Abstract
The implementation of microelectronics, also known as multi-chip modules (MCM), is extensive in automotive, downhole and aerospace applications. MCMs have already demonstrated high-temperature performance, step improvement in reliability, and the potential to reduce product cost through miniaturization and integration of more functions. However, there are barriers preventing wider adoption of MCM technology in downhole applications. High non-recurring expenditures (NRE) charges increase development costs. Long substrate lead times prolong the time to market. Lengthy design iterations make it difficult to apply lean startup methodology to accelerate innovation. The main factor that leads to high NRE and long lead times is the complexity of substrate manufacturing processes. Together with assembly, MCM manufacturing comprises at least 11 steps, 6 different materials, 10 or more different machines, and requires a minimum of 6 supporting employees.
A new concept proposes a simplified process to reduce labor and expenses. With best implementation, this process would require only a single machine capable of cycling through 3-step process of dispensing, placement and cure. Despite the dramatically simplified process, the constructional complexity of circuits can still be very high, such as a 3D multilayer MCM. In this paper, this concept was evaluated, micro-dispensing equipment was used to create basic circuitry blocks. Different materials to create conductive traces, isolation layers and wire bond replacement were evaluated. High-temperature aging tests were conducted to monitor the electrical and mechanical performance under thermal stress. The feasibility of dispensing fine features using dispensing and jetting methods are presented in the study.
Conductors are a critical part in microelectronic assemblies because they create interconnects and thermal dissipation paths for microelectronics. Three different conductor materials were tested for their dispensability, resistance, continuity at temperature, and coefficients of thermal expansion (CTE) compatibility with different materials under thermal cycling. For dielectric materials, the requirements were to create various assembly constructs. The characterization included dispensability, electrical insulation, breakdown voltage, high-temperature performance, and the effects of CTE. Different approaches with different materials were tested for feasibility for wire bonding replacement. The application needs fine feature size with medium resistance lines. Consequently, the criteria for the material selection are fine particle size and medium sheet resistance. For high-power devices where heavy-gauge wires were used, jet dispensing is applicable. For other application with regular wire diameters, direct write is used. The over-all tests demonstrated the feasibility of using dispensed materials to replace wire bonds, which brings better reliability for shock and vibration, as compared to traditional wire bonds. The reliability of this approach requires a set of optimally matched conductive and dielectric materials. Three conductive materials (A, B and C) and three dielectric materials (D, E and F) were evaluated in this study.
Tested conductive epoxy A can be used for attachment of SMT components with non-tin terminals, short traces, and wire bonding replacement for 25-μm wires, but it is not ideal for fine lines(<65um). Tested conductive epoxy B can be used for fine traces (58μm), and wire bonding replacement for 25-μm wires. The resistance of that material is not ideal. Nano-silver paste can be used for long traces, heavy-gauge wire bonding replacement, pads/polygons, the sheet resistance is equivalent to 0.5Oz Cu.
For dielectrics, epoxy C can be used for crossovers, dielectric layers, and components staking. Epoxy D can be used for die edge insulation, but it is not ideal. Epoxy E can be used for crossovers and components staking. Epoxy F can be used for encapsulation and components staking.
The wire bonding replacement concept structure is established with the dielectric forming the insulation around die edge, then the conductive wires dispensed on top of it. Feasibility was confirmed, a proof-of-concept was built, and some level of thermal stress was tested on the samples.
Particle size and viscosity are critical to achieve fine features for micro-dispensing conductors and dielectrics. Periodic evaluations must be conducted to follow up on industry's progress with materials.
I. Introduction
Additive Manufacturing (AM) is the promising technology that brings many benefits to an increasing number of industries. One new direction includes applications in the fields of electronics and microelectronics. Among the potential benefits that AM brings to these industries are increased design density, the ability to manufacture directly from CAD data, reduced non-recurring expenditures, cutting numerous process setups and shortening prototype lead times [1]. The implementation of AM techniques for manufacturing of microelectronic packaging requires availability of suitable conductor and dielectric materials to create the following printed constructs:
Pads
Traces
Polygons/ground planes
Crossovers
Dielectric layers
Wire bond replacements
Vias
3D interconnects are required to connect semiconductor die to the package substrate or from the substrate to the package housing. The most common approaches are wire bonding and flip chip technologies. Wire bonding is a relatively cheap and rapid interconnection option, but it is often susceptible to sweep by encapsulation molding. On the other hand, free-standing bond wires may fail in high-frequency vibration and shock environments. Additionally, relatively high parasitic inductance of bond wires often requires external or on-chip passive components for compensation at high frequencies [2]. Although flip chip technology reduces the interconnection length and its parasitic inductance, but it is plagued by high sensitivity to mismatches in coefficients of thermal expansion (CTE) [3]. 3D printing technology enables selective deposition of electronic materials in a 3D fashion, where dielectric and metallic patterns can be fabricated directly onto largely any surface to create fully-printed, vertically-integrated electronic systems and packages.
Fig. 1 illustrates the cross section of AM-multi-chip module concept, where traces (black), insulations (blue) and interconnects (black) are deposited in the AM approach. Then the components and dies (yellow, green and red) are placed to complete the module. This work covers implementation and testing of various fundamental constructs that are necessary to ascertain applicability of AM to downhole electronics.
II. AM Constructs Materials and Processes
A. Conductors
3 conductive materials (Ag based Epoxy A, Ag based Epoxy B and Nano-Silver C) were tested for their dispensability, resistance and electrical performance under thermal stress. Two deposition methods exist: direct write and jetting. The deposition approach is selected based on the construction requirements. Two types of the tests were conducted on each material: a dispensability test to evaluate the process feasibility and a resistance measurement for electrical performance. Fig. 2 is one of test vehicles in dispensability test.
Dispensability
Conductive material A is a silver-filled one-part epoxy that has excellent adhesion to flex board (e.g., Kapton), low resistance and high-temperature performance. It can be dispensed by direct write or jetting through very fine nozzles (50-μm ID). This material is very resistant to flex and creasing, which is designed to print taller narrow lines with minimal spreading. Conductive material A is well suited for pad or trace deposition where very low resistance is not required. Different constructs were tested, and the results are summarized in Table I.
Conductive material B is a single-component conductive epoxy for fine interconnects. Dispensing fine lines has already been demonstrated in vertical direction using the direct write technique, because the particle size in the epoxy is less than 5μm. As shown in Table 2, with direct writing the material can flow through 25-μm ID dispensing tip, which is the finest tip considered in this study. In addition, per manufacturer's claim, the material can produce 30-μm dots consistently, which allows additional application of interconnects to replace wire bonding, such as bumping.
Conductive material C is a thick paste of nano-sized silver powder in an organic binder formulation. The paste has a very low resistance after fully sintering, and it is suitable for applications like high-power device/modules [4]. The sintering profile is 240°C for 15 minutes. Measured resistivity is 0.01 Ohm/mm, and the smallest feature size can be achieved is 125μm in line width through a 100-μm ID dispensing tip.
Nano silver is a great material for traces, but it can be used only in direct write applications, because it is not well suited for jetting. The issues with jetting this material are most likely caused by pressure during the jetting process. The pressure breaks down the nanoparticles shells leading to some of the Ag particles sintering together. These sintered agglomerates are then clogging the dispensing tip.
Resistance vs. Temperature
To further evaluate the electrical properties under thermal stress, samples were built for resistance measurements. For each conductive material, two lengths of traces were printed with the minimum trace width. A four-point measurement was taken for each time interval at room temperature, 100°C, 150°C, 200°C and 250°C. For epoxy A, the trace width was 65um, while the trace width for epoxy B was 58um. The results are plotted in the following charts (Fig. 3 and Fig. 4). Material C becomes pure Ag after sintered process, therefore no resistance changes occurred under temperature.
For both materials, the resistance dropped within first 200 hours of aging, then stabilized throughout the test up to 500 hours of aging at 200°C. The resistance decrease during the first hours is considered to be the effect of continued curing after the initial build. A higher temperature curing profile at 230°C for 5 minutes on epoxy A is recommended to completely cure the epoxy and minimize the resistance drift. The measured sheet resistance for epoxy A was 25 milliohm per square, with the curing temperature at 230°C for 5 minutes. For epoxy B, with the curing profile of 180°C for 40 minutes, the measured sheet resistance was 54 milliohm per square. For the same conductive epoxy, with every degree of temperature increase, there was approximately a 0.24-percent resistance increase.
B. Dielectrics
Three dielectric materials (dielectric epoxy D, dielectric epoxy E, dielectric epoxy F) were evaluated for different applications. There are three main applications for the use of dielectric materials: 1) crossover for insulation between traces, 2) dielectric layer for multi-layer substrate; 3) die edge insulation for wire bonding replacement. For each construct, different materials were evaluated for its process feasibility and insulation strength under thermal stress. Fig. 5 is one application of dispensing a dielectric as an insulation layer.
Dispensability
Dielectric material D is a red, jettable, low-temperature curing one-part epoxy adhesive. This system features low moisture absorption and high green strength that prevents movement of components during subsequent handling or cure. It also has a snap cure profile of 2 to 5 minutes above 150°C. It is well suited for large area deposits such as dielectric layers. However, batch control is critical because soft agglomerates in one of the supplied batches prevented dispensing fine features. The process test results are summarized in Table III.
Dielectric material E is a blue, jettable, low-temperature curing one-part epoxy. Dielectric material E was found to be not suitable for large-area deposits because cracks developed in large areas under thermal stress. The material does not self-level, making it unsuitable for die insulation. However, small planar deposits are very robust and consistent, making it an excellent material for crossovers and components staking. The process test results are summarized in Table IV.
Dielectric material F is a grey, two-part epoxy adhesive recommended for sealing application. This system features low moisture absorption, and it has excellent solvent and chemical resistance. Dielectric material C has low thixotropic index, and it is good for large-area deposits like substrate attachment and encapsulation. The material's relatively large particle size (under 20μm) requires the dispensing needle ID to stay above 100μm. The process test results are summarized in Table V.
Electrical insulation vs. Temperature
A test substrate was made to evaluate dielectric strength between two conductors. A high-voltage dielectric breakdown test was conducted at room temperature. Increasing potential was applied between two conductors on neighboring layers with a 2-mil dielectric material D between them. The overlap area was 50 mil2. Starting at 1.5 kV, the voltage was raised until breakdown occurred at 6.65 kV (Fig.6).
The obtained number indicates dielectric D has an excellent dielectric strength of 3.3 kV/mil or 133 kV/mm, Dielectric strengths of some common materials provided below for reference and comparison:
Ceramic substrates (H/LTCC, AlN) is 0.4–0.5 (kV/mil),
PCB polyimide & Teflon is 1.5 (kV/mil),
Glass is 2–3 (kV/mil),
Pure polyimide 6.9 (kV/mil)
Table VI lists the DC insulation resistance measurements that were conducted upon completion of 500 hours aging at 200° C. For all measurement temperatures, the resistance was lower than the established passing criteria.
C. Wire bonding replacement
Process feasibility
Jet printing was performed with a PicoPulse pump with a 50-μm nozzle. Two ink materials were printed to determine the wire bonding replacement structure (Fig. 7): dielectric material E (dark green) and conductor material A (silver color). The insulation layer was created around the die edge and substrate, followed by the jetting of conductor material connecting wire bonding pads on the die to the pads on the substrate. The width of jetted interconnect was approximately 350μm.
For regular-size wire bonding replacement, direct write technology was evaluated. An nScrypt desktop 3D printing machine was used with direct-write pen tips. Two ink materials were printed to determine the wire bonding replacement structure (Fig.8): dielectric material D (red) and conductor material A (silver color). The dielectric insulation structure was created in two steps, the insulation around the die, and the thin perimeter line around the die edge. The later one was created with a 50-μm diameter pen tip with the line width of 60um, the insulation around the die was created with 200-μm diameter nozzle to form a slope to the die wall. The conductor was printed from the top of the die along the slope to the substrate using the 50-μm pen tip. The line width was 70 μm.
Resistance measurement
A test vehicle with daisy chain was produced for interface resistance measurements using conductive epoxy A and dielectric epoxy D. Probe station was used to measure the resistance across 8 interfaces. Samples were measured initially and after aging at 200°C. Findings are listed in the following:
For a 70-μm thick connection, the average initial interface resistance between conductive epoxy A and Al wire bonding pad was 10 Ohm.
3–5V were applied on each interface momentarily, and the interface resistance dropped dramatically to milliohms.
After aging at 200°C for 20 hours, the resistance increased back to initial level.
If another voltage shock was applied, the resistance dropped again.
The interface resistance increase was due to the native oxide on Al pad. The thickness of that oxide is from 15 to 50 Å. The voltage applied to the interface breaks down the Al2O3 layer, creating uninsulated tunnels between Al and epoxy, thereby dropping the resistance. During the process of curing and aging at temperature, oxygen penetrated through epoxy and facilitates the oxidization layer growth, causing the resistance to increase again.
Two mitigation approaches can be applied to reduce the interface resistance. One approach is to eliminate the oxygen presence at temperature, the other to pretreat the Al bond pad with an oxide-removing etchant. The best approach however, was to use only Au, Pd, Pt, or Ag for bond pad material for silver-filled epoxy interconnects.
III. Applications
These concept demonstrators fell into 3 categories (Fig. 9 to Fig. 11): assemblies, substrates, and flex applications. Fig. 9 is the half-bridge power module that includes printed conductors(white), 12 interconnected components (SMT components and dies), and high-voltage insulation(green). Fig. 10 is a 3-layer substrate with vias and pads. The image below is the sample circuit printed on Kapton film. Additional flexible applications include flex interconnects, magnetics, sensors and antennae. Additive methods allow fine feature sizes down to 50μm, as well as the ability to print on nearly any surface. The materials qualified are suitable for 200°C and once the process is established, manufacturing will be very economical.
IV. Conclusion
The direct application of dielectrics and conductors in additive manner for use in multi-chip modules was studied and demonstrated to be feasible. Three fundamental building blocks have been fabricated. Both process feasibility and material electrical characterization under temperature was demonstrated. This work identified three main applications of additive manufacturing in microelectronics: assemblies, substrate and flex boards. For each application, proof of concepts was built and process has been developed. Additional testing and process development for AM multichip module on flex board are ongoing.
Direct write allows for very fine features and jetting allows for high-speed deposition. Further research into reliability of AM-MCM including comprehensive testing is required prior to full commercialization of this process. The two biggest challenges are process repeatability and reliability. Reliability aspects include evaluation of interface resistance under thermal stress, structure integrity under thermal and mechanical stress. It is expected that continued research and development of materials and equipment will make this a commercially viable technology within the next few years.
