Asymptotic property of current for a conduction model of
Fermi particles on finite lattice

Let us consider a many body system of Fermi particles moving under various potentials and noise on a one-dimensional finite lattice $[1,N]\cap\mathbb{Z}$ ($N\in\mathbb{N}$). The one-particle Hilbert space is $\mathbb{C}^{N}$. Denote its standard basis by $\{e_{n}\}_{n=1}^{N}$. The many body system is described by the creation and annihilation operators $a^{*}(f),a(f)$, where $f\in\mathbb{C}^{N}$. Let us write $a^{\#}_{n}$ for $a^{\#}(e_{n})$ as usual ($a^{\#}$ is $a^{*}$ or $a$). These operators satisfy the following canonical anti-commutation relations:

$$\{a^{*}_{i},a_{j}\}=\delta_{ij}I,\hskip 10.0pt\{a_{i},a_{j}\}=0,$$

where $\{A,B\}=AB+BA$, $\delta_{ij}$ is the Kronecker delta and $I$ is the identity operator on $\mathbb{C}^{N}$. In the sequel, we write shortly $c\times I$ as $c$ ($c\in\mathbb{C}$). Suppose that the total Hamiltonian is

$$H=\sum_{n=1}^{N-1}\left[-(a_{n}^{*}a_{n+1}+a_{n+1}^{*}a_{n})+v(n)a^{*}_{n}a_{n}\right],$$

where $v(\cdot)$ is a real-valued function called potential. Since we will consider the limit $N\to\infty$, potential $v$ is given as a bounded function on $\mathbb{N}$. Let $\mathcal{A}$ be the algebra generated by the creation and annihilation operators, and $\theta\colon\mathcal{A}\to\mathcal{A}$ be a *-automorphism determined by $\theta(a_{n}^{\#})=-a_{n}^{\#}$.

For real numbers $\alpha_{in}^{l},\alpha_{out}^{l},\alpha_{in}^{r},\alpha_{out}^{r},\beta$ greater than or equal to $0$ (at least one of $\alpha_{in}^{l},\alpha_{out}^{l},\alpha_{in}^{r},\alpha_{out}^{r}$ is not $0$), define a linear map $L\colon\mathcal{A}\to\mathcal{A}$ as

	$\displaystyle L(A)$	$\displaystyle=$	$\displaystyle i[H,A]+$
			$\displaystyle\alpha_{in}^{l}(2a_{1}\theta(A)a_{1}^{*}-\{a_{1}a_{1}^{*},A\})+\alpha_{out}^{l}(2a_{1}^{*}\theta(A)a_{1}-\{a_{1}^{*}a_{1},A\})$
			$\displaystyle+\alpha_{in}^{r}(2a_{N}\theta(A)a_{N}^{*}-\{a_{N}a_{N}^{*},A\})+\alpha_{out}^{r}(2a_{N}^{*}\theta(A)a_{N}-\{a_{N}^{*}a_{N},A\})$
			$\displaystyle+\beta\sum_{n=1}^{N}\left(a^{*}_{n}a_{n}Aa^{*}_{n}a_{n}-\frac{1}{2}\{a^{*}_{n}a_{n},A\}\right).$

Here, $[H,A]=HA-AH$.

It is obvious from the form of each term that $L$ generates a Quantum Dynamical Semigroup $\{e^{tL}\}_{t\geq 0}$ on $\mathcal{A}$ [10]. That is, $e^{tL}$ is a CP (completely positive) map preserving identity (state transformation) for every $t\geq 0$. The physical meaning of each term is as follows:

$i[H,A]$
This term represents the Hamiltonian dynamics of many particles moving independently by a one-particle Hamiltonian

$$(h\psi)(n)=-\psi(n+1)-\psi(n-1)+v(n)\psi(n)$$

($\psi(0)=\psi(N+1)=0$). It operates as

$$i[H,a^{*}(f)a(g)]=ia^{*}(hf)a(g)-ia^{*}(f)a(hg).$$

The terms with coefficients $\alpha_{in}^{l},\alpha_{out}^{l},\alpha_{in}^{r},\alpha_{out}^{r}$
These terms represent the effects of adding a particle to site 1, removing from site 1, adding to site $N$ and removing from site $N$, respectively. Put $p_{n}=|e_{n}\rangle\langle e_{n}|\ (n=1,2,\cdots,N)$, 1-rank projections corresponding to the basis $\{e_{n}\}_{n=1}^{N}$, then these terms operate as

$$2a_{1}\theta(a(f)a^{*}(g))a_{1}^{*}-\{a_{1}a_{1}^{*},a(f)a^{*}(g)\}=-a(p_{1}f)a^{*}(g)-a(f)a^{*}(p_{1}g),$$

$$2a_{1}^{*}\theta(a^{*}(f)a(g))a_{1}-\{a_{1}^{*}a_{1},a^{*}(f)a(g)\}=-a^{*}(p_{1}f)a(g)-a^{*}(f)a(p_{1}g)$$

(replace $p_{1}$ by $p_{N}$ in the case of $N$). The dynamics generated by these terms and $i[H,A]$ is a special case of those in [7, 9].

The term with coefficient $\beta$ (dephasing noise)
This term represents noise from the environment called dephasing noise. Dephasing noise preserves the number of particles and destroys the coherence. Let us check this property. Denote

$$D_{n}(A)=2a^{*}_{n}a_{n}Aa^{*}_{n}a_{n}-\{a^{*}_{n}a_{n},A\}$$

($n=1,2,\cdots,N$), then since $D_{n}$ commutes with each other, we have

$$e^{t\sum_{n=1}^{N}D_{n}}=\prod_{n=1}^{N}e^{tD_{n}}.$$

Easy calculation shows that

$$D_{n}(a^{*}_{i}a_{j})=\begin{cases}0&(i=j=n,\ i,j\neq n)\\ -a^{*}_{i}a_{j}&(\mathrm{otherwise}),\end{cases}$$

$$e^{tD_{n}}(a^{*}_{i}a_{j})=\begin{cases}a^{*}_{i}a_{j}&(i=j=n,\ i,j\neq n)\\ e^{-t}a^{*}_{i}a_{j}&(\mathrm{otherwise}).\end{cases}$$

Recall that for every state $\omega$ on $\mathcal{A}$, its two-point function is described by a positive operator on $\mathbb{C}^{N}$: there is an operator $R\colon\mathbb{C}^{N}\to\mathbb{C}^{N}$ such that $0\leq R\leq I$ and

$$\omega(a^{*}_{i}a_{j})=\langle e_{j},Re_{i}\rangle.$$

In the one-particle system, define a linear map $d_{n}\colon M_{N}(\mathbb{C})\to M_{N}(\mathbb{C})$ as

$$d_{n}(a)=2p_{n}ap_{n}-\{p_{n},a\},\ a\in M_{N}(\mathbb{C})$$

($M_{N}(\mathbb{C})$ is the set of $N\times N$ complex matrices), then

$$e^{td_{n}}(a)=a+(1-e^{-t})d_{n}(a).$$

If $i=j=n$ or $i,j\neq n$,

$$\langle e_{j},e^{td_{n}}(R)e_{i}\rangle=\langle e_{j},Re_{i}\rangle=\omega(e^{tD_{n}}(a^{*}_{i}a_{j})),$$

and if one of $i,j$ is $n$,

$$\langle e_{j},e^{td_{n}}(R)e_{i}\rangle=\langle e_{j},Re_{i}\rangle+(1-e^{-t})\langle e_{j},(-R)e_{i}\rangle=\omega(e^{tD_{n}}(a^{*}_{i}a_{j})).$$

Therefore, the dynamics of the two-point function is described by $\displaystyle\prod^{N}_{n=1}e^{td_{n}}(R)$:

$$\omega\left(e^{t\sum_{n=1}^{N}D_{n}}(a^{*}_{i}a_{j})\right)=\omega\left(\prod_{n=1}^{N}e^{tD_{n}}(a^{*}_{i}a_{j})\right)=\left\langle e_{j},\prod_{n=1}^{N}e^{td_{n}}(R)e_{i}\right\rangle=\left\langle e_{j},e^{t\sum_{n=1}^{N}d_{n}}(R)e_{i}\right\rangle.$$

Set $d(a)=\displaystyle\frac{1}{2}\sum_{n=1}^{N}d_{n}(a)=\sum_{n=1}p_{n}ap_{n}-a$, then $d^{2}=-d$ and

$$e^{td}(a)=a+(1-e^{-t})d(a)=e^{-t}a+(1-e^{-t})\sum_{n=1}^{N}p_{n}ap_{n}.$$

The pure state $|\psi\rangle\langle\psi|$ is transformed to

$$e^{-t}|\psi\rangle\langle\psi|+(1-e^{-t})\sum_{n=1}^{N}|\psi(n)|^{2}p_{n}.$$

Thus, this dynamics destroys the coherence and transform a state to a convex combination of localized states $p_{n}$.

$L$ consists of the above three types of terms. Stationary current induced by the dynamics $e^{tL}$ is the main topic in this paper. From the above discussions it turns out that the dynamics of the two-point function is described by that of the one-particle system. Suppose that the two-point function of the state $\omega\circ e^{tL}$ is expressed as

$$\omega\circ e^{tL}(a^{*}_{i}a_{j})=\langle e_{j},R(t)e_{i}\rangle,$$

then by calculating

$$\frac{d}{dt}\omega\circ e^{tL}(a^{*}_{i}a_{j})=\omega\circ e^{tL}(L(a^{*}_{i}a_{j})),$$

we obtain the following differential equation for $R(t)$:

	$\displaystyle\frac{d}{dt}R(t)$	$\displaystyle=$	$\displaystyle-i[h,R(t)]-\{(\alpha_{in}^{l}+\alpha_{out}^{l})p_{1}+(\alpha_{in}^{r}+\alpha_{out}^{r})p_{N},R(t)\}$
			$\displaystyle+\beta\left(\sum_{n=1}^{N}p_{n}R(t)p_{n}-R(t)\right)+2\alpha_{in}^{l}p_{1}+2\alpha_{in}^{r}p_{N},$

$$R(0)=R.$$

It is easy to check that

$$R(t)=T_{t}(R)+\int^{t}_{0}T_{s}(2\alpha_{in}^{l}p_{1}+2\alpha_{in}^{r}p_{N})ds$$

is a solution of this equation, where $T_{t}$ is an operator semigroup on $M_{N}(\mathbb{C})$ generated by

$$l\colon a\mapsto-i[h,a]-\{(\alpha_{in}^{l}+\alpha_{out}^{l})p_{1}+(\alpha_{in}^{r}+\alpha_{out}^{r})p_{N},a\}+\beta\left(\sum_{n=1}^{N}p_{n}ap_{n}-a\right).$$

$T_{t}=e^{tl}$ is a CP map which does not preserve identity.

Let us consider the long time limit $t\to\infty$. In the case where $\beta=0$,

$$T_{t}(a)=e^{-ith_{D}}ae^{ith_{D}^{*}}$$

for

$$h_{D}=h-i(\alpha_{in}^{l}+\alpha_{out}^{l})p_{1}-i(\alpha_{in}^{r}+\alpha_{out}^{r})p_{N}.$$

Since the imaginary part of every eigenvalue of $h_{D}$ is less than $0$, $\displaystyle\lim_{t\to\infty}e^{-ith_{D}}=0$. Thus, we get

$$\lim_{t\to\infty}R(t)=\int^{\infty}_{0}T_{s}(2\alpha_{in}^{l}p_{1}+2\alpha_{in}^{r}p_{N})ds\equiv R_{\infty}.$$

The integral of the right hand side converges, because

	$\displaystyle 0\leq\int^{t_{1}}_{t_{2}}T_{s}(2\alpha_{in}^{l}p_{1}+2\alpha_{in}^{r}p_{N})ds$	$\displaystyle\leq$	$\displaystyle\int^{t_{1}}_{t_{2}}T_{s}(2(\alpha_{in}^{l}+\alpha_{out}^{l})p_{1}+2(\alpha_{in}^{r}+\alpha_{out}^{r})p_{N})ds$
		$\displaystyle=$	$\displaystyle\left[T_{s}(I)\right]^{t_{1}}_{t_{2}}\to 0\ (t_{1},t_{2}\to\infty).$

Note that $R_{\infty}$ does not depend on $R$. This means that whatever the initial state is, the two-point function converges to the same value $\langle e_{j},R_{\infty}e_{i}\rangle$. Moreover, it can be shown that every state converges to the quasi free state determined by this two-point function [9].

In the case where $\beta>0$, we have the same result for the two-point function.

By this theorem, in the case where $\beta>0$ we also have

$$\lim_{t\to\infty}R(t)=\int^{\infty}_{0}T_{s}(2\alpha_{in}^{l}p_{1}+2\alpha_{in}^{r}p_{N})ds\equiv R_{\infty}.$$

In this subsection we will focus on current. Since current is expressed by two-point function, it converges to a constant in the limit $t\to\infty$. We will consider how the sign of the current is determined by the relation of constants $\alpha_{in}^{l},\alpha_{out}^{l},\alpha_{in}^{r},\alpha_{out}^{r}$. The current is shown to be expressed by a simple formula (Theorem 2.2).

At first, recall that the observable of current from site $n$ to $n+1$ is

$$j_{n}=-i(a_{n}^{*}a_{n+1}-a^{*}_{n+1}a_{n}).$$

As shown in the previous subsection, for any state $\omega$ the limit $\displaystyle\lim_{t\to\infty}\omega\circ e^{tL}(j_{n})$ exists and is independent of $\omega$. In fact it does not depend on $n$. Let us check it. By the definition of generator, for any $\epsilon>0$ there is $h>0$ such that

$$\left\|\frac{e^{hL}(a_{n}^{*}a_{n})-a^{*}_{n}a_{n}}{h}-L(a^{*}_{n}a_{n})\right\|<\epsilon.$$

Thus we have

$$|\omega\circ e^{tL}(L(a^{*}_{n}a_{n}))|<\epsilon+\left|\omega\circ e^{tL}\left(\frac{e^{hL}(a_{n}^{*}a_{n})-a^{*}_{n}a_{n}}{h}\right)\right|,\ ^{\forall}t\geq 0$$

and $\displaystyle\limsup_{t\to\infty}|\omega\circ e^{tL}(L(a^{*}_{n}a_{n}))|\leq\epsilon$. Since $\epsilon$ is arbitrary, $\displaystyle\lim_{t\to\infty}\omega\circ e^{tL}(L(a^{*}_{n}a_{n}))=0$. This equation and

$$L(a^{*}_{n}a_{n})=-i(a_{n-1}^{*}a_{n}-a^{*}_{n}a_{n-1})+i(a^{*}_{n}a_{n+1}-a^{*}_{n+1}a_{n})=j_{n-1}-j_{n},\ n=2,\cdots,N-1$$

show that the limit of the current does not depend on $n$. We denote the limit of current by $\mathcal{J}_{\beta}(N)=\displaystyle\lim_{t\to\infty}\omega\circ e^{tL}(j_{1})$ (it depends on the sample size $N$). Then it is expressed as

	$\displaystyle\mathcal{J}_{\beta}(N)$	$\displaystyle=$	$\displaystyle-i\langle e_{2},R_{\infty}e_{1}\rangle+i\langle e_{1},R_{\infty}e_{2}\rangle$
		$\displaystyle=$	$\displaystyle 2\mathrm{Im}\langle e_{2},R_{\infty}e_{1}\rangle.$

$\mathcal{J}_{\beta}(N)$ has the following simple expression. This is one of our main results in this paper.

Since $\int^{\infty}_{0}\langle e_{1},T_{t}(p_{N})e_{1}\rangle dt>0$, the sign of $\mathcal{J}_{\beta}(N)$ is completely determined by the coefficient $\alpha_{in}^{l}\alpha_{out}^{r}-\alpha_{out}^{l}\alpha_{in}^{r}$. Let us check that $\int^{\infty}_{0}\langle e_{1},T_{t}(p_{N})e_{1}\rangle dt>0$.

	$\displaystyle\langle e_{1},T_{t}(p_{N})e_{1}\rangle$	$\displaystyle=$	$\displaystyle\sum_{n=0}^{\infty}\left\langle e_{1},\frac{(tl)^{n}}{n!}(p_{N})e_{1}\right\rangle$
		$\displaystyle=$	$\displaystyle\left\langle e_{1},\frac{(tl)^{2N}}{(2N)!}(p_{N})e_{1}\right\rangle+\sum_{n=2N+1}^{\infty}\left\langle e_{1},\frac{(tl)^{n}}{n!}(p_{N})e_{1}\right\rangle$
		$\displaystyle=$	$\displaystyle\frac{t^{2N}}{(2N)!}{}_{2N}\mathrm{C}_{N}\langle e_{1},h^{N}p_{N}h^{N}e_{1}\rangle+\sum_{n=2N+1}^{\infty}\left\langle e_{1},\frac{(tl)^{n}}{n!}(p_{N})e_{1}\right\rangle$
		$\displaystyle=$	$\displaystyle\frac{t^{2N}}{(N!)^{2}}+\sum_{n=2N+1}^{\infty}\left\langle e_{1},\frac{(tl)^{n}}{n!}(p_{N})e_{1}\right\rangle.$

Since

$$\left|\sum_{n=2N+1}^{\infty}\left\langle e_{1},\frac{(tl)^{n}}{n!}(p_{N})e_{1}\right\rangle\right|\leq t^{2N+1}\sum_{n=2N+1}^{\infty}\left\langle e_{1},\frac{l^{n}}{n!}(p_{N})e_{1}\right\rangle$$

holds for $0\leq t<1$, for sufficiently small $t>0$ we have

$$\langle e_{1},T_{t}(p_{N})e_{1}\rangle>0.$$

In this subsection, we first prove the following proposition, which is applicable to arbitrary potentials.

Using this formula, we relate the current $\mathcal{J}_{\beta}(N)$ to transfer matrix. In addition, in case of dynamically defined potential, such as the Anderson model, the scaling of the asymptotic behavior is shown to be related with the Lyapunov exponent.

Recall that in noiseless case

$$la=-i(h_{D}a-ah^{*}_{D}).$$

As mentioned before, the imaginary part of every eigenvalue of $h_{D}$ is less than 0, and $h_{D}$ is invertible. Let us prepare a lemma.