Characterizing the properties of thermal interface materials under actual application conditions.
By Dambi Jo, JiHyun Kim, GaHyeon Kim, TaeKyeong Hwang, WonChul Do, and KyungRok Park
In the semiconductor packaging industry, it is common practice to predict warpage and stress levels through simulation prior to actual product fabrication. This approach helps minimize the cost and time associated with prototype production [1]. The need for simulation-driven design and material optimization has grown increasingly critical in recent years, as the cost of producing high-performance package prototypes has increased significantly. However, as packages become more advanced and materials more diverse, accurate prediction through simulation has become increasingly challenging, thereby highlighting the growing importance of simulation validation [2].
Recently, the demand for high-performance semiconductor packages—such as those used in artificial intelligence (AI), high performance computing (HPC), and automotive applications— has been growing rapidly. These advanced packages inherently generate significant heat during operation. Since the heat generated from the die directly affects package performance, efficient thermal dissipation is critically important. In response to this trend, thermal interface materials (TIMs) positioned between the die and the package have emerged as a key factor in thermal management [3]. Metal TIMs, which exhibit higher thermal conductivity than conventional polymer TIMs, are actively being adopted [4].
For packages utilizing polymer TIMs, mechanical behavior is well-characterized due to years of accumulated data, and simulation results tend to correlate well with actual behavior. In contrast, metal TIM packages are relatively new, and relevant data remains limited. Typically, simulations rely on material properties provided by suppliers, which are based on bulk specimens. However, applying these bulk-based properties directly to simulations often leads to discrepancies with actual package behavior. This is primarily because metal TIMs are measured in bulk form with high elastic modulus, whereas in real applications, they are implemented as thin films. Therefore, it is essential to characterize the material properties of metal TIMs under actual application conditions [2].
In this study, a sufficient number of metal TIM packages were evaluated using the shadow moiré technique to quantitatively analyze mechanical behavior across temperature variations [5]. During simulation, the material properties of the TIM were empirically adjusted to inductively derive values that closely match their actual behavior. To experimentally validate the simulation-derived properties, various material characterization techniques—including DMA, tensile testing, and rheometer—were conducted [6]. Through this process, the most suitable testing method showing the highest correlation with simulation results was identified.
The goal of this research is to establish a simulation model capable of accurately predicting the mechanical behavior of metal TIM packages and to support this model with experimental data. This will contribute to enhancing design reliability and reducing development costs and time in the fabrication of high-performance semiconductor packages.
To investigate the mechanical behavior of metal TIM packages, a lidded flip chip ball grid array (FCBGA) test vehicle incorporating a metal TIM was fabricated and warpage measurements were conducted using a shadow moiré system.
The metal TIM test vehicle (TV) used in this study includes a die with dimensions of 25.6 mm × 25.6 mm and a thickness of 0.55 mm. A hat-type lid with a thickness of 1.5 mm was assembled. The metal TIM was applied using a 0.4-mm thick indium 10 silver (In10Ag) preform. In the case of the polymer TIM test vehicle, all other conditions were kept identical, except for the die thickness and TIM material. The die thickness was 0.78 mm and a gel-type polymer TIM with a thickness of 0.07 mm was applied.
The metal TIM test vehicle (TV) used in this study includes a die with dimensions of 25.6 mm × 25.6 mm and a thickness of 0.55 mm. A hat-type lid with a thickness of 1.5 mm was assembled. The metal TIM was applied using a 0.4-mm thick indium 10 silver (In10Ag) preform. In the case of the polymer TIM test vehicle, all other conditions were kept identical, except for the die thickness and TIM material. The die thickness was 0.78 mm and a gel-type polymer TIM with a thickness of 0.07 mm was applied.
Warpage is the most used indicator for evaluating the mechanical behavior of semiconductor packages. In this study, warpage was characterized and quantified in accordance with the JEDEC JESD22-B112 standard [7]. To enable temperature measurements during testing, a dummy sample identical to the actual package was prepared and a thermocouple was attached to the bottom surface of the substrate. The dummy sample was placed at the center of the shadow moiré system, surrounded by actual packages. The temperature was increased from 25°C to 260°C at a rate of 5°C/min, and the warpage of the substrate’s bottom side was measured as a function of temperature. In addition to measurements on the final package, shadow moiré tests were also conducted after the underfill process to observe mechanical behavior changes due to processing.
Shadow moiré measurements revealed that both polymer and metal TIM packages exhibited similar warpage behavior. As shown in Table 1, after the underfill process, the “crying mode” warpage was observed at room temperature, while “smile mode” warpage appeared at elevated temperatures. In the final package state, as shown in Table 2 and Table 3, both TIM types showed a “W”-shaped warpage at room temperature and an “M”-shaped warpage at high temperatures. Notably, both packages exhibited a relatively flat profile around 150°C, suggesting that this temperature can be considered the deformation-free temperature for simulation purposes.
Table 1: After Underfill Status shadow moiré results.
Table 2: Metal TIM Package EOL Status shadow moiré results.
Table 3: Polymer TIM Package EOL Status shadow moiré results.
In this study, mechanical simulations were conducted using material property data provided by the supplier. As shown in Table 4, the elastic modulus of the metal TIM was reported to be 15.5 gigapascals (GPa), which is significantly higher than that of the polymer TIM. All material configurations of the two packages were identical except for the TIM, and the material properties used in the simulation are listed in the table.
The simulation model was constructed using a quarter-symmetry structure, and all material properties were assumed to follow linear elastic behavior as a function of temperature. The substrate was modeled with a full-layered structure, and all bonding interfaces within the package were defined as perfectly bonded. Simulations were performed for different process conditions, and the corresponding models are illustrated in Fig. 1(a) and Fig. 1(b).
Table 4: Supplier provided TIM Material Property.
Fig. 1: Quarter symmetric Lidded FCBGA model, (a) After Underfill status, (b) EOL status.
Simulations were first conducted on the package in the after underfill condition, where the TIM, lid and lid adhesive were not yet applied. As shown in the model illustration of Table 5, the simulation results revealed a distinct crying mode warpage, which was consistent with the shadow moiré measurements. Simulations were performed across the entire temperature range used in the shadow moiré tests, and the results showed a high degree of agreement with the experimental data.
Table 5: After Underfill status simulation results.
Simulation was performed on the final package configuration, which included the TIM, lid, and lid adhesive. As shown in Table 6, the metal TIM package exhibited a complete smile-mode warpage at room temperature. This result differs from the “W”-shaped warpage observed in the actual shadow moiré measurements. As illustrated in Fig. 2, the discrepancy between the simulation and experimental results is most pronounced in the room-temperature range.
In contrast, the polymer TIM package showed excellent agreement between simulation and experimental measurements, as confirmed in Table 7 and Fig. 3. Both the warpage values and shapes across the temperature range matched closely. Since all material properties except for the TIM were kept identical in the simulation, this discrepancy is presumed to originate from the material properties of the TIM itself.
Table 6: EOL status simulation result – Metal TIM Package.
Fig. 2: Simulated warpage of metal TIM package (EOL condition) using supplier-provided properties across temperature range.
Table 7: EOL status simulation results – Polymer TIM Package.
Fig. 3: Simulated warpage of polymer TIM package (EOL condition) using supplier-provided properties across the temperature range.
As shown in Fig. 4, the material properties of the metal TIM provided by the supplier were obtained through tensile testing using bulk dog-bone specimens. However, the metal TIM used in actual packages is a thin film with dimensions of 25 mm × 25 mm and a thickness of 0.4 mm, which differs significantly in geometry and thickness from the test specimens. Due to these differences, it was deemed inappropriate to directly apply the material properties measured from bulk specimens to simulation models.
To address this issue, this study conducted a series of material characterization tests on the film-type metal TIM used in actual packages.
Fig. 4: Supplier used bulk dog-bone specimen of metal TIM.
The film-type metal TIM used in the experiments was a thin film with dimensions of 25 mm × 25 mm and a thickness of 0.4 mm. Each specimen was processed according to the specific conditions required for the respective test methods. Four types of tests were conducted: tensile test, dynamic mechanical analysis (DMA), nano indentation, and rheometer.
The results from the tensile test, DMA, and nano indentation showed modulus values that were either close to or lower than the high values provided by the supplier. In contrast, the rheometer test yielded significantly lower modulus values. The data obtained through the rheometer can be seen in Fig. 5.
The key distinction of the rheometer test lies in its ability to incorporate process effects that occur during actual package fabrication. Unlike the other three methods, the rheometer test was designed to apply forces to the film-type metal TIM in a manner that closely mimics the conditions during package assembly. As shown in Fig. 6, the rheometer test applies force in both the top and bottom directions of the film, reflecting the nature of the process.
Fig. 5: Rheometer result of film-type metal TIM.
Fig. 6: Schematic of the rheometer test setup for film-type metal TIM.
The material characterization results confirmed that the modulus measured via rheometer testing was significantly lower than that obtained from bulk specimens. This rheometer-derived modulus was applied to the simulation to reproduce the warpage behavior of the metal TIM package.
As a result, the simulation using the rheometer-based modulus showed a high degree of agreement with the shadow moiré measurements, both in warpage shape and magnitude. Notably, as shown in Fig. 7 and Table 7, the simulation accurately captured not only the absolute warpage values across temperatures but also the shape transition from “W” to “M” mode.
These findings suggest that, for accurate simulation of metal TIM packages, material properties should be measured based on the actual film-type TIM used in the package rather than bulk specimens. Furthermore, the material data applied in simulations must reflect the directional forces and process effects encountered during real package assembly.
Fig. 7: Simulated warpage of the metal TIM package at EOL condition using rheometer-derived material properties across the temperature range.
Table 8: EOL status simulation results.
In this study, the mechanical behavior of packages incorporating a metal TIM was quantitatively analyzed through shadow moiré testing, and warpage characteristics were evaluated by comparison with polymer TIM packages. The results confirmed that metal TIM packages exhibited similar warpage behavior to polymer TIM packages under both EOL and post-underfill conditions.
Initial simulation studies showed good agreement with shadow moiré measurements under post-underfill conditions. However, for metal TIM packages in the EOL state, the simulation results did not match the actual measurements. In contrast, the simulation results for polymer TIM packages at EOL showed strong correlation with experimental data, indicating that the accuracy of TIM material properties plays a critical role in simulation reliability.
To improve simulation accuracy, it was essential to obtain precise material properties that reflect the actual application conditions of the metal TIM. Therefore, instead of using bulk dog-bone specimens provided by suppliers, various material characterization tests were conducted on film-type metal TIM used in real packages. The results revealed that rheometer testing, which considers process-induced effects such as directional forces during assembly, yielded significantly lower modulus values compared to bulk specimens.
Applying the rheometer-derived modulus to the simulation resulted in a high level of agreement with shadow moiré measurements, both in warpage shape and magnitude. Notably, the simulation successfully reproduced the temperature-dependent warpage transition from “W” to “M” shape.
This study empirically demonstrates the limitations of using bulk specimen-based material properties in simulations and emphasizes the importance of acquiring material data that reflects both the actual specimen and the process effects encountered during package fabrication for metal TIMs. These findings provide a foundation for improving design reliability and effectively reducing prototype fabrication costs and development time in high-performance package development.
Future work will extend this research to include material characterization and simulation studies for various metal TIM compositions (e.g., In-Ag alloy, copper (Cu)-based), contributing to broader optimization in package design.
JiHyun Kim is a researcher at Amkor Technology Korea.
GaHyeon Kim is a researcher at Amkor Technology Korea.
TaeKyeong Hwang is a senior director at Amkor Technology Korea.
WonChul Do is a product development group manager at Amkor Technology Korea.
KyungRok Park is a senior director at Amkor Technology Korea.
Leave a Reply