Raman spectroscopy

Raman spectroscopy measures shifted scattering of light to infer internal energy spacings and the molecular response that makes them visible. In a modern illustrative arrangement, narrowband laser illumination replaces the historical arc, an optical filter rejects the strong excitation line, and a spectrograph resolves the weak shifted spectral lines. The illumination, rejection and spectral dispersion are separate jobs. Increasing rejection can conceal small Raman shifts close to the excitation line.

A proposed measurement should record the excitation wavelength, a dark signal, a blank container and the liquid, then compare calibrated line positions and integrated signals under the same collection conditions. Measure optical filter transmission and detector response before using the Thermal Raman line ratio for temperature. Repeat at lower incident power to look for heating or altered sample response. These are suggested checks, not an assertion that the 1928 apparatus implemented modern calibration procedures.

Its appeal is the conversion of internal molecular energy differences into visible spectral lines: changing the excitation can move the whole pattern while preserving characteristic Raman shifts. Combined with vibrational infrared activity, Raman vibrational activity supplies complementary information about molecular vibration. A line pattern can constrain a material's identity or environment, but overlapping bands, fluorescence backgrounds and the finite spectral resolution element prevent an unqualified identification from one peak alone.

A Raman spectrum records the strength of Raman scattering as a function of scattered optical frequency or, usually, Raman shift relative to the excitation. For vacuum wavelengths in nanometres, a common wavenumber convention isHere the numerical wavelengths are in nanometres. Stokes Raman scattering has positive shift under this convention; Anti-Stokes Raman scattering has negative shift. The strong unshifted Rayleigh scattering line is not a vibrational Raman band.

For an original chosen excitation ${\lambda }_0=500\mathrm{nm}$ and a $1500\mathrm{c}{\mathrm{m}}^{-1}$ vibration, the Stokes wavenumber is $20000-1500=18500\mathrm{c}{\mathrm{m}}^{-1}$, giving ${\lambda }_s\approx 540.54\mathrm{nm}$. The corresponding anti-Stokes wavelength is $10^7/21500\approx 465.12\mathrm{nm}$. Changing the excitation changes these optical wavelengths while the same assumed transition retains its Raman shift. The One-mode Raman spectrum example gives an existing level diagram and a separate numerical spectrum.

The vertical quantity must be stated too: detector counts per spectral bin, calibrated scattered power density and a stimulated gain ratio are different observables. Femtosecond stimulated Raman spectroscopy can plot the Raman pump-on pump-off gain ratio rather than spontaneous photon counts. Its primary apparatus disperses probe and reference beams onto an array and calibrates shifts with known solvent lines.^[ref]

Spectral-density conversion between wavelength and Raman shift also changes the density's units: rebadging the horizontal axis alone need not preserve a spectral integral. Background, excitation rejection, instrument response and spectral resolution affect observed band heights and widths. An integrated band area is not determined by peak height alone. A chemical inference should compare calibrated positions and responses under specified conditions, rather than identify a material from one unqualified peak.

Spectral-density conversion between wavelength and Raman shift preserves the integrated physical quantity while changing its horizontal coordinate. Fix the excitation wavelength ${\lambda }_0$, and let $s$ be the Raman shift. Using numerical vacuum wavelengths in nanometres and shifts in $\mathrm{c}{\mathrm{m}}^{-1}$,Let $D_{\lambda }$ denote detected counts per nanometre and $D_s$ counts per unit Raman shift. Conservation of counts under a change of variables givesThe conversion factor carries the corresponding density-unit change. The same argument applies to calibrated power densities, provided both sides describe the same quantity.

For an original narrow bin centred at $600\mathrm{nm}$, a count density of $1200$ per nanometre becomes $43.2$ per $\mathrm{c}{\mathrm{m}}^{-1}$. A one-nanometre bin there spans approximately $27.78\mathrm{c}{\mathrm{m}}^{-1}$, so its integrated count remains about $1200$. At $800\mathrm{nm}$, the same wavelength density becomes $76.8$ per $\mathrm{c}{\mathrm{m}}^{-1}$. Relabeling bin positions without adjusting the density would conceal this coordinate dependence.

For finite bins, use their exact transformed edges and preserve each bin's integral. Raman spectral rebinning with conserved integrated counts distinguishes this operation from interpolating peak heights onto a new grid. Instrument response correction is another step; the Jacobian alone does not calibrate detector sensitivity or remove background.

The primary FSRS apparatus calibrated Raman shifts using known solvent lines and dispersed the probe spectrum onto an array.^[ref] Its dimensionless Raman pump-on pump-off gain ratio is a different observable: a ratio of matched spectral densities can cancel their common coordinate factor. It should not be treated as a count density merely because it is plotted against Raman shift.

Raman spectral rebinning with conserved integrated counts redistributes finite-bin measurements onto a new spectral grid while preserving their summed quantity over the same covered range. It is different from relabeling bin centres or interpolating peak heights. The parent spectral-density conversion between wavelength and Raman shift derives the continuous Jacobian; finite bins require a stated assumption about unresolved variation inside each bin.

For an original piecewise-constant wavelength-density model, let old bin $I_i$ have width $\Delta {\lambda }_i$ and count $N_i$. Invert the new Raman-shift edges to wavelength intervals $J_j$. Define the overlap matrixWhen the new bins partition the old covered range, ${\sum }_j{A}_{ji}=1$, hence ${\sum }_j{N}_j^{\prime }={\sum }_i{N}_i$. Clipping the new range instead excludes the corresponding counts and must be reported.

For original old wavelength bins $[600,602]$ and $[602,604]\mathrm{nm}$ with counts $100$ and $300$, choose new edges corresponding to wavelengths $600$, $601$, $603$ and $604\mathrm{nm}$. ThenThe total remains $400$. A different within-bin model can distribute that total differently; the original measurements do not resolve every sub-bin feature.

Covariance of count-conserving rebinned Raman channels transforms as $C^{\prime }=AC{A}^T$ for this fixed linear operation. Fractional rebinned estimates are therefore not automatically independent Poisson photon counts. Splitting one measured bin among neighbours creates shared uncertainty; finer output pixels do not create additional spectral resolution.

The primary FSRS instrument records a dispersed spectrum on an array and calibrates its Raman-shift axis.^[ref] The overlap algorithm above is an original illustrative reconstruction, not a claim about that instrument's published processing. Apply radiometric response and background corrections consistently with the quantity whose integral is being conserved.

The covariance of count-conserving rebinned Raman channels records the shared noise created when one measured channel contributes to several output channels. For a fixed deterministic rebinning matrix $A$ and an input count vector $\mathbf{N}$ with covariance matrix $C$, the output is ${\mathbf{N}}^{\prime }=A\mathbf{N}$ andThis identity needs finite second moments, not a Gaussian approximation. Raman spectral rebinning with conserved integrated counts describes the count-preserving map; retaining its full covariance matrix is necessary for uncertainties of sums and fitted spectral features.^[ref]

Consider independent input counts with means $100$ and $300$ and Poisson distribution variances equal to those means. Split them deterministically usingThe three output means are $50,200,150$. Their total has variance $1^T{C}^{\prime }1=400$, exactly the original total variance, and standard deviation $20$. Keeping only diagonal entries would give variance $200$ and standard deviation about $14.14$, understating the uncertainty of the conserved total.

The rank deficiency from deterministic spectral upsampling is visible without matrix inversion: the vector $(1,-1,1{)}^T$ has zero output contrast for every possible input. Three displayed channels therefore contain at most two independent degrees of freedom in this model. Fractional redistribution creates correlated numerical channels; it is different from physically assigning each detected photon randomly to one new channel.

The primary stimulated-Raman apparatus records spectra with array detectors and calibrates their spectral coordinates.^[ref] When those data are placed on a new spectral grid, the original detector counts remain the source of statistical information. Additional interpolation points do not supply independent photon measurements.

Rank deficiency from deterministic spectral upsampling occurs because a deterministic map cannot manufacture independent measurements. For $\mathbf{y}=A\mathbf{n}$ and input covariance matrix $C$,When the output has more channels than the input, its covariance matrix is necessarily singular. Covariance of count-conserving rebinned Raman channels gives the conserved-count example and its full propagated uncertainty.^[ref]

Using that parent's two-to-three channel map, the output always satisfies $y_1-y_2+y_3=0$. If the input readings are $20$ and $30$, the rebinned vector is $(10,25,15{)}^T$. Recovering the original readings as $n_1=2y_1$ and $n_2=2y_3$ shows that the third displayed number is not a third independent observation.

For an original fitting example, suppose both input means equal an unknown $\theta$ and their independent measurement-error variances are known to be $25$. Fitting in the original channels gives $\hat{\theta }=(20+30)/2=25$, with variance $25/2=12.5$. A correct generalized least squares fit after rebinning gives the same result and uncertainty. It must use the singular covariance on its supported range, for example through the Moore-Penrose inverse, together with the exact linear constraints.

The range restriction in singular-covariance spectral fitting matters if a proposed output model fails those constraints. A residual along $(1,-1,1{)}^T$ cannot arise from this rebinning model at all. Merely inserting a pseudoinverse into a quadratic objective would assign zero penalty to that impossible direction. Either enforce the range condition or supply a physically justified additional noise model; adding a small diagonal term solely to make the matrix invertible changes the statistical assumptions.

The motivating Raman apparatus measures detector-array spectra before any subsequent numerical grid change.^[ref] Working on the original grid often avoids this singular fitting problem while retaining the actual measurement information.

Range restriction in singular-covariance spectral fitting requires a residual to lie in the subspace where the measurement model allows random error. Let $\mathbf{y}=\mathbf{m}(\theta )+\mathbf{ϵ}$ with zero-mean error covariance $C$. Every vector $\mathbf{v}$ in the kernel of $C$ satisfiesso ${\mathbf{v}}^T\mathbf{ϵ}=0$ almost surely. Thus a permissible residual obeyswhere $C^{+}$ is the Moore-Penrose inverse. The matrix $C{C}^{+}$ is the orthogonal projector onto the range of the covariance.

Only after enforcing this condition should a generalized least squares objective use $(\mathbf{y}-\mathbf{m}{)}^T{C}^{+}(\mathbf{y}-\mathbf{m})$. The quadratic form alone assigns zero penalty to the forbidden directions and therefore cannot impose the condition. Rank deficiency from deterministic spectral upsampling gives its physical spectral-grid origin through $C=A{C}_{\mathrm{in}}{A}^T$. Covariance propagation preserves those exact dependencies.^[ref]

For an original idealized two-channel example, take $C=\mathrm{diag}(1,0)$, mean $\mathbf{m}(\theta )=(\theta ,2\theta {)}^T$ and observation $(3,4{)}^T$. The second channel has no random error in this stipulated model, so the range condition forces $\theta =2$. The remaining squared residual is one. Minimizing only the pseudoinverse quadratic would instead choose $\theta =3$ and objective zero, while predicting the impossible second-channel value six.

Real measurements need not contain a perfect channel: deterministic rebinning produces exact noise-free combinations of otherwise noisy channels. Additional physical noise in such a combination must be modelled if present. If covariance ranges change with the parameters, pseudoinverse fitting with a parameter-dependent covariance range requires enforcing the appropriate support for each candidate; a single fixed projection need not apply.

Raman detector-array spectra are recorded before numerical redistribution onto another spectral grid.^[ref] Fitting the original independent measurement coordinates often makes these support restrictions explicit and avoids treating interpolation points as new observations.

Pseudoinverse fitting with a parameter-dependent covariance range requires checking the allowed residual subspace separately for each fitted parameter. If the covariance is $C_{\theta }$ and the residual is ${\mathbf{r}}_{\theta }=\mathbf{y}-\mathbf{m}(\theta )$, the range restriction in singular-covariance spectral fitting becomesThe Moore-Penrose inverse quadratic ${\mathbf{r}}_{\theta }^T{C}_{\theta }^{+}{\mathbf{r}}_{\theta }$ alone does not enforce this condition. A projector computed at one trial parameter need not describe another trial's allowed noise.

For an original example, let ${\mathbf{u}}_{\theta }=(\cos \theta ,\sin \theta {)}^T$, zero mean and $C_{\theta }={\sigma }^2{\mathbf{u}}_{\theta }{\mathbf{u}}_{\theta }^T$. Noise lies on a line that rotates with $\theta$. With observation $\mathbf{y}=(2,0{)}^T$ and $\sigma =1$, the model at $\theta =0$ permits the observation and gives quadratic value four. At $\theta =\pi /2$, the pseudoinverse quadratic is zero, yet the observation is outside the allowed vertical line. Minimizing that quadratic without its range constraint would reward an impossible candidate.

The pinned SciPy documentation describes singular-covariance evaluation using pseudo-determinants, pseudoinverses and covariance rank; its implementation additionally applies a numerical support mask and returns zero density, or negative-infinite log density, off support.^[ref] Parameter-dependent supports also require care about which measure defines a density. Values defined on different subspaces are not automatically a common likelihood merely because each uses the same-looking formula.

A physically justified transverse noise resolves the idealized singularity. With perpendicular unit vector ${\mathbf{v}}_{\theta }$, use $C_{\theta ,ϵ}={\sigma }^2{\mathbf{u}}_{\theta }{\mathbf{u}}_{\theta }^T+{ϵ}^2{\mathbf{v}}_{\theta }{\mathbf{v}}_{\theta }^T$. For $ϵ=0.1$, the two quadratic values become four and four hundred. The determinant is the same at these orientations, so this full-rank Gaussian model correctly disfavors the transverse observation.

The transverse-noise limit of a rotating rank-one covariance model is an inference about a stated physical noise model, not permission to add arbitrary diagonal noise for numerical convenience. In spectral work, distinguish a parameter-dependent physical noise direction from a fixed deterministic rebinning matrix; the latter has a fixed output range when its input covariance is positive definite.

The transverse-noise limit of a rotating rank-one covariance model describes how an ordinary two-dimensional noise distribution concentrates onto a line as one physical noise scale tends to zero. For orthonormal vectors ${\mathbf{u}}_{\theta },{\mathbf{v}}_{\theta }$, setA zero-mean Gaussian observation has densityThe pseudoinverse fitting with a parameter-dependent covariance range article explains the range restriction at the exactly singular endpoint.

For a fixed point off the line, ${\mathbf{v}}_{\theta }^T\mathbf{y}\ne 0$, the exponential drives this density to zero as $ϵ\downarrow 0$. On the line the two-dimensional density grows like $1/ϵ$. There is no finite pointwise two-dimensional density limit: probability concentrates onto a one-dimensional Gaussian measure. A density on that line uses a different reference measure. Simply substituting zero for $ϵ$ in the full-rank density formula is invalid.

For an original angular example, let an observed vector have length $R=3$ along orientation ${\theta }_0$, with $\sigma =1$ and $ϵ=0.06$. At angular displacement $\delta =\theta -{\theta }_0$,At $\delta =0.02$ radians the density ratio is about $0.608$. The angular resolution set by transverse spectral noise has characteristic small-angle scale $ϵ/R\approx 0.02$ radians, or $1.15^{\circ }$, when the transverse noise is much smaller than the longitudinal noise.

SciPy's pinned implementation treats singular covariance using its rank and a separate numerical support check, rather than only discarding zero eigenvalues in a quadratic.^[ref] Numerical support tolerance and actual transverse measurement noise are different quantities. Choose the latter from the detector and processing model; a tiny arbitrary regularization can otherwise imply unjustified angular precision.

Angular resolution set by transverse spectral noise quantifies how sharply an observed spectral vector constrains the orientation of an anisotropic two-coordinate noise model. Use the covariance from transverse-noise limit of a rotating rank-one covariance model, with known longitudinal standard deviation $\sigma$ and transverse standard deviation $0<ϵ<\sigma$. For an observation of length $R>0$ aligned with orientation ${\theta }_0$, the Gaussian density givesThe determinant is independent of orientation, so the angular dependence comes entirely from the covariance-weighted residual.^[ref]

Define a local half-width by a chosen log-density drop $c>0$. For $2c\le K$, the lobe around ${\theta }_0$ has half-widthWhen this angle is small and $ϵ\ll \sigma$, it is approximately $\sqrt{2c}ϵ/R$. If $2c>K$, even an orthogonal orientation fails to reach that drop, so this local-width criterion does not exclude any orientation. At $R=0$ or equal noise scales, the observation has no angular preference in this model.

For an original example, choose $R=2$, $\sigma =0.8$ and $ϵ=0.04$ in consistent signal units. Then $K=2493.75$. A drop $c=1/2$ gives half-width about $0.0200$ radians, or $1.15^{\circ }$. Reducing only the transverse standard deviation to $0.02$ approximately halves the width. Increasing a numerical regularization parameter is not an experimental improvement in this noise scale.

The global orientation ambiguity in a rank-one covariance fit remains: orientations separated by $\pi$ give the same covariance. A narrow local lobe does not distinguish those two orientations. Nor is a fixed log-density drop automatically a confidence interval with a stated coverage probability. Multiple observations, unknown noise scales and experimental nuisance parameters require their corresponding inference model.

Repeated reference spectra under stable conditions can estimate the longitudinal and transverse noise components. Their covariance must include shared calibration effects, as required by the measurement-propagation rule.^[ref] The numerical example is a model calculation in spectral coordinates, rather than a specification of a particular optical instrument.

Global orientation ambiguity in a rank-one covariance fit means that covariance identifies an axis without identifying its positive direction. For ${\mathbf{u}}_{\theta }=(\cos \theta ,\sin \theta {)}^T$ and $\lambda >0$,because ${\mathbf{u}}_{\theta +\pi }=-{\mathbf{u}}_{\theta }$. Adding an isotropic transverse-noise term does not remove this equality. The angular resolution set by transverse spectral noise can make each local orientation lobe narrow while leaving the two directed orientations indistinguishable.

For an original example, let $\lambda =4$ and $\theta =30^{\circ }$. Both $30^{\circ }$ and $210^{\circ }$ giveThe eigenvalues are four and zero, and the principal eigenvector can be written as either $(\sqrt{3}/2,1/2{)}^T$ or its negative. More samples can improve estimation of this matrix without choosing between those signs.

In a zero-mean Gaussian covariance model, equal covariance gives equal probability laws, including the supported line in the singular case.^[ref] No fitting procedure using only that law can distinguish the two parameter labels. A fitting range $0\le \theta <\pi$ removes duplicate labels by convention; it does not measure the sign of a physical direction.

For example, local high-likelihood neighborhoods around $30^{\circ }$ and $210^{\circ }$ may each span only a few degrees. Reporting only the first as uncertainty in a directed angle discards an equally supported region unless additional information rules it out. A prior or coordinate convention that selects one sign should be distinguished from an experimental observation.

Sign-sensitive measurements that resolve a covariance-axis ambiguity can supply information outside the covariance-only model, such as an independently calibrated signed mean or phase reference. If the scientific quantity is genuinely an unoriented axis, an angle modulo $\pi$ is already the appropriate parameter. The interpretation of the parameter determines whether the ambiguity needs resolving.

Sign-sensitive measurements that resolve a covariance-axis ambiguity add an observable whose distribution changes when the fitted axis direction is reversed. In the global orientation ambiguity in a rank-one covariance fit, $\mathbf{u}$ and $-\mathbf{u}$ give the same covariance. An independently calibrated signed projection can distinguish them.

Consider two candidates with an additional observation $z=\pm m+\eta$, where $m=a{\mathbf{w}}^T\mathbf{u}\ne 0$ is known, $a$ has known positive sign and $\eta \sim N(0,s^2)$. Assume this calibration reading is independent of the covariance data conditional on the candidate. The Gaussian densities giveThe normal density and cumulative distribution function used here are documented in the pinned numerical reference.^[ref] The likelihood-ratio expression follows by expanding the two squared residuals.

For an original example, let $m=0.20\mathrm{V}$, $s=0.050\mathrm{V}$ and observe $z=0.10\mathrm{V}$. The log likelihood ratio is 16, strongly favoring the positive candidate under this stated model. With equal prior probabilities and equal decision costs, the sign-threshold decision has repeated-measurement error probability $\Phi (-|m|/s)=\Phi (-4)\approx 3.17\times 10^{-5}$. This decision-error probability is different from assigning a posterior error probability to that one observed value.

A practical calibration applies a known signed perturbation and records the signed detector-response component. Reversing the applied perturbation checks readout polarity. A calibrated phase-sensitive readout can provide such a projection; measuring only noise covariance does not supply its sign.

The failure of sign recovery from magnitude-only calibration data is important. For symmetric Gaussian noise, $|z|$ has the same distribution under means $+m$ and $-m$. If $\mathbf{w}$ is orthogonal to the axis, then $m=0$ and even the signed reading adds no separation. If the amplitude sign itself is unknown, reversing both amplitude and axis preserves the mean. The additional measurement resolves the ambiguity only when its reference sign and response model are independently established.

Failure of sign recovery from magnitude-only calibration data occurs when opposite signed calibration models induce exactly the same distribution of every recorded magnitude. Under this symmetry, more repetitions can estimate the magnitude distribution accurately without deciding the missing sign. This is the limitation relevant to sign-sensitive measurements that resolve a covariance-axis ambiguity.

Let $Z=sm+ϵ$ with sign $s\in \{-1,+1\}$, known $m>0$ and symmetric noise $ϵ$ having density $f(x)=f(-x)$. If the recorded value is $Y=|Z|$, then for $y>0$Symmetry gives $g(y\mid +1)=f(y-m)+f(y+m)=g(y\mid -1)$. Independent repetitions multiply identical densities, so their likelihood ratio remains one. With equal sign priors and equal decision costs, no decision rule based solely on these magnitudes can have average sign-error probability below1/2.

For normal distribution noise with standard deviation $\sigma$, the magnitude has a folded normal distribution:where $\varphi$ is the standard normal density. The formula depends on $|m|$, not its sign. The pinned numerical reference provides the equivalent standardized folded-normal density and its scale transformation.^[ref]

For the original calibration scale $m=0.20\mathrm{V}$ and $\sigma =0.050\mathrm{V}$, recording a signed value can distinguish the two models strongly, but recording only its magnitude cannot do so under the symmetric model. This is an information loss in the measurement map, rather than a shortage of averaging trials.

The symmetry assumption is essential. As an original counterexample, let noise equal−1 with probability2/3 and+2 with probability1/3, giving zero mean but an asymmetric distribution. With $m=1$, the positive-sign magnitudes are0 or3, while the negative-sign magnitudes are2 or1. Their disjoint supports permit sign recovery if this asymmetric noise model is independently established. The calibrated asymmetry itself supplies sign-reference information absent from the symmetric case.

Similarly, a genuinely signed projection or a known phase reference can break the original ambiguity. An unverified readout polarity or unknown amplitude sign can preserve it. Specify which signed information the apparatus actually records before interpreting a magnitude fit as an oriented calibration.

Spatially offset Raman spectroscopy, or SORS, collects Raman scattering from a surface region laterally displaced from the illumination spot. In a diffusely scattering material, changing the offset changes the mixture of photon paths and the relative contributions of surface and buried layers. It is not simply a tightly focused measurement at a selected depth.

The2005 experiment of P. Matousek and colleagues used trans-Stilbene powder beneath a $1\mathrm{mm}$ layer of PMMA powder. Its primary report describes a19-fold improvement in buried-layer contrast before numerical separation.^[ref] The physical point is that an opaque-looking scattering overlayer need not contribute the same spectral proportion at every collection offset.

An original two-component linear combination illustrates scaled subtraction of two spatially offset Raman spectra. Let $S(\nu )$ and $B(\nu )$ be fixed surface and buried Raman spectra, and suppose dark-corrected measurements giveIf a surface-specific band determines the scale $3/8$, thenThe surface contribution cancels while a scaled buried spectrum remains. These coefficients are chosen examples, not the paper's measured mixing fractions.

A practical arrangement fixes an excitation spot, moves the collection region laterally, rejects the excitation line with an optical filter, and sends the collected light to a spectrograph. Record comparable exposure, incident radiant power, backgrounds and spectral response at each offset. If the offset changes each component's shape, rather than only its scale, the simple subtraction model fails. Strong absorption, fluorescence, overlapping bands and measurement noise also limit inference. Offset-dependent enrichment provides evidence for depth-dependent contributions; it is not an unqualified depth calibration or unique chemical identification.

Scaled subtraction of two spatially offset Raman spectra suppresses a common surface contribution by matching its amplitude between two measurements. The spatially offset Raman spectroscopy experiment establishes that collection offset can change surface and buried-layer proportions; numerical separation then needs a model of those contributions.^[ref] The following two-component subtraction is an original simplified case, not a claim that it reproduces the paper's full analysis.

Let fixed Raman spectra $S(\nu )$ and $B(\nu )$ describe the surface and buried material. Writewith $a_0\ne 0$. ThenIf the two mixing proportions are identical, so $a_0{b}_1-a_1{b}_0=0$, this cancels the buried component too. Offset diversity must supply different mixtures, not merely a weaker copy of the same spectrum.

For an original example, take $Y_0=10S+B$ and $Y_1=4S+3B$. The correct scale $k=0.40$ gives $Z=2.6B$. Using $k=0.42$ instead gives $Z=-0.2S+2.58B$: an incorrectly subtracted surface band can produce a negative residual even when both true spectra are nonnegative. A surface-specific reference band can estimate the scale only if the assumed common shape and response are adequate.

Noise remaining after scaled Raman subtraction is distinct from the cancelled mean signal. At a surface-only channel, suppose two independent acquisitions have Poisson photoelectron counts with means $10000$ and $4000$. For fixed $k=0.40$, the expected residual is zero, butgiving a standard deviation of about $74.8$ in the scaled count units. Subtraction has not erased shot noise. Correlated acquisitions and uncertainty in $k$ add further terms. Varying shape, wavelength-dependent attenuation or an unmatched fluorescence background can invalidate this simple linear separation.

Noise remaining after scaled Raman subtraction is the random uncertainty left when a surface contribution's mean has been cancelled. For two measured spectral channels $X,Y$ and fixed scale $k$, scaled subtraction of two spatially offset Raman spectra uses $Z=Y-kX$. Its exact variance isThe covariance term describes the actual acquisition relationship; numerical subtraction does not itself make the inputs correlated or independent. General measurement propagation retains such terms.^[ref]

A focused extension allows an uncertain subtraction scale $K$, independent of the pair $(X,Y)$, with mean $k$ and variance ${\sigma }_k^2$. Expanding the product's moments givesThe final term includes uncertainty in the reference scale, not just shot noise in the two spectra.

For an original surface-only example, let independent Poisson photoelectron counts have means ${\mu }_X=20000$ and ${\mu }_Y=6000$, with $k=0.30$. The mean residual is zero. A perfectly known scale gives variance $6000+0.30^2(20000)=7800$ and standard deviation $88.3$. If instead ${\sigma }_k=0.005$, the variance becomes $7800+0.005^2(20000^2+20000)=17800.5$, giving standard deviation $133.4$ in the scaled count units.

If $K$ is estimated from these same counts, the independence assumption needs replacement by the appropriate joint model. A scale shared across wavelengths can also produce correlated spectral residuals from a shared subtraction scale, leaving a structured band-like error rather than independent noise in each bin. The original SORS study demonstrates changing layer contributions with collection offset; the moment calculation here is an independent noise model for a simplified subtraction.^[ref]

Correlated spectral residuals from a shared subtraction scale arise when one uncertain coefficient multiplies many spectral channels. Let a fixed reference profile be $\mathbf{x}$ and let the estimated scale differ from its mean by $\delta k$, with variance ${\sigma }_k^2$. Its contribution to the subtracted spectrum is $\delta \mathbf{z}=-\mathbf{x}\delta k$, givingThis covariance has rank one when $\mathbf{x}\ne 0$. It describes a coherent spectral pattern, not independent noise added separately at each wavelength. The parent noise remaining after scaled Raman subtraction treats a single channel and an uncertain scale.

For an original reference vector $\mathbf{x}=(100,200,100{)}^T$ and ${\sigma }_k=0.02$, the covariance isThe sum of the three residuals has standard uncertainty $0.02(100+200+100)=8$ in the chosen count units. Incorrectly treating the channel errors as independent gives $\sqrt{4+16+4}\approx 4.90$ and understates the uncertainty of that sum.

A weighted spectral contrast ${\mathbf{a}}^T\mathbf{z}$ receives scale-error varianceFor $\mathbf{a}=(1,-1,1{)}^T$ this contribution vanishes in the example. This projection of shared-scale uncertainty onto a spectral contrast also warns that a contrast insensitive to the reference shape can suppress a genuine signal with that same shape. Removing one error direction is not universal denoising.

Independent detector noise adds its own covariance matrix. If the reference profile is itself noisy or the scale is estimated from the same spectra, their correlations must be included too. The measurement propagation rule keeps common influences rather than discarding off-diagonal terms.^[ref] Matching a reference spectrum's mean therefore does not establish independent residuals or a reliable weak-band detection limit.

Projection of shared-scale uncertainty onto a spectral contrast determines which measured spectral features are sensitive to an uncertain common subtraction coefficient. In correlated spectral residuals from a shared subtraction scale, a fixed reference shape $\mathbf{x}$ and scale error $\delta k$ contribute $-\mathbf{x}\delta k$ to the spectrum. For a linear contrast $c={\mathbf{a}}^T\mathbf{z}$,The contrast is insensitive to this particular error when ${\mathbf{a}}^T\mathbf{x}=0$. This is an exact cancellation for a fixed reference shape; it does not require the scale error to have a Gaussian distribution.

Suppose the desired spectral signal has known shape $\mathbf{s}$ and unknown amplitude $\theta$, so that $\mathbf{z}=\theta \mathbf{s}-\delta k\mathbf{x}+\mathbf{ϵ}$. To retain its amplitude while rejecting the scale error, requireFor an original two-channel example, take $\mathbf{x}=(1,1{)}^T$ and $\mathbf{s}=(1,0{)}^T$. The difference $c=z_1-z_2$ has unit signal response and cancels any common scale error. If the two independent reading noises each have standard deviation $3$ in the chosen units, the difference has standard deviation $\sqrt{18}\approx 4.24$. Using channel 1 alone would have reading-noise standard deviation $3$ but would retain the unknown common contribution. Rejection has a noise cost even when the nuisance cancellation is exact.

If $\mathbf{s}$ is proportional to $\mathbf{x}$, the two requirements are incompatible: cancelling that reference direction also cancels the signal. In more channels, a minimum-variance spectral contrast orthogonal to a reference shape can balance detector noise among the remaining directions. The relevant variance is ${\mathbf{a}}^{T}C\mathbf{a}$, including correlations.^[ref] Uncertain reference shape or scale estimates correlated with the measured spectra require their own covariance contributions.

A minimum-variance spectral contrast orthogonal to a reference shape estimates a desired signal amplitude while exactly rejecting one fixed nuisance shape. Let $\mathbf{z}=\theta \mathbf{s}+\beta \mathbf{x}+\mathbf{ϵ}$, with known shapes $\mathbf{s},\mathbf{x}$ and a positive-definite noise covariance matrix $C$. Choose $\hat{\theta }={\mathbf{a}}^T\mathbf{z}$ to minimize ${\mathbf{a}}^{T}C\mathbf{a}$, subject toThe constraints retain unit signal response and reject the reference amplitude, as in projection of shared-scale uncertainty onto a spectral contrast. The quadratic variance includes the full noise covariance.^[ref]

Remove the reference component in the inverse-covariance geometry:For nonzero $\mathbf{x}$ and $D>0$, the minimizing contrast and its variance areFor a short proof, every other feasible contrast is $\mathbf{a}+\mathbf{v}$ with ${\mathbf{v}}^T\mathbf{s}={\mathbf{v}}^T\mathbf{x}=0$. Since $C\mathbf{a}$ is a linear combination of those shapes, the cross term vanishes, leaving the added nonnegative variance ${\mathbf{v}}^{T}C\mathbf{v}$. This is a constrained best linear unbiased estimator.

In an original three-channel example, take $\mathbf{s}=(1,0,0{)}^T$, $\mathbf{x}=(1,1,1{)}^T$ and $C=\mathrm{diag}(1,4,1)$. The optimum is $\mathbf{a}=(1,-1/5,-4/5{)}^T$, with variance $1.8$. Equal subtraction of the two reference channels, $(1,-1/2,-1/2{)}^T$, has variance $2.25$. The quieter third channel receives more of the subtraction.

If the desired and reference shapes are proportional, $D=0$ and no unit-response rejecting contrast exists. The noise penalty from nearly parallel signal and reference shapes already becomes large as $D$ approaches zero. This construction assumes fixed known shapes and covariance; estimating them from the same noisy spectra requires additional uncertainty terms.

The noise penalty from nearly parallel signal and reference shapes is the precision lost when a spectral estimator must reject a nuisance pattern resembling its desired signal. For nonzero shapes $\mathbf{s},\mathbf{x}$ and positive-definite noise covariance $C$, define their inverse-covariance correlationThe Cauchy-Schwarz inequality gives $|r|\le 1$. The minimum-variance spectral contrast orthogonal to a reference shape has varianceIf the nuisance amplitude were known and subtracted without uncertainty, the corresponding optimum variance would be $V_{\mathrm{known}}=1/({\mathbf{s}}^T{C}^{-1}\mathbf{s})$. Exact nuisance rejection therefore multiplies variance by $1/(1-r^2)$ and standard uncertainty by $1/\sqrt{1-r^2}$.

The relevant angle is the noise-whitened angle between spectral signal and nuisance shapes: it is measured after a whitening transformation, not generally from the raw plotted spectra. A channel with large noise contributes little to separating the shapes even if their plotted curves differ there.

For an original example, let $C={\sigma }^{2}I$, $\mathbf{x}=(1,0{)}^T$ and $\mathbf{s}=(1,0.1{)}^T$. Rejecting the reference forces the contrast to $(0,10{)}^T$, with variance $100{\sigma }^2$. With known nuisance amplitude, its variance would be ${\sigma }^2/1.01$. The rejection penalty is therefore $101$ in variance, about $10.05$ in standard uncertainty. Only the weak second-channel component distinguishes signal from nuisance.

The primary stimulated-Raman experiment reported systematic spectral noise above its calculated shot-noise limit and described pure-solvent subtraction of a common pattern.^[ref] A reference can remove that shared pattern while also making an overlapping desired signal hard to estimate. Retain full covariance in the variance calculation.^[ref] At exact proportionality, no unit-response rejecting contrast exists; a low-noise result obtained by suppressing the signal as well is not a valid amplitude estimate.

The noise-whitened angle between spectral signal and nuisance shapes measures their separation in units of measurement noise rather than raw plotting units. For nonzero shapes $\mathbf{s},\mathbf{x}$ and a positive-definite covariance $C$, choose a square whitening transformation $W$ with $WC{W}^T=I$. The angle isIt is the ordinary Euclidean angle after whitening. Both parallel and antiparallel shapes are indistinguishable up to an unknown nuisance amplitude; the precision penalty depends on ${\sin }^2\varphi$, as in noise penalty from nearly parallel signal and reference shapes.

For an original example, take raw shapes $\mathbf{s}=(1,1{)}^T$, $\mathbf{x}=(1,-1{)}^T$ and $C=\mathrm{diag}(1,100)$. The plotted shapes are orthogonal, but whitening gives $(1,0.1{)}^T$ and $(1,-0.1{)}^T$. Thus $\cos \varphi =0.99/1.01\approx 0.98020$ and $\varphi \approx 11.42^{\circ }$. The second channel, which separates their signs, is noisy. Rejecting the nuisance multiplies the optimum variance by $1/{\sin }^2\varphi =25.5025$, or standard uncertainty by $5.05$.

The invariance of noise-whitened spectral angles to the whitening factor follows from $W^{T}W=C^{-1}$. Different valid square whitening factors can rotate the whitened coordinates without changing the angle. Likewise an invertible change of measurement units transforms both shapes and covariance, leaving the inverse-covariance expression unchanged. Rescaling only the displayed curves while leaving their noise model fixed is a different operation.

The covariance must describe the actual measured channels and shared errors, using the full propagation rule.^[ref] If $C$ is singular, first restrict to its physically allowed range; an unrestricted inverse whitening factor does not exist.

The primary stimulated-Raman experiment described systematic spectral noise and subtraction of a pure-solvent reference.^[ref] The angle above assesses which desired patterns remain distinguishable after such nuisance rejection under a stated covariance. It does not establish that an estimated reference shape or noise model is itself accurate.

Invariance of noise-whitened spectral angles to the whitening factor means that any two complete whitening transformations for the same positive-definite covariance give the same angles between transformed shapes. Suppose square invertible matrices satisfy $W_{i}C{W}_i^T=I$. Put $Q=W_2{W}_1^{-1}$. Since $C=W_1^{-1}{W}_1^{-T}$,Thus $Q$ is orthogonal. It may rotate or reflect the whitened coordinates, but preserves their inner products, norms and angles. The noise-whitened angle between spectral signal and nuisance shapes is consequently a property of the shapes and covariance, not a particular whitening algorithm.

For an original example, chooseDirect multiplication gives $W_{1}C{W}_1^T=I$. With $\mathbf{s}=(2,1{)}^T$ and $\mathbf{x}=(2,2{)}^T$, the whitened shapes are $(1,0{)}^T$ and $(1,1{)}^T$, meeting at $45^{\circ }$. Using $W_2=Q{W}_1$ instead gives $(0,1{)}^T$ and $(-1,1{)}^T$, again at $45^{\circ }$, although the coordinate components differ.

SciPy describes covariance whitening as transforming variables to unit covariance and illustrates it with a precision-matrix construction.^[ref] The matrix proof above establishes the invariance independently of an implementation's chosen factor.

The same principle requires covariance-consistent rescaling of whitened spectral coordinates. If raw measurement units change by an invertible matrix $D$, transform both shapes by $D$ and covariance by $C^{\prime }=DC{D}^T$; then $W^{\prime }=W_1{D}^{-1}$ reproduces the same whitened shapes. Changing a plotted channel's scale without transforming its noise model is not this invariant operation.

The claim assumes a full whitening of a positive-definite covariance. Truncating modes can remove information and change angles, while a singular covariance requires restriction to its supported range. Uncertainty in an estimated covariance is also separate from this exact algebraic invariance.

Covariance-consistent rescaling of whitened spectral coordinates changes measurement units or mixes coordinates while transforming their noise model at the same time. If an invertible matrix $D$ maps a spectral vector to ${\mathbf{y}}^{\prime }=D\mathbf{y}$, thenThe covariance follows directly from the expectation of the transformed error outer product. For positive-definite $C$,Thus the inverse-covariance inner product, its norms and the resulting whitened angle are invariant. Invariance of noise-whitened spectral angles to the whitening factor establishes the related freedom to rotate the fully whitened coordinates.

For an original example, let $C=\mathrm{diag}(1,9)$, $\mathbf{s}=(1,1{)}^T$ and $\mathbf{x}=(1,-1{)}^T$. Their noise-whitened angle has cosine $0.8$, or angle about $36.87^{\circ }$. Converting only the first raw coordinate's numerical units by a factor of 1000 gives $D=\mathrm{diag}(1000,1)$ and $C^{\prime }=\mathrm{diag}(10^6,9)$. Using that transformed covariance keeps the cosine exactly $0.8$.

If the shapes are rescaled but the old covariance is mistakenly retained, the calculated cosine becomes approximately $0.99999978$, suggesting an angle only about $0.0382^{\circ }$. A unit conversion would then appear to destroy the separation between signal and nuisance. That apparent change comes from inconsistent uncertainty accounting.

For any full whitening factor $W$, the transformed factor $W^{\prime }=W{D}^{-1}$ satisfies $W^{\prime }{C}^{\prime }(W^{\prime }{)}^T=I$ and gives $W^{\prime }{\mathbf{s}}^{\prime }=W\mathbf{s}$. The pinned SciPy documentation describes whitening and demonstrates a near-identity covariance after transformation.^[ref] Measurement propagation likewise retains all covariance terms under a linear transformation.^[ref]

The invertibility condition matters. Noninvertible spectral compression versus a coordinate change can discard information and need not preserve these angles. A transformation estimated from the same noisy observations also brings estimation uncertainty beyond the fixed-matrix identity proved here.

Noninvertible spectral compression versus a coordinate change distinguishes discarding spectral coordinates from expressing the same measurements in different units. The covariance-consistent rescaling of whitened spectral coordinates article assumes an invertible transformation. A compression $\mathbf{z}=B\mathbf{y}$ with fewer independent rows than input coordinates instead has covariance $C_z=BC{B}^T$ and may lose information even when this covariance is propagated correctly.^[ref]

For positive-definite $C$ and a known signal shape $\mathbf{s}$, compare the squared noise-weighted signal sizesassuming $B$ has independent rows. Put $\mathbf{t}=C^{-1/2}\mathbf{s}$ and $M=B{C}^{1/2}$. ThenThe matrix $P$ is an orthogonal projector. This proves that compression cannot increase this signal measure under the stated covariance model. The condition for preserving a spectral signal under linear compression is that its whitened shape lie entirely in the retained subspace.

For an original example, let $C=I_2$, desired shape $\mathbf{s}=(1,1{)}^T$ and nuisance shape $\mathbf{x}=(1,-1{)}^T$. Keeping only the first channel, $B=(1,0)$, reduces $J$ from two to one and makes both shapes the same scalar. An unknown nuisance amplitude can then imitate the desired signal exactly.

Summing instead, $B=(1,1)$, gives desired response two, nuisance response zero and noise variance two. It retains $J_B=2$ for this particular signal while rejecting that nuisance. Thus a noninvertible map can preserve a chosen estimation task, but cannot in general be reversed to recover every original spectral component.

The primary Raman apparatus records array spectra before subsequent spectral processing.^[ref] Choose compression against the scientific signal and nuisance model, and propagate the full covariance. Correct uncertainty accounting does not restore a spectral distinction removed by the map.

The condition for preserving a spectral signal under linear compression specifies when compression retains the covariance-weighted signal measure of a known shape. Let the mean be $\theta \mathbf{s}$, with fixed positive-definite noise covariance $C$, and compress by a full-row-rank matrix $B$. The measures in noninvertible spectral compression versus a coordinate change obeyThus the rows of the compression must retain the noise-weighted signal direction, not necessarily the unweighted plotted shape.

For a short proof, whiten the shape as $\mathbf{t}=C^{-1/2}\mathbf{s}$ and put $M=B{C}^{1/2}$. The parent's orthogonal projection formula givesEquality holds exactly when $\mathbf{t}$ lies in the row space of $M$. Multiplying that subspace relation by $C^{-1/2}$ yields the stated condition. Linear covariance propagation gives the compressed covariance $BC{B}^T$.^[ref]

One useful construction is the scalar matched filter statistic $z={\mathbf{s}}^T{C}^{-1}\mathbf{y}$. Its mean is $\theta J$ and variance $J$, so $z/J$ has variance $1/J$, the original optimal linear amplitude precision. In the Gaussian known-covariance model, this statistic also retains the likelihood's dependence on that single amplitude.

For an original example, choose $C=\mathrm{diag}(1,4)$ and $\mathbf{s}=(1,2{)}^T$. Then $J=2$ and the retaining row is $B=(1,1/2)$. Its signal response is two and noise variance two, giving $J_B=4/2=2$. An unweighted sum $B=(1,1)$ instead has response three and variance five, giving $J_B=9/5=1.8$: ten percent of this signal measure is lost.

Signal-subspace preservation for several spectral amplitudes requires retaining the corresponding family of weighted mean directions when all those amplitudes are to be preserved. Unknown nuisance shapes can also change the relevant estimation problem. A compression designed for one known shape does not automatically preserve every spectral question or uncertainty component.

Signal-subspace preservation for several spectral amplitudes requires a compression to retain every noise-weighted mean direction relevant to those amplitudes. Let $\mathbf{y}$ have mean $A\mathbf{\theta }$, fixed positive-definite noise covariance $C$, and a full-column-rank shape matrix $A$. Compress it as $B\mathbf{y}$, with $B$ full row rank. The original and compressed amplitude-information matrices areFor Gaussian noise these are Fisher information matrices. Without that distributional assumption, they still describe the corresponding known-covariance linear estimation problem.

Whiten the shapes as $T=C^{-1/2}A$ and put $M=B{C}^{1/2}$. The orthogonal projection onto the row space of $M$ is $P=M^T(M{M}^T{)}^{-1}M$, givingThis difference is positive semidefinite. It vanishes exactly when every column of $T$ lies in that row space, equivalentlyThus preserving one selected shape does not establish preservation of the whole signal subspace. Linear covariance propagation supplies $BC{B}^T$ for the compressed observations.^[ref]

For an original example, take $C=\mathrm{diag}(1,4,9)$ and shape columns $(1,2,0{)}^T$ and $(0,0,3{)}^T$. Then $J=\mathrm{diag}(2,1)$. Choosinggives $BA=\mathrm{diag}(2,1)$ and $BC{B}^T=\mathrm{diag}(2,1)$, hence $J_B=J$. Keeping only the first compressed row retains information2 for the first amplitude but destroys sensitivity to the second. An unweighted first row $(1,1,0)$ would instead reduce the first information entry to1.8.

At least $p$ independent compressed directions are needed to preserve $p$ independent shape amplitudes. One construction is $B=A^T{C}^{-1}$, whose compressed signal has mean $J\mathbf{\theta }$ and covariance $J$. When $J$ is invertible, $J^{-1}B\mathbf{y}$ has mean $\mathbf{\theta }$ and covariance $J^{-1}$. In the fixed-covariance Gaussian model, expanding the likelihood function shows that its parameter dependence is retained through $A^T{C}^{-1}\mathbf{y}$.

Unknown nuisance shapes, changing covariance or additional spectral parameters can alter the subspace that needs preservation. The proof concerns the specified family of mean shapes and fixed noise model, rather than guaranteeing that one compression preserves every future spectral question.

Femtosecond stimulated Raman spectroscopy, or FSRS, measures a Raman spectrum through gain induced in a short broadband probe by a narrowband Raman pump. At a molecular vibration frequency matching the pump–probe frequency difference, Stokes Raman scattering transfers energy from the pump into the probe. The narrow pump and spectrograph supply spectral selectivity while the short probe samples the system in time; the probe duration alone is not the spectral linewidth.^[ref]

The2003 apparatus of David W. McCamant, Philipp Kukura and Richard A. Mathies filtered a Ti:sapphire amplifier output to make a narrowband pump near $793\mathrm{nm}$ and generated a broadband probe in sapphire. Overlapped beams traversed the sample; a spectrograph dispersed probe and reference spectra onto a dual-diode array. The paper demonstrates gain bands in cyclohexane.^[ref]

After dark correction, let $S$ and $R$ denote probe and reference counts. The Raman pump-on pump-off gain ratio isFor an original chosen spectral channel, take $(S,R{)}_{\mathrm{off}}=(10000,10000)$ and $(S,R{)}_{\mathrm{on}}=(11440,11000)$. The normalized ratio is $1.04$, a four-percent Raman gain; the unnormalized probe ratio $1.144$ would mistake a source fluctuation for additional gain. This assumes a linear detector and a reference that tracks that fluctuation.

A second original calculation uses an $800\mathrm{nm}$ pump and a $1000\mathrm{c}{\mathrm{m}}^{-1}$ vibration. The Stokes probe channel is at wavenumber $12500-1000=11500\mathrm{c}{\mathrm{m}}^{-1}$, or about $869.6\mathrm{nm}$. A broadband probe can sample many such differences simultaneously. Spatial and temporal overlap, wavelength calibration, probe chirp and non-Raman nonlinear backgrounds still require characterization. A short pulse label does not by itself establish the time resolution of a changing molecular response.

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.