Gravitational Redshift

Gravitational redshift is the simplest non-trivial prediction of any metric theory of gravity, and the first place a student meets curved spacetime doing real physical work. Fire a beam of light straight up from the surface of a star; a static observer higher in the field receives it at a lower frequency than it was emitted — shifted toward the red. Run the experiment upside down and light falling into the well is blueshifted by the same rule. The effect is not a Doppler shift: emitter and receiver are both at rest, nothing is moving. It is a statement about the relationship between proper time and coordinate time in a spacetime that is curved, or even merely accelerated. The deep origin is the equivalence principle, and this is what makes redshift so revealing. Einstein's 1907–1911 elevator argument shows that a uniformly accelerating rocket in flat spacetime must measure a redshift between floor and ceiling, so a static laboratory in a gravitational field of the same local strength must measure it too. That fixes the effect to first order in Φ/c² without ever solving Einstein's equations — which is exactly why gravitational redshift tests the existence of gravitational time dilation and the metric structure of gravity, while being nearly blind to the detailed form of the field equations (it is largely a test of local position invariance). The geometry sharpens the picture. In a static spacetime with a timelike Killing vector ξ, the projection −ξ·p is conserved along a photon's null geodesic, and the frequency each static observer actually measures differs from it by the local factor √(−g_tt). Take the ratio at two heights and the conserved energy cancels: frequencies scale as √(−g_tt). In Schwarzschild this is exact all the way to the horizon, where the factor vanishes and the redshift diverges — the defining signature of a black hole.

Key takeaways

Gravitational redshift is the drop in frequency (rise in wavelength) of light climbing out of a gravitational potential well, as read by a static observer higher in the field.
It follows directly from the equivalence principle: an accelerating frame predicts a Doppler-like shift that, made local, is a real frequency change in a static gravitational field.
To lowest order the fractional shift is Δν/ν = −ΔΦ/c²; exactly, frequencies measured by static observers scale as √(−g_tt).
It is a gravitational time-dilation effect, not a propagation effect: clocks deeper in the well tick slower, so a photon emitted there is received with fewer cycles per second.
Confirmed by Pound–Rebka (1959–60) at 2.5×10⁻¹⁵ over 22.5 m, by the solar line shift of 2.12×10⁻⁶, and it is the reason GPS clocks are pre-corrected by about +45 μs/day of altitude.
It is a first-order, PPN-insensitive probe of the metric, not of the Einstein field equations themselves.
Key idea: frequencies measured by static observers scale as √(−g_tt), so ν_rec/ν_em = √(g_tt(em)/g_tt(rec)), which to first order is 1 − ΔΦ/c². Light climbing out of a well loses frequency because clocks run slow deep in the well.

The core insight

Frequencies measured by static observers scale as √(−g_tt): ν_rec/ν_em = √(g_tt(r_em)/g_tt(r_rec)), which to first order is 1 − ΔΦ/c². Light climbing out of a potential well loses frequency because the clocks deep in the well run slow.

Where this lives today

The frontier has moved from confirming redshift to using it as a precision tool. Optical lattice clocks are now stable and accurate near the 10⁻¹⁸ level, which corresponds to a height change of about one centimetre in Earth's field — so a redshift is measurable across a tabletop. In 2010 an aluminium-ion clock resolved the shift over a 33 cm elevation change, and by 2018–2022 groups at JILA and elsewhere measured the redshift within a single millimetre-tall cloud of strontium atoms, turning Einstein's tower experiment into a device that fits under a microscope. This underpins the emerging field of relativistic geodesy, where transportable clocks map the geopotential — the shape of Earth's gravity field — by comparing tick rates, complementing satellite gravimetry. In space, the accidentally eccentric orbits of the Galileo satellites GSAT-0201/0202 were exploited in the 2018 GREAT and independent analyses to bound the redshift violation parameter to |α| ≲ 3×10⁻⁵, the best clock test of local position invariance. And in the strong field, the GRAVITY interferometer detected the gravitational (plus transverse-Doppler) redshift of the star S2 as it swung through perihelion around the four-million-solar-mass black hole Sagittarius A* in 2018 — the same √(−g_tt) factor, now read off a star skimming an event horizon.

The physical picture

Two pictures carry the whole effect. The first is Einstein's elevator: a rocket accelerating at g in empty space is locally indistinguishable from a laboratory held static in a gravitational field of strength g. A pulse fired from the floor takes a time h/c to reach the ceiling; in that interval the ceiling has gained an upward speed Δv = gh/c and so recedes from the light it is about to catch, measuring a first-order Doppler redshift Δν/ν = −gh/c². The equivalence principle forbids any local experiment from telling acceleration from gravity, so the same shift must appear in a real static field with gh read as the potential difference ΔΦ. That single argument delivers the answer before any curvature has been computed.

Figure

Einstein's elevator: an accelerating rocket in empty space is locally identical to a lab held static in gravity. Light from the floor is received redshifted at the ceiling by Δν/ν = −gh/c². For g = 9.8 m/s² over h = 22.5 m this is 2.5×10⁻¹⁵, the Pound–Rebka value.

The second picture is the one that survives into the strong field: a photon is a clock. Its oscillation count is fixed at emission, so a receiver higher in the potential — whose own clock runs faster — counts those cycles against a faster standard and records fewer per second, a redshift. The size of the effect is entirely the ratio of clock rates √(−g_tt) at the two heights. In Schwarzschild that factor is √(1 − r_s/r), which is order 10⁻⁶ at the Sun's surface, order 10⁻¹⁵ on an Earth-bound tower, and falls to zero at the horizon r = r_s where the redshift becomes infinite. The second figure plots this exact factor, so that the laboratory, astrophysical, and black-hole regimes all sit on one curve.

Figure

The exact Schwarzschild redshift factor √(1 − r_s/r) for light received far away, versus radius in units of r_s. It rises from 0 at the horizon toward 1 far out; the weak-field regime (Sun, Earth) lives at the far right where the shift is a tiny 10⁻⁶ to 10⁻¹⁵, while at r = r_s the factor is 0 and the redshift z is infinite.

The core equations

The master relation says that the frequency a static observer measures is the conserved photon energy divided by the local time-dilation factor, so the ratio at two heights is a pure ratio of metric components. Everything else in this article is a limit or a numerical evaluation of this one statement.

\frac{\nu_{\rm rec}}{\nu_{\rm em}}=\sqrt{\frac{-g_{tt}(x_{\rm em})}{-g_{tt}(x_{\rm rec})}} \qquad (1)

where: ν_em, ν_rec — frequencies measured locally by static observers at emission and reception; g_tt — time–time component of the static metric (with −g_tt > 0); x_em, x_rec — the two spacetime positions. Reading: light emitted deep in the well, where −g_tt is smaller, arrives higher up, where −g_tt is larger, at a lower frequency — a redshift. Specializing to a mass M in Schwarzschild coordinates makes the factor explicit and shows the horizon divergence. This form is exact — no weak-field approximation — and is what governs light escaping a neutron star or a black hole.

\frac{\nu_{\rm rec}}{\nu_{\rm em}}=\sqrt{\frac{1-r_s/r_{\rm em}}{1-r_s/r_{\rm rec}}},\qquad r_s=\frac{2GM}{c^2} \qquad (2)

where: r_s — Schwarzschild radius of the mass M; r_em, r_rec — areal radial coordinates of emitter and receiver; G, c — Newton's constant and the speed of light. Reading: as r_em → r_s the numerator → 0, so the received frequency vanishes and the redshift diverges. For everyday and astrophysical fields the potential is shallow, and expanding equation (2) to first order in Φ/c² gives the form used at the bench. This is the equivalence-principle result recovered from geometry.

\frac{\Delta\nu}{\nu}\equiv\frac{\nu_{\rm rec}-\nu_{\rm em}}{\nu_{\rm em}}\approx-\frac{\Phi_{\rm rec}-\Phi_{\rm em}}{c^2} \qquad (3)

where: Φ — Newtonian gravitational potential (negative inside a well); Δν/ν — fractional frequency shift; a negative value is a redshift. Reading: climbing out of the well means Φ_rec > Φ_em, so Δν/ν < 0. Spectroscopists quote the effect as a dimensionless redshift z, the fractional stretch of the wavelength, most often for reception at infinity. It is simply the inverse of the frequency ratio minus one.

z\equiv\frac{\lambda_{\rm rec}-\lambda_{\rm em}}{\lambda_{\rm em}}=\frac{\nu_{\rm em}}{\nu_{\rm rec}}-1=\frac{1}{\sqrt{1-r_s/r_{\rm em}}}-1 \qquad (4)

where: z — dimensionless redshift, positive for light leaving a well; λ — wavelength, inversely proportional to frequency. Reading: for a shallow well z ≈ GM/(R c²); at the horizon z → ∞. The physical engine is that a static clock's proper time runs slow relative to coordinate time by exactly the redshift factor, which in the weak field is set by the Newtonian potential. This is the bridge from geometry to the "clocks tick differently" language.

g_{tt}=-\left(1+\frac{2\Phi}{c^2}\right),\qquad \frac{\mathrm{d}\tau}{\mathrm{d}t}=\sqrt{-g_{tt}}\approx1+\frac{\Phi}{c^2} \qquad (5)

where: τ — proper time of a static clock; t — global coordinate time; dτ/dt — the local clock rate relative to coordinate time. Reading: deeper in the well (Φ more negative) the clock runs slow, and the frequency it stamps on emitted light is correspondingly low. Over a short vertical baseline the potential is linear, Φ = gz, and equation (3) collapses to a compact laboratory formula — the one Pound and Rebka set out to test.

\frac{\Delta\nu}{\nu}\approx-\frac{gh}{c^2} \qquad (6)

where: g — local gravitational acceleration; h — height gained by the light. Reading: send light up (h > 0) and it reddens; send it down and it blues by the same amount. Finally, to turn the prediction into a test one attaches a phenomenological knob to the potential term. General relativity, and indeed any theory obeying local position invariance, pins it to zero.

\frac{\Delta\nu}{\nu}=-(1+\alpha)\frac{\Delta\Phi}{c^2} \qquad (7)

where: α — dimensionless violation parameter of local position invariance; GR predicts α = 0. Reading: experiments quote bounds on α, currently |α| ≲ 10⁻⁵ and tightening.

Derivation AEquivalence principle and the accelerating rocket

The goal is to obtain the first-order redshift using nothing but special relativity in an accelerating frame, and then to promote it to a genuine gravitational field by the equivalence principle, so that the answer Δν/ν = −gh/c² is fixed before any spacetime curvature or field equation is introduced. Take a rocket of height h accelerating with constant proper acceleration g in flat, empty spacetime, with a source on the floor and a detector on the ceiling, and work to first order in the small speed the rocket gains during one light-crossing.

Step 1. A pulse leaves the floor and travels the height h to the ceiling; for a slowly moving rocket the light-crossing time is t = h/c to leading order. Why: at first order we may neglect the small change in the rocket's speed during transit, so the light simply covers a fixed distance h at speed c.

Step 2. During that time the rocket, and with it the ceiling, accelerates and gains an upward velocity Δv = g·t = gh/c relative to the state of motion at the instant of emission. Why: the ceiling that finally absorbs the light is moving away from the point in the rocket's history where the light was launched, which is what a Doppler analysis needs.

Step 3. The receding receiver measures a first-order longitudinal Doppler redshift Δν/ν ≈ −Δv/c = −gh/c². Why: to first order the relativistic Doppler formula reduces to the familiar −Δv/c, and the sign is a redshift because emitter and absorber separate during the flight.

Step 4. Invoke the equivalence principle: a static laboratory in a uniform gravitational field of strength g is locally indistinguishable from this rocket, and gh is the potential difference ΔΦ between floor and ceiling, so Δν/ν = −ΔΦ/c². Why: no local experiment can separate acceleration from gravity, so the kinematic shift must be reproduced identically as a static gravitational redshift, which also matches the weak-field metric g_tt = −(1 + 2Φ/c²).

Go deeper · Derivation B · 1 minExact redshift from the timelike Killing vector

Derivation BExact redshift from the timelike Killing vector

The goal is to derive the exact frequency ratio between two static observers in any static spacetime, and then specialize to Schwarzschild, by exploiting the conserved quantity that a timelike Killing vector supplies along the photon's null geodesic. The strategy is to separate two distinct objects that are easy to conflate: the conserved coordinate energy −ξ·p, which is the same everywhere along the ray, and the locally measured frequency −u·p, which depends on the observer's four-velocity and therefore on where the observer sits in the potential.

Step 1. In a static spacetime the vector ξ = ∂_t is a Killing vector because the metric has no explicit t-dependence, and Killing's equation guarantees that E = −ξ·p is conserved along the photon's geodesic, d(ξ·p)/dλ = 0. Why: a symmetry of the metric always yields a constant of geodesic motion, and time-translation symmetry gives a conserved "energy at infinity" that is common to emission and reception.

Step 2. A static observer moves along ξ, so the unit four-velocity is u = ξ/√(−ξ·ξ) = ξ/√(−g_tt), the normalization fixed by u·u = −1 (or −c² in dimensionful units). Why: the observer must stand still in the static coordinates, so their four-velocity points along ξ, and the √(−g_tt) in the denominator is precisely the gravitational time-dilation factor.

Step 3. The frequency this observer actually measures is hν = −u·p = (−ξ·p)/√(−g_tt) = E/√(−g_tt). Why: the measured photon energy is the projection of its four-momentum on the observer's four-velocity, and the observer-dependent √(−g_tt) is the only thing that changes from one static observer to another.

Step 4. Form the ratio at reception and emission; the conserved E cancels, leaving ν_rec/ν_em = √(−g_tt(em))/√(−g_tt(rec)). In Schwarzschild −g_tt = 1 − r_s/r, giving ν_rec/ν_em = √[(1 − r_s/r_em)/(1 − r_s/r_rec)], exact to the horizon where r_em → r_s sends the ratio to zero. Why: the whole redshift is the mismatch between how fast the two static clocks run, encoded in the ratio of time-dilation factors, and its divergence at r_s is the infinite redshift of the black-hole surface.

Go deeper · Derivation C · 1 minWeak field limit and the clock-rate connection

Derivation CWeak field limit and the clock-rate connection

The goal is to show explicitly that the abstract factor √(−g_tt) becomes the everyday statement "clocks run slower deeper in the potential," and to recover the Newtonian-potential redshift (3) from the exact scaling (1), so that geometry and the language of ticking clocks are seen to be one thing. The route is to insert the weak-field metric into the redshift factor, expand, and read off both the clock rate and the frequency shift.

Step 1. In the Newtonian limit the time–time metric component is g_tt = −(1 + 2Φ/c²) with |Φ|/c² ≪ 1, so a static clock's proper time obeys dτ/dt = √(−g_tt) ≈ 1 + Φ/c². Why: this is the content of the Newtonian limit of GR — the single component g_tt carries the Newtonian potential, and its square root is the rate at which a local clock advances relative to coordinate time.

Step 2. Because Φ < 0 inside a well, dτ/dt < 1 there: the deeper clock runs slow relative to a clock higher up. A photon emitted at the lower clock carries a frequency stamped by that slow clock. Why: frequency is cycles per unit proper time at the source, so a slow source clock produces light that a faster receiver clock will judge to be low in frequency.

Step 3. Put the two rates into the exact ratio (1): ν_rec/ν_em = (1 + Φ_em/c²)/(1 + Φ_rec/c²) ≈ 1 + (Φ_em − Φ_rec)/c². Why: to first order the quotient of the two clock rates is just the difference of potentials, which is all the redshift there is in the weak field.

Step 4. Rearranged, Δν/ν = (ν_rec − ν_em)/ν_em ≈ −(Φ_rec − Φ_em)/c² = −ΔΦ/c², and over a short baseline with Φ = gz this is −gh/c². Why: this is equations (3) and (6), now derived as the shallow-well shadow of the exact geometric scaling rather than assumed.

Limiting cases

The exact Schwarzschild factor (2) contains every regime of interest as a limit; the table collects the three that matter.

Case	Result	Meaning
Weak field (lab, Earth, Sun)	Δν/ν ≈ −ΔΦ/c² = −(GM/c²)(1/r_em − 1/r_rec)	First order in Φ/c². Light leaving the Sun for infinity gives Δν/ν ≈ −2.12×10⁻⁶.
Uniform field over height h	Δν/ν ≈ −gh/c²	The Pound–Rebka limit. For g = 9.8 m/s² and h = 22.5 m,	Δν/ν	≈ 2.5×10⁻¹⁵.
Strong field / horizon	ν_rec/ν_em = √(1 − r_s/r_em) → 0 as r_em → r_s	Light from the horizon r_s = 2GM/c² is infinitely redshifted; this defines the black-hole surface for a distant observer.

Common trapRedshift is not the photon losing energy as it climbs

It is tempting to say the photon "spends energy" fighting gravity on the way up and so comes out redder. That picture is wrong, and it leads to contradictions. Locally, light always travels at c, and a photon's energy is not an invariant — it is whatever the local observer measures, hν = E/√(−g_tt). Nothing is being drained from the photon in transit; the conserved quantity E = −ξ·p is in fact constant all the way up. The redshift is entirely a comparison of the clock rates of the two static observers at different potentials: the receiver's faster clock counts the same conserved energy as a lower frequency. Two further confusions to avoid: gravitational redshift is not cosmological redshift (which comes from the expansion of space between comoving galaxies), and it is not a kinematic Doppler shift (which requires relative motion of a fixed emitter and receiver). Here emitter and receiver are both at rest, and space is not expanding.

Worked examples

1 · The solar gravitational redshift (fully worked)

Light leaves the surface of the Sun and is received far away, where the potential is essentially zero. Find the fractional redshift and its velocity equivalent.

Step 1. Use the weak-field formula (3) with the receiver at infinity, Φ_rec = 0, so Δν/ν = −(Φ_em)/c² = −GM⊙/(R⊙c²). The relevant length is GM⊙/c² = 1.477×10³ m (half the Sun's Schwarzschild radius), and the solar radius is R⊙ = 6.96×10⁸ m.

Step 2. Evaluate the ratio: Δν/ν = −(1.477×10³)/(6.96×10⁸) = −2.12×10⁻⁶. The light is redshifted by about two parts in a million — the light climbing out of the Sun's well arrives slightly reddened.

Step 3. Convert to a velocity equivalent so it can be compared with Doppler data: |Δν/ν|·c = 2.12×10⁻⁶ × 3.00×10⁵ km/s = 0.636 km/s ≈ 636 m/s. This is the shift solar spectral lines show once convective blueshifts and rotation are subtracted, and it agrees with helioseismology-corrected measurements. Result: Δν/ν ≈ −2.12×10⁻⁶, equivalent to 636 m/s.

2 · The Pound–Rebka tower (fully worked)

A 14.4 keV gamma ray from ⁵⁷Fe is sent up a tower of height h = 22.5 m at Earth's surface. Predict the fractional shift the experiment had to resolve.

Step 1. The field is uniform over such a short baseline, so use equation (6): Δν/ν = −gh/c² with g = 9.8 m/s², h = 22.5 m, c = 3.00×10⁸ m/s.

Step 2. Numerator gh = 9.8 × 22.5 = 220.5 m²/s². Denominator c² = 9.00×10¹⁶ m²/s². So |Δν/ν| = 220.5/(9.00×10¹⁶) = 2.45×10⁻¹⁵ ≈ 2.5×10⁻¹⁵.

Step 3. Interpret the number: this is a shift of a few parts in 10¹⁵, far finer than any ordinary spectroscopy. Pound and Rebka reached it only because the Mössbauer effect gives a recoil-free line with fractional width near 10⁻¹² and they Doppler-tuned the source to null the shift, confirming the prediction to about 10% (later 1% with Pound and Snider). Result: |Δν/ν| ≈ 2.5×10⁻¹⁵, blueshift downward and redshift upward.

3 · The GPS clock offset (fully worked)

A GPS satellite orbits at altitude ≈ 20,200 km (r ≈ 26,600 km from Earth's centre, Earth radius R⊕ ≈ 6,370 km). Estimate the gravitational clock-rate difference relative to the ground, and combine it with the orbital-speed effect.

Step 1. The fractional rate difference is ΔΦ/c² with ΔΦ = GM⊕(1/R⊕ − 1/r) and GM⊕ = 3.986×10¹⁴ m³/s². Compute 1/R⊕ − 1/r = 1/(6.37×10⁶) − 1/(2.66×10⁷) = 1.570×10⁻⁷ − 0.376×10⁻⁷ = 1.194×10⁻⁷ m⁻¹.

Step 2. So ΔΦ = 3.986×10¹⁴ × 1.194×10⁻⁷ = 4.76×10⁷ m²/s², and ΔΦ/c² = 4.76×10⁷/9.00×10¹⁶ = 5.3×10⁻¹⁰. Over one day (86,400 s) this is 5.3×10⁻¹⁰ × 86,400 s ≈ +46 μs/day — the satellite clock runs fast because it sits higher in the well (textbook value +45 μs/day).

Step 3. Special-relativistic time dilation from the orbital speed (v ≈ 3.9 km/s) subtracts about −7 μs/day, leaving a net +38 μs/day. Uncorrected, the growing timing error would corrupt positions by roughly 10 km per day, so GPS clocks are deliberately offset in frequency before launch. Result: gravitational +45, kinematic −7, net +38 μs/day.

Physics of the portrait

Gravitational redshift is Einstein's effect twice over: he predicted it from the equivalence principle in 1907, years before he had a field equation, and it remained the most stubborn of his three classic tests to confirm. The portrait is his because the phenomenon is, at bottom, the equivalence principle made visible in light.

Origins · 1 minAlbert Einstein and how this was discovered

Albert Einstein

1879–1955 · general relativity and the equivalence principle

Einstein arrived at gravitational redshift not from geometry but from what he later called "the happiest thought of my life" — the realization, in 1907, that a freely falling observer feels no gravity, so acceleration and gravitation are locally the same. In that year's review and again in a sharper 1911 Prague paper, he used the accelerating-frame argument to predict that light escaping the Sun should be redshifted by GM⊙/(R⊙c²) ≈ 2×10⁻⁶, and that clocks lower in a gravitational potential must run slow — all before the field equations of 1915 existed. This was a striking inversion of method: a firm quantitative prediction resting on a symmetry principle rather than on a dynamical law, which is exactly why the redshift tests the metric-and-equivalence foundation of gravity while being nearly blind to the field equations that Mercury's perihelion and light bending probe. The prediction outran the experiments by half a century. Solar-line measurements were confounded by convection and pressure shifts and stayed ambiguous into the 1950s; the white-dwarf test on Sirius B, announced by W. S. Adams in 1925, was later shown to be badly contaminated. Only Pound and Rebka's 1959 Mössbauer experiment gave a clean terrestrial confirmation, and the 1976 Gravity Probe A maser pushed the agreement to 7×10⁻⁵. Einstein did not live to see either, but he had already extracted the answer from a man in a falling elevator.

History

1907

— Einstein's equivalence-principle argument predicts gravitational time dilation and redshift

1911

— Einstein's Prague paper gives the quantitative solar redshift GM/Rc²

1916

— Schwarzschild's solution yields the exact redshift factor √(1 − r_s/r)

1925

— Adams reports a Sirius B redshift, later found badly contaminated

1959

— Pound and Rebka measure the terrestrial shift with the Mössbauer effect

1976

— Gravity Probe A confirms the maser redshift to 7×10⁻⁵

Going further

Texts. Misner, Thorne & Wheeler, Gravitation (1973), §7.2–7.4 and §25.4 · Wald, General Relativity (1984), §6.3 · Carroll, Spacetime and Geometry (2004), §5.4 and §2.5 · Hartle, Gravity (2003), Ch. 6 and §9.2 · Schutz, A First Course in General Relativity (2009), §5.1 and §10.3 · Pound & Rebka, Phys. Rev. Lett. 4, 337 (1960) · Ashby, "Relativity in the Global Positioning System," Living Rev. Relativity 6, 1 (2003). The key numerical anchors used throughout: $GM_\odot/c^2 = 1.477\times10^3$ m and $R_\odot = 6.96\times10^8$ m give the solar shift $2.12\times10^{-6}$ ; $g = 9.8$ m/s² over $h = 22.5$ m gives the Pound–Rebka shift $2.5\times10^{-15}$ ; $GM_\oplus = 3.986\times10^{14}$ m³/s² with $r = 26{,}600$ km and $R_\oplus = 6{,}370$ km gives the GPS gravitational rate $+45\ \mu$ s/day and net $+38\ \mu$ s/day.

Prerequisites. Weak-Field Limit and Linearized Gravity · Newtonian Limit of General Relativity · Schwarzschild Geodesics and Effective Potential. Leads to. Pound–Rebka Experiment · Shapiro Time Delay · Parametrized Post-Newtonian (PPN) Formalism · PPN Parameters Beta and Gamma.

Ask the tutor: Ask me to walk through the Killing-vector cancellation, re-derive the GPS +38 μs/day from scratch, or push the exact √(1 − r_s/r) factor down toward a neutron-star surface or a black-hole horizon.

Is gravitational redshift the same as the Doppler effect?

No, though the equivalence-principle derivation borrows Doppler machinery. A Doppler shift requires relative motion between a fixed emitter and receiver; here both are permanently at rest in the static field, and nothing moves. What the elevator argument shows is that a momentarily accelerating frame can be analyzed with a Doppler shift, and the equivalence principle then translates that into a static gravitational field where the shift persists with no motion at all. The cleanest way to see the difference is that gravitational redshift depends only on the potential difference between two fixed points, not on any velocity — two clocks at fixed heights disagree forever, whereas a Doppler shift vanishes the moment relative motion stops.

Does the photon lose energy as it climbs out of the well?

Not in any invariant sense. The conserved quantity along the ray is E = −ξ·p, the "energy at infinity," and it is exactly the same at the bottom and the top. What differs is the locally measured energy hν = E/√(−g_tt), because the two observers' clocks run at different rates. Saying the photon "loses energy" smuggles in a global notion of energy that does not exist in curved spacetime; energy is only defined relative to an observer, and the redshift is the mismatch between the two observers, not a loss suffered by the light. Locally, at every point, the photon moves at c and carries whatever energy that point's observer assigns.

Why does gravitational redshift barely test general relativity's field equations?

Because it is fixed by the equivalence principle and by a single metric component, g_tt = −(1 + 2Φ/c²), which any metric theory reproducing the Newtonian limit must share. It confirms that gravity curves time and that clocks obey local position invariance, but it says almost nothing about the spatial curvature or the nonlinear structure that the Einstein equations dictate. Those are probed by the PPN parameters β and γ, which govern light bending, the Shapiro delay, and Mercury's perihelion precession — effects that involve the spatial metric. Redshift leaves β and γ essentially unconstrained, which is why it is best described as a test of the foundations of gravity rather than of GR specifically.

What actually happens to light emitted right at a black-hole horizon?

The redshift factor √(1 − r_s/r_em) goes to zero as r_em → r_s, so the received frequency goes to zero and z → ∞. A distant observer sees light from just above the horizon arrive progressively redder, dimmer, and slower, until it fades below detectability. This is the origin of the old name "frozen star": an object falling in appears to a faraway observer to slow, redden, and freeze at the horizon, never quite crossing it, even though in the object's own proper time it sails through in finite time. The infinite redshift is a statement about the distant observer's clock, not about anything singular happening to the infalling matter locally.

How can white-dwarf and neutron-star spectra be used to weigh the star?

The surface redshift is z ≈ GM/(Rc²) to leading order, so a measured line shift directly gives the compactness M/R. For the white dwarf Sirius B the shift is equivalent to about 80 km/s, historically used to estimate its mass-to-radius ratio; for a neutron star the effect is order 10–30%, large enough that the exact factor √(1 − r_s/R) must be used. Because the redshift fixes M/R while other observations (orbital dynamics, thermal emission) fix M or R separately, combining them constrains the radius at fixed mass — and the radius of a neutron star is a sensitive probe of the poorly known equation of state of dense nuclear matter.

If the effect is only 10⁻¹⁵ on Earth, why does GPS care?

Because a navigation system integrates the tiny rate difference over time, and time turns into distance at the speed of light. The gravitational rate offset for a GPS satellite is about +45 μs/day; multiplied by c that is roughly 13 km of ranging error accumulating per day, and even the residual after the −7 μs/day speed effect would drift positions by kilometres within hours. GPS demands nanosecond timing to give metre-level positions, so a part-in-10¹⁰ clock effect is enormous by its standards. The system handles it by pre-tuning each satellite's clock frequency before launch so that, once in orbit, it ticks at the ground rate — a routine engineering correction that is pure general relativity.