Skip to content

Thermally Modulated Specular Phonon Transport in a High‐Debye‐Temperature Diamond Nanobeam

FieldValue
JournalAdvanced Science
Published2026-03-04
AuthorsSeohee Jang, Seung-Woo Jeon, Takuma Shiga, Jeeyoung Shin, Sangwook Han
DOI10.1002/advs.202523242

During cooling from 300 K to about 140 K, measured nanobeam conductivity exceeded a fully diffuse model by up to about 40%. The authors interpret this excess as a growing specular phonon contribution. At 300 K, the nanobeams measured 277 ± 24 W m⁻¹ K⁻¹, about one tenth of the bulk diamond value.

Temperature changes the boundary scattering regime

Section titled “Temperature changes the boundary scattering regime”

The measured conductivity fell with cooling and followed a reported scaling exponent of 1.07. The measured range extended from room temperature to about 140 K. Boundary scattering controlled the mean free path because bulk diamond supports much longer phonon paths than silicon.

The calculated mean free path spectrum explains this contrast. At 300 K, diamond’s heat-carrying phonons occupied a narrower range than silicon. At 100 K, the diamond spectrum was still narrower and extended to much longer paths than silicon. The authors associate this concentrated spectrum with stronger suppression in nanostructures during cooling. As temperature fell, the fully diffuse prediction increasingly undershot the measurement. They attributed this growing gap to specular reflection. The near-linear decline was consistent with a boundary-limited mean free path. Together, these observations indicate a temperature-driven change in the scattering regime of diamond nanobeams.

Nanobeam geometry and electrothermal measurement

Section titled “Nanobeam geometry and electrothermal measurement”

The beams were fabricated from high-purity single-crystal diamond aligned along [100]. Each had an isosceles triangular cross section. Its width was about 185 nm and its height was 120 nm. Boron and nitrogen defect densities were each below 1 ppb. Seven beams had the same cross-sectional area but lengths ranging from about 8.4 to 34.5 µm. A nanomanipulator placed each beam between suspended islands. The authors rotated each beam to vary the heat-flow length without changing its section.

Electrothermal measurements used two suspended islands. A serpentine heater generated heat on one island, and resistive thermometry measured both island temperatures. The measured resistance included the beam and contact resistance at the electrodes. For constant cross section, the authors used the slope of resistance versus beam length. That slope equals 1 divided by kA, so the contact term cancels. The cross-sectional area and inverse slope then gave the intrinsic beam conductivity. Using several lengths prevented contact resistance from dominating the reported conductivity.

Diffuse boundary model matches room temperature

Section titled “Diffuse boundary model matches room temperature”

Transport was calculated with the Boltzmann transport equation and ab initio phonon data. First-principles calculations supplied internal mean free paths, spectral heat capacity, and group velocity. Monte Carlo simulation supplied the boundary mean free path. The conductivity was integrated from the spectral product of heat capacity, group velocity, and nanobeam mean free path. Matthiessen’s rule combined internal and boundary mean free paths. The model treated bulk phonon properties separately from geometry-dependent boundary scattering. This separation combined first-principles bulk properties with a Monte Carlo boundary mean free path in the nanobeam calculation. For the triangular sample geometry, the fully diffuse model gave a boundary mean free path of about 91 nm. The resulting prediction was within about 9% of the measured room-temperature conductivity. This agreement made diffuse scattering a useful baseline at 300 K.

Cooling reveals a non-diffuse boundary contribution

Section titled “Cooling reveals a non-diffuse boundary contribution”

The authors added wavelength-dependent specular scattering with the Ziman model. Its scattering probability depends on phonon wavelength and root-mean-square surface roughness. In each Monte Carlo trial, a phonon continued without diffuse scattering when a random number fell below its wavelength-specific specular probability. The specular models used roughness values from 1.5 to 2 Å. They reproduced the measured excess over the fully diffuse prediction across the temperature range.

Transmission electron microscopy gave a surface roughness of about 0.95 ± 1.03 Å. The authors report agreement between this value and the model-derived roughness. A cited silicon nanofilm study reported 2.5 ± 0.5 Å, which the authors describe as a comparable roughness range. The measured excess was a difference from the diffuse prediction, not a direct observation of specular reflection. The assignment to specular reflection came from comparison with the Ziman model and surface roughness analysis.

Diamond and silicon differ in spectral temperature sensitivity

Section titled “Diamond and silicon differ in spectral temperature sensitivity”

Diamond has a Debye temperature of 2230 K, compared with 645 K for silicon. At a fixed temperature, a lower temperature-to-Debye-temperature ratio leaves fewer high-frequency phonons thermally excited. The heat-carrying spectrum then concentrates in the long-wavelength regime.

At 300 K, the calculated wavelength contributions were similar for diamond and silicon. At 100 K, diamond heat was carried within a much narrower wavelength range. Silicon retained a broader spectral range, which made its boundary specularity less temperature-sensitive in the cited work. The paper associates this restricted diamond spectrum with stronger temperature sensitivity of short-wavelength phonons. They argue that the contracting spectrum makes boundary selection more temperature-sensitive in diamond.

Silicon nanowires with circular sections about 115 nm in diameter showed a linear conductivity decrease below about 100 K. The diamond beams had a hydraulic diameter of about 115 nm and showed the linear trend across the measured range. A Boltzmann model for silicon nanowires agreed with measurements down to cryogenic temperatures. The diamond model instead developed a growing low-temperature gap. The authors therefore place the diffuse-to-specular transition at a higher temperature than in silicon nanowires of similar diameter. These results led them to propose diamond for manipulating wave-like phonon conduction above 100 K. The study shows passive temperature-dependent specularity rather than active temperature control.

Limits and implications for nanobeam design

Section titled “Limits and implications for nanobeam design”

The authors state that deviations between measurements and the fully diffuse model from about 250 to 300 K may have several causes. These causes include errors in the ab initio bulk properties and use of Matthiessen’s rule. The low-temperature specularity result depends on the Ziman boundary model, Monte Carlo treatment, and assumed triangular geometry. The specimens also share one crystal orientation and one cross-sectional geometry. The evidence for wave-like transport comes from transport and model analysis, not a direct wave-packet measurement. The room-temperature fit should not be extrapolated to cooling with the same diffuse assumption. The proposed active modulation remains a suggestion. The transport data support a temperature-dependent specularity interpretation within the stated geometry.

These limits constrain the device claim, although the measured boundary-scattering trend remains clear. The result applies to [100]-oriented triangular beams with the stated surface roughness. A comparison with another diamond nanobeam should include its orientation, cross section, and surface roughness. Thermal models must include boundary scattering and temperature-dependent specularity to reproduce the measured cooling trend. Bulk conductivity alone does not predict the measured nanobeam conductivity. Device engineers should treat surface roughness as a thermal parameter in this regime. The authors suggest a significant specular contribution below 200 K. This temperature provides a concrete reference for interpreting temperature-dependent diamond nanobeam measurements.

The paper requires high-purity single-crystal diamond aligned along [100] for nanobeam phonon-transport measurements. 6CCVD’s diamond materials catalog lists single-crystal diamond grown by MPCVD. The engineering support page can address custom fabrication support for the beam geometry.

View Original Abstract

ABSTRACT While the interplay between phonon and boundary dictates thermal transport at the nanoscale, the spectral manipulation of phonon‐boundary scattering is insufficiently demonstrated yet. Here, we choose a single‐crystal diamond nanobeam, a material with one of the highest Debye temperature materials, to investigate the impact of modulated phonon‐boundary scattering with a temperature knob. The thermal conductivity of nanobeams is measured from room temperature down to ∼140 K, and we find that the value monotonically decreases scaling with T ∼1.07 . Compared to the model prediction based on the Boltzmann transport equation combined with ab initio calculation, we find that the experimental data increasingly deviate from the model prediction with diffuse phonon scattering as the temperature decreases. The deviation indicates the increasing portion of non‐diffuse phonon‐boundary scattering. Our analysis indicates that specular phonon‐boundary scattering is more sensitive to temperature in single‐crystal diamond compared to that of silicon. This work suggests that the diamond is a potential material platform to manipulate wave‐like phonon conduction above 100 K.