Impact of Immersion Cooling on Thermomechanical Properties of Halogen-Free Substrate Core

Massive amounts of data are gathered, stored, processed, and distributed using data centers, which may be used for a variety of purposes, including business, amusement, and other requirements [1,2]. A single CPU's server power will normally be 140–190 W by 2020, whereas a single processor for high-performance computing will exceed 300 W. The heat dissipation of information technology equipment (ITE) accounts for 50–60% of the data center's heat source [3]. In 2025, data centers will consume 4.5% of the world's electricity, of which 40% will be used for cooling systems [4]. Globally, it is predicted that data centers used 91 billion kWh of energy annually in 2013, 140 billion kWh in 2020, and more in 2025. This would result in annual carbon emissions and operating electricity costing more than 150 million metric tons and $13 billion, respectively [5]. Because data centers consume a lot of electricity, particularly for cooling, decreasing energy use is a primary concern for IT companies and policy officials [6].

Despite significant technological advancement over the past few decades, managing the temperature of electronics or microprocessors still presents significant technical difficulties. The effective removal of increasing heat flux and extremely nonuniform power dissipation are the two key cooling problems [6]. Due to inadequate cooling capacity, high energy consumption, and high operational expenses, traditional air cooling has reached its limits. Immersion cooling offers several benefits such as higher thermal mass and high heat dissipation due to the dielectric fluids being in direct contact with all the components [7]. Single-phase immersion has several advantages such as it shields the ITE from the impact of contaminants, reduces failures due to vibration, and lowers the CapEx as it does not require sophisticated liquid distribution manifold construction, thus making data center architecture easier [7].

Reliability issues may arise when the heat-generating components are immersed in the dielectric fluids. Deformation, warpage, or delamination modes of failure that result in the failure of the complete package may be brought on by a slight variation in the material properties of the components [8]. While providing the material properties, most of the fluid suppliers and vendors send the soak test results. However, they do not consider operational reliability like the mechanical and electrical characteristics of the components unique to the application work environment [8]. The material compatibility may be divided into direct and indirect interactions. Direct interactions are processes that are generated by the fluid itself. They can occur in the fluid or when the fluid interacts with a component material. Dissolution, absorption/swelling, chemical contact of material, environmental stress cracking, fluid aging, and oxidation are the six basic ways for direct interactions to occur in single-phase fluids. Chemicals released or generated by direct contact with IT components induce indirect reactions rather than the fluid itself [9].

The thermomechanical characteristics of low loss printed circuit boards immersed in mineral oil for 720 h at 25, 50, 75, and 105 °C were studied by Chauhan et al. [10]. They found that at that the in-plane coefficient of thermal expansion values of the post-aged samples did not change much at these temperatures. Ramdas et al. [11] performed a similar study with FR-4 printed circuit board immersed in EC100 dielectric fluid. The results show that the modulus has decreased for post-aged samples which means that PCBs are less stiff and in turn decrease the warpage. Bhandari et al. [12] investigated the effect of single-phase immersion fluid EC100 on the mechanical properties of nonhalogenated substrates after immersion for 720 h at ambient temperature, 50 °C, and 75 °C. In this study, modulus of the immersed substrate is compared with the modulus of substrates exposed to air at the respective temperatures. It was observed that rising temperatures caused the modulus to decrease but at 75 °C in both air and EC100 fluid, a varying trend was observed.

The fundamental goal of this research is to quantify and characterize the material compatibility of immersed components. As per Open Compute Project (OCP) design guidelines for immersion-cooled IT equipment, the traditional material compatibility tests from standards like ASTM D471 and D2240 can be used with certain appropriate adjustments [9]. The present investigation is bifurcated into two distinct segments: The first stage involves comprehending the effects of thermal aging on the halogen-free substrate core within the dielectric fluid utilized for single-phase immersion cooling, with regards to its thermomechanical characteristics. Another goal of the research is to gain an understanding of the impact of thermal aging on the thermomechanical properties of the substrate core exposed in air at high temperature. The halogen-free laminate (IS550H) is utilized by individuals who necessitate exceptional thermal stability, as well as elevated power and voltage. The present study involves subjecting the substrate core to synthetic hydrocarbon fluid (EC110), polyalphaolefin 6 (PAO 6), and ambient air for a duration of 720 h at two distinct temperatures, namely, 85 °C and 125 °C. The complex modulus and the glass transition temperature (T_g) of the substrate core are subsequently measured using Dynamic Mechanical Analyzer (DMA) before and after the aging process, and a comparative analysis is performed.

The halogen-free substrate samples had a thickness of 0.25 mm and were trimmed to dimensions of approximately 50 mm × 5 mm. A set of 24 specimens were fabricated in preparation for DMA measurements. The selection of experimental parameters, namely, temperature (85 °C and 125 °C) and duration (720 h or one month), was made in accordance with the OCP guidelines and ASTM 3455. Figure 1 depicts the primary constituents of DMA, while Fig. 2 illustrates the sample mounted to the tensile probe. Figure 3 shows the thermal chamber which houses substrate samples that are submerged in EC100, PAO 6, and exposed to air. Table 1 provides information regarding the quantity of samples that are situated in each fluid.

where E* is the elastic modulus, E′ is the storage modulus, E″ is the loss modulus, and tan $δ$ is the damping ratio.

The sample dimensions were measured utilizing a digital caliper that possessed an accuracy of 0.02 mm. The selection of the tensile attachment for the present study was based on the projected modulus derived from the sample dimensions and the material properties. Prior to being mounted onto the DMA for testing, the samples immersed in dielectric were appropriately cleaned using a paper towel. The DMA measurements were conducted with a fixed L amplitude of 10 μm and a force amplitude of 2000 mN for all samples. The experimental data were collected by exposing the specimens to a temperature range spanning from −40 °C to 280 °C, with a consistent ramp rate of 4 °C/min. Additionally, a range of frequencies, including 0.5, 1, 2, 5, and 10 Hz, were applied during the testing process. The auto-LN2 gas cooling unit facilitates the dispensation of liquid nitrogen to lower the temperature of the furnace below ambient temperature [13]. A thermal equilibrium was achieved by subjecting the system to an isothermal hold at the initial temperature, i.e., −35 °C. This resulted in an extension of the measurement duration of the DMA to approximately 2.5 h. Table 2 presents the parameters utilized for the purpose of testing.

The specimens underwent analysis using DMA, and the complex modulus was determined through the examination of the storage modulus and loss modulus data gathered during the testing procedure. Four samples were used for each test case, and the average value was documented. The figures illustrate the values of elastic modulus, with the error bars indicating the standard deviation.

The elastic modulus of the specimens exposed air at 85 °C and 125 °C is illustrated in Fig. 3. The error bar shows the standard deviation with a value of ±1 GPa. The modulus of elasticity of the sample subjected to air at 125 °C is greater than that of the sample exposed to 85 °C. Furthermore, the modulus values of both sample variants exhibit a decreasing trend as the temperature is raised from −35 °C to 180 °C in the DMA. The consistency of the modulus variation between the two samples remains constant within the temperature range of −35 °C to 180 °C. The data depicted in Fig. 4 illustrate an identical trend in samples that were submerged in PAO 6 at both 85 °C and 125 °C. However, the samples that were immersed in EC100 at 85 °C and 125 °C chamber temperature as shown in Fig. 5 exhibit the same modulus at all temperature ranges from −35 °C to 180 °C.

The modulus trends of samples subjected to immersion in EC100, PAO 6, and air at 85 °C are presented in Fig. 6. Analysis reveals that the elastic modulus of substrates immersed in PAO 6 is higher compared to those immersed in EC100. A minor deviation in modulus values is detected between air-exposed samples and those immersed in EC100. As illustrated in Fig. 7, when the substrates were subjected to 125 °C, the modulus of samples immersed in PAO 6 was found to align closely with those exposed to air. Conversely, samples immersed in EC100 exhibited a lower modulus relative to both air-exposed and PAO 6-immersed samples. The observed variations in modulus values across samples subjected to thermal aging and air exposure at 85 °C raise concerns about material performance under varying environmental conditions. These fluctuations echo findings from Bhandari et al. [12], suggesting a consistent pattern across different studies.

Figure 8 displays the results window following the completion of DMA tests conducted at various frequencies. As depicted in the figure, DMA yields curves for damping ratio, loss modulus, and storage modulus. The T_g is determined by identifying the peak temperature of the loss modulus or the onset of storage modulus analysis. These features are often recognized for their sharp and smooth characteristics, which simplify the process of determining T_g [13]. However, the presence of various additives and processing techniques can broaden these peaks, complicating T_g determination. The physical significance of the loss modulus peak lies in molecular motion dynamics. As the material approaches this peak, energy dissipation intensifies due to the cooperative movement of large polymer segments. Paradoxically, this cooperative movement also renders the material more deformable. Thus, the observed peak reflects the intricate interplay between increased viscous behavior and enhanced deformability.

As indicated in Table 1, the substrates underwent an extensive 720-h exposure to an elevated temperature of 85 °C, aimed at scrutinizing the influence of immersion fluid on the T_g. Figure 9 presents the average T_g outcomes for samples immersed in air, EC100, and PAO 6, determined by identifying the peak temperature of the loss modulus. Notably, the average T_g values for air, EC100, and PAO 6 are 203.5 °C, 200.3 °C, and 203.3 °C, respectively. These data show a remarkable consistency in T_g across diverse immersion fluid conditions, suggesting an insubstantial impact on the substrate's T_g. The comprehensive examination underscores the substrate's resilience in maintaining its T_g across diverse environmental conditions, promising versatility for various industrial and technological applications.

The increasing adoption of immersion cooling can be linked to the growing demand for high-performance computing and the rise in power density. Reliability remains a top concern in the electronic packaging industry. This study focused on the reliability of halogen-free substrate cores subjected to thermal aging at 85 °C and 125 °C.

The modulus values of these substrates were then compared with those of substrates that were exposed to air at the same temperatures. Consistent with expectations, the modulus exhibited a negative correlation with rising temperature. A minor deviation in modulus values is detected between air-exposed samples and those immersed in EC100. When the substrates were subjected to 125 °C, the modulus of samples immersed in PAO 6 was found to align closely with those exposed to air. However, for specimens subjected to a temperature of 85 °C in air, PAO 6, and EC100 fluid environments, an irregular pattern was detected. Further analysis is necessary to gain a better comprehension of this phenomenon in the halogen-free substrates aged at 85 °C. The study also explored the impact of immersion fluids on the T_g of the samples. The results showed consistent T_g values across various fluid conditions, indicating minimal influence of immersion fluids on the substrate's T_g. This consistency highlights the resilience of halogen-free substrates across different environmental conditions. Further research is needed to explore the long-term effects of different immersion fluids and environments on these substrates for more comprehensive insights.

The authors attest that all data for this study are included in the paper.

E* =

elastic modulus

E′ =

storage modulus

E″ =

loss modulus

T_g =

glass transition temperature

tan $δ$ =

damping ratio