Metal functionalization of carbon nanotubes for enhanced sintered powder wicks

Abstract

Phase change cooling schemes involving passive heat spreading devices, such as heat pipes and vapor chambers, are widely adopted for thermal management of high heat-flux technologies. In this study, carbon nanotubes (CNTs) are fabricated on a 200 μm thick sintered copper powder wick layer using microwave plasma enhanced chemical vapor deposition technique. A physical vapor deposition process is used to coat the CNTs with a varying thickness of copper to promote surface wetting with the working fluid, water. Thermal performance of the bare sintered copper powder sample (without CNTs) and the copper-functionalized CNT-coated sintered copper powder wick samples is compared using an experimental facility that simulates the capillary fluid feeding conditions of a vapor chamber. A notable reduction in the boiling incipience superheat is observed for the nanostructured samples. Additionally, nanostructured samples having a thicker copper coating provided a considerable increase in dryout heat flux, supporting heat fluxes up to 457 W/cm² from a 5 mm × 5 mm heat input area, while maintaining lower surface superheat temperatures compared to a bare sintered powder sample; this enhancement is attributed primarily to the improved surface wettability. Dynamic contact angle measurements are conducted to quantitatively compare the surface wetting trends for varying copper coating thicknesses and confirm the increase in hydrophilicity with increasing coating thickness.

Introduction

Thermal analysis of three-dimensional circuit architectures, currently being developed for logic, memory, and opto-electronic device integration, points towards significantly higher waste heat energy densities and increased temperature non-uniformities in 3-D devices [1]. Adequate thermal management of these evolving chip architectures is therefore critical to maintaining acceptable device temperatures and improving the reliability and longevity of future electronic systems. Phase-change cooling schemes have emerged as a prominent thermal management solution because of their potential to achieve high heat dissipation rates while maintaining uniform device temperatures. Various cooling schemes with active liquid pumping, such as microchannel cooling [2], [3], jet impingement cooling [4], [5] and spray cooling [6], continue to be topics of intense research, and have considerably reduced the thermal resistance between the chip and the ambient when employing phase change [7]. However, implementation of pumped liquid cooling schemes is becoming increasingly difficult in real systems due to demanding space and weight constraints. In contrast, devices that employ passive phase-change cooling, such as heat pipes and vapor chambers, can be more easily incorporated into realistic systems for efficient heat removal.

The operating principle of a vapor chamber is analogous to that of a heat pipe; however, the term vapor chamber typically refers to spreading devices that spread heat from a small evaporator area to a large condenser surface over a short fluid working distance. In a typical vapor chamber, an optimized charge of working fluid is housed in a thin-walled copper shell at a reduced pressure corresponding to the desired operating saturation temperature. Evaporation/boiling occurs in a liquid-saturated porous wick structure lining the heated evaporator side. The resulting vapor pressure differential drives flow towards the condenser side. Capillary pressure passively transports the condensed working fluid back to the heat input area of the wick structure, closing the fluid-thermal cycle and allowing continuous removal of heat. Fig. 1 depicts a schematic diagram of the internal transport processes in a vapor chamber with a sintered copper powder layer as the wick structure on the evaporator side; the sintered wick is coated with carbon nanotubes (CNTs) in the figure. Vapor chambers are relatively inexpensive, offer high reliability, and allow for packaging adaptability, making them an attractive and reliable solution for next-generation electronic systems. However, current-generation vapor chambers are limited by the maximum supported thermal loads and heat flux.

The maximum heat flux supported by a vapor chamber is dictated primarily by the liquid replenishment capability of the evaporator wick structure. The thermal resistance posed by the conduction path through the wick material and by phase change at the liquid–vapor interface is critical in determining the overall device resistance and the maximum amount of heat that can be transported by the vapor chamber. Hence, it is critical that a thorough understanding of the evaporative/boiling heat transfer mechanisms in the evaporator wick structure be obtained.

The boiling incipience superheat and the dryout heat flux constitute two important design criteria for a wick structure. With increasing temperature non-uniformities and the presence of local hot spots in dense circuit-packaging designs, operation of a vapor chamber in the boiling regime proves advantageous because of the enhanced heat transfer coefficients and lower overall thermal resistance associated with the wick structure [8]. Reduction of the surface superheat temperature at which the transition from evaporation to boiling occurs is desirable because it would provide operation in the boiling regime over a wider range of operational heat fluxes and avoid the temperature excursion associated with this transition. Extending the dryout heat flux, which is the maximum supported heat flux, to a level of 1 kW/cm² or greater [9] is decisive for thermal management of future electronic systems. A high-heat-flux vapor chamber should maximize the permeability of the wick to liquid flow to minimize the pressure drop, while simultaneously providing sufficient capillary pressure to draw enough liquid to the evaporator to support the maximum heat flux desired.

A number of studies dedicated to understanding the influence of wick parameters (such as wick thickness, sintered particle diameter and porosity) on thermal performance have imparted a good understanding of the respective performance trends for monoporous wicks that possess a single characteristic pore size [8], [10], [11]. Several modifications to traditional monoporous wicks have been proposed in prior work [12], [13], [14], [15], [16] to improve the dryout heat flux and to decrease the thermal resistance of the wick structure to meet the cooling requirements for ultra-high heat flux applications. Semenic and Catton [12] compared the critical heat flux (CHF) of monoporous and biporous wicks (wicks with two different pore sizes) using water as the working fluid, under sub-atmospheric saturation pressure (<13 kPa) conditions and reported high CHF values of 520 and 990 W/cm² at superheats of 50 and 147 °C for 800 and 3000 μm thick biporous wicks, respectively. They argued that monoporous wicks are not suitable for supporting high heat dissipation rates because as the pore size is decreased to yield improved capillary forces required at high heat fluxes, this also leads to more efficient vapor trapping and the prevent of liquid feeding; this supports local dryout. In contrast, the two characteristic pore sizes of biporous wicks may be optimized to provide adequate vapor release sites via the larger pores while maintaining sufficiently high capillarity to prevent dryout via the smaller pores. Nevertheless, thick biporous wicks may not find practical use due to the high wick resistance and the high resulting wall superheat. Zhao and Chen [13] investigated the thermal performance of sintered copper wicks with microgrooves in the direction of liquid flow of widths ranging from 150–500 μm. Compared to a randomly distributed pattern of vapor exit pathways in biporous wicks, microgrooves provided aligned passages for vapor to exit the wick, and extended the CHF by as much as 350% compared to monoporous wicks.

Nanostructuring of the heat transfer surface has been pursued as an alternate approach for achieving superior thermal performance in pool boiling and vapor chambers. Nanostructuring could provide an additional capillary length scale for improved fluid wicking while preserving the micron-scale vapor escape locations available in traditional microstructured wicks. Nanostructured coatings provide pore radii on the order of a few tens of nanometer, at least two orders of magnitude less than the few tens of micron pore sizes prevalent in traditional porous surfaces, and hence provide high capillary pressures to overcome the viscous drag on fluid flow at the microscale and delay capillary dryout. Chen et al. [17] reported high CHF values (∼200 W/cm²) and heat transfer coefficients (∼6 W/cm²K) in pool boiling studies on flat silicon surfaces individually coated with dense arrays of Si nanowires fabricated by electro-less etching and Cu nanowires synthesised by electroplating of Cu into nanoporous alumina templates. The observed enhancement was attributed to superhydrophilicity and enhanced capillary pumping of the nanowire arrays. Yao et al. [18] investigated the effect of the height of Cu and Si nanowires fabricated by inexpensive electrochemical methods and observed that taller nanowire arrays enhanced pool boiling performance by providing stable active nucleation sites. Nam et al. [19] nanostructured copper micropost wicks by chemical oxidation to form CuO nanostructures. CHF values greater than 500 W/cm² were reported for a 2 mm × 2 mm heat input area but decreased to 200 W/cm² in the case of a 5 mm × 5 mm heat input area. Ding et al. [20] fabricated 3 cm× 3cm × 600 μm thick vapor chamber devices with hair-like nanostructured titania grown on Ti pillars, though only very low heat fluxes (<10 W/cm²) were examined during thermal testing.

Complete coverage and patterning of wick surfaces with carbon nanotubes (CNTs) has found considerable interest in the research community [14], [15], [21]. CNTs are cylindrical nanostructures of graphitic carbon with outer diameters ranging from 1 to 100 nm and typical lengths from 1 to 50 μm. Apart from the benefits of nanostructuring discussed above, CNTs also possess high intrinsic thermal conductivity and a tunable porosity. Ujereh et al. [22] performed pool boiling experiments on CNT-coated silicon and copper substrates with distinct patterned areas of coverage and differing CNT array densities. Surfaces coated entirely with CNTs eliminated the large incipience superheat overshoot and the resulting substantial temperature drop upon incipience observed in the case of bare silicon and copper surfaces. CNT forests grown on silicon substrates [23] were reported to augment CHF values by 25% compared to bare silicon surfaces in addition to enhancing heat flux dissipated in the film boiling regime. Mechanisms proposed for this augmentation included increased surface areas for heat transfer provided by CNTs in the form of ‘nanofins’ and disruption and possible collapse of the vapor film in the film boiling regime resulting in improved transient quenching of the CNT surface compared to a bare silicon substrate. While these two studies utilized highly wetting dielectric fluids, water is the preferred working fluid for use in vapor chambers for electronics cooling applications. Pool boiling experiments on hybrid sintered/CNT surface [24] employing deionized water as the working fluid revealed a significant reduction in boiling incipience superheat compared to a plain flat copper surface.

It is essential to distinguish the heat transfer mechanisms encountered in pool boiling experiments from those seen in vapor chamber operation. In a vapor chamber, liquid migration to the heat input area occurs through capillary action in the wick structure and heat transfer occurs by evaporation at the liquid–vapor free interface at lower surface superheats and by boiling at the solid–liquid interface at higher surface superheats. In the case of pool boiling, the surface is completely submerged in liquid and heat is dissipated by natural convection until the surface temperature is high enough to initiate nucleate boiling. Therefore, any performance augmentation due to CNTs that were previously shown for pool boiling must be evaluated under capillary feeding conditions that exist in a vapor chamber.

The inherent hydrophobic nature of CNTs, however, limits their usage with water in vapor chambers. Several surface functionalization techniques for promoting CNT surface wetting behavior have been explored. Cai and Chen [14] used a 2% hydrochloric acid-treatment process to improve the surface wettability of CNTs. Using this technique, a CNT biwick structure consisting of parallel CNT stripes 100 μm wide and 50 μm apart was fabricated and heat dissipation capacities of up to 600 W/cm² were demonstrated. Hashimoto et al. [15] deposited CNTs on a 1.5 mm thick layer of sintered copper powder by thermal chemical vapor deposition (CVD) and rendered them hydrophilic via exposure to ultraviolet (UV) radiation. Experiments were conducted to demonstrate that CNT-coated wicks removed heat fluxes up to 500 W/cm² at a superheat of ∼33 °C, compared to a baseline sample without CNT-coating at a superheat of ∼45 °C at the same heat flux. The prevailing hypothesis to explain the surface wettability switch of CNTs in response to UV light exposure is based on the assisted adsorption of hydrophilic hydroxyl groups [25]. This method could conceivably minimize the surface degradation of CNTs compared to acid treatment processes, and potentially facilitate scalability. Nevertheless, this methodology is not expected to be feasible for continuous operation in thin vapor chambers due to likely degradation of surface wettability of UV-exposed CNT-coated wicks over time. The hydrophilic behavior of CNTs conformally coated with copper was previously demonstrated by our group [16], [21]. This method is advantageous for use in vapor chambers due to long-term compatibility of the copper coating with water.

In recent work [16], the thermal characteristics of 1 mm thick monolithic and patterned sintered copper powder wicks were evaluated with and without a functionalized CNT coating. Patterning of the wick structure with a grid of square recesses reaching down to the substrate provided a low resistance vapor exit path to improve performance during vigorous boiling. Growth of a CNT array directly on the substrate in the patterned recesses in the wick structure did not provide any enhancement as they remained flooded during testing. A similar study was then conducted with 200 μm thick sintered powder layers only a few particle layers deep and compatible for use in ultra-thin vapor chambers [26]. The intention was to compare the heat transfer enhancement mechanisms for the different wick layer thicknesses. While patterned samples without CNT coating reduced the dryout heat flux due to capillary fluid starvation over the surface, a CNT array grown in the patterned recesses resulted in active wicking of the fluid and increased the dryout heat flux. However, this study revealed that for 200 μm thick wicks, monolithic samples generally outperformed grid-patterned samples due to the increase in the available liquid feeding area. In all test cases in the above studies [16], [26], CNT-coated wick structures were shown to reduce the superheat at boiling incipience, motivating a detailed study of boiling incipience for CNT-coated monolithic wicks [26]. In this study, a functionalized CNT coating was shown to reduce the mean surface superheat at the boiling incipience point by 5.6 °C compared to monolithic samples.

In the present work, we explore means of increasing the dryout heat flux of nanostructured monolithic wicks to accommodate the ever-increasing thermal loads of high-performance power electronics, while preserving the reduction in the boiling incipience superheat achieved in the previous work. We evaluate 200 μm thick nanostructured wicks with different nominal thicknesses of the copper coating deposited on the CNTs, and compare their thermal and surface wetting characteristics with a bare sintered powder sample (without CNTs).