JOVANA
Library Glossary Getting Started Three Levels Fields How it works Mission
Join the mission
Back to the library
數學 1963

確定性的非週期流

愛德華·洛倫茲

簡單而確定的方程,也能永遠無法預測。

Choose your version
In depth · the introduction

三個簡潔的方程,在 1960 年代一台小電腦上運行,證明了天氣可以是確定的,卻依然無法長期預測。

核心思想

愛德華·洛倫茲是一位氣象學家,他想找一個最簡單的對流玩具模型——暖空氣上升、冷空氣下沉。他把它濃縮成三個數,由三條精確的規則支配,處處沒有隨機。運行這些規則,未來在原則上是完全確定的。

可答案卻是不可預測的。若他讓模型從兩個幾乎相同的狀態出發,它們會同行一陣,然後漸漸剝離,直到毫無相似之處。起點上極小的不確定會不斷放大,最終淹沒一切——所以即便是一個完美的、確定性的模型,也只能看清前方一小段。

事情的經過

這個發現純屬偶然。1961 年,洛倫茲想重跑模擬的一部分,為省時間,他照著列印紙重新輸入了數字。列印紙只顯示三位小數,而電腦記憶體裡存的是六位。這點差別——百萬分之幾——竟足以讓重跑出來的「天氣」過一陣後變得面目全非。

起初他懷疑是真空管壞了。隨後他意識到:是機器對、自己的直覺錯了——這個系統對起點極其敏感。他把模型削減到三個方程,好把這一效應展示得盡可能乾淨,並於 1963 年發表在一份氣象學期刊上——而此後約十年間,幾乎沒有數學家注意到它。

為何重要

自牛頓以來的三個世紀裡,科學一直假設:精確的定律意味著可預測的世界——只要精確地知道現在,未來便會展開。洛倫茲指出其中有個陷阱。許多尋常的系統會把最微小的誤差放大得如此之快,以致長期預測在原則上、而非僅在實踐上不可能。「確定」與「可預測」就此悄然分家。

這重塑了天氣預報,也給了整個想法一個流行的名字——「蝴蝶效應」,源自洛倫茲後來的意象:一隻蝴蝶搧動翅膀,竟能推動遠方的一場風暴。

一個日常畫面

想像兩顆一模一樣的彈珠,在一座陡坡的最頂端、相距一髮之微處放手。起初它們並肩滾下。但坡上佈滿石子,每塊石子都把兩顆彈珠彈得略有不同。彈夠多次之後,一顆落進池塘,另一顆落進灌木叢——同樣的起點、同樣的坡、同樣的規則,命運卻天差地別。

大氣就是那道佈滿無數石子的山坡。蝴蝶的振翅,就是那一髮之微的領先——它決定了幾週之後,彈珠會落進哪一道山谷。

兩個彩色的點在同一條蝴蝶形曲線上奔跑。它們起初幾乎在一起、一同運動,隨後分開,最終落到兩翼。滑桿可改變它們起始的間距。

它的位置

洛倫茲把龐加萊在 1890 年代最早瞥見的隱憂——太陽系的三體問題或許不可預測——化為一個具體而可計算的東西。他的蝴蝶吸引子開啟了混沌這一領域,與曼德博的碎形、費根鮑姆通向混沌的普適道路並肩而立。

它也為本館中牛頓(1687)那座鐘錶般的宇宙悄悄劃下邊界:定律依舊精確,長期的預報卻不再可靠。洛倫茲所簡化的對流模型,倚仗的正是傅立葉(1822)所引入的那種逐模分解。

The original document
Original source text
Edward N. Lorenz · Journal of the Atmospheric Sciences 20 (1963): 130–141 · received 18 November 1962
Abstract
Finite systems of deterministic ordinary nonlinear differential equations may be designed to represent forced dissipative hydrodynamic flow. Solutions of these equations can be identified with trajectories in phase space. For those systems with bounded solutions, it is found that nonperiodic solutions are ordinarily unstable with respect to small modifications, so that slightly differing initial states can evolve into considerably different states. Systems with bounded solutions are shown to possess bounded numerical solutions.
A simple system representing cellular convection is solved numerically. All of the solutions are found to be unstable, and almost all of them are nonperiodic. The feasibility of very-long-range weather prediction is examined in the light of these results. (Concluding sentences of the abstract.)
1 · Phase space and the choice of a system
Lorenz frames the weather as a point moving through a phase space: every instantaneous state of the atmosphere is one point, and the governing equations carry it along a single trajectory. He restricts attention to forced dissipative systems — energy pumped in, energy bled out by friction — because their trajectories are eventually trapped inside a bounded region of that space.
[ … ]
6 · A simple convection model
Following Barry Saltzman, Lorenz reduces Rayleigh–Bénard convection — a fluid heated from below — to just three numbers. X measures how vigorously the fluid is rolling over; Y, the temperature difference between the rising and falling currents; Z, how far the vertical temperature profile departs from a straight line. Their rates of change are three coupled nonlinear equations (the convection constants enter as σ, r and b).
7 · Numerical integration
With σ = 10, r = 28 and b = 8/3, Lorenz integrates the system step by step on a Royal McBee LGP-30 desk computer. The trajectory never repeats and never settles; it loops about one centre, then crosses to loop about another, alternating an unpredictable number of times — the orbit that would later be drawn as the two-winged 'butterfly' attractor.
[ … ]
Conclusion
Because two trajectories that begin imperceptibly close drift apart until all resemblance is lost, any forecast started from imperfect measurements must eventually fail. Lorenz concludes that prediction of the sufficiently distant future is impossible by any method, unless the present conditions are known exactly — and in a system this sensitive, 'exactly' is unattainable.
Massachusetts Institute of Technology · 1963