Ramsey frequency precision Created 2026-10-18 Updated 2026-10-25
The full width at half maximum of a fringe, the lifetime of quantum coherence, and the precision of an estimated frequency are different quantities. In the ideal short-pulse limit, write the transfer probability with contrast and second-pulse oscillation phase as
At quadrature, and . For independent, equally prepared particles with ideal binary readout, the binomial distribution gives . Local propagation of uncertainty therefore yields the displayed . For chosen , and , it is approximately , although the ideal fringe width is . This is an original counting-noise estimate at a known quadrature, not the accuracy of a historical clock or a guarantee of finding the correct fringe.
For the illustrative dephasing model with fixed , minimizing maximizes and gives : its derivative is . This conclusion changes if longer interrogation reduces the number of accepted particles or the repetition rate. Quantum decoherence, a distribution of flight times, reference oscillator noise and readout errors can reduce contrast; a fitted decay time need not identify one microscopic mechanism. More counting can improve precision while a reproducible cavity quantum phase offset still biases the frequency, as in atomic clock frequency shifts.
Ramsey fringe envelope 2026-10-18
The exact finite-pulse Ramsey transition probability separates a slowly varying pulse response from rapid quantum interference. Define the single-zone transfer probability and a pulse-induced quantum phase by
Since , factorizing the real bracket in the original formula gives
Where , the probability is zero and need not be defined. Elsewhere this identity proves the envelope bound without a weak-drive assumption. The envelope is not the single-zone response : two equal coherent interactions have different transfer weights. A contiguous resonant Rabi pi pulse of duration provides another useful comparison with the same total drive duration, but no dark interval. Its line is broad because it has no long flight phase.
Figure 1.
Finite pulses set the envelope while dark flight sets the fine fringes
. Original exact two-level system predictions for , and , matching the existing velocity-averaged Ramsey line reference particle. The envelope is a bound, not a measured curve or the single-zone line. The central panel changes only the second pulse's oscillation phase. No velocity average or experimental data are plotted. Details.
Ramsey interferometry 2026-10-18
Ramsey interferometry uses two coherent pulses with a free interval between them: the first creates a quantum superposition, the interval accumulates relative quantum phase, and the second converts that quantum phase into measurement probabilities. Its attraction is that an ordinary population readout can reveal the quantum coherence hidden from an energy measurement during the free interval.
Here is an ideal qubit model with an explicit pulse convention. Let , prepare , apply , evolve freely with , and apply the inverse pulse before a computational-basis measurement. The first pulse gives . Direct matrix multiplication gives
Thus the Born rule gives and . Before the second pulse, the same computational-basis measurement would give for both outcomes for every . At , the second pulse changes these probabilities to and . If the two pulses have the same sign instead, the output labels interchange: the pulse convention is part of the prediction.
Figure 1.
An ideal Ramsey sequence turns relative phase into population
. Original pulse-sequence schematic and exact probabilities for . Dashed lines show a separate readout with the second pulse omitted. Instantaneous perfect pulses and noiseless free evolution are idealizations, not experimental measurements. Details.
In a pulse experiment the phase is measured relative to the oscillator supplying the coherent pulses. In a rotating frame, write the free Hamiltonian operator as with signed angular frequency difference in the chosen labeling convention. A free interval then accumulates ; changing the second pulse's phase changes the analysis direction. This comparison makes Ramsey interferometry useful for measuring frequency differences and diagnosing loss of quantum coherence.
For example, suppose the first pulse has prepared a density matrix with equal populations and transverse Bloch vector length :
The same inverse analysis pulse yields . The pure state case gives the full ideal fringe; the mixed state gives the flat result . This is a controlled mathematical comparison, not a fit to an experiment. Experimental contrast alone does not identify quantum decoherence: imperfect preparation, pulse errors and readout errors can also reduce it.
A moving chemical atom or molecule can experience the two pulses of Ramsey interferometry by flying through two short oscillatory-field regions. Their common oscillator supplies the reference quantum phase; the flight between them supplies the interrogation time. The long interval need not contain a driving field. This turns a spatial apparatus into a coherent pulse sequence and makes narrow spectral lines possible without filling the entire flight path with a uniform oscillatory field.
Norman Foster Ramsey, Jr.'s 1950 paper calculates the two-state transition probability and averages it over molecular velocities. It is a theoretical investigation, rather than a report of an atomic clock's measured accuracy.[ref]
His 1989 lecture contrasts an electric-resonance apparatus photograph with a magnetic-resonance schematic. Coherent drive regions bracket a static-field flight region; state-dependent deflection converts transitions into detector changes, while differing flight times suppress side fringes.[ref]
The readout infers a change of internal quantum state through state-dependent transmission, as in the Rabi molecular beam magnetic resonance experiment. A useful control is to block one drive region: a narrow fringe requiring both interactions must disappear. Changing the second region's oscillation phase should move the fringe, whereas an incoherent background need not move. These are consequences of the model below, not additional historical measurements. A detector dip alone neither proves perfect preparation nor determines all apparatus-induced frequency shifts; cesium-beam clock procedure adds the controls needed for a practical atomic clock.
The publication sequence begins with Norman Foster Ramsey, Jr.'s 1949 article A New Molecular Beam Resonance Method, Physical Review 76, p. 996, published 1 October; the fuller theoretical paper followed in 1950.[ref]
NIST-7 supplied a practical realization: optical pumping moved caesium populations into the manifold, without isolating . Two separated microwave interactions probed the clock transition; fluorescence detected transferred atoms. A synthesized reference, tied to a hydrogen maser, sampled opposite sides of resonance for feedback control. Beam reversal tested cavity quantum phase bias; independent velocity determinations checked motion corrections. The observed central fringe resolved the transition, but power-dependent velocity weighting and light shifts still required evaluation.[ref]
Ramsey spectroscopy 2026-10-18
Ramsey spectroscopy measures a transition frequency by comparing the relative quantum phase acquired between two coherent interactions with the reference oscillator. In a beam, a state selector and apertures admit particular internal quantum states and trajectories; an analyzer and detector measure state-dependent transmission. In an optically prepared atomic clock, optical pumping and fluorescence can perform those functions. The preparation and readout determine which part of the beam contributes; they are not themselves the two coherent interactions.
For an ideal accepted particle, the first resonant half-duration Rabi pi pulse splits the probability amplitude between and . The second combines two indistinguishable alternatives: a transition in the first region followed by no transition in the second, or the reverse. Their relative quantum phase depends on the dark flight and the drive oscillation phases. Adding probability amplitudes gives quantum interference; adding only the two alternatives' probabilities would erase the fringe. “Dark” means that the transverse drive is absent: a static magnetic field and the internal Hamiltonian operator can still act.
Figure 1.
Two spatial interaction regions become two timed pulses for each accepted particle
. Original functional Ramsey spectroscopy model, not a reconstruction of NIST-7 or a historical apparatus. Equal zone lengths and edge-to-edge gap give and for constant speed . Preparation and state-sensitive readout are abstract functions; the diagram does not claim perfectly pure preparation. Details.
The useful separation is between pulse duration and phase-accumulation time. Short regions can share one coherent oscillator while a long drive-free interval resolves small frequency differences. A scan over reference frequency locates a fringe pattern; controlled changes of the second pulse's oscillation phase distinguish that coherent signal from phase-insensitive background. A reference scan must also identify the central fringe before atomic clock frequency lock, because neighboring fringes provide competing lock points.
Random-cluster inverse correlation length Created 2026-10-14 Updated 2026-11-05
Take and for the infinite-volume wired random-cluster measure on the cubic lattice. Positive association of random variables and translation invariance give for axis connection probabilities. The Fekete lemma applied to proves the displayed limit of a sequence and the bound . The lower single-edge conditional bound gives , so the decay rate is finite and nonnegative. It may be zero.
At , every distinct-vertex connection probability is zero, so the inverse correlation length is under the convention . At , every connection probability is one and the rate is zero. Thus finiteness of the displayed rate requires the positive- hypothesis.
Random-cluster model Created 2026-09-24 Updated 2026-11-05
On a finite graph , the random-cluster model weights an open-edge configuration by
The partition function is the sum of these weights, and counts the open connected components of a graph, including isolated graph vertices. The endpoint laws at and are obtained by limits. The factor is the cluster weight in the random-cluster model.[ref]
At , the component factor disappears and the law is independent bond percolation. For other , is a model parameter rather than necessarily the marginal opening probability of each edge. For a single edge joining two vertices, with , the closed and open weights are and , so the edge is open with probability .
The random-cluster single-edge conditional probability describes the local mechanism. The Edwards-Sokal coupling connects integer cluster weights with the Potts model, while random-cluster boundary conditions specify how boundary vertices are identified before component counting.
Random-cluster single-edge conditional probability Created 2026-10-06 Updated 2026-11-05
For an edge with endpoints already connected without that edge, including any boundary wiring, the conditional open probability is . Otherwise opening the edge merges two components, so the conditional probability is . For , this conditional probability increases when more other edges are open or more boundary vertices are identified. This is the local mechanism of boundary monotonicity of the random-cluster measure.
Random cuts of a cycle 2026-10-25
Choose distinct edges uniformly from an -vertex cycle graph, with , and delete them. Two specified vertices at oriented distance remain connected exactly when one of their two complementary edge-disjoint paths is uncut. All cuts must then lie in the other path, giving the displayed probability. The two events are disjoint because at least one edge is cut. Set when .
Summing the connectivity indicators for all other vertices gives an expected number
of other vertices in a tagged connected component. The sum identity follows by classifying -element subsets according to their greatest element. For three cuts this is other vertices.
Let be a nonzero multivariate polynomial over a field , with total degree at most . Choose a nonempty finite set . For independent random variables , each with a uniform distribution on a finite set on , the Schwartz-Zippel lemma gives
Nonzeroness is essential; the zero polynomial always evaluates to zero. The total degree sums exponents within each monomial, rather than taking their maximum: has total degree five. The field need not be finite, but must be finite. The estimate reduces error only when .[ref]
The proof begins with the root bound for a polynomial. For several variables, write , with . The total degree of is at most . By induction, vanishes with probability at most . Otherwise the remaining univariate polynomial has at most polynomial roots. Adding the two contributions gives .[ref]
The algorithm samples a point, performs exact polynomial evaluation, rejects the identity if the value is nonzero, and otherwise accepts. The lemma bounds false acceptance for every fixed nonzero input. For the polynomial identity testing over a finite field example, has and , so its exact error attains the bound. Numerical near-zero values from floating-point arithmetic do not satisfy this exact algebraic contract. A polynomial-time algorithm claim additionally needs efficient sampling and evaluation, including the bit cost of representing and operating on field elements. Counting field operations alone does not establish that cost.
A coordinate descent method that randomly selects a block and updates only its coordinates. Suppose each coordinate is selected with probability , every block has Lipschitz constant at most , and the objective function is a strongly convex function with modulus . The block descent lemma and the strong convexity gradient inequality give
For uniform blocks of size in coordinates, updates constitute an optimization epoch, and the epoch complexity is . Here is the initial objective function gap, and the target is its expected gap at most . A coordinate-based common bound is ; actual constants can improve it.
Target: gradient-descent.bigb, under Gradient descent.
Randomized controlled trial Created 2026-09-24 Updated 2026-11-05
A randomized controlled trial assigns experimental units to competing treatments by a specified random mechanism and compares their outcomes. Under complete randomization of units, exactly receive assignment one and assignment zero. With fixed potential outcomes, each unit has assignment probability . Therefore
This proves design-unbiasedness of the mean difference for the finite-population average treatment effect, assuming consistency in causal inference and no interference in causal inference.
Randomization controls the assignment distribution; it does not force every realized baseline covariate difference to be zero. Equal assignment probabilities yield unconditional exchangeability in the design, whereas covariate-dependent randomization can require conditional exchangeability and an analysis respecting that design.[ref]
When assigned and received treatments differ, an intention-to-treat analysis estimates an assignment-policy effect. A treatment-receipt effect needs additional assumptions. Allocation concealment, blinded outcome assessment and handling of missing follow-up address other sources of distortion beyond random assignment itself.
Randomized minimization in clinical trials Created 2026-10-14 Updated 2026-11-05
For each new eligible participant, compare the imbalance that would result from each treatment assignment across prespecified baseline covariate margins. Assign with a probability favoring the less imbalanced choice, while retaining a random component and allocation concealment. This targets marginal balance when a large number of joint strata makes stratified randomization with blocks inefficient. It is a sequential allocation rule, and inference should respect its design.
Randomized play-the-winner rule Created 2026-10-05 Updated 2026-11-05
In a two-treatment urn design, begin with balls of each type, draw and replace a ball to allocate treatment, then add balls of the same type after success or of the opposite type after failure. This response-adaptive randomization favours treatments with more observed successes. For RPW, type-0 probability after counts is .
Randomized response Created 2026-10-31 Updated 2026-11-05
Randomized response masks which of two complementary binary questions a person answered. Suppose the sensitive property has probability , and independently the direct question is selected with known probability , while the complementary question is selected otherwise. Truthful answers then have yes probability . The investigator observes only the answer, preserving uncertainty about an individual's status while permitting population statistical estimation when .
For independent responses, , and the affine estimator is an unbiased estimator. Its variance is , by the mean and variance of a binomial distribution. Equivalently the variance is , exposing the extra sampling uncertainty caused by the masking. At , all produce and identifiability fails. At or , there is no masking and the extra variance vanishes.
Because , maximum likelihood estimation gives . This follows by maximizing the binomial likelihood over the interval and then inverting its monotone affine parametrization. Clipping can introduce bias: for , and , the clipped estimate has mean .
A randomized statistical test specifies a measurable rejection probability for each possible data value. It can be implemented by rejecting when an independent uniform random variable on satisfies . A deterministic test has only the values zero and one. Under a simple hypothesis , its rejection probability is , which is its test size for and its statistical power for .
The Neyman-Pearson lemma permits randomization on an equality set of the form to reach a prescribed test size. Continuous observations do not rule out an equality set of positive probability: if both hypotheses give the same probability density function, their likelihood ratio is identically one. This explains why a precise statement of the lemma includes boundary randomization even for continuous data.
Let have the uniform distribution on and place a point at every , . This is a simple point process, locally finite, and a stationary point process: a translation only changes the phase modulo one, preserving its uniform distribution. For every Borel set , the Tonelli theorem gives
Its consecutive points are exactly one unit apart, despite having the same intensity measure as a unit-rate Poisson point process. This is the one-dimensional translated-grid construction.[ref]
Every half-open interval of length one contains exactly one point. For and , , each count has the Bernoulli distribution with parameter , and
More generally, an interval of length , with a nonnegative integer and , contains or points with probabilities and . Its count variance is , while a unit-rate Poisson point process has count variance .
Random path in a binary tree Created 2026-09-24 Updated 2026-11-05
In a rooted full binary tree, suppose each internal vertex selects either child with Conditional probability . A specified leaf at depth is reached with probability : reaching it requires its particular sequence of choices, and the chain rule multiplies their conditional probabilities.
The resulting leaves need not be uniformly distributed. A tree whose root has a leaf on one side and a two-leaf subtree on the other gives leaf probabilities . For a finite full binary tree these probabilities sum to one, giving . For an infinite tree the sum over finite leaves can be smaller than one, because a path can continue forever. If a vertex has only one available child, the fair-two-child formula needs an explicit convention; a forced transition contributes probability one instead of one half.
Random quadratic with uniform coefficients Created 2026-10-07 Updated 2026-11-05
For independent uniform on , real roots occur on , a region of probability . On that region both roots are nonpositive. Both lie in exactly when . Its area is , so the conditional probability of both absolute values being at most one is . Joint probability density integration turns algebraic discriminant and root constraints into simple planar areas.
Random-restart optimization Created 2026-10-20 Updated 2026-11-05
Random-restart optimization repeatedly runs a local optimization algorithm from independently sampled starting points and retains the best result. It exploits multiple basins of attraction without requiring the local solver itself to escape a local extremum.
If each restart has probability of entering a desired basin, independent restarts miss it with probability after attempts. This formula quantifies the benefit only for the specified starting distribution and basin; neither nor a finite run certifies a global extremum. Recording the starting distribution and stopping rule makes computational comparisons reproducible.