1983. Varenna. Italy, 84-87. This is an Appendix to the paper ‘Fully Developed Turbulence and Intermittency’ by Frisch, U. pp. 71-88.

[15] Benzi, Paladin, Parisi and Vulpiani (1984) “On the multifractal nature of fully developed turbulence and chaotic systems”, J. Phys. A: Math. Gen. 17, 3521.

[16] Frisch, (1995) “Turbulence, the Legacy of A.N. Kolmogorov”, Cambridge University Press.

[17] Meyer, Y. (2021) “Giorgio Parisi et la turbulence”, Available in French on the site of the “Institut national des sciences mathématiques et de leurs interactions” https://www.insmi.cnrs.fr/fr/cnrsinfo/giorgio-parisi-et-la-turbulence-par-yves-meyer.

[18] Jaffard, (2000) “On the Frisch-Parisi conjecture”, J. Math. Pures Appl. 69, 6, 525-552.


环境百科全书由环境和能源百科全书协会出版 (www.a3e.fr),该协会与格勒诺布尔阿尔卑斯大学和格勒诺布尔INP有合同关系,并由法国科学院赞助。

引用这篇文章: BENZI Roberto, FRISCH Uriel (2024年4月12日), 2021年诺贝尔奖得主乔治·帕里西对统计物理学的贡献, 环境百科全书,咨询于 2024年11月21日 [在线ISSN 2555-0950]网址: https://www.encyclopedie-environnement.org/zh/physique-zh/parisi-nobel-prize-physics-2021/.

环境百科全书中的文章是根据知识共享BY-NC-SA许可条款提供的,该许可授权复制的条件是:引用来源,不作商业使用,共享相同的初始条件,并且在每次重复使用或分发时复制知识共享BY-NC-SA许可声明。

On the contributions to statistical physics of Giorgio Parisi, Nobel Prize winner 2021

Three  Laureates  share  the 2021 Nobel  Prize  in  Physics  for  their  studies  of  complex  phenomena. Syukuro Manabe and Klaus Hasselmann laid the foundation of our knowledge of Earth’s climate and how humanity influences it. The third laureate, Giorgio Parisi is rewarded for his revolutionary contributions to the theory of disordered and random phenomena. In this article the authors Roberto Benzi and Uriel Frisch, close collaborators and friends of Giorgio Parisi, offer their vision of two of his major contributions, in which they have been respectively involved. Roberto Benzi explains the contribution of Parisi and collaborators to the conceptual understanding of the transitions from temperate to ice age climates. Uriel Frisch discuss one of the oldest unsolved problem in science, concerning the dynamics of turbulence and its proliferation of “fractals”.  The contributions of the two other laureates are presented in another article under the heading ‘climate’.

1.  A Nobel Prize for groundbreaking contributions to understanding of complex systems

While IPCC received the Nobel Prize of peace in 2007, climate science is honored by the Nobel price of physics for the first time in 2021. This science relies on classical fluid mechanics and thermodynamic processes, apparently far from the current frontier of physics. The difficulty however arises from the complex interaction of these various processes, occurring over a huge range of scales in space and time.

Among the three laureates, Syukoro Manabe and Klaus Hasselmann pioneered the modeling approach that led to the current climate models (read “On the contributions to climate sciences of Klaus Hasselmann et Syukuro Manabe, Nobel prize winner 2021”). This required a good physical insight to evaluate the dominant effects with very limited computer power. A more fundamental limitation is due to the unstable behaviour of the system, exemplified by the difficulty of weather forecast, even with the most powerful computers available nowadays. However such a chaotic behaviour leads to well defined statistics on long time scale, as requested for climate studies. This is the realm of statistical physics, the field of research to which Giorgio Parisi has provided several important contributions.

Statistical physics was initially developed to derive the properties of matter from first principles. After the pioneering contributions of Ludwig Boltzmann who clarified the concept of entropy, Albert Einstein and Paul Langevin quantified the Brownian motion of colloidal particles under the effect of random molecular collisions. Klaus Hasselmann has applied similar approaches to the climate systems, in which fast weather fluctuations replace molecular collisions. Giorgio Parisi and colleagues has analysed how such fluctuations trigger transitions between two stable states, as discussed in section 2.

Section 3 deals with a different topic, the old problem of turbulence, an intricate interaction of eddies at different scales which has defied physics and mathematics for more than a century. Giorgio Parisi introduced the concept of multifractals to account for experimental data.

Note that these two topics are not the only contributions for which Giorgio Parisi received the Nobel Prize. His most famous work is the theory of disordered magnetic systems, as explained by the official reports [1] of the Royal Swedish Academy. His various contributions provide insight in complex systems beyond physics. They also influence mathematics, biology, neuroscience and machine learning. This influence was favoured by his stimulating and friendly leadership among students and colleagues.

2. Stochastic resonance and interglacial periods (by R. Benzi)

Scientists analysing the ice composition in Greenland and Antarctica provided good evidence that, during the last million years or so, Earth’s climate experienced glacial and interglacial periods with a periodicity of about 100,000 years and a temperature difference of about 10 degrees. It was a challenging question how to explain the observed quasi-periodic alternation of climate behavior (read “Astronomical theories of climate: a long history”).

Figure 1. Schematic view of the Earth energy budget. Earth receives Sun solar radiation (Rin) which can be exactly computed from Earth-Sun distance. Most of the Sun radiation is absorbed by Earth ground while part of it is reflected back by ice and clouds (Rout). The Earth is emitting infrared radiation due to its temperature, which balances RinRout in a stationary regime. [Source: authors]
To fix the question, let us consider the simplest possible climate model which deals with the global Earth energy budget (see figure 1). Sun provides incoming radiation (Rin) which is partly reflected (Rout) to the space by Earth’s surface (ice, clouds and other surface effects). The net incoming radiation Rin-Rout warms up Earth’s surface which emits infrared radiation. Knowing this net incoming radiation, one can estimate the ground temperature (on a global scale) using our knowledge on how radiative emission depends on temperature. In a stationary state, this infrared radiation just balances the net incoming radiation Rin-Rout. This provides a rather good estimate of the Earth mean global temperature corresponding to the present interglacial period around 15 degrees Celsius (read “The average temperature of the Earth”).

Although the emitted infrared flux is partly absorbed by green house gas in the atmosphere, the net emission is an increasing function of temperature, close to the black body expression in T4(read “The thermal radiation of the black body”). This is sketched as a linear function of T in figure 2. As for the reflected radiation, it strongly depends on the area covered by snow, which reflects light more efficiently than ground and free water. Therefore the incoming flux Rin-Rout also increases with temperature. As shown by Budyko [2] and Sellers, this increase is rather sharp with the melting of large ice cap over North America and Eurasia. This result in the black curve shown in figure 2.

Stationary states correspond to the three intersections of the two curves. The middle point represents however an unstable state: a small increase of temperature away from this stationary point indeed leads to an excess of incoming radiation Rin-Rout, resulting in a further increase of temperature, until the system reaches the warm stable point. Similarly a slight decrease of temperature is amplified and leads to the cold stable point representing a glacial climate. Therefore this model provides a simple conceptual explanation of two stable states.  In the original Budyko-Sellers model the glacial state referred to ice-covered Earth. It corresponds to extremely low temperature, never observed in paleoclimatic records of the last million years.

Figure 2. Sketch of the radiation balance model of Benzi et al. inspired from the model of Budyko Sellers. The black line shows the net incoming radiation (Rin-Rout) while the red line is the infrared emission. The two curves meet in three points: two points (in green) represent stable climate states while the grey point at the middle is an unstable state. In the right hand panel focused on the warmer state, the Milanković change in the solar radiation is shown by the dashed line: the mean climate state becomes colder by 0.2 degree. [Source: authors]
In the late 1920s the Serbian geophysicist Milutin Milanković performed an accurate and cumbersome computation of the Earth orbital variation around the Sun. He discovered that due to the eccentricity variation, the global incoming radiation was changing with approximately 100,000 years cycle. It was later discovered by John Imbrie and others that oscillation of the incoming sun radiation, as computed by Milanković, was almost in phase with the paleoclimatic Earth temperature during the glacial-interglacial oscillations.

At first sight, one may think that the Milanković cycle is able to explain the observed almost periodic behaviour in paleoclimatic records. However, there was a major problem puzzling scientists for decades: the point is that the Milanković fluctuations in the solar incoming radiation was extremely weak, of the order of 0.1 percent. This very small modulation of Rin-Rout would lead to a temperature modulation of about 0.2 degree, much smaller than the recorded temperature difference between glacial and interglacial states.

In the early eighties, Giorgio Parisi and his co-workers developed a new theoretical approach to solve the puzzle [3]. The starting point was again the surface energy budget previously considered. Next, looking at climate records and numerical circulation models, they understood that the Earth mean temperature is fluctuating from year to year due to internal and nonlinear dynamics of the climate. The crucial point was to consider these fluctuations as “noise” and in particular “internal noise” embedded in climate dynamics. Parisi and co-workers went on with this idea and observed that if there exist two climate states with a temperature difference of about 10 degrees, than Earth climate, due to internal climate fluctuations, can exhibit a random excursion from one state to another. This would randomly occur once every 50,000 years in average if the two states correspond to the observed glacial and interglacial states. This fits with the typical interval of 50 000 years between transitions, but without the true periodicity observed in paleoclimatic records.

The existence of two states can be understood, again, as the change of the out coming reflection of Earth surface as a function of temperature: upon decreasing temperature, it is no hard to imagine an increase of both ice and cloud cover in the climate and an equilibrium climate state can be reached.

Figure 3. Time behaviour of the temperature simulated with the model proposed by Parisi and co-workers using the mechanism of stochastic resonance. [Source: authors]
The final step was to consider the effect of Milanković cycle on the system. At this point they discovered what is now called the mechanism of stochastic resonance: even if small, the Milanković cycle was able to change the probability of a transition from the present interglacial climate to the glacial state. These transitions were occurring almost in phase with the Milanković cycle. Indeed the probability of transition exponentially decreases with the distance to the intermediate unstable state, so a small reduction of this distance by the Milanković cycle is sufficient to strongly favour the transition. In figure 3 we show the temperature behaviour in the model used in the original publication [3] of Benzi, Parisi, Sutera and Vulpiani.

Beside the highly non trivial point concerning the existence of two climatic states (glacial and interglacial), the theory proposed by Parisi and coworkers is based on two fundamental concepts: noise effects on long time scales and noise-forcing cooperation.

For the first time in climate theories, noise was providing the driving mechanism for climate change on very long time scales. In climate, what we call noise is the effect of small scales nonlinear climate variables on the so called “coarse-grained” variables like the Earth global temperature. By now there are well defined and quantitative examples which allow us to speak of noise even if the word by itself can be misleading. The effect of the noise makes any system to perform fluctuations around some stable state (if it exists). However, if the system is non-linear and there exist more than one stable state, noise, with an exponential small probability, can drive the system from one state to another. Although small, on a very long time scale, as in the climate case, one should observed random (non-periodic) transition among the states.

The second point, namely cooperative effects of noise and external forcing, was entirely new in physics not only within climate theorists: nobody ever thought about this possibility. The paper[3] by Parisi and colleagues shows for the first time that this cooperation may occur in non-linear system like Earth’s climate. Later on, this cooperative effect was generalised for chaotic system [4] and, by now, there exist thousands of applications which exploit the mechanism of stochastic resonance in physics, biology and other scientific research.

3. Multifractal models for turbulence (by U. Frisch)

Figure 4. Sketch of turbulence  by Leonardo da Vinci, showing the interaction of eddies at different scales (source https://www.oist.jp/photo/eddies-turbulent-pool-sketch-15th-century-leonardo-da-vinci )

The very similar aspect of turbulent motion in air or water already interested Lucretius at the time of the Romans.  In the early 16th century, Leonardo da Vinci studied the decay of eddies in rivers, see figure 4. All this makes turbulence one of the oldest unsolved problems in science.

Here, we begin with the oldest contribution, directly relevant to the work of Giorgio Parisi, namely that of A.N. Kolmogorov (1941) [5], abbreviated as K41. It deals with “fully developed turbulence”, which means that eddies can freely interact over a wide range of scales. These interactions transfer energy to very small eddies, which are damped by viscosity. This so-called energy cascade is at stake for instance in the process of turbulent drag (read “Drag suffered by moving bodies”). Specific flow features at the largest scales get scrambled by turbulent interactions through this cascade process. Kolmogorov then observed that, ignoring viscosity all the way together could lead to a dynamical equation in which the smaller and smaller eddies would just be similar – in a statistical sense – to those governing larger eddies. The correct level of similarity is determined by demanding a finite energy dissipation rate as the viscosity tends to zero. This self-similar picture (cf. figure 5, left) had an amazing success.

Figure 5. The cascade from large to small eddies according to Kolmogorov (1941), left hand side. The spatial scales form a geometric series with multiplier r. On the right hand side, the intermittent cascade in the fractal model of Frisch, Sulem, Nelkin (1978)

Although Kolmogorov was foremost a mathematician, one of the greatest in the 20th century, he wanted to confirm his theory by experimental data collected in his Moscow laboratory and nearby. Such work provided strong evidence that turbulence is far from the above self-similar picture: the small eddies are much less space-filling than the larger ones and display intermittency (cf. figure 5, right). This led Kolmogorov and collaborators to propose a model for intermittency in 1961 [6]

An intermittency phenomenon had also been discovered by G.K. Batchelor and A.A. Townsend (1949) [7] and the question arose whether this phenomenon would be or not compatible with the self-similar picture of K41. For describing Kolmogorov’s intermittency, B. Mandelbrot (1974) [8] proposed using his “fractal” model. Fractals are geometrical objects obtained by iteratively inserting a similar pattern at smaller and smaller scales. Mathematicians had introduced this concept since the end of the 19th century. In 1950, L.F. Richardson, another giant of turbulence and atmospheric dynamics, has used fractals in a completely different context, namely how to avoid conflicts between nations. It seems that a fractally convoluted border between countries will indeed be more prone to developing conflicts than a straighter one.

Figure 6. Examples of multifractal signal, from turbulence (top) and from the financial market (bottom). The signal is analysed by taking the histogram of the difference between two points with different separations indicated on the curves. For turbulence the separation is expressed as multiples of the Kolmogorov scale (the size of the smallest eddies). For the financial market (here an exchange rate between two currencies), the time interval is expressed in hours. (histograms taken from  S. Ghashghaie, W. Breymannt, J. Peinke, P. Talkner &Y. Dodgell (1996), Nature 381)

Returning to turbulence, a velocity probe crossing the turbulent field yields a typical “noisy” cut as shown in figure 6, top. Similar noisy signals are encountered in many complex systems, for instance financial markets (figure 6, bottom). They fascinate mathematicians as they are everywhere rough (the slope is not defined).

To grasp the multiscale structure of the signal, a natural approach is to study the velocity difference Δv between two points separated by a given distance Δr. The variance (the average of |Δv|2) indicates how the energy is distributed [9] among the different scales. However this is not the whole story: it does not distinguish rare and intense eddies from space-filling eddies with the same global energy. Such information is provided by the probability distributions (histograms) shown in the right hand side of figure 6. Long “tails” of these histograms represent rare and intense events. Those become stronger for small separation Δr, revealing the small eddies are rather sparse. This is the essence of intermittency. Similar analysis can be done for any time series by considering a time interval Δt instead of a distance, as shown in figure 6 for a financial market. Then the tails represent the probability of large drop or rise in a given time interval, which is of direct relevance for risk management.

Coming back to turbulence, these tails can be evaluated by considering the average of |Δv|p (the moment of order p). Those are indeed more and more sensitive to large fluctuations as their order p is higher. Their dependency on the separation Δr, called the structure functions, therefore characterises intermittency. In the truly self-similar cascade described by K41, the structure function of order p behaves as the power law Δrp/3 (figure 7). Experiments indicate a power law |Δr|ζ(p) with exponents ζ(p) smaller than p/3 :  high moments are relatively stronger in the limit of small separation Δr, which quantifies intermittency.

The model proposed by Kolmogorov in 1961 is closer to the experimental results, but it leads to a decrease of ζ(p) for large index p which turns out to be mathematically inconsistent. As an alternative, fractal models making some contact with the dynamical equations for turbulence were introduced, particularly by Frisch, Sulem and Nelkin (1978) [10]. This yields a linear increase of ζ(p) with p, with slope smaller than 1/3. The slope depends on a single parameter, such as the fractal dimension, quantifying how the active eddies get more and more sparse at smaller and smaller scales. Experiments however reveal a convex curve, as shown in figure 7.

Figure 7. The scaling exponent ζ(p) for the “structure functions” Spr), defined by Kolmogorov as the (statistical) moments of order p of the velocity increments over a distance Δr (projected along the direction Δr). Here the turbulence is assumed to be isotropic (rotation invariant) and homogeneous (translation invariant). The structure function  Spr),  is roughly a power-law |Δr|ζ(p). K41 refers to the first Kolmogorov theory, where Spr) is proportional to |Δr|p/3. The β-model shows the Frisch-Sulem-Nelkin (1978) results which has a single fractal dimension. The black triangles are the experimental data of Anselmet et al. (1984), which are still now benchmarks for the multifractal model. The lognormal model (K61) provided a first approach to intermittency but the decrease of ζ(p) for large index p is mathematically inconsistent.

In the summer 1983, G. Parisi and U. Frisch attended a Varenna (Italy) summer school on “Turbulence and predictability in geophysical fluid dynamics and climate dynamics.” Experimental data by F. Anselmet, Y. Gagne, E.J. Hopfinger, R.A. Antonia [11] were hardly compatible with a single fractal, as shown in figure 7. With a remarkable physical intuition, G. Parisi suggested a multifractal model to account for these results. This agreement relies on fitting parameters but the underlying analysis revealed deep mathematical properties.

This idea of having more than one fractal dimension – even infinitely many fractal dimensions – has its origin in a very difficult subject: the evaluation of large risk or of financial ruin. The pioneering work had been done by a Swedish mathematician, H. Cramér (1938) [12] with strong background in finance. The standard vision before Cramér was that, by adding n independent and equally distributed random variables, one would get (with suitable conditions) two major properties : the mean would tend to the average (law of large numbers), and deviations from the mean, divided by n1/2, would tend to a Gaussian law (the Central Limit Theorem).

But what happens when we multiply the variables instead of adding them (or equivalently when we add variables and take an exponential of the result)? This is precisely what happens in the pulverization of rocks, a problem studied by Kolmogorov [13] before the results of Cramér were well known. The probability of a given grain size indeed results from a product of probabilities of successive independent breaking events. It is then exceptional to obtain Gaussian limits. All these problems involve a strange function called the large deviation function, which characterises the deviation from the law of large numbers. The name changes depending on the field: “the rate function”, the “Cramér function” or the “entropy”. Yes, Boltzmann entropy is just a particular case; understanding entropy, without any advanced probability theory is remarkable, unless one is Boltzmann.

Here are some key references on multifractals. The initial reference is actually a 2.5 page Appendix by Parisi and Frisch to a paper by Frisch [14] published in the proceedings of the 1983 Varenna meeting.  Immediately following, a more detailed paper was written by Benzi, Paladin, Parisi and Vulpiani [15]. It included a crucial extension from fully developed turbulence to strange attractor in chaotic dynamical system. The key argument of these papers of the eighties make use of the Legendre transformation, the same transformation that appears in statistical thermodynamics when defining the entropy, but Cramér’s large deviations are not mentioned. A much more detailed presentation, centered on an elementary presentation of large deviations is in the book of Frisch (1995) [16]. Finally, there is an excellent online text by Y. Meyer (2021) [17], which is in principle intended for mathematician, but can be read also by persons without advanced mathematical knowledge.

Figure 8. From Jonathan Swift’s cascade of fleas to turbulence and multifractals. [Source: authors]
We wish now to return to L.F. Richardson who was certainly aware of fractals already in the 1920s, when he introduced cascades in turbulence. Richardson borrowed a poem from Jonathan Swift, which is reproduced in figure 8 together with an illustration by Jérémie Bec. Unless the head-to-head distance between the left and the right fleas is made much smaller, the flea model is clearly fractal (and easily multifractal). As pointed out by Benoit Mandelbrot, fractals are quite common in natural phenomena and that is what we normally crave for. Mandelbrot objected against those making everything straight and rigid.

Finally, we want to stress a paradox. It is clear that increasingly accurate measurements of turbulent flow have been a driving force in the development of multifractal analysis. There are some solid theorems [18] concerning multifractal singularities arising in model equations. Nevertheless we do not have a single proven theorem applicable to Navier-Stokes equations, which describes three-dimensional (3D) strong turbulence. In spite of this precarious mathematical basis, multifractal analysis is now routinely used to quantitatively analyse a wide variety of complex chaotic or turbulent phenomena observed in the environment, like climatic time records, wind gusts, cloud shapes, solar wind.

4. Messages to remember

  • Giorgio Parisi has been awarded the Nobel price of physics in 2021 for his leading contributions to major advances in statistical physics.
  • Two of these contributions are relevant to climate and environmental flows. Those are the concepts of “stochastic resonance” and “multifractals”;
  • Stochastic resonance is a cooperative effect of noise and external periodic forcing which can lead to nearly periodic switching between two stable states. It was initially proposed as a conceptual model for paleoclimatic records of the last million years but it turned out to have many applications in physics, biology and other scientific fields;
  • Multifractals have been introduced to describe the multi-scale structure observed in turbulent flows. The concept turned out to have deep relevance for other chaotic systems and it triggered progress in the mathematical analysis of random processes.

 


Notes and references

Cover image. The three laureates of the Nobel Prize in physics, from left to right, Syukuro Manabe, Klauss Hasselmann, Giorgio Parisi [Source : © illustration Niklas Elmehed, Nobel Prize Outreach]

[1] Popular science background :  https://www.nobelprize.org/uploads/2021/10/popular-physicsprize2021.pdf , higher level version: https://www.nobelprize.org/uploads/2021/10/sciback_fy_en_21.pdf

[2] Budyko, M. I.  (1969) “The  effect of solar  radiation variations  on  the  climate  of  the  earth.”  Tellus 21, 611-619.

[3] Benzi R., Parisi G., Sutera A., Vulpiani A. (1982) “Stochastic resonance in climatic change”, Tellus 34, 10-16.

[4] Benzi R., Sutera A., Vulpiani A. (1981) “The mechanism of stochastic resonance”, J. Phys. A: Math. Gen. 14, L453-457.

[5] Kolmogorov, A.N. (1941a) “The local structure of turbulence in incompressible viscous fluid for very large Reynolds number”. Dokl. Akad. Nauk SSSR 30, 9-13 (translated in Proc. R. Soc. Lond. A 434, 9-13 (1991)).

[6] Kolmogorov, A.N.  (1961) “Précisions sur la structure locale de la turbulence dans une fluide visqueux aux nombres de Reynolds élevés”, in “La Turbulence en Mécanique  des Fluides”, 447-451, eds. A. Favre, L.S.G. Kovasznay, R. Dumas, J. Gaviglio & M. Coantic. Gauthiers-Villard, Paris. An English translation can be found in Kolmogorov, A.N.  (1962) J. Fluid Mech. 13, 82 – 85.

[7] Batchelor, G.K and Townsend, A.A. (1949) “The nature of turbulent motion at large wave-numbers”. Proc. R. Soc. Lond. A 199, 238-255.

[8] Mandelbrot, B. (1974) “Intermittent turbulence in self-similar cascades: divergence of high moments and dimension of the carrier”. J. Fluid Mech. 62, 331-358.

[9] This variance is related to the « energy spectrum » which can be obtained also from the Fourier transform of the signal.

[10] Frisch, U., Sulem, P.L. and Nelkin, M (1978) “A simple dynamical model of intermittent fully developed turbulence”. J. Fluid Mech. 736, 5-23.

[11] Anselmet, F., Gagne, Y., Hopfinger, E.J. and Antonia, R.A. (1984) “High-order velocity structure function in turbulent shear flow”. J. Fluid Mech. 140, 63-89.

[12] Cramér, H. (1938) “Sur un nouveau théorème-limite de la théorie des probabilités”, Actualités Scientifiques et Industrielle 736, 5-23.

[13] Kolmogorov, A.N. (1941) “On the logarithmic normal law of distribution of the size of particles under pulverization” Dokl. Akad Nauk SSSR 31, 99-101.

[14] Parisi, G. and Frisch, U. (1985) On the singularity structure of fully developed turbulence, in Turbulence and Predictability in Geophysical  Fluid Dynamics. Proceed Intern. School of Physics  ‘E. Fermi’. 1983. Varenna. Italy, 84-87.  This is an Appendix to the paper ‘Fully Developed Turbulence and Intermittency’ by Frisch, U. pp. 71-88.

[15] Benzi, Paladin, Parisi and Vulpiani (1984) “On the multifractal nature of fully developed turbulence and chaotic systems”, J. Phys. A: Math. Gen. 17, 3521.

[16] Frisch, U. (1995) “Turbulence, the Legacy of A.N. Kolmogorov”, Cambridge University Press.

[17] Meyer, Y. (2021) “Giorgio Parisi et la turbulence”,  Available in French on the site of the  “Institut national des sciences mathématiques et de leurs interactions” https://www.insmi.cnrs.fr/fr/cnrsinfo/giorgio-parisi-et-la-turbulence-par-yves-meyer .

[18] Jaffard, S. (2000) “On the Frisch-Parisi conjecture”, J. Math. Pures Appl. 69, 6, 525-552.


环境百科全书由环境和能源百科全书协会出版 (www.a3e.fr),该协会与格勒诺布尔阿尔卑斯大学和格勒诺布尔INP有合同关系,并由法国科学院赞助。

引用这篇文章: BENZI Roberto, FRISCH Uriel (2021年12月3日), On the contributions to statistical physics of Giorgio Parisi, Nobel Prize winner 2021, 环境百科全书,咨询于 2024年11月21日 [在线ISSN 2555-0950]网址: https://www.encyclopedie-environnement.org/en/physics/parisi-nobel-prize-physics-2021/.

环境百科全书中的文章是根据知识共享BY-NC-SA许可条款提供的,该许可授权复制的条件是:引用来源,不作商业使用,共享相同的初始条件,并且在每次重复使用或分发时复制知识共享BY-NC-SA许可声明。