Conformal Transformations and Causal Invariance

A conformal transformation is a pointwise rescaling of the metric by a strictly positive smooth scalar field, $\tilde g_{ab}(x)=\Omega^2(x)\,g_{ab}(x)$ . It is not a diffeomorphism — it does not relabel points — and it is not, in general, a symmetry: it genuinely changes the geometry, stretching and shrinking every proper length and duration by a position-dependent factor. What it conspicuously leaves alone is the ratio of any two infinitesimal displacements at a point. It preserves angles, and in Lorentzian signature it preserves the local light-cone. That single fact is the seed of the entire subject: conformal transformations are the causality-preserving rescalings of spacetime. The reason is almost embarrassingly simple. Whether a vector $V^a$ is timelike, null, or spacelike is decided by the sign of $g_{ab}V^aV^b$ , and multiplying a number by $\Omega^2>0$ can never flip its sign. So two conformally related metrics classify every vector identically: they share the same light-cones, the same futures and pasts $I^\pm$ , $J^\pm$ , the same achronal sets, domains of dependence, and verdict on global hyperbolicity. Causality is a conformal invariant — it depends only on the conformal class $[g_{ab}]$ , not on the representative metric. This invariance is the engine behind conformal compactification. Choose $\Omega\to 0$ at infinity and the infinite reaches of a spacetime are pulled into the interior of a finite, unphysical spacetime, with the limit set $\Omega=0$ attached as a genuine boundary — conformal infinity. Because null geodesics survive the rescaling, the causal diagram drawn on that finite canvas faithfully records who can signal whom in the physical spacetime. This is the origin of Penrose–Carter diagrams and the rigorous definitions of null infinity $\mathscr{I}^\pm$ , spatial infinity $i^0$ , and timelike infinity $i^\pm$ that underpin asymptotic flatness and the theory of gravitational radiation.

Key takeaways

A conformal transformation rescales the metric by a positive smooth function, $\tilde g_{ab}=\Omega^2 g_{ab}$ , leaving angles and the light-cone structure untouched while altering every proper length and time.
Because the sign of $g_{ab}V^aV^b$ survives multiplication by $\Omega^2>0$ , the timelike/null/spacelike split — and hence all of $I^\pm$ , $J^\pm$ , achronal sets and domains of dependence — is conformally invariant.
Null geodesics map to null geodesics (only the affine parameter changes); timelike and spacelike geodesics are generally destroyed, so light rays alone carry causal connectivity across the conformal class.
Conformal compactification uses an $\Omega$ that vanishes at infinity to bring asymptotic regions to a finite boundary — the Penrose diagram — with $\mathscr{I}^\pm$ , $i^0$ , $i^\pm$ appearing as edges and points.
The Weyl tensor $C^a{}_{bcd}$ is conformally invariant in its mixed form; the Ricci and scalar curvatures transform inhomogeneously, carrying all the $\Omega$ -dependence.
Key idea: multiplying the metric by a positive function $\Omega^2$ leaves the sign of $g_{ab}V^aV^b$ untouched, so the timelike/null/spacelike split — and hence the entire causal structure — is invariant; only distances, durations, and the affine (not the path) parametrization of null geodesics change.

The core insight

Multiplying the metric by a positive function $\Omega^2$ leaves the sign of $g_{ab}V^aV^b$ untouched, so the timelike/null/spacelike split — and hence the entire causal structure — is invariant; only metric distances, durations, and the affine (not the path) parametrization of null geodesics change.

Where this lives today

Conformal invariance of causal structure is the working foundation of how relativists reason about the large-scale universe and about gravitational radiation. Every Penrose–Carter diagram in a modern paper on Schwarzschild, Reissner–Nordström, Kerr, de Sitter, or FLRW spacetimes relies on it: the diagrams are trustworthy precisely because a conformal rescaling cannot alter who can signal whom, so a horizon, a trapped region, or a failure of global hyperbolicity can be read off by inspection. The coordinate-free definition of asymptotic flatness — a smooth conformal completion with boundary $\mathscr{I}^\pm$ where $\Omega=0$ , $\mathrm{d}\Omega\neq0$ — is the setting in which Bondi mass, the BMS asymptotic symmetry group, and the memory effect are defined, all alive in current gravitational-wave theory because outgoing radiation is characterized by its data at $\mathscr{I}^+$ . In cosmology the same idea appears as conformal time $\eta$ : FLRW spacetimes are conformally flat, $g_{ab}=a^2(\eta)\bar g_{ab}$ , so particle horizons are computed in the frame where null cones are trivial and then dressed with the scale factor — the logic behind the horizon problem that motivates inflation.

The physical picture

Picture the light-cone at each point as a rigid geometric object — a double cone of directions splitting the tangent space into future, past, and elsewhere. A conformal transformation is like re-upholstering the spacetime fabric with a position-dependent magnifying glass: it changes how far apart things really are, but it cannot tip, open, or close any cone. Tipping a cone would mean turning some timelike direction spacelike, which means flipping the sign of $g_{ab}V^aV^b$ , and a strictly positive factor $\Omega^2$ never flips a sign. The first figure makes this concrete: the same three test vectors keep their classifications, and the sign of each contraction merely scales, never turns over.

Figure

Causal classification is conformally invariant — under g̃ = Ω²g with Ω = 2 the three contractions scale by Ω² = 4 (−0.64 → −2.56, 0 → 0, +0.75 → +3.0) but never change sign, so the timelike/null/spacelike split and the whole light-cone are unchanged

Now think of conformal compactification as drawing the whole infinite universe on a finite sheet using a rubber ruler that shrinks faster and faster the further out you go. Infinitely distant events are pulled in to a finite edge, yet because the rubber ruler preserves angles, the 45-degree light-cones drawn on the paper still tell you exactly who can send a signal to whom — that fidelity is the entire point of a Penrose diagram. The second figure is the canonical example: all of Minkowski spacetime, mapped into a finite triangle inside the Einstein static universe, with its infinity resolved into the null surfaces $\mathscr{I}^\pm$ , the point $i^0$ where space ends, and the points $i^\pm$ where time ends.

Figure

The conformal (Penrose) diagram of Minkowski spacetime: arctan compactification u = tan q, v = tan p with Ω = 2 cos p cos q maps the infinite spacetime to a finite triangle where every light ray runs at 45°; null infinity 𝓘± are the slanted null edges (Ω = 0), i⁰ is spatial infinity, i± are timelike infinity, and the left edge is the worldline r = 0

Curvature splits cleanly the same way: the Weyl tensor is the shape part that survives the magnifying glass, while the Ricci and scalar parts — the trace, sourced by matter through Einstein's equations — get distorted. This is why source-free, purely tidal gravitational physics is naturally conformal while matter is not. Relatedly, massless particles carry no clock, no proper time to set a scale, so their physics cannot notice a rescaling and is conformally invariant; a massive particle carries the scale $1/m$ and therefore breaks conformal symmetry.

The core equations

The definition is a single line, and everything downstream is a consequence of it. A conformal transformation multiplies the metric by the square of a smooth, strictly positive function, with the inverse metric scaling oppositely so that indices still raise and lower consistently.

\tilde g_{ab}=\Omega^2\,g_{ab},\qquad \tilde g^{ab}=\Omega^{-2}\,g^{ab},\qquad \Omega(x)>0 \qquad (1)

where: $g_{ab}$ — physical (original) metric; $\tilde g_{ab}$ — conformally rescaled metric; $\Omega(x)$ — smooth, strictly positive conformal factor. Reading it: at each point the metric is stretched by $\Omega^2$ ; positivity is the whole story, because it is what protects every sign. The causal payoff follows immediately. The character of a vector is the sign of its norm, and that sign is inert under multiplication by a positive number.

\mathrm{sign}\big(\tilde g_{ab}V^aV^b\big)=\mathrm{sign}\big(\Omega^2\,g_{ab}V^aV^b\big)=\mathrm{sign}\big(g_{ab}V^aV^b\big) \qquad (2)

where: $V^a$ — an arbitrary tangent vector at a point. Reading it: timelike ( $<0$ ), null ( $=0$ ), spacelike ( $>0$ ) are defined purely by this sign, so the classification — and every causal set built from it — is identical for $g$ and $\tilde g$ . The Levi-Civita connection is not conformally invariant; it picks up a difference tensor built from the gradient of $\ln\Omega$ . This is the object that controls how geodesics transform.

\tilde\Gamma^a{}_{bc}=\Gamma^a{}_{bc}+\delta^a{}_b\,\partial_c\omega+\delta^a{}_c\,\partial_b\omega-g_{bc}\,g^{ad}\,\partial_d\omega,\qquad \omega\equiv\ln\Omega \qquad (3)

where: $\Gamma^a{}_{bc}$ — Christoffel symbols of $g_{ab}$ ; $\tilde\Gamma^a{}_{bc}$ — Christoffel symbols of $\tilde g_{ab}$ ; $\partial_c\omega$ — gradient of the log conformal factor. Reading it: the correction $C^a{}_{bc}=\tilde\Gamma^a{}_{bc}-\Gamma^a{}_{bc}$ is a genuine tensor, so the difference between the two geometries' notions of "straight" is coordinate-independent. Because the correction is contracted twice with the tangent along a geodesic, only null curves are protected: the null condition kills the term that would otherwise spoil the geodesic equation, leaving a curve that is still geodesic after reparametrization.

k^a\nabla_a k^b=0\ \Longrightarrow\ \tilde k^a\tilde\nabla_a\tilde k^b=0,\qquad \tilde k^a=\Omega^{-2}k^a \qquad (4)

where: $k^a$ — affinely parametrized null geodesic tangent for $g$ ; $\tilde k^a$ — rescaled tangent, affine for $\tilde g$ . Reading it: the light ray is identical as a point set; only its clock changes, via $\mathrm{d}\tilde\lambda=\Omega^2\,\mathrm{d}\lambda$ . Curvature splits into a conformally invariant part and an inhomogeneously transforming part. The mixed Weyl tensor is genuinely unchanged, while the Ricci tensor transforms with derivatives of $\Omega$ .

\tilde C^a{}_{bcd}=C^a{}_{bcd},\qquad \tilde R_{ab}=R_{ab}-(n-2)\nabla_a\nabla_b\omega-g_{ab}\Box\omega+(n-2)(\nabla_a\omega)(\nabla_b\omega)-(n-2)g_{ab}(\nabla\omega)^2 \qquad (5)

where: $C^a{}_{bcd}$ — Weyl (trace-free) conformal tensor; $R_{ab}$ — Ricci tensor; $\Box=g^{ab}\nabla_a\nabla_b$ — d'Alembertian of $g$ ; $n$ — spacetime dimension. Reading it: the $(1,3)$ Weyl tensor encodes the conformal class of the curvature; the trace part is where all the $\Omega$ -dependence lives. Finally, the massless field equations that inherit conformal invariance. The source-free Maxwell field is invariant in four dimensions, and the conformally coupled scalar is invariant in every dimension once the field is given the right conformal weight.

\tilde\nabla^a\tilde F_{ab}=\Omega^{-4}\nabla^aF_{ab}\ \ (n=4),\qquad \Big(\Box-\tfrac{n-2}{4(n-1)}R\Big)\phi=0\ \Longrightarrow\ \Big(\tilde\Box-\tfrac{n-2}{4(n-1)}\tilde R\Big)\tilde\phi=0,\ \ \tilde\phi=\Omega^{-\frac{n-2}{2}}\phi \qquad (6)

where: $F_{ab}$ — electromagnetic field strength; $\phi$ — conformally coupled scalar; $R$ — Ricci scalar; $\tilde\phi$ — conformally weighted field. Reading it: the field equations in the unphysical, compactified spacetime stay regular at $\mathscr{I}$ , which is exactly why radiative data can be read off on the boundary.

Derivation ANull geodesics survive a conformal rescaling but with a new affine parameter

The goal is to establish the precise mechanism that makes causal structure conformally invariant while metric structure is not: a curve that is null and geodesic for $g_{ab}$ remains null and geodesic for $\tilde g_{ab}=\Omega^2 g_{ab}$ once we rescale its tangent and change its affine parameter, whereas a timelike geodesic is generally converted into an accelerated, non-geodesic worldline. Take $k^a$ tangent to an affinely parametrized null geodesic of $g$ , with $g_{ab}k^ak^b=0$ and $k^a\nabla_a k^b=0$ , and track which terms the null condition removes.

Step 1. Nullity transfers for free: $\tilde g_{ab}k^ak^b=\Omega^2 g_{ab}k^ak^b=\Omega^2\cdot 0=0$ . Why: a null vector for $g$ is null for every metric in the conformal class, because $\Omega^2$ multiplies a zero — this is the same sign-preservation that protects the whole cone.

Step 2. Relate the two covariant derivatives through the difference tensor from equation (3): $\tilde\nabla_a k^b=\nabla_a k^b+C^b{}_{ac}k^c$ with $C^b{}_{ac}=\delta^b_a\partial_c\omega+\delta^b_c\partial_a\omega-g_{ac}g^{bd}\partial_d\omega$ . Why: the connection is where "geodesic" is defined, so any change in the geometry's straight lines is carried entirely by this tensor built from $\nabla\omega$ .

Step 3. Contract with $k^a$ along the curve: $k^a\tilde\nabla_a k^b=k^a\nabla_a k^b+(\delta^b_a\partial_c\omega+\delta^b_c\partial_a\omega-g_{ac}g^{bd}\partial_d\omega)k^ak^c$ . The first term vanishes because the curve is a geodesic of $g$ ; the third term in the bracket vanishes because $g_{ac}k^ak^c=0$ by nullity. Why: this is the crucial cancellation — the nullity condition is precisely what annihilates the one term that could not be absorbed into a reparametrization.

Step 4. Collect what survives: $k^a\tilde\nabla_a k^b=(k^c\partial_c\omega)k^b+(k^a\partial_a\omega)k^b=2(k^c\partial_c\omega)\,k^b$ . The right-hand side is proportional to $k^b$ , so the curve is still a geodesic of $\tilde g$ — merely a non-affinely parametrized one, with proportionality function $f=2\,k^c\partial_c\omega$ . Why: a geodesic that satisfies $k^a\nabla_a k^b=f k^b$ is the same path as an affine geodesic; only the parameter is stretched, so the light ray as a set of points is untouched.

Step 5. Restore affinity by rescaling the tangent: setting $\tilde k^a=\Omega^{-2}k^a=e^{-2\omega}k^a$ exactly cancels the inhomogeneous term, giving $\tilde k^a\tilde\nabla_a\tilde k^b=0$ with a new affine parameter obeying $\mathrm{d}\tilde\lambda=\Omega^2\,\mathrm{d}\lambda$ . Why: this is why light rays are the natural carriers of causal information across a conformal class — their paths are invariant, and only the clock ticking along them is rescaled, so they can be integrated all the way to a boundary at infinity. *Verified: matches Wald App. D (D.14–D.18) and Carroll App. G; for a timelike geodesic Step 3's third term does not vanish, confirming timelike geodesics are not preserved.*

Go deeper · Derivation B · 2 minConformal invariance of the Weyl tensor and the Ricci transformation law

Derivation BConformal invariance of the Weyl tensor and the Ricci transformation law

The goal is to show that Riemann curvature splits into a conformally invariant Weyl part and a Ricci-built remainder that carries the entire conformal weight — establishing that the conformal class of a metric is exactly the data stored in the Weyl tensor and that vanishing Weyl is the invariant criterion for conformal flatness. Write $\omega=\ln\Omega$ , decompose Riemann into Weyl plus a Schouten-tensor piece in $n$ dimensions, and feed the connection difference tensor through the curvature formula.

Step 1. Recall the decomposition $R^a{}_{bcd}=C^a{}_{bcd}+(\delta^a_c P_{bd}-\delta^a_d P_{bc}+g_{bd}P^a{}_c-g_{bc}P^a{}_d)$ with the Schouten tensor $P_{ab}=\frac{1}{n-2}\big(R_{ab}-\frac{R}{2(n-1)}g_{ab}\big)$ . Why: this separates curvature into a trace-free conformal piece and a trace piece before any transformation is applied, so we can track exactly where the $\Omega$ -dependence goes.

Step 2. Transform Riemann by inserting $C^a{}_{bc}$ into the curvature formula: $\tilde R^a{}_{bcd}=R^a{}_{bcd}+2\nabla_{[c}C^a{}_{d]b}+2C^a{}_{e[c}C^e{}_{d]b}$ . This generates corrections linear in $\nabla\nabla\omega$ and quadratic in $\nabla\omega$ . Why: every $\Omega$ -dependent term in the curvature is produced here, so if they all reassemble into a single tensor's transformation we will have isolated the invariant part.

Step 3. Show the corrections are precisely the transformation of the Schouten tensor: $\tilde P_{ab}=P_{ab}-\nabla_a\nabla_b\omega+(\nabla_a\omega)(\nabla_b\omega)-\tfrac12 g_{ab}(\nabla\omega)^2$ . Every inhomogeneous term needed to build $\tilde R^a{}_{bcd}$ is exactly the term needed to build $\tilde P_{ab}$ from $P_{ab}$ . Why: the Schouten tensor is the exact repository of the conformal weight of curvature — contracting it gives the Ricci transformation law of equation (5).

Step 4. Subtract the transformed Schouten part from $\tilde R^a{}_{bcd}$ and find no $\Omega$ -dependence left: $\tilde C^a{}_{bcd}=C^a{}_{bcd}$ . The mixed $(1,3)$ Weyl tensor is conformally invariant, while the fully covariant form scales as $\tilde C_{abcd}=\Omega^2 C_{abcd}$ . Why: index placement matters — only the vector-to-vector map $C^a{}_{bcd}$ is strictly unchanged, and this is the object that encodes the conformal class.

Step 5. Read off the flatness criterion: $C^a{}_{bcd}=0\Longleftrightarrow g_{ab}=\Omega^2\eta_{ab}$ for $n\ge 4$ . Vanishing Weyl is necessary and sufficient for conformal flatness; in $n=3$ the Weyl tensor vanishes identically and the Cotton tensor takes over this role. Why: this is the operational test used everywhere — for instance every FLRW cosmology has $C^a{}_{bcd}=0$ and is therefore conformally flat. *Verified: consistent with Wald App. D, Carroll App. G, and MTW Box 22.2; the Ricci law here is the trace of the Schouten transformation in Step 3.*

Go deeper · Derivation C · 2 minConformal compactification of Minkowski spacetime

Derivation CConformal compactification of Minkowski spacetime

The goal is a fully explicit construction that turns the abstract "choose $\Omega\to0$ at infinity" into the concrete Penrose diagram: starting from four-dimensional Minkowski space we pick a divergent coordinate transformation and a conformal factor carrying the entire infinite spacetime into a finite triangular region of the Einstein static universe, then read off the boundary pieces $\mathscr{I}^\pm$ , $i^0$ , and $i^\pm$ . Begin with Minkowski in spherical coordinates, $\mathrm{d}s^2=-\mathrm{d}t^2+\mathrm{d}r^2+r^2\mathrm{d}\Omega^2$ , and pass to null coordinates $u=t-r$ , $v=t+r$ .

Step 1. Rewrite the line element in null coordinates: with $r=(v-u)/2\ge 0$ so $v\ge u$ , one finds $\mathrm{d}s^2=-\mathrm{d}u\,\mathrm{d}v+\tfrac14(v-u)^2\,\mathrm{d}\Omega^2$ , valid over the whole range $u,v\in(-\infty,\infty)$ with $v\ge u$ . Why: null coordinates are adapted to the light-cones, which are the objects the compactification must draw faithfully.

Step 2. Compactify each null coordinate with an arctangent: set $u=\tan q$ , $v=\tan p$ with $p,q\in(-\tfrac\pi2,\tfrac\pi2)$ and $p\ge q$ . Infinity is mapped to the finite endpoints $\pm\tfrac\pi2$ . Why: arctan is the simplest monotone map from the whole real line to a finite interval, so the infinite null coordinates now live on a bounded square.

Step 3. Extract the conformal factor. Using $\mathrm{d}u=\sec^2q\,\mathrm{d}q$ , $\mathrm{d}v=\sec^2p\,\mathrm{d}p$ and $v-u=\sin(p-q)/(\cos p\cos q)$ , the metric becomes $\mathrm{d}s^2=\frac{1}{4\cos^2 p\,\cos^2 q}\big[-4\,\mathrm{d}p\,\mathrm{d}q+\sin^2(p-q)\,\mathrm{d}\Omega^2\big]$ . Why: the prefactor is exactly $\Omega^{-2}$ with $\Omega=2\cos p\cos q$ , and the bracket is the finite unphysical metric $\mathrm{d}\tilde s^2=\Omega^2\mathrm{d}s^2$ — the rescaling that tames infinity.

Step 4. Recognize the target geometry. Setting $T=p+q$ , $\chi=p-q$ turns the bracket into $\mathrm{d}\tilde s^2=-\mathrm{d}T^2+\mathrm{d}\chi^2+\sin^2\chi\,\mathrm{d}\Omega^2$ , part of the Einstein static universe $\mathbb{R}\times S^3$ ; Minkowski occupies a finite triangle within it. Why: the unphysical spacetime is a smooth, compact-in-space geometry, so the "edge" of Minkowski is now a genuine boundary one can attach and study.

Step 5. Read off the boundary structure at $\Omega=0$ : future null infinity $\mathscr{I}^+$ ( $p=\tfrac\pi2$ ) and past null infinity $\mathscr{I}^-$ ( $q=-\tfrac\pi2$ ) are null hypersurfaces — the slanted 45-degree edges; spatial infinity $i^0$ ( $p=\tfrac\pi2,q=-\tfrac\pi2$ ) is a single point; future and past timelike infinity $i^\pm$ ( $p=q=\pm\tfrac\pi2$ ) are single points. Why: outgoing light rays end on $\mathscr{I}^+$ , massive particles end at $i^+$ , and spacelike slices end at $i^0$ — the causal fate of every kind of curve is now visible as a boundary point or surface. *Verified: reproduces the canonical Minkowski Penrose diagram in Hawking & Ellis §5.1, Wald §11.1, and Carroll §5.7; the $-4\,\mathrm{d}p\,\mathrm{d}q=-\mathrm{d}T^2+\mathrm{d}\chi^2$ identity checks with $T=p+q,\ \chi=p-q$ .*

Limiting cases

Conformal transformations interpolate between several familiar special cases according to the choice of $\Omega$ .

Case	Result	Meaning
Constant rescaling	$\Omega=\mathrm{const}$	A uniform change of units: $\Gamma^a{}_{bc}$ , Riemann and all geodesics are unchanged, but every length and time scales by $\Omega$ . The trivial limit that isolates what a genuine conformal factor adds.
Conformally flat spacetime	$g_{ab}=\Omega^2\eta_{ab}\ \Leftrightarrow\ C^a{}_{bcd}=0$	Metric conformal to Minkowski; every FLRW cosmology is of this type. Detected exactly by the vanishing of the Weyl tensor ( $n\ge 4$ ).
Conformal Killing field	$\mathcal{L}_\xi g_{ab}=2\phi\,g_{ab}$	An infinitesimal conformal symmetry; $\xi^a$ is a conformal Killing vector. When $\phi=0$ it reduces to an ordinary Killing field, and null geodesics then conserve $k_a\xi^a$ .
Compactifying factor	$\Omega>0$ inside, $\Omega=0,\ \mathrm{d}\Omega\neq0$ on $\mathscr{I}$	The defining choice for conformal completion: the locus where $\Omega$ vanishes with nonzero gradient is conformal infinity, drawn as the edge of a Penrose diagram.

Common trapA conformal transformation is not a diffeomorphism and not a symmetry

The single most common error is to treat a conformal rescaling like a coordinate change or a symmetry of Einstein's equations. It is neither. It changes proper lengths and times; it maps timelike and spacelike geodesics to non-geodesics; it does not preserve the Ricci tensor, the Ricci scalar, or curvature invariants like $R_{abcd}R^{abcd}$ . Only the light-cone structure, the null geodesics (as paths), and the mixed Weyl tensor $C^a{}_{bcd}$ come through unchanged. Two further traps: do not confuse a conformal isometry — a vector field generating a one-parameter family of conformal maps, a conformal Killing field — with a mere global rescaling; and remember that index placement is decisive for the Weyl tensor, since the fully covariant $C_{abcd}$ scales as $\Omega^2 C_{abcd}$ while only the $(1,3)$ form is strictly invariant.

Worked examples

1 · Causal classification is unchanged by a rescaling (confidence builder)

In Minkowski coordinates with signature $(-,+,+,+)$ , test three vectors at a point against both $g$ and the rescaled metric $\tilde g=\Omega^2 g$ with $\Omega=2$ . Take $V=(1,\,0.6,\,0,\,0)$ , $k=(1,\,1,\,0,\,0)$ , and $s=(0.5,\,1,\,0,\,0)$ .

Step 1. Compute the physical norms from $g_{ab}V^aV^b=-(V^0)^2+(V^1)^2+\cdots$ . For $V$ : $-(1)^2+(0.6)^2=-1+0.36=\mathbf{-0.64}$ , so $V$ is timelike. For $k$ : $-(1)^2+(1)^2=\mathbf{0}$ , so $k$ is null. For $s$ : $-(0.5)^2+(1)^2=-0.25+1=\mathbf{+0.75}$ , so $s$ is spacelike.

Step 2. Apply the rescaling. Since $\tilde g_{ab}=\Omega^2 g_{ab}=4\,g_{ab}$ , each norm is multiplied by $4$ : $\tilde g(V,V)=4(-0.64)=\mathbf{-2.56}$ , $\tilde g(k,k)=4(0)=\mathbf{0}$ , $\tilde g(s,s)=4(+0.75)=\mathbf{+3.0}$ .

Step 3. Compare signs. The magnitudes changed — proper lengths and times are genuinely rescaled — but the signs are untouched: timelike stays timelike, null stays null, spacelike stays spacelike. The light-cone at the point is exactly the same set of null directions, because $\tilde g(k,k)=0$ iff $g(k,k)=0$ . Every causal relation survives; only metric magnitudes scale, by exactly $\Omega^2=4$ .

2 · Locating events on the Minkowski Penrose diagram

Using the compactification of the third derivation ( $u=\tan q$ , $v=\tan p$ , $\Omega=2\cos p\cos q$ ), find the diagram coordinates $(T,\chi)$ and the conformal factor for two events on the slice $t=0$ : event A at $r=1$ and event B at $r=1000$ (geometric units).

Step 1. Event A. With $t=0,r=1$ : $u=t-r=-1$ , $v=t+r=+1$ . Then $q=\arctan(-1)=-\tfrac\pi4=\mathbf{-0.7854}$ and $p=\arctan(1)=+\tfrac\pi4=\mathbf{+0.7854}$ . So $T=p+q=\mathbf{0}$ and $\chi=p-q=\tfrac\pi2\approx\mathbf{1.5708}$ — the event sits on the equator of the diagram, halfway out.

Step 2. Conformal factor at A: $\Omega=2\cos p\cos q=2\cos(\tfrac\pi4)\cos(-\tfrac\pi4)=2\cdot\tfrac{\sqrt2}{2}\cdot\tfrac{\sqrt2}{2}=2\cdot\tfrac12=\mathbf{1}$ . The rubber ruler is essentially undistorted this close in — the finite interior is a mild rescaling of the physical metric.

Step 3. Event B. With $t=0,r=1000$ : $u=-1000$ , $v=+1000$ , so $p=\arctan(1000)=\mathbf{1.5698}$ and $q=-1.5698$ , giving $T=0$ and $\chi=p-q=\mathbf{3.1396}\approx\pi$ . The event has been pulled almost all the way to the right-hand vertex.

Step 4. Conformal factor at B: $\Omega=2\cos^2(1.5698)=2(0.0010)^2\approx\mathbf{2\times10^{-6}}$ , vanishingly small. As $r\to\infty$ the whole spacelike slice $t=0$ limits onto the single point $\chi=\pi$ with $\Omega\to0$ — this is spatial infinity $i^0$ . Finite $(T,\chi)$ and $\Omega\to0$ at large $r$ are precisely how an infinite spacetime is drawn on a finite page.

Physics of the portrait

Conformal infinity is one person's idea before it is anyone else's. The notion that a spacetime's infinity can be studied as a finite boundary by rescaling the metric, and that light-cones drawn there record causal relations exactly, is due to Roger Penrose in the early 1960s — the conceptual leap that turned "infinity" from a limit one takes into a place one draws.

Origins · 1 minRoger Penrose and how this was discovered

Roger Penrose

born 1931 · conformal treatment of infinity and the geometry of causal structure

Roger Penrose, an English mathematical physicist trained in algebraic geometry, reshaped general relativity in the 1960s by insisting on the global, coordinate-free geometry of spacetime — and conformal methods were his signature tool. In 1963–1965 he introduced the idea of conformally rescaling the metric so that infinity is brought to a finite boundary, showing that because the rescaling preserves null cones, the causal structure of the physical spacetime is faithfully drawn on the compactified picture. This gave the world the Penrose diagram and the rigorous notions of null infinity $\mathscr{I}$ (he coined the script letter and the pronunciation "scri"), spatial infinity $i^0$ , and timelike infinity $i^\pm$ . He used the same philosophy to define asymptotic flatness, radiation fields, and the peeling theorem, in which massless fields fall off in a hierarchy of powers governed by their conformal weights near $\mathscr{I}^+$ . In 1965 he combined causal-structure arguments with global topology to prove the first singularity theorem, for which he shared the 2020 Nobel Prize in Physics. The through-line of his work — from twistor theory to the singularity theorems to conformal cyclic cosmology — is the conviction that the conformal class of the metric, not the metric itself, carries the deepest physical information: the light-cones, not the rulers.

History

1921

— Weyl isolates the conformally invariant part of curvature, the Weyl tensor

1963

— Penrose introduces conformal rescaling to bring infinity to a finite boundary

1964

— Penrose defines null infinity 𝓘 and the conformal treatment of asymptotic fields

1965

— Penrose proves the first singularity theorem using global causal structure

1965

— Carter and Penrose systematize conformal diagrams for stationary black holes

Going further

Texts. Wald, General Relativity (1984), App. D (conformal transformations) and §11.1 (asymptotic flatness, conformal completion) · Misner, Thorne & Wheeler, Gravitation (1973), §22.4 and Box 22.2, and Ch. 34 (conformal infinity) · Carroll, Spacetime and Geometry (2004), App. G and §5.7 (Penrose diagrams) · Hawking & Ellis, The Large Scale Structure of Space-Time (1973), §5.1 and §6.8–6.9 · Penrose, Relativity, Groups and Topology, ed. DeWitt & DeWitt (1964) · Schutz, A First Course in General Relativity (2009), Ch. 11.

Provenance. Every transformation law here (connection, Riemann, Ricci, Weyl, the scalar and Maxwell equations) is standard and recomputable from the definition $\tilde g_{ab}=\Omega^2 g_{ab}$ : check the connection difference tensor and Weyl invariance against Wald App. D (D.1–D.18) and Carroll App. G, the null-geodesic preservation against Wald (D.14ff), and the Minkowski compactification with its identification of $\mathscr{I}^\pm$ , $i^0$ , $i^\pm$ against Hawking & Ellis §5.1, Wald §11.1, and Carroll §5.7. The worked-example arithmetic ( $-0.64,0,+0.75$ scaling by $4$ ; $\arctan 1=\pi/4$ , $\arctan 1000=1.5698$ , $\Omega=2\cos p\cos q$ ) is elementary and self-checking. No empirical numbers are quoted; all content is mathematical and textbook-verifiable.

Builds on: Causal Structure of Spacetime · Null Geodesics and Light Cones · Chronological and Causal Futures and Pasts Leads to: Penrose-Carter Conformal Diagrams · Conformal Infinity and Asymptotic Structure · Null Infinity (Scri) · Asymptotic Flatness

Ask the tutor: Want to see how this same conformal machinery builds the Penrose diagram of a Schwarzschild black hole, horizon and singularity and all?

Is a conformal transformation the same as a change of coordinates?

No, and conflating the two is the single most common mistake. A diffeomorphism (coordinate change) is a passive relabeling of the same points; it leaves the geometry — all lengths, times, and curvatures — physically unchanged, merely expressed in new labels. A conformal transformation $\tilde g_{ab}=\Omega^2 g_{ab}$ actively changes the metric at each point, so proper lengths and durations really are different, curvature invariants change, and geodesics are generally rerouted. The two agree only in what they preserve about causality: both keep the light-cones, so both are legitimate tools for drawing causal diagrams. But a conformal transformation produces a genuinely different geometry, one in the same conformal class as the original.

Why is only the mixed Weyl tensor invariant, not the fully covariant one?

Because lowering an index is done with the metric, and the metric itself carries conformal weight. The strictly invariant statement is $\tilde C^a{}_{bcd}=C^a{}_{bcd}$ for the $(1,3)$ tensor — the one that acts as a linear map taking vectors to vectors, an operation that does not reference an overall scale. The moment you lower the upper index with $\tilde g_{ae}=\Omega^2 g_{ae}$ , you pick up a factor: $\tilde C_{abcd}=\Omega^2 C_{abcd}$ . So both statements are true and consistent; you just have to be explicit about index placement. The physical content — that the conformal class of curvature is encoded by Weyl — is carried by the invariant mixed form.

If null geodesics are preserved, why aren't timelike ones?

The proof hinges on a cancellation that only the null condition provides. When you contract the connection-difference tensor twice with the tangent, one of the resulting terms is proportional to $g_{ac}k^ak^c$ , the norm of the tangent. For a null geodesic that norm is zero, so the term vanishes and what remains is proportional to the tangent itself — the hallmark of a geodesic (after reparametrization). For a timelike geodesic that norm is nonzero, the term stays, and it is not proportional to the tangent, so the curve acquires a genuine four-acceleration for $\tilde g$ : it is no longer a geodesic. Physically, a massive particle carries a scale (its mass sets a proper-time clock) that a rescaling disturbs, while a light ray carries no such scale.

What exactly is 𝓘 (scri) and how is it different from a coordinate infinity?

Null infinity $\mathscr{I}$ (read "scri") is the boundary an outgoing or incoming light ray reaches in the unphysical, conformally rescaled spacetime — the locus $\Omega=0$ with $\mathrm{d}\Omega\neq0$ . In the physical spacetime these are points infinitely far away that no finite construction reaches; the conformal rescaling with $\Omega\to0$ attaches them as a genuine, differentiable hypersurface. That is the crucial difference from a mere coordinate infinity: $\mathscr{I}$ is not an artefact of bad coordinates but a real geometric boundary of the completed manifold, with its own induced structure. It is where massless radiation "arrives," where Bondi mass and gravitational-wave data are defined, and where the BMS symmetry group acts.

Are Einstein's equations conformally invariant?

No — and this is important. General relativity is not a conformally invariant theory: the Einstein equations $G_{ab}=8\pi T_{ab}$ single out one metric in the conformal class through the Ricci and scalar curvatures, which transform inhomogeneously under rescaling. Conformal invariance is a property of causal structure and of certain massless matter equations (source-free Maxwell in 4D, the conformally coupled scalar), not of the gravitational field equations themselves. This is exactly why conformal methods are used as a tool — to draw causal diagrams and to attach a boundary at infinity — rather than as a symmetry of the dynamics. The physical metric and the unphysical (rescaled) metric are related by a chosen factor $\Omega$ , and only the physical one satisfies Einstein's equations.

Why must the conformal factor vanish at infinity, and what is the regularity condition?

To bring points at infinite distance to a finite location you must shrink the ruler without bound out there, which means $\Omega\to0$ as you approach infinity. But you cannot let $\Omega$ vanish carelessly: the boundary is differentiable only if $\Omega=0$ together with $\mathrm{d}\Omega\neq0$ there, so that $\Omega$ is a good coordinate cutting transversally across $\mathscr{I}$ . That gradient condition guarantees the unphysical metric $\tilde g_{ab}=\Omega^2 g_{ab}$ extends smoothly to the boundary and that $\mathscr{I}$ is a regular hypersurface rather than a singular pinch. In the Minkowski example, $\Omega=2\cos p\cos q$ vanishes to first order at $p=\pi/2$ or $q=-\pi/2$ , exactly meeting the condition.