What is Quantum Field Theory really about? And what does it tell us about reality? The only way to answer these questions is to go to the formalism itself - but that's really hard.
I already asked an earlier chatbot version to sketch out an answer, which I guess was pushing at the state-of-the-art back in October 2025. What a difference ten months makes!
So here is GPT-5.6 with this prompt: Is it possible to rewrite the original post incorporating the insights of this thread (our prior discussion this evening) and have the resulting text understandable by a good physics graduate (using sufficiently student-friendly explanatory text)?
What Exactly Does a Quantum Field Operator Do?
Quantum field theory is often introduced by saying that particles are “excitations of fields”. This is useful shorthand, but it conceals a conceptual problem. What sort of thing is a quantum field, mathematically? What does it take as input? What does it return? And at what point in the mathematics does anything resembling an actually observable electron appear?
One useful way of keeping the story straight is to keep track of the type signatures all the way through. The signatures tell us what kind of mathematical object is being supplied at each stage and what kind of object comes back. They also expose several places where ordinary QFT language slides rather casually from one ontological category to another.
1. Start with a classical field
A classical scalar field has the familiar form
φ : M → ℝ,
where M is a spacetime manifold. Give the field a spacetime point x,
x ∈ M,
and it returns a real number:
φ(x) ∈ ℝ.
The type signature is therefore completely straightforward:
spacetime point → field value.
This makes it tempting to imagine a quantum field in the same way, perhaps as
Φ : M → ℂ, (complex numbers)
so that each spacetime point carries some sort of quantum amplitude.
That is not what a quantum field is.
2. In QFT spacetime enters through a test function
We commonly write a quantum scalar field as Φ(x), where x is a spacetime point. But this notation is formal. Quantum fields are generally too singular to be ordinary operator-valued functions defined independently at every point.
The mathematically better object is an operator-valued distribution.
Instead of giving the field a single spacetime point, we give it a suitable test function f defined over spacetime and assigning real or complex numerical weights across a spacetime region:
f : M → ℂ.
More precisely, f belongs to some chosen space of test functions. For example,
f ∈ C∞c(M),
the smooth functions of compact support. One may instead use Schwartz functions, which are smooth and rapidly decreasing; a Gaussian belongs to that latter class.
The field is then smeared with f:
Φ(f) = ∫ d4x f(x)Φ(x).
The crucial point is that spacetime has not disappeared from the type signature. It has moved one level down. The argument supplied to the quantum field is itself a function whose domain is spacetime.
So the quantum field has, schematically, the higher-order signature
Φ : (M → ℂ)test → {operators}.
Or, making the usual test-function space explicit,
Φ : C∞c(M) → {operators}.
This is already quite different from the classical case.
The classical field maps
spacetime point → number.
The quantum field maps
spacetime test function → operator.
3. What kind of operator?
Quantum states are represented by vectors in a Hilbert space H. Field operators are generally unbounded, so they are not literally defined on every vector in H. We therefore introduce a suitable dense domain
D ⊂ H.
Then, schematically,
Φ(f) : D → D,
or in some formulations D → H.
Combining the two stages gives the curried type signature
Φ : C∞c(M) → (D → D).
Since a test function is itself a spacetime function, we can display the same idea more explicitly as
Φ : (M → ℂ)test → (D → D).
This is perhaps the most revealing single type signature in the whole discussion.
Give the quantum field a spacetime test function f, and it gives you an operator:
f → Φ(f).
Give that operator a quantum state |Ψ⟩, and it gives you another quantum state:
|Ψ⟩ → Φ(f)|Ψ⟩.
So the complete curried chain is
f → Φ(f) → Φ(f)|Ψ⟩.
The equivalent uncurried form is
Φ̃ : C∞c(M) × D → D,
with
Φ̃(f, |Ψ⟩) = Φ(f)|Ψ⟩.
Nothing here has returned either an amplitude or an electron. The output is another state vector.
4. The familiar Φ(x) notation
Physicists nevertheless routinely write
Φ(x).
Here x is indeed a spacetime variable:
x = (t, x) ∈ M.
Formally this makes the field look as though it had the signature
Φ : M → (D → D).
That is a very useful mnemonic, but it is not the rigorous type. The pointwise object Φ(x) is distributional. The mathematically controlled object is the smeared operator Φ(f).
So one should mentally read
Φ(x)
as convenient notation for the spacetime dependence of an operator-valued distribution, not as an ordinary function which takes a spacetime point and returns a perfectly well-defined operator.
5. Where do creation and annihilation operators enter?
For a free scalar field, the same quantum field Φ introduced above can be decomposed formally into positive- and negative-frequency parts:
Φ(x) = Φ(+)(x) + Φ(−)(x).
Here Φ(+)(x) contains annihilation operators, while Φ(−)(x) contains creation operators. Schematically,
Φ(x) ∼ a + a†.
The creation and annihilation operators are therefore not different fields. They are components appearing in the mode expansion of the same free quantum field.
A creation operator has, again schematically, the signature
a† : D → D.
Applied to the vacuum,
a†|0⟩ = |1⟩.
Physicists naturally say that a† has “created a particle”. But look at the type signature. Its domain is D, and its codomain is also D, a suitable dense domain of state vectors in the Hilbert space.
It does not have the signature
D → {physical particles}.
The mathematical output is a state vector.
6. The electron field
The electron field is more complicated because it is a Dirac spinor field, but the same distinction applies.
Formally one writes a mode expansion of the form
ψ(x) ∼ a u e−ipx + b†v eipx,
Here u and v are Dirac spinors carrying the spinor structure of the electron and positron modes; a and b† are the operators acting on Fock space. Momentum and spin labels and normalisation factors have been suppressed.
Here a annihilates an electron mode, while b† creates a positron mode. The adjoint field contains the corresponding electron-creation operator a†.
Again the fundamental type structure is not
spacetime → electron.
It is closer to
ψ : (M → spinor)test → (D → D).
The details of spinor test-function spaces need not concern us here. What matters is the ontological sequence:
spacetime dependence → test function → operator → quantum state.
When an electron creation operator acts on the vacuum,
a†s(p)|0⟩ = |p,s; electron⟩,
the output is a state with the mass, charge, spin and momentum quantum numbers associated with an electron.
We call it a one-electron state.
But a one-electron state and an electron are not the same kind of thing. One is a vector in a mathematical state space. The other is what we take that vector to represent physically.
7. This is the first ontological jump
The innocent-looking sentence
“the operator creates an electron”
therefore compresses two claims.
The mathematical claim is:
|0⟩ → |one-electron state⟩.
The interpretative claim is:
this state represents a physical situation in which an electron exists.
The first follows from the formalism. The second connects the formalism to the world.
For routine calculations the distinction is harmless. For ontology it is crucial.
No operator encountered so far has electrons as elements of its codomain.
8. Where do complex amplitudes appear?
We now add another state, represented by a bra ⟨Χ|. Given an operator A and states |Ψ⟩ and |Χ⟩, we can form the matrix element
⟨Χ|A|Ψ⟩.
This time the output really is a complex number:
⟨Χ|A|Ψ⟩ ∈ ℂ.
Its type signature is therefore schematically
D* × (D → D) × D → ℂ.
Substituting a smeared field operator gives
(⟨Χ|, f, |Ψ⟩) → ⟨Χ|Φ(f)|Ψ⟩ ∈ ℂ.
So, if we curry everything, the journey can be displayed as
(M → ℂ)test → (D → D) → ℂ
once the initial and final states have also been supplied.
This is where amplitudes enter. They are not values of the quantum field at spacetime points. They arise after operators and states have been combined.
9. Correlation functions have the same lesson
Consider a two-point function:
⟨0|Φ(f)Φ(g)|0⟩.
Here we have supplied two spacetime test functions, obtained two operators, acted on the vacuum and finally paired the resulting state with the vacuum bra.
The type journey is roughly
(f,g) → (Φ(f),Φ(g)) → Φ(f)Φ(g)|0⟩ → ℂ.
The result is a complex-valued correlation.
Similarly, the familiar time-ordered expression
⟨0|T{Φ(x)Φ(y)}|0⟩
is the Feynman two-point function or propagator, again understood distributionally.
It is therefore misleading to think of the propagator as something that the field simply “has” at two points. It is the result of several layers of mathematical construction.
10. An amplitude is still not an observation
There is another ontological step to keep separate.
A complex number such as
⟨Χ|A|Ψ⟩
is not itself an observed event.
In a scattering calculation, suitable amplitudes are used, together with normalisation and phase-space factors, to calculate measurable cross-sections or decay rates.
More generally, a possible measurement outcome can be represented by an effect E. For a normalised state |Ψ⟩,
P(E) = ⟨Ψ|E|Ψ⟩.
Here P(E) is already a probability, a real number between 0 and 1, not a complex amplitude.
So the relevant type transition is
ℂ (amplitude) → [0,1] (probability) → actual event.
P(E) = ⟨Ψ|E|Ψ⟩.
The corresponding type is
(E, |Ψ⟩) → P(E) ∈ [0,1].
We have now moved from operators and state vectors to a real probability.
But even this probability is not the observed event.
It is a number predicted by the theory.
11. Where is the observable electron?
Suppose an electron reaches a detector.
The experiment may produce a charge pulse, an ionisation track, a bright pixel or a stored digital count. Those are physical events.
The QFT description may contain a prepared state, field operators, interaction Hamiltonians, amplitudes and probabilities. Schematically:
preparation → state → QFT dynamics → amplitude → probability.
Then, in the physical world,
a detector event occurs.
This final event is not an element of the Hilbert space, not an operator and not a complex number.
Nor did any function in the QFT formalism ever have the signature
F : something → {actual electrons}.
There is no electron-valued codomain.
This is the point at which the ontology becomes impossible to conceal beneath notation.
12. What, then, is a “one-electron state”?
A one-electron state is a state carrying the characteristic quantum numbers of the electron: its mass, electric charge, spin and momentum.
Within the theory we identify such states by their transformation properties and conserved quantum numbers.
So the mathematical classification
|Ψ⟩ ∈ {one-electron states}
is extremely precise.
The statement
“there is an electron”
is the physical interpretation placed upon such a state.
That interpretation is overwhelmingly successful experimentally. But it is still an interpretation relating one ontological category, a mathematical state, to another, a physical entity or event.
13. Interacting theories make the distinction sharper
The free-field picture can tempt us into thinking that particles are simply produced by applying creation operators (a vacuum state goes to a one-electron state, then to a two-electron state in distinct modes):
|0⟩ → a†|0⟩ → (a†)2|0⟩.
That picture is exact for free Fock space and enormously useful pedagogically.
But real electrons interact with the electromagnetic field. A physical electron is not simply the pristine state a†|0⟩ of a non-interacting Dirac theory. It is dressed by its electromagnetic interaction, and in QED the presence of massless photons makes the asymptotic particle story technically subtle.
Nevertheless, particle states remain extraordinarily effective in appropriate regimes, particularly in scattering experiments. The S-matrix and LSZ machinery connect local quantum fields to asymptotic particle states.
This suggests a reversal of the naïve textbook picture.
Instead of thinking
particles are primary and fields create and destroy them,
the formalism itself is better described as
local quantum fields generate relations between quantum states, from which particle descriptions emerge in appropriate regimes.
That is a statement about the structure of QFT, not a proof that fields are the ultimate furniture of reality.
14. The complete signature journey
We can now assemble the whole chain.
First comes spacetime, represented by a four-dimensional Lorentzian manifold M:
x ∈ M.
A test function is defined on spacetime:
f : M → ℂ.
The quantum field takes an admissible spacetime test function and returns an operator:
Φ : (M → ℂ)test → (D → D).
Supplying f gives
Φ(f) : D → D.
Supplying an initial state gives another state:
Φ(f)|Ψ⟩ ∈ D.
Supplying a final-state bra gives a complex matrix element:
⟨Χ|Φ(f)|Ψ⟩ ∈ ℂ.
A suitable measurement construction gives a probability:
P(E) ∈ [0,1].
And finally, outside the mathematical codomains of the theory, the laboratory produces an actual physical event.
The ontological journey is therefore:
spacetime → function on spacetime → operator → quantum state → amplitude → probability → observed event.
15. Keeping everything really clear
Keeping the signatures visible prevents several familiar but misleading identifications.
A quantum field is not a function from spacetime to complex amplitudes.
A field operator does not return a particle.
A creation operator does not have electrons in its codomain.
A one-electron state is not the same category of thing as a physical electron.
An amplitude is not an observation.
A probability is not the event whose probability it gives.
And a detector event is not a vector in Hilbert space.
QFT relates these different levels with extraordinary quantitative success. But the familiar phrase “the electron field creates an electron” compresses a long mathematical and interpretative journey:
spacetime localisation → operator → state → physical interpretation.
For calculation, the shorthand is harmless. For understanding what the theory is actually saying, the type signatures are really essential.
