Pair Production from Gamma Rays

Light colliding with light, turning into matter

Einstein's E = mc² is usually read as a recipe for getting energy out of matter — fission, fusion, the Sun. Run it backwards and it is just as valid: pour enough energy into a small enough volume and matter appears. The cleanest version of this is pair production, where electromagnetic energy converts into an electron and its antiparticle, a positron. The most striking astrophysical channel needs no matter at all to start with. Two photons — pure light — collide and vanish, and where they were there are now two massive particles flying apart. This is photon-photon, or γγ, pair production, the Breit-Wheeler process named for the 1934 paper of Gregory Breit and John Wheeler.

It sounds like a laboratory curiosity, and on Earth it nearly is: getting two beams of real photons dense enough to collide measurably is brutally hard. But the universe is enormous and full of light. A gamma ray with the energy of a TeV — a trillion electron-volts, a million times more energetic than an X-ray — travels for hundreds of millions of light-years through a faint sea of starlight that fills all of intergalactic space. Given enough path length, it will eventually meet a soft optical or infrared photon head-on with just the right energy, and the two will turn into an electron-positron pair. The gamma ray is gone. This single process, repeated across cosmic distances, makes the high-energy universe partly opaque — it draws a gamma-ray horizon beyond which the most energetic photons simply cannot reach us.

How it works: two channels, one principle

Conservation of energy and momentum forbids a lone photon from spontaneously becoming an e⁺e⁻ pair in empty space. The pair would have to carry the photon's momentum, but a pair at rest (the minimum-energy configuration) has zero momentum, and any moving configuration carries more energy than a single on-shell photon of that momentum can supply. Something must absorb the recoil. There are two ways the universe arranges this.

Channel 1 — nuclear-field (one-photon) pair production. A photon passing close to an atomic nucleus can transfer recoil momentum to the heavy nucleus, which barely moves. This unlocks the conversion once the photon clears the rest-mass cost of the pair:

E_γ > 2 m_e c² = 1.022 MeV

This is the dominant interaction of photons with matter above about 10 MeV, and it is the engine of electromagnetic showers: a high-energy photon pair-produces, the electron and positron radiate bremsstrahlung photons, those pair-produce again, and an exponential cascade of particles develops — the principle behind every gamma-ray detector and the air showers used by ground-based telescopes.

Channel 2 — photon-photon (γγ) pair production. The recoil-absorbing body is a second real photon. Two photons of energies E₁ and E₂ meeting at an angle θ form a system whose invariant center-of-momentum energy must reach the pair rest mass:

s = 2 E₁ E₂ (1 − cos θ) ≥ (2 m_e c²)²
  ⟹  E₁ E₂ (1 − cos θ) ≥ 2 (m_e c²)²

For a head-on collision, cos θ = −1 and the (1 − cos θ) factor is 2, giving the cleanest form of the threshold:

E₁ E₂ ≥ (m_e c²)² ≈ (0.511 MeV)² ≈ 0.261 MeV²

This is the costed claim at the heart of the topic. The two energies enter as a product, so a very high-energy photon needs only a very low-energy partner, and vice versa. That trade-off is what couples gamma rays to the diffuse background light, and it is what the visualization above is built around: two photons sweep in, their combined energy clears the threshold, and where they meet an electron and a positron spring outward.

Worked example: which photon does a TeV gamma ray pair with?

Take a gamma ray of E₁ = 1 TeV = 10⁶ MeV, the canonical energy for ground-based Cherenkov telescopes like H.E.S.S., MAGIC, and VERITAS. For a head-on collision the most efficient target photon — the one that sits at the threshold — has energy

E₂ ≈ (m_e c²)² / E₁
   = 0.261 MeV² / 10⁶ MeV
   = 2.6 × 10⁻⁷ MeV
   = 0.26 eV

A 0.26 eV photon has a wavelength of about 4.8 microns — the near-to-mid infrared. The cross section actually peaks a few times above threshold, near a target of ~1 eV (wavelength ~1 micron), which is precisely the near-infrared starlight peak of the extragalactic background light. So a TeV gamma ray is tuned, almost perfectly, to be absorbed by the accumulated light of all the stars that ever shone. Scale the energies and the target wavelength scales inversely:

Gamma-ray energy	Optimal target energy E₂ ≈ (m_ec²)²/E₁	Target wavelength	EBL component
10 GeV	26 eV	48 nm (extreme UV)	essentially transparent
100 GeV	2.6 eV	480 nm (optical)	starlight, blue
1 TeV	0.26 eV	4.8 μm (near-IR)	starlight peak
10 TeV	26 meV	48 μm (far-IR)	dust emission
100 TeV	2.6 meV	480 μm (sub-mm)	far-IR / CMB tail
1 PeV	0.26 meV	4.8 mm (microwave)	cosmic microwave background

At PeV energies the optimal target is a CMB photon — and the CMB has an enormous photon density (about 411 per cm³), so PeV gamma rays are absorbed within a few tens of kiloparsecs, never escaping even our own Galaxy. This is one reason PeV astronomy is done with neutrinos rather than photons. The lesson of the table is the inverse scaling: the higher the gamma-ray energy, the longer the wavelength of the light that destroys it, and the denser that target field becomes.

The gamma-ray horizon and the optical depth

The probability that a gamma ray of energy E emitted at redshift z survives its journey to us is e^(−τ_γγ), where the optical depth is the line-of-sight integral of the EBL photon density times the γγ cross section over all angles and energies above threshold:

τ_γγ(E, z) = ∫₀ᶻ dl/dz' dz' ∫dΩ (1−cos θ)/2 ∫ dε  n_EBL(ε, z') σ_γγ(E, ε, θ)

The cross section σ_γγ rises from zero at threshold, peaks at about 0.2 σ_T (a fifth of the Thomson cross section, ≈ 1.3 × 10⁻²⁵ cm²) when the center-of-momentum energy is roughly 1.4 times the threshold, then falls slowly at high energies. Defining the gamma-ray horizon as the redshift where τ_γγ(E, z) = 1, the picture is:

• Below ~10 GeV: τ < 1 out to z ≳ 3. The universe is transparent; Fermi-LAT sees blazars to z > 4.
• At 100 GeV: the horizon is around z ≈ 1.
• At 1 TeV: the horizon collapses to z ≈ 0.1–0.2 (a few hundred Mpc). Beyond it the TeV sky is nearly dark.
• At 10 TeV: only the very nearest blazars (z ≲ 0.03, like Mrk 421 and Mrk 501) are detectable.

The horizon is not a hard wall but an energy-dependent exponential fog. Its practical signature is that an observed blazar spectrum is the intrinsic spectrum multiplied by e^(−τ), producing a characteristic steepening and curvature at the high-energy end. Because we can model the intrinsic spectrum physically (it cannot rise without bound and cannot be harder than the emission mechanism allows), the observed curvature is a direct readout of τ — and hence of the EBL density. This is how gamma-ray astronomy measures the integrated light of all galaxies without ever resolving them: the absorption of a few dozen blazars at known redshifts pins down the EBL, and through it the cosmic star-formation history, in agreement with deep galaxy counts.

Pair cascades: the energy doesn't disappear

Absorbing a TeV gamma ray does not erase its energy; it transfers it to an ultrarelativistic electron-positron pair, each particle carrying roughly half the original TeV. These pairs are immersed in the densest photon field in the universe — the cosmic microwave background — and they immediately inverse-Compton scatter CMB photons, boosting microwave photons (≈ 6 × 10⁻⁴ eV) up to the GeV gamma-ray band. A pair from a 1 TeV photon, with Lorentz factor γ ≈ 10⁶, upscatters the CMB to energies of order γ² × 6×10⁻⁴ eV ≈ hundreds of GeV. Those regenerated gamma rays can pair-produce again, and the process repeats as an electromagnetic pair cascade that degrades one hard photon into a spray of softer ones.

The geometry of the regenerated emission is a powerful probe. If the intergalactic magnetic field (IGMF) in cosmic voids is essentially zero, the pairs travel straight and the cascade GeV emission arrives along the same line of sight, just delayed. If the IGMF is non-zero, the pairs are deflected before they upscatter, smearing the regenerated GeV photons into an extended, time-delayed pair halo around the source. The non-detection of this expected GeV cascade emission from several hard-spectrum TeV blazars (such as 1ES 0229+200) by Fermi-LAT implies the pairs were deflected — placing a widely cited lower bound of B ≳ 10⁻¹⁶ G on the magnetic field of the cosmic voids, one of the only ways we have to weigh magnetism in the emptiest parts of the universe.

Variants and regimes

• Internal vs. external absorption. γγ absorption can happen inside the source (a compact gamma-ray emitter is opaque to its own highest-energy photons — the "compactness" constraint used in gamma-ray bursts) or externally on the diffuse EBL/CMB during propagation. The external version sets the cosmological horizon.
• The compactness argument in GRBs. If a gamma-ray burst were not moving relativistically toward us, its luminosity and short variability would imply a photon density so high that γγ pair production would make it opaque to its own MeV photons — yet we see them. The resolution is bulk Lorentz factors of Γ ≳ 100, which blueshift and de-beam the radiation enough to lower the rest-frame photon energies below threshold. γγ opacity is thus a key piece of evidence that GRB jets are ultrarelativistic.
• Triplet and double pair production. At very high energies a photon can pair-produce off an electron (γ + e⁻ → e⁻ + e⁺ + e⁻, "triplet" production) or two photons can make two pairs; both are higher-order corrections, generally subdominant to the leading processes.
• Lorentz-invariance tests. Some quantum-gravity models predict an energy-dependent shift in the photon dispersion relation that would alter the γγ threshold at the highest energies. The clean detection of multi-TeV photons from cosmological distances therefore constrains Lorentz-invariance violation at and beyond the Planck scale.
• Magnetic pair production. Near magnetars and pulsar polar caps, a single photon can pair-produce in an ultrastrong magnetic field (which absorbs the transverse momentum), a distinct one-photon channel that governs pulsar gamma-ray cutoffs.

Observational status and applications

• EBL measurement. The Fermi-LAT collaboration (2018) and the H.E.S.S./MAGIC/VERITAS teams have used the γγ absorption imprint on stacked blazar spectra to reconstruct the EBL intensity across redshift, matching the integrated light of resolved galaxy surveys — a triumph of using the highest-energy photons to count the contribution of every star.
• The Hubble constant from gamma rays. Because the horizon depends on the geometry and expansion of the universe, the energy at which blazar spectra steepen has been used to estimate H₀ and the matter density Ω_m independently of the distance ladder (Domínguez et al. 2019).
• Cosmic voids' magnetism. As above, cascade non-detections set B ≳ 10⁻¹⁶ G in voids, a unique window on primordial magnetogenesis.
• The "TeV transparency anomaly." Several very hard, distant blazars appear slightly less absorbed than the standard EBL predicts. Proposed explanations range from underestimated source hardness to exotic physics such as photon-to-axion-like-particle oscillations that let gamma rays skip the absorbing medium — an active and contested research frontier.
• Future instruments. The Cherenkov Telescope Array (CTA) will detect hundreds of blazars and map τ_γγ(E, z) with the precision needed to settle the EBL, the void magnetic field, and the transparency anomaly.

Common pitfalls and misconceptions

• Thinking a single photon can pair-produce in vacuum. It cannot — momentum conservation forbids it. A nucleus, a second photon, or a strong magnetic field must absorb recoil.
• Confusing the two thresholds. The nuclear-field (one-photon) threshold is E_γ > 1.022 MeV. The γγ (two-photon) threshold is on the product E₁E₂ > (m_ec²)² ≈ 0.261 MeV², which is an entirely different condition — a 1 TeV photon and a 0.26 eV photon clear it even though neither alone is near 1 MeV.
• Believing the gamma-ray horizon is a fixed distance. It is strongly energy-dependent: transparent at 10 GeV to z > 3, opaque at 1 TeV beyond z ≈ 0.15. There is no single horizon, only a τ_γγ(E, z) surface.
• Assuming absorbed energy is lost. It is reprocessed into pairs that upscatter the CMB and regenerate GeV gamma rays — the cascade. The energy reappears at lower photon energies, possibly as a halo.
• Treating γγ absorption and the photoelectric/Compton processes as the same kind of thing. Photoelectric absorption and Compton scattering involve photons interacting with matter; γγ pair production needs only light, and is the only one that operates over intergalactic distances on the diffuse radiation field.
• Forgetting the angle factor. The threshold E₁E₂(1 − cos θ) means glancing collisions (small θ) need much higher energies; the head-on case is the easiest and dominates an isotropic target field's contribution.

Quantitative analysis: the threshold from invariants

The cleanest derivation uses the relativistic invariant s, the squared total four-momentum of the two-photon system. For photons with four-momenta p₁ and p₂,

s = (p₁ + p₂)² = p₁² + p₂² + 2 p₁·p₂ = 0 + 0 + 2 p₁·p₂

since photons are massless. With energies E₁, E₂ and a collision angle θ between their directions, the dot product gives

p₁·p₂ = (E₁E₂/c²)(1 − cos θ)   ⟹   s = (2 E₁E₂/c²)(1 − cos θ)

Pair production becomes kinematically allowed when s reaches the squared rest energy of the lightest pair we can make, the electron-positron pair, produced at rest in the center-of-momentum frame:

s c² ≥ (2 m_e c²)²
2 E₁E₂ (1 − cos θ) ≥ 4 (m_e c²)²
E₁E₂ (1 − cos θ) ≥ 2 (m_e c²)²

Setting θ = π (head-on, 1 − cos θ = 2) recovers the headline result E₁E₂ ≥ (m_e c²)² ≈ 0.261 MeV². Plugging numbers for a 1 TeV gamma ray gives a minimum partner energy of 0.26 eV; for the cross section to be near its peak (s ≈ 1.4 × threshold) the ideal partner is closer to 1 eV, the near-IR starlight band. The cross section itself, derived by Breit and Wheeler, is

σ_γγ(β) = (3 σ_T / 16)(1 − β²)[ (3 − β⁴) ln((1+β)/(1−β)) − 2β(2 − β²) ]

where β = √(1 − 1/x) is the velocity of the produced electron in the center-of-momentum frame and x = s c²/(2 m_ec²)² is the dimensionless energy above threshold. σ_γγ vanishes at β = 0 (threshold), rises to a maximum of ≈ 0.26 σ_T at x ≈ 3.5 (β ≈ 0.7, center-of-momentum energy ≈ 1.4× threshold), and falls as (ln x)/x at high energies. The peak value ≈ 0.26 × 6.65 × 10⁻²⁵ cm² ≈ 1.7 × 10⁻²⁵ cm² is what, multiplied by the EBL photon density and a Gpc path length, drives τ_γγ to order unity for TeV photons across the local universe.