Created: 2026-03-06 07:59:30
Updated: 2026-03-06 08:16:07

3 PI quantization of a Bose particle

Suppose at tit_{i}, the particle is at xix_{i}. The probability amplitude of finding particle at xfx_{f} at tft_{f} is

<xf,tfxi,ti>\begin{equation} \left<x_{f},t_{f}|x_{i},t_{i}\right> \end{equation}

where the vectors are defined in Heisenberg picture

x^(ti)xi,ti=xixi,ti,x^(tf)xf,tf=xfxf,tf\begin{equation} \begin{aligned} \hat{x}(t_{i})\ket{x_{i},t_{i}} = x_{i}\ket{x_{i},t_{i}} , \\ \hat{x}(t_{f}) \ket{x_{f},t_{f}} = x_{f} \ket{x_{f},t_{f}} \end{aligned} \end{equation}

Rewrite the probability amplitude in terms of Schrodinger picture: let x^=x^(0)\hat{x}=\hat{x}(0) be position operator with x^x=xx\hat{x}\ket{x}=x\ket{x}, since x^\hat{x} has no time dependence, its eigenvector should be time independent. If

x^(ti)=eiH^tix^eiH^ti\begin{equation} \hat{x}(t_{i}) = e^{i \hat{H} t_{i}} \hat{x} e^{-i \hat{H} t_{i}} \end{equation}

is substituted, we have

eiH^tix^eiH^tixi,ti=xixi,tix^[eiH^tixi,ti]=xi[eiH^tixi,ti]\begin{equation} \begin{aligned} e^{i \hat{H}t_{i}}\hat{x} e^{-i \hat{H}t_{i}}\ket{x_{i},t_{i}} & = x_{i}\ket{x_{i},t_{i}} \\ \hat{x} [e^{-i \hat{H}t_{i}}\ket{x_{i},t_{i}} ] & = x_{i} [e^{-i \hat{H}t_{i}}\ket{x_{i},t_{i}} ] \end{aligned} \end{equation}

so xi,ti=eiHtixi\ket{x_{i},t_{i}}=e^{iH t_{i}}\ket{x_{i}} and xf,tf=eiHtfxf\ket{x_{f},t_{f}}= e^{iHt_{f}}\ket{x_{f}}, so

xf,tfxi,ti=xfeiH^(tfti)xi=h(xf,xi;i(tfti))\begin{equation} \bra{x_{f},t_{f}} x_{i},t_{i}\rangle = \bra{x_{f}} e^{-i \hat{H}(t_{f}-t_{i})} \ket{x_{i}} =h(x_{f},x_{i};i(t_{f}-t_{i})) \end{equation}

define heat kernel of H^\hat{H} h(x,y;β)=xeHβyh(x,y;\beta)=\bra{x}e^{-H\beta}\ket{y}.

Suppose Hamiltonian is of the form H=p22m+V(x)H=\frac{p^{2}}{2m}+V(x)

\fcolorbox{pink}{}{Lemma 1.2}
Let a>0a>0. Then

exp(iap2)dp=πia\begin{equation} \int _{-\infty}^{\infty} \exp\left(-iap^{2} \right)\, dp = \sqrt{ \frac{\pi}{ia} } \end{equation}

Note: use the following contour.

\fcolorbox{pink}{}{End Lemma}

For any x,yRx,y\in \mathbb{R},

xeiHϵy=12πiϵexp[iϵ{m2((xy)ϵ)2V(x+y)2}+O(ϵ2)+O(ϵ(xy)2)]\begin{equation} \bra{x} e^{-iH\epsilon}\ket{y} = \frac{1}{\sqrt{ 2\pi i\epsilon }} \exp\left[i\epsilon\left\{ \frac{m}{2}\left( \frac{\left(x-y\right)}{\epsilon} \right)^{2} - \frac{V(x+y)}{2}\right\} + O(\epsilon ^{2}) + O(\epsilon(x-y)^{2})\right] \end{equation}

(Simpler derivation without order consideration):

<xexp(iHϵ)y>=dk<xexp(iϵH)k><ky>=dk2πeikyeikxeiϵHx\begin{equation} \left<x|\exp(-iH\epsilon)|y\right> = \int dk \left<x| \exp(-i\epsilon H) |k\right> \left<k|y\right> = \int \frac{dk}{2\pi} e^{-iky}e^{ikx}e^{-i\epsilon H_{x}} \end{equation}

where Hx=k22m+V(x)H_{x} = \frac{k^{2}}{2m} +V(x) since

iϵk22m+ik(xy)iϵV(x)=iϵ2m(kmϵ(xy))2+i2ϵm(xy)2iϵV(x)\begin{equation} -\frac{i\epsilon k^{2}}{2m} + ik(x-y) - i\epsilon V(x) = -\frac{i\epsilon}{2m}\left( k- \frac{m}{\epsilon}(x-y) \right)^{2} + \frac{i}{2\epsilon}m(x-y)^{2} - i\epsilon V(x) \end{equation}

after integration,

LHS=πi(ϵ/2m)12πexp[iϵV(x)+iϵ(m2(xyϵ)2)]=m2πiϵexp[iϵV(x)+iϵ(m2(xyϵ)2)]\begin{equation} \begin{aligned} LHS & = \sqrt{ \frac{\pi}{i (\epsilon / 2m)} } \frac{1}{2\pi} \exp \left[ -i\epsilon V(x) + i\epsilon \left( \frac{m}{2}\left( \frac{x-y}{\epsilon}\right)^{2} \right)\right] \\ & = \sqrt{ \frac{m}{2\pi i\epsilon} } \exp \left[ -i\epsilon V(x) + i\epsilon \left( \frac{m}{2}\left( \frac{x-y}{\epsilon}\right)^{2} \right)\right] \end{aligned} \end{equation}

After repeating,

<xf,tfxi,ti>=Dxexp[ititfdtL(x,x˙)],L(x,x˙)m2x˙2V(x)\begin{equation} \left<x_{f},t_{f}|x_{i},t_{i}\right> = \int \mathcal{D}x \exp\left[ i \int _{t_{i}}^{t_{f}} \, dt L(x,\dot{x}) \right] , L(x,\dot{x})\equiv \frac{m}{2}\dot{x}^{2} - V(x) \end{equation}

3 Imaginary time, partition function

Suupose HH is positive definite: SpecH={0<E0E1}\mathrm{Spec} H = \{0<E_{0}\leq E_{1}\leq\dots\}, suppose for simplicity that ground state is not degenerate. Then the spectral decomposition of eiHte^{-iHt} is given by

eiHt=neiEntnn\begin{equation} e^{-iHt}=\sum_{n} e^{-iE_{n}t} \ket{n} \bra{n} \end{equation}

and is analytic in lower half-plane of tt, where Hn=EnnH\ket{n}=E_{n}\ket{n}. Introduce wick rotation

t=iτ,τR+\begin{equation} t=-i\tau, \tau\in \mathbb{R}_{+} \end{equation}

where τ\tau is regarded as imaginary time, which is also known as Euclidean time since t2x2=(τ2+x2)t^{2}-\vec{x}^{2}= -(\tau ^{2}+\vec{x}^{2}).

x˙=dxdt=idxdτeiHt=eHτ\begin{equation} \begin{aligned} \dot{x}= \frac{dx}{dt}=i \frac{dx}{d\tau} \\ e^{-iHt}=e^{-H\tau} \end{aligned} \end{equation}

and thus

<xf,τfxi,τi>=<xfeH(τfτi)xi>=Dˉxeτiτfdτ[(1/2m)(dx/dτ)2+V(x)]\begin{equation} \left<x_{f},\tau_{f}|x_{i},\tau_{i}\right> = \left<x_{f}|e^{-H (\tau_{f}-\tau_{i})}|x_{i}\right>= \int \mathcal{\bar{D}}x e^{-\int _{\tau_{i}}^{\tau_{f}} \, d\tau \left[ \left(1/2 m\right) (dx/d\tau)^{2} + V(x)\right]} \end{equation}

For a given Hamiltonian, the partition function is

Z(β)=TreβH(β>0)\begin{equation} Z(\beta)= \mathrm{Tr} e^{-\beta H}\qquad\left(\beta>0\right) \end{equation}

in eigenvector basis,

Z(β)=dx<xeβHx>=periodicDˉxexp{0βdτ(12mx˙2+V(x))}\begin{equation} Z(\beta) = \int \, dx \left<x | e^{-\beta H}|x\right> = \int _{\text{periodic}} \mathcal{\bar{D}}x \exp \left\{ - \int _{0}^{\beta} \, d\tau \left( \frac{1}{2}m\dot{x}^{2} + V(x) \right)\right\} \end{equation}

Time ordered product, generating functional

T[A(t1)B(t2)]=A(t1)B(t2)θ(t1t2)+B(t2)A(t1)θ(t2t1)\begin{equation} T[A(t_{1})B(t_{2})] = A(t_{1})B(t_{2})\theta(t_{1}-t_{2})+ B(t_{2})A(t_{1})\theta(t_{2}-t_{1}) \end{equation}

Generalization to more than 3 operators should be trivial. Operators in the bracket are rearranged so that the time parameters decrease from left to right. The nn-operator T-product has n!n! terms, each with n1n-1 heaviside functions.

COnsider

<xf,tfT[x^(t1)x^(t2)x^(tn)]xi,ti>,\begin{equation} \left<x_{f},t_{f}\mid T[\hat{x}(t_{1})\hat{x}(t_{2})\dots \hat{x}(t_{n})]\mid x_{i},t_{i}\right> ,\qquad \end{equation}

Suppose (ti<t1<<tn<tf)(t_{i}<t_{1}<\dots<t_{n}<t_{f}). Inserting completeness relation

1=dxkxk,tkxk,tk\begin{equation} 1= \int _{-\infty}^{\infty} \, dx_{k} \ket{x_{k},t_{k}} \bra{x_{k},t_{k}} \end{equation}

,

<xf,tfx^(tn)x^(t1)xi,ti>=dx1dxnx1xn<xf,tfxn,tn><x1,t1xi,ti>=Dxx(t1)x(tn)eiS\begin{equation} \begin{aligned} \left<x_{f},t_{f}| \hat{x}(t_{n})\dots \hat{x}(t_{1}) | x_{i},t_{i}\right> & =\int \, dx_{1}\dots dx_{n} x_{1}\dots x_{n} \left<x_{f},t_{f} | x_{n},t_{n}\right> \dots \left<x_{1},t_{1}| x_{i},t_{i}\right> \\ & = \int \mathcal{D}x x(t_{1}) \dots x(t_{n}) e^{iS} \end{aligned} \end{equation}

in the second equality we expressed each <xk,tkxk1,tk1>\left<x_{k},t_{k}|x_{k-1},t_{k-1}\right> as path integral. Note that x^(tk)\hat{x}(t_{k}) in the LHS is a Heisenberg operator, and x(tk)=xkx(t_{k})=x_{k} on the RHS is a real value of path x(t)x(t).

Define generating functional Z[J]Z[J] to obtain the matrix elements of TT-products efficiently. Couple J(t)J(t) to coordinate x(t)x(t):

S[x(t),J(t)]=titfdt[12mx˙2V(x)+xJ]\begin{equation} S[x(t),J(t) ] = \int _{t_{i}}^{t_{f}} \, dt \left[ \frac{1}{2}m\dot{x}^{2}-V(x)+xJ\right] \end{equation}

The translation amplitude in presence of J(t)J(t) is denoted by xf,tfxi,ti\langle x_{f},t_{f}|x_{i},t_{i}\rangle, so

<xf,tfT[x(tn)x(t1)]xi,ti>=(i)nδnδJ(t1)δJ(tn)DxeiS[x(t),J(t)]J=0\begin{equation} \left<x_{f},t_{f}|T[x(t_{n})\dots x(t_{1})]|x_{i},t_{i}\right> = (-i)^{n} \frac{\delta ^{n}}{\delta J(t_{1})\dots\delta J(t_{n})} \left. \int \, \mathcal{D}x e^{iS[x(t),J(t)]} \right|_{J=0} \end{equation}

Suppose the system under consideration is in ground state 0\ket{0} at tit_{i} and calculate probability amplitude with which the system is also in ground state at tft_{f}. Suppose J(t)0J(t)\neq 0 only for [a,b][ti,tf][a,b] \subset [t_{i},t_{f}]. The transition amplitude in presence of J(t)J(t) may be obtained from HJ=Hx(t)J(t)H^{J}=H-x(t)J(t) and the unitary operator UJ(tf,ti)U^{J}(t_{f},t_{i}) of Hamiltonian.

<xf,tfxi,ti>=<xfUJ(tf,ti)xi>=xfeiH(tfb)UJ(b,a)eiH(ati)xi=mnxfeiH(tfb)mmUJ(b,a)nneiH(ati)xi=eiEm(tfb)eiEn(ati)<xfm><nxi>mUJ(b,a)n\begin{equation} \begin{aligned} \left<x_{f},t_{f}|x_{i},t_{i}\right> & = \left<x_{f}| U^{J}(t_{f},t_{i})|x_i\right> \\ & = \bra{x_{f}} e^{-iH(t_{f}-b)} U^{J}(b,a) e^{-iH(a-t_{i})} \ket{x_{i}} \\ & = \sum_{mn} \bra{x_{f}} e^{-iH(t_{f}-b)}\ket{m} \bra{m} U^{J}(b,a)\ket{n} \bra{n} e^{-iH(a-t_{i})} \ket{x_{i}} \\ & = e^{-iE_{m}(t_{f}-b)}e^{-iE_{n}(a-t_{i}) } \left<x_{f}|m\right> \left<n|x_{i}\right> \bra{m} U^{J}(b,a) \ket{n} \end{aligned} \end{equation}

where m,n\ket{m},\ket{n}are energy vectors. Next rotate the time variable tiτt\to-i\tau, eiEteEτe^{-iEt}\to e^{-E\tau}, then take limit τf,τi\tau_{f}\to \infty,\tau_{i}\to-\infty, which picks up ground states m=n=0m=n=0. Thus

limtf,ti<xf,tfxi,ti>J=<xf0><0xi>Z[J]\begin{equation} \lim_{ t_{f} \to \infty ,t_{i}\to-\infty} \left<x_{f},t_{f}|x_{i},t_{i}\right>_{J} = \left<x_{f}|0\right> \left<0|x_{i}\right>Z[J] \end{equation}

where we have defined generating functional Z[J]=0UJ(b,a)0=limtf,ti0UJ(tf,ti)0Z[J]= \bra{0} U^{J}(b,a) \ket{0} = \lim_{ t_{f} \to \infty, t_{i} \to -\infty } \bra{0}U^{J}(t_{f},t_{i})\ket{0}, and

Z[J]=limtf,ti<xf,tfxi,ti>J<xf0><0xi>\begin{equation} Z[J] = \lim_{ t_{f} \to \infty ,t_{i}\to-\infty} \frac{\left<x_{f},t_{f}|x_{i},t_{i}\right>_{J}}{\left<x_{f}|0\right> \left<0|x_{i}\right>} \end{equation}

The denominator is a constant independent of JJ. Now we have PI representation of Z[J]Z[J]:

Z[J]=NDx eiS[x,J]\begin{equation} Z[J] = \mathcal{N} \int \mathcal{D}x\ e^{iS[x,J]} \end{equation}

where nomalization constant N\mathcal{N} is chosen so that Z[0]=1Z[0]=1. Z[J]Z[J] generates matrix elements of TT-product between ground states:

<0T[x(t1)x(tn)]0>=(i)nδnδJ(t1)δJ(tn)Z[J]J=0\begin{equation} \left<0| T[x(t_{1})\dots x(t_n)]|0\right> = (-i)^{n} \frac{\delta ^{n}}{\delta J(t_{1})\dots\delta J(t_{n})}Z[J] |_{J=0} \end{equation}

Leave a Comment

captcha
Fontsize