A. Dubietis 2026-11-05
A. Dubietis coauthored the 1992 paper introducing chirped and stretched pulse parametric amplification in beta barium borate with G. Jonušauskas and A. Piskarskas. Its publisher Abstract reports compressed pulses of and approximately output power.[ref]
The approach connects optical parametric amplifier with the central logic of chirped pulse amplification: amplify a temporally stretched signal, then compensate its spectral phase to recover a short optical pulse. The amplified optical bandwidth matters because a narrow spectrum cannot support an arbitrarily short pulse, even with ideal compression.
For an original Gaussian pulse example, an intensity full width at half maximum and ordinary-frequency intensity width satisfy for a transform-limited pulse. Thus a Gaussian spectrum of width supports a transform-limited intensity duration of about . A measured duration of with that same spectrum would have time-bandwidth product one, exceeding the transform-limited value. Residual spectral phase can account for such broadening, though the product alone does not uniquely reconstruct that phase. The stipulated bandwidth and durations are illustrative, not additional experimental data from the paper.
Airy disk Created 2026-10-14 Updated 2026-10-14
The Airy disk is the central bright region, up to the first dark ring, in the scalar diffraction pattern of a uniformly illuminated, unobscured circular aperture. The whole pattern includes the weaker surrounding rings; its radius is distinct from its full width at half maximum.
For a circular optical pupil of diameter under normal, uniform monochromatic plane wave illumination, Fraunhofer diffraction gives
Here is the on-axis light intensity, is the angle from the optical axis, is the wavelength in the propagation medium, and is the order-one Bessel function of the first kind. The paraxial approximation replaces by , measured in radians. The first positive zero gives
This is the angular radius of the Airy disk; the angular diameter is twice as large. The Airy disk derivation explains the origin of the Bessel function of the first kind and of the numerical coefficient.[ref]
Figure 1.
Ideal circular-aperture intensity: central disk, dark rings and radial profile
. Details.
The image and both radial curves show the same calculated light intensity, divided by its peak. The image's logarithmic scale makes the weak rings visible; its colour is a display choice. The linear curve shows how small the ring peaks are compared with the centre. A dark ring marks a zero of wave amplitude, produced by cancellation across the optical pupil; it is not the edge of all transmitted light. The first Airy disk contains about of the total radiant flux, as calculated in Airy disk normalization and encircled energy.
For an original design example, take an ideal optical lens in air with focal length , circular optical pupil diameter and wavelength . In the focal plane, the radial coordinate gives
The first-zero diameter is . Solving gives , so the diameter measured by full width at half maximum is only
Thus a quoted “spot diameter” is incomplete without its definition. Doubling halves both widths at fixed wavelength and focal length; increasing magnification alone changes the image scale without improving object-space discrimination.
The Airy disk supplies a calculable reference for a point spread function, allowing diffraction to be distinguished from defocus and optical aberrations. Its beauty is that one circular boundary determines an entire sequence of bright and dark rings through a Fourier transform. The ideal formula assumes an unobscured pupil, uniform illumination, ideal optical focus, a single wavelength, and a scalar description with small output angles. A blocked centre, nonuniform pupil transmission or a significant pupil wave phase error changes the response. High numerical aperture may require polarization of an electromagnetic wave and a vector field description. The circular-aperture diffraction laboratory experiment tests the far-field pattern directly; the Rayleigh criterion uses its first zero to define a two-source convention.
Axial resolution Created 2026-10-20 Updated 2026-11-05
Axial resolution describes the ability to distinguish features separated along an imaging system's axis. In optical microscopy, the three-dimensional point spread function is usually longer along that axis than across it, so two layers can overlap even when comparable sideways separations are distinguishable.
A small fluorescent bead moved through focus provides an experimental response curve: record the detected intensity at successive axial positions and measure a stated width, such as its full width at half maximum. The result depends on numerical aperture, wavelength, specimen refractive index and the detection method. Confocal microscopy can reject much of the signal from outside the focal plane, improving optical sectioning.
Chirped pulse amplification energy budget Created 2026-10-18 Updated 2026-11-05
Let be the pulse's power integrated over its transverse section, and its energy. Its peak power is . For a chosen Gaussian power envelope with full width at half maximum ,
The numerical shape factor is approximately . It matters when comparing quoted peak powers; the estimate assumes a shape and a width convention.
If an otherwise unchanged pulse is stretched by a factor without loss, its energy stays fixed and its peak power falls by . For energy gain and compressor energy transmission , ideal restoration of the original shape gives
These relations require a specified shape and preserved spectrum; they do not model amplifier saturation. For the diagram's chosen , seed, , and , the successive peak powers are , , and . The last pulse contains : the compressor increases peak power while losing some energy.
Fluence is time-integrated light intensity, , measured in . For a uniform beam of area , and . At fixed transverse profile, lossless temporal stretching lowers peak light intensity but leaves fluence unchanged. Absorbed energy, heat removal, beam size, coatings and pulse duration still constrain optical components. For nonoverlapping identical pulses at repetition frequency , the average power is ; it is a different quantity from peak power.
Choose equal-amplitude, monochromatic odd components with orders , , and odd . Fix the complex-field convention
Here is a normalized light intensity after averaging the fast carrier oscillations. Setting every and summing a geometric series gives
At the removable singularities take the continuous limit. Maxima repeat every ; the first zeros on either side of a maximum lie at . The zero-to-zero width is , which is not the full width at half maximum.
Choose orders and driving wavelength . Using with the speed of light gives and burst spacing . Solving the exact half-maximum equation gives a full width at half maximum of approximately attoseconds. These are values of the specified five-component model, not measurements from the 2001 experiment.
A deliberately chosen spectral phase translates the whole attosecond pulse train by without changing its shape. Adding makes neighboring phase increments unequal: the same component intensities then produce a different temporal profile. For the choice and radians per squared harmonic order, the adjacent-component delay estimates are
namely , , and for . This finite difference estimates the group delay at the pair's midpoint. It is not a single common arrival time.
Figure 1.
Identical harmonic intensities can encode different pulse shapes and delays
. Original idealized calculation for five monochromatic odd components. The top panels compare equal component intensities and the resulting normalized light intensities. The lower panels compare a chosen delay-dependent two-path signal and the group delays recovered from neighboring spectral phases. All values are stipulated model inputs or calculated outputs; none are experimental traces. Details.
The infinite monochromatic sum is periodic forever and has infinite total energy. A real attosecond pulse train has a finite envelope and finite component linewidths. This calculation isolates coherent addition; it cannot describe an isolated burst or establish its contrast against satellites. As a further check, averaging over one half-period gives for any constant component spectral phases, because the cross terms integrate to zero by Fourier orthogonality. Spectral phase redistributes the light intensity in time while leaving that average fixed.
If phase-randomizing collisions form a Poisson process of mean interval , field correlation survives only when no such collision occurs during the delay. Thus its magnitude decays as . The Wiener-Khinchin theorem gives the normalized Lorentzian line shape
Its full width at half maximum is in angular frequency or in ordinary frequency. A population lifetime is a different time constant: lifetime-limited amplitude decay at half the population rate introduces the familiar factor of two. The definition of the supplied lifetime must therefore accompany any numerical linewidth.
For an isolated fluorescent molecule at displacement from the scanning optical focus, write for relative excitation light intensity and for the fraction of its collected emission passed by a centred detection pinhole of radius . The expected response, apart from a common brightness and collection factor, is
Here is position in the detection-pinhole plane and is the emitted light intensity distribution there conditional on collection, with units of inverse area. This point spread function normalization retains throughput: is a fraction, not a curve rescaled to have peak one. The product expresses excitation followed by detection; it is an intensity/probability product, not multiplication of coherent wave amplitudes from different molecules. [ref]
Assume linear, unsaturated fluorescence, independent emitters whose light intensities add, a stationary specimen, fixed collection efficiency and quantum efficiency, and negligible changes from scattering, absorption or bleaching during a scan. Different excitation and emission wavelengths generally give different excitation and detection responses. For a sufficiently small pinhole, is approximately its area times at the centre; this is the familiar excitation-times-detection point spread function. It is not universally the square of one point spread function. A spatially invariant response further allows the image to be expressed as a convolution with the emitter distribution; depth-dependent optical aberrations can invalidate that approximation.
For an explicitly synthetic axial example, put and choose an on-axis excitation , inspired by a Gaussian beam. Stipulate a circular Gaussian function for the detection-plane light intensity, with standard deviation . In dimensionless coordinates , put . Then
The expression for follows by integrating over a disk using polar coordinates. The assumptions specify a toy model with matched axial scales; its parameters are not a calibration for a real objective optical lens. It omits Airy disk rings, vector diffraction, and unequal excitation and emission wavelengths.
Figure 1.
Synthetic axial response: narrower pinholes reduce throughput and suppress defocus
. Original Gaussian toy model, not measurements. The left panel keeps a common signal scale; the right divides each response by its own peak to compare widths. Here is a radius in units of detection-plane Gaussian standard deviation, not an Airy-unit diameter. The dashed reference is the same scanned excitation with all collected emission accepted, not conventional uniform widefield illumination. Both panels retain nonzero out-of-focus tails. Details.
For a vanishingly small pinhole, the peak-normalized synthetic response tends to , whose full width at half maximum is . With all collected emission accepted it tends to , of width . Meanwhile as the pinhole closes. The narrower normalized curve therefore accompanies a vanishing detected signal; peak normalization must not conceal that loss.
At one depth, scan the illumination spot across and and assign the detector reading during each dwell interval to a pixel. Move the specimen or objective optical lens axially to acquire the next depth. A stack of these optical sections samples three-dimensional structure. With beam scanning, the returning fluorescence commonly retraces the scanning optics: this descanning holds the image of the currently illuminated point at the detection pinhole. With stage scanning, the fixed spot and pinhole remain aligned while the specimen moves. [ref]
An optical section is a depth-weighted image of an intact specimen, with a finite axial response and tails. A physical cut actually separates a slice of material; its thickness and cutting damage are properties of the preparation. Changing optical focus selects another optical section without removing a slice, although illumination can still damage a specimen. Neither a thin physical slice nor a sharply displayed optical section makes the imaging point spread function infinitely narrow. [ref]
Point response and section thickness answer different questions. For a spatially uniform, very thin fluorescent sheet, its response versus axial displacement is proportional to the lateral integral of the three-dimensional confocal response,
Consequently a trace through a single bead, , does not by itself give the sheet's optical-section thickness. State which object and which full width at half maximum are being measured. A smaller axial sampling step gives more samples of the same response, rather than automatically improving diffraction-limited resolution. [ref]
Construct an original illustrative spectrum using ordinary frequency detuning . Write the longitudinal Doppler shift in frequency units as , where . Choose a normal distribution with standard deviation and a weak-probe Lorentzian line shape with half width at half maximum :
Thus . Choose and natural linewidth , so . The Gaussian-only full width at half maximum is ; the averaged profile also includes the finite Lorentzian line shape width.
Add a deliberately phenomenological Lamb dip:
Here is the absorption coefficient normalized to the pump-off line-centre value. The factor is chosen, not inferred from populations or a fit. The dip's assigned full width at half maximum borrows the single-atom power broadening scale purely to illustrate contrast versus width. It is not the preceding overlap integral or a fit to historical spectra. There are no crossover features in this isolated-transition toy spectrum.
For , the subtracted dip has depth and width ; for , depth and width . Its half-depth points are . Measure that width relative to the curved background , not a horizontal line through the total absorption minimum.
Figure 1.
Chosen Doppler envelope and phenomenological Lamb dips
. Original numerical Gaussian velocity average with two explicitly chosen Lorentzian absorption dips. Every numerical setting is illustrative. The right panel magnifies line centre; these curves are not experimental data or fits. Details.
For a chosen cell optical depth , the line-centre pump-off transmission is . At it is . The transmitted probe therefore increases by about percent relative to pump off. Absorption contrast, transmission contrast and spectral width are three different quantities.
For an ideal air-gap Fabry–Pérot etalon at normal incidence, the free spectral range is and the peak full width at half maximum is . To separate a pair of lines spaced by by at least peak widths, require . Putting both lines inside one unambiguous free spectral range additionally requires , so . In terms of their vacuum wavelengths, .
The strict order condition needs a margin if tails from adjacent periodic peaks are to be negligible. If every neighbouring peak in a scanned two-line periodic spectrum must be separated by widths, the stronger condition is . Ordinary resolving power alone does not prevent ambiguous repeated orders. A fixed etalon's transmission as a function of wavelength is a periodic-in-frequency filter; it is not automatically a plot of the incident source's spectral lines.
Finesse 2026-10-14
The finesse of a Fabry–Pérot interferometer is its free spectral range divided by a transmission peak's full width at half maximum. Large finesse means a narrow accepted frequency interval compared with the spacing between repeated orders. Here half maximum means of the ideal unit peak, with zero detector background subtracted; it does not mean halfway between the transmission minimum and maximum.
Write for the coefficient multiplying in Fabry–Pérot transmission from multiple reflections. This coefficient is sometimes called the coefficient of finesse; it is not the finesse . About a resonance , put . Solving gives
and, since one free spectral range advances the wave phase by ,
These are exact for the ideal nondispersive model with constant and linear frequency dependence of . The inverse sine requires , or . At equality, neighbouring peaks merely touch half maximum at their shared minima: the nominal width is one full free spectral range and . Isolated half-maximum peaks require the strict inequality. Below this threshold the transmission never falls to , so this peak-width definition supplies no finesse; at the transmission is flat.
For large reflectivity, the small-angle approximation gives
This is a high-reflectivity approximation, not an identity for every . The original dimensionless examples below have exact full width at half maximum values obtained from the inverse sine formula.
Mirror intensity reflectivity Exact Exact finesseApproximate finesse
Figure 1.
Increasing mirror reflectivity narrows Fabry–Pérot interferometer transmission peaks
. Original stipulated dimensionless model for , with and modulo . All curves peak at unity and repeat after . The second panel enlarges the central resonance; coloured segments at mark each exact full width at half maximum. The inset compares exact finesse with the high-reflectivity approximation. These are calculations, not measured scans. Details.
The beauty of the narrowing is visible directly in the geometric series: large preserves contributions from many round trips, so a small phase difference accumulates across many terms and spoils their addition. For the stipulated vacuum gap, gives an ideal full width at half maximum of about . This instrumental width is not a measured spectral line width. A real scan also depends on the source profile, accepted angles, mirror shape, alignment, and scan stability. The University of Washington handout separately discusses mirror-reflectivity, mirror-flatness, and detector-pinhole contributions; increasing alone cannot remove the others.[ref]
Use the Driven two-level Rabi model in the ordered basis , with Pauli matrices satisfying . Let the Rabi detuning be , in angular frequency, while and are ordinary frequencies in hertz. Assume two identical square pulses, equal drive oscillation phases, constant positive Rabi angular frequency , constant during pulses and flight, and no quantum decoherence or loss. The rotating-wave approximation and isolation of the chosen two-level system are required, just as for the Driven two-level Rabi model. Free flight means no transverse drive; the reference oscillator continues running.
The pulse and free-flight Hamiltonian operators in the rotating frame are
The Pauli matrix multiplication law gives . Thus the exact special unitary group pulse matrix for this effective Hamiltonian operator, and the flight matrix, are
The sequence is . Starting in , matrix multiplication and the Born rule give
This retains the tilted rotation axis and quantum phase accumulated during each finite pulse. At resonance it gives ; with no flight it reduces to the single square-pulse response for duration . Both checks would fail if arbitrary ideal rotations were silently substituted at nonzero Rabi detuning.
For resonant pulse area , and , the central fringe approaches
Its half-height points are , so the central peak's full width at half maximum is in hertz, or in angular frequency. This width assumes an effectively flat pulse envelope across that central peak, one flight time and unit contrast; it is neither a universal finite-pulse width nor a frequency uncertainty. Longer flight lets a smaller frequency difference accumulate the same observable quantum phase.
The two pulses here have the same sign and the plotted outcome is transfer to . The earlier Ramsey interferometry example uses an inverse analysis pulse and reports return to its initial label. Specifying the measured outcome and pulse oscillation phases reconciles the two conventions. In a beam the lengths and speed determine and ; in the Quantronium Ramsey interference experiment electronics set the pulse times for a stationary circuit.
Choose an initially transform-limited pulse with Gaussian power and complex envelope
Here is the standard deviation of the power-weighted time distribution; is its full width at half maximum. Neither is a field-amplitude width. The spectral amplitude is proportional to , so the spectral power has standard deviation . The Gaussian time-bandwidth product in ordinary frequency is , where both widths are power FWHMs.
Multiplication by preserves spectral power and energy. Substituting into the moment identity gives
This is an exact result for a Gaussian spectrum with quadratic spectral phase. It also follows by completing the square in its inverse Fourier transform.
Figure 1.
Phase cancellation recovers a Gaussian pulse, while residual phase and narrowed bandwidth limit it
. Original complex-envelope Fast Fourier transform calculations for a chosen power FWHM, pulse. The stretcher adds of group delay dispersion. Separate output cases show exact cancellation, residual GDD, residual third-order dispersion, and a Gaussian optical gain profile. The finite-gain curve divides out the central-frequency gain and retains the resulting energy change. All curves are calculations, not measured pulse traces. Details.
For these chosen inputs, . The stretched pulse has and root mean square width . Adding recovers ; leaving gives . All phase-only cases retain .
If GDD is cancelled but remains, the pulse is asymmetric. Gaussian spectral moments give
Thus the centroid is and the root mean square duration is . Converting this last value to a FWHM with the Gaussian factor would be wrong: the power envelope is no longer Gaussian. The calculation's integrated energy agrees in time and frequency through Parseval's identity, while its centroid and variance agree with these independent moment formulas.
A real amplifier also changes spectral amplitude. For a simple small-signal example, stipulate an intensity gain profile
This specifies finite gain bandwidth, with no saturation or extra amplifier phase. Multiplying two Gaussian spectra gives
For chosen , the amplified spectrum is narrower by , the shortest Gaussian power FWHM becomes , and the total energy is . Perfect spectral phase correction cannot recreate the suppressed spectral amplitude. Real gain narrowing and saturation require a material-dependent amplifier model; a constant multiplier is only a limited idealization. Gérard Mourou's 2018 lecture discusses gain narrowing as a limit on the shortest amplified pulses.[ref]
Gaussian instrument broadening in third-order pulse correlation is the widening produced when an ideal third-order intensity autocorrelation is convolved with an additional Gaussian instrument response. It is a model of the measurement, not a universal correction for every correlator. The finite width of the optical nonlinear gate is already part of the ideal third-order trace and must not be counted again as an independent response.
Let a Gaussian pulse have intensity standard deviation . Its ideal trace is a Gaussian with . Take the additional normalized impulse response to be
The convolution is
The response preserves integrated trace area while reducing peak height and increasing variance. Since every Gaussian full width at half maximum has the same conversion factor from standard deviation,
For an original example, and give an ideal correlation width about and a measured width about . The peak falls to approximately 0.853 of the ideal peak under this unit-area response. Dividing the measured width only by would overestimate the pulse duration.
The correct Gaussian inversion is when the radicand is positive. Its conditioning worsens near the response floor: the derivatives with respect to and are and . Small width errors can therefore cause large errors in a narrowly resolved pulse. Gaussian pulse-width deconvolution uncertainty must include uncertainty of the characterized response, with covariance if relevant.
The primary correlator used second-harmonic generation, sum-frequency generation, crossed beams and spatial imaging to encode delay. It did not assert that every additional broadening mechanism was a Gaussian convolution.[ref] To apply this model, characterize the response under the same optical alignment, spectral acceptance and detector conditions. Non-Gaussian wings, clipping or unresolved side lobes require a different model; a smaller sampling step cannot by itself remove optical broadening.
Gaussian pulse Created 2026-10-20 Updated 2026-11-05
A Gaussian pulse has a field or intensity envelope specified by a Gaussian function; the distinction must be stated. For the displayed intensity, the temporal full width at half maximum is . A transform-limited pulse with this shape has intensity spectral and temporal widths satisfying . A quadratic spectral phase can lengthen the pulse without changing its spectral intensity.
For an isolated Gaussian pulse, the relation between laser pulse energy and peak power depends on the pulse-width convention. Here is the total incident radiant power, the background is zero, and denotes the power full width at half maximum. Integrating over the entire pulse defines its positive energy .
Pulse energy, peak power and average power resolve different information. Consider a Gaussian pulse with
The chosen pulse duration is the full width at half maximum of radiant power: . It refers to radiant power, rather than the wave amplitude of the electric field. The Gaussian integral gives the relation between Gaussian pulse energy and peak power:
The pulse energy shape factor is . It equals for this power profile; a rectangular pulse of duration has . Knowing only a width and an integrated energy therefore requires a pulse-shape assumption before inferring peak power.
An equivalent form makes the normalization transparent. Write the pulse centre as , and let denote the standard deviation of its energy-normalized pulse profile:
The ratio is a probability density function in the mathematical sense: it is nonnegative, has units , and integrates to one. Here it weights the pulse's energy over time. Its normal distribution profile has mean and variance . This gives without a separate area estimate.
For a stipulated and , the power full width at half maximum is and the peak power is . Confusing the standard deviation with the full width at half maximum would insert the wrong duration into the conversion.
The Gaussian pulse field-envelope width is a second distinct convention. If a fixed spatial field profile gives for a constant , the wave amplitude envelope obeys
Consequently its amplitude full width at half maximum is . Inserting into the formula for a power-width input underestimates the true peak power by a factor . State which measured waveform defines the width.
An electrical photodetector trace can also be broadened by its impulse response. The optical pulse-width inference therefore needs a characterized detector response.[ref]
The Gaussian pulse field-envelope width is the full width at half maximum of the magnitude of a Gaussian electric-field envelope. It differs from the FWHM of radiant power. For a fixed spatial mode with and constant ,
The Gaussian pulse energy and peak power article derives this relation. Power-FWHM is also the convention explicitly used for PULSELENGTH in the primary source documentation.[ref]
For an original example, a power width of corresponds to field-envelope width . At the field-envelope half-maximum points, the power is one quarter of its peak, since . At the power half-maximum points, the field magnitude is instead of its peak. A threshold applied to one measured quantity therefore does not define the other quantity's FWHM.
Figure 1.
Squaring a Gaussian field envelope narrows its full width at half maximum by a factor of square root two.
Details.
The envelope magnitude is distinct from the rapidly oscillating signed electric field. A direct photodetector generally supplies a power-related electrical trace, further modified by its impulse response; it is not a direct recording of the optical carrier's field envelope. Inserting an amplitude width into a power-width peak power formula consequently uses the wrong convention. State the observable and the response model with the reported pulse duration.
The Gaussian pulse time-gate energy fraction is the share of an isolated pulse's laser pulse energy admitted by a finite time window. The result below acts on incident optical radiant power. Let be the power full width at half maximum, and first place the pulse centre and the detection time gate centre at zero.
A detection time gate need not collect the whole pulse. Integrating the same profile only over gives the Gaussian pulse time-gate energy fraction
where is the error function. A gate as wide as the power full width at half maximum collects only , or 76.10% of the pulse energy. The pulse has nonzero Gaussian tails outside that interval. A real readout's impulse response can extend the required integration window further.
Alignment adds a parameter. Let the pulse centre be , the gate centre be , and write the Gaussian gate centring error as . Put . Integrating the normalized Gaussian pulse profile over the gate gives
At the two error function terms reduce to the centred formula above. With and , the captured fraction drops to . The pulse centre then lies on a gate edge. Merely keeping the gate width fixed does not keep its collected energy fixed.
The gate works best when centred. With the centred energy-normalized pulse profile and , differentiation of gives
A Gaussian function is an even function about its centre and decreases with distance from that centre. Thus for , and symmetry gives the opposite sign for . The same argument proves that a centred interval maximizes mass for an even unimodal density; unimodality supplies the required monotonicity.
A centred gate of width captures of the energy. Solving gives for 99% capture in this ideal model. The requirement changes if the centring error is nonzero.
Figure 1.
Original Gaussian pulse gate model: width and timing offset jointly determine captured energy.
Details.
Pulse timing jitter can therefore create variation in gated energy even when every full pulse has the same . A wider gate reduces this loss but also changes admitted background and the chance of including neighbouring pulses. Measure the timing and detector response associated with the chosen gate.
For a detector modeled as a linear time-invariant system, its recorded waveform is a convolution of the optical input with its impulse response. NIST tested avalanche photodiode radiometers with roughly optical pulses, recorded their responses with an oscilloscope, and convolved measured responses with longer Gaussian inputs to estimate corrections. The detector and amplifier were characterized together.[ref] A gate applied to that electrical waveform must use the readout model rather than assuming the optical fraction is unchanged.
Gaussian pulse-width deconvolution uncertainty is the uncertainty in an inferred pulse duration after subtracting a characterized Gaussian instrument contribution. For the Gaussian instrument broadening in third-order pulse correlation model, let be measured correlation full width at half maximum, the additional response width and the inferred pulse width:
These widths must use the same convention. The relation already includes the ideal nonlinear gate; represents an additional response.
First-order propagation uses the sensitivity coefficients
Thus the standard uncertainty obeys
The negative covariance term follows from the subtraction, not a rule that uncertainties should simply be subtracted. The GUM's general law retains covariance and requires considering higher-order terms when nonlinearity is significant.[ref]
For an original example, and give . With independent , the formula gives . If instead the two width estimates have correlation coefficient 0.8 with those same uncertainties, the covariance is and . The assumed relationship between the estimates changes the result.
Figure 1.
Original Gaussian pulse-width inversion and first-order uncertainty curves for a 60 fs response and independent 2 fs width uncertainties
. Details.
Near , both derivatives grow because approaches zero. With , and independent 2 fs uncertainties, the central inferred value is only about while the first-order uncertainty is about . This is an unresolved Gaussian pulse-width estimate under these measurement conditions, rather than a reliably resolved 9 fs pulse.
The quantity being subtracted is , with approximate variance . If plausible width values include , linear symmetric error bars on can be misleading. Report the response assumptions and consider an appropriate nonlinear inference or bound; simply discarding negative radicands would silently change the statistical model. Even a precise inversion remains conditional on the Gaussian response model being physically appropriate.
A spectral doublet comprises two neighbouring spectral lines. Resolving it requires information about the line response, relative strengths and noise, not only precision in reading an angle. Choose two equal angular responses shaped as Gaussian functions with common standard deviation and separation . If is angular displacement from their midpoint, their summed signal is
These are assumed response shapes, not a measured Siegbahn frequency spectrum. By symmetry , and
The midpoint changes from a maximum to a minimum at . For this particular equal-width, equal-height model, two distinct maxima occur when . Each isolated Gaussian function has full width at half maximum ; the two-maxima threshold is therefore about times that width.
Figure 1.
Angular error and the visibility of a chosen spectral doublet
. Original calculated models. Left: magnitude of the angular term in Bragg wavelength error propagation, alongside a separately chosen fractional spacing error of parts per million; the curves are not added. Right: explicitly chosen interplanar spacing , first diffraction order, and lines at and . Equal Gaussian function responses use angular standard deviations and ; dotted curves show the narrower isolated components. None of these numbers is a historical crystal calibration or measured spectrum. Details.
In this chosen model, the two Bragg angles are and . Their separation is times for , giving one maximum, and times for , giving two. Sharpening the response can reveal the doublet without changing either emitted wavelength.
The Rayleigh criterion uses a first-zero convention for specified diffraction patterns. A Gaussian function has no finite zero, so its two-maxima threshold is not the Rayleigh criterion. Neither convention is a universal inference limit: known response shapes and adequate signal-to-noise ratio can support estimating a separation before two visible maxima appear. Unequal strengths, background, finite sampling and uncertainty in the response can make that inference much harder. Measurement resolution, the precision of an isolated line's centre and absolute wavelength accuracy should therefore be reported separately.