Skip to content

Questioning the near-intrinsic thermal conductivity of suspended graphene membranes fabricated via a cyclododecane-based transfer method

FieldValue
JournalNature Communications
Published2026-03-27
AuthorsMohsen Moazzami Gudarzi, Seyed Hamed Aboutalebi
DOI10.1038/s41467-026-70213-9

A reanalysis finds that the thermal conductivity calculated from the published Raman thermometry data is more than seven times smaller than the reported 4914 W/m K value at 338 K. The critique identifies absorbed laser power and heat flow through the support grid as unresolved parts of the conversion from temperature rise to conductivity.

Reported value and disputed interpretation

Section titled “Reported value and disputed interpretation”

The challenged result came from Wang et al.’s 2024 study of suspended graphene made by chemical vapor deposition and transferred with a cyclododecane-based method. They reported 4914 W/m K near room temperature and attributed it to an optimized, “contamination-free” transfer. The present Nature Communications paper argues that the published evidence does not substantiate that attribution.

The original measurement and the disputed property are different quantities. Raman thermometry measured a temperature response under laser heating. Thermal conductivity was then inferred with a model that included incident and absorbed heat. The critique accepts the temperature data as the starting point but questions the assumptions used to convert it.

That distinction matters because 4914 W/m K was presented as evidence for high-quality, near-intrinsic suspended graphene. The reanalysis does not replace it with a new direct conductivity measurement. It shows that the same temperature data can lead to a much lower value once absorbed power is stated.

The original experiment heated suspended CVD graphene with a laser and tracked temperature through Raman thermometry. Its thermal impedance treatment relates temperature rise to the heat input at each laser power. The present analysis uses the reported temperature and power data from Supplementary Fig. 18 of the original study.

The conversion in Eq. (1) includes R, the radius of the suspended area, (r_0), the laser beam size, and d, the graphene thickness. A numerical factor α is close to unity. The reanalysis applies the same equation to the published data.

Sample details remain material to that equation. The graphene was CVD-grown and transferred onto a silicon nitride grid, following the cyclododecane-based preparation. The grid thickness was not specified in the original work.

The original paper also did not explicitly state the absorbed laser power. Raman temperature rise alone therefore does not provide a complete thermal input for Eq. (1). The critique fills that gap with an absorption estimate rather than an independently reported absorbed-power measurement.

For the recalculation, the authors assume that graphene absorbs about 3% of the light at 532 nm. They describe that value as an upper estimate. With this absorbed power, Eq. (1) gives a conductivity at 338 K that is more than seven times smaller than 4914 W/m K.

The authors state that uncertainty in absorbed power cannot by itself explain a discrepancy of this size. A lower absorption fraction would change the calculated value, but it would not remove the conflict without changing another part of the analysis. The missing absorbed-power value therefore remains central.

Heat loss to air would reduce the calculated conductivity further. The critique says this effect is minimal for graphene, so air loss does not account for the disagreement. The discrepancy therefore remains after the effect is considered.

The temperature dependence provides a second internal check. The 2D Raman band shift varies linearly with incident laser power. The authors interpret this relation as making the temperature-rise-to-power ratio constant, which would imply constant thermal conductivity over the measured range.

The maximum temperature rise in that range is approximately 450 K. A constant ratio would therefore predict a flat conductivity trend, provided the thermal model and absorbed power remain valid.

Figure 3f of the original study instead shows a more than two-fold conductivity decrease as temperature rises from 338 K to 398 K. The critique plots this behavior on a logarithmic scale and identifies a nearly (T^{-5}) trend. These data conflict with the linear power relation in Supplementary Fig. 18b.

Most earlier Raman thermometry measurements of graphene showed a nearly (T^{-1}) trend near room temperature, similar to graphite. The authors also state that phonon-phonon scattering dominates near room temperature, so the reported (T^{-5}) decrease is not expected there.

A temperature-independent inferred conductivity instead points to phonon-defect scattering as the dominant process. The authors use that implication to question the claim that the suspended graphene had high quality and near-intrinsic thermal behavior.

The support grid creates a separate thermal path. The original paper used a silicon nitride grid with unspecified thickness. For the critique’s comparison, the authors assume a typical grid thickness of 100 to 200 nm and a thermal conductivity of about 2 W/m K.

As a reference point in that comparison, monolayer graphene has κ of about 2000 W/m K and a thickness of 0.335 nm. The authors argue that the grid’s thermal impedance is comparable to or greater than that of the graphene. Even with ideal thermal contact, they judge silicon nitride unsuitable as an effective heat sink.

If the grid heats with the graphene during laser exposure, its temperature rise changes the heat flow used in Eq. (1). The authors regard a significant substrate temperature rise as highly probable under these conditions. This makes the original application of Eq. (1) questionable.

They connect that support effect to the apparent low conductivity of 675 W/m K in their analysis. The paper presents silicon nitride heating as the likely explanation, not as a directly measured result. No independent temperature measurement of the grid is reported.

The numerical comparison broadens the challenge to the claimed benchmark. Wang et al. stated that 4914 W/m K exceeded all previous measurements for isotopically pure diamond and graphene. They also placed it above theoretical predictions for pristine, infinitely large graphene sheets.

Figure 1 of the critique plots the reported and recalculated conductivities against spot temperature. It also shows them on a logarithmic scale with literature graphene and graphite data. The recalculated values fall far below the reported 4914 W/m K point. Their calculation adds an explicit 3% absorption assumption that the original paper did not state.

The comparison does not establish an exact corrected conductivity. The authors’ calculation depends on 3% absorption at 532 nm, while the support contribution depends on a grid thickness that the original study did not state. Their main claim is that the available data cannot support the near-intrinsic value.

The critique does not treat Raman thermometry as generally unreliable. It identifies an absorbed-power gap and a conflict between the linear Raman response and the sharply decreasing conductivity.

Implications for material selection and measurement

Section titled “Implications for material selection and measurement”

For engineers selecting diamond thermal material, the paper offers a warning about benchmark values rather than a product comparison. No diamond specimen was tested here. The disputed 4914 W/m K value should not serve as a validated upper benchmark for the cyclododecane-transferred membrane or for diamond selection.

A Raman thermometry comparison involving graphene or diamond should separate heat spreading in the measured membrane from heat rejection through its support. The paper does not report the grid’s thickness or temperature directly. It also does not resolve the inconsistent temperature trends.

The main checks would report the absorbed laser power, characterize the support’s thermal impedance, and apply a consistent temperature model across all power points. These checks follow directly from the issues identified in the critique.

The authors conclude that the published temperature response does not substantiate 4914 W/m K. Their recalculated conductivity is lower under the stated 3% absorption assumption, and its exact value remains tied to the unresolved support response.

View Original Abstract

A recent study by Wang et al. 1 reported a remarkably high thermal conductivity of 4914 W/m K near room temperature (338 K) in suspended graphene synthesized via chemical vapor deposition (CVD). This value notably exceeds all previous measurements for isotopically pure diamond 2 and graphene 3 and even surpasses theoretical predictions for pristine, infinitely large graphene sheets 4 . Wang et al. attribute this exceptional thermal conductivity to an optimized, contamination-free transfer process of their CVD-grown graphene 1 . However, based on our critical analysis of their data, we contend that the evidence presented does not substantiate this claim. Our assessment suggests that the actual thermal conductivity of their sample is substantially lower than reported, raising important questions about the reliability of these findings and the conditions necessary to achieve such high thermal transport in graphene.