Ward-Takahashi identity

From Physics wiki

Jump to: navigation, search

Consider the time-ordered vacuum expectation value of some operator \mathcal{O}(\phi)\,,

\left\langle \mathcal{O} (\phi)\right\rangle = \int\!\mathcal{D}\phi\, \mathcal{O}(\phi) e^{i S[\phi]}\,,

and consider some infinitesimal transformation of the fields Q[\phi(x)] = \phi'(x) - \phi(x) = \delta\phi(x)\,, e.g., a gauge transformation. We may write Q[\phi(x)] = i \epsilon_a(x) G_a \phi(x) \, where G_a\, is the generator of the transformation, and \epsilon_a(x)\, is some infinitesimal parameter. Then

\left\langle \mathcal{O} (\phi')\right\rangle = \int\!\mathcal{D}\phi\, \mathcal{O}(\phi') e^{i S[\phi']}\,,

but for most cases under consideration, the functional measure is invariant, i.e., \mathcal{D}\phi' = \mathcal{D}\phi\,. Thus

0\,  = \left\langle \mathcal{O} (\phi') - \mathcal{O} (\phi)\right\rangle\,,
 = \int\!\mathcal{D}\phi\, \left(  \mathcal{O}(\phi') e^{i S[\phi']} - \mathcal{O}(\phi) e^{i S[\phi]}\right)\,,
 \approx \int\!\mathcal{D}\phi\, \left[  (\mathcal{O}(\phi) + Q[\mathcal{O}(\phi)]) e^{i S[\phi]} ( 1 + i Q[S[\phi]] + ...) - \mathcal{O}(\phi) e^{i S[\phi]}\right]\,.

If the transformation is a classical symmetry, i.e., S[\phi] \to S[\phi] - \int_M \epsilon_a(x) \, \partial_\mu J_a^\mu(x)\, (see Noether current)

 \int\!\mathcal{D}\phi\, e^{i S[\phi]} ( Q[\mathcal{O}(\phi)]  + i \mathcal{O} (\phi) Q[S[\phi]] ) = 0\,,

or

 \left\langle Q[\mathcal{O}(\phi)]\right\rangle  -  i  \int_{ M}\! \epsilon_a(x) \partial_\mu \left\langle  \mathcal{O} (\phi)  J_a^\mu(x) \right\rangle = 0\,,

or

 \left\langle Q[\mathcal{O}(\phi)] \right\rangle =  i \int_{ M} \epsilon_a(x)\, \partial_\mu \left\langle \mathcal{O} (\phi)   J_a^\mu(x) \right\rangle \,.

Conserved charges

Suppose \mathcal{O} (\phi) = \phi(x_1)\phi(x_2)...\phi(x_n)\,. Then

i \int_{ M} \epsilon_a(x)\, \partial_\mu \left\langle \phi(x_1)\phi(x_2)...\phi(x_n) J_a^\mu (x)\right\rangle\, = i \sum_i \left\langle \phi(x_1)... G_a \phi(x_i) \epsilon_a(x_i)...\phi(x_n)\right\rangle\,,
= i \int_{ M} \epsilon_a(x) \sum_i \delta^{(d)}(x-x_i)\left\langle \phi(x_1)... G_a \phi(x_i) ...\phi(x_n)\right\rangle\,,

and, since \epsilon_a(x)\, is an arbitrary parameter,

i \partial_\mu \left\langle \phi(x_1)\phi(x_2)...\phi(x_n) J_a^\mu (x)\right\rangle =  i  \sum_i \delta^{(d)}(x-x_i) \left\langle \phi(x_1)... G_a \phi(x_i) ...\phi(x_n)\right\rangle.

Now suppose \mathcal{O}(\phi) = \phi(x_1) \mathcal{A}(\phi)\, where the time-component x_1^0\, differs from all those in \mathcal{A}(\phi)\,. Then, suppose we integrate the previous expression over an infinitesimally thin volume bounded by times t_- < x_1^0 < t_+\, and by spatial infinity such that none of the points in \mathcal{A}(\phi)\, lie in this volume. Then, by the divergence theorem, defining

Q_a =  \int\!d^{d-1}x\,J_a^0 (\mathbf{x})\,, we have
i \left\langle \mathcal{A}(\phi) G_a \phi(x_1) \right\rangle = \lim_{t_- \to t_+} i \left.\int\!d^{d-1}x\, \left\langle \mathcal{A}(\phi) \phi(x_1)  J_a^0 (\mathbf{x}, t)\right\rangle\right|_{t_-}^{t_+}\,
= \lim_{t_- \to t_+}i  \left\langle \mathcal{A}(\phi) \phi(x_1)  Q_a(t_+)\right\rangle - i \int\!d^{d-1}x\, \left\langle \mathcal{A}(\phi) \phi(x_1)  Q_a(t_-)\right\rangle\,
= \lim_{t_- \to t_+} i \left\langle \mathcal{A}(\phi) Q_a(t_+) \phi(x_1)\right\rangle - i \int\!d^{d-1}x\, \left\langle \mathcal{A}(\phi) \phi(x_1)  Q_a(t_-)\right\rangle\, (since the correlation function is time-ordered),
= i \left\langle \mathcal{A}(\phi) \left[ Q_a(x_1^0) , \phi(x_1) \right]\right\rangle\,.

Since \mathcal{A}(\phi)\, is arbitrary, we may further argue, by generalizing the path integral to go between arbitrary initial and final states, that

G_a \phi(x) = \left[ Q_a , \phi \right]\,.

Thus conserved charges are the generators of the symmetry transformations.

References

Further reading:[1]

  1. P. Di Francesco, P. Mathieu, and D. Sénéchal, Conformal Field Theory, Springer Science, New York, 1997. ISBN 0-387-94785-X.
Personal tools