RELIABILITY TESTING OF Ni-MODIFIED SnCu AND SAC305 – ACCELERATED THERMAL CYCLING Joelle Arnold, Nathan Blattau, and Craig Hillman DfR Solutions College Park, MD USA
[email protected] Keith Sweatman Nihon Superior Co., Ltd. Osaka, Japan
ABSTRACT Interest in alternate lead free solder alloys has initiated a life prediction study of SN100C. This paper outlines the results of the thermal cycling portion of the testing and demonstrates that its robustness under accelerated life testing is somewhere between that of SnPb and SAC305 with its performance being dependent on the amount of strain to which it is subjected in the particular test. A low cycle fatigue exponent of 2.38 is proposed, and the acceleration factor is presumed to be dependent upon maximum temperature and relatively insensitive to the dwell time. Key words: SN100C, Thermal cycling, Modified NorrisLandzberg INTRODUCTION Since the institution of the Restriction Of Hazardous Substances (ROHS) Directive in the European Union on July 1, 2006, and mounting legislation in Asia and the rest of the world, concerns related to the reliability and manufacturabilty of lead-free (Pb-free) solders, particularly on large surface mount devices, have been growing. While much of the consumer industry has embraced SAC305 the concern over its drop/shock performance in mobile applications, it's effect on wave soldering equipment, and the cost of silver has led to research in other Pb-free tincopper (Sn-Cu) metallurgies with little or no silver, such as SAC105, SAC0307 (0.3% silver) SN100C and other tinsilver-copper alloys (SACX).
mechanical shock after aging, and vibration tests were all performed. Both SnPb and SAC305 were tested in parallel under the most severe conditions of each experiment type. For detailed procedures, please reference Accelerated Life Testing of SN100C for Surface Mount Devices1. EXPERIMENTAL PROCEDURE The test vehicles selected for this activity consisted of three package styles: a leadless 2512 ceramic chip resistor, a 44 Lead, Type 1, Thin Scale Outline Package (TSOP) with an Alloy 42 leadframe (gull-wing leads), and a small ball grid array Chip Scale Package (CSP). These types of packages represent the three major families of attachment styles, leadless, ball grid array and leaded. These particular parts were selected due to their tendency to fail relatively rapidly when subjected to cyclic thermal or mechanical loads. Test boards are visible in Figure 1, Figure 2 and Figure 3. 2512 Resistors Therminator heater
Monitoring channels
Figure 1: Resistor test board with 40 components The Sn-Cu without silver was quickly rejected as a viable alterative to Sn-Pb solder due to an undesirable grain structure. Joints formed of Sn-Cu alone demonstrated a cracked and dull surface finish with visible dendrites, and had poor strength properties. However, with the addition of silver or nickel, the appearance, grain structure and mechanical behavior of Sn-Cu can be improved. Nihon Superior contracted DfR Solutions to test their SN100C (Sn-0.7Ni-0.05Cu+Ge) alloy in order to generate an acceleration factor (AF) and other models for failure prediction and accelerated life testing. Thermal cycling,
Chip scale
Therminator heater
Resistor SN100C TSSOP CSP Resistor SAC305 TSSOP CSP SnPb Resistor TSSOP CSP SUM
10min @ 125°C 5 min @ -40°C 1 2 1 1 2 1 1 2 1 12
Thermal Cycling 30min @ 125°C 10min @ 125°C 10min @ 100°C 5min @ 25°C 5min @ 25°C 5min@25°C 1 1 1 2 2 2 1 1 1
4
4
4
Figure 4: Boards subjected to each thermal profile by component type and solder Monitoring channels
Figure 2: CSP test board with 24 components TSOP components Heater circuit
Monitoring channels
Figure 3: TSOP test board with 12 components. Two TSOP boards were tested for each condition to achieve 24 samples. Because SN100C is currently not offered by part suppliers as a BGA metallurgy the chip scale packages (CSP) were ordered without solder spheres and then balled with SN100C spheres. THERMAL CYCLING The more benign thermal profiles (25-100ºC, 25-125ºC short dwell, and 25-125ºC long dwell) were performed using an environmental testing system designed at DfR Solutions. This system is a printed wiring board level power cycler and allows for precise temperature control, minimal thermal gradients and the ability to apply rapid ramp rates (20ºC/minute) for more manageable cycle times. Please reference Rapid Power Cycling of Pb-Free Soldered Components2 for details on the functionality of the system. The more extreme profile (-40-125ºC) was conducted in a thermal chamber with use of liquid nitrogen to facilitate cooling below room temperature. The number of boards subjected to each profile is displayed in Figure 4.
RESULTS Thermal cycling results have been discussed in greater detail in Accelerated Reliability Testing of Ni-Modifified SnCu and SAC3053 and are reviewed in Figure 7 through Figure 8.
Figure 7: Test results for 98 I/O CSP reflowed with SN100C subjected to temperature cycles of 25-100°C, 25-125°C (long and short dwells) and -40-125°C ∆T
Figure 5: Test results for 2512 resistors reflowed with SN100C subjected to temperature cycles of 25-100°C, 25-125°C (long and short dwells) and -40-125°C
75 100 100 165
short dwell long dwell
CSP TTF (cycles) 4792 1757 1975 867
TSOP TTF (cycles) 6061 3584 2467 1071
Resistor TTF (cycles) 6274 897 1342 628
Figure 8: Summary of characteristic life for thermal cycling CORRELATION OF THERMAL CYCLING DATA TO LIFE PREDICTION MODELS Life prediction models that are temperature dependent rely upon an acceleration factor, AF, to scale the laboratory data to field prediction. One approach for relating solder fatigue behavior under different thermal cycles is the NorrisLandzberg model:
f N AF = O = t Nt fO Figure 6: Test results for 44 I/O TSOP reflowed with SN100C subjected to temperature cycles of 25-100°C, 25-125°C (long and short dwells) and -40-125°C
−1 / 3
1 ∆Tt 1 exp 1414 − ∆TO Tmax,O Tmax,t 2
The above displayed formula for acceleration factor has terms specific to SnPb. The acceleration factor depends upon the maximum temperature experienced, both under field conditions (O) and testing (t), the change in temperature from high to low during the cycle, and the frequency of the cycles. The acceleration factor, AF, is simply the ratio of number of cycles survived under operating conditions (NO) and under test conditions (Nt). For SAC305, a modified Norris-Landzberg model was recently proposed4: AF =
N O tt = N t tO
0.136
∆Tt ∆TO
2.65
1 1 exp 2185 − T T max,t max,O
While follow up analyses5,6, which incorporated a broader range of test data, found poor correlation, the basic trends identified by this equation still bear some merit. Given that
SN100C is a high tin, lead-free alloy, it was decided to base the development of an acceleration factor on the NorrisLandzberg model.
Temperature to Test Time to Failure Correlation CSP
TSOP
Resistor
MAXIMUM TEMPERATURE (TMAX) While maximum temperature is likely to have an influence of time to failure for SN100C, just as it has for SnPb and SAC305, the test variation chosen for this experiment was too limited to extrapolate this information. Future work in the regime of 0 to 100C and -40 to 60C, to correlate with 25 to 125C, should be sufficient to capture the appropriate constants.
Time to Failure (Cycles)
7000 6000 5000 4000 3000 2000
y = 2E+08x-2.3805
1000 0
TEMPERATURE RANGE (∆T) The Norris-Landzberg model is an approximate acceleration factor for a solder alloy irrespective of part type. For this reason, when deriving such a model, it is key to use a number of different part types. By plotting the characteristic cycles to failure of each component type with respect to the thermal cycle experienced, curve fit equations were generated to capture the influence of temperature range (Figure 9). While trendlines were developed for each part type, the components were treated equally and a trendline was fitted to all SN100C components (Figure 9). Temperature to Test Time to Failure Correlation CSP
TSOP
Resistor
Power (CSP)
Power (TSOP)
Power (Resistor)
50
70
90
110
130
150
170
190
Change in Temperature (oC)
Figure 10: Fit of power equation to all data points Due to the limited scope of the tests a full implementation of the Norris-Landzberg equation (modified Coffin-Manson) was not done and only a Coffin-Manson equation was developed where the acceleration factor is merely a ratio between the number of cycles under operating conditions and the number of cycles under laboratory conditions raised to a power. These are inverted from the ratios within the equation. For this reason, it’s not uncommon to normalize and use the reciprocal. Thus, the plot in Figure 10 can also be plotted as Figure 11. Normalized Look at Power Equation 4.5 4
Normalized Cycles to Failure
DWELL TIME AT MAXIMUM TEMPERATURE (t) Previous work on SAC305, and to a lesser extent on SnPb, have identified that the dwell time at maximum temperature has a definitive influence on time to failure. Given this history, it would be expected that SN100C would also display some dependence on dwell time at maximum temperature. However, within the limited range of dwell times, 10 and 30 minutes, the influence seems to be marginal. This limitation effectively prevents the development of a constant within the life model to describe dwell time effects.
3.5 y = x2.3801 3 2.5 SN100C Components (all)
y = 0.9353x2.3801
2 1.5 1 0.5
7000
Time to Failure (Cycles)
0
6000
0.6 y = 7E+07x-2.172
1
1.2
1.4
1.6
1.8
Normalized Temperature Gradient
5000
Figure 11: Normalized cycles to failure with respect to normalized temperature range cycled
y = 3E+07x-2.0622
4000
0.8
3000 2000 1000 -2.6202
y = 3E+08x
0 50
70
90
110
130
150
170
190
o
Change in Temperature ( C)
Figure 9: Correlation of change in temperature to cycles to failure for SN100C assembled devices.
From Figure 11, the value “x” in the trend line is the relationship between the temperature range (∆T) under test conditions and under operating conditions. The rest of the formula for the acceleration factor is relatively constant; the driving force for acceleration of failures is the temperature range. Thus, the rest of the relationship can be set equal to 1 with very little change of fit. As a result, the current acceleration factor model for SN100C is:
AFSN 100C
∆T = t ∆TO
2.38
This is the same manner in which the acceleration factor for SnPb and SAC are often simplified
AFSnPb
∆T = t ∆TO
2
AFSAC and
∆T = t ∆TO
2.65
.
The fatigue exponent, 2.38, indicates that the acceleration of life under thermal cycling of SN100C is somewhere between that of SnPb and SAC. DISCUSSION Sufficient data was obtained from temperature cycling to provide initial correlation for SN100C behavior under test conditions to field environments. Not surprisingly, the data seems to suggest that the SN100C will demonstrate a dependence upon temperature range, roughly between SAC305 and SnPb. It is important to note that this attempt at a Norris-Landzberg model will require additional test data to obtain dependence upon maximum temperature and dwell at maximum temperature, though initial indications are that SN100C is less sensitive to dwell at maximum temperature than SAC305. In a similar vein to vibration, improvements in prediction can possibly be obtained through the use of first order, second order or FEA modeling. The ability to develop these models will require information on stress-strain and creep behavior of the SN100C alloy and this data will shortly become available.
1
Arnold, et al. “Accelerated Life Testing of SN100C for Surface Mount Devices,” IPC/JEDEC Global Conference on Lead Free Reliability and Reliability Testing for ROHS Lead Free Electronics, Boston, MA, 2007. 2 Blattau, et al. “Rapid Power Cycling of Pb-Free Soldered Components,” Procedings: IPC Printed Circuits Expo, APEX and Designers Summit, Los Angeles, CA, 2007. 3 Arnold and Sweatman, “Accelerated Reliability Testing of Ni-Modified SnCu and SAC305,” IPC/JEDEC International Conference on Reliability, Rework and Repair of Lead-Free Electronics, Raleigh, NC 2008. 4 Pan et. al., “An Acceleration Model for Sn-Ag-Cu Solder Joint Reliability Under Various Thermal Cycle Conditions,” SMTAI 2005 5 C. Hillman, “Assessment of Pb-Free Norris-Landzberg Model to JG-PP Test Data, http://www.acqp2.nasa.gov/LFS%20Reliability/JGPP%20and%20NL%20Model%20AnalysisII.pdf 6 Olli Salmela, “Acceleration Factors for Lead-Free Solder Materials,” SMTAI 2006
