These opening two pages are Feynman’s roadmap for a new way of calculating quantum electrodynamics. He is not yet deriving the machinery; he is explaining what problem it solves and why the older formulation was cumbersome.

The paper’s central idea

The paper is Richard Feynman’s 1949 article, “Space-Time Approach to Quantum Electrodynamics.” Its essential claim is:

Instead of describing a quantum process as a long sequence of intermediate states evolving moment by moment, calculate the amplitude of the complete process directly in spacetime.

This is the conceptual foundation of what we now call the Feynman-diagram approach.

---

Page 1: What Feynman says the paper will accomplish

Feynman announces two main objectives.

1. Simplify calculations in quantum electrodynamics

In the conventional method, a physical process was expanded into many separate mathematical terms. Each term corresponded to a particular intermediate state or ordering of events.

For example, suppose two electrons exchange a photon. The older calculation might separately consider:

• Electron 1 emits the photon and electron 2 absorbs it.
• Electron 2 emits the photon and electron 1 absorbs it.
• One event occurs before the other.
• The opposite time ordering occurs.
• Additional virtual particles appear in intermediate states.

These alternatives are closely related physically, but the Hamiltonian method often treats them as separate terms.

Feynman’s approach combines them into a single spacetime expression. Each term in the perturbation expansion can then be understood as a physical spacetime process—what would later be represented by a Feynman diagram.

What is a “matrix element”?

A matrix element is the quantum amplitude for a system to go from an initial state to a final state:$$\mathcal{M}_{fi} = \langle f|\hat{S}|i\rangle.$$

It is not itself a probability. Roughly speaking, the measurable probability or cross section is obtained from$$|\mathcal{M}_{fi}|^2.$$

When Feynman says he wants to “write down matrix elements directly,” he means that one should be able to look at a physical process and construct its amplitude without first listing every possible intermediate state.

---

Expansion in powers of the electromagnetic coupling

Feynman says that QED calculations are performed as an expansion in powers of$$\frac{e^2}{\hbar c}.$$

In the units used in the paper, this is essentially the fine-structure constant:$$\alpha \approx \frac{1}{137}.$$

Because $\alpha$ is small, increasingly complicated interactions are normally less important:$$\mathcal{M} = \mathcal{M}^{(0)} + \alpha\,\mathcal{M}^{(1)} + \alpha^2\mathcal{M}^{(2)} +\cdots.$$

Each higher-order term involves more interaction vertices, loops, or virtual particles. This is called perturbation theory.

Feynman’s advance is not the invention of the expansion itself. It is the creation of a much clearer way to organize and interpret its terms.

---

What are “virtual quanta”?

The paper uses the older phrase “virtual quanta.” In modern language, these are virtual photons or other internal particles appearing inside a perturbative calculation.

They are not photons directly observed by a detector. They are mathematical components of the amplitude connecting observable initial and final particles.

For instance:$$e^-+e^- \longrightarrow e^-+e^-$$

can be described as two electrons exchanging a virtual photon:$$e^- \longrightarrow e^-+\gamma^*, \qquad e^-+\gamma^* \longrightarrow e^-.$$

The asterisk reminds us that the photon is virtual and does not have to satisfy the energy–momentum relation of a real photon.

---

The “overall spacetime view”

This is perhaps the most important phrase on the first page.

The Hamiltonian formulation asks:

Given the state of the system at one time, how does it evolve to the next time?

Feynman’s spacetime formulation instead asks:

What is the amplitude connecting the complete initial configuration with the complete final configuration?

The latter viewpoint treats the experiment as a four-dimensional spacetime process. Intermediate histories contribute to the final amplitude, but they do not need to be interpreted as directly observable stages.

This also makes relativistic symmetry more visible. Space and time enter the expressions together, rather than time being singled out as the parameter driving the calculation.

---

Combining electron and positron processes

Feynman says that processes involving virtual electron–positron pairs can be combined with processes involving only positive-energy electrons.

This reflects his famous interpretation:

A positron can be represented mathematically as an electron propagating backward in time.

This does not mean that a laboratory positron literally travels into yesterday. It means that the mathematical propagator describing a negative-energy electron moving one way through spacetime can be reinterpreted as a positive-energy antiparticle moving in the opposite temporal direction.

The benefit is that electron and positron contributions can be described by one unified propagator rather than by several disconnected rules.

---

The second objective: dealing with infinities

QED calculations produced divergent quantities—integrals whose values appeared to be infinite.

One notorious example is the electron’s self-energy. An electron interacts with its own electromagnetic field, schematically:$$e^- \longrightarrow e^-+\gamma^* \longrightarrow e^-.$$

The loop correction to the electron propagator contains an integral over all possible virtual momenta. At very large momentum, corresponding to extremely short distances, the integral diverges.

Feynman temporarily modifies the interaction at extremely short distances. This introduces a cutoff: the theory no longer allows arbitrarily short-distance contributions to grow without limit.

The modified calculation is finite, but the cutoff is not supposed to remain part of the observable prediction.

---

Renormalization in the paper

Feynman explains that the divergent self-energy can be absorbed into a redefinition of the electron’s mass.

The basic idea is$$m_{\text{physical}} = m_{\text{bare}}+\delta m.$$

Here:

• $m_{\text{bare}}$ is a parameter appearing in the original equations.
• $\delta m$ is the mass correction generated by interactions.
• $m_{\text{physical}}$ is the mass actually measured.

A similar procedure applies to electric charge:$$e_{\text{physical}} = e_{\text{bare}}+\delta e.$$

The separate quantities on the right may depend on the regulator or cutoff. But after expressing the answer in terms of the measured mass and charge, observable predictions can remain finite as the cutoff is removed.

That is what Feynman means when he says the cutoff width may be taken to zero for real processes.

---

Feynman is candid about the weakness

The proposed short-distance modification is not presented as a fundamental description of nature. Feynman explicitly admits that:

• Its physical basis is unclear.
• It can create difficulties with energy conservation.
• It is mainly a device for defining divergent calculations.
• The final observable results should not depend on its detailed form.

This is important. Feynman is not claiming that nature necessarily possesses the particular cutoff he introduces. He is saying that it gives a controlled route to finite answers.

Modern QED expresses this more systematically through regularization and renormalization.

---

Page 2: What happens after renormalization?

The second page continues the argument:

1. Regulate the divergent expressions.
2. Identify the divergent parts with corrections to mass and charge.
3. Rewrite the theory using the measured mass and charge.
4. Remove the regulator.
5. Obtain finite predictions for observable processes.

Feynman contrasts his method with Schwinger’s. Schwinger’s formulation identifies and removes the mass and charge corrections before evaluating the remaining physical quantities. Feynman’s method regulates the whole calculation first and then separates out the renormalizations.

The final predictions should agree.

Feynman notes that Freeman Dyson would provide a more systematic proof that the Schwinger and Feynman approaches were equivalent.

---

What does Feynman mean by “real processes”?

He does not mean “real” as opposed to imaginary numbers. He means observable processes, such as:

• Electron scattering
• Photon emission
• Pair creation
• Energy-level shifts
• Measurable cross sections

Quantities such as a bare electron mass or an isolated divergent loop are not directly observable. The requirement is that measurable predictions be finite and independent of the artificial regulator.

---

Two limitations Feynman acknowledges

Feynman says the theory is not yet mathematically complete.

1. It is an order-by-order expansion

The method can calculate$$\mathcal{M}^{(0)},\quad \mathcal{M}^{(1)},\quad \mathcal{M}^{(2)},\ldots$$

but it does not provide a single closed expression containing all orders simultaneously.

Even today, perturbative QED is generally used as an asymptotic series rather than as an ordinary convergent infinite series.

2. Equivalence with conventional QED was not yet fully proved

Feynman believed his results agreed with the conventional Hamiltonian theory, but this particular paper did not contain a complete mathematical proof. Dyson’s work subsequently clarified the equivalence among the Feynman, Schwinger, and Tomonaga formulations.

---

How Feynman says the method originated

Feynman briefly describes the development of his approach:

1. He began with the Lagrangian formulation of quantum mechanics.
2. He learned how to eliminate—or “integrate out”—electromagnetic field oscillators.
3. This produced a delayed interaction between charged particles.
4. He modified the short-distance interaction to control divergences.
5. He extended the formalism from the Schrödinger equation to the relativistic Dirac equation.
6. He incorporated electron–positron pair creation.
7. He expanded the result in powers of the electromagnetic coupling.
8. Each term acquired a simple spacetime interpretation.

The phrase “integrating out the field” means that the photon field is mathematically eliminated as an independent variable. Its effect remains encoded in an interaction connecting charged particles at different spacetime points.

Schematically, instead of explicitly describing$$\text{electron} \rightarrow \text{photon field} \rightarrow \text{electron},$$

one writes an effective interaction of the form$$J_\mu(x)\,D^{\mu\nu}(x-y)\,J_\nu(y),$$

where:

• $J_\mu(x)$ is the electromagnetic current at $x$,
• $D^{\mu\nu}(x-y)$ is the photon propagator,
• $J_\nu(y)$ is the current at $y$.

The propagator carries the electromagnetic influence from one spacetime point to another.

---

Field description versus direct interaction

At the end of page 2, Feynman begins comparing two descriptions of electromagnetism.

Field viewpoint

A charge produces an electromagnetic field, and another charge responds to that field:$$\text{source charge} \rightarrow \text{field} \rightarrow \text{absorbing charge}.$$

This is Maxwell’s familiar description.

Direct-interaction viewpoint

One may instead regard the source and absorber as interacting directly, with the interaction delayed by the finite speed of light:$$\text{charge at }x \longleftrightarrow \text{charge at }y.$$

The photon propagator mathematically represents this connection.

Feynman considers the two viewpoints equivalent and complementary. The field picture is convenient for radiation emitted by complicated sources. The direct-interaction picture can be more natural when calculating how a small number of charged particles scatter from one another.

---

The main takeaway from these pages

The opening pages are telling us that QED can be reorganized around complete spacetime processes:$$\boxed{ \text{Initial particles} \;\longrightarrow\; \text{all allowed intermediate histories} \;\longrightarrow\; \text{final particles} }$$

Each history contributes an amplitude. Related time orderings and particle interpretations are combined by propagators. Divergent short-distance contributions are regulated, absorbed into measured masses and charges, and removed from observable predictions.

That is the conceptual bridge from traditional Hamiltonian perturbation theory to modern Feynman diagrams and propagators.