1: \begin{proof}

2: 	Denote $\barPhi_n = \Expect[ \Phi_n]$, and similarly, $\barPhi^1_n = \Expect[ \Phi^1_n]$, $\barPhi^2_n = \Expect[ \Phi^2_n]$.

3: 	Based on \eqref{e:PhinAn}, we have:

4: 	\begin{equation}

5: 	\begin{aligned}

6: 	\barPhi^2_{n+1} = & \,\, \Big (1 + \frac{1}{n + 1} \Big ) \big(I + A) \barPhi_n^2 + \frac{1}{\sqrt{n}} A \barPhi_n^1

7: 	\\

8: 	\approxeq  &  \sum_{k = 0}^{n - 1} \big(I + A)^k \frac{1}{\sqrt{n - k}} A \barPhi_{n - k}^1

9: 	\\

10: 	\approxeq  &   \frac{1}{\sqrt{n}} \bigg(  \sum_{k = 0}^{\infty} \big(I + A)^k \bigg) A \barPhi_{n}^1

11: 	\\

12: 	=  &   - \frac{1}{\sqrt{n}} A^{-1} A \barPhi_{n}^1

13: 	\\

14: 	=  &   - \frac{1}{\sqrt{n}} \barPhi_{n}^1

15: 	\end{aligned}

16: 	\label{e:Phin2}

17: 	\end{equation}

18: 	In \eqref{e:Phin2}, each of the approximations can be shown rigorously, using techniques that are very similar to the ones that were used in \Section{s:recursions}: The error in each approximation can be bounded as a constant times $\|\Phi_n\|/n$. %Specufically, the error terms can be bounded as ).

19: 	

20: 	Similarly, the recursion for $\barPhi^1_n$ satisfies:

21: 	\begin{equation}

22: 	\begin{aligned}

23: 	\barPhi^1_{n+1} = & \,\, \barPhi_n^1 + \frac{1}{\sqrt{n}} (I + A) \barPhi_n^2 +  \frac{1}{n + 1} \big( \half I + A) \barPhi_n^1

24: 	\\

25: 	\approxeq  &  \barPhi_n^1 - \frac{1}{{n}} (I + A) \barPhi_n^1 +  \frac{1}{n + 1} \big( \half I + A) \barPhi_n^1

26: 	\\

27: 	\approxeq  &  \barPhi_n^1 -  \frac{1}{2(n + 1)} \barPhi_n^1

28: 	%\\

29: 	%=  &   - \barPhi_n^1 + \frac{1}{\sqrt{n}} (I + A) \barPhi_n^2 +  \frac{1}{n + 1} \big( \half I + A) \barPhi_n^1

30: 	%\\

31: 	%=  &   - \barPhi_n^1 + \frac{1}{\sqrt{n}} (I + A) \barPhi_n^2 +  \frac{1}{n + 1} \big( \half I + A) \barPhi_n^1

32: 	\end{aligned}

33: 	\end{equation}

34: 	The above recursion can be viewed as a stochastic approximation algorithm (with $0$ noise). This implies:

35: 	$

36: 	\lim_{n\to\infty}  

37: 	\Expect[\Phi^1_n] =0

38: 	$.  Using this in \eqref{e:Phin2}, we also have $

39: 	\lim_{n\to\infty}  

40: 	\Expect[\Phi_n] =0

41: 	$.

42: \end{proof}

43: