The Stress-Energy Tensor

What the stress-energy tensor is

In special relativity, energy and momentum merge into a single four-vector, the four-momentum p^μ = (E/c, p_x, p_y, p_z). For a continuous distribution of matter and fields, one four-vector per particle is not enough — we need to know how much four-momentum sits in each little volume and how it flows across each little surface. The object that answers both questions at once is the stress-energy tensor (also called the energy-momentum tensor), written T^μν.

The definition is compact: T^μν is the flux of the μ-component of four-momentum across a surface of constant x^ν. Written as a 4×4 matrix with the index order (t, x, y, z), it has a natural block structure:

        ┌ T⁰⁰   T⁰ˣ   T⁰ʸ   T⁰ᶻ ┐   ┌ energy      energy-flux / c        ┐
Tᵘᵛ  =  │ Tˣ⁰   Tˣˣ   Tˣʸ   Tˣᶻ │ = │ density   (= c · momentum density) │
        │ Tʸ⁰   Tʸˣ   Tʸʸ   Tʸᶻ │   │  ————————   ———————————————————————│
        └ Tᶻ⁰   Tᶻˣ   Tᶻʸ   Tᶻᶻ ┘   └ momentum      momentum flux         ┘
                                        density     (pressure + shear)

Reading the blocks:

• T⁰⁰ — energy density (J/m³). For ordinary matter this is dominated by rest-mass energy, ρc².
• T⁰ⁱ = Tⁱ⁰ — momentum density × c, equivalently energy flux / c. The relativistic identity T⁰ⁱ = Tⁱ⁰ is E = mc² in tensor clothing: energy flux and momentum density are the same physical thing.
• T^ij — momentum flux, the 3×3 Cauchy stress tensor. Its diagonal entries T^xx, T^yy, T^zz are pressures (normal stresses); its off-diagonal entries are shear stresses. For an isotropic fluid, T^ij = p δ^ij.

Two structural facts matter. First, T^μν is symmetric: T^μν = T^νμ. Symmetry of the spatial block enforces conservation of angular momentum; symmetry of the mixed block is the statement that energy flux equals c² × momentum density. Second, every component carries the same units — energy per unit volume, J/m³, which is identical to pressure, Pa. That single unit is the tell that pressure and energy density belong in the same object.

Why energy — not just mass — is the source of gravity

Newton's theory has a scalar source: Poisson's equation ∇²Φ = 4πGρ ties the gravitational potential Φ to mass density ρ. But relativity forbids a scalar source, for a simple reason. Mass and energy are equivalent (E = mc²), and energy is not the same in every reference frame — a moving object carries more energy, and a box of gas carries energy in its pressure and heat. If the source were a single number, different observers would disagree about how strong gravity is at the same event, which is impossible.

The fix is to promote the source to a tensor that transforms correctly under Lorentz boosts. The stress-energy tensor is precisely that object: boosting mixes its components — energy density picks up contributions from momentum flux, pressure feeds into energy density — in exactly the way E = mc² requires. Einstein's insight, completed in November 1915, was that the geometry side of the equation must be a symmetric, divergence-free, rank-2 tensor built from curvature. The unique such object (up to a cosmological term) is the Einstein tensor G^μν = R^μν − ½ R g^μν. Matching it to matter gives:

G^μν  +  Λ g^μν  =  (8πG / c⁴) T^μν

Here G^μν is the Einstein tensor (built from the Ricci tensor R^μν and Ricci scalar R), Λ is the cosmological constant, g^μν is the inverse metric, G = 6.674 × 10⁻¹¹ m³·kg⁻¹·s⁻² is Newton's constant, and c = 2.998 × 10⁸ m/s. The coupling 8πG/c⁴ ≈ 2.08 × 10⁻⁴³ s²·kg⁻¹·m⁻¹ is minuscule, which is why you need something as extreme as a neutron star or the entire mass of the Sun to bend light by a measurable arcsecond.

How it works, component by component

1. Pick an observer. The split of T^μν into energy, momentum, and stress is observer-dependent. For an observer with four-velocity u^μ (normalized u^μu_μ = −c²), the energy density they measure is ε = T_μν u^μ u^ν / c² (equal to ρc² for ordinary matter), and the momentum density is a projection of T^μν u_ν onto their spatial slice.
2. Read the diagonal. In the matter's own rest frame the tensor is diagonal for a fluid: T^μν = diag(ρc², p, p, p). The corner is energy density; the three remaining diagonals are the pressure pushing outward in each direction.
3. Add the off-diagonals for real materials. A solid under shear, a viscous fluid, or an electromagnetic field fills in the T^ij off-diagonals (shear/Maxwell stresses) and the T⁰ⁱ entries (Poynting flux for light).
4. Feed it into Einstein's equations. Every non-zero component sources curvature. Pressure appears with the same weight as energy density in the trace, so it genuinely bends spacetime.
5. Check conservation. Take the covariant divergence: ∇_μT^μν = 0 must hold automatically, because the geometry side satisfies the contracted Bianchi identity ∇_μG^μν = 0.

The perfect-fluid form and its special cases

The workhorse of relativistic astrophysics and cosmology is the perfect fluid: no viscosity, no heat conduction, isotropic pressure. Its stress-energy tensor is

T^μν = (ρ + p/c²) u^μ u^ν + p g^μν     (signature −+++)

where ρ is the mass-energy density, p is the isotropic pressure, u^μ is the fluid four-velocity (normalized u^μu_μ = −c²), and g^μν is the inverse metric. In the fluid rest frame u^μ = (c, 0, 0, 0) and this collapses to diag(ρc², p, p, p). A single formula, tuned by the equation of state p = p(ρ), then describes wildly different contents of the universe:

Component	Equation of state	What it is	Cosmological role
Dust (pressureless matter)	p = 0	Galaxies, cold dark matter, slow-moving mass	ρ ∝ a⁻³ (dilutes with volume)
Radiation / ultrarelativistic gas	p = ρc²/3	Photons, neutrinos, hot early universe	ρ ∝ a⁻⁴ (dilutes + redshifts)
Stiff fluid	p = ρc²	Maximally rigid matter (sound speed = c)	ρ ∝ a⁻⁶
Cosmological constant / vacuum	p = −ρc²	Dark energy, Λ	ρ = constant (drives acceleration)
Curvature-like fluid	p = −ρc²/3	Effective term for spatial curvature	ρ ∝ a⁻²

The parameter w = p/(ρc²) is the equation-of-state parameter: w = 0 (dust), w = 1/3 (radiation), w = −1 (dark energy). The measured value from Planck and supernova surveys is w ≈ −1.03 ± 0.03, consistent with a pure cosmological constant.

Worked point: pressure gravitates

Take the Newtonian limit of Einstein's equations for a static, weak field. The source of the potential is not ρ alone but the combination that appears in the Tolman–Whittaker (active gravitational) mass density:

ρ_active = ρ + 3p/c²      →      ∇²Φ = 4πG (ρ + 3p/c²)

The factor of 3 comes from the three spatial pressure terms in the trace of T^μν. Consequences:

• Positive pressure adds to gravity. Squeezing matter harder makes it gravitate more, not less. This is why you cannot indefinitely hold up a collapsing star by raising its internal pressure — beyond a point (the Tolman–Oppenheimer–Volkoff limit, ≈ 2.2–2.9 M_⊙ for neutron stars) the extra pressure only accelerates collapse into a black hole.
• Radiation gravitates more per joule than dust. For radiation ρ + 3p/c² = ρ + 3(ρ/3) = 2ρ, twice the naive value.
• Large negative pressure repels. For dark energy p = −ρc², so ρ + 3p/c² = ρ − 3ρ = −2ρ < 0. The active mass is negative, gravity effectively pushes, and the expansion of the universe accelerates — the phenomenon discovered in 1998 by the Supernova Cosmology Project and the High-z Supernova Search Team (Nobel Prize 2011).

Conservation: ∇_μT^μν = 0

The single equation ∇_μT^μν = 0 is the covariant law of energy-momentum conservation. Splitting by the free index ν:

• ν = 0 → conservation of energy (a continuity equation for T⁰⁰ plus its flux).
• ν = i → conservation of momentum (this is the relativistic Euler equation for a fluid).

In flat spacetime the covariant derivative becomes an ordinary one and the law reads ∂_μT^μν = 0, exactly the four-dimensional continuity/Euler equations. In curved spacetime the Christoffel terms Γ hidden inside ∇ represent energy and momentum exchanged with the gravitational field itself — which is why gravitational potential energy is not part of T^μν and why energy is only locally conserved, not globally, in general relativity.

Remarkably, you do not have to impose ∇_μT^μν = 0 as a separate axiom. The Einstein tensor obeys the contracted Bianchi identity ∇_μG^μν ≡ 0 as a geometric fact. Since G^μν = (8πG/c⁴)T^μν, conservation of the matter source is forced by the geometry. Geometry tells matter how to move, and it does so consistently.

Representative energy densities

Because every component of T^μν is an energy density (J/m³), it is instructive to compare the rest-energy density ρc² of common systems with their ordinary pressures — the huge gap is why rest mass, not pressure, dominates gravity in everyday life.

System	Rest-energy density ρc² (J/m³)	Typical pressure p (Pa)
Air at sea level	≈ 1.1 × 10¹⁷	1.0 × 10⁵
Liquid water	≈ 9.0 × 10¹⁹	1.0 × 10⁵
Sun's core	≈ 1.4 × 10²²	≈ 2.5 × 10¹⁶
Neutron-star interior	≈ 4 × 10³⁴	≈ 10³⁴ (pressure ~ ρc²)
CMB radiation (today)	≈ 4.2 × 10⁻¹⁴	≈ 1.4 × 10⁻¹⁴ (= ρc²/3)
Cosmological dark energy	≈ 5.3 × 10⁻¹⁰	≈ −5.3 × 10⁻¹⁰ (= −ρc²)

Notice the last two rows: for radiation p is exactly one-third of the energy density, and for dark energy it is minus the energy density. In dense stellar matter pressure climbs to the same order as ρc², and that is exactly the regime where its gravitational contribution becomes decisive.

The electromagnetic stress-energy tensor

Fields carry energy-momentum too. The electromagnetic field has

T^μν_EM = (1/μ₀) [ F^μα F^ν_α − ¼ g^μν F_αβ F^αβ ]

where F^μν is the Faraday tensor and μ₀ = 1.257 × 10⁻⁶ T·m/A is the vacuum permeability. Its T⁰⁰ is the familiar field energy density ½(ε₀E² + B²/μ₀); its T⁰ⁱ is the Poynting vector S = E × B/μ₀ divided by c; and its T^ij is the Maxwell stress tensor. This tensor is traceless (T^μ_μ = 0), which is precisely why pure radiation has the equation of state p = ρc²/3: tracelessness forces 3p = ρc².

Common misconceptions

• "The stress-energy tensor is just mass density." No — that is only T⁰⁰ in the low-velocity limit. The whole point is that momentum density, energy flux, pressure, and shear all sit in the same object and all gravitate.
• "Pressure doesn't affect gravity." It does. The active gravitational mass density is ρ + 3p/c². Everyday pressures are negligible only because they are ~10¹⁵ times smaller than ρc², not because pressure is exempt.
• "∇_μT^μν = 0 means energy is globally conserved." Only locally. In a general curved or expanding spacetime there is no conserved total energy of matter, because energy is exchanged with the gravitational field — photon redshift in an expanding universe is the standard example.
• "T^μν includes the energy of gravity itself." It does not. The gravitational field has no local energy density in general relativity (the equivalence principle lets you transform gravity away at a point). Any "gravitational energy" is captured by pseudotensors, not T^μν.
• "The tensor could be asymmetric." The metric (Hilbert) definition T^μν = (−2/√−g) δ(√−g L_matter)/δg_μν is automatically symmetric; symmetry is required for angular-momentum conservation.
• "Vacuum has zero stress-energy." The quantum vacuum has T^μν = −ρ_vac g^μν with p = −ρc² — this is the cosmological constant, and getting its value to match observation (rather than the naive field-theory estimate 10¹²⁰ times too large) is the cosmological constant problem.

A short history

The notion of a stress tensor for the electromagnetic field goes back to Maxwell (1873) and its four-dimensional form to Minkowski (1908) and Max Abraham. Einstein, guided by Marcel Grossmann's tensor calculus, arrived at the field equations G^μν = (8πG/c⁴)T^μν in November 1915, choosing the geometry side precisely so that the Bianchi identity would guarantee ∇_μT^μν = 0. David Hilbert derived the same equations days later from a variational principle, and it was this action approach that gave the clean, manifestly symmetric definition of T^μν as the response of the matter action to a change in the metric. Richard Tolman worked out the pressure contribution to active gravitational mass in the 1930s, and the perfect-fluid tensor became the foundation of Friedmann–Lemaître cosmology and of the Tolman–Oppenheimer–Volkoff equation for relativistic stars in 1939.