An exterior differential ideal on a smooth manifold is a graded ideal $\mathcal{I}$ in the exterior algebra ${\Omega }^{\ast }(M)$ of differential forms that is closed under the exterior derivative:Being an ideal also requires $\alpha \wedge \beta \in \mathcal{I}$ for every $\beta \in {\Omega }^{\ast }(M)$. Thus stability under differentiation is an additional condition, not a consequence of wedge closure.

For example, the ideal generated by $dz$ in ${\mathbb{R}}^3$ is differential because $d(fdz)=df\wedge dz$. Its integral surfaces include the planes $z=\text{constant}$. In contrast, the ideal generated by $\alpha =dz-xdy$ is not differential: $d\alpha =-dx\wedge dy$ and $\alpha \wedge d\alpha =-dz\wedge dx\wedge dy\ne 0$, whereas wedging $\alpha$ with any element of its generated ideal gives zero. The Frobenius theorem interprets this obstruction as failure of the associated tangent-plane distribution to be integrable.

For a small-noise system with Freidlin--Wentzell action $I_T$, its quasipotential from $x$ to $y$ is the free-time minimum actionOnly admissible paths enter the inner infimum. The optimization over duration is essential: a finite-time action and a quasipotential answer different questions. A minimum need not be attained at a finite duration.^[ref]

For $dX=-aXdt+\sqrt{\epsilon }dW$, $a>0$, the action from zero is $\frac{1}{2}{\int }_0^T(\dot{\varphi }+a\varphi {)}^{2}dt$. Its minimum for endpoint $y$ is $a{y}^2/(1-e^{-2aT})$, obtained by minimizing the squared control driving $\dot{\varphi }=-a\varphi +u$. Taking $T\to \infty$ gives $V(0,y)=a{y}^2$. This shows how a free-time cost records the difficulty of leaving a stable equilibrium and why the noise normalization affects the numerical coefficient.

Freeze-etching removes surface ice from a frozen specimen by sublimation under vacuum, often after freeze-fracture microscopy has exposed a new face. Removing that ice can reveal structures initially hidden beneath it. Fracturing opens the specimen mechanically; etching subsequently changes the exposed surface by removing material.^[ref]

A physical preparation holds the specimen cold in a controlled vacuum, permits a chosen amount of ice removal, and then applies the coating appropriate for electron microscopy. The exposed block face can be examined with a scanning electron microscope, or a supported replica can be examined by transmission electron microscopy. Coating and transfer preserve information about a prepared face rather than making the original bulk material transparent.^[ref]

For an own geometrical model, suppose a flat ice surface recedes uniformly at $v=0.5\mathrm{nm}{\mathrm{s}}^{-1}$ while an embedded feature does not sublime. A feature initially $d=10\mathrm{nm}$ below the surface is first exposed atAfter $40\mathrm{s}$ the ice has receded $20\mathrm{nm}$, leaving the feature protruding $10\mathrm{nm}$ in this model. The numbers are stipulated, not a measured etching rate or a recommended preparation setting.

A useful control compares matched specimens after different etching durations while recording specimen temperature, transfer conditions and coating geometry. Progressive exposure would be consistent with ice removal, but a different appearance can also arise from changed preparation or contamination. Condensation deposits material rather than removing it, so contamination must not be mistaken for an etched feature.

The ideal recession calculation assumes a uniform rate and an unchanged embedded feature. Real local composition and heat transfer can violate both assumptions. Interpret image contrast together with etching and coating history rather than assigning every relief feature directly to the initial specimen.

Fragaria vesca, the woodland strawberry, is a small perennial flowering plant in Fragaria and Rosaceae. It is a diploid experimental model, with $2n=14$ chromosomes: seven corresponding pairs. Like other strawberries, it bears small dry achenes on an enlarged fleshy receptacle.^[ref]

Its genome illustrates the difference between a cell's chromosome complement and an assembly's reference representation. A draft published online in 2010 and in the 2011 journal volume combined extensive DNA sequencing with a genetic linkage map to anchor sequence into seven reference pseudochromosomes. Seven reference sequences therefore do not mean that ordinary diploid body cells contain only seven chromosomes.^[ref]

The plant also supports experiments connecting anatomy to molecular activity. A 2013 study compared five fruit tissue types across five developmental stages using RNA sequencing. Keeping achenes and receptacle tissues separate allowed gene expression patterns to be assigned to particular parts of the developing structure.^[ref]

For an illuminating comparison, imagine two fruits with the same average expression of a gene but different proportions of seed-bearing and fleshy tissues. Their whole-fruit averages can agree even when corresponding tissue types differ. The relatively tractable genome and dissectible fruit make Fragaria vesca useful for testing such developmental questions.

A fluorescence calibration curve relates a measured fluorescence signal to known amounts or concentrations of a fluorescent substance under specified conditions. Standards measured with the same fluorometer settings and a matched sample background provide the comparison for an unknown. In a verified linear range, a useful model iswhere $S$ is the measured signal, $b$ the background intercept, $k$ the sensitivity and $c$ the concentration. The units of $k$ depend on the chosen signal and concentration units.

For a synthetic mass concentration example, standards at $0$, $1$ and $2\mathrm{mg}{\mathrm{L}}^{-1}$ give signals $10$, $60$ and $110$ in arbitrary instrument units. The line has $b=10$ and $k=50$ instrument units per $\mathrm{mg}{\mathrm{L}}^{-1}$. An unknown reading of $85$ then givesThese are stipulated numbers illustrating the inference, not reported measurements. Repeated standards and a linear regression allow scatter and departures from the straight-line model to be assessed; three perfectly chosen numbers do not establish a real instrument's linear range.

A fluorescence blank tests the background. If the true intercept rises from $10$ to $20$ while the old line is still used, the inferred concentration is too high by $10/50=0.2\mathrm{mg}{\mathrm{L}}^{-1}$. If only signal uncertainty contributes and $b$ and $k$ are treated as exact, a signal standard uncertainty of $2$ instrument units gives a concentration standard uncertainty of $2/50=0.04\mathrm{mg}{\mathrm{L}}^{-1}$. Actual propagation of uncertainty also includes the fitted slope and intercept, their covariance, and standard preparation.

A calibration applies to its tested conditions. Excitation and emission bands, detector gain, solvent, temperature and sample background can change the response. At larger concentrations, absorption of excitation or emitted light produces an inner filter effect; fluorescence quenching can also prevent proportionality. Bracketing an unknown with standards in the tested range gives a stronger basis for inference than extending a line far beyond the standards.

One physical application is dye-based flow measurement: the Bureau of Reclamation describes fluorometer readings for detecting downstream tracer concentrations. The optical calibration supplies a concentration; a separate mixing and conservation model is then required to infer flow. An accurate calibration cannot compensate for inadequate tracer mixing.^[ref]

A fluorescence blank is a control measurement intended to estimate the optical and instrumental background accompanying a fluorescence measurement without the target fluorescent substance. A useful blank matches the solvent, sample vessel, other relevant constituents and fluorometer settings. It establishes what part of the measured signal can arise without the intended emitter.

For spectra, measure the blank at the same excitation and emission settings and subtract it point by point. HORIBA describes solvent-blank subtraction for removing a solvent Raman scattering contribution from a fluorescence spectrum. The control must represent that contribution under the sample's measurement conditions.^[ref]

For a synthetic scalar example, a sample measurement gives $180$ counts and an equal-duration blank gives $30$ counts. The background-subtracted result is $180-30=150$ counts. If the measurements are independent with ideal Poisson distributions, an estimated variance for the difference is the sum,Subtracting the mean background does not remove its photon shot noise. If the blank is measured for a different duration, scale its count rate appropriately and propagate its uncertainty with the same scaling.

The subtraction can also produce a negative value when two noisy readings are close. For example, $25-30=-5$ counts is a possible difference, not evidence for negative emitted light. Interpreting such a result requires the measurement uncertainty rather than silently treating it as a precise zero.

A fluorescence blank primarily addresses additive background. It does not by itself correct an inner filter effect, fluorescence quenching or a changed detector response. A fluorescence calibration curve still requires standards and unknowns measured under comparable conditions. If adding the target changes the background-producing material, a separately prepared blank may no longer reproduce the sample's background exactly.

A fluorescent tracer is a marker detected through fluorescence whose distribution is used to infer transport, mixing or location. In a fluid experiment, a small amount of fluorescent dye can reveal where an injected portion of water travels. The marker's own motion and chemical interactions determine what it faithfully tracks.

A possible optical arrangement illuminates a sample at an excitation wavelength, rejects most excitation light with an emission filter, and records the remaining fluorescence with a detector. A fluorometer can turn this into a concentration measurement. For fixed optical conditions in a calibrated dilute regime, an illustrative model iswhere $c$ is dye mass concentration. A fluorescence calibration curve establishes the usable range and background. Assuming linearity at every concentration would hide effects of absorption, quenching or detector response.

The same marker can measure a volumetric flow rate. For steady injection at rate $q$ and stock concentration $c_s$ into an upstream flow $Q$ with background concentration $c_0$, complete downstream mixing givesFor $q=0.010\mathrm{L}{\mathrm{s}}^{-1}$, $c_s=1000\mathrm{mg}{\mathrm{L}}^{-1}$, $c_d=2\mathrm{mg}{\mathrm{L}}^{-1}$ and $c_0=0$, the inferred upstream flow is $4.99\mathrm{L}{\mathrm{s}}^{-1}$. The extra $0.01$ litre per second belongs to the injected stream.

The Bureau of Reclamation's manual describes this constant-injection method, requiring known injection rate, adequate downstream mixing and concentration measurement; fluorometers detect dye at levels too faint for visual comparison.^[ref] Adsorption, photobleaching or incomplete mixing can invalidate an ideal conserved-tracer inference. Observing a bright pattern alone does not establish that the dye is a quantitatively faithful measure of every constituent's motion.

A Doppler profile correction factor converts a specified sampled velocity into the area-mean normal velocity needed to calculate volumetric flow rate. For a fixed opening of cross-sectional area $A$ and nonzero sampled trace $u_s(t)$, defineThe definition depends on both the spatial flow profile and what the instrument's sampled trace represents. A centerline trace, spectral mode and fastest detected envelope need not have the same correction. The primary ASE report distinguishes these spectral quantities and requires matched locations for velocity and area measurements.^[ref]

Even if each instantaneous correction is known, the appropriate correction for a whole velocity-time integral is generally not its ordinary time average. Provided the denominator is nonzero,This velocity-weighted Doppler profile correction gives greater influence to intervals carrying more of the sampled velocity-time integral.

In an original two-interval calculation, each interval lasts one second. Let sampled velocities be 10 and 30 cm/s and instantaneous corrections 0.5 and 1.0. The sampled velocity-time integral is 40 cm, while its corrected counterpart is $0.5\times 10+1\times 30=35$ cm. Thus ${\alpha }_{\mathrm{eff}}=0.875$, whereas the ordinary average correction is 0.75. For $A=1\mathrm{c}{\mathrm{m}}^2$, the latter would predict 30 mL instead of the model's 35 mL.

When the sampled velocity is nonnegative, this is a weighted mean and lies between the instantaneous corrections. Reversing fluid flow can create cancellation in the denominator, so that conclusion need not hold. A factor measured in a steady tube also cannot be assigned to a changing cardiac profile without evidence. The calculation explains an averaging requirement; it supplies no universal clinical correction.

A floating-pinion balancing mode is a reduced mechanical motion in which a floating pinion shifts between mating meshes and redistributes their reactions. A one-coordinate model can connect the static balance set by floating-pinion centering stiffness with oscillation, damping and the possibility of losing a contact.

For two compressed linear contacts of stiffness $k$, mismatch $e$, support stiffness $k_s$, effective mass $m$ and viscous damping coefficient $c$, the illustrative equation isThe static position is $x_{\mathrm{eq}}=-2ke/K$. Setting $y=x-x_{\mathrm{eq}}$ givesFor $0<\zeta <1$, the oscillation frequency is ${\omega }_d={\omega }_n\sqrt{1-{\zeta }^2}$ and its amplitude envelope decays as $e^{-ct/(2m)}$. These frequencies have angular-frequency units; ordinary frequency is $f_n={\omega }_n/(2\pi )$.

An original example uses $k=10^{6}N/m$, $k_s=0.5\times 10^{6}N/m$, $m=0.25\mathrm{kg}$ and $c=250\mathrm{N}s/m$. It gives $K=2.5\times 10^{6}N/m$, $f_n\simeq 503.3\mathrm{Hz}$, $\zeta \simeq 0.158$ and an envelope reduced to $e^{-1}$ after $2\mathrm{ms}$. With mismatch $e=2\mu \mathrm{m}$, the equilibrium displacement is $-1.6\mu \mathrm{m}$. These are chosen model parameters, not measured gearbox constants.

A floating-pinion contact-loss check must still enforce the assumptions. For reactions $F_1=k(\delta +e+x)$ and $F_2=k(\delta -e-x)$, both contacts remain compressed only while $|e+x|<\delta$. If the oscillation leaves that interval, the two-contact linear equation must change; extrapolating a sinusoid into tensile tooth forces is inconsistent.

A motivating experiment perturbs the pinion slightly about a controlled loaded state, records its displacement and mesh-force proxies, and compares frequency and decay with the model while checking contact continuity. The primary face-gear report used slow rolling to study torque division and mounting freedom; it is evidence for that tested balancing arrangement, not a measurement of the hypothetical resonant frequency above.^[ref]

A subterm of a first-order term is a term appearing at a node of its syntactic tree, including the whole term. Recursively, a variable or constant has only itself as a subterm; the subterms of $f(t_1,\dots ,t_n)$ are the full term together with every subterm of every $t_i$. A proper subterm excludes the full term.

For $t=f(g(x),g(x))$, the distinct subterms are $t$, $g(x)$ and $x$. The subterm $g(x)$ occurs twice, so a subterm value and a subterm occurrence are different notions. Replacing a particular occurrence requires its position in the tree, not only its text.

The subterm relation organizes structural induction, recursive evaluation and term rewriting. Every proper subterm has smaller syntactic size in the usual finite term grammar, so recursive definitions can terminate by descending through the tree. A substring of printed notation is not automatically a subterm: the tree structure determines which expressions are complete terms.

A sporadic simple group is one of the twenty-six exceptional finite nonabelian simple groups that lie outside the alternating and Lie-type families in the classification of finite simple groups. “Sporadic” describes their place in that classification; it does not mean randomly chosen or lacking internal structure.^[ref]

For example, the Mathieu group $M_{24}$ acts by coordinate permutations preserving the extended binary Golay code on twenty-four positions. Thus an exceptional group can emerge as the exact symmetry of a highly organized error-correcting code. The Monster group is another sporadic example, vastly larger. In contrast, the alternating group $A_5$ is simple but belongs to an infinite family, so is not sporadic.

The Monster group $\mathbb{M}$ is the largest of the twenty-six sporadic simple groups. Its order isIt is a finite nonabelian simple group: it has no nontrivial proper normal subgroups. Size alone does not specify its multiplication; its structure is studied through representations and subgroups.^[ref]

The ATLAS specifies generators $a,b$ in conjugacy classes $2A,3B$, with $ab$ of order $29$. The letters distinguish different conjugacy classes having the same element order; the relations $a^2=b^3=(ab{)}^{29}=1$ alone are not a presentation of the Monster group. This illustrates how computational data about enormous groups can still start from two carefully chosen elements.

The mass-transfer Biot number for drug releasecompares internal diffusive resistance with external mass-transfer resistance in a specified release model. Here $h$ is a receiving-side transfer coefficient, $L$ the chosen matrix transport length and $D$ the internal diffusion coefficient. Its value is dimensionless. The definition requires compatible concentration scales; interfacial partitioning must be included if the phases do not share one.

For the steady planar model in finite mass-transfer boundary in drug release,A large ${\mathrm{Bi}}_m$ means external resistance is small relative to the specified internal resistance. A small ${\mathrm{Bi}}_m$ means receiving-side transport can dominate. The comparison says nothing by itself about whether the bulk concentration $c_b$ is near zero, whether $D$ is constant or whether the chosen steady model applies.

The sink-boundary approximation error in drug release can be quantified in this model. With fixed $c_i-c_b$, let $J_{\infty }=D(c_i-c_b)/L$ be the negligible-external-resistance flux. ThenRequiring this last error to be no more than $5\%$ gives ${\mathrm{Bi}}_m\ge 19$. If the error is instead divided by the actual finite-resistance flux $J$, the ratio is $1/{\mathrm{Bi}}_m$ and the corresponding threshold is $20$. A numerical rule is incomplete without its denominator and model assumptions.

For an original example with $D=10^{-12}{\mathrm{m}}^2/s$, $L=100\mu \mathrm{m}$ and $h=2\times 10^{-7}m/s$, ${\mathrm{Bi}}_m=20$. The external-resistance-free approximation exceeds the model flux by $5\%$ relative to $J$, or about $4.76\%$ relative to $J_{\infty }$.

A release experiment can change receiving flow or membrane transport while keeping the formulation fixed. The FDA’s adaptive-perfusion account specifically contrasts passive dialysis-membrane diffusion with pressure-driven separation.^[ref] Such observations motivate checking external resistance; a fitted release curve alone does not measure $h$, $D$ and $L$ independently.

Salt fingering is double-diffusive convection in which narrow rising and sinking fluid structures develop because heat and dissolved salt diffuse at different rates. Warm, salty water above colder, fresher water can have a stable total mass density stratification: the stabilizing temperature difference outweighs the destabilizing salinity difference.^[ref]

A small upper parcel displaced downward enters colder surroundings. It loses heat relatively quickly while retaining more of its salt, becoming denser and continuing downward. A fresher parcel displaced upward can gain heat while remaining relatively fresh and light. The distinction between thermal diffusivity and solutal diffusivity supplies the feedback; stability of the initial combined mass density profile does not exclude this mechanism. The density ratio records the competing contributions, while the salt-finger instability criterion specifies the parameter regime.

A laboratory analog can replace heat and salt by two dissolved substances. David Osmond describes a 1820-by-80-by-120 millimetre tank with salt and sugar solutions entering at opposite ends and an overflow at the centre. The supplied solutions had matching mass density, but salt diffused faster than sugar. The salt-rich plume could lose salt faster than it gained sugar, producing a lighter core and denser surroundings. Fingering and diffusive layers appeared in different regions, showing that identical source densities need not remain dynamically equivalent when composition diffuses.^[ref]

In this analog, salt plays the faster-diffusing role played by heat in the ocean, while sugar plays the slower-diffusing role of oceanic salt. Keeping that correspondence explicit prevents reversing the physical explanation.

A Raman pump-on pump-off gain ratio compares the probe spectrum with and without the narrowband Raman pump, after correcting detector background and normalizing by a reference. For positive dark-corrected probe values $S$ and reference values $R$,Unity denotes no pump-induced change in this model; $G-1$ is the fractional gain. The primary FSRS experiment explicitly performs dark subtraction, reference normalization and the on/off division.^[ref] These operations define a measured optical gain observable, not a direct count of Raman photons alone.

An original offset example has true probe readings $100000$ with pump off and $102000$ with pump on, with an unchanged reference ratio. The true gain is $1.02$. A common unremoved detector offset of $50000$ changes the raw ratio toReference normalization does not automatically remove an additive detector background. Uncertain subtraction also contributes uncertainty to the corrected ratio.

The shot-noise precision of a Raman gain ratio gives a separate statistical limit. If the four inputs are independent Poisson photoelectron counts with large means, first-order propagation givesFor weak gain and all four means approximately $N$, $u(G)\simeq 2/\sqrt{N}$. At $N=250000$, that is about $0.004$: a one-percent gain is only about2.5times this standard uncertainty. These are chosen photoelectron-count assumptions, not an uncertainty specification for arbitrary digitized detector counts.

Shared laser fluctuations, read noise, uncertain references and non-Raman nonlinear effects require their own covariance and response model. The ratio can reveal a weak band only when the acquisition and background control justify its precision. The parent femtosecond stimulated Raman spectroscopy article explains the pump/probe apparatus and supplies a distinct reference-fluctuation example.

Precision limits of weighting correlated repeats describes what a weighted mean can achieve when repeated observations share an uncertain contribution. Supposewith independent, zero-mean repeat errors ${ϵ}_i$ independent of $D$. An estimate $\hat{S}={\sum }_i{w}_i{X}_i$ must satisfy ${\sum }_i{w}_i=1$ to preserve the unknown signal $S$. The shared contribution then has coefficient one as well:This is a special case of correlated-input measurement uncertainty propagation. The covariance matrix is $\Sigma ={\tau }^{2}11^{\mathsf{T}}+\mathrm{diag}({\sigma }_i^2)$; omitting its off-diagonal terms would incorrectly treat the common uncertainty as independently redrawn.^[ref]

The Cauchy-Schwarz inequality gives a sharp bound:Equality occurs for inverse-variance weighting,For equal repeat variances this reduces to equal weights and $u_{\min }^2={\tau }^2+{\sigma }^2/n$. Allowing negative weights cannot improve this bound within the stated model.

For an original body surface area prediction example, take shared standard uncertainty $\tau =0.050{\mathrm{m}}^2$ and two repeat-noise values ${\sigma }_1=0.040{\mathrm{m}}^2$ and ${\sigma }_2=0.080{\mathrm{m}}^2$. The optimal weights are $0.8$ and $0.2$, giving $u_{\min }=\sqrt{0.0025+0.00128}\approx 0.0615{\mathrm{m}}^2$. Equal weights give approximately $0.0671{\mathrm{m}}^2$. Weighting helps, but it does not erase the shared $0.050{\mathrm{m}}^2$ contribution. These values are stipulated, not clinical performance data.

For a fixed individual, $D$ can represent an unknown constant model discrepancy rather than a fluctuation between readings. No weights summing to one can distinguish that discrepancy from $S$, because both enter every observation identically. Independent reference information or a better anatomical measurement model is required to constrain it. If the residual errors themselves have further covariance, their full matrix replaces the diagonal repeat-noise model; the displayed inverse-variance weights no longer follow automatically.

Chromatographic baseline-area uncertainty quantifies uncertainty in the integrated baseline subtracted when assigning a chromatographic extracted-ion chromatogram peak area. For an extracted-ion chromatogram baseline $b(t)$ over fixed limits $t_1,t_2$, let $B={\int }_{t_1}^{t_2}b(t)dt$. The baseline's uncertainty is shared across time whenever fitted parameters or other common inputs are shared.

For sampled times with integration weights $w_i$ and baseline estimates $b_i$,The weights include the actual sampling intervals and chosen integration rule. This is correlated-input uncertainty propagation, not an assumption that each drawn baseline point is independently uncertain.^[ref]

An original constant-baseline example uses 100 equal intervals of $0.1\min$ and one common fitted baseline with standard uncertainty0.2 signal units. Its integrated uncertainty is $10(0.2)=2$ signal-unit minutes. Treating those 100 appearances of the same fitted quantity as independent would instead give $\sqrt{100}(0.1)(0.2)=0.2$, underestimating uncertainty tenfold. More sampling points do not average away a shared offset.

For total recorded area $P$ and assigned peak area $Q=P-B$,With stipulated $u_P=0.5$, $u_B=0.4$ and correlation coefficient0.6, the result is $u_Q=\sqrt{0.17}\approx 0.412$ area units, compared with 0.640 under independence. The appropriate covariance depends on how the trace and its baseline were estimated; it cannot be selected to obtain a preferred uncertainty.

Changing integration limits or choosing a different baseline form can create additional effects beyond parameter uncertainty within one fixed model. Inspect neighboring peaks and baseline regions, and record the fitted model and boundary choices. A precisely computed extracted-ion chromatogram peak area can still depend on an unsuitable baseline model.

The exterior cube of a vector space $V$ over a field is its third exterior power ${\Lambda }^{3}V$. It is generated by products $u\wedge v\wedge w$, multilinear in the three vectors and zero whenever two inputs coincide. Every alternating trilinear map out of $V^3$ factors uniquely through a linear map from ${\Lambda }^{3}V$.

If $e_1,\dots ,e_n$ is a basis, the products $e_i\wedge e_j\wedge e_k$ for $i<j<k$ form a basis, so $\dim {\Lambda }^{3}V=\left(\genfrac{}{}{0ex}{}{n}{3}\right)$. In particular it vanishes for $n<3$ and is one-dimensional for $n=3$.

In an oriented three-dimensional coordinate space,Thus the exterior product packages oriented volume; swapping two vectors reverses its sign. The alternating condition, including vanishing repeated inputs, remains the definition even in characteristic two, where minus and plus signs agree.

For every real number $a>0$ and nonnegative integer $m$, $x^m{e}^{-ax}\to 0$ as $x\to \infty$. Thus every fixed polynomial grows more slowly in magnitude than a positive-rate exponential function, regardless of the polynomial's coefficients.

A short proof comes from the power series of the exponential function. For $x>0$, $e^{ax}\ge (ax{)}^{m+1}/(m+1)!$, soApply this estimate term by term to a polynomial. For example, $x^{100}/e^x$ eventually tends to zero even though it can be extremely large at moderately sized $x$. This is an asymptotic statement, not a claim that the exponential function is larger at every positive input. It explains why exponential tails have finite moments of every fixed order and why the proof that the exponential function is a transcendental function works.

An etalon delay inferred from free spectral range estimates an etalon round-trip delay from the frequency spacing of adjacent transmission resonances. When round-trip group delay varies little across that spacing,The measured quantity is resonance spacing, not the width of one resonance. The manufacturer tutorial explicitly distinguishes free spectral range from optical finesse, which compares that spacing with an individual peak's width.^[ref]

The relation follows from wave phase. Let $\varphi (\nu )$ be the total etalon round-trip phase, including relevant reflection phases. Adjacent resonances satisfy $\varphi ({\nu }_{m+1})-\varphi ({\nu }_m)=2\pi$. Defining ${\tau }_{\mathrm{rt}}(\nu )=(2\pi {)}^{-1}d\varphi /d\nu$ gives the exact interval relationThus the reciprocal spacing is the delay averaged over that frequency interval. Interpreting it as a local pulse-envelope delay requires sufficiently weak variation across the interval and relevant pulse spectrum.

For a uniform plate at normal incidence with negligible frequency-dependent reflection phase, $\varphi =4\pi \nu n(\nu )L/c$. Differentiation gives ${\tau }_{\mathrm{rt}}=2n_{g}L/c$, where $n_g=n+\nu dn/d\nu$ is the group refractive index. Treating $n$ as frequency independent recovers the manufacturer tutorial's elementary plate formula; using a phase index in a strongly dispersive material can instead give the wrong delay.

An original measured-spacing example takes $\Delta {\nu }_{\mathrm{FSR}}=25.0\mathrm{GHz}$. The inferred delay is $40.0\mathrm{ps}$. A spacing standard uncertainty of $0.2\mathrm{GHz}$ contributes approximately $0.32\mathrm{ps}$ to the delay uncertainty through $u(\tau )=u(\Delta \nu )/(\Delta \nu {)}^2$. With stipulated $n_g=1.50$ and negligible coating delay, the corresponding plate thickness is about $4.00\mathrm{mm}$.

In a spectral experiment, identify adjacent peaks of the same resonance family and calibrate the frequency coordinate. If the coordinate is wavelength, use $\Delta \nu =c|1/{\lambda }_1-1/{\lambda }_2|$ rather than treating a wavelength interval as a frequency interval. Beam angle, dispersion and reflection phase belong in the propagation model.

A time-domain check looks for replicas separated by the predicted extra round trip. The historical single-shot correlator experiment inserted a known-thickness BK7 etalon to calibrate delay and separately placed an uncoated etalon in the input to create a double-reflection postpulse.^[ref] The single-pass delay caused by inserting glass into an arm is distinct from the additional round-trip separation of successive transmitted replicas.