Doubt in Example 4.3.4 of Brémaud's Markov Chains book

28 Views Asked by At

I'm reading Brémaud's Markov Chains, Gibbs Fields, Monte Carlo Simulation and Queues 2nd edition and I'm not following example 4.3.4 in page 159. Here's what's in it:

Let $A$ be a square matrix of dimension $r$ with $r$ distinct eigenvalues $\lambda_1 , \dots , \lambda_r$, and let $u_1 , \dots , u_r$ be the associated left eigenvectors and
$v_1 , \dots , v_r$ be the associated right eigenvectors. Then both the systems $u_1 , \dots , u_r$ and $v_1 , \dots , v_r$ are linearly independent. Moreover, $u^T_i v_j =0$ if $i \neq j$,and we can choose them in a way that $u^T_i v_i =1$ for each $1\leq i \leq r$, in which case we have the spectral decomposition: \begin{equation} A^n = \sum_{i=1}^r \lambda^n_i v_i u^T_i. \end{equation}

Example 4.3.4: The transition matrix on $E=\{1,2\}$, $$ P=\begin{bmatrix} 1-\alpha & \alpha\\ \beta & 1-\beta \end{bmatrix} $$ where $\alpha,\beta \in (0,1)$ has for characteristic polynomial $(1-\alpha-\lambda)(1-\beta-\lambda)-\alpha\beta$, which admits the roots $\lambda_1 =1$ and $\lambda_2 = 1-\alpha-\beta$. The right eigenvector associated with $\lambda_1 =1$ is $v=\mathbf{1}$, with all entries equal to $1$. Also, the stationary distribution $$ \pi^T = \left(\frac{\beta}{\alpha + \beta} , \frac{\alpha}{\alpha + \beta}\right) $$ is the left eigenvector corresponding to the eigenvalue $1$. In this example, the spectral decomposition takes the form: $$ P^n =\frac{1}{\alpha + \beta} \begin{bmatrix} 1-\alpha & \alpha \\ \beta & 1-\beta \end{bmatrix} + \frac{(1-\alpha-\beta)^n}{\alpha + \beta} \begin{bmatrix} \alpha & -\alpha \\ 1-\beta & -\beta \end{bmatrix} $$ and therefore, since $|1-\alpha-\beta|<1$, $$ \lim_{n \to \infty} P^n = \frac{1}{\alpha + \beta} \begin{bmatrix} 1-\alpha & \alpha \\ \beta & 1-\beta \end{bmatrix}. $$ In particular, the result of convergence to steady state, $$ \lim_{n \to \infty} {P}^n = \mathbf{1} \pi^T $$ was recovered for this special case in a purely algebraic way.

If I understood correctly, by the last line of the example we should have: $$ \lim_{n \to \infty} {P}^n = \mathbf{1} \pi^T = \frac{1}{\alpha + \beta}\begin{bmatrix} 1 \\ 1 \end{bmatrix}\begin{bmatrix} \beta & \alpha \end{bmatrix}=\frac{1}{\alpha + \beta}\begin{bmatrix} \beta & \alpha\\ \beta & \alpha\end{bmatrix}, $$ but this obviously doesn't coincide with the expression in the 3rd to last line: $$ \lim_{n \to \infty} P^n = \frac{1}{\alpha + \beta} \begin{bmatrix} 1-\alpha & \alpha \\ \beta & 1-\beta \end{bmatrix}. $$

Also, I've done the calculations myself and the spectral decomposition I got is: $$ P^n = \frac{1}{\alpha+\beta}\begin{pmatrix}{\beta} & {\alpha}\\{\beta} & {\alpha} \end{pmatrix} +\frac{(1-\alpha-\beta)^n}{\alpha+\beta}\begin{pmatrix}{\alpha} & -{\alpha}\\-{\beta} & {\beta}\end{pmatrix} $$ which, taking limit when $n \to \infty$, coincides with $\mathbf{1} \pi^T$. Are the spectral decomposition and the equation in the 3rd to last line of the example wrong?