pith. sign in

arxiv: 2604.20186 · v1 · submitted 2026-04-22 · ⚛️ physics.chem-ph

Chromatographic Peak Shape from a Stochastic-Diffusive Model with Multiple Retention Mechanisms: Analytic Time-Domain Expression and Derivatives

Hern\'an R. S\'anchez This is my paper

Pith reviewed 2026-05-09 23:34 UTC · model grok-4.3

classification ⚛️ physics.chem-ph
keywords chromatographypeak shapestochastic modelanalytic expressionretention mechanismsdiffusionexponentially modified Gaussian
0
0 comments X

The pith

A stochastic-diffusive model produces an analytic time-domain expression for chromatographic peak shapes that handles multiple independent slow retention mechanisms.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper establishes a closed-form expression for the shape of chromatographic peaks by combining axial diffusion, finite initial band variance, fast short-duration retention events, and an arbitrary number of independent slow retention mechanisms. This expression comes with an efficient evaluation method and analytic derivatives for every parameter. When tested on literature peaks, the model achieves lower full-profile root-mean-square error than the exponentially modified Gaussian in every case. The improvement becomes especially large once more than one slow mechanism is allowed.

Core claim

Within a stochastic-diffusive framework that includes axial diffusion, finite initial spatial variance, a high rate of short-duration retention events, and any number of independent slow retention mechanisms each characterized by infrequent long-duration events, a time-domain analytic expression for the peak shape is derived together with a fast evaluation scheme and closed-form derivatives with respect to all parameters.

What carries the argument

The stochastic-diffusive model with multiple independent slow retention mechanisms, which yields the time-domain analytic peak-shape expression and its derivatives.

If this is right

  • For the single-slow-mechanism case the expression evaluates two to four orders of magnitude faster than the prior analytic route.
  • Analytical derivatives for all parameters can be obtained at essentially the same cost as the peak shape itself, enabling direct gradient-based optimization.
  • Allowing more than one slow mechanism produces data-dependent error reductions that can exceed one order of magnitude for some peaks.
  • Full-profile RMSE values drop to 0.03–0.14 percent of peak height, compared with 0.43–5.57 percent for the exponentially modified Gaussian on the same data.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The framework could be used to test whether observed peak asymmetries arise from a small number of discrete slow processes rather than a continuous distribution of rates.
  • Because the expression is analytic, it supplies an explicit route to compute moments or other derived quantities without numerical integration or simulation.
  • The independence assumption between mechanisms suggests a natural test: whether adding a second mechanism improves fit only when the column chemistry actually contains two distinct slow sites.

Load-bearing premise

The slow retention mechanisms are treated as fully independent and the stochastic events are described solely by their rate parameters without additional correlations or memory effects from the column packing.

What would settle it

A set of experimental chromatograms recorded on columns whose packing is known to contain two or more chemically distinct slow retention sites, with direct comparison of the model's predicted versus observed full peak profiles to check whether the reported RMSE values hold or systematic residuals appear.

Figures

Figures reproduced from arXiv: 2604.20186 by Hern\'an R. S\'anchez.

Figure 1
Figure 1. Figure 1: Illustrative fits to three literature peaks (see text for details on each case). Upper panels: experimental data (black crosses, regularly subsampled for clarity) and fitted profiles. Lower panels: residuals (dots), computed from the full data. Fitted profiles and their corresponding residuals: EMG ( ), 𝑀 = 1 ( ), 𝑀 = 2 ( ), 𝑀 = 3 ( ). Only model orders producing a visually appreciable improvement are show… view at source ↗
read the original abstract

A time-domain analytic expression for chromatographic peak shapes is derived within a stochastic-diffusive framework that incorporates axial diffusion (molecular and multipath/Eddy), finite initial spatial variance, a retention mechanism characterized by a high rate of short-duration events, and an arbitrary number of independent slow retention mechanisms, each characterized by its own rate of infrequent, long-duration events. A highly efficient evaluation scheme is derived for this expression. In the single-slow-mechanism case, it is two to four orders of magnitude faster than the previously available analytic route. Analytical derivatives with respect to all model parameters are also obtained, and each can be evaluated at computational cost comparable to that of the peak-shape expression. Illustrative fits to three literature peaks yielded full-profile RMSE values lower than those of the exponentially modified Gaussian in all tested cases, with minima ranging from 0.03 to 0.14 percent of peak height, compared with 0.43 to 5.57 percent for the reference model. Relative to the one-slow-mechanism formulation, allowing more than one slow mechanism produced a data-dependent improvement that exceeded one order of magnitude for one of the peaks.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper derives an analytic time-domain expression for chromatographic peak shapes from a stochastic-diffusive model incorporating axial diffusion (molecular and multipath), finite initial spatial variance, a fast retention mechanism (high rate of short-duration events), and an arbitrary number of independent slow retention mechanisms (infrequent long-duration events). It provides a highly efficient evaluation scheme (2-4 orders of magnitude faster than prior routes for the single-slow-mechanism case), analytic derivatives w.r.t. all parameters at comparable cost, and shows that fits to three literature peaks yield lower full-profile RMSE (0.03-0.14% of peak height) than the exponentially modified Gaussian (0.43-5.57%).

Significance. If the derivation holds under the stated assumptions, the work supplies a physically grounded, extensible peak-shape model with practical computational efficiency and derivatives suitable for routine fitting and optimization in chromatography. The multi-mechanism flexibility and reported RMSE gains over EMG represent a concrete advance for modeling asymmetric or complex peaks.

major comments (2)
  1. [§2 (model setup) and the multi-mechanism extension] §2 (model setup) and the multi-mechanism extension: the closed-form time-domain expression and the efficient evaluation scheme (claimed 2-4 orders faster) are obtained only by treating the slow retention mechanisms as strictly independent so that the overall retention-time density is the convolution of the individual densities (equivalently, product of characteristic functions). The manuscript provides no justification, sensitivity test, or discussion of possible spatial correlations or memory effects in real column packings; violation of independence would introduce cross terms that invalidate both the analytic form and the efficiency claim.
  2. [Results on illustrative fits] Results on illustrative fits: the reported RMSE improvements (0.03-0.14 % vs. EMG) and the data-dependent gain exceeding one order of magnitude when adding mechanisms are presented without specifying the number of slow mechanisms per peak, the resulting parameter count, or any uncertainty quantification or overfitting check. These details are required to establish that the gains are attributable to the model rather than added flexibility.
minor comments (2)
  1. [Abstract] Abstract: the phrase 'full-profile RMSE values ... percent of peak height' should explicitly define the normalization (e.g., RMSE divided by maximum peak height) and confirm whether the metric is computed over the entire time window or a defined interval.
  2. [Notation] Notation: ensure consistent symbols for the rate constants of the fast and slow mechanisms across equations and text; a table summarizing all free parameters would aid readability.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments on our manuscript. We address each of the major comments in detail below and outline the revisions we will make to improve the paper.

read point-by-point responses

  1. Referee: [§2 (model setup) and the multi-mechanism extension] §2 (model setup) and the multi-mechanism extension: the closed-form time-domain expression and the efficient evaluation scheme (claimed 2-4 orders faster) are obtained only by treating the slow retention mechanisms as strictly independent so that the overall retention-time density is the convolution of the individual densities (equivalently, product of characteristic functions). The manuscript provides no justification, sensitivity test, or discussion of possible spatial correlations or memory effects in real column packings; violation of independence would introduce cross terms that invalidate both the analytic form and the efficiency claim.

    Authors: The independence of the slow retention mechanisms is a foundational assumption of the stochastic model, as each mechanism is represented by an independent Poisson process governing the occurrence of retention events. This allows the characteristic function of the total retention time to be the product of the individual characteristic functions, which in turn yields the convolution for the density and enables the efficient numerical evaluation scheme. Physically, this corresponds to the solute molecules experiencing distinct retention sites or mechanisms (e.g., different chemical interactions) that do not influence each other. We agree that in real column packings, spatial correlations or memory effects could exist due to heterogeneity. We will revise the manuscript to include an explicit discussion of this assumption in §2, providing justification based on the separation of timescales and the multi-mechanism framework, while acknowledging potential limitations and noting that the model is an approximation. A comprehensive sensitivity analysis involving correlated processes would require additional Monte Carlo simulations and is left for future work. revision: partial

  2. Referee: [Results on illustrative fits] Results on illustrative fits: the reported RMSE improvements (0.03-0.14 % vs. EMG) and the data-dependent gain exceeding one order of magnitude when adding mechanisms are presented without specifying the number of slow mechanisms per peak, the resulting parameter count, or any uncertainty quantification or overfitting check. These details are required to establish that the gains are attributable to the model rather than added flexibility.

    Authors: In the illustrative fits, we employed one or two slow mechanisms depending on the peak, with the largest improvement observed when using two mechanisms for one of the peaks. Each additional slow mechanism introduces two parameters (the rate constant and the mean event duration). We will update the results section to explicitly state the number of mechanisms and total parameters for each fit, and include a table summarizing these details along with the RMSE values. The fits are presented as illustrative examples of the model's capability rather than a statistical study; consequently, formal uncertainty quantification (e.g., parameter covariances) and overfitting diagnostics (e.g., AIC or cross-validation) were not performed. We will add a brief discussion noting the increased flexibility with more parameters and the physical motivation for additional mechanisms, while cautioning that the reported gains should be interpreted in light of this. revision: yes

Circularity Check

0 steps flagged

Derivation self-contained from stochastic model assumptions with no reduction to inputs

full rationale

The paper presents a derivation of the time-domain analytic expression directly from the stochastic-diffusive model assumptions (axial diffusion, finite initial variance, fast retention events, and independent slow mechanisms), without any step that equates the output expression or its derivatives to fitted parameters, data, or prior self-citations by construction. The efficient evaluation scheme and analytical derivatives follow from the same model framework. Illustrative fits and RMSE comparisons to literature peaks are post-derivation validations on held-out data and do not enter the derivation chain. No load-bearing step reduces the claimed result to its inputs; the independence assumption is stated explicitly as a modeling choice rather than derived from the result itself.

Axiom & Free-Parameter Ledger

2 free parameters · 2 axioms · 0 invented entities

The derivation rests on standard stochastic assumptions for retention events and diffusion; no new particles or forces are introduced. Model parameters such as retention rates and diffusion coefficients are fitted to data in the examples.

free parameters (2)
  • retention rates for each slow mechanism
    Fitted to experimental peak data; central to the multi-mechanism improvement shown in the abstract.
  • axial diffusion coefficients
    Included as adjustable parameters in the stochastic-diffusive framework.
axioms (2)
  • domain assumption Retention events follow independent Poisson-like processes with distinct rates for fast and slow mechanisms.
    Invoked to obtain the closed-form time-domain solution.
  • domain assumption Axial diffusion (molecular plus Eddy) can be treated as a Gaussian spreading process independent of retention.
    Required for the stochastic-diffusive framework to separate diffusion and retention contributions.

pith-pipeline@v0.9.0 · 5510 in / 1448 out tokens · 153631 ms · 2026-05-09T23:34:15.134954+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

23 extracted references · 21 canonical work pages

  1. [1]

    Mathematical functions for the representation of chromatographic peaks

    Valerio B. Di Marco and G.Giorgio Bombi. “Mathematical functions for the representation of chromatographic peaks”. In:Journal of Chromatography A931.1 (2001), 1–30.doi: 10.1016/S0021-9673(01)01136-0. 6

    work page doi:10.1016/s0021-9673(01)01136-0 2001

  2. [2]

    The use ofpredictive models to developchromatography-basedpurification processes

    C. R. Bernau,M. Knödler,J. Emonts,R. C. Jäpel,andJ. F. Buyel. “The use ofpredictive models to developchromatography-basedpurification processes”. In:Frontiers in Bioengineering and BiotechnologyVolume 10 - 2022 (2022).doi: 10.3389/fbioe.2022.1009102

    work page doi:10.3389/fbioe.2022.1009102 2022

  3. [3]

    A MolecularDynamic Theory of Chromatography

    J. Calvin Giddings andHenry Eyring. “A MolecularDynamic Theory of Chromatography”. In:The JournalofPhysicalChemistry59.5 (1955), 416–421.doi: 10.1021/j150527a009

    work page doi:10.1021/j150527a009 1955

  4. [4]

    Stochastic Considerations on Chromatographic Dis- persion

    J. Calvin Giddings. “Stochastic Considerations on Chromatographic Dis- persion”. In:The Journal of Chemical Physics26.1 (1957),169–173.doi: 10.1063/1.1743244

    work page doi:10.1063/1.1743244 1957

  5. [5]

    On the StochasticTheoryofChromatogra- phy

    Donald A. McQuarrie. “On the StochasticTheoryofChromatogra- phy”. In:The JournalofChemicalPhysics38.2 (1963),437–445.doi: 10.1063/1.1733677

    work page doi:10.1063/1.1733677 1963

  6. [6]

    Contributions to the StochasticTheoryofChro- matographicKinetics

    George H. Weiss. “Contributions to the StochasticTheoryofChro- matographicKinetics”. In:Separation Science5.1 (1970),51–62.doi: 10.1080/00372367008057946

    work page doi:10.1080/00372367008057946 1970

  7. [7]

    Stochastic-Dispersive Theory ofChromatography

    Attila Felinger, Alberto Cavazzini, Maurizio Remelli, and Francesco Dondi. “Stochastic-Dispersive Theory ofChromatography”. In:AnalyticalChemistry 71.20 (1999), 4472–4479.doi: 10.1021/ac990412u

    work page doi:10.1021/ac990412u 1999

  8. [8]

    The characteristicfunction methodin the stochastictheoryofchromatography

    Francesco Dondi andMaurizio Remelli. “The characteristicfunction methodin the stochastictheoryofchromatography”. In:The Journalof Physical Chemistry90.9 (1986), 1885–1891.doi: 10.1021/j100400a029

    work page doi:10.1021/j100400a029 1986

  9. [9]

    Stochastic theory of two-site adsorption chromatography by the characteristic function method

    Alberto Cavazzini,Maurizio Remelli,andFrancesco Dondi. “Stochastic theory of two-site adsorption chromatography by the characteristic function method”. In:JournalofMicrocolumn Separations9.4 (1997),295–302. doi: 10.1002/(SICI)1520-667X(1997)9:4<295::AID-MCS7>3.0.CO;2-X

    work page doi:10.1002/(sici)1520-667x(1997)9:4 1997

  10. [10]

    Stochastic Theory of Multiple-Site Linear Adsorption Chromatography

    Alberto Cavazzini, Maurizio Remelli, Francesco Dondi, and Attila Felinger. “Stochastic Theory of Multiple-Site Linear Adsorption Chromatography”. In: Analytical Chemistry71.16 (1999), 3453–3462.doi: 10.1021/ac990282p

    work page doi:10.1021/ac990282p 1999

  11. [11]

    Microscopicmodels ofliquidchromatography: From ensemble-averaged information to resolution of fundamental viewpoint at single-molecule level

    Luisa Pasti, Nicola Marchetti, Roberta Guzzinati, Martina Catani, Valentina Bosi,Francesco Dondi,Annamária Sepsey,Attila Felinger,andAl- berto Cavazzini. “Microscopicmodels ofliquidchromatography: From ensemble-averaged information to resolution of fundamental viewpoint at single-molecule level”. In:TrAC Trends in AnalyticalChemistry81 (2016), 63–68.doi: ...

    work page doi:10.1016/j.trac.2015.08.007 2016

  12. [12]

    Estimation ofchromatographicpeakshape param- eters in Fourierdomain

    Attila Felinger. “Estimation ofchromatographicpeakshape param- eters in Fourierdomain”. In:Talanta83.4 (2011),1074–1078.doi: 10.1016/j.talanta.2010.10.001

    work page doi:10.1016/j.talanta.2010.10.001 2011

  13. [13]

    Chromatographic peak shape from a stochastic model: Analytictime-domain expression in terms ofphysicalparameters and conditions under which heterogeneity reduces tailing

    Hernán R. Sánchez. “Chromatographic peak shape from a stochastic model: Analytictime-domain expression in terms ofphysicalparameters and conditions under which heterogeneity reduces tailing”. In:Journal of Chro- matography A1775 (2026), 466896.doi: 10.1016/j.chroma.2026.466896

    work page doi:10.1016/j.chroma.2026.466896 2026

  14. [14]

    The distribution of the sum of independent gamma random variables

    Peter G Moschopoulos. “The distribution of the sum of independent gamma random variables”. In:Annals ofthe Institute ofStatisticalMathematics 37.1 (1985), 541–544.doi: 10.1007/BF02481123

    work page doi:10.1007/bf02481123 1985

  15. [15]

    Reconstruction ofchromatographic peaks using the exponen- tially modified Gaussian function

    Yuri Kalambet,Yuri Kozmin,Ksenia Mikhailova,Igor Nagaev,and Pavel Tikhonov. “Reconstruction ofchromatographic peaks using the exponen- tially modified Gaussian function”. In:Journal of Chemometrics25.7 (2011), 352–356.doi: 10.1002/cem.1343

    work page doi:10.1002/cem.1343 2011

  16. [16]

    ExponentiallymodifiedGaussian func- tions—a goodmodelforchromatographic peaks in isocratic HPLC?

    Pamela J NaishandS Hartwell. “ExponentiallymodifiedGaussian func- tions—a goodmodelforchromatographic peaks in isocratic HPLC?” In: Chromatographia26.1 (1988), 285–296.doi: 10.1007/BF02268168

    work page doi:10.1007/bf02268168 1988

  17. [17]

    Multidimensionaltranscriptome datasetforsystematicevaluation of Jakyakgamcho-tang-induced cell signatures

    Su-Jin Baek,Haeseung Lee,Sang-Min Park,Aeyung Kim,No Soo Kim, Eun-Hye Seo,A Yeong Lee,Yu Ri Kim,Wook Jin Kim,Kyu-Won Seo, et al. “Multidimensionaltranscriptome datasetforsystematicevaluation of Jakyakgamcho-tang-induced cell signatures”. In:Scientific Data(2026). doi: 10.1038/s41597-026-06759-6

    work page doi:10.1038/s41597-026-06759-6 2026

  18. [18]

    Raw HPLC chromatographic data forJakyamgamcho-tang

    SuJin Baek. “Raw HPLC chromatographic data forJakyamgamcho-tang”. In: (2025).doi: 10.6084/m9.figshare.30962618.v4

    work page doi:10.6084/m9.figshare.30962618.v4 2025

  19. [19]

    Simple and Effective HPLC Method for Elucidating Glycerol Oxidation Products

    Eva Ng, Camilo A. Mesa, Elena Mas-Marzá, and Sixto Giménez. “Simple and Effective HPLC Method for Elucidating Glycerol Oxidation Products”. In:ACS MeasurementScience Au5.3 (2025),367–377.doi: 10.1021/ac- smeasuresciau.5c00024

    work page doi:10.1021/ac- 2025

  20. [20]

    S¨ uzen, Supplement and dataset for bootstrapped time-integrated spread complexity, 10.5281/zen- odo.19707590 (2026)

    Eva Ng,Camilo Mesa,Elena Más-Marzá,andSixto Giménez.Kinetic Insights into GlycerolElectrooxidation on Nickel: Current-Dependent ProductDistribution andReaction Mechanism. 2024.doi: 10.5281/zen- odo.13374357.url:https://doi.org/10.5281/zenodo.13374357

    work page doi:10.5281/zen- 2024

  21. [21]

    Mark Galassi,Jim Davies,James Theiler,Brian Gough,Gerard Jungman, Patrick Alken, Michael Booth, Fabrice Rossi, and Rhys Ulerich.GNU Scien- tific Library Reference Manual. Third. Bristol,U.K.: Network Theory Ltd., 2009.isbn: 9780954612078.url: https://www.gnu.org/software/ gsl/

    2009

  22. [22]

    doi:10.1038/s41592-019-0686-2 , eprint =

    Pauli Virtanen,Ralf Gommers,Travis E. Oliphant,Matt Haberland,Tyler Reddy,David Cournapeau,Evgeni Burovski,Pearu Peterson,Warren Weckesser, Jonathan Bright, et al. “SciPy 1.0: Fundamental Algorithms for Scientific Computing in Python”. In:Nature Methods17 (2020), 261–272. doi: 10.1038/s41592-019-0686-2

    work page doi:10.1038/s41592-019-0686-2 2020

  23. [23]

    Sánchez.STEP: StochasticTime-domain Elution Profile

    Hernán R. Sánchez.STEP: StochasticTime-domain Elution Profile. GitHubrepository. 2026.url: https://github.com/hernan3009/ STEP. 7

    2026