Reliability of Novel Ceramic Encapsulation Materials for Electronic Packaging Open Access

The development of current and future power electronic systems focusses on miniaturization and increased power density, especially for efficient automotive electronics in harsh environments. Wide bandgap (WBG) power semiconductors, such as silicon carbide (SiC) and gallium nitride (GaN), are able to process electric power at higher voltages and temperatures with reduced die areas and less power losses. Combined with decreased cooling requirements and smaller passive components, it is possible to receive significantly reduced volumes of power electronic devices at lower costs [1–4].

However, the size reduction and the higher switching frequencies of WBG semiconductors result in an increased power loss density and higher operating temperatures up to 300°C. Therefore, the development of innovative materials, which are able to operate at higher temperatures, is a major focus within the research of advanced electronic packaging technologies [5–7]. Especially, encapsulation materials for active and passive electronic components have to be improved in terms of operating temperatures above 175°C. Common epoxy compounds or silicone gels show limited lifetime above 200°C because of thermal degradation effects. Thus, improved polymeric mold compounds, operating even up to 300°C, are under development [8]. However, polymer-based encapsulation materials are limited in thermal conductivity (<2 W/[m·K]) or workability. Especially, for epoxy resins with a higher thermal conductivity above 1 W/(m·K), a special processing, applying heat and pressure, is obligatory for proper gap filling behavior.

Therefore, novel ceramic encapsulation (CE) materials, based on calcium aluminate cement (CAC) and phosphate cement (PC), with increased thermal conductivity above 5 W/(m·K), temperature stability up to 300°C, and enhanced thermomechanical properties have been developed to realize maximum efficiency and reliability for future power electronic devices [9, 10]. It has been pointed out that the processing of CAC-ceramic composites (CCC) or similar encapsulants, compared with common filled epoxy resins or silicone gels, is significantly simplified. Also, CCC encapsulation of miniaturized passive components such as transformers or coils working below 150°C seems to be promising. However, the dehydration behavior of CCC revealed critical aspects for the encapsulation of semiconductors and at temperatures above 210°C [9]. Furthermore, it was possible to identify the failure mechanism of CCC on Si semiconductors after extensive thermomechanical and electrical stressing, as the cement-based encapsulant reacted with Al bonding wires and the Al topside metallization (AlSiCu) [10].

Therefore, optimizations on the encapsulants formulation are necessary to enhance the compatibility with electronic materials such as Al bonding wires and Al topside metallizations, to avoid degradation effects or chemical interactions. Increased reliability and lifetime of power modules with CE should be the consequence.

The present work describes recent material developments and adaptions of CCC encapsulants, renamed to “CE” as it better describes the bulk characteristics. Enhancements of the CE materials composition, compared with the original CCC encapsulation material, will be illustrated to understand the presented advances in environmental testing and compatibility with semiconductor surfaces.

The focus of this article amplifies the improvements on CE reliability, tested by high-temperature reverse bias (HTRB), and correlations with material properties, such as microstructure, thermal behavior, thermal conductivity, and electrical resistivity. The ceramic encapsulants are also compared with common epoxy compounds or silicone gels and PC encapsulants.

In addition, promising high-humidity high-temperature reverse bias (H³TRB) results of PC encapsulants will be highlighted. This encapsulation material already performed 3.5 times better in power cycling testing (compared with silicone gel) after HTRB stability was proved. For the PC-encapsulated modules, failure analysis indicated that the failure mode shifted from a wire bond liftoff (failure mode of power modules encapsulated with common silicone gel) to a solder and die attach degradation [10].

Because of the fact that in terms of mechanical behavior, PC-based encapsulants belong to the same type of material class as CE does, these results are also relevant for the discussed CE material and upcoming H³TRB testing with CE.

The developed CE materials can be separated from nonpolymeric encapsulation systems by their matrix chemistry. Hydration reactions, where the matrix material incorporates water, represent the hardening and binding mechanism in all of the developed nonpolymeric materials (CCC, CE, and PC). The unique curing at 60°C for 3 h with further postcuring at 150°C for 3 h ensures a stable encapsulation material with proper strength values. Despite PC-based encapsulants, alumina represents the filler, promoting high thermal conductivities.

Calcium aluminate cement ceramic composites, addressed in Refs. [9] and [10], contain cement as the binding phase. In contrast to CCC, the matrix of the CE encapsulant is characterized by a hydraulic ceramic binder, named reactive alumina [11].

The main differences between CCC and the improved CE represent the lower pH value and the resulting binder phases (mostly amorphous aluminum hydroxide) during curing.

Phosphate cement encapsulation materials (PC), as used in dental applications, are formed by the reaction between an acid (e.g., phosphoric acid) with a base (commonly a metal oxide). This hydration reaction forms insoluble hydrogen phosphate salts [12].

The most relevant material properties of the discussed encapsulants are given in Table I. The comparison toward filled epoxy and silicone gels points out the main differences and highlights the advantages of CE and PC, such as higher thermal conductivities and a simplified processing. Furthermore, it is possible to fit the overall low coefficient of thermal expansion within a certain range, while maintaining the other properties.

Generally, soldered power diode (1700V Semikron CAL-diode, type SKCD81C170IHD) and sintered insulated-gate bipolar transistors (IGBTs) (1200V Infineon IGBT4, type IGC50T120T8RL) on direct copper-bonded (DCB) substrate were applied as test vehicles. The topside of the power semiconductors was contacted by 400 μm Al wire bonding. This specific design enabled a decent investigation of potential chemical and electrochemical interactions of the developed encapsulants with the involved contact interfaces of the electronic components. Fig. 1 shows the encapsulated diode and IGBT test vehicles.

For encapsulation material preparation, a premixed powder of ceramic filler and binder (cement for CCC and PC or reactive alumina for CE) was mixed with deionized water (for details, see [9]) and casted on the test vehicles using a silicone casting mold and a syringe.

The whole encapsulation process can be easily transferred to similar 1K or 2K potting systems known from silicone or epoxy processing. Necessary adaptions on such potting machines are also subject of present research activities on CE materials.

For optical investigations of contact surfaces between the electronic components and the different encapsulants, a Zeiss Axiophot light microscope with a Leica DFC450 camera was used. Microstructural features of CE compared with CCC and PC were obtained on a Zeiss Supra 35VP scanning electron microscope (SEM) with INCA Energy EDS-System from Oxford Instruments. For both analyses, the samples were prepared by dry grinding and polishing with alcohol-based lubricants down to 1/4 μm with additional ion etching and Pt/Pd-sputtering (only for SEM). A major difference of CE to CCC encapsulation materials consists in different crystalline and/or amorphous matrix phases. Therefore, qualitative and quantitative x-ray diffraction analyses were conducted using a Bruker D8 Advance x-ray diffractometer at 40 kV/20 mA from 10–50 °2θ. Adequate count statistics were achieved by a typical setup with a Cu-sealed tube, a Göbelspiegel, a .2° divergence slit and a knife edge on the primary site and a VÅNTEC-1 detector providing .4 s per step with .008° increments. All measurements were taken on untreated surfaces of cured and tempered bulk encapsulation material. Phase contents were finally calculated via Rietveld refinement [13] with SiroQuant 4.0 software.

For understanding the thermal behavior of CE and CCC encapsulants, thermal gravimetry-differential scanning calorimetry-mass spectra (TG-DSC-MS) analysis up to operating temperatures of 300°C was conducted. Thus, it is possible to determine mass changes via TG, thermal effects via DSC, and MS of degassing phases simultaneously. Measurements were conducted on an STA 409CD-QMS422 from Netzsch Gerätebau GmbH with synthetic air (75 mL/min) from 30°C to 300°C.

According to the cement nomenclature, chemical compositions are abbreviated when needed: A = Al₂O₃, C = CaO, and H = H₂O.

Environmental testing of power electronics is obligatory for many applications and especially in the automotive sector. The different qualification tests and relevant parameters are defined accordingly [14]. Therefore, interface reactions under different loading conditions can be studied and compared with untreated components.

With the HTRB test, the topside isolation properties of the semiconductor can be evaluated. Potential defects due to ionic impurities within semiconductor structures will be identified by the failure of this qualification method. Testing with high temperature (150°C) and the applied electrical field (80% of the IGBT blocking voltage) enables accelerated conditions and higher ion mobility rates (ion migration) [15]. The results of this state of high stress could even destroy the device [16].

Because of the progress in PC development, it is possible to show the improvements of nonpolymeric encapsulants in terms of stability and functionality under humid conditions. Respectively, the H³TRB test reveals deterioration effects of nonhermetic devices influenced by humidity. After the isolating encapsulation materials pass an HTRB test, H³TRB testing detects degradation effects, triggered by moisture vapor penetrating through the electronic package. As a consequence, surface corrosion and ionic migration are common failure mechanisms under this harsh environment.

All H³TRB measurements were performed on PC-encapsulated power diodes (see detailed description in the corresponding paragraph) in a climate chamber at 85% relative humidity, 85°C, and with a blocking voltage of 80 V [17]. The test fails with an increase in leakage current of more than 100% compared with its initial value [14]. HTRB and H³TRB results will be discussed and correlated with the material analysis of CE to improve the functionality of these novel and promising nonpolymeric encapsulation materials.

The microstructure of nonpolymeric encapsulants are totally different from polymers such as epoxy resins or silicone gels. Cement-based encapsulants such as CCC and the presented CE material show crystalline and amorphous phases, which build up the binding matrix around the Al₂O₃ filler particles. For best packaging density and thermal conductivity, these filler particles have a broad particle size distribution [18]. Fig. 2 shows the recently developed CCC encapsulant [9].

The SEM images illustrate a crystalline composite material with differently sized Al₂O₃ filler particles and two different hydrate phases (C₃AH₆ and γ-AH₃), resulting from the cement hydration reaction. Points 1 and 2 indicate the areas of EDX (energy dispersive X-ray spectroscopy) point analysis. It can be verified that the big lamellar shaped hydrates represent the hydrogarnet (C₃AH₆), whereas the small needle-shaped hydrate phases (darker gray between the small filler particles) are gibbsite (γ-AH₃). Furthermore, a significant amount of porosity with pore sizes up to 800 nm (measured with mercury intrusion porosimetry [MIP]) is obvious in the polished SEM image (Fig. 2a). Environmental tests such as HTRB and H³TRB should clarify whether the observed porosity has to be considered critical.

Because of the incompatibility of CCC with the chip surfaces, improvements on the encapsulant chemistry had to be considered [10]. The result of this material development represents the introduced CE material. As mentioned in the Introduction, CE consists of an amorphous aluminum hydroxide matrix surrounding the filler particles. Fig. 3 shows the CE microstructure, also on polished and fracture surfaces. Next to the filler particles, the matrix consists of two differently shaped hydrates. Bigger round particles with some needle shaped at the particle boundaries and very small needle-shaped hydrates surrounding the smallest filler particles (Fig. 3a, H and h) are visible. The phase analysis detects no significant amount of crystalline matrix phases (see next paragraph). Therefore, both of these hydrate phases have to be poorly crystalline to amorphous, regardless of their specific morphology.

CE represents a no-cement composite and there is no C₃AH₆, which especially characterizes the unique fracture surfaces of CCC (Fig. 2b). Compared with Fig. 3b, it seems like pore sizes are smaller in CE because of the absence of the calcium-bearing C₃AH₆. EDX point analysis for the different hydrates is not marked because all measurements detected similar chemical compositions. This fact is also dedicated to the absence of calcium, which made the distinction between C₃AH₆ and γ-AH₃ possible.

As porosity in general maintains at higher values (comparable with CCC), the maximum pore size decreases drastically to <100 nm. This fact, also revealed by MIP, is in good agreement with the illustrated SEM images of CE. Small pore sizes also support the conclusions on low permeability of reactive alumina composites in the literature [11].

Finally, the reported critical microcrack formation after dehydration of CCC [9] cannot be observed in the CE material. However, further thermal investigations have to be evaluated.

Changing the matrix in CE from CAC to reactive alumina drastically changes the phase content and thermal behavior. Fig. 4 shows the diffraction patterns of CCC and CE after curing and after tempering at 300°C for several hours.

Apart from the alumina filler, no other matrix phases can be detected in the CE encapsulants. In addition, only reflexes with minor intensities marked with “?” can be observed, which probably point to Na-bearing aluminum hydroxides with amounts <3 wt.%. Generally, the absence of higher crystalline aluminum hydroxide amounts supports the assumption that the CE matrix consists of amorphous aluminum hydroxides. This assumption fits to according literature data, which also predicts high amorphous contents of hydrated reactive alumina at pH values above 9.5 [19]. First in-house chemical analysis revealed significant amounts of impurities (mostly Na+-ions) due to the reactive alumina powder. Detailed studies on the formed amorphous hydrate phases and the ion mobility under different environmental conditions are part of future research work. By contrast, CCC matrix phases (γ-AH₃ and C₃AH₆) are almost completely crystalline after curing. Next to these hydrates and the Al₂O₃ filler, β-A (Na-bearing alumina) show up in CCC. The influence of heat treatment up to 300°C leads to the dehydration of all hydrate phases in CCC, which results in the extinction of the corresponding XRD (X-ray diffraction) reflexes. In addition, small amounts of CA₂ and C₁₂A₇ (cement clinker phases) are formed because of dehydration of the hydrates.

For the CE material, the diffraction pattern does not change. This fact fits to the previously stated differences to CCC matrix phases. Amorphous hydrates are not obviously detectable in XRD measurements. They only show up with a broad higher background compared with the measurement of a standard. Therefore, it can be concluded that the crystalline phases in CE after curing are also stable after heat treatment up to 300°C. For evaluating the thermal behavior of the amorphous phase content in CE, the illustrated XRD results are not sufficient.

Therefore, TG-DSC-MS measurements revealed extremely different dehydration temperatures of both encapsulation materials up to 300°C. CE dehydrates mostly below 200°C with small amounts of water loss up to 280°C. The dehydration of CCC, as discussed in [9], starts at about 210°C and ends at nearly 300°C because of the high crystallinity of such phases. For the electronic package, the thermal behavior of CE could be an advantage, as it would be possible to dehydrate the material below 200°C in a postcuring step. Thus, mobilization of crystalline-bound water, which would support critical ion migration effects, can be avoided. Furthermore, the observation of Na-bearing phases and the small amounts of water loss up to 280°C have to be evaluated critically in HTRB and future H³TRB tests of ceramic encapsulants. In terms of CCC, it is not possible to add such a postcuring step because of the needed temperature of about 300°C to lose all of the crystalline-bound water.

First material characterizations of CCC-encapsulated IGBT chips uncovered critical reactions between CCC and the chip metallization (AlSiCu) after environmental testing [10]. Consequently, the metallization corroded drastically to partly complete dissolution of the AlSiCu (Fig. 5). Al-bearing materials such as bond wires or the chip metallization are very sensitive to pH values. However, such a strong chemical interaction was not observed at the boundary surface between Al bond wires and CCC. The lower electrochemical activity of Al compared with an AlSiCu alloy could be the reason [20, 21].

HTRB test results, as discussed in the next paragraph, would also support the assumption that generally, the pH of the encapsulation material has to stay below a certain value, even before applying electronic fields to the chip. As the DCB substrate indicates no reaction with the cement-based CCC encapsulants, the lower redox potential of Al (E1/2 = −1.66 V) compared with Cu (E1/2 = +.34 V) explains the reactivity of AlSiCu, whereas the DCB copper shows no reaction at all [22]. Therefore, the attack of the AlSiCu should be avoided by reducing the pH of encapsulants. Apart from that, Cu wire bonds and optimized interconnection technologies on power modules [23] could level the critical pH aspect.

CE represent the enhancements of CCC in terms of pH values. By reducing the matrix pH from 11.5 to below 10, it was possible to create a material, which shows no negative reaction with the chip, Al bond wires, or any other part of the electronic package after hardening (Fig. 6). Changing the matrix to reactive alumina reduces the composite pH from 11.3 down to almost 8.5.

Before testing CE encapsulants in HTRB, none of the CCC-encapsulated IGBTs survived the applied voltage for 1,000 h. Only PC-encapsulated test vehicles (cured and dehydrated) passed the HTRB test from the beginning. For chips packaged with the CE material, a special postcuring treatment was successful for passing the final HTRB test. First results of cured CE-encapsulated IGBTs also failed the HTRB test after several hours. This failure could not be related to pH values of the encapsulation or any critical interactions with contact surfaces of the IGBT test vehicle (see Fig. 6).

Cured CE, as stated in the material section, hardens through the formation of amorphous hydrates after curing. Applying an electronic field to the encapsulant might have mobilized bound water, which could have led to conductivity pathways through the material. Therefore, the dehydration of CE finally enabled the success of passing the HTRB test. Furthermore, the detected impurities of CE did not affect the HTRB performance negatively.

Fig. 7 exemplarily illustrates the leakage current, measured via a curve tracer test, of a CE- and PC-encapsulated IGBT (dashed black) compared with a CCC-encapsulated test vehicle (red). The CE and PC leakage current stays at a level of about 2 μA for the whole voltage sweep. In addition, the applied electrical field stays constant. It has to be pointed out that the leakage current of CE- and PC-encapsulated IGBT test vehicles equals the nonencapsulated leakage current of the IGBT.

By contrast, it is significant how the measured current rises for the test vehicles with CCC encapsulation at a voltage level of about 250 V up to a value of 190 μA. This even represents the blocking current of the used curve tracer. Therefore, the effects mentioned previously (critical pH value) can be the reason for a failure with CCC. The chip has lost its capability to block any voltage, thus is not functional any more.

Finally, the investigations regarding HTRB with CE (and PC) encapsulants show a very high potential of improving the thermal stability and reliability of power electronic devices.

HTRB results finally prove that a volume resistivity of 10⁷ Ω·cm and a dielectric loss of .15 at 300 kHz (see Table I) are sufficient for passing the test. However, dielectric loss and resistivity is significantly higher, compared with the silicone gel data.

In terms of general specifications, final conclusions on the minimal volume resistivity and dielectric loss are not possible because of further required environmental testing of ceramic encapsulants.

Initial active power cycling experiments have been performed to investigate the potential of new PC encapsulants for thermomechanical stabilization of the complete stack of a power module based on a frame module with soldered dice, Al bond wires, DCB substrate, and Cu base plate. In contrast to state-of-the-art silicon materials, the novel developed PC materials are capable of prolongation of module lifetime by a factor of 3.5. Thorough analysis of electrically stressed devices revealed a change in failure mechanism. As already mentioned in Introduction, the well-known bond wire liftoff, typically observed with silicone gel potted devices, did not occur when the encapsulant was a PC material.

By contrast, degradation of the solder layers (die attach and base plate fixation) was observed after a significant longer lifetime. Surprisingly, a further degradation on the chip surfaces and bonding wires was observed, which did not occur in the absence of electrical stress on dice within the same module [10]. These degradation pathways were subject of investigations on ionic migration pathways of PC-encapsulated samples, applying an electric field within a climate chamber to facilitate ionic mobility.

As described for CE- and CCC-encapsulated test vehicles, also PC glob-tops were potted on test vehicles comprising power diodes (1,700 V) soldered on DCB substrates and the topside connected with 400 μm Al bonding wires (Fig. 8).

For H³TRB tests, such devices were electrically contacted on surface isolation resistance (SIR) testing boards and exposed to an atmosphere of 85°C and 85% relative humidity and stressed with a reverse bias of 80 V (according to [14] and [17]). All vehicle resistances were monitored over the total testing time to identify fails after a resistance decrease below 10⁶ Ω.

The first formulations under test failed directly after 10 h of exposure to electrical fields. These short testing times indicate quick penetration of the PC bulk with moist atmosphere and produce pore solutions under condensation conditions directly at the chip surface.

Such effects are well known for silicon encapsulants; however, failures are prevented by low ionic contents within the formulations. By contrast, the matrix of all cements is a complete ionic composition. Accordingly, the resulting pore solutions were investigated by monitoring the electric conductivity of aqueous suspensions of PC in deionized water (Fig. 9).

Conductivity of such aqueous solutions increases over time and allow preselecting PC formulations on their potential for stability against moisture. This property agrees well with decreasing solubility products of the respective PC components. Thoroughly, phase analysis allows for binder and additive choice that results in encapsulation materials with a reduced potential for increased ion conductivity in solution. These PC encapsulants show high stability under humid atmosphere, also under electrical stress, and thus allow for passing 1,200 h without failure in the H³TRB test (Fig. 10).

The described results of PC encapsulants are promising in terms of the behavior of CE-encapsulated devices under humid conditions and future H³TRB, respectively. PC and CE encapsulants are both nonpolymeric, nonmetallic, and brittle materials. Because of their chemical composition, also PC belongs to the general class of ceramic materials. Therefore, future research on CE has to focus on humidity tests, such as H³TRB, and the influence of CE solubility and corresponding ion migration effects.

In addition, further environmental investigations on CE and PC materials regarding the SIR [24] will be conducted to finally qualify these novel encapsulation materials for their first applications on power electronic devices.

The presented article introduces a novel CE material, optimized on the basis of CCC encapsulants. The CE material shows enhanced thermal conductivities and simple processing compared with common filled epoxy resin or silicone gel. Furthermore, the current advances on reliability reveals the potential of CE-encapsulated power electronic systems, equipped with wide bandgap semiconductors, operating at high temperatures. The main results are as follows.

1. Changing the matrix of CE from CAC to reactive alumina significantly reduces the pH (down to 8.5) and the permeability. Latter is correlated with the composite pore size reduction (<100 nm) and the change from crystalline matrix phases to almost amorphous aluminum hydroxides.

2. Thermal behavior of CE seems to be promising, in contrast to CCC at higher temperatures up to 300°C.

3. The reduced pH of CE avoided critical interactions with the chip surface (AlSiCu).

4. HTRB tests with CE encapsulation were passed. Former interactions of CCC with the chip metallization represents the corresponding failure mechanism. However, dehydration of CE is necessary to pass the tests. Otherwise, mobilized bound water could cause ion migration effects. A volume resistivity of 10⁷ Ω·cm is not critical for passing the HTRB test.

5. H³TRB tests of PC-encapsulated power diodes were passed. The improvements to former failures are correlated with enhanced insolubility of all phases under humid condition. Therefore, ion migration effects can be avoided. Power cycling with the PC encapsulation already revealed 3.5 times better lifetime compared with silicone gel. Both H³TRB and power cycling results are promising for future developments on the presented CE materials.

Ongoing research with improved CE materials is focused on different environmental testing, including thermal shock tests, SIR tests, and H³TRB tests. Furthermore, power cycling of H³TRB-passed PC-encapsulated devices will be tested. After preventing a failure mechanism during the H³TRB test by enhancing the insolubility of PC, further increases in lifetime (beyond 3.5 times of silicone gel) can be assumed. Accordingly, CE materials have to be evaluated by power cycling tests, as they show PC-comparable material properties and also higher thermal conductivities (4.6 W/[m·K]). Thus, a further lifetime enhancement could be possible.

The authors specially thank Prof. Ronald Eisele for the impulse of developing novel ceramic and cement-based encapsulation materials for future power electronic systems. The presented research on ceramic encapsulation materials for electronic packaging received financial support by the Federal Ministry of Education and Research BMBF (Contract: 16EMO0226).