1 INTRODUCTION: Thermoelectric cooling modules are considered to be highly reliable components due to their solid-state construction. For most applications they will provide long, trouble-free service. There have been many instances where TE modules have been used continuously for twenty or more years and the life of a module often exceeds the life of the associated equipment. The specific reliability of thermoelectric devices tends to be difficult to define, however, because failure rates are highly dependent upon the particular application. For applications involving relatively steady-state cooling where DC power is being applied to the module on more-or-less continuous and uniform basis, thermoelectric module reliability is extremely high. Mean Time Between Failures (MTBFs) in excess of 200,000 hours are not uncommon in such cases and this MTBF value generally is considered to be an industry standard. On the other hand, applications involving thermal cycling show significantly worse MTBFs especially when TE modules are cycled up to a high temperature.

The publishing of thermoelectric module reliability data entails some risk because there are numerous application parameters and conditions that will affect the end result. Although reliability data is valid for the conditions under which a test was conducted, it is not necessarily applicable to other configurations. Module assembly and mounting methods, power supply and temperature control systems and techniques, and temperature profiles, together with a host of external factors, can combine to produce failure rates ranging from extremely low to very high. In an effort to provide users with certain basic information about thermoelectric module life and to aid engineers in designing systems for optimum reliability, we instituted several test programs to acquire the necessary reliability data. Test results to date are presented for several situations that may be useful to end-users having similar or related applications. This data will be shared on a case-by-case basis depending on application and availability.

General requirements for the proper installation of thermoelectric modules may be found in Section 6 of this technical manual. It is important that modules are installed in accordance with these general requirements in order to minimize the possibility of premature module failure due to faulty assembly techniques. Some installation related factors that can affect module reliability include:

a) Thermoelectric modules exhibit relatively high mechanical strength in a compression mode but shear strength is comparatively low. A TE cooler should not be designed into a system where it serves as a major supporting member of the mechanical structure. Furthermore, in applications where severe shock and vibration will be present, a thermoelectric cooling module should be compression-mounted, i.e., installed by the clamping method. When properly mounted, thermoelectric coolers have successfully met the shock and vibration requirements of aerospace, military, and similar environments.

b) Although the maximum recommended compression loading for thermoelectric modules is 15 kilograms per square centimeter (200 pounds per square inch) of module surface area, tests have shown that well over 75 kilograms per square centimeter (1000 pounds per square inch) compression normally can be applied to most of our modules without causing failure. It is important to ensure that when modules are installed using the clamping method, sufficient pressure is maintained so that a module is not "loose" whereby it may easily be moved by applying a small sideways or lateral force. Loose modules may be a particular problem when several modules are grouped together in the same cooling assembly. In this situation, the lack of adequate clamping pressure may result in both reduced cooling performance and early module failure. When multiple modules are mounted in an array, modules with a close height tolerance of +/- .03mm (.001") are recommended. In all cases, clamping pressure must be applied uniformly and mating surfaces must be flat (see section 6 for Installation Guidelines).

c) A large unsupported mass should not be directly bonded to a module's cold surface to prevent the possible fracture of module components when subjected to significant mechanical shock. Where a large mass is involved, thermoelectric modules should be clamped between the heat sink and either the mass itself or an intermediate "cold plate" on which the mass is mounted. In this arrangement, the clamping screws will effectively increase shear strength of the overall mechanical system.

d) Moisture should not be allowed to enter the inside of a thermoelectric module in order to prevent both a reduction in cooling performance and the possible corrosion of module materials through electro-chemical action or electrolysis. When cooling below the dew point, a moisture seal should be provided either on the module itself or between the heat sink and cooled object in the area surrounding the TE module. An electronic-grade silicone rubber RTV may be used to directly seal a thermoelectric module. Flexible closed-cell foam insulating tape or sheet material, possibly combined with RTV to fill small gaps, may be used for a seal between the cold object and the heat sink.

e) When an application will involve large temperature changes or thermal cycling, thermoelectric modules should not be installed using solder or epoxy whereby an object is rigidly bonded to the module. Unless the thermal coefficients of expansion of all system components are similar, rigid bonding combined with temperature cycling often will result in early module failure due to the induced thermal stresses. Rigid bonding to the module's hot side generally is less of a problem because the hot side temperature tends to be relatively constant during operation. When significant temperature variation or temperature cycling is involved, we strongly recommend that modules be mounted by clamping (compression) using a flexible mounting material such as thermal grease or foils of graphite or indium. In addition, rigid mounting to both sides of modules is not recommended for devices larger than about 15mm (5/8") square.

Temperature control methods also have an impact on thermoelectric module reliability. Linear or proportional control should always be chosen over ON/OFF techniques when prolong life of the module is required.

2 MODULE RELIABILITY RELATED TO HIGH TEMPERATURE EXPOSURE 

Thermoelectric module failures typically may be classified into two groups: catastrophic failures and degradation failures. Degradation failures tend to be long-term in nature and usually are caused by changes in semiconductor material parameters or increases in electrical contact resistance. High temperature exposure may lead to material parameter changes and, therefore, reduced thermoelectric performance. A test was conducted to study this effect. Ferrotec's 95-Series TE modules were subjected to long-term, continuous exposure to an elevated temperature of 150°C in a normal air atmosphere. During the test period, relevant module parameters were regularly measured and recorded. One parameter that is a good indicator of overall module performance is the maximum temperature differential (DTmax). This parameter was tracked over a 42-month period with the average value being shown the graph of Figure (1). It can be seen that a small (2.5%) decline in DTmax, with a decreasing rate of change, occurred in the first 12 months of high temperature exposure. In the remaining 30 months, however, the additional reduction in DTmax was only about 1.3% as semiconductor material characteristics stabilized.

Figure (1)

3 MODULE RELIABILITY RELATED TO THERMAL CYCLING 

The continuous thermal cycling of thermoelectric modules over a wide temperature range effectively constitutes a module "torture test," especially when the modules are raised to a relatively high temperature at one end of the cycle. Except for a few unusual applications, module failure rates are higher for this mode of operation than for any other operating condition. The basis for most thermal cycling failures is the unavoidable mismatch of thermal expansion coefficients of the various module components and materials. Such failures tend to be catastrophic in nature but some degradation normally may be observed prior to failure.

It is necessary, at this point, to define thermal cycling. Many thermoelectric applications involve the periodic raising and lowering of the control temperature, sometimes over a fairly wide temperature range. Although there often is not a well defined line between a cycling and noncycling application, thermal cycling usually is considered to be an operation where the temperature is regularly, and more or less continuously, raised and lowered over a long period of operation. A cycling application tends to suggest automatic or machine control of the temperature excursion as opposed to manual control. If the temperature of an apparatus is temperature-cycled up and down a few times each day, this generally is not considered to be a temperature-cycling application for the purpose of this discussion. If you are uncertain about the status of your particular application, please do not hesitate to contact us for assistance.

At least four factors relate to failure rate in thermal cycling including (1) the total number of cycles, (2) the total temperature excursion over the cycle, (3) the upper temperature limit of the cycle, and (4) the rate of temperature change. Highest reliability and module life is seen when the number of cycles is small, the temperature excursion or range is narrow, the upper temperature limit is relatively low and the rate of temperature change is minimalized. (Conversely, a large number of cycles over a wide temperature range with a rapid rate of change and a high temperature value on the up cycle results in significantly lower module life.) It is important to note that absolute module life is dependent upon the total number of cycles rather that the total time required to accrue those cycles. Consequently, when discussing thermal cycling, MTBF is best stated in terms of number of cycles instead of hours; we will take the liberty of using MTBF in this manner in the following discussion.

The type of module used in thermal cycling applications also is important in respect to failure rate. Modules rated at higher maximum operating temperatures provide substantially better life than do lower rated devices. This is true even though the upper temperature of the cycle is well below the maximum rated module temperature. In one application involving a two-stage thermoelectric assembly that was being cycled between -55°C and +125°C, a 150°C rated module provided a MTBF of 8100 cycles while a module rated at 200°C exhibited a MTBF of 17,500 cycles. Modules rated at even lower maximum operating temperatures should only be used for relatively low temperature cycling applications. In general, we recommend the SuperTEC series modules (rated for 200°C) be used for thermal cycling applications exceeding 90°C.

It should be mentioned that two other factors also may affect thermal cycling MTBFs. Physically smaller modules having fewer couples appear to provide improved life as do modules having larger elements or "dice." Sufficient data is available to suggest that modules having a size of 30mm (1.17") square or less exhibit better reliability in thermal cycling applications than do physically larger modules. Thermally induced mechanical stresses are greater in larger modules and such modules generally have a greater number of couples resulting in many more individual solder connections which may become fatigued by thermal stress.

In order to better define module failure rates under high temperature thermal cycling conditions, a test was conducted involving the continuous cycling of SuperTEC™ Series modules between +30°C and +100°C. Modules were mounted on a forced convection heat sink and covered with an insulated aluminum plate. Polarity of the applied DC power was alternately reversed to provide active heating and cooling and the cover-plate temperature was measured to determine cycling limits. The total time period of the cycle was 5 minutes (2.5 minutes from 30°C to 100°C and 2.5 minutes from 100°C to 30°C) resulting in 288 cycles per day or 2016 cycles per week. Module parameters were measured weekly and a failure was indicated by a sharp rise in electrical resistance.

Modules showed a slow and predictable rise in electrical resistance until reaching a point where a rapid resistance increase occurred indicating failure. All modules achieved a minimum of 25,000 cycles without failure, see Figure (2), and the test was continued until 50% of the modules failed. MTBF of the module group was calculated to be 68,000 cycles. Once again it is important to note that mounting methods, and overall assembly details are important factors when the application involves thermal cycling. Some applications have been tested between 5°C and 95°C exhibiting MTBF's over 100,000 cycles.

Figure (2) Before leaving the subject of thermal cycling it might be worthwhile to mention a practical use for this process. Because of the resulting mechanical stresses within a thermoelectric module, thermal cycling has been shown to be an effective "burn-in" technique. By subjecting thermoelectric devices to a well controlled cycling program, it is possible to identify potentially unsatisfactory modules thereby reducing the likelihood of infant mortality failures. There obviously is some cost associated with this operation but it may be useful when extremely high reliability is required.

4 MODULE RELIABILITY RELATED TO ON/OFF POWER CYCLING 

As discussed previously, the accepted industry standard for thermoelectric module MTBF is 200,000 hours minimum. This MTBF value is based on relatively steady-state module operation where system power is occasionally (typically a few times per day) turned on and off. In some applications power is turned on and off more frequently especially where thermostatic temperature control is used. A test was conducted using ValueTEC™ Series modules to study the effects of ON/OFF power cycling at a relatively constant temperature. Modules were mounted between a pair of forced convection cooled heat sinks using thermal grease at the module/heat sink interfaces. Full rated current was supplied to the modules for a period of 7.5 seconds followed by a 7.5 second "off" period that resulted in one complete ON/OFF cycle every 15 seconds. The input current to each module was monitored and a failure was indicated by an appreciable current decrease resulting from an increase in module electrical resistance. The test was run until an arbitrary total of 25,000 hours or approximately 6 million cycles was accrued. For these test conditions, the calculated MTBF was 125,000 hours or 3x107 on off cycles.

CAUTION: Most conventional thermostats inherently have moderately large open/close temperature differentials. This condition may effectively set up a thermal cycling situation where the temperature of the TE module is continuously varying between the upper and lower differential limits. Since thermal cycling is known to reduce the life of thermoelectric modules, the use of traditional ON/OFF thermostatic temperature control schemes is not recommended for high-reliability applications.

5 ENVIRONMENTAL CONSIDERATIONS 

Thermoelectric modules often are installed in systems that are subject to significant shock, vibration, and/or other potentially detrimental environmental conditions. As mentioned earlier in this report, modules will withstand moderate compression forces but shear strength is relatively low. However, when thermoelectric modules are properly mounted within a mechanical subassembly, they will withstand substantial mechanical stress without failure.

Ferrotec's modules have been subjected to a number of environmental/mechanical test conditions and have successfully met those conditions without failure. Such tests include:

High Temperature Operations and Storage:

150°C for 30,000+ hours

Low Temperature Operations and Storage:

-40°C for 1000+ hours

Thermal Shock:

(a) 100°C (15 sec)/100°C (15 sec), 10 cycles 

(b) 150°C (5 min)/-65°C (5 min)/ 150°C, 10 cycles 

(c) MIL-STD-202, Method 107

Range for ValueTEC™ Series modules: -55oC to +85oC 

Range for SuperTEC™ Series modules: -65oC to 150oC

Mechanical Shock:

(a) 100G, 200G 2 6msec; 500G, 1000G @1 sec 3-axis, three shocks each axis 

(b) MIL-STD-202, Method 213, Test Condition I

Vibration:

(a) 10/55/10 Hz,1 minute cycle, 9.1G, 3-axis, 2-hours each axis 

(b) MIL-STD-202, Method 204A, Test Condition B, 15G Peak

6 STANDARD QUALITY CONTROL PROCEDURES 

Thermoelectric device manufacturers have independently developed quality control and test procedures to insure that products meet published specifications and exhibit acceptable standards of workmanship. While few formal standards (Military Specifications, etc.) exist within the industry, there have emerged certain minimum recognized criteria to which most major thermoelectric manufacturers adhere. However, if users have particular concerns about quality-related issues that may affect their specific application, it generally is desirable for users to discuss their concerns with individual thermoelectric manufacturers.

Ferrotec America's test and quality program has evolved from many years of industry experience covering an extensive range of thermoelectric cooling applications. General aspects of this program include 100% electrical and mechanical testing/inspection of products prior to shipment; in-process testing and screening using either 100% inspection or sampling inspection as per MIL-STD-105; and the use of statistical process control techniques on various critical operations. The overall quality assurance program is structured in accordance with MIL-Q-9858A.

7 CONCLUDING REMARKS 

In the foregoing discussion we have emphasized the great dependence of thermoelectric module reliability on application conditions. By following some basic guidelines, and with knowledge of how certain factors tend to affect module life, it should be possible for designers to optimize system reliability. While some may wish to perform a comprehensive analysis and model all relevant parameters, many users having unusual requirements or nontraditional configurations often turn to an empirical approach for determining the reliability of their specific application.