cphysics.

org article 10254 2023-11-12

Conformal Isometries
and the Wave Equation
Randy S

Abstract A conformal isometry is a transformation of spacetime that

leaves the metric invariant up to an overall scale factor that may vary from
one point to the next. This article shows that when N ≥ 2, the classical wave
equation in N -dimensional flat spacetime has symmetries corresponding to all
conformal isometries.

Contents

1 Conformal isometries 3

2 Symmetries of the wave equation 4

3 Symmetries of the wave equation: examples 5

4 Composing fieldomorphisms 6

5 Strategy 8

6 Direct approach 9

7 Approach using the action principle 11

8 Approach using the action principle: details 12

1
© 2018-2023 Randy S
For noncommercial use only
cphysics.org article 10254 2023-11-12

9 Condition (10) 14

10 Condition (11) 15

11 The main result 16

12 A quick review of the embedding space formalism 17

13 Embedding space and the wave equation 18

14 Embedding space and the wave equation: details 19

15 Weyl invariance 21

16 General covariance 23

17 References 24

18 References in this series 24

2
cphysics.org article 10254 2023-11-12

1 Conformal isometries
Let x = (x0 , x1 , ..., xN −1 ) denote a point in N -dimensional spacetime,1 and write
∂a for the partial derivative with respect to xa . For most of this article, spacetime
is taken to be flat, and the components of the metric are
(
±1 if a = b,
ηab =
0 otherwise.

This includes the Minkowski metric as a special case, but the signature is arbitrary
in this article. Use the abbreviations2
x · y ≡ ηab xa y a ∂ · ∂ ≡ η ab ∂a ∂b ,
where η ab are the components of the inverse metric, which in this case happen to
be the same as the original components ηab .
In this context, a conformal isometry3 is a diffeomorphism4 x → x̂ for which
dx̂ · dx̂ = Ω2 (x) dx · dx. (1)
Examples:3
• A conformal isometry for which Ω(x) = 1 is called an (ordinary) isometry.
Poincaré transformations are ordinary isometries in Minkowski spacetime.
• x → λx with λ > 0 is called a dilation. In this case, Ω(x) = λ.
• x → x/(x · x) is called an inversion. In this case, Ω(x) = 1/(x · x).
When N ≥ 3, all other conformal isometries are generated by these three examples.3
1
The superscripts are indices, not exponents.
2
I’m using the standard summation convention, with an implied sum over each index that occurs as both a
superscript and subscript in the same term.
3
Article 38111
4
In this context, a diffeomorphism is a smooth rearrangement of the points of spacetime (article 93875), but
not necessarily defined at all points in spacetime. The inversion x → x/(x · x) is defined only where x · x 6= 0.

3
cphysics.org article 10254 2023-11-12

2 Symmetries of the wave equation

This article is about symmetries of the wave equation

∂ · ∂φ(x) = 0 (2)

in N -dimensional spacetime, where φ(x) is a classical scalar field. The goal is to

show that if φ(x) satisfies the wave equation with N ≥ 2 and if x → x̂ is any
conformal isometry (equation (1)), then the field φ̂(x) defined by5


(N −2)/2
φ̂(x) ≡ ω(x)φ x̂(x) ω(x) ≡ Ω(x) (3)

also satisfies the wave equation at all points where x̂ is defined:

∂ · ∂ φ̂(x) = 0. (4)

The transformation φ(x) → φ̂(x) defined by equation (3) is sometimes called a

conformal transformation, but that name is also used for other things.6 To
avoid equivocation, this article will call it a conformal fieldomorphism.7,8 The
goal is to show that if N ≥ 2, then all conformal fieldomorphisms are symmetries
of the wave equation.9

5
Equation (1) doesn’t specify the sign of Ω(x), but equation (3) assumes the convention Ω(x) > 0.
6
Appendix D in Wald (1984) uses the name conformal transformation for what many physicists (including this
article) call a Weyl transformation. Section 7.6.2 in Nakahara (1990) uses the name conformal transformation for
what this article calls a conformal isometry.
7
This name is not standard. It builds on the name fieldomorphism that article 00418 used for a transformation
like φ(x) → φ̂(x) ≡ φ(x̂), where x → x̂ is an arbitrary diffeomorphism.
8
Section 15 reviews the concept of a Weyl transformation, mainly for the purpose of distinguishing it from what
this article calls a conformal fieldomorphism.
9
When N ≥ 3, they are also symmetries of the more general equation ∂ · ∂φ ∝ φ(N +2)/(N −2) . Notice that the
exponent is an integer only if N = 4 or N = 6.

4
cphysics.org article 10254 2023-11-12

3 Symmetries of the wave equation: examples

Article 49705 showed that if x → x̂ is any diffeomorphism for which dx̂·dx̂ = dx·dx,
then the corresponding fieldomorphism (3) is a symmetry of the wave equation. In
this case, the factor ω(x) is equal to 1.
The easiest example with ω(x) 6= 1 is the conformal fieldomorphism correspond-
ing to a dilation x → λx with constant scale factor λ 6= 1. The fact that this is a
symmetry of the wave equation should be clear by inspection. The factor ω(x) is
still independent of x in this case.
The easiest example in which ω(x) is not independent of x is the conformal
fieldomorphism correponding to an inversion x → x/(x · x). In this case, the scale
function in equation (1) turns out to be10 Ω(x) = 1/(x · x), so the transformed field
is  x  1
φ̂(x) = ω(x) φ with ω(x) = . (5)
x·x (x · x)(N −2)/2
This is defined wherever x · x 6= 0. Straightforward calculation shows that this
function ω(x) satisfies

∂ · ∂ω(x) = 0


∂ · ∂ ω(x)x̂ = 0

wherever x · x 6= 0. Section 6 will use these properties of ω(x) to show that the
transformation φ(x) → φ̂(x) defined by (5) is a symmetry of the wave equation.

10
Article 38111

5
cphysics.org article 10254 2023-11-12

4 Composing fieldomorphisms
The composition of two conformal isometries (the result of applying them sequen-
tially) is another conformal isometry. This is clear from the definition (1). This
section shows that the composition of two conformal fieldomorphisms (3) is another
conformal fieldomorphism that agrees with the composition of the corresponding
conformal isometries.
Since we’re dealing with two conformal isometries now, we need to use a notation
that distinguishes between them. If we use the operator-like notation
x → σ1 x x → σ2 x
for the two conformal isometries, then their composition (the result of applying σ1
and then σ2 ) is
x → σ12 x ≡ σ2 σ1 x.
These are conformal isometries, so the effect of each one on the line element may
be written
d(σ1 x) · d(σ1 x) = Ω21 (x) dx · dx
d(σ2 x) · d(σ2 x) = Ω22 (x) dx · dx
d(σ2 σ1 x) · d(σ2 σ1 x) = Ω212 (x) dx · dx.
Use the first two equations to get
d(σ2 σ1 x) · d(σ2 σ1 x) = Ω22 (σ1 x) d(σ1 x) · d(σ1 x)
= Ω22 (σ1 x)Ω21 (x) dx · dx,
which gives this equation for the scale function Ω12 of the composite transformation:
Ω12 (x) = Ω2 (σ1 x)Ω1 (x). (6)
Now, for any function φ(x) and any n ∈ {1, 2, 12}, define a transformation φ → σn φ
by
σn φ(x) ≡ ωn (x)φ(σn x) (7)

6
cphysics.org article 10254 2023-11-12

with
ωn (x) ≡ (Ωn (x))(N −2)/2 , (8)
as in equation (3). Equations (7) with n ∈ {1, 2} imply

σ2 σ1 φ(x) = ω1 (x)σ2 φ(σ1 x) = ω1 (x)ω2 (σ1 x)φ(σ2 σ1 x),

and comparing this to equation (7) with n = 12 gives

ω12 (x) = ω2 (σ1 x)ω1 (x), (9)

which is consistent with equations (6) and (8). This completes the derivation.

E
This would all still be true if ωn (x) = Ωn (x) for any exponent E, but the
exponent shown in (8) is special because it makes ωn (x) satisfy the conditions (10)
and (11). This is easy to check by direct calculation when σn is the inversion
σn x = x/(x · x). The following sections explain how it can be inferred for other
conformal isometries.
The result derived in this section can be expressed using the language of category
theory.11 Conformal isometries may be regarded as the morphisms in a category
with just one object. That one object is the smooth manifold (spacetime) on
which the conformal isometries act. Conformal fieldomorphisms may be regarded
as the morphisms in another category with just one object. That one object is the
set of scalar fields that satisfy the wave equation (wherever they’re defined). In
both cases, the composition of two morphisms is another morphism, as required by
the definition of category. By expressing conformal fieldomoprhisms in terms of
conformal isometries, equations (3) define a functor from the conformal-isometries
category to the conformal-fieldomorphisms category. A functor converts morphisms
in one category to morphisms in another category, respecting composition.

11
Spivak (2013) and McLarty (1992) are relatively inviting introductions to category theory.

7
cphysics.org article 10254 2023-11-12

5 Strategy
The goal is to show that the transformations φ → φ̂ defined in equation (3) –
conformal fieldomorphisms – are symmetries of the wave equation (2) if N ≥ 2.
Here’s an outline:

• Section 6 will show that a conformal fieldomorphism is a symmetry of the

wave equation if ω satisfies the conditions

∂ · ∂ω(x) = 0 (10)


∂ · ∂ ω(x)x̂ = 0 (11)

wherever it is defined.

• Sections 7-8 will use a different method, involving the action principle, to
show that a conformal fieldomorphism is a symmetry of the wave equation if
ω satisfies the condition (10).12

• Sections 9-10 will show that if two conformal isometries both satisfy the
conditions (10)-(11), then so does their composition.13

• Sections 11 use that result to show that every conformal isometry satisfies
the conditions (10)-(11) if N ≥ 2. This implies that all of the corresponding
conformal fieldomorphisms are symmetries of the wave equation.

For extra fun, sections 12-14 review how the embedding space formalism can be
used to study conformal fieldomorphisms.14
12
This approach doesn’t use the condition (11) explicitly, but it does use the fact that ω(x) is defined by equations
(1) and (3), which the subsequent sections use to show that ω satisfies the conditions (10) and (11).
13
Actually the proof shown here is incomplete, because the composition of two conformal isometries may be defined
everywhere (by continuation) even if the conformal isometries that went into it are not. Example: composing the
inversion x → x/(x · x) with itself gives the identity transformation, which is defined everywhere, in the same sense
that (x · x)/(x · x) = 1. The proof shown here doesn’t include the continuation step.
14
Article 38111 uses the embedding space formalism to study conformal isometries (equation (1)).

8
cphysics.org article 10254 2023-11-12

6 Direct approach
This section shows that if φ(x) is any solution of the wave equation (2), and if ω
satisfies the conditions (10) and (11),15 then the new field φ̂(x) defined in (3) is
another solution of the wave equation.
Start with the elementary identities


∂ · ∂ φ̂(x) = ∂ · ∂ ω(x)φ(x̂)



= ∂ · ∂ω(x) φ(x̂) + 2 ∂ω(x) · ∂φ(x̂) + ω(x)∂ · ∂φ(x̂). (12)

Use the abbreviation ∂ˆa ≡ ∂/∂ x̂a to get

∂φ(x̂) = (∂ x̂a )∂ˆa φ(x̂) (13)

∂ · ∂φ(x̂) = ∂ · (∂ x̂a )∂ˆa φ(x̂)

= (∂ · ∂ x̂a )∂ˆa φ(x̂) + (∂ x̂a ) · ∂ ∂ˆa φ(x̂)

= (∂ · ∂ x̂a )∂ˆa φ(x̂) + (∂ x̂a ) · (∂ x̂c )∂ˆc ∂ˆa φ(x̂) (14)
To continue, use the general identity
dx̂a = dxb (∂b x̂a ) (15)
to see that (1) implies
ηcd (∂a x̂c )(∂b x̂d ) = Ω2 (x)ηab , (16)
which in turn implies16
(∂ x̂a ) · (∂ x̂b ) = Ω2 (x)η ab . (17)
Equation (17) implies

(∂ x̂a ) · (∂ x̂c )∂ˆc ∂ˆa = Ω2 (x)∂ˆ · ∂,

ˆ
15
For the rest of this article, the qualification “wherever the conformal isometry is defined” is understood.
16
To derive this, let η be the matrix with components ηab , and let M be the matrix with components Mab ≡ ∂a x̂b .
Then equation (16) is M ηM T = Ω2 η. Take the matrix inverse of both sides and then re-arrange to get M T η −1 M =
Ω2 η −1 , which is equation (17).

9
cphysics.org article 10254 2023-11-12

and the assumption that φ(x) satisfies the wave equation (∂ · ∂φ(x) = 0) implies
∂ˆ · ∂φ(x̂)
ˆ = 0 just by relabeling the coordinates, so the last term in equation (14)
is zero. Use these results in (12) to get

∂ · ∂ φ̂(x) = ∂ · ∂ω(x) φ(x̂) + Γa ∂ˆa φ(x̂)

(18)

with
Γa ≡ 2 ∂ω(x) · (∂ x̂a ) + ω(x)(∂ · ∂ x̂a ).


(19)
The condition (10) implies that the quantity Γa may also be written

Γa = ∂ · ∂ ω(x)x̂a .


(20)

Equations (18) and (20) show that φ̂(x) satisfies the wave equation if ω(x) satisfies
the conditions (10) and (11).

10
cphysics.org article 10254 2023-11-12

7 Approach using the action principle

This section shows that if φ(x) is any solution of the wave equation (2), and if ω
satisfies the condition (10), then the new field φ̂(x) defined in (3) is another solution
of the wave equation. In contrast to the approach that was used in section 6, the
approach used here involves only first-order derivatives of the field.
The wave equation (2) can be written as17

δS[φ]
=0
δφ(x)

where S[φ] is the action18

Z
S[φ] = dN x (∂φ) · (∂φ). (21)

Now suppose that x → x̂ is any conformal isometry, not necessarily satisfying the
condition (10). Section 8 shows that replacing the original scalar field φ(x) with
the new scalar field (3) has this effect on the action:
Z Z
S[φ̂] = ±S[φ] + d x ∂ · (something) − dN x φ2 (x̂)ω(x)∂ · ∂ω(x).
N
(22)

Adding a total-derivative term to S doesn’t affect the equation of motion (21), so

the result (22) implies that if φ(x) satisfies the wave equation and ω satisfies the
condition (10), then φ̂(x) also satisfies the wave equation.

17
Article 49705
18
I’m omitting a conventional but inconsequential overall factor of 1/2.

11
cphysics.org article 10254 2023-11-12

8 Approach using the action principle: details

To derive (22), start with
Z
S[φ̂] = dN x ∂ φ̂(x)) · ∂ φ̂(x)). (23)

Equation (3) gives

∂a φ̂(x) = ω(x)∂a φ(x̂) + φ(x̂)∂a ω(x),
and using this in (23) gives

S[φ̂] = S1 [φ] + S2 [φ] (24)

with Z
S1 [φ] ≡ dN x ω 2 (x) ∂φ(x̂)) · ∂φ(x̂)) (25)

and
Z
dN x 2ω(x)φ(x̂) ∂φ(x̂) · ∂ω(x) + φ2 (x̂) ∂ω(x) · ∂ω(x)





S2 [φ] ≡
Z Z
= d x ∂ · ω(x)φ (x̂)∂ω(x) − dN x φ2 (x̂)ω(x)∂ · ∂ω(x).
N 2


(26)

To get a more useful expression for S1 [φ], use the general identity

∂a = (∂a x̂c )∂ˆc

to write it as
Z
dN x ω 2 (x)(∂ x̂a ) · (∂ x̂b ) ∂ˆa φ(x̂) ∂ˆb φ(x̂) .



S1 [φ] = (27)

Use (17) in (27) to get

Z
ˆ
dN x ω 2 (x)Ω2 (x) ∂φ(x̂) ˆ



S1 [φ] = · ∂φ(x̂) . (28)

12
cphysics.org article 10254 2023-11-12

The identity (15) implies

dN x̂
dN x = (29)
|∂ x̂|
where |∂ x̂| is the determinant of the matrix with components ∂b x̂a , and equation
(16) implies

N
|∂ x̂|2 = Ω2 (x) . (30)
Use (29), (30), and the definition of ω in (28) to get
Z
ˆ
S1 [φ] = ± dN x̂ ∂φ(x̂) ˆ



· ∂φ(x̂) = ±S[φ]. (31)

Combining (24), (26), and (31) gives the promised result (22).

13
cphysics.org article 10254 2023-11-12

9 Condition (10)
As in section 4, consider two conformal isometries σ1 and σ2 . This section shows
that if ω1 (x) and ω2 (x) both satisfy condition (10), then so does ω12 (x). The
notation here is the same as in section 4, and the abbreviations
∂
x̂ ≡ σ1 x ∂ˆ ≡
∂ x̂
will also be used.
Equation (9) combined with ∂ · ∂ω1 (x) = 0 implies


∂ · ∂ω12 (x) = 2 ∂ω1 (x) · ∂ω2 (σ1 x) + ω1 (x)∂ · ∂ω2 (σ1 x)
= 2 ∂ω1 (x) · (∂ x̂a )∂ˆa ω2 (x̂) + ω1 (x)∂ · (∂ x̂a )∂ˆa ω2 (x̂)

= 2 ∂ω1 (x) · (∂ x̂a )∂ˆa ω2 (x̂) + ω1 (x)(∂ · ∂ x̂a )∂ˆa ω2 (x̂)

+ ω1 (x)(∂ x̂b ) · (∂ x̂a )∂ˆb ∂ˆa ω2 (x̂).

Use equation (17) and ∂ˆ · ∂ω

ˆ 2 (x̂) = 0 to see that the last term is zero, which leaves

∂ · ∂ω12 (x) = 2 ∂ω1 (x) · (∂ x̂a ) + ω1 (x)(∂ · ∂ x̂a ) ∂ˆa ω2 (x̂)




= ∂ · ∂ ω1 (x)x̂a − x̂a ∂ · ∂ω1 (x)) ∂ˆa ω2 (x̂).




The assumption that ω1 (x) satisfies the conditions (10) and (11) implies that the
quantity in square brackets is zero, so this proves that ω12 (x) satisfies condition
(10).

14
cphysics.org article 10254 2023-11-12

10 Condition (11)
This section shows that if ω1 (x) and ω2 (x) both satisfy condition (11), then so does
ω12 (x). More explicitly: if


∂ · ∂ ω1 (x)σ1 x = 0 (32)


∂ · ∂ ω2 (x)σ2 x = 0, (33)

then

∂ · ∂ ω12 (x)σ12 x = 0. (34)
The notation here is the same as in section 9.
Equation (9) combined with ∂ · ∂ω1 (x) = 0 implies

∂ · ∂ ω12 (x)(σ12 x)a = ∂ · ω2 (x̂)(σ2 x̂)a ∂ω1 (x) + ω1 (x)∂ ω2 (x̂)(σ2 x̂)a



h
i
a b ˆ a
= ∂ · ω2 (x̂)(σ2 x̂) ∂ω1 (x) + ω1 (x)(∂ x̂ )∂b ω2 (x̂)(σ2 x̂)

Use equations (17) and (33) in the last term to get

∂ · ∂ ω12 (x)(σ12 x)a = ∂ · [ω2 (x̂)(σ2 x̂)a ∂ω1 (x)] + ∂ · ω1 (x)(∂ x̂b ) ∂ˆb ω2 (x̂)(σ2 x̂)a .

 

Use ∂ · ∂ω1 (x) = 0 in the first term to get

∂ · ∂ ω12 (x)(σ12 x)a = ∂ [ω2 (x̂)(σ2 x̂)a ] · ∂ω1 (x) + ∂ · ω1 (x)(∂ x̂b ) ∂ˆb ω2 (x̂)(σ2 x̂)a

 

= (∂ x̂b )∂ˆb [ω2 (x̂)(σ2 x̂)a ] · ∂ω1 (x) + ∂ · ω1 (x)(∂ x̂b ) ∂ˆb ω2 (x̂)(σ2 x̂)a ,
 

and use ∂ · ∂ω1 (x) = 0 again to get

∂ · ∂ ω12 (x)(σ12 x)a = ∂ˆb ω2 (x̂)(σ2 x̂)a ∂ · ∂ ω1 (x)x̂b .




Equation (32) implies that the quantity in square brackets is zero, which proves
(34).

15
cphysics.org article 10254 2023-11-12

11 The main result

When N ≥ 3, the group of conformal isometries is generated by ordinary isometries,
dilations, and inversions.19 We already know20 that these transformations satisfy
the conditions (10) and (11), so the results derived in sections 6-10 imply that all
conformal isometries correspond – via (3) – to symmetries of the wave equation
when N ≥ 3.
When N = 2, ω is independent of x, so the conditions (10) and (11) reduce to
the condition
∂ · ∂ x̂ = 0. (35)
The line element may be written

dx · dx ∝ du dv

with
u ≡ x0 + x1 v ≡ x0 − x1
if the signature is lorentzian (Minkowski spacetime), or with

u ≡ x0 + ix1 v ≡ x0 − ix1

if the signature is euclidean, where i2 = −1. In either case, the condition (35) may
be written21
∂ ∂ ∂ ∂
û = 0 v̂ = 0.
∂u ∂v ∂u ∂v
According to sections 2.3 and 2.5 Schottenloher (2008), all conformal isometries
satisfy these conditions when N = 2, so the conclusion of the preceding paragraph
extends to N ≥ 2.

19
Section 1
20
Section 3
21
In the euclidean case, ∂/∂u ≡ 12 (∂0 − i∂1 ) and ∂/∂v ≡ 12 (∂0 + i∂1 ).

16
cphysics.org article 10254 2023-11-12

12 A quick review of the embedding space formalism

Article 38111 introduces the embedding space formalism, which relates at least
some conformal isometries of the original N -dimensional spacetime M to origin-
preserving ordinary isometries of an (N + 2)-dimensional spacetime in which M is
embedded.22 This section reviews the idea, and section 13 applies it to the wave
equation.
Consider an (N +2)-dimensional manifold with coordinates X ≡ (X 0 , X 1 , ..., X N −1 )
and Y and Z, where Y and Z are individual real-valued coordinates, and with the
metric defined implicitly by the line element23
dX · dX + dY 2 − dZ 2 . (36)
I’ll call this the ambient manifold.24,25 Let C be the “cone” defined by
X · X + Y 2 − Z 2 = 0, (37)
and let P be the hyperplane defined by
Y +Z =R (38)
for some constant R. We can use
Xa a
x ≡ (39)
Y +Z
as coordinates on the N -dimensional intersection C ∩ P, everywhere except where
Y + Z = 0. Article 38111 shows that the induced metric in this N -dimensional
manifold is conformally equivalent the Minkowski metric dx · dx and that origin-
preserving ordinary isometries of the ambient manifold correspond to conformal
isometries of the N -dimensional manifold.
22
Weinberg (2010) describes how tensor fields in Minkowski spacetime can be represented using this formalism.
23
As in the preceding sections, X · X ≡ ηab X a X b for any quantity X with N components.
24
It’s also called the embedding space, because the spacetime of interest – Minkowski spacetime – will be
embedded inside it.
25
Mnemonic: I’m using uppercase letters for coordinates in the bigger (higher-dimensional) manifold, and lowercase
letters for coordinates in the smaller manifold (Minkowski spacetime).

17
cphysics.org article 10254 2023-11-12

13 Embedding space and the wave equation

The embedding space formalism can be used to relate the symmetries of the
wave equation in N -dimensional spacetime to ordinary isometries in the (N + 2)-
dimensional embedding space. Let Φ(X, Y, Z) be a function that satisfies this
(N + 2)-dimensional version of the wave equation:

∂X · ∂X Φ + ∂Y2 Φ − ∂Z2 Φ = 0. (40)

All origin-preserving isometries of (36) correspond to symmetries of equation (40)

in the usual way. To relate these to the symmetries (3) of the wave equation in
N -dimensional Minkowski spacetime, consider functions of the form26
 
−(N −2)/2 X Y −Z
Φ(X, Y, Z) = (Y + Z) f , . (41)
Y +Z Y +Z

Section 14 shows that applying the differential operator

∂X · ∂X + ∂Y2 − ∂Z2 (42)

to such a function and then imposing the constraints (37) and (38) gives the same
result as applying the differential operator ∂x · ∂x to
 
−(N +2)/2 −x · x
φ(x) ≡ R f x, . (43)
R2

In particular, if the function (41) satisfies (40) everywhere on C ∩ P, then the

function (43) satisfies the usual wave equation in Minkowski spacetime.

26
This is equation (37) in Bars (2000), but beware: that paper uses the same symbol λ for two different things on
the same page. The two different meanings of λ are introduced implicitly in equations (26) and (31) of that paper.

18
cphysics.org article 10254 2023-11-12

14 Embedding space and the wave equation: details

This section shows that applying the differential operator (42) to any function of
the form (41) and then imposing the constraints (37) and (38) gives the same result
as applying the differential operator ∂x · ∂x to the function (43).27
The coordinates X, Y, Z can be written as

Y +Z =σ Y −Z =ρ X = σx

wherever σ 6= 0. This is consistent with (39). Use the generic identity28

X ∂
a
X
a ∂
dX + dY ∂Y + dZ ∂ Z = dx + dρ ∂ρ + dσ ∂σ (44)
a
∂X a a
∂x a

to infer29
∂ ∂
= σ
∂xa ∂X a
∂ 1
= (∂Y − ∂Z )
∂ρ 2
∂ X ∂ 1 1X a ∂ 1
= xa a
+ (∂ Y + ∂ Z ) = x a
+ (∂Y + ∂Z )
∂σ a
∂X 2 σ a ∂x 2

and uses these to infer that the differential operator in (40) may be written

∂X · ∂X + ∂Y2 − ∂Z2 = ∂X · ∂X + (∂Y + ∂Z )(∂Y − ∂Z )

!
1 ∂ 1 X ∂ ∂
= 2 ∂x · ∂x + 4 − xa a .
σ ∂σ σ a ∂x ∂ρ
27
The approach used here is a slight variation of the approach used in Bars (2000), section 4.1.
28
This identity holds for any two coordinate systems X, Y, Z and x, ρ, σ, no matter how they’re related to each
other (as long as each one can be written in terms of the other).
29
To deduce this, write dX, dY, dZ in terms of dx, dρ, dσ and substitute those expressions for dX, dY, dZ into the
left-hand side of (44).

19
cphysics.org article 10254 2023-11-12

Equation (41) may also be written

Φ = σ −(N −2)/2 f (x, ρ/σ).

Any function of this form satisfies

! !
∂ 1 X ∂ ∂ ∂ 1X a ∂ 1
− xa a Φ= − x Φ1
∂σ σ a ∂x ∂ρ ∂σ σ a ∂xa σ
!
−1 N ρ X ∂
= Φ1 + Φ2 + xa Φ1 (45)
σ2 2 σ a
∂xa

with
∂
Φ1 ≡ σ −(N −2)/2 f (x, y)
∂y y=ρ/σ
∂2
Φ2 ≡ σ −(N −2)/2 f (x, y) .
∂y 2 y=ρ/σ

This is true even if Φ doesn’t satisfy (40). On the other hand, any function of the
form (43) satisfies
" #
N x·x X ∂
∂x ·∂x f (x, y) = ∂x · ∂x f (x, y) − f1 (x, y) + 2 f2 (x, y) − xa a f1 (x, y)
2 R a
∂x
y=−x·x/R2

with
∂ ∂2
f1 (x, y) ≡ f (x, y) f2 (x, y) ≡ 2 f (x, y).
∂y ∂y
According to equations (37) and (38), points on the intersection C ∩ P satisfy
−x · x
σ=R ρ= ,
R
Use this in the last line of equation (45) to complete the derivation.

20
cphysics.org article 10254 2023-11-12

15 Weyl invariance
Section 2 defined some symmetries of the wave equation. Those symmetries mix
the scalar field’s values at different points of spacetime with each other in the same
way that conformal isometries mix the points of spacetime itself, and they don’t
transform the metric field at all.
In contrast, a Weyl transformation30 does affect the metric field but doesn’t
mix the fields’ values in different regions with each other:
φ(x) → Ωs (x)φ(x) gab (x) → Ω2 (x)gab (x) (46)
where s ∈ R is called the conformal weight of the scalar field. To say anything
about what a Weyl transformation does to the action (or equation of motion) for
a scalar field, we first need to specify how the action depends on the metric field.
One natural choice is31
Z p
S[g, φ] = d x |g| g ab (∂a φ)(∂b φ)
N
(47)

where gab (x) are the components of a metric tensor, g ab (x) are the components of
its inverse, and |g|(x) is its determinant. When N = 2, the action (47) is invariant
under Weyl transformations (46) with s = 0.32 When N 6= 2, the action (47) is not
invariant under Weyl transformations (46) for any s, but the modified action
 
N −2
Z p
0 N ab 2
S [g, φ] = d x |g| g (∂a φ)(∂b φ) + Rφ (48)
4(N − 1)
is invariant up to a total derivative under Weyl transformations with33,34 s =
30
Article 38111
31
Section 16 reviews why this choice is natural.
32
p In abmore detail: the only part of the integrand affected by this transformation is the metric-dependent factor
|g| g , and the transformation (46) leaves this invariant when N = 2.
33
Section 22.3 in Blau (2021). Beware that Blau quietly ignores the total-derivative term, acknowledging its
existence only in the text above equation (22.115).
34
Appendix D in Wald (1984) derives this using the equation of motion instead of the action. The equation of
motion is not affected by the total-derivative term that a typical Weyl transformation adds to the action.

21
cphysics.org article 10254 2023-11-12

(2 − N )/2 when N ≥ 2, where R is the Ricci scalar constructed from the cur-
vature tensor.35 This property of (48) is called Weyl invariance or conformal
invariance,36 but beware that the names conformal transformation, conformal
invariance, and conformal symmetry are all overloaded in the physics literature.
A symmetry that doesn’t mix the fields’ values in different regions with each
other is called an internal symmetry, so Weyl invariance is an example of an
internal symmetry.

35
This is not related to the constant R in section 12.
36
Wald (1984), appendix D

22
cphysics.org article 10254 2023-11-12

16 General covariance
The action (47) is a natural choice because it has a property that is sometimes
called general covariance: for any diffeomorphism x → x̂, replacing the original
fields φ(x) and gab (x) with the new fields

φ̂(x) ≡ φ(x̂) ĝab (x) ≡ gcd (x̂)(∂a x̂c )(∂b x̂d ) (49)

leaves the action invariant:

S[ĝ, φ̂] = S[g, φ].
To derive this, use the identities37

∂a φ̂(x) = (∂a x̂b )∂ˆb φ(x̂) ĝ ab (x) = g cd (x̂)(∂ˆc xa )(∂ˆd xb ) (50)

to get
ĝ ab (x) ∂a φ̂(x) ∂b φ̂(x) = g ab (x̂) ∂ˆa φ(x̂) ∂ˆb φ(x̂) ,

and use the definition of ĝab (x) to get

p p
dN x |ĝ(x)| = dN x̂ |g(x̂)|.

The modified action (47) has this property, too.

37
To derive the second identity, let G(x̂) and Ĝ(x) be the matrices with components gab (x̂) and ĝab (x), respectively.
Let M be the matrix with components Mab = ∂a x̂b . Then the second equation in (49) is Ĝ(x) ≡ M G(x̂)M T . Take
the matrix inverse of both sides to get Ĝ−1 (x) = (M T )−1 G−1 (x̂)M −1 . This is the second equation in (50).

23
cphysics.org article 10254 2023-11-12

17 References
Bars, 2000. “Two-Time Physics in Field Theory” Phys. Rev. D 62: 046007,
https://arxiv.org/abs/hep-th/0003100

Blau, 2021. “Lecture Notes on General Relativity (Last update November 15,
2021)” http://www.blau.itp.unibe.ch/GRLecturenotes.html

McLarty, 1992. Elementary Categories, Elementary Toposes. Clarendon Press

Nakahara, 1990. Geometry, Topology, and Physics. Adam Hilger

Schottenloher, 2008. A Mathematical Introduction to Conformal Field The-

ory (Second Edition). Springer, http://www.mathematik.uni-muenchen.
de/~schotten/LNP-cft-pdf

Spivak, 2013. “Category Theory for Scientists” https://ocw.mit.edu/courses/

mathematics/18-s996-category-theory-for-scientists-spring-2013/textbook

Wald, 1984. General Relativity. University of Chicago Press

Weinberg, 2010. “Six-dimensional Methods for Four-dimensional Conformal Field

Theories” Phys. Rev. D 82: 045031, https://arxiv.org/abs/1006.3480

18 References in this series

Article 00418 (https://cphysics.org/article/00418):
“Diffeomorphisms, Tensor Fields, and General Covariance” (version 2022-02-20)

Article 38111 (https://cphysics.org/article/38111):

“Conformal Isometries in the Embedding Space Formalism” (version 2023-11-12)

Article 49705 (https://cphysics.org/article/49705):

“Classical Scalar Fields and Local Conservation Laws” (version 2023-11-12)

24
cphysics.org article 10254 2023-11-12

Article 93875 (https://cphysics.org/article/93875):

“From Topological Spaces to Smooth Manifolds” (version 2023-11-12)

25