RF/Microwave Power Package Reliability: The Effects of Component Materials and Assembly Processes

The mechanical performance and reliability of various Al2 O3 ceramics were evaluated with intended applications as insulator frames for RF/microwave power packages. Considering variations in base-flange material properties (CuW, CuMoCu), base-flange thickness (0.040", 0.020"), assembly process material (AgCu, Cu), assembly process temperature (860°C, 1080°C) and Al2 O3 insulator material (96, 99, and 20 percent ZrO2 /Al2 O3 , zirconia-toughened Al2 O3 (ZTA)), finite element analysis (FEA) spatially resolved the fabrication-induced stresses during the assembly of CuW/Al2O3 and CuMoCu/Al2 O3 structures and showed the critical regions to be the frame corners at the metal base-flange/Al2 O3 interface. Bend strengths

(four-point) and Weibull distributions were determined for each Al2 O3 material and, coupled with the FEA-predicted stresses for the various package configurations and assembly processes, the failure probabilities (Pf) for the various CuW/Al2 O3 and CuMoCu/Al2 O3 structures were calculated. The strength and reliability data were considered in conjunction with relative cost for the Al2 O3 ceramics to select an optimum frame insulator for the application.

Philip M. Fabis
St. Gobain Industrial Ceramics, Norton Diamond Film
Northboro, MA

In addition to the electrically functional device (Si, GaAs), RF/microwave packages utilize a wide spectrum of material types ranging from traditional ceramics (Al2 O3 , BeO, AlN) and metals/alloys (Cu, Au, Ag, Al, CuAg, MoMn, Kovar) to specialized metal-matrix composites (AlSiC, CuW), clads

(CuMoCu, MoNiMo), polymers (polytetrafluoroethylene, polyimide), ceramic-filled polymers and glass-cloth laminates. Reliable joining of these dissimilar materials at device and system levels presents challenges in tailoring assembly processes and material properties to produce functional and reliable packages. Ultimately, the materials compatibility engineering of the macro-composite structures requires collective satisfaction of electrical, thermal, mechanical and economic specifications with minimal design compromise.

The ceramic component of these packages can be a significant contributor to enhanced package performance. Depending on design, the ceramic may serve, individually or collectively, as an electrical and environmental isolator, heat spreader/sink and dielectric medium. However, brittle mechanical characteristics and distributed fracture strengths typically require an assessment of their reliability through probabilistic methods based on Weibull statistics — weakest-link theory.1,2 Although Al2 O3 has been extensively used as an electronic substrate in monolithic plate morphology, little has been reported on its compatibility with new base-flange materials (clads, composites) or as a shaped structure (frame). Since this shape variable of a material can be related to its stiffness and strength through section moments and shape factors,3 such consideration is important when assessing the performance and reliability of the material in an application environment. This article discusses the combination of predictive finite element modeling and empirical analysis of package component materials and assembly processes as a performance and reliability assessment of alumina ceramic insulator frames in RF/microwave power packages.

Experimental Analysis

A typical RF/microwave power package schematic with labeled components is shown in Figure 1 . The component materials with pertinent properties as well as experimental package configurations and assembly process conditions are listed in Tables 1 and 2 , respectively. Mechanical property measurements for the 96 percent Al2 O3 , 99 percent Al2 O3 and ZTA (20 wt. percent ZrO2) insulator frame varieties were performed on 30 specimen populations at room temperature in a four-point bend geometry flexure test to obtain the flexure strength statistics and Weibull modulus. Fracture surface morphologies were examined using scanning electron microscopy (SEM) to determine failure-mode characteristics.

Table I
Package Component Material Properties

Material

K (W/mK)

u (x 10-6 /°C)

E (GPa)

u

Ys(MPa)

r (kg/m3 )

Cp (J/kg°C)

CuMoCu (131)

218

6.5

263

-

518

9660

126

CuW (15/85)

248

7.0

390

0.30

400

17,300

223

Cu

393

17.0

129

0.34

100

8900

385

AgCu (85/15)

300

19.6

83

0.36

260

9900

279

Al2 O3 *

21

6.6

372

0.25

-

3750

880

*<8% variation of thermomechanical properties for all Al2 O3 materials considered.
k=thermal conductivity; a=thermal expansion coefficient;E=Young's Modulus;r=density,u=Poisson's Ratio;Ys=yield stress;Cp=heat capacity-

Utilizing these material and process characteristics, FEA models were generated using ANSYS software, assuming elastic-plastic materials behavior, to predict the magnitude and spatial location of fabrication-induced stresses as well as Pf for three varieties of Al2 O3 insulator frame components. The mesh resolution was optimized to provide adequate computational resolution in reasonable times; element volumes in the observed fracture sites were on the order of 10–13 m3 and increased to 10–10 m3 when removed from these sites. The thermomechanical properties of the selected Al2 O3 materials were very similar over the assembly process temperature range and, therefore, were not considered as variable parameters in the model.

Table II
Experimental Package Structures, Dimensions and Assembly Processes

Package Assembly

Base-flange Material

Base-flange Thickness (")

        Assembly Materials/               Process Temp (°C)      

1

CuMoCu (131)

0.040

DBC/1080

2

CuMoCu (131)

0.040

AgCu/860

3

CuMoCu (131)

0.020

DBC/1080

4

CuMoCu (131)

0.020

AgCu/860

5

CuW (15/85)

0.040

DBC/1080

6

CuW (15/85)

0.040

AgCu/860

7

CuW (15/85)

0.020

DBC/1080

8

CuW (15/85)

0.020

AgCu/860

Based on a best-/worst-case ranking of the FEA results for stress magnitude and failure probability, experimental package structures (CuMoCu/Al2 O3 and CuW/Al2 O3 ) were fabricated in conventional belt furnaces. The AgCu processing was performed at 860°C in flowing H2N2 forming gas and the DBC process was performed at 1080°C in flowing dry N2. Both processes followed standard manufacturing time/temperature profiles. Nondestructive evaluations were performed with particular attention paid to locations predicted by FEA to be highly stressed. Metallographic preparation and SEM analysis of cross-sectioned package structures were performed to determine the integrity of the Al2 O3 frame and the frame/base-flange interfaces.

Results

Strength-degrading flaws in and on ceramics can originate in the initial materials (binders, surfactants, sintering aids) from forming-related cracks, voids and metallic impurities introduced during grain growth and densification and final shaping and machining. The Al2 O3 varieties examined for this study possessed nearly equivalent chemical purity. Analyses indicated Na, Si, Fe and Ca (which were all less than 20 ppm) to be the dominant impurities. The microstructural characteristics of grain size and porosity that evolve from the specific fabrication parameters can significantly influence the mechanical properties. The strengths, Weibull parameters, fracture toughness, and grain sizes and porosities for the Al2 O3 varieties are listed in Table 3 . It is evident that the 96 percent Al2 O3 material exhibited the lowest and least reproducible strength characteristics while the 99 percent Al2 O3 , although possessing superior strength, had a Weibull modulus indicating more variability than the ZTA.

Table III
Al2 O3 Frame Insulator Mechanical and Microstuctural Characteristics

Al2 O3 (t=0.025")

96%

99%

ZTA (20 wt. % Zr02 )

MOR strength (MPa)

390

660

501

Weibull modulus

8.2

7.7

10.0

Porosity (%)

1.50

0.38

0.51

Grain size (µm)

9

6

4

Fracture Toughness (MPa m-1/2 )

3.4

4.8

7.0

Relative Cost

1.0

1.4

1.7

4 pt. bend - specimen size = 9mm x 1.5mm  x 0.6mm bar; fracture toughness by short-bar specimen

The mechanical property trends followed observed differences in grain size and porosity characteristics with smaller grain size and lower porosity exhibiting improved mechanical properties. All of the Al2 O3 materials exhibited intergranular fracture features, as shown in Figure 2 , with critical flaw dimensions on the order of the grain size.

Finite element modeling of warp deformation, principal stress S1 at insulator frame critical corner sites and  Pf for each Al2 O3 variety was performed using the previously listed package structure components subjected to the assembly brazing processes outlined earlier. The single-quadrant finite element mesh structure for the package (dimensionally and morphologically symmetric in the other quadrants) is shown in Figure 3 , along with a magnified view of the critical corner region of the frame. A cross-sectional model view of a CuMoCu base flange, Al2 O3 insulator and Cu I/O lead is shown in Figure 4 . The dashed-line silhouette of a prefabricated assembly and a full schematic of a postfabrication package show the warp deformation following assembly processing. A maximum concave upward warp of 61 mm was predicted over the 20 mm longitudinal dimension of the base flange. Under such conditions, the maximum predicted stress in the Al2 O3 was 516 MPa, spatially located at the interface between the base flange and Al2 O3 .

A stress contour map of the Al2 O3 (quadrant) is shown in Figure 5 , illustrating the maximum stress at the critical corner site and a decreasing gradient to the free surface. A plot of the through-thickness S1 at the critical corner site of the Al2 O3 insulator frame vs. the thickness position in the frame is shown in Figure 6 , indicating the decreasing stress gradient from the bonded interface to the free surface. The results summary for all assembly scenarios is listed in Table 4 for CuMoCu and CuW base flanges assembled to each Al2 O3 variety using the DBC/1080 and AgCu 860 processing schemes.

Table IV
FEA Results for Package Deformation and Al2 O3 Stress and Failure Probabilities

Package Assembly

Warp Distortion

Al2O3 Frame Stress*

Failure Probability Fraction

 

(µin/µm)

(MPa)

96% Al2 O3

ZTA

99% Al2 O3

1

(+)1.5x103 /38

396

0.03739

0.006103

0.003593

2

(+)1.3x103 /34

369

0.02042

0.003261

0.001913

3

(+)2.4x103 /61

516

0.3316

0.0610

0.01693

4

(+)2.0x103 /52

455

0.1226

0.02029

0.006927

5

(-)162/4.1

32

4.758x10-12

8.593x10-13

7.973x10-10

6

(-)95/2.4

29

1.033x10-11

1.859x10-12

1.427x10-9

7

(-)297/7.5

36

1.267x10-11

2.280x10-12

1.693x10-9

8

(-)190/4.8

34

2.369x10-11

4.297x10-12

2.570x10-9

(+) = concave up, (-) = concave down
* Al2O3 frame stress is reported at critical corner site

CuMoCu Structures

As the base-flange thickness is reduced from 0.040" to 0.020", assembly using the DBC/1080 produced a 38 percent increase in warp deformation and a 23 percent increase in stress in the Al2 O3 insulator frame material at the critical corner site. The Pf increased nearly an order of magnitude for the 96 percent Al2 O3 and ZTA while the increase was 4x for the 99 percent Al2 O3 . A similar trend of lower increments was observed for AgCu/860 processing with a 35 percent increase in warp deformation,  19 percent increase in stress in the critical corner region of the Al2 O3 insulator frame material and 4 to 5¥ increase in the Pf for each Al2 O3 variety. Maintaining the base-flange thickness constant and changing the assembly material/process temperature from DBC/1080 to AgCu/860 produced a 12 percent decrease in warp deformation, seven percent decrease in stress in the critical corner region of the Al2 O3 insulator frame material and nearly a 2x decrease in Pf for all Al2 O3 varieties for the 0.040" base-flange thickness. The thinner 0.020" base flange displayed slightly more advantageous figures with a 15 percent reduction in warp deformation, 12 percent reduction in Al2 O3 insulator frame stress at the critical corner region and approximately a 3x decrease in Pf for all Al2 O3 varieties. The absolute magnitudes of warp deformation ranged from 1.3 ¥ 103 min (33.5 mm) for a 0.040" base flange assembled using AgCu/860 processing (package assembly no. 2) to 2.4  x 103 min (61.4 mm) for the 0.020"-thick base flange assembled using DBC/1080 (package assembly no. 3).

The warping polarity for all structures was concave upward. Verification of these features through profilometry on the base-flange backside showed agreement within 12 percent. Principal stresses in the Al2 O3 insulator frame materials at the critical corner region followed the deformation trend, from a minimum of 369 MPa for package assembly no. 2 to a maximum of 516 MPa for package assembly no. 3. The Pf was lowest (2 x 103 ppm) for package assembly no. 2 with the 99 percent Al2 O3 in a 0.040" base flange assembled using AgCu/860 and highest (3 x 105 ppm) for package assembly no. 3 with the 96 percent Al2 O3 in a 0.020" base flange using DBC/1080 processing.

CuW Structures

Base-flange thickness reduction from 0.040" to 0.020" concurrent with using the DBC/1080 processing resulted in a 45 percent increase in warp deformation, 15 percent increase in Al2 O3 insulator frame stress at the critical corner region and approximately a 2¥ increase in Pf for the three varieties of Al2 O3 . The identical reduction in base-flange thickness coupled with the AgCu/860 assembly process increased warp deformation 50 percent, increased Al2 O3 insulator frame stress 11 percent at the critical corner region and increased Pf of all Al2 O3 varieties approximately two-fold. Maintaining base-flange thickness at 0.040" and changing the assembly material/process temperature from DBC/1080 to AgCu/860 produced a decrease in warp deformation of 42 percent, stress decrease in the Al2 O3 insulator frame of nine percent at the critical corner region and approximately a 2¥ increase in Pf for all Al2 O3 varieties. A 0.020" base flange exhibited similar trends with warp deformation decreasing 36 percent, Al2 O3 insulator frame stress decreasing five percent at the critical corner region and an approximate 1.7x increase in Pf for all Al2 O3 varieties.

The absolute magnitudes of warp deformation were an order of magnitude lower and of opposite polarity (concave down) relative to the CuMoCu structures, and were in 10 percent agreement with profilmetric measurements. A 0.040" base flange assembled with AgCu/860 to Al2 O3 showed a 95 min (2.5 mm) warp (package assembly no. 6), while the 0.020" base flange assembled to Al2 O3 using DBC/1080 showed a 297 min (7.5 mm) warp (package assembly no. 7). Principal stress in the Al2 O3 critical corner region ranged from 29 MPa for a 0.040" base flange assembled to Al2 O3 using DBC/1080 (package assembly no. 6) to 36 MPa for a 0.020" base flange assembled to Al2 O3 using AgCu/860 (package assembly no. 7) — more than 10¥ lower than observed for Al2 O3 in the CuMoCu structures. The Pf was lowest at 10–7 ppm for a 0.040" base flange assembled to ZTA using DBC/1080 (package assembly no. 5) and highest at 10–3 ppm for a 0.020" base flange assembled to 99 percent Al2 O3 using AuCu/860 (package assembly no. 8).

Nondestructive and destructive evaluation of selected package assembly schemes (2 and 3 for CuMoCu and 6 and 8 for CuW) using each Al2 O3 variety showed insulator frame cracking (postfabrication) for CuMoCu/96 percent ZTA structures, as shown in Figure 7 , and base-flange/

bond interface voiding, as shown in Figure 8 . The observed crack location in the Al2 O3 insulator frames was in good agreement with the location predicted by FEA. The fracture surface morphology of cracked insulator frames (CuMoCu/Al2 O3 ), shown in Figure 9 , displayed intergranular fracture similar to the four-point bend test specimens and suggests assembly-induced loading characteristics to be tensile. Examination of CuMoCu/99 percent Al2 O3 and CuW/96 percent ZTA, 99 percent Al2 O3 structures revealed neither insulator frame cracking nor bond interface voiding. Stress (principal, S1) in the Al2 O3 varied most significantly with base-flange material and to lesser degrees with base-flange thickness and assembly braze material/process temperature.

Compared to the measured flexure strengths, only the 99 percent Al2 O3 material exhibited a strength in excess of the assembly-induced stresses for all CuMoCu structures. The 96 percent, 99 percent and ZTA materials possessed sufficient strength to be considered in the application with CuW structures for all thickness and process variations. The rating hierarchy for performance/cost of the Al2 O3 materials differed significantly between CuMoCu and CuW structures. An order-of-magnitude difference of approximately 10 in Pf figures for Al2 O3 in these structures posed difficulties in assigning comparable failure threshold tolerability limits. For the CuMoCu structures, a one percent Pf showed that three of four assembly/processing scenarios for the 99 percent Al2 O3 material would achieve the designated survivability threshold. For the ZTA material, only two of four assembly/ processing scenarios would achieve the designated survivability threshold, and 96 percent Al2 O3 would not survive any designated assembly/processing scenario, the lowest Pf being approximately two percent. Therefore, using the relative cost comparison, a strength-driven performance/cost selection of 99 percent Al2 O3 followed by ZTA would be appropriate. If a higher failure probability threshold would be tolerable (from two to 33 percent), appropriate selection of an assembly/processing scenario could accommodate 96 percent Al2 O3 material with concurrent cost reduction. For all listed assembly/process scenarios, the Al2 O3 materials in CuW structures showed Pf figures < 10–9, suggesting survivability of the Al2 O3 in all cases. A performance ranking hierarchy of 99 percent Al2 O3 < 96 percent Al2 O3 < ZTA and a relative cost ranking of 96 percent Al2 O3 < 99 percent Al2 O3 < ZTA suggests a cost-driven performance-cost selection of 96 percent Al2 O3 followed by 99 percent Al2 O3 .

Conclusion

For a given thickness and assembly brazing/process temperature, a base-flange material change from CuMoCu to CuW produced a 10¥ decrease in warp deformation, 10 to 15x decrease in stress in the Al2 O3 critical corner region and 7 to 10 order of magnitude decrease in the Pf of the Al2 O3 . For a given base-flange material and assembly brazing/process temperature, a base-flange thickness change produced approximately a 30 to 50 percent change in warp deformation, 10 to 20 percent change in stress in the Al2 O3 critical corner region and factor of 2 to 10 change in the Pf of the Al2 O3 . For a given base-flange material and thickness, the assembly brazing/process temperature was least influential in CuMoCu structures, exhibiting approximately a 12 to 15 percent change in warp deformation, approximately 10 percent change in stress in the Al2 O3 critical corner region and approximately a factor of 2 change in the Pf of the Al2 O3 . However, although an order of magnitude lower in absolute magnitude of warp deformation and stress, CuW structures exhibited significant changes in warp deformation (36 to 50 percent) for both base-flange thickness and assembly brazing/process temperature. The stress in the Al2 O3 at critical corner regions decreased with decreasing warp deformation and although Pf of the Al2 O3 changed only a factor of 2 for the base-flange thickness and assembly braze material/process temperature changes, the absolute magnitudes of Pf ranged from 10–3 to 10–7 ppm.

The judicious selection of base-flange material and thickness as well as assembly process material and temperature can minimize structure warp deformation, Al2 O3 material stress and Al2 O3 material failure probabilities. The cumulative performance (fracture strength, Weibull modulus, Pf)/cost rating suggested the selection of the Al2 O3 material for CuMoCu structures to be strength dominated; a cost partitioning was not possible unless higher Pf could be tolerated. Conversely, for CuW structures the low Pf figures offered numerous successful performance/cost selections for the Al2 O3 materials incorporating the various options of base-flange thickness and assembly process/material.

Acknowledgment

Special thanks are extended to Kiran Dalal and Jeff Karker of Brush-Wellman, Newburyport, MA, for their assistance with this work. The ANSYS version 5.5.2 software is a product of ANSYS Inc., Houston, PA.

References

1.         W. Weibull, “A Statistical Distribution of Wide Applicability,” Journal of Applied Mechanics, Vol. 18, No. 3, 1951, p. 293.

2.         A.G. Evans, “A General Approach for the Statistical Analysis of Multiaxial Fracture,” Journal of the American Ceramic Society, Vol. 61, No. 7–8, 1978, p. 302.

3.         M.F. Ashby, Materials Selection in Mechanical Design, Oxford, Butterworth-Heinemann, Reed-Elsevier PLC Group, 1992.

Philip M. Fabis is a senior materials engineer and manager, Metalization Facility at St. Gobain Industrial Ceramics, Norton Diamond Film. His expertise is in materials deposition and modification and analysis of thin films, surfaces and interfaces.