REVIEW 4 major objections 5 minor 11 references
1/f and Random Telegraph Noise of Single-Layer Graphene Devices with Interdigitated Electrodes
T0 review · 4 major / 5 minor · reviewed 2026-08-09 · deepseek-v4-flash
Pith's one-line read Single-layer graphene devices with interdigitated electrodes show 1/f and random telegraph noise whose normalized amplitude scales inversely with the active graphene area, implying the noise comes from conductance fluctuations in the…
desk verdict Useful empirical noise data on graphene IDE devices, but the headline 1/A scaling rests on a seven-point fit with two post-hoc outliers, so the design rule needs replication before it hardens into a claim. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
The load-bearing object is the active-area product $A = G \times W \times (N_{\mathrm{fingers}}-1)$, with $G$ the gap between interdigitated fingers, $W$ the graphene micro-ribbon width, and $N_{\mathrm{fingers}}$ the number of electrodes; the device is modeled as $(N_{\mathrm{fingers}}-1)$ bottom-gate graphene FETs in series, which licenses the use of standard MOSFET low-frequency-noise formalism. The argument runs by extracting the equivalent resistance noise $S_R$ from the slope of $S_I^{1/2}$ versus voltage at 20 Hz and then fitting normalized noise $S_R/R^2$ versus $1/A$. The linearity of that fit is what converts individual spectra into a geometric scaling law.
What would settle it
Re-measure the two excluded devices on fresh chips and redo the $S_R/R^2$ versus $1/A$ fit with all nine points; if the linear trend disappears or the slope stops matching $1/A$, the central claim fails. A second check is to hold $G$ and $W$ fixed while varying $N_{\mathrm{fingers}}$ so that $A$ changes by a controlled integer factor and confirm that the noise follows $1/A$ point-for-point, rather than depending on individual device variability.
Extended reading notes
Core claim
The central claim is that normalized low-frequency resistance noise in these IDE graphene devices obeys a $1/A$ area law: plotting $S_R/R^2$ against $1/[G W (N_{\mathrm{fingers}}-1)]$ for the nine devices yields a linear correlation, and the paper states that the higher the active graphene area, the higher the SNR. Because the trend holds across different finger gaps and ribbon widths, the paper concludes that the noise is resistive conductance fluctuations distributed over the whole graphene area rather than noise generated at the graphene–aluminium contacts. Random telegraph noise observed in some devices, with two current levels whose histogram fits two Gaussians, is attributed to capture and emission of carriers by traps at the SLG/SiO$_2$ interface, and its amplitude shows the same increase for smaller ribbon widths as the $1/f$ noise.
Load-bearing premise
The area-scaling law depends on treating two of the nine devices (G15W50 and G25W100) as outliers and removing them from the linear fit; if they are included, the claimed $1/A$ correlation is significantly weakened, and the paper offers no independent evidence that grain boundaries, wrinkles, or ribbon edges caused their deviation.
Editorial extensions
If this is right
- Sensor design can use active area as a tunable knob: making $A$ larger lowers normalized $1/f$ noise and raises SNR, all else equal.
- Noise characterization of such devices should normalize by $1/A$ so results from different IDE layouts are directly comparable.
- RTN amplitude follows the same geometry trend, so interface traps at the SLG/SiO$_2$ boundary are part of the same area-dependent noise budget.
- The series-GFET model implies that adding fingers (increasing $N_{\mathrm{fingers}}$) should reduce normalized noise in proportion to the added area, a prediction testable without changing material growth.
Reading between the lines
- The $1/A$ law implies a resolution-versus-noise tradeoff for graphene sensors: shrinking the sensing area to improve spatial localization raises normalized noise, so applications must balance SNR against footprint; the slope of the reported fit gives the quantitative exchange rate.
- If the outlier explanation is correct, wafer-scale defect maps such as Raman D-band intensity or grain-boundary density should predict which identical-geometry devices depart from the area law; this is a testable prediction the paper does not make.
- The voltage-independence of $\Delta I/I_H$ suggests each RTN event is dominated by a single effective trap; combining the measured $\Delta I/I_H$ with capture and emission time statistics could convert the area scaling into an interface trap density estimate.
- Because the devices are modeled as series-connected GFETs, the same data imply that the gate-referred voltage noise spectral density should also scale with $1/A$; re-analyzing already-published IDE graphene noise data with this normalization would provide an independent check.
Signed reviews
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper reports low-frequency noise measurements (1/f and random telegraph noise) on nine single-layer graphene micro-ribbon devices with interdigitated electrodes of varying gap (8, 15, 25 μm) and width (50, 100, 200 μm). The authors observe 1/f behavior in all devices and RTN in several, extract the equivalent resistance noise SR from the slope of sqrt(SI) versus applied voltage, and plot the normalized resistance noise SR/R^2 against 1/A, where A = G × W × (Nfingers − 1). A linear trend is reported for seven of the nine devices, with two devices (G15W50, G25W100) labeled outliers. The paper concludes that the higher the active graphene area the higher the SNR, and that the low-frequency noise is conductance noise originating from the active graphene area rather than the contact regions; it also claims that RTN shows the same area dependence.
Significance. If the area-scaling result is correct, the paper would provide a useful design rule for graphene sensors with interdigitated electrodes and evidence about the physical origin of low-frequency noise in this geometry. The experimental effort is systematic in covering a 3×3 grid of device geometries, and the noise extraction method (linear regression of sqrt(SI) versus V) is standard. However, the central empirical claim is statistically fragile: the area-scaling conclusion rests on a fit to seven of nine points, with exactly one device per geometry, no error bars, and no quantitative outlier analysis. Because the paper's main conclusion depends on this fit, the result needs additional support before it can be considered established.
major comments (4)
- [Section III.B, Fig. 6] The central claim that SR/R^2 scales with 1/A rests on a linear fit to seven of nine devices, with G15W50 and G25W100 excluded as outliers. The manuscript gives no quantitative outlier criterion (e.g., residual threshold, studentized residual, robust regression) and no independent evidence, such as Raman mapping or AFM, that these two devices are structurally different. With one device per geometry and no measurement repeats, the two excluded points may simply be device-to-device scatter. The authors should either justify the exclusion statistically (for example, by showing that the remaining points remain significant under robust regression and that the excluded points fall outside a stated confidence band) or provide multiple nominally identical devices per geometry. Without this, the conclusion that the noise scales strictly with 1/A is not supported.
- [Section III.B, Fig. 6, Eq. (implicit)] The fit to the 1/A dependence is presented without fit statistics: no R^2, no confidence interval for the slope, and no number of degrees of freedom are reported. Additionally, the area definition A = G × W × (Nfingers − 1) is asserted without justification. It is not explained why the product of gap, width, and finger number is the correct 'active area' for conductance fluctuations, nor whether the graphene under the metal fingers contributes. To support the conclusion that the noise is area-dominated rather than contact-dominated, the authors should report the regression parameters and compare with an alternative model that includes a constant contact-noise contribution independent of A.
- [Section III.C, Fig. 8 and Section IV] The conclusion that RTN exhibits the same area dependence as 1/f noise is not supported by the presented data. Figure 8 shows only two devices, both with gap 8 μm and widths 50 and 100 μm; no RTN was observed for G8W200, and no RTN results are shown for gaps of 15 and 25 μm. Two points are insufficient to establish an area-scaling law for RTN amplitude. This claim should either be removed from the conclusions or explicitly qualified as preliminary.
- [Section III.B] The sentence 'It can be concluded without doubt that the higher the active graphene area, the higher the SNR' overstates the evidence. The paper does not directly measure signal-to-noise ratio, but rather the current noise PSD normalized by resistance; a high-SNR claim requires a defined signal level. Furthermore, since the underlying correlation is weakened by the outlier issue, the phrase 'without doubt' is inappropriate. The conclusion should be reworded to reflect the statistical uncertainty.
minor comments (5)
- [Section II.A] The sentence 'Finally, e-beam resist was removed with standard acetone/IPA cleaning steps using.' is incomplete; the trailing 'using' should be removed.
- [Section III] The heading 'RESULTS AND DISCUSSSION' contains a typo: 'DISCUSSSION' should be 'DISCUSSION'.
- [Section II.B] The phrase 'The same triplet of devices was studied in detail' is unclear because the previous sentence refers to the nine devices in Table I. Please specify which three devices are meant.
- [Section III.B] In the sentence 'The latter are called outliers...', the antecedent of 'the latter' is ambiguous; G15W50 and G25W100 are actually two devices, not 'the latter' of a list of two items. Please rewrite to name the devices explicitly.
- [Fig. 6 caption] The caption says 'Line indicates the linear fit to normal data.' Please define what 'normal data' means and mark the excluded points explicitly in the plot (e.g., open symbols) to make the outlier treatment transparent.
Circularity Check
No circularity: the 1/A scaling is an empirical correlation, not a fitted parameter renamed as a prediction.
full rationale
The paper's central claim, that normalized resistance noise SR/R2 correlates with 1/A where A = G × W × (Nfingers−1), is an empirical scaling observation extracted from nine measured devices. No parameter is fitted to a subset of the noise data and then used to predict another closely related noise quantity; the linear fit in Fig. 6 is a direct description of the measured SR/R2 versus 1/A points. The use of 'standard MOSFET LFN noise theory' is interpretive and does not enter the derivation as an input that already contains the claimed area dependence. There are no load-bearing self-citations: all cited works are external, and none is invoked as a uniqueness theorem or as the source of the scaling law being 'derived'. The post-hoc exclusion of G15W50 and G25W100 as outliers is a data-selection concern about robustness of the correlation, not a circularity: the paper does not define the outliers in terms of the conclusion, nor does it use the excluded points as an input to a prediction. Because the derivation chain contains no step where a result reduces by definition or by fitted input to its own premise, the appropriate circularity score is 0.
Assumptions & free parameters
free parameters (2)
- slope of SR/R2 vs 1/A
- per-device slope of sqrt(SI) vs V
assumptions (4)
- domain assumption Standard 1/f noise theory (Hooge/McWhorter) applies to these graphene devices.
- domain assumption The IDE device can be modeled as (Nfinger-1) bottom-gate GFETs in series with floating bottom gate.
- domain assumption The active graphene area is A = G x W x (Nfingers-1).
- domain assumption Self-heating begins at 20 uA and is width-dependent.
Cite this review
Pith. "Pith review of 1/f and Random Telegraph Noise of Single-Layer Graphene Devices with Interdigitated Electrodes." pith.science (2026). https://pith.science/paper/LTVTATUE
@misc{pith2026250201348,
author = {Pith},
title = {Pith review of: 1/f and Random Telegraph Noise of Single-Layer Graphene Devices with Interdigitated Electrodes},
year = {2026},
howpublished = {\url{https://pith.science/paper/LTVTATUE}},
note = {Machine review of arXiv:2502.01348}
}
read the original abstract
Single-layer Graphene (SLG) is a promising material for sensing applications. High performance graphene sensors can be achieved when Interdigitated Electrodes (IDE) are used. In this research work, we fabricated SLG micro-ribbon (GMR) devices with IDE having different geometric parameters. 1/f noise behavior was observed in all of the examined devices, and in some cases random telegraph noise (RTN) signals suggesting that carrier trapping/de-trapping is taking place. Our experimental results indicate that the geometrical characteristics can have a crucial impact on device performance, due to the direct area dependence of the noise level.
Figures
Reference graph
Works this paper leans on
-
[1]
F. Catania et al., “A Review on Recent Advancements of Graphene and Graphene-Related Materials in Biological Applications,” Applied Sciences, vol. 11, no. 2, p. 614, Jan. 2021, doi: 10.3390/app11020614
-
[2]
Practical Chemical Sensors from Chemically D erived Graphene,
J. D. Fowler, M. J. Allen, V. C. Tung, Y. Yang, R. B. Kaner, and B. H. Weiller, “Practical Chemical Sensors from Chemically D erived Graphene,” ACS Nano, vol. 3, no. 2, pp. 301–306, Feb. 2009, doi: 10.1021/nn800593m
-
[3]
Low-frequency noise in gas sensors: A review,
W. Shin et al., “Low-frequency noise in gas sensors: A review,” Sensors and Actuators B: Chemical, vol. 383, p. 133551, May 2023, doi: 10.1016/j.snb.2023.133551
-
[4]
J. Nah et al., “Electrical and Low Frequency Noise Characterization of Graphene Chemical Sensor Devices Having Different Geometries,” Sensors, vol. 22, no. 3, p. 1183, Feb. 2022, doi: 10.3390/s22031183
-
[5]
Low -frequency 1/f noise in graphene devices,
A. A. Balandin, “Low -frequency 1/f noise in graphene devices,” Nature Nanotech , vol. 8, no. 8, pp. 549 –555, Aug. 2013, doi: 10.1038/nnano.2013.144
-
[6]
Vertical graphene spin valve with Ohmic contacts,
J. Meng, J.-J. Chen, Y. Yan, D.-P. Yu, and Z.-M. Liao, “Vertical graphene spin valve with Ohmic contacts,” Nanoscale, vol. 5, no. 19, p. 8894, 2013, doi: 10.1039/c3nr03168c
-
[7]
Role of Joule Heating on Current Saturation and Transient Behavior of Graphene Transistors,
S. Islam, Z. Li, V. E. Dorgan, M. -H. Bae, and E. Pop, “Role of Joule Heating on Current Saturation and Transient Behavior of Graphene Transistors,” IEEE Electron Device Lett., vol. 34, no. 2, pp. 166–168, Feb. 2013, doi: 10.1109/LED.2012.2230393
-
[8]
1 / f noise in van der Waals materials and hybrids,
P. Karnatak, T. Paul, S. Islam, and A. Ghosh, “1 / f noise in van der Waals materials and hybrids,” Advances in Physics: X, vol. 2, no. 2, pp. 428–449, Mar. 2017, doi: 10.1080/23746149.2017.1314192
Show all 11 references
-
[9]
Materials Science Challenges to Graphene Nanoribbon Electronics,
V. Saraswat, R. M. Jacobberger, and M. S. Arnold, “Materials Science Challenges to Graphene Nanoribbon Electronics,” ACS Nano, vol. 15, no. 3, pp. 3674–3708, Mar. 2021, doi: 10.1021/acsnano.0c07835
2021 doi
-
[10]
Impacts of graphene/SiO2 interaction on FET mobility and Raman spectra in mechanically exfoliated graphene films,
K. Nagashio, T. Yamashita, J. Fujita, T. Nishimura, K. Kita, and A. Toriumi, “Impacts of graphene/SiO2 interaction on FET mobility and Raman spectra in mechanically exfoliated graphene films,” in 2010 International Electron Devices Meeting , San Francisco, CA, USA: IEEE, p. 23...
2010
-
[2748]
V.REFERENCES
which are funded by the Operational Programme NSRF 2014-2020 and the Hellenic Foundation of Research and Innovation (HFRI) respectively. V.REFERENCES
2014
Reviewed August 9, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.