Ramsey Interferometry and Spectroscopy

A beam of atoms crosses not one microwave zone but two, with a long stretch of empty space between them — and out of that emptiness comes the narrowest spectroscopic line physics knows how to draw. Ramsey interferometry takes the two-level atom and its Bloch-sphere dynamics — the resonant Rabi rotation and the free precession of a coherent superposition, both owned by the preceding optical-Bloch articles — and arranges them into an interferometer for the atom's own internal phase. The first zone splits the atom into an equal superposition; during the field-free drift the two amplitudes tick at different rates because the atom is detuned by δ; the second zone recombines them so the accumulated phase δT becomes a measurable population. The payoff is entirely in that drift time T. A single interaction of duration τ can only resolve frequencies to about 1/τ, and stretching τ means building a long region whose field is uniform in amplitude and phase over its whole length — an engineering nightmare. Ramsey's 1949 insight was to demand instead only two short, well-characterised zones and let the atom coast, phase-coherent, between them. The interference fringes then have a central width of order 1/(2T): the resolution is set by how long the atom can be made to remember its phase, not by how long the field acts on it. That single idea is why the caesium beam and the caesium fountain that realise the SI second both work this way, and why the same interference, moved onto an optical transition and read with an ultrastable laser, defines today's most accurate clocks. What follows develops the Bloch-sphere mechanism, the full lineshape, the fringe-width scaling, and the systematic shifts that pull the fringe centre.

Key takeaways

Ramsey's separated oscillatory fields replace one long interaction with two short pulses bracketing a free-evolution time T, so coherence accumulates phase over T, not over the pulse duration τ.
Two π/2 pulses give $P_e=\tfrac12[1+\cos(\delta T+\varphi)]$ — cosine fringes whose central FWHM scales as $\approx 1/(2T)$ , not $1/\tau$ , beating the single-zone Rabi lineshape by $\sim T/\tau$ .
The quality factor is $Q=\nu_0/\Delta\nu\approx 2\nu_0 T$ : long coherence and high transition frequency both buy resolution, which is why optical clocks reach uncertainties below $10^{-18}$ .
The picture is a Bloch-sphere interferometer — first pulse to the equator, free precession by δT, second pulse converts azimuth into population.
Second-order Doppler, Zeeman, collisional, AC-Stark and cavity-phase shifts all pull the fringe centre; taming them is the craft of precision Ramsey spectroscopy.
Key idea: two short, phase-coherent pulses separated by a long free-evolution time T turn the atom into an internal-state interferometer whose fringe spacing is 1/T in frequency, so resolution is set by the storage time T, not the pulse duration.

The core insight

Two short, phase-coherent pulses separated by a long free-evolution time T turn the atom into an internal-state interferometer whose fringe spacing is 1/T in frequency; the resolution is set by the storage time T, not by the pulse duration, so long coherence gives narrow lines.

Where this lives today

The caesium fountain is the working realisation of Ramsey's idea and the primary standard behind international atomic time. Laser-cooled caesium is gathered in an optical molasses, launched upward in a "fountain", and passed through a single microwave cavity twice — once rising, once falling — so the two Ramsey zones are the same cavity crossed half a second apart. That geometry both lengthens T and, because the atoms are microkelvin-cold, all but erases the second-order Doppler shift; the best fountains realise the SI second to a few parts in $10^{16}$ and discipline the timescale behind GPS and the power and financial grids. The frontier is the optical clock: the same Ramsey (or Rabi) interrogation on a narrow optical transition, read with a laser locked to that line and counted by a frequency comb. Strontium and ytterbium atoms in magic-wavelength lattices — depths tuned so the light shifts $|g\rangle$ and $|e\rangle$ equally — and single ions in Paul traps now support free times of seconds without motional dephasing, reaching the $10^{-18}$ range. That is precise enough to see the gravitational redshift across a centimetre of table height, turning the clock into a relativistic altimeter and a probe for drifting constants, dark matter, and a possible redefinition of the second.

The physical picture

Picture two runners who start side by side. The first pulse splits them into the two internal-state amplitudes; during the free time T they run at slightly different phase rates because the atom is detuned by δ; the second pulse brings them back together, and whether they arrive in step or out of step depends entirely on the phase δT they piled up while apart. Sweep the detuning and you sweep that phase, tracing cosine fringes of period 2π/T in δ, or 1/T in ordinary frequency. A tiny frequency error, given a long wait, winds up into a large, easily measured phase — the whole reason long T means high resolution: the drift time is a lever arm turning frequency error into readable phase. On the Bloch sphere the mechanism is exact and three-step: a resonant π/2 pulse tips the vector from the south pole (ground state) to the equator, making $\tfrac{1}{\sqrt2}(|g\rangle+|e\rangle)$ ; the drift precesses it about the vertical axis by the azimuthal angle δT; the second π/2 pulse tips it back toward a pole by an amount set by that azimuth — converting azimuth into height $s_z$ , hence into $P_e=\tfrac12(1+\cos\delta T)$ , as the next figure shows.

Figure

The Ramsey sequence on the Bloch sphere for a caesium fountain with pulse time τ ≈ 10 ms and free-evolution time T ≈ 0.5 s. A π/2 pulse tips the vector from the south pole to the equator, free precession winds its azimuth by δT, and the second π/2 pulse maps that azimuth onto the height s_z, so Pₑ = ½（1 + cos δT）

A second, broader structure hides under the fast fringes. Each short pulse can only excite atoms whose detuning lies inside its Rabi lineshape — a window of width $\sim\Omega\sim 1/\tau$ . So the true signal is fast cosine fringes (fine structure set by T) modulated by a slow single-pulse envelope (set by τ): the pulses decide which frequencies get excited at all, and the free evolution carves the sharp fringes within that window.

Figure

Ramsey fringes Pₑ versus detuning for T ≈ 0.5 s: a fast cosine of period 1/T = 2 Hz (blue) sitting under the broad single-pulse Rabi envelope (amber dashed, width ≈ 1/τ). The central fringe has full width at half maximum ≈ 1/（2T） = 1 Hz on the 9.19 GHz caesium clock line

The core equations

Everything begins with the exact transition probability for two identical square pulses of duration τ and Rabi frequency Ω separated by a drift time T — the product of two Rabi rotations and a free-precession phase, projected onto the excited state. It carries both the fast fringes and the slow envelope, and every later result is a limit of it.

P_e(\delta) = 4\,\frac{\Omega^2}{\Omega_R^2}\sin^2\!\Big(\frac{\Omega_R\tau}{2}\Big)\left[\cos\!\Big(\frac{\Omega_R\tau}{2}\Big)\cos\!\Big(\frac{\delta T}{2}\Big)-\frac{\delta}{\Omega_R}\sin\!\Big(\frac{\Omega_R\tau}{2}\Big)\sin\!\Big(\frac{\delta T}{2}\Big)\right]^2 \qquad (1)

where: $P_e$ — excited-state probability after the second zone; $\Omega$ — resonant Rabi frequency; $\delta=\omega-\omega_0$ — detuning; $\Omega_R=\sqrt{\Omega^2+\delta^2}$ — generalised Rabi frequency; $\tau$ — pulse duration; $T$ — free-evolution time. The prefactor is the single-pulse Rabi envelope; the squared bracket carries the δT interference. Well inside that envelope — for π/2 pulses ( $\Omega\tau=\pi/2$ ) and $|\delta|\ll\Omega$ — the prefactor is flat and the bracket collapses to a pure cosine, the working form for a clock — a raised cosine in the accumulated phase.

P_e(\delta)=\tfrac{1}{2}\big[\,1+\cos(\delta T+\varphi)\,\big]\qquad(\text{two }\tfrac{\pi}{2}\text{ pulses, }\tau\ll T) \qquad (2)

where: $\delta$ — detuning; $T$ — free-evolution time; $\varphi$ — controllable phase difference between the pulses (zero for identical zones). Scanning δ sweeps the fringe; scanning φ slides it sideways without touching the atom. The central fringe reaches half height when the cosine argument hits ±π/2, i.e. $2\pi\Delta\nu\,T=\pi/2$ on each side; doubling that half-width gives the full width at half maximum in ordinary frequency — the most quoted number in the subject.

\Delta\nu_{\text{Ramsey}} \approx \frac{1}{2T} \qquad (3)

where: $\Delta\nu_{\text{Ramsey}}$ — FWHM of the central fringe (Hz); $T$ — free-evolution time (s). A caesium fountain with $T\approx0.5$ s gives a fringe roughly 1 Hz wide on a 9.19 GHz line, narrowing without bound as T grows — capped only by coherence and oscillator stability. What that resolution buys a clock is captured by the quality factor and the Allan deviation (fractional frequency instability) — a clock is good when its line is narrow (large Q), its signal clean (high S/N), and it averages down with time.

Q=\frac{\nu_0}{\Delta\nu}\approx 2\nu_0 T,\qquad \sigma_y(t)\;\approx\;\frac{1}{\pi Q}\,\frac{1}{S/N}\,\sqrt{\frac{T_c}{t}} \qquad (4)

where: $Q$ — quality factor; $\nu_0$ — clock transition frequency; $\sigma_y(t)$ — Allan deviation at averaging time $t$ ; $S/N$ — signal-to-noise per interrogation; $T_c$ — measurement cycle time. Both a high transition frequency and a long T raise Q — why the field pushed from microwave to optical clocks. The moving atom carries its own clock, and relativity slows it. With the field perpendicular to the beam the first-order Doppler shift cancels, leaving the second-order time-dilation term that pulls the fringe centre downward — the systematic that most rewards cold atoms.

\frac{\Delta\nu_{\text{2nd}}}{\nu_0} = -\frac{\langle v^2\rangle}{2c^2},\qquad \langle v^2\rangle=\frac{3k_BT_{\text{gas}}}{m} \qquad (5)

where: $\Delta\nu_{\text{2nd}}$ — relativistic shift of the fringe centre; $\langle v^2\rangle$ — mean-square speed; $c$ — speed of light; $k_B$ — Boltzmann constant; $T_{\text{gas}}$ — sample temperature; $m$ — atomic mass. A thermal Cs beam sits near $-3\times10^{-13}$ ; cooling to microkelvin drives it toward parts in $10^{18}$ . Finally, no real coherence lasts forever: field noise, collisions and trap inhomogeneity relax the transverse coherence over a time $T_2$ , so the fringe contrast decays as $e^{-T/T_2}$ — a direct ruler of coherence time.

P_e(\delta)=P_0\big[\,1+C\,e^{-T/T_2}\cos(\delta T+\varphi)\,\big] \qquad (6)

where: $C$ — ideal fringe contrast ( $\le 1$ ); $T_2$ — transverse coherence time; $P_0$ — mean excited-state population. Scanning T and watching the contrast fall reads $T_2$ straight off.

Derivation ARamsey fringes from two pi-over-two pulses on the Bloch sphere

The goal is to obtain the ideal fringe $P_e=\tfrac12(1+\cos\delta T)$ from pure geometry, carrying a unit Bloch vector through the four operations — pulse, drift, pulse, readout — and reading its final height. We use only that a resonant π/2 pulse is a 90° rotation about a horizontal axis and free evolution a rotation about the vertical axis at rate δ; reaching a cosine without a single propagator is the point, since the interferometer is just a rigid-body rotation.

P_e(\delta)=\cos^2\!\Big(\frac{\delta T}{2}\Big)=\tfrac{1}{2}\big[1+\cos(\delta T)\big]\qquad(\text{ideal fringes})

Step 1. Start in $|g\rangle$ : the Bloch vector is $\vec{s}=(0,0,-1)$ , the south pole. Why: ground state means $s_z=-1$ under the mapping $P_e=\tfrac12(1+s_z)$ .

Step 2. First π/2 pulse about x: $R_x(\pi/2)$ takes $(0,0,-1)\to(0,-1,0)$ , onto the equator. Why: a 90° rotation about x carries the −z pole to the −y axis, the superposition $\tfrac{1}{\sqrt2}(|g\rangle+|e\rangle)$ .

Step 3. Drift for time T: a detuning δ rotates the vector about z by δT, $(0,-1,0)\to(\sin\delta T,\,-\cos\delta T,\,0)$ . Why: off resonance the atomic phase winds relative to the drive, which is exactly a rotation about the vertical axis — phase stored as azimuth.

Step 4. Second π/2 pulse about x rotates y into z, giving $s_z=\cos\delta T$ . Why: $R_x$ lifts the equatorial $-\cos\delta T$ component onto the pole axis — azimuth becomes height.

Step 5. Read out $P_e=\tfrac12(1+s_z)=\tfrac12(1+\cos\delta T)$ . Why: a cosine of period 2π/T in δ, peaking at δ=0 and averaging to ½ off resonance, matching formula (1) as τ≪T.

Go deeper · Derivation B · 1 minFringe width and why a long free time wins

Derivation BFringe width and why a long free time wins

The goal is to show quantitatively that the central-fringe width is set by the free time T, not the pulse time τ, and to size the advantage over single-zone spectroscopy. We impose the half-maximum condition on the ideal fringe, solve for the detuning at half height, double it, and compare against the Rabi linewidth of a single pulse of the same duration — the comparison that makes the method worth building.

\Delta\nu_{\text{FWHM}}=\frac{1}{2T}\qquad(\text{central fringe})

Step 1. Write the fringe in ordinary frequency, $P_e=\tfrac12[1+\cos(2\pi\Delta\nu\,T)]$ , and demand half height: $\cos(2\pi\Delta\nu\,T)=0$ . Why: the signal falls from 1 to ½ exactly where the cosine crosses zero, so the half-maximum points depend only on the phase argument.

Step 2. Solve on one side: $2\pi\Delta\nu\,T=\pi/2$ gives a half-width at half maximum $\Delta\nu=1/(4T)$ , so the full width is $1/(2T)$ . Why: the fringe is symmetric about δ=0, so the FWHM is twice the one-sided half-width.

Step 3. Compare with a single zone: one pulse of duration τ has a Rabi linewidth $\Delta\nu_{\text{Rabi}}\approx0.8/\tau$ , so Ramsey is narrower by $\approx 1.6\,T/\tau$ . Why: T can far exceed any τ, so the same atoms give a far sharper line — a long uniform field swapped for a long empty gap.

Step 4. Read the metrology: $Q=\nu_0/\Delta\nu\approx2\nu_0T$ grows linearly with T, so a fountain with $T\approx0.5$ s reaches $\sim1$ Hz fringes and $Q\approx10^{10}$ on the 9.19 GHz line. Why: Q feeds straight into the Allan deviation (4).

Go deeper · Derivation C · 1 minSecond-order Doppler shift of the fringe centre

Derivation CSecond-order Doppler shift of the fringe centre

The goal is to derive the relativistic time-dilation shift that pulls the fringe centre off the true atomic frequency, and to size it for a thermal beam so the case for cold atoms is quantitative. A moving atom's proper clock runs slow by the Lorentz factor; we expand to lowest order in v/c, average over the thermal velocity distribution, and read off the fractional shift — the dominant velocity-dependent systematic in beam clocks.

\frac{\Delta\nu}{\nu_0}=-\frac{\langle v^2\rangle}{2c^2}\qquad(\text{time dilation})

Step 1. With the field perpendicular to the beam the linear Doppler term averages away, leaving $\nu_{\text{obs}}=\nu_0\sqrt{1-\beta^2}$ , $\beta=v/c$ . Why: symmetrising the two zones cancels the first-order shift, leaving the quadratic relativistic term.

Step 2. Expand: $\nu_{\text{obs}}\approx\nu_0\big(1-\tfrac{v^2}{2c^2}\big)$ . Why: beam speeds have $\beta\ll1$ , so the $O(\beta^2)$ binomial term suffices.

Step 3. Average over Maxwell–Boltzmann, $\langle v^2\rangle=3k_BT_{\text{gas}}/m$ , giving $-3k_BT_{\text{gas}}/(2mc^2)$ . Why: the fringe records the ensemble, so the shift follows the mean-square speed set by temperature and mass.

Step 4. Estimate: a thermal beam near 250 m/s gives $\langle v^2\rangle\approx6\times10^4\ \text{m}^2\text{s}^{-2}$ and a shift $-3.5\times10^{-13}$ ; cooling to $\sim1\ \mu$ K collapses it by fifteen orders of magnitude. Why: this is why the standard moved from hot beams to fountains — cold atoms lengthen T and suppress the shift at once.

Limiting cases

The ideal two-π/2-pulse Ramsey signal $P_e(\delta)$ in the key regimes, with pulse duration τ, free time T, and Rabi frequency Ω.

Case	Result	Meaning
On resonance, π/2 pulses	$P_e(0)=1$	Two 90° rotations sum to a π rotation; constructive interference drives the atom fully to $\|e\rangle$ .
Central fringe, $T\gg\tau$	$\Delta\nu_{\text{FWHM}}\approx \dfrac{1}{2T}$	Fringe width is set by the free time; longer T gives narrower fringes and larger Q.
Zero free time, $T\to0$	$P_e\to\sin^2(\Omega\tau)$	The pulses merge into one of area $2\Omega\tau$ ; fringes vanish, only the Rabi lineshape remains.
Far detuning, $\|\delta\|\gg\Omega$	$P_e\to 0$	Outside the Rabi envelope neither pulse excites the atom, so no amplitude is left to interfere.
Finite coherence, $T\gtrsim T_2$	contrast $\propto e^{-T/T_2}$	Fringes wash out as T nears $T_2$ ; going past it buys resolution you cannot read.

Common trapThe narrow central fringe sits exactly at the atomic transition frequency

It very nearly does — but "nearly" is where clocks live or die. The sharp central fringe rides under the broad Rabi envelope, and its centre is displaced by every systematic in the budget — the second-order Doppler (time-dilation) shift, the quadratic Zeeman shift, collisional (density) shifts, the AC-Stark light shift from the probe itself, and the cavity or end-to-end phase difference between the zones, which pulls the centre by $\Delta\nu_{\text{cav}}=\varphi_{\text{end-to-end}}/(2\pi T)$ . A naive fit to the peak inherits all of these at once; finding the true centre means modelling the full lineshape and cancelling each shift in turn. The fringe tells you where the dressed transition is; only the systematics budget tells you where the atom is.

Worked examples

1 · Fringe width and Q of a caesium fountain (confidence builder)

A caesium fountain launches laser-cooled atoms up through a microwave cavity and catches them on the way down, giving a free-evolution time $T=0.5$ s on the $\nu_0=9\,192\,631\,770$ Hz hyperfine clock transition; each cavity transit lasts $\tau\approx10$ ms. Find the central-fringe FWHM, the quality factor, and the improvement over a single zone.

Step 1. Central fringe width from equation (3): $\Delta\nu=1/(2T)=1/(2\times0.5)=\mathbf{1\ Hz}$ . Half a second of free flight resolves the 9.19 GHz line to a fringe one hertz wide.

Step 2. Quality factor from equation (4): $Q=\nu_0/\Delta\nu=9\,192\,631\,770/1=\mathbf{9.19\times10^{9}}$ , identical to $2\nu_0T=2(9.19\times10^9)(0.5)$ — the number the Allan deviation divides by.

Step 3. Compare with a single zone. One 10 ms pulse gives a Rabi linewidth $\Delta\nu_{\text{Rabi}}\approx0.8/\tau=0.8/0.010=\mathbf{80\ Hz}$ , so the Ramsey sequence is narrower by $\mathbf{80\times}$ , consistent with $\sim1.6\,T/\tau=80$ — eighty times the resolution from the same atoms, bought by letting them coast for half a second. FWHM = 1 Hz, Q = 9.19×10⁹, an 80× gain over single-zone Rabi.

2 · The second-order Doppler shift, hot beam versus cold fountain

Estimate the fractional time-dilation shift of the fringe centre for (a) a thermal Cs beam with $v_{\text{rms}}=250$ m/s, and (b) a laser-cooled sample at $T_{\text{gas}}=1\ \mu$ K. Take $m_{\text{Cs}}=2.21\times10^{-25}$ kg, $c=3.00\times10^8$ m/s, $k_B=1.38\times10^{-23}$ J/K.

Step 1. Hot beam, equation (5): $\Delta\nu/\nu_0=-v_{\text{rms}}^2/(2c^2)=-(250)^2/[2(3.00\times10^8)^2]=\mathbf{-3.5\times10^{-13}}$ — an absolute pull of about $-3\times10^{-3}$ Hz, small in hertz but a thousandfold larger than the fractional fringe width, so it dominates a hot-beam clock's uncertainty.

Step 2. Cold sample, mean-square speed: $\langle v^2\rangle=3k_BT_{\text{gas}}/m=3(1.38\times10^{-23})(10^{-6})/(2.21\times10^{-25})=\mathbf{1.9\times10^{-4}\ m^2 s^{-2}}$ , i.e. $v_{\text{rms}}\approx1.4$ cm/s.

Step 3. Fractional shift: $\Delta\nu/\nu_0=-1.9\times10^{-4}/(1.8\times10^{17})=\mathbf{-1.0\times10^{-21}}$ — cooling to microkelvin shrinks the time-dilation systematic by nine orders of magnitude, below every other term in the budget. Hot beam ≈ −3.5×10⁻¹³; 1 μK fountain ≈ −1.0×10⁻²¹ — the quantitative case for cold-atom clocks. The same interferometry at optical frequencies pays off enormously: put $\nu_0=4.29\times10^{14}$ Hz (the $^{87}$ Sr line) and $T=1$ s into equation (4) and $Q=8.6\times10^{14}$ , about $9\times10^4$ times the fountain's Q — nearly all of it from the higher frequency, which is why optical clocks reach below $10^{-18}$ .

Physics of the portrait

Ramsey interferometry is named for a single person whose one idea reshaped a whole field and kept paying dividends for seventy years.

Origins · 1 minNorman Ramsey and how this was discovered

Norman Ramsey

1915–2011 · invented the method of separated oscillatory fields; shared the 1989 Nobel Prize in Physics

Norman Ramsey trained under I. I. Rabi at Columbia, where Rabi's molecular-beam magnetic-resonance method — a single oscillating field driving transitions in a beam of atoms — had just opened precision atomic spectroscopy. Rabi's technique had a hard ceiling: narrowing the line meant a longer interaction region, and a long region demanded a field uniform in amplitude and phase over its whole length, which no apparatus could guarantee. In 1949 Ramsey saw the way around it — two short zones separated by a field-free gap, with the atom's coherence coasting between them, so the resolution is set by the drift time and the punishing uniformity requirement evaporates because the empty middle needs no field at all. He published the method in Physical Review in 1950, and it became the backbone of atomic beam spectroscopy: it made the caesium clock practical — Essen and Parry built the first at NPL in 1955 — and underwrote the 1967 redefinition of the SI second. Ramsey went on to co-invent the hydrogen maser and to lead decades of neutron electric-dipole-moment searches, and shared the 1989 Nobel Prize for the method — which today, on optical transitions and ultracold atoms, defines the most accurate measurements ever made.

History

1938

— Rabi demonstrates molecular-beam magnetic resonance with a single oscillating field

1950

— Ramsey publishes the method of separated oscillatory fields in Physical Review 78, 695

1955

— Essen and Parry operate the first caesium-beam frequency standard at NPL

1967

— the 13th CGPM defines the SI second as 9 192 631 770 periods of the caesium transition

1989

— Ramsey shares the Nobel Prize in Physics for separated oscillatory fields

Going further

Texts. Ramsey, N. F., Molecular Beams (Oxford, 1956) · Ramsey, N. F., Phys. Rev. 78, 695 (1950), "A Molecular Beam Resonance Method with Separated Oscillating Fields" · Foot, C. J., Atomic Physics (Oxford, 2005), Ch. 6–7 · Demtröder, W., Laser Spectroscopy (Springer) · Riehle, F., Frequency Standards (Wiley, 2004) · Loudon, R., The Quantum Theory of Light, 3rd ed. · Saleh & Teich, Fundamentals of Photonics.

Provenance. A verifier should confirm: (1) the two-pulse $P_e$ formula (1) and its $\cos^2(\delta T/2)$ short-pulse limit against Ramsey (Phys. Rev. 78, 695, 1950) and Foot Ch. 7; (2) FWHM $=1/(2T)$ from $\cos(2\pi\Delta\nu\,T)=0$ ; (3) the Bloch-sphere sequence reproducing $P_e=\tfrac12(1+\cos\delta T)$ ; (4) the second-order Doppler $-\nu_0\langle v^2\rangle/2c^2$ with $\langle v^2\rangle=3k_BT_{\text{gas}}/m$ (Demtröder; Foot Ch. 6); (5) the Cs hyperfine frequency $9\,192\,631\,770$ Hz as the exact SI value. Numbers recomputed here: $T=0.5$ s → 1 Hz, $Q=9.19\times10^9$ ; beam shift $-3.5\times10^{-13}$ ; 1 μK shift $-1.0\times10^{-21}$ ; Sr $\nu_0=4.29\times10^{14}$ Hz → $Q=8.6\times10^{14}$ . Clock uncertainties ( $10^{-16}$ fountains, below $10^{-18}$ optical) are standard values from Riehle; no precise constants fabricated beyond the SI definition.

Prerequisites. Two-level atom and Rabi oscillations · Bloch sphere and optical Bloch equations · Doppler Cooling and the Doppler Limit · Coherence and interference of amplitudes. Leads to. Bose–Einstein Condensation in Dilute Gases · Evaporative Cooling · Magneto-Optical Trap.

Ask the tutor: Ask me to walk the Bloch vector through the pulse–drift–pulse sequence, or to size the free time and systematics budget for any clock.

Why are the Ramsey fringes narrower than the single-pulse Rabi line if the pulses are identical in both?

Because the two lines are set by different times. A single pulse of duration τ can only define frequency to about $1/\tau$ — its Fourier width. The Ramsey fringe is an interference between two amplitudes that accumulated phase δT over the whole free time T, so a fixed change in $P_e$ needs only a detuning of order $1/T$ . Since T can be hundreds of times τ, the fringe is hundreds of times narrower — the pulses set the broad envelope, but the drift time governs the fine structure inside it.

If the middle region is field-free, how does anything happen to the atom there?

Plenty happens — just no driven transitions. After the first π/2 pulse the atom is in a superposition of $|g\rangle$ and $|e\rangle$ , and because those states have different energies their amplitudes accumulate phase at different rates; in the drive frame that is a steady precession of the Bloch vector at the detuning δ. The free region does no work and adds no light shift — its virtue is that it stores the phase cleanly while the atomic oscillator is compared against the local oscillator.

What is the deliberate phase φ between the two pulses, and why step it?

The two pulses share a phase reference; shift the second by φ and the fringe becomes $\tfrac12[1+\cos(\delta T+\varphi)]$ — the same fringe slid sideways. A servo wants to sit not on the peak (where the slope $\mathrm{d}P_e/\mathrm{d}\delta$ is zero and it cannot tell which way it drifted) but on the steep flank. Stepping φ by ±π/2 places the interrogation on alternating steep sides of the central fringe, the sharpest discriminator. Composite and "hyper-Ramsey" sequences go further, adding pulses with engineered phases to cancel probe-induced light shifts — vital in optical clocks.

Why is a cavity or "end-to-end" phase difference so dangerous for a clock?

If the field the atom sees in the first zone is not exactly in phase with the second, the atom mistakes that fixed offset for accumulated atomic phase, and the centre pulls by $\Delta\nu_{\text{cav}}=\varphi_{\text{end-to-end}}/(2\pi T)$ . In a fountain the two zones are the same cavity crossed twice, nulling much of it by symmetry, and reversing the horizontal launch lets you measure and subtract the residual. The $1/T$ is a small mercy — longer free times dilute a fixed phase error — but it stays a leading systematic, which is why cavity design and beam-reversal are obsessed over.

Does the Ramsey method require a single two-level atom, or does it work with many atoms?

Both — and the distinction matters for the noise. The lineshape here is single-atom, but a clock interrogates $N$ atoms at once to beat down quantum projection noise, improving the signal-to-noise as $\sqrt N$ and hence the Allan deviation (4). The catch is that they must stay coherent over T: mutual collisions cause density-dependent shifts, and any spread in environment — field, trap depth, light shift — smears the fringe. Much of the art of a modern clock is preparing many atoms that all see the same T and environment, so the ensemble fringe stays as sharp as the single-atom one.

How does scanning T instead of δ turn the interferometer into a coherence meter?

Fix the detuning and vary the free time: the fringe amplitude in equation (6) rides an envelope $C\,e^{-T/T_2}$ , so plotting contrast against T reads the coherence time $T_2$ straight off. This is how experimenters characterise decoherence from field noise, collisions, or trap inhomogeneity in cold-atom and qubit platforms. Adding a π pulse in the middle (a spin echo) refocuses the reversible dephasing, so comparing plain-Ramsey with echo decay separates recoverable inhomogeneous dephasing from genuine relaxation.