Excellent heat dissipation properties of the super-aligned carbon nanotube films

Abstract

Carbon nanotubes (CNTs) and related macroscopic materials could be widely used as thermal management materials because of their high thermal conductivity and light weight at ambient temperature. Meanwhile, heat dissipation properties also play a significant role in applications of thermal management materials. Measurement of the heat dissipation properties of different layers of super-aligned carbon nanotube (SACNT) films has been reported in this work. The equilibrium temperature of a hot copper plate could be reduced about 13.3% by the multilayer SACNT films. This feature indicates that multilayer SACNT films have very good heat dissipation properties, leading to the appearance of novel light-weight and high-performance heat dissipation materials. Furthermore, the radiative heat dissipation and the natural convective heat dissipation of the SACNT films are respectively calculated by the measurement of surface emissivity. It is found that the radiative heat dissipation of multilayer SACNT films play an important role in the total heat dissipation properties. This advantage would be more promising when they are applied in aerospace science and technology. It is further found that the SACNT films have much better natural convective heat dissipation properties than the several-layer graphene. In view of the high thermal conductivity and low density of compacted multilayer SACNT films reported in our previous work, a novel CPU-radiator made of compacted multilayer SACNT films has been presented to pave the way for practical thermal management applications of the compacted SACNT films. In addition to this application, our results reported here also indicate that the SACNT films would have a variety of applications such as shells of portable electronic devices and thermal management coatings of aerospace devices.

---

Introduction

The rapid development of electronic technology with micro-nanoscale-related devices continues to dimensionally scale down with the on-going increase of generated heat.¹ The consequent generated heat could seriously influence the operating performance of related devices, and even damage themselves. Much attention has been paid to the crucial issue of heat removal for the development of integrated electronic industries,² and the thermal phenomenon in low-dimensional materials also shows some novel and intriguing features.³ Moreover, rapid developments of aerospace technology and increasingly popular portable electronic devices such as smartphones and laptops raise the urgent demand for light-weight and high-performance thermal management materials.^4,5 Nowadays, thermal properties and the weight of thermal management materials have posed important obstacles for the rapid development of portable electronic devices.

Individual carbon nanotubes (CNTs), monolayer graphene and their macroscopic materials have very high thermal conductivity and low density at ambient temperature.^4,6–11 A lot of papers have been published to study the thermal conductivity of these novel carbon-based macroscopic materials. In our previous work, based on super-aligned carbon nanotube (SACNT) arrays and films,^12,13 a kind of SACNT buckypaper^4,14 with very high axial thermal conductivities (about 800 W m⁻¹ K⁻¹) and of light weight (about 1.39 g cm⁻³) was fabricated. Recently, Xin et al. reported a kind of graphene fiber¹⁵ with very high thermal conductivity (about 1300 W m⁻¹ K⁻¹) and outstanding mechanical properties. These results show that the novel carbon-based materials could be widely used as high-performance thermal management materials. In addition, it is expected that the thermal management materials could not only conduct heat but also effectively dissipate heat to the surrounding atmosphere. Thus the heat dissipation properties of these novel carbon-based materials also play a critical role in the thermal management applications. Up to now, several works have been published to study the heat dissipation properties of CNTs^16–19 and graphene.^20–22 For examples, Jiang et al. reported the relatively high natural convective heat transfer coefficient (∼60 W m⁻² K⁻¹) of the suspended single-layer SACNT films.¹⁸ The heat dissipation properties of surface nano-coatings consisting of CNTs and graphene have been serially studied by Lin’s group.^16,21,23 Moreover, several factors have been investigated to optimize the cooling performance and efficiency of the nano-coatings.²³ However, few works have studied the total heat dissipation properties of the multilayer SACNT films and the influence of the radiative heat dissipation properties on the total heat dissipation properties. The radiative heat dissipation would be of great importance in a variety of applications such as aerospace technology and infrared imaging technology.

In the present work, the heat dissipation properties and the surface emissivity of different layers of SACNT films have been measured by an as-assembled thermal analyzer. Our results demonstrate that multilayer SACNT films have much better heat dissipation properties than pure copper, which has been widely used as a thermal management material. The cooling efficiency of the multilayer SACNT films could be as high as about 13.3% in comparison with pure copper. In addition, we also calculate the radiative and natural convective heat dissipation properties of different layers of SACNT films, respectively. We find that the radiative heat dissipation could play an important role in the total heat dissipation properties for multilayer SACNT films because of their very high surface emissivity. Furthermore, calculations of radiative and convective heat dissipation properties also indicate that the improvement of total heat dissipation properties of SACNT films is mainly due to the enhancement of radiative heat dissipation properties of SACNT films. It is also found that the SACNT films have much better natural convective heat dissipation properties than the several-layer graphene. More importantly, a novel CPU-radiator made of compacted multilayer SACNT films with a light weight and excellent thermal properties has been presented. In addition to the CPU-radiator, we believe that the compacted SACNT films could be widely applied in some other fields such as the shells of portable electronic devices and the thermal management coatings of aerospace devices.

Experimental methods

Heat dissipation principle

Heat can dissipate from a solid surface to the surrounding air by means of conduction, radiation and convection. By comparison, the conductive heat dissipation could be ignored because of the very low thermal conductivity (k_air ≈ 0.026 W m⁻¹ K⁻¹) of air.^24,25 Thus, at natural conditions, the heat can dissipate from a solid surface to the surrounding air mainly by means of natural convection and radiation. The natural convective heat dissipation could be expressed by the equation:

P_c = h_cA(T_h − T_a) (1)

where P_c, h_c and A are the natural convective heat dissipation power (W), natural convective heat transfer coefficient (W m⁻² K⁻¹) and surface area (m²), respectively. T_h and T_a are the temperatures (K) of a hot surface and ambient air, respectively. The radiative heat dissipation is determined by the Stefan–Boltzmann law, that is:

P_r = εσA(T_h⁴ − T_a⁴) (2)

where P_r, ε and σ are the radiative heat dissipation power (W), surface emissivity (0 < ε < 1), and the Stefan–Boltzmann constant (5.67 × 10⁻⁸ W m⁻² K⁻⁴), respectively. At the steady state, the total heat dissipation power (P) is:

P = P_c + P_r = εσA(T_h⁴ − T_a⁴) + h_cA(T_h − T_a) (3)

For materials with very low surface emissivity, such as pure metals and transparent thin films, P_r is negligible in comparison to P_c. In this case, the radiative heat dissipation could be ignored.^20,26 However, when the surface emissivity values of the materials are big enough (ε > 0.90), the radiative heat dissipation power could reach about 30% of the total heat dissipation power.^16,24 Therefore, radiative cooling could be an effective way to improve the thermal dissipation properties.¹⁶ Furthermore, the radiative cooling effect could even play a dominant role in the fields without airflows such as aerospace science and technology.

As-assembled thermal analyzer

The basic strategy for quantifying the heat dissipation properties is to record the surface temperatures of a sample and the heating power of a heater. Fig. 1a shows an as-assembled thermal analyzer consisting of a flexural electrical heater (5 cm × 3 cm × 1.5 mm), a heat insulator (5 cm × 3 cm × 3 cm), a pure copper plate (5 cm × 3 cm × 1.5 mm), and different layers of SACNT films. The electrical heater was made of nichrome resistance wires and silicon rubber layers with a maximum input power of 10 W and highest heating temperature of about 520 K. The input heating powers (P_in) and heating temperatures of the heater were controlled by a voltage source (KEITHLEY 2400), that is, P_in = voltage × current. Porous foam with very low thermal conductivity (k ∼ 0.02–0.05 W m⁻¹ K⁻¹) was used as the heat insulator and was pasted on one side of the electrical heater by heat-resistant adhesive. The heater and the copper plate were adhered by high-purity silver paste in order to reduce the heater–copper thermal boundary resistances and achieve good heat conduction conditions. In this way, almost all the heat was dissipated through the copper side. In spite of this, the heat dissipation power on the side wall and the insulator side was estimated and subtracted during our calculations, that is:

P_in − P_s = P = εσA(T_h⁴ − T_a⁴) + h_cA(T_h − T_a) (4)

where P_s is the heat power dissipating on the side wall and the insulator side.

Fig. 1 The as-assembled thermal analyzer and information of samples. (a) The schematic diagram of the as-assembled thermal analyzer. An electrical heater supplied the heating power. Four k-type thermocouples collected temperatures on different regions of the analyzer. The copper plate was adhered to the heater by some high-purity silver paste. An infrared (IR) thermometer was installed to measure the outer surface temperatures of the setup. (b–d) The scanning electron microscopy (SEM) images of different layers of SACNT films. (b) The SEM image of the single-layer SACNT film, which shows that the film is very thin and a lot of inter-spaces exist between individual CNTs. (c) The SEM image of the 3-layer SACNT films, which shows that some inter-spaces still exist in the films and a little part of the copper substrate could be observed. (d) The SEM image of the 25-layer SACNT films, which shows that the films are thick enough and the copper substrate couldn’t be observed anymore. (e–j) The photograph of a copper plate covered by different layers of SACNT films. These photographs (with the same scale bar) clearly show that the surface emissivity increases with the layer number of SACNT films. The copper plate is completely covered when the number of SACNT films reaches 25 layers.

Several k-type thermocouples with a diameter of 0.3 mm were used to measure the temperatures on different regions of the as-assembled thermal analyzer. The temperature measuring range of these thermocouples is 233–1173 K. The temperature data were collected by an eight-channel data collector (ART DAM 3039F) with a recording precision of 0.1 K, which was connected to a computer by an RS-485 bus. Three thermocouples were embedded between the heater and the copper plate to measure the inside temperatures of the copper plate. Another thermocouple was adhered on the outer surface of the copper plate or SACNT films by high-purity silver paste to measure the surface temperatures. The surface emissivity of a sample would be determined by the simultaneous temperature measurements by a k-type thermocouple and an infrared (IR) thermometer.²⁴ Specifically, the surface emissivity of a sample is necessary to measure its surface temperature by an IR thermometer. Reversibly, when the temperature of a tested sample reaches a steady state, a thermocouple would be used to measure the true surface temperature. Then the emissivity setting of the IR thermometer was adjusted to get the same temperature with the thermocouple. Thus the emissivity setting of the IR thermometer is the surface emissivity value of the sample.²⁴ The IR thermometer (OPTRIS LS) is a kind of hand-held thermometer, and the spectral range and temperature range are 8–14 μm and 240–1170 K, respectively. By the above method, the total emissivity of every sample was measured, which could be treated as independent of wavelength. In order to eliminate the influence of different airflows during the measurements, all measurements were performed in a big plastic box (1 × 1 × 1 m³), which is big enough and could avoid the boundary effect between our measuring setup and the box.

Preparation of SACNT films

The CNTs used in this work are all multi-walled CNTs (MWNTs), which were fabricated by chemical vapor deposition (CVD).^12,13 The diameters of these individual MWNTs are about 10–20 nm. Nowadays, the SACNT arrays can be massively produced in our laboratory by the CVD method (detailed information are described in the ESI†).¹² Based on the SACNT arrays, SACNT films²⁷ and other macroscopic CNT materials^4,28,29 have been fabricated. The single-layer SACNT films can be drawn from the SACNT arrays by an electrical motor with a constant speed because of very strong van der Waals forces between individual CNTs.^12,30 The thickness of single-layer SACNT films is about tens of nanometers and the optical transmission is about 80%.²⁷ The thermal and electrical conductivity of the SACNT films is anisotropic because of the super-aligned arrangements of individual CNTs.

The few-layer (e.g. single-layer, 3-layer, 8-layer and 15-layer) SACNT films are very soft and thin. It is difficult to directly measure the emissivity and heat dissipation properties of these soft and thin films. So supporting substrates are needed to support the few-layer SACNT films during the sample fabrication and measurements. On the other hand, copper has been widely used as a thermal management material. In view of this, the heat dissipation properties of pure copper plates were measured as a comparison experiment. In order to avoid the influence of different substrates on the heat dissipation properties, the same copper plate was chosen as the supporting substrates of the SACNT films during our measurements. In this work, different layers of SACNT films on a copper plate were prepared by the following steps. Firstly, the single-layer SACNT films were drawn from the SACNT arrays by an electrical motor and spread on the smooth copper plate. Then the SACNT films were cut by the cutting laser. The cutting laser is a kind of fiber laser with a central wavelength ∼1070 nm and its cutting power is about 100 W. Finally, some ethanol was dropped on the SACNT films and copper plate. After the evaporation of ethanol, the SACNT films could adhere on the surface of the copper plate because of tensions of ethanol molecules. By repeating the steps, different layers of SACNT films were formed on the same copper plate, and multilayer SACNT films also bonded together by van der Waals forces after ethanol dropping. Fig. 1b–d show scanning electron microscopy (SEM) images of the different layers of SACNT films on the copper plate. It can be clearly seen that inter-spaces between individual CNTs decrease with the layer number of SACNT films. Fig. 1b gives the SEM image of the single-layer SACNT films spreading on the copper plate, which clearly shows that numerous inter-spaces exist between individual CNTs. Some inter-spaces still exist in the 3-layer SACNT films, as is shown in Fig. 1c. However, when the number of SACNT films reaches 25 layers, as is shown in Fig. 1d, the multilayer SACNT films are thick enough and the copper plate is completely covered. Fig. 1e–j show photographs of the copper plate covered by different layers of SACNT films. These photographs clearly indicate that the surface emissivity values increase with the layer number of SACNT films.

Results and discussion

Total heat dissipation properties of different samples and cooling efficiency of multilayer SACNT films

The total heat dissipation properties of eight samples covered by different layers of SACNT films (i.e. pure copper, single-layer SACNT films, 3-layer SACNT films, 8-layer SACNT films, 15-layer SACNT films, 25-layer SACNT films, 100-layer SACNT films and 500-layer SACNT films) were measured in this work. Four different heating powers of the electrical heater were sequentially exerted, that is 1.1 W, 2.4 W, 3.6 W and 5.9 W. The electrical resistance of the heater was kept almost the same at the four heating powers. A computer software recorded the increasing processes and steady states of temperatures. As is shown in Fig. 2, in order to clearly show the results, four typical temperature curves of four corresponding samples are presented here (other curves are shown in the ESI†). These temperature curves in Fig. 2a–d demonstrate that those samples covered by SACNT films have lower equilibrium temperatures than the pure copper sample at the same heating power, which clearly demonstrates that multilayer SACNT films have much better heat dissipation properties than pure copper.

Fig. 2 The temperature curves of pure copper, 8-layer SACNT films, 15-layer SACNT films, and 100-layer SACNT films at a heating power of (a) 1.1 W, (b) 2.4 W, (c) 3.6 W, and (d) 5.9 W. All of these curves demonstrate that the SACNT films have lower equilibrium temperatures than the pure copper at the same heating power.

Moreover, the reduction of equilibrium temperatures increases with the heating powers. It can be seen that, at the heating power of 1.1 W, the equilibrium temperatures of pure copper and the 100-layer SACNT films are about 326 K and 322.5 K, respectively. So the reduction of equilibrium temperatures is 3.5 K. When the heating power reaches 5.9 W, the reduction of equilibrium temperatures increases to about 17 K. In order to quantitatively compare the total heat dissipation properties of the SACNT films with that of pure copper, the cooling efficiency (δ) of the multilayer SACNT films as compared with that of the pure copper is defined as δ = (ΔT/ΔT_c) × 100%,^23,31 where ΔT is the reduction of equilibrium temperature, and ΔT_c is the temperature rise from room temperature (295 K during our measurement) of the pure copper plate. Under the four heating powers, the equilibrium temperatures are reduced about 3.5 K, 8 K, 10 K and 17 K by the 100-layer SACNT films, and the temperature rise of pure copper is 31 K, 64 K, 88 K and 128 K, respectively. So the cooling efficiency of the 100-layer SACNT films is 11.3%, 12.5%, 11.4% and 13.3%, respectively. These efficiencies are obviously higher than that of disordered MWNT composites, which is only about 5.4%.³¹ This result proves that the SACNT materials have better heat dissipation properties than disordered CNT composites. In addition, some other previous results^16,23 show that the disordered CNTs may have better heat dissipation properties than the SACNT films. So the previous results demonstrate that the heat dissipation properties of disordered CNTs may be unreliable.

Furthermore, the SACNT films have many advantages in the application of thermal management materials compared with the disordered CNT film. Firstly, the compacted multilayer SACNT films have a much higher axial thermal conductivity than the disordered CNT film. This feature could ensure good thermal transport properties, which are also very important when they are used as thermal management materials. Secondly, the compacted SACNT films could have superior mechanical properties and a stable surface. Conversely, the structure of disordered CNT films may be fragile and unstable, which would increase the difficulty and cost of production. Finally, the random oriented CNTs must be coated on other materials when they are used as thermal management materials. In contrast, SACNT films could be directly used as thermal management materials or coated on other materials according to different situations. This feature could ensure the good thermal properties as well as reduce the weight of thermal management devices. The weight also plays an important role in portable electronic devices and aerospace devices. During our measurements, the ambient temperature kept relatively stable (±0.7 K).

Measurement of the surface emissivity

As mentioned above, the surface emissivity of every sample could be determined by temperature measurements using a thermocouple and an IR thermometer simultaneously. The surface emissivity of every sample was repeatedly measured at four heating powers. The average values of the surface emissivity for the samples are shown in Fig. 3. The error bars capture the uncertainty due to variations of different measurements at different heating powers. There are decreasing errors of the surface emissivity values for the pure copper and other samples, as the layer number of the SACNT films increases. This could be explained by the copper oxidation at different operating powers and low coverage rate of few-layer SACNT films. When the SACNT films are thick enough (e.g. 100-layer and 500-layer SACNT films), the variations of the surface emissivity values become very small. In this work, the experimental surface emissivity of the pure copper is 0.14 ± 0.07. This value is in the range of other reported results.^32,33 The specific value of the surface emissivity of a typical metal is significantly influenced by the surface roughness, oxide layer and other factors. It should be noted that previous works have shown that the spectral emissivity of metals usually decreases as the wavelength increases.³² In the present work, the measured emissivity is the total emissivity at the wavelength range 8–14 μm, which could be treated as independent of wavelength.

Fig. 3 Surface emissivity values (ε) of pure copper and samples covered by different layers of SACNT films. These error bars capture the uncertainties due to variations of different measurements at different heating powers.

The single-layer SACNT film coating sample has a surface emissivity value of 0.15 ± 0.07, which is essentially the same as that of the pure copper. It is reasonable that the thin single-layer SACNT films have a light transmissivity of about 80%. The surface emissivity values for different layers of SACNT films rapidly increase until the layer number of the SACNT films reaches 100 layers. It can be illustrated that the surface emissivity of 100-layer and 500-layer SACNT films is 0.91 ± 0.02 and 0.95 ± 0.02, respectively. These results indicate that the 100-layer and 500-layer SACNT films could completely cover the copper plate. In addition, the thickness of the 500-layer SACNT films is about 100 μm, which was measured by an optical microscope. The above results further demonstrate that SACNT films with hundreds of layers have an excellent coverage level when they are used as functional coatings.

Calculation of the radiative heat dissipation and the natural convective heat dissipation

The radiative heat dissipation power (P_r) and its percentage ((P_r/P)%) of the total heat dissipation power (P) were calculated using the ε values and eqn (2). All the values of (P_r/P)% of the eight samples at four heating powers are shown in Fig. 4a. The value of P_r increased with the heating powers, but there was no obvious difference in (P_r/P)% of every sample at four heating powers. For the pure copper, the (P_r/P)% value is only about 5.4% because of its very low surface emissivity. Thus, the radiative heat dissipation power of the pure copper could be ignored in comparison to the total heat dissipation power. When the copper was covered by 500-layer SACNT films, the (P_r/P)% value increases to about 32%. In this case, the radiative heat dissipation plays a significant role in the total heat dissipation.^16,24 This suggests that the radiative cooling could be an effective way to enhance the total heat dissipation properties. This aspect would be more promising in an environment without airflow, such as in aerospace science and technology. Furthermore, because of the chemical durability of CNTs, the SACNT films have advantages over other organic coatings when they are used as radiative heat dissipation coatings.

Fig. 4 The percentage of the radiative heat dissipation power and the natural convective heat transfer coefficient of every sample at four heating powers. (a) The percentage of the radiative heat dissipation power ((P_r/P)%) at four heating powers. For multilayer SACNT films, these curves show that the (P_r/P)% value could exceed 32%, which indicates that the radiative cooling could be an effective way to enhance the heat dissipation properties. (b) The calculated natural convective heat transfer coefficient (h_c) by subtracting the radiative heat dissipation power.

The error bars in Fig. 4a capture the propagating errors of temperature measurements and surface emissivity measurements and the calculation errors of the heat dissipation power (P_s) on the side wall and insulator wall. The measured values of the surface emissivity were combined results of the SACNT films and the copper surface. The apparent surface area of every sample was used to calculate the radiative heat dissipation power. It is worth noting that the actual radiative area might be different from the apparent surface area. In particular, for the few-layer SACNT film (e.g. single-layer and 3-layer SACNT films) coating samples, the actual radiative area is smaller than the apparent surface area because of the numerous inter-spaces between individual CNTs. On the other hand, the actual radiative area of the multilayer SACNT film coating samples could be a little larger than their apparent surface area due to the tight arrangements and the cylindroid surface of individual CNTs. It is extremely difficult to directly measure or calculate the actual radiative area of every sample, and the apparent surface area is thereby used to calculate the radiative heat power. In spite of this, the calculated radiative heat dissipation powers are also apparent and combined values. This simplified model could only result in some calculation errors without changing the fundamental characteristics of our results in Fig. 4a. Additionally, Hsiao et al. discussed the radiative cooling mechanisms of nano-coatings consisting of CNTs and graphene.²¹ They proposed that the radiative cooling of nano-coatings could have two different paths, which are named radiative cooling and non-radiative cooling, respectively. The dispersion level of nanoparticles in nano-coatings could have an obvious influence on the radiative cooling mechanisms. Specifically, well-quantized lattice motions (such as individual CNTs and monolayer graphene) are excited to the higher vibrational states by absorbing heat from the heating source, and then the excess energy is released by emitting IR radiation. Thus the heat source is effectively cooled down.²¹ This situation is named as the radiative cooling path. On the other hand, when continuum or quasi-continuum materials absorb heat from a heating source, excited lattice motions will relax non-radiatively to return the excess energy back to the heating source, which is named as the non-radiative cooling path.²¹ The SACNT films studied here are a kind of continuous sheet in the aligned direction of individual CNTs, and the individual CNTs are separated in the perpendicular direction. So we believe that two paths of the radiative cooling would exist simultaneously. That is, the radiative cooling in the aligned direction of individual CNTs occurs mainly by means of self-heating accumulation, whereas the radiative cooling in the perpendicular direction is mainly by means of radiation emission.

Using eqn (1)–(4), we also tried to calculated the natural convective heat transfer coefficient (h_c) of every sample by subtracting the radiative heat dissipation power from the total heat dissipation power (Fig. 4b). The error bars capture the calculation errors of P_s and the propagating errors of temperature measurements. A small amount of input power was dissipated on the side wall and the insulator side, which was estimated and removed. It has to be emphasized that the calculations of h_c values contain uncertainties because of intangible heat transfer coefficients of the side wall and the insulator side. Systemic and random errors also exist at different measuring points of the four thermocouples, which have been integrated in the error bars in Fig. 4b. It can be summarized that h_c increases with the heating power. This can be ascribed to the fact that the higher heating power leads to higher surface temperatures. Thus, the higher surface temperatures result in bigger natural convective forces. Fig. 4b shows that h_c of the few-layer SACNTs is in the range of 17–26 W m⁻² K⁻¹, which is obviously higher than the h_c value of of several-layer graphene (8–12 W m⁻² K⁻¹ when the heating temperature is about 393 K).^20,34 This superiority indicates that the CNTs would have more advantages than graphene when they are used as heat dissipation materials. In addition, h_c decreases with the layer number of SACNT films. Combined with the results in Fig. 4a, the multilayer SACNT films could enhance the radiative heat dissipation, and weaken the natural convective heat dissipation. These features indicate that there could be a competitive relationship between the radiative heat dissipation and the natural convective heat dissipation. Here, we suggest that these characteristics are mainly due to the following aspects.

Firstly, the temperature curves in Fig. 2 show that the equilibrium temperatures decrease with the layer number of SACNT films at the same heating power, which declares that the total heat dissipation properties increase with the layer number of SACNT films. Meanwhile, the natural convective driving force is decreasing with the layer number of SACNT films, which could cause a decreasing natural convective heat transfer coefficient. Secondly, several reported works pointed out that the natural convective heat dissipation between a solid and gas could be influenced by the thermal accommodation coefficient.^20,35,36 Fan et al.³⁶ compared the different convective heat dissipation mechanisms, which are known as the directly scattering and trapping-desorption of ambient gas respectively.^20,36 Similarly, the adsorption energies of ambient gas molecules (such as O₂, N₂, and H₂O) on SACNT films are relatively small (about 20 meV) and therefore the ambient gas molecules could be adsorbed but instantaneously released due to the comparable adsorption and thermal energies. So the thermal accommodation coefficient is low and the thermal interfacial conductance is limited.²⁰ Conversely, the adsorption energies of ambient gases on the oxidized copper surface are very high (several hundreds of meV), which are much larger than the adsorption energies of ambient gas molecules on SACNT films.²⁰ The ambient gas molecules would be trapped for a longer time to absorb heat from the copper surface before scattering. So the thermal interfacial conductance is large in this case.²⁰ Thus, the natural convective heat dissipation is much stronger.

Additionally, we suggest that the thermal boundary resistance (TBR) between individual CNTs^37–40 also has a slight influence on the convective heat dissipation. The reported order of magnitude of the TBR between CNTs and air is about 10⁻⁶ to 10⁻⁵ m² K W⁻¹.^17,35 On the other hand, people also have studied the TBR between individual CNTs by theoretical and experimental methods, which is in the range of 10⁻⁸ to 10⁻⁵ m² K W⁻¹.^37,39,40 For the few-layer (e.g. single-layer and three-layer) SACNT film coating samples, there is almost no overlap of individual CNTs because of numerous inter-spaces between individual CNTs. However, the total CNT–CNT TBR increases with the number of SACNT films. The increasing CNT–CNT TBR could have a slight influence on the convective heat dissipation properties. Finally, the non-radiative heat accumulation²¹ proposed by Hsiao et al. would also influence the convective heat dissipation. Briefly, the portion of non-radiative heat accumulation would increase with the layer number of SACNT films. Hence, the convective heat dissipation is weakened by the non-radiative heat accumulation (more details are discussed in ref. 21). We have to admit that we just propose several possible reasons for the results reported here. Detailed work needs to be carried out to verify the above reasons.

The design of a CNT CPU-radiator

As mentioned above, the multilayer SACNT films have better total heat dissipation properties than common metals. On the other hand, the compacted multilayer SACNT films (buckypapers) reported in our previous works have very high axial thermal conductivity and low density.^4,14 Therefore, we believe that SACNT buckypapers could be directly used as raw materials to make a CPU-radiator for laptops that is conventionally made of common metals. However, as mentioned above, the SACNT films would have many other applications and would not be limited to this application. According to the structure of a traditional CPU-radiator of a laptop, a novel CNT CPU-radiator, as shown in Fig. 5, has been designed and tested. The CNT buckypapers are too flexible, so we set a metal framework to support the CNT CPU-radiator. The metal framework was made of a piece of titanium plates, which has a relatively low density and the highest specific strength among common metals. In this way, the CNT CPU-radiator could be self-standing and keep a stable shape. The CNT CPU-radiator could be directly pasted on the CPU because it is all solid state (more detailed preparation processes of the CNT CPU-radiator are described in the ESI†).

Fig. 5 The schematic diagram of a CNT CPU-radiator. This novel CNT CPU-radiator has better heat dissipation properties and a much lighter weight than the traditional metallic CPU-radiator.

The mass of the CNT CPU-radiator is only about one fifth of the traditionally metallic CPU-radiator. Our results in Fig. 2 demonstrate that the CNT CPU-radiator could have better heat dissipation properties than the metallic CPU-radiator. Nevertheless, the superior heat dissipation properties of the novel CNT CPU-radiator have been confirmed by monitoring the working temperatures of a CPU when it is respectively equipped with the two radiators. In most cases, the CPU equipped with the CNT CPU-radiator has lower and more stable working temperatures (as shown in Fig. S5 of the ESI†), which demonstrates that the CNT CPU-radiator has much better thermal properties than the metallic CPU-radiator. Additionally, the CNT CPU-radiator also has a lighter weight, indicating that the compacted SACNT films could also be widely used as light-weight and high-performance thermal management materials in other cases such as shells of portable electronic devices and thermal management coatings of aerospace devices.

Conclusions

In summary, we measured the heat dissipation properties of different layers of SACNT films. The results demonstrate that the multilayer SACNT films have much better heat dissipation properties than pure copper. The cooling efficiency of the multilayer SACNT films could be as high as about 13.3%. On the other hand, the natural convective and radiative heat dissipation properties of every sample were respectively determined. Our results indicate that the radiative heat dissipation properties of multilayer SACNT films play an important role in the total heat dissipation properties. This advantage would be more promising in the environment without airflow, such as aerospace science and technology. It is further found that the SACNT films have much better natural convective heat dissipation properties than the several-layer graphene. Moreover, a novel CPU-radiator consisting of the compacted multilayer SACNT films has been proposed and designed to have advantages of a lighter weight and better heat dissipation properties over a traditional metallic radiator. In view of these results reported here, we believe that the compacted SACNT films would be widely used as light-weight and high-performance thermal management materials.

Acknowledgements

This work was supported by the National Basic Research Program of China (2012CB932301) and the Natural Science Foundation of China (51276191, 51572146).

Notes and references

1. E. Pop, Nano Res., 2010, 3, 147–169.
2. A. A. Balandin, Nat. Mater., 2011, 10, 569–581.
3. D. G. Cahill, P. V. Braun, G. Chen, D. R. Clarke, S. Fan, K. E. Goodson, P. Keblinski, W. P. King, G. D. Mahan, A. Majumdar, H. J. Maris, S. R. Phillpot, E. Pop and L. Shi, Appl. Phys. Rev., 2014, 1, 11305.
4. L. Zhang, G. Zhang, C. Liu and S. Fan, Nano Lett., 2012, 12, 4848–4852.
5. A. L. Moore and L. Shi, Mater. Today, 2014, 17, 163–174.
6. P. Kim, L. Shi, A. Majumdar and P. L. McEuen, Phys. Rev. Lett., 2001, 87, 215502.
7. L. Dai, D. W. Chang, J. Baek and W. Lu, Small, 2012, 8, 1130–1166.
8. E. Pop, D. Mann, Q. Wang, K. Goodson and H. Dai, Nano Lett., 2006, 6, 96–100.
9. A. A. Balandin, S. Ghosh, W. Bao, I. Calizo, D. Teweldebrhan, F. Miao and C. N. Lau, Nano Lett., 2008, 8, 902–907.
10. X. Zhang, L. Song, L. Cai, X. Tian, Q. Zhang, X. Qi, W. Zhou, N. Zhang, F. Yang, Q. Fan, Y. Wang, H. Liu, X. Bai, W. Zhou and S. Xie, Light: Sci. Appl., 2015, 4, e318.
11. R. H. Baughman, A. A. Zakhidov and W. A. de Heer, Science, 2002, 297, 787–792.
12. K. Jiang, J. Wang, Q. Li, L. Liu, C. Liu and S. Fan, Adv. Mater., 2011, 23, 1154–1161.
13. K. Jiang, Q. Q. Li and S. S. Fan, Nature, 2002, 419, 801.
14. G. Zhang, C. Liu and S. Fan, Sci. Rep., 2013, 3, 2549.
15. G. Xin, T. Yao, H. Sun, S. M. Scott, D. Shao, G. Wang and J. Lian, Science, 2015, 349, 1083.
16. C. N. Suryawanshi and C. Lin, ACS Appl. Mater. Interfaces, 2009, 1, 1334–1338.
17. I. Hsu, M. T. Pettes, M. Aykol, C. Chang, W. Hung, J. Theiss, L. Shi and S. B. Cronin, J. Appl. Phys., 2011, 110, 44328.
18. S. Jiang, C. Liu and S. Fan, ACS Appl. Mater. Interfaces, 2014, 6, 3075–3080.
19. Z. Huang, M. Gao, T. Pan, Y. Zhang, B. Zeng, W. Liang, F. Liao and Y. Lin, J. Appl. Phys., 2015, 117, 24901.
20. J. J. Bae, S. C. Lim, G. H. Han, Y. W. Jo, D. L. Doung, E. S. Kim, S. J. Chae, T. Q. Huy, N. Van Luan and Y. H. Lee, Adv. Funct. Mater., 2012, 22, 4819–4826.
21. T. Hsiao, T. Eyassu, K. Henderson, T. Kim and C. Lin, Nanotechnology, 2013, 24, 395401.
22. N. Han, T. Viet Cuong, M. Han, B. Deul Ryu, S. Chandramohan, J. Bae Park, J. Hye Kang, Y. Park, K. Bok Ko, H. Yun Kim, H. Kyu Kim, J. Hyoung Ryu, Y. S. Katharria, C. Choi and C. Hong, Nat. Commun., 2013, 4, 1452.
23. T. Eyassu, T. Hsiao, K. Henderson, T. Kim and C. Lin, Ind. Eng. Chem. Res., 2014, 53, 19550–19558.
24. C. N. Suryawanshi, T. Kim and C. Lin, Rev. Sci. Instrum., 2010, 81, 35105.
25. J. Strand and A. Vengar, Eur. J. Phys., 1984, 5, 9–12.
26. T. J. Kang, T. Kim, S. M. Seo, Y. J. Park and Y. H. Kim, Carbon, 2011, 49, 1087–1093.
27. L. Xiao, Z. Chen, C. Feng, L. Liu, Z. Bai, Y. Wang, L. Qian, Y. Zhang, Q. Li, K. Jiang and S. Fan, Nano Lett., 2008, 8, 4539–4545.
28. D. Wang, P. C. Song, C. H. Liu, W. Wu and S. S. Fan, Nanotechnology, 2008, 19, 75609.
29. K. Liu, Y. Sun, R. Zhou, H. Zhu, J. Wang, L. Liu, S. Fan and K. Jiang, Nanotechnology, 2010, 21, 45708.
30. X. B. Zhang, K. L. Jiang, C. Teng, P. Liu, L. Zhang, J. Kong, T. H. Zhang, Q. Q. Li and S. S. Fan, Adv. Mater., 2006, 18, 1505.
31. T. Teng and T. Teng, Appl. Therm. Eng., 2013, 51, 1098–1106.
32. D. P. Dewitt and G. D. Nutter, Theory and Practice of Radiation Thermometry, Wile, New York, 1988, pp. 915–931.
33. A. Sala, Radiant Properties of Materials, Elsevier, New York, 1986, pp. 175–257.
34. J. Kang, H. Kim, K. S. Kim, S. Lee, S. Bae, J. Ahn, Y. Kim, J. Choi and B. H. Hong, Nano Lett., 2011, 11, 5154–5158.
35. M. Hu, S. Shenogin, P. Keblinski and N. Raravikar, Appl. Phys. Lett., 2007, 90, 231905.
36. J. R. Manson and G. Fan, Phys. Rev. B: Condens. Matter Mater. Phys., 2009, 79, 45424.
37. G. Zhang, C. Liu and S. Fan, ACS Nano, 2012, 6, 3057–3062.
38. L. Hu and A. J. H. McGaughey, Appl. Phys. Lett., 2014, 105, 193104.
39. J. K. Yang, S. Waltermire, Y. F. Chen, A. A. Zinn, T. T. Xu and D. Y. Li, Appl. Phys. Lett., 2010, 96, 231092.
40. H. Zhong and J. R. Lukes, Phys. Rev. B: Condens. Matter Mater. Phys., 2006, 74, 125403.

---

Footnote

† Electronic supplementary information (ESI) available: Detailed fabrication process of the SACNT buckypapers, all the temperature curves of the eight measured samples, detailed preparation process of the CNT CPU-radiator, and comparison of the heat dissipation properties of the two CPU-radiators. See DOI: 10.1039/c6ra07143k

---