Abstract
With IP traffic increasing by 10-fold over the last decade, together with limitation and cost increase due to shrinking semiconductor nodes have led to requiring technological breakthrough in the packaging of semiconductor devices especially those used in high performance computing (HPC).This increase in IP traffic has led to requirement for higher data speed transmission in these devices, and consequently packaging technologies that enable those solutions such as 2.5D packaging utilizing silicon interposers. Furthermore, in recent years, increasing number of dies are placed in a single package for these devices thereby making the size of silicon interposers larger. Thus, the design of organic substrates used in these devices, are also becoming ever complex often with multiple layers with long trace lengths for routing increased number of IOs and allowing for power and signal control management. In order to facilitate the high speed data transmission requirement with longer trace lengths, stable low insertion loss design of organic substrates are becoming significantly important even when devices are exposed at elevated humidity or higher temperatures due to surrounding environment or from dies heating. Additionally, as silicon interposers are increasing in size, preventing stress build-up, which can cause cracking between the interposer and the organic substrate, is also becoming paramount. These requirements have led to innovative materials to be developed to enable organic substrates to have these properties. In this paper, we present a new dielectric build-up material for use in advanced organic substrates, by combining newly developed original resin with existing formulation technology that meet these criteria of enabling lower insertion loss with design that reduces deterioration even at elevated humidity and temperature, and furthermore having high crack resistance during temperature cycle testing.
I. Introduction
Internet Protocol (IP) data traffic has increased 10-fold over the last decade and is projected to continuously increase driven by increased number of connected devices especially Machine to Machine (M2M) devices, uploading and streaming of higher definition videos, use of artificial intelligence (AI) and big data technology [1]–[4]. To handle the increased IP data traffic, infrastructures utilizing high performance computing (HPC) with state-of-the-art semiconductor devices are becoming more prominent where these semiconductor devises require technologies that enable higher speed data transmission. This drive in semiconductor device technology has meant that technological breakthrough in the packaging of these devices has become more important and new solutions such as 2.5D packaging utilizing silicon interposer has emerged in recent years [5]. To satisfy the ever-increasing computation required from these devices, 2.5D devices have also evolved with greater numbers of High Bandwidth Memories (HBM) placed, utilizing larger sized logic dies and different types of chiplet designs being incorporated. This has led to packaging of these devices with larger form factors and requiring greater routing from the die to the PCB with greater IO counts. Furthermore, greater thermal emission from the dies due to the more powerful chip requiring greater power consumption meant that thermal management of these devices are also becoming more important.
Flip-chip packaging utilizing organic build-up laminate substrates fabricated using semi-additive process are still commonly used for these high-end devices [6]. Semi-additive process manufacturing allows for fine line and space designs leading to high density, large IO count packaging with high yields and lower cost as several substrate manufacturers already have tools necessary for the manufacturing process. However, as mentioned earlier, as the technical requirements for the packaging of these devices increases, greater challenges have emerged on the organic build-up substrates also. Large IO counts require greater number of routing lines which increases the necessity for more build-up layers in the substrate as well as larger sized substrates. This has led to several times longer trace length and coupled with the faster signal transmission requirement with signals beyond 10GHz now commonly used, a design that enables much lower insertion loss is required. Furthermore, the thermal emissions from the dies means that the insertion loss of the signals need to be low even at elevated temperatures. Larger sized substrates with high layer counts have also increased the stress level on the substrates both during the assembly process of the devices as well as during the operation of the devices. Combined with the requirements for the substrates to connect not only dies and PCBs but Si interposer and PCBs for 2.5D packaging have meant that there are higher degree of coefficient of thermal expansion (CTE) mismatching with the potential of higher risks of reliability troubles such as bump joint reliability, die cracking and substrate cracking. These technological requirements necessitate better performance from the materials that comprises the organic build-up laminate substrates.
One of the major component of the build-up laminate substrates is the build-up dielectric material. Epoxy based build-up dielectric materials with no fiber-glass reinforcement are currently commonly used enabling finer trace pitch designs. The ideal material for these technical requirements will have
Design to enable low insertion loss performance even at elevated temperatures
Low CTE to match that of copper or silicon
High fracture toughness as previously experiment has shown high fracture toughness reduces cracking especially with large sized substrate during temperature cycle test (TCT) [7]
Have good semi-additive process manufacturability
However, achieving all of these performances have proven to be a continuous challenge. This paper introduces a new build-up dielectric material that has been developed with the goal to meet all of the above technical targets by combining newly developed original resin with existing formulation technology.
II. Material Target Specification
As mentioned previously, epoxy based build-up dielectric materials have been commonly used in advanced organic build-up laminate substrates for over 20 years [6]. In order to develop a new build-up dielectric material, it was important to make sure that the material had good semi-additive process manufacturability as well as having the targeted performance while not using exotic materials that may increase the cost.
One of the steps during semi-additive process is using CO2 or YAG lasers to etch vias on the build-up dielectric material. During this process, smears from melted resins are formed which need to be removed. Permanganate solutions are often used to remove these smears in a process called the desmear process. During the desmear process, the surface of the dielectric material is also etched and roughened. After the desmear process, Cu traces are plated and the roughened surface of the dielectric material allows for adhesion strength with the Cu traces. Epoxy based build-up dielectric material has shown to have both good desmear-ability (the ability for the permanganate solution to remove any of the smears formed) and a relatively controlled reactivity with the permanganate solution to form a roughened surface to maintain good adhesion with the Cu traces and this relatively good manufacturability has kept the build-up materials to be epoxy based for many years. However, designing epoxy based materials which enable low insertion loss both at room temperature / standard humidity as well as when temperature and humidity are elevated have been challenging.
Although several other dielectric materials that has the potential to lower insertion loss, such as polyimide (PI), polyphenylene ether (PPE) and liquid crystal polymer (LCP) have been examined previously, all of them had issues with the manufacturability, for example the desmear-ability as materials which enable low insertion loss tends to have low polarity which restrict reactivity during the desmear process.
Another challenging aspect of developing a new build-up dielectric material is achieving low CTE similar to that of copper while maintaining high fracture toughness. Nevertheless, for this study, epoxy based build-up dielectric material was chosen to be developed for the potential ease of semi-additive process manufacturability. The target of the material was to enable low insertion loss beyond the signal frequency of 10GHz (equivalent of beyond 20Gbps when signal is transmitted using NRZ or beyond 40Gbps when signal is transmitted using PAM4) even during elevated temperatures and humidity thus having similar insertion loss performance as LCP [8]. Furthermore, the material needed to have low CTE similar to that of copper and have high fracture toughness, which is a trade-off relationship.
III. Material Design for Development
Epoxy based build-up dielectric materials have material matrix consisting of epoxy resins, hardeners (curing agent) and polymers. In order to achieve low dielectric loss tangent and subsequently low insertion loss, it is important to reduce the responsiveness of the molecules to the alternating current (AC) signals. Choosing the right type of hardener and the backbone system is key for this. As epoxy resins are thermosetting polymer, hydroxy groups are formed during standard curing and this leads to higher Df and consequently higher insertion loss of the substrate shown in Fig. 1 (A). Therefore, molecules that forms lower numbers of polar chemical groups were selectively picked as shown in Fig. 1 (B). However, with this approach, due to reduction in the number of cross-linking points, or cross-linking bonds, the molecular motion increases as environmental temperature increases. Therefore, even though the material with this design may lead to low insertion loss at room temperature, as the molecular chain expands through thermal expansion, the molecules will start to respond to the AC signal causing worsening of insertion loss. Thus, in order to achieve the desired material properties, a unique material shown in Fig. 1 (C), which is both symmetric and bulky, was specifically designed internally at Sekisui and incorporated. By using this, it was hoped that when thermal expansion causes the polymer chain to expand, it can lead to lower insertion loss due to the molecules having less responsiveness to the AC signal. Additionally, this design reduces the concentration of polar chemical groups in the molecules enabling lower moisture absorption and should lead to stability against any humidity.
Furthermore, in order to increase the toughness of the polymer, a flexible segment was introduced to the overall structure. Generally, addition of flexible segment into this type of polymers separates the segment from the thermosetting component. This makes it difficult to incorporate this type of segment into these types of polymers. However, by optimizing the structure and proportion of each of the segments of the polymer to homogenize the flexible component into the thermosetting component, the target was to achieve a material with high fracture toughness.
IV. Evaluation Result
A. Material Properties
The physical and electrical properties of the newly developed material are listed in Table 1. All measurements were taken after the material was fully cured. Df and dielectric constant (Dk) were measured using a cavity-type resonator. Mechanical properties were measured using a tensile tester, coefficient of thermal expansion (CTE) was measured using Thermo Mechanical Analyzer (TMA), and water absorption was measured using the JIS K7209A standard. The result shows that the developed material achieves a Df value of 0.0023 when measured at 5.8GHz similar to that of LCP and also exhibit low water absorption rate of 0.06%. Furthermore, the material has CTE of 24ppm/°C, very similar to that of copper while elongation was over 3%.
B. Insertion Loss measurement using Strip Line Coupon
To confirm the effect of the low Df of the newly developed material, an electrical measurement coupon was fabricated with a stripline design to measure the insertion loss. Fig. 3 shows the schematic diagram of the structure of the test coupon. The following condition was used to measure the insertion loss.
Environmental temperature: at 25°C and at 110°C
Frequencies: Between 0.1 to 40GHz
Fig. 4 shows the insertion loss of the newly developed material measured under insertion loss parameter S21 at different frequencies and temperatures. The measurement was compared with currently available epoxy based low Df build-up dielectric material (Conventional Market Material). The result shows that compared with the Conventional Market Material, the newly developed material has lower insertion loss and the difference is more significant as the frequency region exceeds 5 GHz. Moreover, the developed material exhibits lower temperature dependence compared to the Conventional Market Material. The developed material has insertion loss of less than −5dB/inch at 110°C with frequency of 40GHz. Compared to the Conventional Market Material, this is a 22% reduction in insertion loss. This could lead to more freedom of substrate design as there are less concern for insertion loss at different operating temperatures.
Characteristics of parameter S21 with frequency and temperature dependence
Insertion loss was also measured before and after highly accelerated temperature and humidity stress test (HAST) at each frequencies under the following condition.
Environment temperature: 25°C
Frequencies: Between 0.1 to 40GHz
HAST Condition: 130°C and 85% RH for 96 hours
As shown in Fig. 5, the developed material exhibited insertion loss of less than −5dB/inch at 40GHz after the material went through HAST condition. This is a 23% reduction in insertion loss compared to the Conventional Market Material under the same condition.
Characteristics of parameter S21 with frequency dependence after HAST (130°C and 85% RH for 96 hours)
Characteristics of parameter S21 with frequency dependence after HAST (130°C and 85% RH for 96 hours)
This improvement in insertion loss even after the material going through HAST is attributed to the fact that the developed material has a new original resin. As the data suggest, the developed material is stable during environmental loading when used as a package material and may allow its application in a wider field in the future especially in areas of high frequency transmission range and consequently faster data transmission.
C. Elongation Curve (S-S Curve)
Fig. 6 shows the stress-strain curve of the developed material and Conventional Market Material measured by a tensile tester. The area under the stress-strain curve represents the energy required to break the material. The result shows that the developed material exhibits nearly twice as much elongation and the fracture toughness (i.e. the energy required to break the material) more than 2.2 times that of the Conventional Market Material.
The reason why the developed material exhibits high fracture toughness can be explained by the distribution of the polymer structure. Fig. 7 is the cross section SEM image of the polymer structure before optimization and the polymer structure after optimization. The left image (a) shows a polymer structure where it is largely separated after thermosetting. With this type of distribution, the material exhibits low fracture toughness. Meanwhile, the right image (b) shows the polymer structure homogenized. This homogenization of the polymer structure leads to the high fracture toughness of the developed material.
Cross-section SEM after thermosetting (Left): Before optimization of polymer structure(a) (Right): After optimization of polymer structure(b)
Cross-section SEM after thermosetting (Left): Before optimization of polymer structure(a) (Right): After optimization of polymer structure(b)
D. Folding Endurance Test
By introducing a flexible segment structure to the resin, the developed material exhibited better elongation as shown by the stress-strain curve in Fig. 6. Oftentimes, polymers with high elongation are effective in significantly relieving the fatigue caused by the elongation and contraction forces. These properties are especially important for materials used in large sized substrate as these substrates go through multiple heat cycle during manufacturing, assembly and reliability testing. These temperature cycle can put significant strain on the substrate and consequently material used in the substrate causing cracks and delamination. To evaluate the fatigue properties of the developed material, the fold endurance was tested using the MIT test method shown in Fig. 8. Fold numbers based on ISO standard in different bend radius were examined. The result outlined in Table 4. showed that the developed material exhibited much larger folding endurance compared to the Conventional Market Material proving that the longer elongation of the developed material can contribute in reducing the fatigue of the material when going through multiple contraction and elongation forces.
MIT test machine and measurement conditions (Equipment: Toyo Seiki Seisaku-sho, Ltd.)
MIT test machine and measurement conditions (Equipment: Toyo Seiki Seisaku-sho, Ltd.)
E. Surface Roughness after Desmear Process
Fig 9. shows the SEM image of the surface roughness of the developed material after the desmear process and the via image fabricated by CO2 laser and its corresponding properties.
Surface roughness observation after the desmear process (SEM), and 60 um diameter via hole by CO2 laser after desmear process (SEM)
Surface roughness observation after the desmear process (SEM), and 60 um diameter via hole by CO2 laser after desmear process (SEM)
As mentioned earlier, it is difficult to obtain good adhesion strength with Cu when the surface roughness is low. In addition, when the material has low water absorption, it becomes difficult to remove any of the smears formed on the bottom of the via using permanganate solutions as low water absorption material has lower polarity which means less reactivity to other chemicals.
However, as the developed material has high elongation, an adhesion strength of greater than 0.45kgf/cm was observed despite having a low surface roughness. Furthermore, as shown in the via bottom image, the introduction of an original resin skeleton into the material structure allowed for high smear removability even though the material maintained the low water absorption properties.
V. Conclusion
A new epoxy based build-up dielectric material with low Df, as low as LCP, coupled with low water absorption was developed.
The developed material was able to demonstrate stable low insertion loss even under high temperature and high humidity conditions. It also showed the material has high durability against repeated expansion and contraction, because of the strong energy required to break the material. In addition, the material was designed to be compatible with conventional semi-additive process manufacturing.
The newly developed material, therefore, has a potential to be a useful material for organic substrates used in advanced device packaging, including 2.5D packaging which requires high-speed data transmission and high crack resistance.