Design of Cu/Zr Alloy Interface for Enhanced Thermal Fatigue Performance in Electronic Packaging

Abstract

This study addresses a critical limitation in direct bonded copper (DBC) materials used in power electronics by introducing a copper–zirconium (Cu/Zr) alloy interposing layer at the copper-ceramic interface. This novel design aims to mitigate mechanical stress induced by mismatched material properties, such as the coefficient of thermal expansion (CTE) and elastic modulus, during thermal cycling. The key findings of this study are (1) thermal fatigue improvement: Test samples with the Cu/Zr interface layer (Cu–Cu/Zr–AlN) three times enhanced thermal fatigue resistance, surviving 30 thermal cycles from −55 to 300 °C before delamination, while standard DBC substrates without the Cu/Zr layer failed after just 10 cycles, indicating a performance improvement with the Cu/Zr alloy, (2) durability projections: Based on the Coffin–Manson model, if the upper temperature is capped at 150 °C, the Cu–Cu/Zr–AlN substrates are projected to survive approximately 1372 cycles, underscoring their potential for long-term reliability, and (3) stress mitigation: The Cu/Zr alloy layer bridges the CTE disparity between copper and ceramic, reducing mechanical stress and improving structural integrity across a broad temperature range (−55 to 300 °C). This study reveals, incorporating the Cu/Zr interposing layer between Cu and AlN substrates significantly enhances the operational lifespan and reliability, making them well-suited for high-temperature and high-stress environments in advanced power electronics. This innovation demonstrates the feasibility of improving thermal fatigue performance while maintaining mechanical integrity, representing a substantial advancement over conventional DBC materials.

1. Introduction

Power electronics modules, which integrate a ceramic substrate with copper sheets, are extensively used in industrial applications such as renewable energy systems, hybrid-electric vehicles, locomotives, and space missions. , These modules experience various stressesthermal, chemical, electrical, and mechanicalwhich cumulatively lead to degradation and eventual failure. Figure A shows a typical schematic diagram of a power electronic module system. Understanding the failure mechanisms of these components is critical for improving reliability and extending their lifetimes. ,− In addition, vibration or shock can accelerate the degradation process. A primary cause of failure in power electronics modules upon thermal cycling is the mismatch in the coefficient of thermal expansion (CTE) between dissimilar materials. This mismatch induces stress buildup at the interfaces of the electronic chip (die), the metallized substrate, and the packaging board for example heat sink or heat spreader. Electronic devices that can function reliably at elevated temperatures exceeding 125 °C are critical for high-performance and harsh-environment applications. These devices must withstand extreme thermal stress in areas such as aerospace, space exploration, geothermal energy, and oil/gas exploration.

1.

(A) A schematic diagram of a thin film sample cross section with a direct bonded copper (DBC) circuit board, copper sheets on both sides of the ceramic in a typical application, where a base plate or heat spreader is soldered to one side and an electronics device is soldered to the other. (B) A Cu/Zr/Ti alloy deposited on a clean 630 μm AlN substrate. A 5 nm Ti and 5 nm Cr layers were deposited for adhesion. Nominal thickness of deposited alloy is approximately 250 nm. An approximate 300 μm thick Cu top sheet was electroplated. Note: A 50 nm Cu seed layer for electroplating is not shown.

Previous research has categorized the failure mechanisms of power electronic devices into two distinct types: (1) chip-related (intrinsic) failures and (2) package-related (extrinsic) failures. This classification is essential for understanding and addressing the challenges these devices face. ,, While both failure types are important to investigate, this study focuses specifically on the package-related failures. To address the package-related failures, direct-bonded copper (DBC) AlN substrates are effective due to their excellent thermal conductivity. These substrates outperform conventional ceramics like Al₂O₃ in heat dissipation for high-power module applications. However, they are prone to cracking, particularly at the edges of the copper film during thermal cycling, due to the CTE mismatch between AlN and copper. Some studies have reported the fabrication of AlN-DBC ceramic substrates by reducing CuO to improve the interface microstructure of the bonding zone. Analyzing the failure modes at the Cu ceramic interface is crucial for enhancing reliability and performance. Additionally, investigating interfacial thermal transport during thermal shock is essential for effective thermal management in aerospace devices. Accordingly, this study addresses the package-related failures by focusing on the copper-ceramic (AlN) interface.

To mitigate failure mechanisms at the Cu/AlN interface, research has been dedicated to enhancing the conductor material through additives or alloying strategies. In this regard, Cu/Zr-based amorphous alloys are notable for their exceptional glass-forming ability, , high thermal stability, and desirable mechanical properties, such as high strength and ductility. , These characteristics make Cu/Zr-based alloys ideal for replacing the traditional DBC interface between ceramic and copper sheets. In addition, a previous modeling study suggests that Cu/Zr bulk amorphous alloys can serve as a conductive and durable interface material, effectively reducing the CTE mismatch in power electronics packaging.

Beyond addressing the CTE mismatch, it is also important to shield electronic devices from extreme temperatures and oxidation. This study explores innovative strategies to enhance oxidation resistance and thermal stability in power electronics by employing an interposing Cu/Zr thin film in place of the conventional DBC material configuration. The proposed approach involves integrating copper conductors with a titanium adhesion layer on AlN substrates, forming a Cu–Cu/Zr–AlN trilayer architecture. The primary objectives of this study are to (1) synthesize a Cu–Cu/Zr–AlN trilayer thin film that integrates copper conductors with a titanium adhesion interface layer on AlN substrates; (2) evaluate the thermal cycling performance of the Cu–Cu/Zr–AlN system and compare its durability and resistance with the conventional Cu–AlN system; and (3) predict survival cycles using the Coffin–Manson model to estimate the operational lifespan and thermal fatigue resistance of the Cu/Zr thin film. This approach is expected to significantly enhance the durability and reliability of power electronic applications in challenging environments. By improving oxidation resistance and better performance under thermal stress, this study contributes to extending the operational lifespan of power electronics modules.

2. Materials and Methods

2.1. Materials

Copper sulfate (CuSO₄), hydrochloric acid (H₂SO₄), and copper chloride (CuCl₂) were purchased from Fisher Scientific, USA. All glassware, coverslips, and silicon wafers were washed with Millipore water and acetone and dried well before use. The four different metal targets e.g. Zr target (99.99% purity, 2 in. diameter × 0.250 in. thickness) and Cu target (99.99% purity, 2 in. diameter × 0.250 in. thickness), Ti target (2 in. diameter × 0.250 in. thickness), and Si target (2 in. diameter × 0.250 in. thickness) were obtained from Kurt J. Lesker, USA. Direct bonded copper (DBC) is obtained from Stellar Industries Corp. (MA, USA).

2.2. Synthesis of Cu–Cu/Zr–AlN Ceramics

A multisource magnetron sputtering system was used to fabricate the engineered Cu–Cu/Zr–AlN thin film, in which the Cu/Zr layer is approximately 250 nm thick. Subsequently, a thick copper sheet was electroplated on top, as depicted in Figure B. AlN ceramic substrates, with a thickness of 630 μm and a thermal conductivity of 170 W/m K, were purchased from Stellar Industries Corp (MA, USA). The AlN substrates (34 mm by 34 mm) were cleaned in a March Plasmod plasma cleaner for approximately 1 min in oxygen plasma (100 W, 0.7 Torr). The substrates were not heated but nominally maintained at room temperature during the magnetron sputtering deposition process. Since adhesion of the engineered interface to the ceramic surface is essential for optimal strength, a 5 nm titanium layer was utilized to ensure adhesion at the ceramic–alloy interface. Among various metallic interlayers explored to enhance bonding at ceramic-metal interfaces, Jin et al. reported that Ti and Cr interlayers introduced the fewest interfacial defects and yielded the highest bonding strength. An approximately 250 nm thick Cu/Zr (Ti) alloy was deposited by magnetron cosputtering with each individual source set to the appropriate power to achieve the desired alloy concentrations (refer to Supporting Information).

A 50 nm pure Cu seed layer was deposited to initiate electroplating, following the deposition of a 5 nm chromium adhesion layer onto the oxidized top surface. Both depositions were performed using a Denton deposition system with e-beam sources (refer to Supporting Information). The copper sheet was deposited by electroplating in an acidic copper sulfate electrolyte (refer to Supporting Information). The final engineered material structure is depicted in Figure B.

2.3. Characterizations. Thermal Mechanical Analyzer (TMA)

A TA Instruments TMA Q400 with an MCA 70 chiller provided repetitive thermal cycling between −55 and 300 °C at a ramp rate of 20 °C/min. A 3 min hold time at the upper temperature and a 10 min hold at the lower temperature ensured that the chamber and sample stabilized at the maximum and minimum temperatures, respectively. The chamber size limits to one sample at a time (sample size approximately 10 mm × 12 mm × 1 mm). A standard expansion probe monitored changes in sample thickness throughout the test. The accuracy of the TMA dimensional change measurement was less than 0.1 μm, whereas the CTE-induced thickness change of the specimen was expected to be 3 μm over the 355 °C temperature range. Height changes due to asymmetric specimen expansion, which caused curvature changes, was 16 μm over a 100 °C temperature cycle. Thermal cycling fatigue failure, indicated by sample fracture, was detected by a significant change in sample thickness (∼100 μm) and was easily identified in the sample thickness versus time plot.

2.4. Scanning Transmission Electron Microscopy (STEM)

Imaging and energy dispersive spectroscopy (EDS) elemental analysis were utilized to verify and characterize the nanostructure of thin alloy films. High-resolution STEM characterization and EDS mapping were performed with an FEI Talos operated at 200 kV. Cross sectional TEM specimens were prepared and extracted using an FEI NOVA focused ion beam (FIB) system. A lamella specimen, 20–30 nm thick and approximately 15 μm by 10 μm in area, was extracted from the TEM specimen using an Omniprobe micromanipulator, and then mounted for investigation. In addition, optical images of the fracture surfaces were obtained with a digital microscope (model UHM350-11, AmScope).

3. Results and Discussion

3.1. Cu–Cu/Zr–AlN Samples STEM-EDS Investigations

The STEM elemental mapping was performed on a cross section of the Cu–Cu/Zr–AlN film sample with 50% Cu and 50% Zr to analyze its composition. The STEM image in Figure shows distinct Cu, Cr, Cu/Zr, Ti, and Al layers. Although nitrogen (N) was present with Al, it was not explicitly reported. The thickness of Cu/Zr layer is approximately 250 nm. Fourier transform analysis indicates that the Cu/Zr alloy remains amorphous (confirmed by TEM electron diffraction analysis-Figure inset) or glassy even after 35 thermal cycles between −55 and 300 °C. Amorphous alloys, also called metallic glasses, have attracted extensive attention due to the structural and compositional homogeneity as compared to crystalline phase. However, the electrolytically deposited Cu sheet is polycrystalline. TEM electron diffraction image (inset of Figure ) confirmed the presence of crystalline form of Cu and AlN and amorphous phase of Cu/Zr. The amorphous phase of the Cu/Zr alloy is evident from the diffuse torus in the diffraction pattern whereas the copper and AlN have strong crystalline diffraction patterns without any diffuse background.

2.

STEM Image of the Cu–Cu/Zr–AlN thin film, with Cu shown in red, Al in violet, Ti in yellow, and Cr in orange. The greenish color represents Cu/Zr alloy. Insets display the electron diffraction patterns of the thin film in the Cu, Cu/Zr, and AlN layers.

The STEM-EDS line scans of the cross section, along the growth direction of the pulsed DC magnetron sputtering (PDCMS)-deposited Cu/Zr thin film, revealed the nanoscale structure of the alloy film. Figure shows the EDS elemental line scans for three (Cu/Zr/Ti) compositions: (50/50/0), (50/40/10), and (35/55/10), where the numbers in parentheses are the percentage of each component. For all the samples, a 5 nm pure Ti adhesion layer is observed at the ceramic–alloy interface, which remains confined at the interface without migrating into the alloy or ceramic. The thicknesses of the Cu/Zr/Ti alloy region were 260, 230, and 260 nm for the (50/50/00) (Figure A), (50/40/10) (Figure B), and (35/55/10) (Figure C) samples compositions, respectively. The compositional distribution of Cu and Zr is uniform throughout the thickness, except at the top surface. In all three samples, the top surface of the alloy features a thin 20 nm Zr-enriched layer followed by a 20 nm Cu-enriched layer underneath. The EDS line scan also indicates that oxygen is present in the top Zr-enriched layer but absent in the Cu-enriched region. This selective oxidation of Zr over Cu occurs due to its higher oxygen affinity, forming a lower stoichiometry oxide compared to ZrO₂. Similar oxidized surface structures have been reported for Zr–Cu alloys. , The underlying Cu-enriched layer suggests that Zr atoms diffuse to the surface and react with ambient oxygen. As this oxide layer grows, it effectively slows further oxidation by limiting oxygen penetration and thus passivates the metal surface against further oxidation.

3.

EDS line scans from STEM for Cu(%)/Zr(%)/Ti(%). All three samples were analyzed after the Cu/Zr/Ti alloy was deposited on AlN but before the top copper sheet was electroplated: (A) (50/50/0), (B) (50/40/10), and (C) (35/55/10). The line scans represent the atomic concentration percentage, with the AlN substrate on the left and the alloy deposition progressing from left to right.

The EDS line scans also provided elemental concentration profiles through the alloy thickness for each composition: 49% Cu and 45% Zr for (50/50/0) sample; 48% Cu, 38% Zr, and 7% Ti for (50/40/10) sample; and 37% Cu, 50% Zr, and 6% Ti for (35/55/10) sample. All three samples contain minor impurities (N, Al, and O) at levels below 4%. Considering this small impurity background, the measured alloy compositions align with the target concentrations within a ±4% uncertainty. A similar elemental concentration pattern was observed in the AlN ceramic substrate. A slight enhancement of Al and a decrease in N were detected within 50 nm of the substrate surface, likely due to environmental exposure after manufacturing and polishing-induced nitrogen loss near the surface. A minor oxygen peak near the surface aligns with the 5 nm Ti adhesion layer, which acts as an oxygen getter during the TEM sample preparation and transfer.

3.2. Thermo-Mechanical Analyzer (TMA) Investigations

Samples with the Cu/Zr alloy interface were subjected to thermal cycling in the TMA until fracture failure occurred. Figure shows the average number of cycles to failure for each alloy, based on two to four samples per type: 24, 30, and 7 cycles for (35/55/10), (50/50/0), and (50/40/10), respectively, compared to 8 cycles for the baseline DBC. The (35/55/10) and (50/50/0) alloys exhibited 2 and 3 times, the fatigue resistance of the baseline DBC, respectively. However, the (50/40/10) composition showed limited fatigue life, averaging only seven cycles, which was comparable to the baseline DBC. These results indicate that the Cu/Zr alloys with 50% or greater Zr content have improved thermal fatigue resistance.

4.

Data points are the mean number of thermal fatigue cycles (−55 to 300 °C) before the onset of fracture failure versus thin film interface alloy type and the DBC baseline, where nomenclature is atomic percent: (Cu (%)/Zr (%)/Ti (%)). The horizontal lines are minimum and maximum number of cycles survived for different samples (total 13 samples) of the same type. Statistics are not shown due to limited number of specimens.

To further understand the thermal cycling failure process, optical microscopy was used to examine the fracture planes, which were parallel to the ceramic-metal interface and exhibited brittle characteristics. Optical images of the fracture planes were similar for all three sample types (Figure ). In some regions of the (50/50/0) sample, fractures propagated slightly away from the interface, leaving a thin ceramic layer attached to the metal surface. Higher-resolution imaging of the metal side of the fracture revealed microscopic AlN ceramic particles adhered to the entire metal surface, confirming that the fracture predominantly occurred within the ceramic rather than the metal. These results are consistent with both experimental and finite element modeling results reported for thermal fatigue failure of commercial DBCs. , The finite element simulations predicted that the fracture initiates at the sample corners due to stress concentration at the free edge corner caused by the CTE mismatch between the metal and ceramic layers and then propagates toward the center. This pattern is consistent with the observed fracture contours (Figure ).

5.

Optical images of fractured thin film samples (sample width: 1 cm) with varying Cu (%)/Zr (%)/Ti (%) compositions. Images (A,C,E) show the ceramic sheet side, while (B,D,F) display the copper sheet side. A mirror symmetry in surface structure is apparent. Samples 35/55/10 and 50/40/10 show successive fracturing, originating in approximately the corners and propagating to the approximate center. Note: (A,C,E) are optical images of the ceramic side, while (B,D,F) are those of the metallic side. (A,B): (50/50/0), (C,D): (60/40/10), (E,F): (35/55/10).

The improved thermal fatigue resistance of the (35/55/10) and (50/50/0) alloys can be attributed to stress redistribution at the interface. By incorporating an interposing Cu/Zr alloy with an intermediate CTE, the stress is distributed across two interfaces rather than being concentrated at a single boundary. This redistribution reduces tensile and compressive stresses within both the ceramic and metal layers, thereby delaying the onset of fatigue failure. Moreover, our calculations using molecular dynamics simulation and previous study show that the CTE of Cu/Zr alloy and the density dependencies on composition are highly correlated and even proportional, as illustrated in Supplementary Figure S1. Computed CTE values indicate that the addition of zirconium to copper decreases the CTE, enhancing compatibility with AlN substrates. This decrease alleviates thermal loads during cycling, as demonstrated by the improved thermal fatigue resistance of Cu–Cu/Zr–AlN substrates.

The samples were fabricated near room temperature, resulting in minimal residual stress under room temperature conditions. However, as the temperature deviates from room temperature, residual stress accumulates at the ceramic-metal interface due to the CTE-induced differential expansion between the ceramic and metal, which is constrained at the interface. As a result, the material with the higher CTE undergoes compression, while the material with the lower CTE experiences tension parallel to the interface. This combined compression and tension also induce curvature. As the temperature changes further from equilibrium, the curvature increases until fracture occurs at the corresponding maximum or minimum temperature, reaching the ultimate breaking stress. Introducing an interposing alloy layer at the interface with a CTE between that of the ceramic and copper helps distribute strain and stress across two interfaces instead of localizing them at a single interface. This transition smooths out tensile and compressive stresses in both the ceramic and copper layers. As a result, the material can withstand a broader temperature range before reaching the critical stress level that leads to thermal fatigue failure.

Notably, the (50/40/10) alloy fractured after only seven cycles on average between −55 and 300 °C. This sample had the lowest Zr concentration (40%) compared to the other two alloys (50% and 55%). As listed in Table , the CTE of Zr (5.9 ppm/K) is much closer to that of AlN (4.7 ppm/K) than that of Cu (16.6 ppm/K). Since fracture occurred within the AlN at the interface, having a CTE closer to that of AlN is more effective in reducing interfacial residual stress and enhancing thermal fatigue resistance than having a CTE closer to that of Cu. Therefore, the lower Zr concentration in the (50/40/10) alloy resulted in the shortest thermal fatigue life compared to the (50/50/0) and (35/55/10) alloys.

1. Thermal Conductivity, Flexural Strength, and Coefficient of Thermal Expansion (CTE) Data for Ceramics Commonly Used in Power Electronic Circuit Boards ,,− ^,

material	thermal conductivity (W/m K)	strength (MPa)	CTE (×10–6 K–1)
ceramic		flexural Strength
Al2O3	24	322	6.8
Si3N4	90	604	2.5
AlN	170	350	4.7
metal		tensile strength
Cu	400	235	16.6
Zr	23	275	5.9
Ti	22	316	9.2

^aAdditionally, thermal conductivity, tensile strength, and CTE data for the metals utilized in this study are provided.

Another factor contributing to the enhanced thermal fatigue resistance is the increased fracture toughness at the metal/alloy/ceramic interfaces provided by the intermediate Cu/Zr layer, compared to the metal/ceramic interfaces of the baseline DBC. Although not reported in this paper, finite element simulations indicate that the reduction in interlaminar tensile and shear stresses due to the smoother CTE transition is not significantly pronounced, given the thinness of the interface layers (250 nm) relative to the bulk Cu (300 μm) and AlN (630 μm). Therefore, the improvement in fracture toughness may play a more significant role in the observed test results. Efforts to directly quantify the fracture toughness of both the baseline and modified interfaces are currently underway through experimental and computational studies using a notched beam under four-point bending.

3.3. Accelerated Testing Prediction

Accelerated thermal cycling tests are commonly used in electronics reliability studies, to predict fatigue failure in significantly less time. Tests are often conducted at increased thermal stress via higher cycle frequency, larger temperature range, and elevated maximum temperature. An acceleration factor (AF) derived from parameters fitted to an appropriate model allows for extrapolation to the number of cycles at lower thermal stress levels. The Coffin–Manson (C–M) model, widely used for low-cycle fatigue analysis in solder joints, − was applied to predict the fatigue life of the Cu/Zr alloy interface. This model applies to solder used as a die attach material situated between a brittle crystalline semiconductor and a metal contact or wire. When applying the C–M model for failure analysis, the thin engineered interface alloy in this study is considered analogous to a thin solder or die attach material.

The model accounts for thermal stress dependency on three factors: temperature cycle frequency, temperature range, and maximum temperature in Kelvin, allowing for the prediction of the number of cycles to failure (N _i ). The C-M model’s power-law terms describe the relationship with cycle frequency (f _i ) and temperature range (ΔT _i ), while the dependence on the maximum temperature is captured through a thermal activation term ( $\exp \left(\frac{E_{\mathrm{A}}}{k{T}_{i,\max }}\right)$ ), as shown in eq .

$$N_i=A{f_i}^{-a}{{\mathrm{\Delta }T}_i}^{-p}\exp \left(\frac{E_{\mathrm{A}}}{k{T}_{i,\max }}\right)$$ 1

where a and p are exponents for f and ΔT, A is a constant, E _A is the activation energy, and k is the Boltzmann constant. The subscript (i) is either (L) or (H) for low- and high-stress conditions, respectively. The AF is then determined as the ratio of the number of cycles to failure under low-stress conditions (N _L) to those under high-stress conditions (N _H), as given by eq

$$\mathrm{A}\mathrm{F}=\frac{N_{\mathrm{L}}}{N_{\mathrm{H}}}={\left(\frac{f_{\mathrm{L}}}{f_{\mathrm{H}}}\right)}^{-a}{\left(\frac{{\mathrm{\Delta }T}_{\mathrm{L}}}{{\mathrm{\Delta }T}_{\mathrm{H}}}\right)}^{-p}\exp \left[\frac{E_{\mathrm{A}}}{k}(\frac{1}{T_{\mathrm{L},\mathrm{m}\mathrm{a}\mathrm{x}}}-\frac{1}{T_{\mathrm{H},\mathrm{m}\mathrm{a}\mathrm{x}}})\right]$$ 2

Literature values suggest exponents a ≈ 1/3 and p ≈ 2 for solder failures, , which were used in our calculations. Applying eq to experimental data from two baseline DBC fatigue tests: i.e., 8.3 average cycles to failure at −55 to 300 °C, and 83 cycles at −55 to 200 °C (Table ) yielded E _A = 0.38 eV, which falls within the reported range of 0.12–0.58 eV for the Coffin–Manson model. −

2. Coffin–Manson Model Fitted to the DBC Baseline and Cu/Zr/Ti Samples .

		number of cycles to failure
		DBC baseline	Cu (%)/Zr (%)/Ti (%)
temperature range	acceleration factor (AF)		50/50/00	35/55/10	50/40/10
–55 to 300 °C	1	8.3#	30.5#	23.7#	6.8#
–55 to 200 °C	10	83#	305*	237*	68*
–55 to 150 °C	45	374*	1372*	1066*	306*

^aAcceleration factors (AF) are calculated for the DBC baseline using eq and are used to predict the number of cycles to failure for Cu/Zr/Ti samples at different temperature ranges. Note: # denotes experimental values, and * denotes predicted values.

Since all samples (baseline DBC and current engineered interface materials) exhibited a similar failure mechanism with fractures occurring in the AlN ceramic, we confidently applied the DBC-fitted AF to predict the fatigue failure of three-layer engineered samples at lower T _max. The calculated results (Table ) indicate that the best performing (50/50/0) alloy is predicted to survive over 1372 cycles for a temperature range of −55 to 150 °C. The (35/55/10) sample is predicted to survive 1066 cycles under the same conditions. Conducting a thermal cycle test for 1400 cycles would require 33 days, which demonstrates the advantage of accelerated testing. For applications where the maximum temperature is limited to ∼150 °C, the predicted 1400-cycle survival suggests that this engineered Cu/Zr interface can extend thermal fatigue life beyond 1000 cycles, making it promising for applications requiring high thermal conductivity of AlN and thermal fatigue durability. However, for a temperature range of −55 to 200 °C, the predicted number of cycles to failure is only in the hundreds, which may not be sufficient for most applications. Thus, further optimization is needed to enhance the interface configuration, especially for applications operating above 150 °C. To our literature search, no direct comparable studies could be found in the literature. However, some investigations on active metal brazing (AMB) substrates, such as thermal cycling of Si₃N₄ from −30 to 150 °C have reported failure predictions of up to 6685 cycles. These results show greater cycle endurance than observed in our study. While it is noteworthy to point out the higher flexural strength of Si₃N₄ (604 MPa) compared to AlN (350 MPa) (Supporting Information: Table S1), our results indicate higher thermal fatigue is probably due to increased fracture toughness. Our approach, however, retains AlN for its superior thermal conductivity relative to Si₃N₄, while incorporating the interposed Cu/Zr alloy interface to improve thermal fatigue resistance. ,

4. Conclusions

In this work, we have engineered an improved material interface to reduce thermal stress during temperature cycling of ceramic-copper bilayer sheets. The interface incorporates a Cu/Zr metal alloy to bridge the CTE mismatch and enhance interfacial adhesion between AlN and Cu. The unique properties of Cu/Zr alloys provide an opportunity to leverage the low CTE of Zr with the mechanical toughening of a bulk metallic glass. With the Cu/Zr thin film at the interface, resistance to thermal cycling fatigue fracture over the temperature range of −55 to 300 °C improved by a factor of 3 compared to the baseline DBC. The nanostructure of the Cu/Zr alloy was investigated with STEM/EDS, while the structure and chemical composition of the fracture surfaces were analyzed with optical microscopy. For the Cu/Zr ratios examined, fracture occurring within the ceramic suggests that further optimization of material morphology is needed to enhance the fracture toughness of the hybrid interface and improve thermal fatigue life. Application of the Coffin–Manson accelerated testing model shows promise for increasing thermal fatigue resistance beyond 1000 cycles within the temperature range of −55 to 150 °C.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsomega.5c06959.

• Synthesis of ceramic-Cu/Zr–Cu thin film; deposition of copper seed layer; deposition of thick copper sheet; CTE calculation using molecular dynamics simulation. (PDF)

The manuscript was written through the contributions of all authors. All authors have given approval to the final version of the manuscript.

The authors declare no competing financial interest.