Ward-Takahashi Identity and QED Renormalization

1. The Renormalizability Puzzle of QED

Before diving into the identity itself, let's appreciate why we need it. A naive analysis of QED's divergences suggests the theory might be in trouble.

As you know from your physics background, dimensional analysis is a powerful tool. In QFT, we can use a version of it called "power counting" to estimate the degree of UV divergence ($D$) of a Feynman diagram. For QED in 4 dimensions, the formula is $D=4-\frac{3}{2}{E}_e-E_{\gamma }$, where $E_e$ and $E_{\gamma }$ are the number of external electron and photon lines. A diagram with $D\ge 0$ is superficially divergent.

This analysis flags several amplitudes as problematic. The resource below details this power-counting and identifies the apparent need for counterterms that QED's Lagrangian does not possess.

As you just read, power counting suggests we need counterterms like ${\delta }_{m_{\gamma }}{A}_{\mu }{A}^{\mu }$ to cancel a quadratic divergence ($\propto {\Lambda }^2$) in the photon self-energy. But such a term is forbidden by gauge invariance! The theory appears to be non-renormalizable unless something forces these problematic divergences to vanish on their own. That "something" is the Ward-Takahashi identity.

2. The Identity: From Current Conservation to Green's Functions

The Ward-Takahashi identity is a direct consequence of the conservation of the electromagnetic current, ${\partial }_{\mu }{J}^{\mu }=0$. In the quantum theory, we must be careful about how this operator equation translates to expectation values of time-ordered products of fields, which form our Green's functions.

The derivation involves applying the divergence ${\partial }_{\mu }$ to a time-ordered product containing the current operator $J^{\mu }(y)$. The derivative does not simply pass through the time-ordering operator $T$. When the time component $y^0$ crosses the time of another field operator, the derivative brings down a term related to their equal-time commutator.

This procedure, which is based on the canonical quantization formalism you're familiar with, leads to a master identity relating different Green's functions. The following resource provides a clear, formal proof.

The central result from that reading, written in momentum space, is the Ward-Takahashi identity:

Let's break this down:

• ${\mathcal{M}}^{\mu }$ is the amputated Green's function (or S-matrix element) for a process involving an external photon with momentum $k$ and index $\mu$.
• The left side is the contraction of this amplitude with the photon's momentum.
• ${\mathcal{M}}_0$ represents the same process but without the external photon.
• The right side is a sum over all external charged fermion lines $j$. The two terms represent the amplitude where the photon's momentum $k$ is absorbed by an outgoing fermion line versus an incoming fermion line. The signs differ depending on whether the particle is incoming or outgoing.

This abstract equation is beautifully captured by a diagrammatic representation.

The identity provides a powerful, non-perturbative constraint relating an amplitude with $N$ photons to an amplitude with $N-1$ photons.

3. Consequences for Renormalization

This identity is the linchpin of QED's renormalizability. It enforces the cancellation of the "forbidden" divergences we identified earlier.

A. Photon Self-Energy and Masslessness

Let's apply the identity to the photon self-energy, ${\Pi }^{\mu \nu }(k)$. This is an amplitude with two external photons and no external charged particles. Since there are no external fermion lines, the sum on the right-hand side of the WT identity is empty, so it equals zero. This gives the simpler Ward Identity:

This tells us that the vacuum polarization tensor must be transverse. The most general Lorentz-covariant form for ${\Pi }^{\mu \nu }(k)$ is ${\Pi }^{\mu \nu }(k)=A(k^2)g^{\mu \nu }+B(k^2)k^{\mu }{k}^{\nu }$. Applying the Ward identity gives:

This must hold for any $k$, which implies $A(k^2)=-k^{2}B(k^2)$. Therefore, the vacuum polarization tensor must take the form:

(Note: Authors often define $\Pi (k^2)=-B(k^2)$ to get the familiar form ${\Pi }^{\mu \nu }(k)=(k^2{g}^{\mu \nu }-k^{\mu }{k}^{\nu })\Pi (k^2)$).

Now, recall the power-counting result from section 1. The superficial degree of divergence for ${\Pi }^{\mu \nu }$ is $D=+2$, suggesting a divergent part of the form ${\Pi }_{div}^{\mu \nu }=C_1{\Lambda }^2{g}^{\mu \nu }+C_2\log (\Lambda )(k^2{g}^{\mu \nu }-k^{\mu }{k}^{\nu })+\dots$.
The Ward identity $k_{\mu }{\Pi }_{div}^{\mu \nu }=0$ forces the coefficient of the $g^{\mu \nu }$ term to be zero. This means the quadratic divergence $C_1{\Lambda }^2$ must vanish. This is a profound result: gauge symmetry, via the Ward identity, forbids the photon from acquiring a mass through radiative corrections. The "missing" photon mass counterterm isn't needed because the divergence it would cancel is already zero.

This is explicitly explained in the second part of the first resource.

B. Vertex Correction and the Universality of Charge

Another crucial consequence arises when we apply the WT identity to the electron-photon vertex. This relates the vertex function ${\Gamma }^{\mu }$ to the full electron propagator $S(p)$. A detailed derivation connects the vertex renormalization constant $Z_1$ to the electron field strength renormalization constant $Z_2$.

The result of this application is the famous identity:

This means the counterterms for the vertex (${\delta }_1$) and the electron wave function (${\delta }_2$) must be identical, ${\delta }_1={\delta }_2$.

Why is $Z_1=Z_2$ so important? Recall the relation between the bare charge $e_B$ and the renormalized physical charge $e_{phys}$:

The Ward-Takahashi identity, by ensuring $Z_1=Z_2$, simplifies this to:

This means that the charge renormalization comes only from the vacuum polarization ($Z_3$), which affects the photon. It is independent of the specifics of the charged particle (electron, muon, etc.). This guarantees that if all bare charges are equal, all physical charges are equal. In other words, it ensures the universality of electric charge, a fundamental experimental fact.

Conclusion

Today we have seen that the renormalizability of QED is not an accident but a direct consequence of its gauge symmetry, mathematically encoded in the Ward-Takahashi identity.

Key Takeaways:

• The Ward-Takahashi identity is a non-perturbative result stemming from current conservation. It relates Green's functions with an external photon to Green's functions without it.
• It serves as a powerful constraint that guarantees the consistency of QED renormalization.
• Consequence 1: For the photon self-energy, it implies transversality ($k_{\mu }{\Pi }^{\mu \nu }=0$). This forces the cancellation of a quadratic divergence, ensuring the photon remains massless to all orders in perturbation theory.
• Consequence 2: For the vertex function, it leads to the crucial relation $Z_1=Z_2$. This ensures that charge renormalization is universal for all charged particles and depends only on the photon's field renormalization ($Z_3$).

You have now completed the core of QED renormalization within the canonical quantization framework. We have built everything from fields, operators, and commutators. While powerful, this approach can become quite cumbersome, as the derivation of the WT identity demonstrates.

Preview of the Next Lesson:
In the next module, we will pivot to a completely different formulation of quantum field theory: the path integral. We will see that it offers a more direct and often more intuitive route to deriving Green's functions and understanding symmetries. Our first lesson will build the formalism from the ground up, starting with the derivation of the path integral for a simple quantum mechanical propagator using the time-slicing method. This will open the door to a new and elegant perspective on the principles we have established so far.