Physics-Guided Concentration Inference from Resistance Transients in a Mixed-Phase SnO-SnO$_2$ Carbon Monoxide Sensor with p-n Switching

Amit Kumar Gangwar; Preetam Singh; Sani Biswas

arxiv: 2605.23971 · v1 · pith:KVW5EOQWnew · submitted 2026-05-13 · ⚛️ physics.chem-ph · cs.LG· physics.app-ph

Physics-Guided Concentration Inference from Resistance Transients in a Mixed-Phase SnO-SnO₂ Carbon Monoxide Sensor with p-n Switching

Sani Biswas , Preetam Singh , Amit Kumar Gangwar This is my paper

Pith reviewed 2026-06-30 21:26 UTC · model grok-4.3

classification ⚛️ physics.chem-ph cs.LGphysics.app-ph

keywords carbon monoxide sensorresistance transientsp-n switchingphysics-guided descriptorsmachine learningSnO-SnO2concentration inferencetransient dynamics

0 comments

The pith

Cycle-level resistance transients in a mixed-phase SnO-SnO2 sensor encode CO concentration through physically interpretable dynamics, with p-type favoring classification and n-type favoring regression.

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

This paper develops a machine-learning framework that extracts physically interpretable descriptors from resistance transients measured on a mixed-phase tin oxide sensor that switches between p-type and n-type conduction with temperature. These descriptors are combined with compact FFT and DWT summaries and fed to Random Forest models under leakage-aware grouped cross-validation, separately for each regime. The physics-guided descriptors alone remain competitive with the full fused feature set, indicating that the dominant concentration signal already resides in the transient shapes. The p-type regime delivers the strongest class discrimination at roughly 96.5 percent accuracy while the n-type regime supplies the strongest quantitative estimates with mean absolute error near 1.48 ppm and R-squared near 0.992.

Core claim

In a mixed-phase SnO-SnO2 carbon monoxide sensor that exhibits temperature-dependent p-n switching, cycle-level transient responses represented by physically interpretable descriptors, together with FFT and DWT summaries, support both multi-class concentration classification and continuous regression. Across both regimes the fused features perform best overall, yet the physics-guided descriptor block alone stays highly competitive, showing that the essential concentration information is already captured in the transient dynamics. The p-type branch achieves the best classification performance while the n-type branch achieves the best regression performance.

What carries the argument

Physics-guided descriptor block derived from cycle-level resistance transients, augmented by FFT and DWT summaries and evaluated with leakage-aware grouped cross-validation in separate p-type and n-type regimes.

If this is right

The dominant concentration information resides in physically meaningful transient dynamics rather than in additional spectral summaries.
Fused features give the strongest performance, but the physics-guided block alone is nearly as effective.
p-type sensing is particularly favorable for concentration-class discrimination.
n-type sensing is particularly favorable for high-fidelity quantitative estimation.
Cycle-level analysis with leakage-aware validation extends conventional single-response metrics while preserving interpretability.

Where Pith is reading between the lines

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

Sensors could be deliberately operated at different temperatures to switch between classification mode and high-precision regression mode depending on the application need.
The same transient-feature approach may transfer to other mixed-phase or temperature-switchable gas sensors without requiring new material synthesis.
Real-time deployment could pre-select the operating regime on the basis of measured temperature to optimize either classification accuracy or regression precision.
Testing the framework on multi-gas mixtures or varying humidity levels would reveal how robust the transient descriptors remain under more complex field conditions.

Load-bearing premise

The temperature-dependent p-n switching produces cleanly separable regimes whose transient responses can be analyzed independently without significant overlap or misassignment of cycles.

What would settle it

A new set of cycles in which p-n regime labels are ambiguous or in which forced regime misassignment produces large drops in model accuracy would falsify the separability premise.

read the original abstract

This work presents a physics-guided machine-learning framework for carbon monoxide concentration inference from experimentally measured resistance transients of a mixed-phase SnO-SnO$_2$ material gas sensor exhibiting temperature-dependent p-n switching behavior. Cycle-level transient responses are represented through physically interpretable descriptors and complemented by compact fast Fourier transform (FFT) and discrete wavelet transform (DWT)-based summaries. Using leakage-aware grouped cross-validation, we study both multi-class concentration classification and continuous concentration regression for the p-type and n-type sensing regimes separately. Across both regimes, fused features provide the strongest overall performance, while the physics-guided descriptor block remains highly competitive, indicating that the dominant concentration information is already encoded in physically meaningful transient dynamics. The p-type branch shows the best concentration-class discrimination, with the fused Random Forest classifier reaching approximately $96.5\%$ accuracy, whereas the n-type branch yields the best quantitative concentration estimation, with the fused Random Forest regressor achieving an MAE$\approx 1.48$ ppm and an R$^2$ $\approx 0.992$. These results reveal a clear dual-regime behavior: p-type sensing is particularly favorable for classification, whereas n-type sensing is more favorable for high-fidelity regression. More broadly, the study demonstrates that leakage-aware, cycle-level, physics-guided machine learning can extend conventional gas-sensing analysis beyond single-response metrics while preserving physical interpretability

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.

Desk Editor's Note private letter to a colleague

The paper shows physics-guided transient descriptors can match fused ML features on this mixed-phase sensor's data with a p-n regime split, but the separation method is not described.

read the letter

The main point is that they measured resistance transients on a SnO-SnO2 sensor that switches p-n with temperature, pulled out physically motivated cycle descriptors plus FFT and DWT summaries, and ran random forests for both classification and regression of CO levels, keeping the regimes separate. The physics block stays competitive with the full fused set, and they report a split where p-type does better at classification while n-type does better at regression.

That split and the fact that the transient descriptors carry most of the signal without new algorithms are the parts worth noting. The grouped cross-validation is a sensible choice for cycle data.

The soft spot is the regime handling. The abstract says the regimes are studied separately but gives no rule for assigning cycles, no mention of hysteresis checks, and no test for overlap or leakage between p-type and n-type responses. If assignment is noisy, the separate accuracy and R2 numbers become harder to compare. Feature selection details are also missing, so it is not obvious whether the physics descriptors were fixed in advance or adjusted after seeing results.

This is for people who already work on gas-sensor transients and want to see how standard feature sets perform on a real p-n switching material. The experimental grounding is there and the claims stay within what the data can support.

Send it to review if the full manuscript spells out the assignment rule and shows the descriptors were not tuned post-hoc. The work is incremental but the data and setup are concrete enough to deserve referee time.

Referee Report

1 major / 2 minor

Summary. The paper proposes a physics-guided machine learning framework for inferring carbon monoxide concentrations from resistance transients in a mixed-phase SnO-SnO2 gas sensor exhibiting temperature-dependent p-n switching. Cycle-level transients are represented via physically interpretable descriptors supplemented by FFT and DWT summaries. Using leakage-aware grouped cross-validation, the work performs multi-class classification and continuous regression separately in the p-type and n-type regimes, reporting that fused features yield the strongest performance while physics-guided descriptors remain competitive. It claims p-type regime favors classification (fused RF ~96.5% accuracy) and n-type favors regression (fused RF MAE≈1.48 ppm, R²≈0.992), revealing dual-regime behavior.

Significance. If the regime partitioning holds and descriptors are not post-hoc tuned, the results indicate that transient dynamics already encode the dominant concentration information in physically meaningful terms. This extends conventional steady-state gas-sensing analysis while retaining interpretability and suggests regime-specific task optimization.

major comments (1)

[Abstract] Abstract (third paragraph) and the central dual-regime claim: the manuscript states that regimes are studied 'separately' but provides no explicit assignment rule for partitioning cycles into p-type versus n-type (temperature threshold, resistance sign change, or hysteresis handling) and no validation that switching is sharp without overlap or misassignment. This is load-bearing because the reported p-type accuracy and n-type R² become incomparable if cycles leak between regimes, undermining the conclusion of distinct optimal behaviors.

minor comments (2)

[Abstract] Abstract: performance numbers (96.5%, MAE≈1.48 ppm, R²≈0.992) are stated without accompanying error-bar methodology, number of cycles, or cross-validation fold details.
[Abstract] The abstract does not indicate whether the physics-guided descriptor set was fixed prior to ML evaluation or selected after observing which features improved scores.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for their careful and constructive review. We address the single major comment below and will revise the manuscript accordingly to strengthen the presentation of the dual-regime analysis.

read point-by-point responses

Referee: [Abstract] Abstract (third paragraph) and the central dual-regime claim: the manuscript states that regimes are studied 'separately' but provides no explicit assignment rule for partitioning cycles into p-type versus n-type (temperature threshold, resistance sign change, or hysteresis handling) and no validation that switching is sharp without overlap or misassignment. This is load-bearing because the reported p-type accuracy and n-type R² become incomparable if cycles leak between regimes, undermining the conclusion of distinct optimal behaviors.

Authors: We agree that an explicit, reproducible partitioning rule is essential for validating the dual-regime conclusions and for allowing readers to assess potential leakage between regimes. While the manuscript describes the temperature-dependent p-n switching behavior and the separate analysis of the two regimes, we acknowledge that the precise cycle-assignment criterion (e.g., temperature threshold, sign of resistance change, or handling of any hysteresis) is not stated with sufficient clarity or accompanied by validation of regime separation. In the revised manuscript we will add a dedicated paragraph (or subsection) that (i) states the exact assignment rule used, (ii) reports the number of cycles assigned to each regime, and (iii) provides supporting evidence—such as a resistance-versus-temperature plot or a table of transition points—demonstrating that the switch is sharp and that overlap or misassignment is negligible. This addition will make the reported performance metrics directly comparable and will reinforce the claim of distinct optimal behaviors in the two regimes. revision: yes

Circularity Check

0 steps flagged

No circularity; claims rest on experimental data and cross-validated ML performance

full rationale

The paper reports ML classification and regression results on experimentally measured resistance transients, using leakage-aware grouped cross-validation. Physics-guided descriptors are extracted from observed transient dynamics and evaluated for predictive utility on held-out cycles; they are not fitted to the target concentration values. Regime separation into p-type and n-type is an experimental partitioning step whose validity is external to the reported metrics. No equations, self-citations, or ansatzes are shown that reduce any performance claim to a definitional identity or fitted input. The derivation chain is therefore self-contained against the data.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

Only the abstract is available, so the ledger is necessarily incomplete. The work implicitly relies on standard supervised learning assumptions (i.i.d. cycles after grouping, feature relevance) and the domain assumption that resistance transients contain concentration information separable by regime. No invented entities are introduced.

axioms (2)

domain assumption Resistance transients contain extractable, physically meaningful descriptors that correlate with gas concentration independently of the ML model chosen.
Stated in the abstract as the basis for using physics-guided descriptors; if false the competitiveness claim collapses.
domain assumption The p-type and n-type regimes can be cleanly partitioned without cycle misassignment.
The separate analysis of regimes presupposes this partition is reliable.

pith-pipeline@v0.9.1-grok · 5806 in / 1554 out tokens · 24127 ms · 2026-06-30T21:26:43.091627+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

14 extracted references · 7 canonical work pages

[1]

Single resistive sensor for selective detection of multiple VOCs employing SnO2 hollow spheres and machine learning algorithm: A proof of concept,

S. Acharyya, B. Jana, S. Nag, G. Saha, and P. K. Guha, “Single resistive sensor for selective detection of multiple VOCs employing SnO2 hollow spheres and machine learning algorithm: A proof of concept,”Sens. Actuators B Chem., vol. 321, Art. no. 128484, 2020, doi: 10.1016/j.snb.2020.128484

work page doi:10.1016/j.snb.2020.128484 2020
[2]

Ultra-selective tin oxide-based chemiresistive gas sensor employing signal transform and machine learning techniques,

S. Acharyya, S. Nag, and P. K. Guha, “Ultra-selective tin oxide-based chemiresistive gas sensor employing signal transform and machine learning techniques,”Anal. Chim. Acta, vol. 1217, Art. no. 339996, 2022, doi: 10.1016/j.aca.2022.339996

work page doi:10.1016/j.aca.2022.339996 2022
[3]

Characteristics of semiconductor gas sensors I. Steady state gas response,

P. K. Clifford and D. T. Tuma, “Characteristics of semiconductor gas sensors I. Steady state gas response,”Sens. Actuators, vol. 3, pp. 233– 254, 1982, doi: 10.1016/0250-6874(82)80026-7

work page doi:10.1016/0250-6874(82)80026-7 1982
[4]

Preparation of nanocrystalline Pd/SnO 2 thin films deposited on alumina substrate by reactive magnetron sputtering for efficient CO gas sensing,

A. K. Gangwar, R. Godiwal, S. Srivastava, P. Pal, G. Gupta, and P. Singh, “Preparation of nanocrystalline Pd/SnO 2 thin films deposited on alumina substrate by reactive magnetron sputtering for efficient CO gas sensing,”Mater . Res. Bull., vol. 148, Art. no. 111692, 2022

2022
[5]

Room temperature sputtered nanocrystalline SnO2 thin film sensitized with Pd nanoparticles for high- performance CO gas sensing application,

A. K. Gangwar, S. Srivastava, R. Godiwal, J. Jaiswal, P. Vashishtha, S. Pal, P. Pal, G. Gupta, and P. Singh, “Room temperature sputtered nanocrystalline SnO2 thin film sensitized with Pd nanoparticles for high- performance CO gas sensing application,”Opt. Mater ., vol. 128, Art. no. 112362, 2022

2022
[6]

Temperature-dependent p-n switching for highly selective CO gas sensing based on mixed phases of magnetron sputtered (p)SnO-(n)SnO2 thin films,

A. K. Gangwar, R. Godiwal, U. Varshney, S. Das, J. S. Tawale, G. Gupta, and P. Singh, “Temperature-dependent p-n switching for highly selective CO gas sensing based on mixed phases of magnetron sputtered (p)SnO-(n)SnO2 thin films,”Appl. Surf. Sci., vol. 655, Art. no. 159607, 2024

2024
[7]

Enhancement of p-type mobility in tin monoxide by native defects,

D. B. Granato, J. Caraveo-Frescas, H. Alshareef, and U. Schwingen- schl¨ogl, “Enhancement of p-type mobility in tin monoxide by native defects,”Appl. Phys. Lett., vol. 102, 2013

2013
[8]

A mini-review on metal oxide semiconductor gas sensors for carbon monoxide detection at room temperature,

Y . He and M. Jiao, “A mini-review on metal oxide semiconductor gas sensors for carbon monoxide detection at room temperature,” Chemosensors, vol. 12, no. 4, Art. no. 55, 2024, doi: 10.3390/chemosen- sors12040055

work page doi:10.3390/chemosen- 2024
[9]

Low temperature NO2 sensing properties of RF-sputtered SnO- SnO2 heterojunction thin-film with p-type semiconducting behavior,

H.-S. Jeong, M.-J. Park, S.-H. Kwon, H.-J. Joo, S.-H. Song, and H.-I. Kwon, “Low temperature NO2 sensing properties of RF-sputtered SnO- SnO2 heterojunction thin-film with p-type semiconducting behavior,” Ceram. Int., vol. 44, pp. 17283–17289, 2018

2018
[10]

Prepara- tion and characterization of SnO nanoplatelets by microwave innovative technique,

T. Krishnakumar, N. Pinna, K. Perumal, and R. Jayaprakash, “Prepara- tion and characterization of SnO nanoplatelets by microwave innovative technique,” inAIP Conf. Proc., 2008, pp. 122–125

2008
[11]

Metal-oxide semiconductors for carbon monoxide (CO) gas sensing: A review,

S. Mahajan and S. Jagtap, “Metal-oxide semiconductors for carbon monoxide (CO) gas sensing: A review,”Appl. Mater . Today, vol. 18, Art. no. 100483, 2020, doi: 10.1016/j.apmt.2019.100483

work page doi:10.1016/j.apmt.2019.100483 2020
[12]

Synthesis of CuO thin films by a direct current reactive sputtering process for CO gas sensing application,

D. Mahana, A. K. Mauraya, S. Kumaragurubaran, P. Singh, and S. K. Muthusamy, “Synthesis of CuO thin films by a direct current reactive sputtering process for CO gas sensing application,”Phys. Scr ., vol. 98, Art. no. 035709, 2023

2023
[13]

Machine learning-based discrimination of indoor pollutants using an oxide gas sensor array: High endurance against ambient humidity and temperature,

J. Oh, S. H. Kim, M.-J. Lee, H. Hwang, W. Ku, J. Lim, I.-S. Hwang, J.-H. Lee, and J.-H. Hwang, “Machine learning-based discrimination of indoor pollutants using an oxide gas sensor array: High endurance against ambient humidity and temperature,”Sens. Actuators B Chem., vol. 364, Art. no. 131894, 2022, doi: 10.1016/j.snb.2022.131894

work page doi:10.1016/j.snb.2022.131894 2022
[14]

Review: Influ- ences of semiconductor metal oxide properties on gas sensing characteristics,

B. Saruhan, R. Fomekong, and S. Nahirniak, “Review: Influ- ences of semiconductor metal oxide properties on gas sensing characteristics,”Front. Sens., vol. 2, Art. no. 657931, 2021, doi: 10.3389/fsens.2021.657931

work page doi:10.3389/fsens.2021.657931 2021

[1] [1]

Single resistive sensor for selective detection of multiple VOCs employing SnO2 hollow spheres and machine learning algorithm: A proof of concept,

S. Acharyya, B. Jana, S. Nag, G. Saha, and P. K. Guha, “Single resistive sensor for selective detection of multiple VOCs employing SnO2 hollow spheres and machine learning algorithm: A proof of concept,”Sens. Actuators B Chem., vol. 321, Art. no. 128484, 2020, doi: 10.1016/j.snb.2020.128484

work page doi:10.1016/j.snb.2020.128484 2020

[2] [2]

Ultra-selective tin oxide-based chemiresistive gas sensor employing signal transform and machine learning techniques,

S. Acharyya, S. Nag, and P. K. Guha, “Ultra-selective tin oxide-based chemiresistive gas sensor employing signal transform and machine learning techniques,”Anal. Chim. Acta, vol. 1217, Art. no. 339996, 2022, doi: 10.1016/j.aca.2022.339996

work page doi:10.1016/j.aca.2022.339996 2022

[3] [3]

Characteristics of semiconductor gas sensors I. Steady state gas response,

P. K. Clifford and D. T. Tuma, “Characteristics of semiconductor gas sensors I. Steady state gas response,”Sens. Actuators, vol. 3, pp. 233– 254, 1982, doi: 10.1016/0250-6874(82)80026-7

work page doi:10.1016/0250-6874(82)80026-7 1982

[4] [4]

Preparation of nanocrystalline Pd/SnO 2 thin films deposited on alumina substrate by reactive magnetron sputtering for efficient CO gas sensing,

A. K. Gangwar, R. Godiwal, S. Srivastava, P. Pal, G. Gupta, and P. Singh, “Preparation of nanocrystalline Pd/SnO 2 thin films deposited on alumina substrate by reactive magnetron sputtering for efficient CO gas sensing,”Mater . Res. Bull., vol. 148, Art. no. 111692, 2022

2022

[5] [5]

Room temperature sputtered nanocrystalline SnO2 thin film sensitized with Pd nanoparticles for high- performance CO gas sensing application,

A. K. Gangwar, S. Srivastava, R. Godiwal, J. Jaiswal, P. Vashishtha, S. Pal, P. Pal, G. Gupta, and P. Singh, “Room temperature sputtered nanocrystalline SnO2 thin film sensitized with Pd nanoparticles for high- performance CO gas sensing application,”Opt. Mater ., vol. 128, Art. no. 112362, 2022

2022

[6] [6]

Temperature-dependent p-n switching for highly selective CO gas sensing based on mixed phases of magnetron sputtered (p)SnO-(n)SnO2 thin films,

A. K. Gangwar, R. Godiwal, U. Varshney, S. Das, J. S. Tawale, G. Gupta, and P. Singh, “Temperature-dependent p-n switching for highly selective CO gas sensing based on mixed phases of magnetron sputtered (p)SnO-(n)SnO2 thin films,”Appl. Surf. Sci., vol. 655, Art. no. 159607, 2024

2024

[7] [7]

Enhancement of p-type mobility in tin monoxide by native defects,

D. B. Granato, J. Caraveo-Frescas, H. Alshareef, and U. Schwingen- schl¨ogl, “Enhancement of p-type mobility in tin monoxide by native defects,”Appl. Phys. Lett., vol. 102, 2013

2013

[8] [8]

A mini-review on metal oxide semiconductor gas sensors for carbon monoxide detection at room temperature,

Y . He and M. Jiao, “A mini-review on metal oxide semiconductor gas sensors for carbon monoxide detection at room temperature,” Chemosensors, vol. 12, no. 4, Art. no. 55, 2024, doi: 10.3390/chemosen- sors12040055

work page doi:10.3390/chemosen- 2024

[9] [9]

Low temperature NO2 sensing properties of RF-sputtered SnO- SnO2 heterojunction thin-film with p-type semiconducting behavior,

H.-S. Jeong, M.-J. Park, S.-H. Kwon, H.-J. Joo, S.-H. Song, and H.-I. Kwon, “Low temperature NO2 sensing properties of RF-sputtered SnO- SnO2 heterojunction thin-film with p-type semiconducting behavior,” Ceram. Int., vol. 44, pp. 17283–17289, 2018

2018

[10] [10]

Prepara- tion and characterization of SnO nanoplatelets by microwave innovative technique,

T. Krishnakumar, N. Pinna, K. Perumal, and R. Jayaprakash, “Prepara- tion and characterization of SnO nanoplatelets by microwave innovative technique,” inAIP Conf. Proc., 2008, pp. 122–125

2008

[11] [11]

Metal-oxide semiconductors for carbon monoxide (CO) gas sensing: A review,

S. Mahajan and S. Jagtap, “Metal-oxide semiconductors for carbon monoxide (CO) gas sensing: A review,”Appl. Mater . Today, vol. 18, Art. no. 100483, 2020, doi: 10.1016/j.apmt.2019.100483

work page doi:10.1016/j.apmt.2019.100483 2020

[12] [12]

Synthesis of CuO thin films by a direct current reactive sputtering process for CO gas sensing application,

D. Mahana, A. K. Mauraya, S. Kumaragurubaran, P. Singh, and S. K. Muthusamy, “Synthesis of CuO thin films by a direct current reactive sputtering process for CO gas sensing application,”Phys. Scr ., vol. 98, Art. no. 035709, 2023

2023

[13] [13]

Machine learning-based discrimination of indoor pollutants using an oxide gas sensor array: High endurance against ambient humidity and temperature,

J. Oh, S. H. Kim, M.-J. Lee, H. Hwang, W. Ku, J. Lim, I.-S. Hwang, J.-H. Lee, and J.-H. Hwang, “Machine learning-based discrimination of indoor pollutants using an oxide gas sensor array: High endurance against ambient humidity and temperature,”Sens. Actuators B Chem., vol. 364, Art. no. 131894, 2022, doi: 10.1016/j.snb.2022.131894

work page doi:10.1016/j.snb.2022.131894 2022

[14] [14]

Review: Influ- ences of semiconductor metal oxide properties on gas sensing characteristics,

B. Saruhan, R. Fomekong, and S. Nahirniak, “Review: Influ- ences of semiconductor metal oxide properties on gas sensing characteristics,”Front. Sens., vol. 2, Art. no. 657931, 2021, doi: 10.3389/fsens.2021.657931

work page doi:10.3389/fsens.2021.657931 2021