---
title: Constriction Thermal Resistance
url: https://www.emergentmind.com/topics/constriction-thermal-resistance
type: topic
---

# Constriction Thermal Resistance

Constriction thermal resistance is the temperature drop produced by nonuniform heat flow through a limited contact region, an interface populated by discrete asperities, or a geometrically narrowed heat path. In its most generic form, thermal contact resistance is defined as \(R_c=\Delta T/Q\), where \(\Delta T\) is the temperature drop across the interface and \(Q\) is the total heat current. In area-normalized form, it is also written as \(R_c=\Delta T/q=\Delta T\,A_a/Q\), with \(q=Q/A_a\) and \(A_a\) the apparent contact area. Across the literature, constriction resistance appears in classical spreading problems, rough-surface contact mechanics, nanoscale interfaces such as carbon nanotube supports, thermal interface materials, and Joule-heated nanoconstrictions in two-dimensional materials [1007.4364] [1809.09277] [2106.02180] [2106.05129].

## 1. Classical definition and canonical geometries

The classical origin of constriction thermal resistance is the additional temperature drop generated when heat enters or leaves a body through a contact area smaller than the surrounding heat-spreading volume. For a circular contact of radius \(a\) on a semi-infinite substrate of thermal conductivity \(k\), the spreading resistance of one body is
\[
R_{sp}=\frac{\Delta T}{Q}=\frac{1}{4ka}.
\]
For two identical semi-infinite media in series, the total constriction resistance doubles to
\[
R_{sp}=\frac{1}{2ka}.
\]
The same two-body result appears in single-spot contact models for identical isotropic solids, where the thermal resistance of a single circular contact is written as
\[
R_{c,\mathrm{spot}}=\frac{1}{2ak}.
\]
These expressions establish the central scaling: reducing the contact radius increases the temperature drop per unit heat current [1007.4364] [2106.02180].

A generalized representation introduces geometry explicitly:
\[
R_c=\frac{1}{2kA}f(a/b),
\]
where \(A\) is the apparent contact area, \(b\) is a characteristic dimension such as layer thickness, and \(f(a/b)\) is a dimensionless spreading-resistance function. This form is used when the interface is not well approximated by a single circular contact on a semi-infinite body, for example in finite-thickness layers or cylindrical geometries. It emphasizes that constriction resistance is not solely a materials parameter; it is jointly controlled by conductivity, contact size, and the surrounding geometry [1007.4364].

A related device-level definition appears in Joule-heated nanoconstrictions, where the quantity of interest is the thermal resistance associated with the hottest point of a constricted region:
\[
R_{\rm th}=\frac{T_{\max}-T_0}{P}.
\]
Here \(T_{\max}\) is the peak temperature, \(T_0\) the ambient temperature, and \(P=IV\) the total Joule-heating power. This definition extends the constriction concept from a contact interface to a geometrically narrowed heat-flow path [2106.05129].

## 2. Rough interfaces, asperity ensembles, and contact mechanics

For nominally flat solids, true intimate contact occurs only at discrete asperities. In a simple multiple-asperity model with \(N\) identical circular contacts of radius \(a\) distributed over an apparent area \(A\), the net constriction resistance is
\[
R_c=\frac{1}{2akN}=\frac{1}{2kA_r},
\]
where \(A_r=N\pi a^2\) is the real contact area. Since \(A_r\ll A\) in most practical interfaces, the difference between apparent and real contact area is the primary source of large contact resistance [2106.02180].

More elaborate statistical models incorporate surface roughness and deformation mechanics. The summarized framework includes Gaussian asperity-height distributions characterized by rms roughness \(\sigma\) or areal rms roughness \(S_q\), together with elastic Hertzian or plastic asperity deformation. Under elastic Hertzian contact of identical asperities,
\[
A_r \propto P^{2/3}R^{1/3}/E^*,
\]
where \(P\) is the applied pressure, \(R\) the asperity radius of curvature, and \(E^*\) the composite modulus. In the plastic limit, the real contact area scales as \(A_r\propto P/H\), with \(H\) the hardness of the softer material. A closely related force-balance statement used in rough-contact theory is \(A_r/A_a \simeq P/H\) [1809.09277] [2106.02180].

For two nominally flat, microscopically rough solids in vacuum, the Cooper–Mikic–Yovanovich correlation gives the constriction resistance as
\[
R_{cc}
=\frac{2\,\sigma\,\nu}{\pi\,k_s}\,
\exp\!\Bigl(\frac{\eta^2}{2}\Bigr)\,
\bigl[1-\tfrac12\,\mathrm{erfc}(\eta/\sqrt2)\bigr]^{-1.5},
\]
where
\[
k_s=\frac{2k_1k_2}{k_1+k_2},
\qquad
\nu=\sqrt{m_1^2+m_2^2},
\qquad
\sigma=\sqrt{\sigma_1^2+\sigma_2^2},
\]
and
\[
\eta=-\sqrt2\,\mathrm{erfc}^{-1}(2P/H).
\]
Here \(k_s\) is the harmonic-mean conductivity, \(\nu\) the effective absolute slope, and \(\eta\) the dimensionless load parameter. For Gaussian surfaces, the empirical slope–roughness relation is
\[
\nu(\sigma)\simeq
\begin{cases}
0.124\,\sigma^{-0.743}, & \sigma\le 1.6\,\mu\mathrm{m},\\
0.076\,\sigma^{-0.520}, & \sigma>1.6\,\mu\mathrm{m}.
\end{cases}
\]
The mean-plane separation is approximated by
\[
Y \simeq 1.185-\ln[3.132-(P/H)^{0.547}]
\]
in micrometers. These expressions couple interfacial mechanics and transport by making contact resistance a function of roughness, slope, load, hardness, and mean separation rather than a fixed interfacial constant [1809.09277].

## 3. Conductive constriction and near-field radiative transfer as parallel channels

In vacuum, rough interfaces exchange heat through two parallel channels: conductive constriction through real microcontacts and radiative transfer through noncontact gaps. The combined resistance is therefore
\[
\frac{1}{R_{\mathrm{total}}}=\frac{1}{R_{cc}}+\frac{1}{R_x},
\qquad
R_{\mathrm{total}}=\bigl(R_{cc}^{-1}+R_x^{-1}\bigr)^{-1}.
\]
This formulation departs from the common simplification that thermal contact resistance is purely conductive. In the summarized treatment, the radiative channel is evaluated with fluctuational electrodynamics, in which the spectral heat flux between two planar, semi-infinite media separated by a vacuum gap contains both propagating and evanescent contributions, and the total radiative heat flux per unit area is
\[
Q_r=\int_0^\infty \frac{d\omega}{2\pi}\,q_{\omega,12},
\qquad
R_x=\frac{\Delta T}{Q_r}.
\]
The evanescent contribution is the mechanism responsible for near-field enhancement below submicron separations [1809.09277].

The parameter regime in which radiation becomes important is explicit. The effective rms roughness spans approximately \(1\,\mathrm{nm}\) to \(10\,\mu\mathrm{m}\), while the mean-plane separation \(Y\) rises from a few nanometers at high pressure to more than \(100\,\mathrm{nm}\) at low pressure. Near-field radiation dominates when \(Y\lesssim 0.1\,\mu\mathrm{m}\); for \(\sigma_{\mathrm{eff}}<0.1\,\mu\mathrm{m}\), it can exceed conductive heat transfer even at room temperature. For amorphous \(\mathrm{Al_2O_3}\), strong surface phonon polaritons at \(\omega\approx 1\text{–}2\times 10^{14}\,\mathrm{rad/s}\) produce radiative resistances orders of magnitude below the blackbody limit for \(d<10\,\mathrm{nm}\). Aluminum shows a more modest near-field effect at sub-\(100\,\mathrm{nm}\) gaps [1809.09277].

Illustrative numerical trends reinforce the point. At \(\sigma_{\mathrm{eff}}=1\,\mathrm{nm}\), \(\Delta T=10\,\mathrm{K}\), and \(P=10\,\mathrm{MPa}\), the conductive heat transfer coefficient is approximately \(h_{cc}\approx 10^7\,\mathrm{W/m^2K}\), while \(h_{\mathrm{rad}}(\mathrm{Al})\approx 10^5\text{–}10^6\,\mathrm{W/m^2K}\) and \(h_{\mathrm{rad}}(\mathrm{Al_2O_3})\approx 10^7\,\mathrm{W/m^2K}\). At \(P<1\,\mathrm{MPa}\), where \(Y>10\,\mathrm{nm}\), \(h_{cc}\) falls to \(10^5\text{–}10^3\,\mathrm{W/m^2K}\), while \(h_{\mathrm{rad}}(\mathrm{Al_2O_3})\) remains above \(10^4\,\mathrm{W/m^2K}\), so radiative transfer can dominate. In micro- and nanoscale interfaces with \(\sigma_{\mathrm{eff}}\sim 10\text{–}100\,\mathrm{nm}\) and \(P<1\,\mathrm{MPa}\), neglecting near-field radiation underestimates heat transfer by up to \(50\text{–}100\%\) [1809.09277].

## 4. Carbon nanotube contacts: switching between weak and strong thermal coupling

The carbon-nanotube study provides a direct nanoscale demonstration that constriction resistance can be tuned over more than two orders of magnitude by altering how a nanotube is coupled to its support. In the weakly coupled, “unanchored” configuration, multiwall carbon nanotubes with diameter approximately \(14\,\mathrm{nm}\) are drop-cast onto free-standing, electron-transparent SiNx membranes. Pd heater wires and a single Pd contact pad are defined by e-beam lithography, but the nanotube lies on the SiNx only by van der Waals forces, touching tangentially along a tiny arc. The effective contact width, and therefore the contact area, is minimal. In the strongly coupled, “Pd-anchored” configuration, a second Pd pad is deposited on the free end of a similar nanotube of diameter approximately \(17\,\mathrm{nm}\). Cross-sectional imaging shows the nanotube embedded beneath the Pd pad over roughly half its circumference, increasing the contact width \(w\) and giving \(A\approx (\pi d/2)w\) [1007.4364].

The thermal contact resistance was extracted by in-situ electron thermal microscopy using a discontinuous indium layer on the back side of the SiNx membrane as a distributed thermometer. Each In island melts abruptly at \(T_m=156.6\,^\circ\mathrm{C}\). A slowly ramped DC voltage was applied across the Pd heater wire, bright-field TEM was used to record the melting voltage \(V_{\mathrm{melt}}\) for each island, and the resulting melting-voltage profile was compared between the CNT side and the opposite side. For each sample type, approximately 30 In islands were selected in two symmetric regions, linear regression of melting voltage versus distance provided an expectation value and 95% confidence interval, and COMSOL finite-element modeling of
\[
\nabla\!\cdot(k\nabla T)+P/V=0
\]
with \(R_c\) as the sole adjustable parameter yielded the best-fit contact resistance [1007.4364].

The measured values were sharply different. For weak coupling, no detectable asymmetry in melting voltages was observed within the measurement resolution, and modeling imposed the lower bound \(R_c>250\,\mathrm{K\!\cdot\!m/W}\) at 95% confidence; the actual value could be much higher. For strong coupling, clear melting-voltage asymmetry was observed, and the extracted value was
\[
R_c=4.2\,\mathrm{K\!\cdot\!m/W}
\]
with a 95% confidence interval of \(+5.6/-2.1\,\mathrm{K\!\cdot\!m/W}\). The study attributes the reduction to contact-area scaling, noting that for phonon-boundary-limited transport \(R_c\propto 1/A\), and reports quantitative agreement with the observed \(>60\times\) reduction. It also states that the Pd–CNT contact is dominated by phonon coupling, with the electronic contribution estimated via the Wiedemann–Franz law as less than \(1\%\) of the total conductance. A central implication is that high intrinsic CNT conductivity does not by itself ensure low interfacial resistance; atomic bonding and effective contact width can dominate the thermal behavior [1007.4364].

## 5. Graphene thermal interface materials and the roughness–loading trade-off

In noncured graphene thermal interface materials, constriction resistance appears as one component of the total sandwich resistance
\[
R_{\mathrm{tot}}=\frac{\mathrm{BLT}}{K_{\mathrm{TIM}}}+2R_c,
\]
where BLT is the bond-line thickness and \(K_{\mathrm{TIM}}\) is the bulk thermal conductivity of the composite. The summarized semi-empirical roughness model is
\[
R_{c1+2}=2R_c
=
c\cdot \frac{S_q^n}{K_{\mathrm{TIM}}\,G^n\,P^{\,n-1}},
\]
with \(S_q\) the areal RMS roughness, \(G=\sqrt{G'^2+G''^2}\) the complex shear modulus, and \(P\) the applied pressure. This form captures the way roughness, TIM stiffness, conductivity, and pressure jointly determine the interfacial temperature drop [2106.02180].

The experimental system consisted of graphene/few-layer-graphene fillers dispersed in silicone oil at loading fractions \(w=0\text{–}40\,\mathrm{wt\%}\). Copper plates of size \(1''\times 1''\times 1.09\,\mathrm{mm}\) were polished with 180-grit SiC paper for \(0\text{–}3.5\,\mathrm{min}\) at \(100\,\mathrm{RPM}\), producing \(S_q=0.05\), \(1.2\), \(2.5\), and \(3.1\,\mu\mathrm{m}\). Thermal measurements used the ASTM D5470 steady-state method with an industrial LongWin tester, very flat steel source and sink plates of roughness \(\lesssim 5\,\mathrm{nm}\), applied pressure \(P=0.55\,\mathrm{MPa}\) \((80\,\mathrm{psi})\), and \(80\,^\circ\mathrm{C}\) for \(40\,\mathrm{min}\). BLT was controlled between \(25\) and \(350\,\mu\mathrm{m}\) using plastic shims, and \(R_c\) was obtained from the intercept of the measured \(R_{\mathrm{tot}}\) versus BLT line [2106.02180].

The principal result is that contact resistance depends on graphene loading non-monotonically. It decreases from approximately \(0.90\times 10^{-3}\,\mathrm{K\,cm^2/W}\) at \(w=5\,\mathrm{wt\%}\) to a minimum of \((0.45\pm 0.02)\times 10^{-3}\,\mathrm{K\,cm^2/W}\) at \(w=15\,\mathrm{wt\%}\), then slowly rises to approximately \(0.65\times 10^{-3}\,\mathrm{K\,cm^2/W}\) at \(w=40\,\mathrm{wt\%}\). Over the same loading range, \(K_{\mathrm{TIM}}\) increases super-linearly up to \(w\approx 10\text{–}15\,\mathrm{wt\%}\), consistent with percolation, and saturates at approximately \(4.2\,\mathrm{W/m\!\cdot\!K}\) by \(w=40\,\mathrm{wt\%}\). The study concludes that the combined evolution of \(K_{\mathrm{TIM}}\) and \(R_c\) yields an optimum loading around \(15\,\mathrm{wt\%}\), where \(K_{\mathrm{TIM}}\approx 2.5\,\mathrm{W/m\!\cdot\!K}\) [2106.02180].

Surface roughness imposes a distinct penalty. At fixed loading, increasing \(S_q\) from \(0.05\) to \(1.2\,\mu\mathrm{m}\) roughly doubles \(R_c\). For \(w=15\%\), \(R_c\) rises from \(0.45\times 10^{-3}\) to \(0.85\times 10^{-3}\,\mathrm{K\,cm^2/W}\); for \(w=30\%\), it rises from \(0.55\times 10^{-3}\) to \(1.0\times 10^{-3}\,\mathrm{K\,cm^2/W}\). The study fits a nearly linear relation,
\[
R_c(w,S_q)\simeq R_{c0}(w)+\beta(w)S_q,
\]
and gives, for \(w=15\,\%\),
\[
R_c(15\%,S_q)\simeq (0.45\pm 0.02)\times 10^{-3}\,\mathrm{K\,cm^2/W}
+(0.36\pm 0.04)\times 10^{-3}\,\mathrm{K\,cm/W}\cdot S_q(\mu\mathrm{m}),
\]
and for \(w=30\,\%\),
\[
R_c(30\%,S_q)\simeq (0.55\pm 0.03)\times 10^{-3}\,\mathrm{K\,cm^2/W}
+(0.40\pm 0.05)\times 10^{-3}\,\mathrm{K\,cm/W}\cdot S_q(\mu\mathrm{m}).
\]
For rougher copper with \(S_q=3.1\,\mu\mathrm{m}\), choosing \(\mathrm{BLT}\approx 300\,\mu\mathrm{m}\) at \(w=30\%\) keeps \(R_{\mathrm{tot}}\approx 2\times 10^{-3}\,\mathrm{K\,cm^2/W}\), matching smoother \(S_q=1.2\,\mu\mathrm{m}\) cases. The study therefore treats constriction resistance not as a fixed penalty, but as a quantity that can be traded against conductivity and bond-line thickness by formulation and surface preparation [2106.02180].

## 6. Nanoconstrictions in two-dimensional materials with self-heating

In monolayer two-dimensional materials, constriction thermal resistance can be formulated from the steady-state heat equation with substrate heat loss and Joule self-heating:
\[
\nabla^2T(x,y)+\frac{T_0+T_J(x,y)-T(x,y)}{L_H^2}=0,
\]
subject to \(T=T_0\) at the longitudinal contacts and \(\partial_n T=0\) at free edges. The thermal-healing length is \(L_H=\sqrt{\kappa R_B}\), where \(\kappa\) is the 2D thermal conductivity and \(R_B\) the thermal boundary resistance to the substrate. Within this framework, a constant-width nanowire and a wedge-shaped bow-tie constriction admit approximate one-dimensional descriptions, and the bow-tie geometry leads to a modified Bessel equation of zero order [2106.05129].

The constriction thermal resistance is defined as
\[
R_{\rm th}=\frac{T_{\max}-T_0}{P}.
\]
For the nanowire, the maximum temperature occurs at the center, and the leading width dependence is
\[
R_{\rm th}^{\rm nw}\propto \frac{1}{w}.
\]
For fixed current, the peak temperature diverges as \(\Delta T\propto I^2/w^2\). For the bow-tie constriction, the device resistance scales as
\[
R_{\rm bowtie}\approx \frac{L}{\sigma W}\ln\!\frac{W}{w}
\]
for \(w\ll W\), and the constriction thermal resistance behaves as
\[
R_{\rm th}^{\rm bt}\propto \frac{1}{w^2\ln(W/w)}\times \frac{W}{L}.
\]
The corresponding peak temperature divergence is only logarithmic in \(\ln(W/w)\) for fixed current. The abstract summarizes the comparison succinctly: at fixed potential difference, the temperature remains finite as the width shrinks, whereas for fixed current it diverges, logarithmically with width for the bow tie and with an inverse-square dependence in a nanowire [2106.05129].

The geometrical distinction is therefore central. A constant-width nanowire funnels current and heat into a fixed narrow area, while a wedge-shaped bow tie progressively widens away from the constriction, reducing the severity of the hotspot. The material parameters enter through \(L_H=\sqrt{\kappa R_B}\): larger \(\kappa\) increases lateral heat spreading, lower \(R_B\) improves heat sinking into the substrate, and both lower \(R_{\rm th}\). A plausible implication is that “constriction thermal resistance” in device physics is best understood as a coupled electrothermal quantity, not merely a local interfacial property, because it depends simultaneously on geometry, substrate coupling, and the bias condition [2106.05129].

## 7. Cross-scale interpretation and recurring misconceptions

Across classical contacts, rough interfaces, nanotube supports, TIM sandwiches, and 2D nanoconstrictions, a common structure emerges: constriction thermal resistance increases when heat is forced through a smaller effective area, a smaller real contact fraction, or a more weakly coupled thermal escape path. In the nanotube case, increasing contact width from a tiny tangential arc to roughly half the circumference lowers \(R_c\) by over two orders of magnitude. In rough-interface theory, increasing load or reducing roughness enlarges the real contact area and lowers \(R_{cc}\). In graphene TIMs, minimizing total resistance requires balancing contact resistance, bulk conductivity, and bond-line thickness rather than maximizing filler loading alone. In 2D nanoconstrictions, changing the constriction shape from a straight nanowire to a bow tie alters the divergence law of the hotspot temperature [1007.4364] [1809.09277] [2106.02180] [2106.05129].

Several misconceptions are explicitly contradicted by the cited studies. One is that thermal contact resistance is purely a materials constant; the summarized models instead make it a function of contact radius, apparent and real area, roughness, slope, pressure, hardness, thickness, and substrate coupling. A second is that interfacial heat transfer in vacuum is purely conductive; the rough-interface analysis shows that near-field radiation can dominate for \(Y\lesssim 0.1\,\mu\mathrm{m}\), especially in dielectric systems with surface phonon polaritons. A third is that higher intrinsic conductivity or higher filler content necessarily produces lower interface resistance; the CNT study shows that poor bonding can suppress heat transfer despite the high conductivity of the nanotube, and the graphene TIM study shows a minimum \(R_c\) near \(15\,\mathrm{wt\%}\) rather than at the highest loading [1809.09277] [1007.4364] [2106.02180].

The collected results therefore position constriction thermal resistance as a multiscale transport problem. At one end are continuum spreading solutions such as \(1/(4ka)\) and \(1/(2ka)\); at another are roughness- and mechanics-controlled microcontacts; at another are nanoscale phonon-coupled interfaces; and at the device level are electrothermal constrictions governed by geometry and substrate heat leakage. This suggests that accurate prediction and control of constriction resistance generally require simultaneous treatment of contact mechanics, interfacial coupling, and thermal geometry rather than reliance on a single nominal contact area or a single bulk conductivity value.

Source: https://www.emergentmind.com/topics/constriction-thermal-resistance