---
title: 'Free Scalar Fields: Definition, Applications, Implications'
url: https://www.emergentmind.com/topics/free-scalar
type: topic
---

# Free Scalar Fields: Definition, Applications, Implications

Free scalar denotes a quantum field theory whose dynamical field is scalar and whose action is quadratic, so that the equations of motion are linear and correlation functions are determined by two-point functions through Wick contraction. In its ordinary conformal realization, the Euclidean action is
\[
S\propto\int d^d x\,\phi^\dagger\square\phi,
\qquad
\square=\delta^{ij}\partial_i\partial_j,
\]
with the real-field version obtained by replacing \(\phi^\dagger\) by \(\phi\). The term also encompasses higher-derivative \(\square^k\) theories, free scalar fields on curved or asymptotically anti-de Sitter backgrounds, defect and boundary realizations, nonrelativistic and BMS-invariant limits, and discretized or constrained formulations. These extensions differ substantially in unitarity, locality, representation theory, and even the meaning of “free.”

## 1. Ordinary free scalar conformal field theory

For the ordinary theory, corresponding to \(k=1\), the flat-space action is
\[
S\propto\int d^dx\,\phi^\dagger\square\phi,
\]
and the equation of motion is
\[
\square\phi=0.
\]
On a curved background, conformal invariance requires the conformally coupled action
\[
S=\int d^dx\sqrt g\left[
\phi^\dagger\square\phi
-\frac{d-2}{4(d-1)}R|\phi|^2
\right].
\]
The scalar transforms under Weyl rescaling according to
\[
g_{ij}\to e^{2\sigma(x)}g_{ij},
\qquad
\phi\to e^{-\Delta_\phi\sigma(x)}\phi,
\qquad
\Delta_\phi=\frac d2-1.
\]
The curved-space equation is the Yamabe equation,
\[
P_1[g]\phi\equiv
\left(\square-\frac{d-2}{4(d-1)}R\right)\phi=0.
\]

The conformal two-point function is
\[
\langle\phi^\dagger(x)\phi(0)\rangle
=\frac{1}{|x|^{d-2}},
\]
with the other charge assignments vanishing in the complex theory. All higher correlation functions follow by Wick contraction. For a real scalar, the corresponding propagator is the usual Green function of the conformal Laplacian.

The singlet sector of a formal \(U(N)\) or \(O(N)\) model is obtained by restricting to operators invariant under the global group, which is not dynamically gauged. In the \(U(N)\) theory, bilinears contract \(\phi^\dagger_a\phi^a\); in the \(O(N)\) theory, one uses \(\phi_a\phi_a\). The singlet single-trace primaries are bilinear currents,
\[
J^{(k=1)}_{i_1\cdots i_s}
\sim
\phi^\dagger\partial_{i_1}\cdots\partial_{i_s}\phi+\cdots,
\]
with dimensions
\[
\Delta=d-2+s.
\]
In the \(U(N)\) model there is one primary for every spin \(s=0,1,2,\ldots\), whereas the \(O(N)\) model contains only even spins. For \(s\geq1\), the currents saturate the spin-\(s\) unitarity bound,
\[
\Delta=d+s-2,
\]
and obey
\[
\partial^{i_1}J_{i_1\cdots i_s}=0.
\]
The spin-two current is the stress tensor, while the spin-one current in the \(U(N)\) theory is the global \(U(1)\) current associated with \(\delta\phi=i\phi\). These currents generate the ordinary free-scalar higher-spin algebra [1607.07439].

The theory also contains multi-trace operators, formed from products of singlet bilinears and their derivatives. Their correlation functions and OPE coefficients are completely determined by Wick contraction. The higher-spin currents can be contracted with symmetric traceless conformal Killing tensors to form conserved Noether currents.

On the cylinder, state-operator correspondence identifies energies with scaling dimensions. For a real scalar in three spacetime dimensions quantized on \(S^2_R\times\mathbb R\), the conformal curvature coupling gives
\[
M=\frac{1}{2R},
\qquad
\xi=\frac18,
\]
and the oscillator energies are
\[
\omega_\ell=\frac{\ell+\frac12}{R}.
\]
Thus the scalar has \(\Delta_\phi=\frac12\), while \(\phi^2\) has \(\Delta_{\phi^2}=1\). A fuzzy-sphere construction reproduces this spectrum, the correlators, and the harmonic-oscillator algebra in a controlled low-angular-momentum, large-system limit [2506.14904].

## 2. Higher-derivative free scalar theories

A broad generalization replaces the kinetic operator by a power of the Laplacian:
\[
S\propto\int d^dx\,\phi_a^\dagger\square^k\phi^a,
\qquad k\geq1.
\]
Conformal invariance requires
\[
\Delta_\phi=\frac d2-k,
\]
and the equation of motion is
\[
\square^k\phi=0.
\]
The conformal two-point function is
\[
\langle\phi^\dagger(x)\phi(0)\rangle
=\frac{1}{x^{2\Delta_\phi}}
=\frac{1}{x^{d-2k}},
\]
apart from special even dimensions in which the analytic propagator is not a Green function.

The curved-space completion is given by the GJMS operator,
\[
P_k[g]=\square^k+\text{curvature terms},
\]
which obeys the conformal covariance relation
\[
P_k[e^{2\sigma}g]
\left(e^{-\Delta_\phi\sigma}\phi\right)
=
e^{-(\Delta_\phi+2k)\sigma}P_k[g]\phi.
\]
For \(k=2\), this is the Paneitz operator. On an Einstein metric, the higher-order operator factorizes into second-order factors involving the scalar curvature.

The singlet single-trace spectrum contains \(k\) Regge trajectories, indexed by
\[
b=0,1,\ldots,k-1.
\]
A representative primary is
\[
j^{(b)}_{i_1\ldots i_s}
\sim
\phi^\dagger\partial_{i_1}\cdots\partial_{i_s}\square^b\phi+\cdots,
\]
with dimension
\[
\Delta_{s,b}=d-2k+s+2b.
\]
Each trajectory has fixed twist
\[
\tau_{s,b}=d-2k+2b.
\]
The highest trajectory, \(b=k-1\), has
\[
\Delta=d+s-2
\]
and is the ordinary conserved-current trajectory; its spin-two member is the stress tensor when a genuine stress tensor exists. The lower trajectories are generally below the unitarity bound.

More generally, the currents are multiply conserved:
\[
\partial^{i_1}\cdots\partial^{i_c}
j^{(b)}_{i_1\cdots i_c\cdots i_s}=0,
\qquad
c=2(k-b)-1.
\]
Thus \(b=k-1\) gives ordinary conservation, \(b=k-2\) gives triply conserved currents, and \(b=0\) gives \((2k-1)\)-fold conservation. In the \(\square^2\) theory, the two trajectories have
\[
\Delta_{s,0}=d+s-4,
\qquad
\Delta_{s,1}=d+s-2.
\]
The first is triply conserved for \(s\geq3\), while the second is ordinarily conserved for \(s\geq1\).

These theories are generally non-unitary for \(k>1\). Negative- and zero-norm states occur, and changing the overall sign of the action cannot remove relative sign differences among trajectories. The higher-derivative character therefore changes the representation theory rather than merely adding additional free modes [1607.07439].

## 3. Higher-spin algebras, null states, and extensions

The multiply conserved currents of the \(\square^k\) theories generate generalized higher-spin symmetries. A generalized conformal Killing tensor satisfies
\[
\partial^{(i_1}\cdots\partial^{i_c}
K^{i_{c+1}\ldots i_s)}_{(s,c),T}=0.
\]
Its contraction with a \(c\)-conserved current produces an ordinary conserved Noether current. The leading action of the corresponding transformation on the scalar is
\[
\delta_K\phi
=
K^{i_1\ldots i_{s-c}}
\partial_{i_1}\cdots\partial_{i_{s-c}}
\square^{k-b-1}\phi+\cdots.
\]
The generalized transformations close, up to on-shell trivial terms, into a higher-spin algebra denoted \(hs_k\):
\[
[\delta_K,\delta_{K'}]\phi
=
\delta_{[K,K']}\phi+\text{on-shell trivial terms}.
\]
For \(k=1\), this reduces to the ordinary free-scalar higher-spin algebra. For \(k=2\), the lower trajectory generates new triply conserved symmetries.

The conformal modules can be nonstandard. In ordinary representation theory, a primary state lies in the kernel of \(K_i\), while descendants lie in the image of translations \(P_i\). In higher-derivative theories, a state can be both primary and descendant. For \(d=6\), \(k=2\),
\[
j^{(1)}_0=2\square j^{(0)}_0,
\]
so \(j^{(1)}_0\) is both primary and descendant and has zero norm. It is not a conventional null vector because it is not orthogonal to every state.

The operator space may contain extension states that are neither primary nor descendant. In the same example,
\[
\widetilde j^{(1)}_0
=
\partial_i\phi^\dagger\partial^i\phi
-\frac12(\square\phi^\dagger\,\phi+\phi^\dagger\square\phi)
\]
satisfies
\[
K_i|\widetilde j^{(1)}_0\rangle
\propto
|\partial_i j^{(0)}_0\rangle\neq0,
\]
while it cannot be generated by translations acting on an existing state. Its inner product with the primary descendant is nonzero, producing an extension of the original conformal module.

A related degeneration occurs in \(d=3\), \(k=2\), where
\[
j^{(0)}_{2,ij}
=
\partial_{(i}\partial_{j)_T}j^{(0)}_0
\]
is simultaneously primary and descendant with zero norm. The corresponding extension operator is neither primary nor descendant. The general classification associates such module gluings with type-I and type-III conformal-algebra degeneracies; no type-II module gluings were found in the cited analysis [1607.07439].

These structures modify OPE decompositions. Apparent poles in ordinary primary-descendant bases can cancel only after combining the degenerate primary module with its extension. Consequently, conformal blocks may need to be replaced by extended blocks containing both modules.

## 4. Special dimensions, finite algebras, anomalies, and free energies

In even dimensions
\[
d=2,4,\ldots,2k,
\]
the scalar dimension
\[
\Delta_\phi=\frac d2-k
\]
is a non-positive integer. The conformal propagator becomes analytic rather than a Green function, and sufficiently differentiated correlators vanish. After quotienting by null operators, the single-trace sector becomes finite-dimensional and generally non-unitary.

For \(k=2\), in \(d=4\),
\[
\langle\phi^\dagger(x)\phi(0)\rangle=1,
\]
and only
\[
j^{(0)}_0=\phi^\dagger\phi
\]
has nonzero correlators. The single-trace space is one-dimensional. In \(d=2\), the nonzero single-trace primaries include
\[
j^{(0)}_0,\qquad j^{(1)}_0,\qquad j^{(0)}_1,
\]
with only finitely many descendants. More generally, analyticity requires a spin-\(s\) primary to satisfy
\[
\Delta+s=0,-1,-2,\ldots,
\]
so only finitely many primaries survive. At the critical dimension \(d=2k\), only
\[
j^{(0)}_0\sim\phi^\dagger\phi,
\qquad
\Delta=0,
\]
survives, giving a one-dimensional algebra. As \(d\) decreases below \(2k\), a finite pyramid of states appears. These algebras are finite quotients or degenerations of \(hs_k\).

There is a distinct logarithmic possibility. One may replace the analytic propagator by logarithmic Green functions, such as
\[
\langle\phi^\dagger(x)\phi(0)\rangle\sim x^2\log|x|
\]
for \(d=2,k=2\), or
\[
\langle\phi^\dagger(x)\phi(0)\rangle\sim\log|x|
\]
for \(d=4,k=2\). In this realization, \(\phi\) is not a scaling operator and must be removed or subject to a shift symmetry. The finite analytic algebras and logarithmic theories are therefore distinct choices.

The sphere effective action determines the \(a\)-type anomaly in even dimensions. For the ordinary scalar,
\[
a_{k=1,d=4}=\frac1{90}.
\]
For higher-derivative theories, the eigenvalues of the sphere operator are
\[
\lambda_\ell
=
r^{-2k}
\prod_{j=1}^{2k}
\left(\ell+\frac d2+j-k-1\right),
\]
with degeneracies
\[
d_\ell
=
\binom{d+\ell}{d}-\binom{d+\ell-2}{d}.
\]
The resulting anomaly coefficients can have either sign and require caution in the finite or logarithmic dimensions \(d\leq2k\), where coupling to an arbitrary background metric is not generally well-defined [1607.07439].

For an ordinary conformally coupled scalar on \(S^d\), dimensional continuation gives the generalized free energy
\[
\widetilde F_{\rm free}
=
\Gamma(1+d)
\int_0^{d/2}du\,u\sin(\pi u)
\Gamma\!\left(\frac d2+u\right)
\Gamma\!\left(\frac d2-u\right).
\]
In \(d=3\),
\[
F_{\rm free}
=
\frac{\log2}{8}
-\frac{3\zeta(3)}{16\pi^2}.
\]
This Gaussian determinant is the exact reference contribution for cubic scalar theories; interaction corrections arise from vacuum diagrams evaluated at nonzero fixed-point couplings [2412.14086].

## 5. Entanglement, defects, and boundary conditions

Spherical entanglement of a free scalar in Euclidean AdS can be mapped by the replica construction to thermal free energy on a topological black hole. The \(n\)-sheeted geometry has Euclidean time periodicity
\[
\tau_E\sim\tau_E+2\pi n,
\]
so the temperature is
\[
T_n=\frac1{2\pi n}.
\]
The Rényi entropy is therefore
\[
S_n=\frac{2\pi n}{n-1}(F_n-F_1).
\]
A heat-kernel calculation yields the leading area law
\[
S_n
=
\frac{1+n}{12n(d-2)(4\pi)^{(d-2)/2}}
\frac{\mathcal A^{(d-2)}(\Sigma)}{\delta^{d-2}}
+\mathcal O(\delta^{-d+4}),
\]
and
\[
S_1
=
\frac{1}{6(d-2)(4\pi)^{(d-2)/2}}
\frac{\mathcal A^{(d-2)}(\Sigma)}{\delta^{d-2}}
+\mathcal O(\delta^{-d+4}).
\]
The leading term is independent of the AdS curvature and reflects ultraviolet locality [1608.00305].

For odd-dimensional AdS, exact finite Rényi terms are available in terms of the effective mass parameter
\[
\nu=\sqrt{m^2\ell_{\mathrm{AdS}}^2+\left(\frac{d-1}{2}\right)^2}.
\]
The dual operator dimensions are
\[
\Delta_\pm=\frac{d-1}{2}\pm\nu.
\]
When \(0<\nu<1\), both standard and alternate quantizations are allowed. A double-trace deformation drives the flow
\[
\mathrm{CFT}^{(N)}\longrightarrow\mathrm{CFT}^{(D)},
\]
and the leading entropy difference is proportional to the shift of the A-type central charge. The classical area remains unchanged at order \(G_N^{-1}\); the difference is quantum and occurs at order \(G_N^0\).

Free scalar defects are constrained by the bulk equation of motion and defect unitarity. A codimension-\(q\) defect preserves
\[
SO(p+1,1)\times SO(q),
\qquad q=d-p.
\]
The bulk-to-defect spectrum consists of two families,
\[
\widehat\Delta_s^{(+)}=\Delta_\phi+s,
\qquad
\widehat\Delta_s^{(-)}=\Delta_\phi+2-q-s.
\]
The \(+\) family is always present, while the \(-\) family can encode nontrivial defect dynamics. For a real scalar, unitary non-monodromy defects can be nontrivial only in codimension three with \(d\geq5\); monodromy defects can be nontrivial in codimension two for \(d\geq4\). Other unitary codimension-\(q\geq2\) defects have bulk-scalar correlators that are Gaussian and fixed by the compatible propagator [2005.02413].

The free scalar can also be viewed through conformal boundary conditions on
\[
H^{p+1}\times S^{q-1}.
\]
Dirichlet-type conditions select the faster falloff and correspond to the trivial defect. Neumann-type conditions select the slower falloff where it satisfies defect unitarity and correspond to nontrivial non-monodromy defects. Such conditions exist only for
\[
q=1,2,3,4.
\]
The Neumann-to-Dirichlet flow is generated by a defect-localized double-trace deformation and supports a defect \(C\)-theorem: the interpolating defect free energy decreases along the flow [2101.02399].

The entanglement structure also probes geometric singularities. For two disjoint spheres in even-dimensional free scalar CFTs, the Rényi mutual information has a spherical-twist OPE and a conformal-block expansion. The leading exchange is the replica bilinear of dimension
\[
\Delta=d-2,
\]
so
\[
I_{A_1,A_2}^{(n)}\sim z^{d-2}
\]
at small cross ratio. The next contributions arise from descendants, replica-antisymmetric spin-one currents of dimension \(d-1\), and dimension-\(d\) scalar and spin-two operators [1612.00114].

For a cubic trihedral corner in \(3+1\) dimensions, the universal logarithmic term is
\[
8v_\alpha\log(L/\delta)
\]
for a cube, with
\[
v_1=0.00286.
\]
The coefficient is positive and has opposite sign and larger magnitude than the smooth-sphere coefficient. Its normalized Rényi dependence is numerically consistent with
\[
\frac{v_\alpha}{v_1}
\simeq
\frac{(\alpha+1)(\alpha^2+1)}{4\alpha^3}.
\]
The similarity to the spherical dependence does not imply equality: the sharp trihedral singularity and the smooth sphere have different universal normalizations and signs [1703.03413].

## 6. Quantization, defects with degrees of freedom, and nonstandard free-scalar models

The ground states of two free scalar Hamiltonians with masses \(m_1\) and \(m_2\) are related by a Bogoliubov transformation. With
\[
\omega_i(p)=\sqrt{p^2+m_i^2},
\]
the oscillators satisfy
\[
a_2(p)=u(p)a_1(p)+v(p)a_1^\dagger(-p),
\]
where
\[
u(p)=\frac12\left(
\sqrt{\frac{\omega_2(p)}{\omega_1(p)}}
+
\sqrt{\frac{\omega_1(p)}{\omega_2(p)}}
\right),
\]
\[
v(p)=\frac12\left(
\sqrt{\frac{\omega_2(p)}{\omega_1(p)}}
-
\sqrt{\frac{\omega_1(p)}{\omega_2(p)}}
\right).
\]
A unitary squeeze operator maps the vacua:
\[
\widehat S|0\rangle_1=|0\rangle_2,
\]
with squeeze parameter
\[
f(p)=\frac12\log\frac{\omega_2(p)}{\omega_1(p)}.
\]
The Schrödinger wavefunctional is Gaussian,
\[
\Psi_0^{(m)}[\widetilde\varphi]
=
\operatorname{Det}\left(\frac{\omega}{\pi}\right)^{1/4}
\exp\left[
-\frac12\int\frac{dp}{2\pi}\,
\omega(p)\widetilde\varphi(p)\widetilde\varphi(-p)
\right].
\]
The construction is formal in infinite volume and requires finite-volume regularization and the usual Bogoliubov implementability condition [1909.13497].

The BMS-invariant free scalar is a different degenerate theory. Its action is
\[
S=\frac1{4\pi}\int d\sigma\,d\tau\,(\partial_\tau\phi)^2,
\]
with no spatial-gradient term, so
\[
\partial_\tau^2\phi=0.
\]
Canonical quantization in the highest-weight vacuum gives
\[
c_L=2,\qquad c_M=0.
\]
The field has a rank-two derivative primary multiplet,
\[
O_0=i\partial_y\phi,
\qquad
O_1=i\partial_x\phi,
\]
and a basic two-point function
\[
\langle\phi(x_1,y_1)\phi(x_2,y_2)\rangle
=
-\frac{y_{12}}{x_{12}}.
\]
The vanishing of \(c_M\) does not truncate the representation theory because an additional quasi-primary
\[
K=-\frac12:(\partial_x\phi)^2:
\]
produces enlarged staggered BMS modules. The torus partition function factorizes as
\[
Z(a,b)=\sqrt{\frac{2}{|b|}}\frac1{\eta(ia)^2}
\]
and is invariant under
\[
a\to\frac1a,
\qquad
b\to-\frac b{a^2}
\]
[2111.04701].

A free bulk scalar can also coexist with genuinely interacting defect degrees of freedom. Couplings to lower-dimensional CFTs, unitary minimal models \(\mathcal M(p,p+1)\), two-dimensional compact bosons, fermions, and \(U(1)\) monodromy sectors produce defect fixed points while leaving the bulk field equation linear. The bulk remains free, but defect-localized interactions modify bulk-to-defect coefficients, displacement correlators, stress-tensor one-point functions, and defect anomalies. In this sense, a free bulk does not imply a trivial defect CFT [2410.01716].

Other uses of “free scalar” depart further from the conventional theory. In quantum free streaming, the free Klein–Gordon field is represented by a Wigner function satisfying
\[
\hbar^2\Box W=4(k^2-m^2c^2)W,
\qquad
k\cdot\partial W=0.
\]
The transport equation is exact, while the classical collisionless Boltzmann equation arises after dropping the \(\hbar^2\Box W\) term and obtaining on-shell support. The exact quantum Wigner function is generally off shell and sign-indefinite [2304.00109].

Projective constructions define free scalar fields through compatible Gaussian measures on discretizations. For massive fields on compact Riemannian manifolds, Whitney finite-element maps compress the continuum covariance
\[
(1+\Delta)^{-1}
\]
to finite-dimensional spaces, producing a projective family whose limit is identified with the continuum massive Gaussian field. For massless fields on polyhedral cell complexes, projective consistency is obtained by renormalizing inner products on discrete gradients, although identification with the standard continuum massless field remains unresolved [2509.12498].

By contrast, constrained mimetic dark-matter models are not free scalar field theories. They contain a mimetic constraint, auxiliary fields, Lagrange multipliers, derivative-dependent functions, and interacting scalar sectors. Their dust-like or dark-energy-like behavior is emergent from constraints rather than from a free quadratic action [2507.05290]. Similarly, scalar extensions of electrodynamics that retain the divergence of the four-potential as a physical scalar degree of freedom define a generalized electrodynamics, not an ordinary free scalar CFT [2608.26839]. A scalar coupled to Einstein gravity through
\[
\Box\Phi=0
\]
is a free matter field in the matter sector, but its gradients back-react on the geometry and generally convert would-be rotating horizons into curvature singularities [2501.10223].

Source: https://www.emergentmind.com/topics/free-scalar