1. The document provides a historical introduction to stochastic processes and Brownian motion. It discusses key figures like Robert Brown, Louis Bachelier, and Albert Einstein and their contributions to understanding random particle motion.
2. Einstein developed the idea that Brownian motion results in a Gaussian distribution, with variance increasing linearly over time. Bachelier also noticed properties like the Markov property of Brownian motion.
3. Formally, a stochastic process is defined as a collection of random variables indexed by time, defined on a common probability space. The challenge is developing properties like how sample paths vary over time from a fixed starting point.
1. The document provides a historical introduction to stochastic processes and Brownian motion. It discusses key figures like Robert Brown, Louis Bachelier, and Albert Einstein and their contributions to understanding random particle motion.
2. Einstein developed the idea that Brownian motion results in a Gaussian distribution, with variance increasing linearly over time. Bachelier also noticed properties like the Markov property of Brownian motion.
3. Formally, a stochastic process is defined as a collection of random variables indexed by time, defined on a common probability space. The challenge is developing properties like how sample paths vary over time from a fixed starting point.
1. The document provides a historical introduction to stochastic processes and Brownian motion. It discusses key figures like Robert Brown, Louis Bachelier, and Albert Einstein and their contributions to understanding random particle motion.
2. Einstein developed the idea that Brownian motion results in a Gaussian distribution, with variance increasing linearly over time. Bachelier also noticed properties like the Markov property of Brownian motion.
3. Formally, a stochastic process is defined as a collection of random variables indexed by time, defined on a common probability space. The challenge is developing properties like how sample paths vary over time from a fixed starting point.
that the word “speed” is not really appropriate.) Ein-
IV.24 Stochastic Processes stein’s method was based on considerations of statis- Jean-François Le Gall tical physics, which led him to the heat equation [I.3 §5.4] and then to the Gaussian density that solves this equation (see section 5.2). 1 Historical Introduction A few years before Einstein, the French mathemati- Stochastic processes are one of the major themes of cian Louis Bachelier, in his work about the mathemat- modern probability theory. Roughly speaking, they are ical modeling of stock markets, had already noticed mathematical models that describe the evolution of the Gaussian distribution of Brownian motion. How- random phenomena as time goes by. In this article, we ever, Bachelier was dealing not with the physical phe- shall introduce and illustrate the fundamental ideas of nomenon known as Brownian motion, but rather with the theory of stochastic processes by concentrating on random walks where the step size was very small. As the single most important example: Brownian motion. we shall see in sections 2 and 3, the two concepts are essentially equivalent from a mathematical viewpoint. We start with a brief historical introduction, in order to Bachelier pointed out what we call today the Markov provide some motivation for the mathematical theory property of Brownian motion: if we wish to predict the that follows. displacement after time t of a Brownian particle, then In 1828, the British botanist Robert Brown observed knowledge of the path followed by the particle before the very irregular and wiggly motion of small particles time t does not help us any more than just knowing of pollen suspended in water. Brown pointed out the the position at time t. Bachelier’s arguments were not unpredictable character of the motion, which appeared completely satisfactory, and his ideas were not fully to obey no known physical rule. During the nineteenth appreciated in his time. century, several physicists tried to understand the ori- How does one go about modeling a particle that gin of this “Brownian motion,” which turned out to moves in a random way? A first remark is that the posi- be present in many other physical phenomena. Several tion of the particle at time t will be a random variable theories were proposed, some of them rather fanci- [III.71 §4] Bt . But these random variables will depend on ful: perhaps Brownian particles were living microscopic each other: if you know where the particle is at time t, animals, or perhaps the motion was due to electro- it will affect your knowledge of how likely it is to be in static forces. By the end of the century, however, physi- a certain region at some later time. These two consid- cists had concluded that the constant changes of direc- erations can be accommodated if we take as our basic tion in Brownian motion could be explained by the model a set of random variables Bt , one for each non- impacts on a particle from the molecules of the sur- negative real number, all defined on the same underly- rounding medium. If the particle was sufficiently light, ing probability space. This, formally speaking, is what then these numerous collisions could have a macro- a stochastic process is. scopic influence on its displacement. This explanation This may seem a rather simple definition, but in order was also consistent with the experimental observation for a stochastic process to be interesting it needs to that the motion became faster if the temperature of the have additional properties, and difficult mathematical water, and thus the thermal agitation of its molecules, questions arise as soon as one tries to obtain them. Let increased. us write Ω for the underlying probability space. Then Albert Einstein, in one of his three famous 1905 each of the random variables Bt is a function from Ω to papers, was responsible for a major step forward in R3 , and therefore we associate a point in R3 with each the understanding of Brownian motion. He worked out pair (t, ω) (where t is a positive real number and ω that if a Brownian particle starts at the origin, then belongs to Ω). So far we have thought about the prob- after a fixed time t its position should be randomly dis- ability distribution of Bt , so we have been focusing on tributed according to the (three-dimensional) gaussian what happens when we fix t and let ω vary. However, distribution [III.71 §5] with mean 0 and variance σ 2 t, we must also consider what happens when we look at where σ 2 is a constant, called the diffusion constant, a “single instance” of a stochastic process, by fixing ω that measures how quickly the distribution spreads and letting t vary. For fixed ω, the function that takes t out with time. (One can think of this loosely as the to Bt (ω) is called a sample path. If we want a rigorous speed of the Brownian motion, but we shall see later mathematical theory of Brownian motion, then a very