91Ó°ÊÓ

Markov chains are a fundamental concept in probability theory and can be thought of as systems that undergo transitions from one state to another according to certain probability rules. Each state in a Markov chain represents a possible condition of the system, and the transitions between states are determined by a set of probabilities. These probabilities are often represented in the form of a matrix known as a stochastic matrix, where each element in the matrix indicates the probability of moving from one state to another.

For a matrix to be stochastic, each row must sum to one since it represents the total probability distribution of moving from an initial state to all possible subsequent states. When a stochastic matrix includes certain states that once entered, cannot be left—they are labelled as 'absorbing states'. In such matrices, certain rows will consist of a probability of 1 for the absorbing state and 0s for all other transitions. Understanding the structure and behavior of absorbing stochastic matrices is critical when studying Markov chains as they can model systems that will eventually reach an absorbing state, a common scenario in real-life processes.

Matrix partitioning is a useful technique in matrix algebra that involves breaking down a matrix into several smaller matrices or blocks. When dealing with absorbing stochastic matrices in the context of Markov chains, partitioning the matrix helps to isolate the parts of the matrix that correspond to the absorbing states from those that don't.

By rearranging rows and columns so that absorbing states come first, we divide the matrix into four blocks: two along the diagonal and two off-diagonal. The top-left block represents the absorbing states themselves, typically containing '1's down the main diagonal. The bottom-right block, often denoted as submatrix 'S', represents transitions between non-absorbing states. Meanwhile, the other two blocks represent the probabilities of transitions from non-absorbing states to absorbing states ('R'), and vice versa—which, in the case of an absorbing matrix, is always a block of zeros because once an absorbing state is reached, transitions out of it do not occur.

Once an absorbing stochastic matrix has been properly partitioned, two critical submatrices can be identified: 'R' and 'S'. Submatrix 'R' contains the transition probabilities from each non-absorbing state to each absorbing state—reflecting the potential for the system to be 'absorbed' into a final state from which it will not leave. The 'S' matrix, on the other hand, describes transitions between non-absorbing states only. This matrix is particularly important because it reveals the system's internal dynamics before any absorption happens.

Understanding these submatrices is essential in analyzing Markov processes because they allow us to calculate long-term behaviors, such as absorption probabilities and expected times until absorption. For instance, to find the absorption probabilities, one would need to calculate the fundamental matrix, typically found by taking the inverse of the matrix 'I - S', where 'I' stands for the identity matrix of the corresponding size. The 'R' matrix will then be used in conjunction with the fundamental matrix to determine the specific probabilities of ending up in the various absorbing states from any given non-absorbing state.