pith. sign in

arxiv: 2604.08859 · v2 · submitted 2026-04-10 · 🧮 math.DS

Computing with reaction networks at input-independent speed: exponential and logarithmic functions

David F. Anderson , Badal Joshi , Tung D. Nguyen This is my paper

Pith reviewed 2026-05-10 17:40 UTC · model grok-4.3

classification 🧮 math.DS
keywords chemical reaction networksmass-action kineticschemical computationexponential functionlogarithmic functioninput-independent speedanalog computingtranscendental functions
0
0 comments X

The pith

Chemical reaction networks compute the exponential and logarithmic functions at speeds independent of input concentrations.

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

The paper constructs dedicated reaction network modules for the exponential and logarithmic functions that operate directly under mass-action kinetics rather than through series approximations. These modules reach arbitrary accuracy after a fixed amount of time that does not grow with the size of the input values. The modules combine with existing arithmetic modules so that any composite computation keeps a total runtime bounded independently of both the input concentrations and the number of operations performed. A reader would care because this removes a key scaling obstacle in chemical analog computing, where time and species count previously grew with desired accuracy or computational depth.

Core claim

The authors build specific mass-action reaction networks that serve as modules for the exponential and logarithmic functions. These networks achieve any prescribed accuracy after sufficient time while maintaining computation speed independent of the initial concentrations of the input species. The modules compose freely with each other and with the arithmetic modules from prior work; every resulting composite system runs at input-independent speed whose total duration is bounded independently of the input values and independently of the number of elementary computational steps.

What carries the argument

Mass-action reaction network modules for the exponential and logarithmic functions, equipped with fixed rate constants chosen so that the dynamics produce the target functions in time bounded independently of inputs.

If this is right

  • Any composite system built from the exponential, logarithmic, and arithmetic modules runs with total computation time bounded independently of inputs and of the number of steps.
  • Arbitrary accuracy for the exponential or logarithmic functions is obtained simply by allowing more computation time while keeping all rate constants fixed.
  • The same modules can be reused in larger networks without introducing input-dependent slowdowns.
  • The constructions provide templates that can be adapted for direct computation of additional transcendental functions.

Where Pith is reading between the lines

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

  • Such modules could allow biochemical circuits to perform nonlinear transformations on concentration signals while preserving predictable timing across different signal strengths.
  • Implementation in synthetic biology might yield cellular controllers whose response time remains stable even when input molecule counts fluctuate widely.
  • The fixed-time property opens the possibility of chaining many transcendental operations into larger chemical programs without the total runtime becoming unpredictable.

Load-bearing premise

The particular reaction networks and fixed rate constants chosen for the exponential and logarithmic modules actually generate the claimed output concentrations and time bounds under mass-action kinetics.

What would settle it

A simulation or experiment on the proposed networks showing that the time to reach a fixed accuracy threshold changes substantially when the input concentration is varied over several orders of magnitude.

Figures

Figures reproduced from arXiv: 2604.08859 by Badal Joshi, David F. Anderson, Tung D. Nguyen.

Figure 1
Figure 1. Figure 1: Comparison of the logarithm constructions developed in this paper. Each ellipse represents a desirable [PITH_FULL_IMAGE:figures/full_fig_p018_1.png] view at source ↗
read the original abstract

The concept of input-independent computational time for chemistry-based analog computers was introduced in Anderson-Joshi (2025), where it was shown that arithmetic operations can be computed in a fixed time independent of the input values. Here, by inputs we mean the numerical values encoded by the initial concentrations of designated input species, with the underlying reaction network and rate constants held fixed. Combining these operations via power series to approximate transcendental functions is possible in principle, but the number of chemical species required grows with the number of terms retained, and achieving sufficient accuracy may demand many terms -- a burden that is especially severe for slowly converging series such as the power series for the logarithm. In this paper, we begin the program of directly computing transcendental functions by chemical reaction networks by focusing on the exponential and logarithmic functions, two widely used transcendental functions, and constructing reaction network modules that compute them without relying on truncated power series. We show that the resulting modules are mass-action systems, and prove that they achieve arbitrary accuracy given sufficient time while operating at input-independent speed. Moreover, these modules can be freely combined with each other and with the arithmetic modules of Anderson-Joshi (2025): any such composite computation also runs at input-independent speed, with the total computation time bounded independently of the input values \emph{and} of the number of elementary steps in the computation. Logarithm and exponential functions serve as foundational cases, and the constructions developed here are intended to serve as templates for the direct computation of more general transcendental functions by chemical reaction networks.

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

3 major / 3 minor

Summary. The paper constructs explicit mass-action reaction network modules for the exponential and logarithmic functions that achieve arbitrary accuracy in finite time. It proves that these modules run at input-independent speed (time bounds independent of initial input concentrations) and that they compose freely with each other and with the arithmetic modules of Anderson-Joshi (2025), yielding composite computations whose total time remains bounded independently of both the input values and the number of elementary operations.

Significance. If the explicit networks, rate constants, and convergence proofs hold, the work provides a concrete advance in chemical analog computation by supplying direct, non-series-based implementations of transcendental functions that preserve the input-independent timing property across arbitrary compositions. This removes a key scalability barrier for building more complex chemical computers and supplies reusable modules whose correctness can be verified by standard mass-action analysis.

major comments (3)
  1. [§4, Theorem 4.1] §4 (Exponential module), Theorem 4.1 and the subsequent Lyapunov analysis: the claimed uniform lower bound on convergence time independent of input concentration x requires an explicit spectral-gap estimate or comparison lemma showing that the Jacobian eigenvalues around the target equilibrium remain bounded away from zero for all x > 0. The mass-action ODEs contain bilinear terms whose linearization can produce eigenvalues scaling as O(x) or O(1/x); without a uniform bound or a proof that the chosen fixed rate constants cancel this dependence, the input-independent speed claim is not yet established.
  2. [§5, Proposition 5.2] §5 (Logarithmic module), Proposition 5.2: the same uniform spectral-gap requirement applies. The construction must exhibit a network whose linearized rate matrix at equilibrium has a spectral gap bounded below by a positive constant independent of the input; the current argument appears to rely on a comparison lemma whose constants may still depend on the input magnitude.
  3. [§6] §6 (Composition theorem): the inductive argument that any finite composition of the new modules with Anderson-Joshi arithmetic modules inherits an input-independent time bound inherits the same gap requirement from the individual modules. If the base modules lack a uniform gap, the composite bound cannot be guaranteed.
minor comments (3)
  1. [Abstract, §2] The abstract and introduction refer to “arbitrary accuracy given sufficient time” but do not state the precise error metric (e.g., absolute vs. relative, or concentration vs. logarithmic scale) used in the convergence statements; this should be fixed in §2.
  2. [Throughout] Notation for the input species concentrations is occasionally overloaded (x used both for the numerical value and the species); a consistent convention would improve readability.
  3. [Introduction] The paper cites Anderson-Joshi (2025) but does not indicate whether that work is already published or still in preprint form; adding the arXiv identifier or journal status would help readers locate the referenced arithmetic modules.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the careful and constructive report. The comments correctly identify that our convergence arguments would benefit from more explicit uniform spectral-gap estimates to fully substantiate the input-independent timing claims. We will revise the manuscript to supply these estimates (via added lemmas and expanded calculations) while preserving the existing network constructions and overall results. Point-by-point responses follow.

read point-by-point responses

  1. Referee: [§4, Theorem 4.1] §4 (Exponential module), Theorem 4.1 and the subsequent Lyapunov analysis: the claimed uniform lower bound on convergence time independent of input concentration x requires an explicit spectral-gap estimate or comparison lemma showing that the Jacobian eigenvalues around the target equilibrium remain bounded away from zero for all x > 0. The mass-action ODEs contain bilinear terms whose linearization can produce eigenvalues scaling as O(x) or O(1/x); without a uniform bound or a proof that the chosen fixed rate constants cancel this dependence, the input-independent speed claim is not yet established.

    Authors: We agree that an explicit uniform spectral-gap bound is needed for full rigor. The proof of Theorem 4.1 uses a Lyapunov function whose derivative yields a decay rate controlled by fixed rate constants chosen to dominate the bilinear contributions; the comparison lemma already produces a positive lower bound independent of x because the equilibrium scaling and rate choices cancel the O(x) and O(1/x) terms. However, the bound is not written out as a standalone lemma. In revision we will insert an explicit lemma computing the minimal eigenvalue gap (via Gershgorin or direct characteristic polynomial analysis) and verify it is bounded below by a positive constant depending only on the fixed rates. revision: yes

  2. Referee: [§5, Proposition 5.2] §5 (Logarithmic module), Proposition 5.2: the same uniform spectral-gap requirement applies. The construction must exhibit a network whose linearized rate matrix at equilibrium has a spectral gap bounded below by a positive constant independent of the input; the current argument appears to rely on a comparison lemma whose constants may still depend on the input magnitude.

    Authors: We accept the observation. The logarithmic module employs a similar Lyapunov comparison, but the constants in the decay estimate are not shown to be independent of the input in the current text. Revision will add an explicit spectral-gap lemma for the log network, confirming that the chosen fixed rates produce a uniform positive lower bound on the real parts of the relevant eigenvalues for all positive inputs, again via direct linearization and comparison. revision: yes

  3. Referee: [§6] §6 (Composition theorem): the inductive argument that any finite composition of the new modules with Anderson-Joshi arithmetic modules inherits an input-independent time bound inherits the same gap requirement from the individual modules. If the base modules lack a uniform gap, the composite bound cannot be guaranteed.

    Authors: We agree that the composition theorem in §6 relies on the base modules possessing uniform gaps. Once the explicit uniform bounds are added to Theorems 4.1 and Proposition 5.2, the inductive argument carries over directly; the total time bound remains the sum of the individual module times, each independent of inputs and of the number of steps. We will add a clarifying sentence in §6 stating this dependence and referencing the new lemmas. revision: yes

Circularity Check

0 steps flagged

No circularity: new modules and proofs are independent of prior arithmetic work

full rationale

The paper constructs explicit mass-action reaction networks for exp and log, then proves directly that they achieve arbitrary accuracy at input-independent speed. The combination claim with Anderson-Joshi (2025) arithmetic modules inherits the input-independent property only after the new modules are shown to satisfy it; the time bounds are not obtained by renaming or fitting parameters from the prior paper. No self-definitional equations, fitted inputs presented as predictions, or load-bearing uniqueness theorems imported via self-citation appear in the derivation. The self-citation is limited to the foundational arithmetic modules and does not reduce the central claims to tautology.

Axiom & Free-Parameter Ledger

1 free parameters · 2 axioms · 1 invented entities

The central claims rest on choosing particular reaction networks and fixed rate constants whose mass-action ODEs are asserted to realize the functions with the stated time properties; these choices are not derived from first principles but introduced to achieve the desired behavior.

free parameters (1)
  • rate constants of the exp and log modules
    Fixed values chosen so that the resulting ODEs produce the exponential or logarithmic mapping after a bounded time.
axioms (2)
  • domain assumption The evolution of concentrations obeys the mass-action law, yielding a system of ODEs
    Invoked to prove convergence to the target function value.
  • domain assumption Initial concentrations are positive and the networks are closed under the stated reactions
    Required for the time-bound arguments to hold.
invented entities (1)
  • exp and log reaction network modules no independent evidence
    purpose: Direct computation of the transcendental functions without series truncation
    Newly constructed sets of reactions and species introduced in the paper.

pith-pipeline@v0.9.0 · 5576 in / 1508 out tokens · 108653 ms · 2026-05-10T17:40:13.725210+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

24 extracted references · 24 canonical work pages

  1. [1]

    Anderson and Badal Joshi

    David F. Anderson and Badal Joshi. Chemical mass-action systems as analog computers: Implementing arithmetic computations at specified speed.Theoretical Computer Science, 1025:114983, 2025

    work page 2025

  2. [2]

    Programmability of chemical reaction networks

    Matthew Cook, David Soloveichik, Erik Winfree, and Jehoshua Bruck. Programmability of chemical reaction networks. In Anne Condon, David Harel, Joost N. Kok, Arto Salomaa, and Erik Winfree, editors,Algorithmic Bioprocesses, Natural Computing Series, pages 543–584. Springer, Berlin, 2009

    work page 2009

  3. [3]

    Deterministic function computation with chemical reaction networks.Natural computing, 13:517–534, 2014

    Ho-Lin Chen, David Doty, and David Soloveichik. Deterministic function computation with chemical reaction networks.Natural computing, 13:517–534, 2014

    work page 2014

  4. [4]

    Anderson, and Erik Winfree

    Daniele Cappelletti, Andres Ortiz-Mu˜ noz, David F. Anderson, and Erik Winfree. Stochas- tic chemical reaction networks for robustly approximating arbitrary probability distributions. Theoretical Computer Science, 801:64–95, 2020

    work page 2020

  5. [5]

    Strong Turing completeness of continuous chemical reaction networks and compilation of mixed analog-digital programs

    Fran¸ cois Fages, Guillaume Le Guludec, Olivier Bournez, and Amaury Pouly. Strong Turing completeness of continuous chemical reaction networks and compilation of mixed analog-digital programs. InInternational Conference on Computational Methods in Systems Biology, pages 108–127. Springer, 2017

    work page 2017

  6. [6]

    Rate-independent compu- tation in continuous chemical reaction networks.Journal of the ACM, 70(3):1–61, 2023

    Ho-Lin Chen, David Doty, Wyatt Reeves, and David Soloveichik. Rate-independent compu- tation in continuous chemical reaction networks.Journal of the ACM, 70(3):1–61, 2023

    work page 2023

  7. [7]

    Anderson, Badal Joshi, and Abhishek Deshpande

    David F. Anderson, Badal Joshi, and Abhishek Deshpande. On reaction network implemen- tations of neural networks.Journal of the Royal Society Interface, 18(177):20210031, 2021

    work page 2021

  8. [8]

    DNA as a universal substrate for chemical kinetics.Proceedings of the National Academy of Sciences, 107(12):5393–5398, 2010

    David Soloveichik, Georg Seelig, and Erik Winfree. DNA as a universal substrate for chemical kinetics.Proceedings of the National Academy of Sciences, 107(12):5393–5398, 2010

    work page 2010

  9. [9]

    Scaling up digital circuit computation with DNA strand dis- placement cascades.Science, 332(6034):1196–1201, 2011

    Lulu Qian and Erik Winfree. Scaling up digital circuit computation with DNA strand dis- placement cascades.Science, 332(6034):1196–1201, 2011

    work page 2011

  10. [10]

    Neural network computation with DNA strand displacement cascades.Nature, 475(7356):368–372, 2011

    Lulu Qian, Erik Winfree, and Jehoshua Bruck. Neural network computation with DNA strand displacement cascades.Nature, 475(7356):368–372, 2011

    work page 2011

  11. [11]

    Campagnolo, Daniel S

    Olivier Bournez, Manuel L. Campagnolo, Daniel S. Gra¸ ca, and Emmanuel Hainry. Polynomial differential equations compute all real computable functions on computable compact intervals. Journal of Complexity, 23(3):317–335, 2007

    work page 2007

  12. [12]

    Gra¸ ca, and Amaury Pouly

    Olivier Bournez, Daniel S. Gra¸ ca, and Amaury Pouly. Polynomial time corresponds to solu- tions of polynomial ordinary differential equations of polynomial length.Journal of the ACM, 64(6):38:1–38:76, 2017

    work page 2017

  13. [13]

    Claude E. Shannon. Mathematical theory of the differential analyser.Journal of Mathematics and Physics, 20(1–4):337–354, 1941

    work page 1941

  14. [14]

    Glenn E. R. Cowan, Robert C. Melville, and Yannis P. Tsividis. A VLSI analog com- puter/digital computer accelerator.IEEE Journal of Solid-State Circuits, 41(1):42–53, 2006. 20

    work page 2006

  15. [15]

    Energy-efficient hybrid analog/digital approximate computation in continuous time.IEEE Journal of Solid-State Circuits, 51(7):1514–1524, 2016

    Ning Guo, Yipeng Huang, Tao Mai, Sharvil Patil, Chi Cao, Mingoo Seok, Simha Sethumad- havan, and Yannis Tsividis. Energy-efficient hybrid analog/digital approximate computation in continuous time.IEEE Journal of Solid-State Circuits, 51(7):1514–1524, 2016

    work page 2016

  16. [16]

    A survey on analog models of computation

    Olivier Bournez and Amaury Pouly. A survey on analog models of computation. In Vasco Brattka and Peter Hertling, editors,Handbook of Computability and Complexity in Analysis, Theory and Applications of Computability, pages 173–226. Springer, Cham, 2021

    work page 2021

  17. [17]

    Rubens, Rahul Sarpeshkar, and Timothy K

    Ramiz Daniel, Jacob R. Rubens, Rahul Sarpeshkar, and Timothy K. Lu. Synthetic analog computation in living cells.Nature, 497:619–623, 2013

    work page 2013

  18. [18]

    Analog synthetic biology.Philosophical Transactions of the Royal Society A, 372(2012):20130110, 2014

    Rahul Sarpeshkar. Analog synthetic biology.Philosophical Transactions of the Royal Society A, 372(2012):20130110, 2014

    work page 2012

  19. [19]

    Chemical reaction networks for computing logarithm.Synthetic Biology, 2(1):ysx002, 2017

    Chun Tung Chou. Chemical reaction networks for computing logarithm.Synthetic Biology, 2(1):ysx002, 2017

    work page 2017

  20. [20]

    Klinge, James I

    Xiang Huang, Titus H. Klinge, James I. Lathrop, Xiaoyuan Li, and Jack H. Lutz. Real-time computability of real numbers by chemical reaction networks.Natural Computing, 18(1):63–73, 2019

    work page 2019

  21. [21]

    Klinge, and James I

    Xiang Huang, Titus H. Klinge, and James I. Lathrop. Real-time equivalence of chemical reaction networks and analog computers. InInternational Conference on DNA Computing and Molecular Programming, pages 37–53. Springer, 2019

    work page 2019

  22. [22]

    Klinge, James I

    Willem Fletcher, Titus H. Klinge, James I. Lathrop, Dawn A. Nye, and Matthew Rayman. Real-time computing and robust memory with deterministic chemical reaction networks.Nat- ural Computing, 24:383–397, 2025

    work page 2025

  23. [23]

    Juris Hartmanis and Richard E. Stearns. On the computational complexity of algorithms. Transactions of the American Mathematical Society, 117:285–306, 1965

    work page 1965

  24. [24]

    On the inverse problem of reaction kinetics.Qualitative theory of differential equations, 30:363–379, 1981

    Vera H´ ars and J´ anos T´ oth. On the inverse problem of reaction kinetics.Qualitative theory of differential equations, 30:363–379, 1981. 21

    work page 1981