Proving and improving electronics reliability

TWI Bulletin, September - October 1997

Nihal is a Principal Consultant in the Advanced Materials and Processing Department of TWI, providing technical skills in electronics technology and reliability.

In the electronics industry the length of time between product innovation and market is rapidly shrinking. Nihal Sinnadurai describes the increasing importance of accelerated reliability testing within a product's life cycle.

Reliability testing is one method used to help determine and improve component or equipment reliability. Today, there is strong interest in reliability development for applications in severe climates eg tropical launch sites of spacecraft, aerospace applications (Fig.1), rural applications of telecommunications, global location of oil rigs (Fig.2), and global use of cars and trucks, for which valid methods of modelling reliability are essential in order to develop technologies and products capable of meeting the requirements. Such high reliability has already been engineered for telecommunications systems by co-operative development between users and manufacturers of the materials and components.

Different types of testing will provide information on the different reliability regimes:

• infant mortality (early life failures due to induced faults or fabrication weaknesses)
• random failures (process related failures throughout life)
• wearout (the final cause of end of life)

Both infant mortalities and random failure rates benefit from good quality control practices. Advanced statistical process control (SPC) methods at the component level result in a high yield and improvements in random failure rates. Improved automation control and use of in-line vision systems in packaging board level assembly result in improved product quality and lowered infant mortalities. However, wearout failures are more dependent on technology choices and design rules. For example, plastic encapsulation by silicone applied to a fully activated clean silicon integrated circuit (IC) surface can safeguard the reliability of the IC for more than 100 years in harsh tropical climates. ^[1,2]

Over the four decades since invention of the IC, technology has progressed dramatically and has led to the development and verification of reliability, especially for applications in severe environments. Many thousands of millions of device hours of reliability testing have been accumulated, which has established undisputed and disputed models for the mechanisms of failure of electronics components.

Probably the most authentic and proven method of accelerating ageing of semiconductor component wearout is the use of elevated temperature, which is correlated with normal ageing according to the Arrhenius equation for acceleration of ageing. For environments which also include high humidities, a number of alternative models have emerged. These have varying degrees of authenticity depending on whether they are based on purely statistical correlation or supported by physical and chemical analyses of the failures and appropriate diagnostics. Reliability development for operation in severe climates necessarily makes use of highly accelerated stress testing ^[3] (HAST). For instance, the S-H (Sinnadurai-HAST) model provides authentic acceleration of ageing within identified limits of validity. Ignorance of valid scientific models has led to non-valid use of HAST.

TWI is actively engaged in researching and developing the methods of highly accelerated reliability testing (HART) to serve modern industry which requires rapid feedback. This work will validate designs and assemblies in a wide range of application environments.

The benefits of using the S-H expression have been realised in applications in tropical and subtropical climates where the hostile climates led to moisture leakage into MIL-STD-883 qualified hermetic packages. Consequently the components failed due to severe corrosion of the metallisations with failure of some 10 000 devices over five years resulting in a failure rate of 1755 FITs. (One FIT=one failure in 10 ⁹ hours). Reliability verification of alternative commercial plastic packaged ICs using HAST tests and the S-H expression led to tremendous improvements in system design and reliability as well as cost, resulting in failure rate improvements by more than an order of magnitude.

Accelerated ageing to prove and improve reliability has traditionally been undertaken by applying thermal storage (dry heat without bias) thermal overstress (dry heat with bias) and damp heat storage and damp heat overstress (including HAST). However, two important aspects of reliability testing are often overlooked when carrying out so-called reliability or accelerated ageing tests, namely knowing the 'normal operating environment' and knowing the 'validation limits' of the accelerated ageing model. Both of these are addressed below.

Accelerated Ageing Models

Equations that describe accelerated ageing due to overstress are as follows:

The effect of temperature on the ageing of electronics components including plastic packaged devices has been shown to conform to the well established Arrhenius equation which translates to:

t _amb/t _s = exp { (E _A/k) (l/T _amb - 1/T _s)} (1)

in which t _s is duration of the stress test, t _amb is the required lifetime at ambient temperature T _amb, T _amb and T _s are the absolute temperature, subscripts 'amb' and 's' refer to 'ambient' and 'stress' respectively, E _A is the activation energy for the specific failure mechanism, and k is the Boltzmann constant.

Where the applications or transportation environments also contain moisture, then moisture will not only permeate plastic packages but will also leak into MIL-STD-883 qualified hermetic packages. The acceleration of ageing by nonsaturated vapour is given by the following S-H equation: ^[3-5]

t _amb/t _s = exp {X [(RH _s) ⁿ - (RH _amb) ⁿ] + (E _A/k) (l / T _amb - 1 /T _s)} (2)

in which RH _amb is the ambient humidity (determined for each application environment), and n is an empirical constant empirically determined for each type of technology (eg n=2 for ICs and n=1 for thick film resistors).

The equation has been verified by extensive tests conducted from 1968 through to 1990 on integrated circuits, hybrid microcircuits and thick film circuits and supported by science of failure models and failure mechanisms studies. ^[3-5] Such science of failure models are naturally bounded by the range of stresses over which they are modelled, unlike statistically derived expressions which do not take account of the realities that a vapour that changes its state behaves dramatically differently.

When applied to packaged integrated circuits, substituting 175 000 hours (20 years) for t _amb, and using experimentally determined data, the S-H expression becomes:

t _s=175000/exp {0.00044 [(RH _s) ²- (RH _amb) ²] + 7000 (1/T _amb - 1 /T _s)} (3)

Validating the Environments and the Models

Because RH _s, and T _s are the applied stresses and are therefore known, the remaining unknowns in equation (3) are RH _amb and T _amb. These must be determined for each location and application. Climatic conditions for sheltered but otherwise uncontrolled climates in temperate and tropical regions, averaged over 10 years, have therefore been assessed ^[6] as shown in Table 1.

Table 1: Standard environmental conditions

The stress condition applied in the most comprehensive humidity and HAST tests, conducted over 20 years at BT Labs, was 108°C, 90%RH, in a non-saturating autoclave, because of the chosen HAST pressurisation of two atmospheres. The HAST test gives more severe, but more dependable damp heat acceleration of ageing than is obtainable with humidity chambers ( Table 2, Fig 4 and 5).

The HAST equipment may be pressurised to many atmospheres, and enables non-saturating humidity testing to be conducted up to much higher temperatures than the initial pioneering work. Hence, the test durations for THB (temperature + humidity + bias) overstress tests to simulate 20 years (175 000 hrs) operation in various climates has been calculated (Table 2).

Table 2: THB reliability tests for 20 year survival in different climatic conditions

The calculations which have also been proven by extensive experimental evidence, show the tremendous advantage of HAST over conventional humidity testing. The S-H model has been verified by humidity chamber and HAST testing, from low relative humidities and temperatures at 40°C and 50%RH to beyond 150°C and 95%RH. From such work, the limit of validity (ie validation limit), for genuine acceleration of ageing, was found to be 130°C and 95%RH. Beyond this, the adsorption isotherms no longer obey an RH exponent, and some polymers exposed to high humidities above 130°C are found to de-bond. Consequently, calculations for the above tabulation were extended up to 125°C and 95%RH, but no higher. Even at this level of accelerated ageing, some test durations are decreased below 50 hours. This is the recommended minimum for humidity testing, to allow for induction time (ie time for adsorption followed by permeation or percolation to the active areas) before the onset of failure mechanisms.

Thick film resistors in microcircuits are also accelerated by HAST conditions according to a modification of the generalised S-H expression as follows:

t _s=175 000/exp{0.025 [X[ (RH _s) - (RH _amb)] + 8120 (1 /T _amb - 1 /T _s)} (4)

Another expression for accelerated ageing by moisture is that of Peck ^[7] (a variation of the S-H RHn model) formulated by regression analyses of published data, including reliability results from tests conducted in saturating pressure cookers. The Peck version is:

t=A.RH ⁿ•exp (E _A/kT) (5)

in which n was stated to be -2.66 and E _A was 0.79eV.

There is a disparity between the interpretations of Peck and Sinnadurai because the regression by Peck spans both saturated and unsaturated water vapour and thereby inevitably spanning a change of state of water vapour from vapour to liquid. It does not take account of the consequent significant changes in failure mechanisms that arise. There also appears to be no validation constraints placed on application of the expression.

Due to the lack of constraint in many published expressions for humidity testing, there are a number of reliability investigations that have been carried out at excessively elevated temperatures (as high as 159°C and 95%RH ^[8] ). This results in component failures which may not be correlated with normal ageing behaviour. Unfettered acceleration can be extrapolated to such an extent that a few seconds of testing at high temperatures and 100% RH could be considered to represent whole life behaviour, which is clearly unrealistic. This is a good reason to recognise limits to the validity of all accelerated ageing conditions and to reject expressions based purely on statistical manipulation of data.

Ongoing Work

While the above benefits are clear, more research into environments and HART methods is essential as electronics products are increasingly required to be portable for use globally, thereby exposing them to vibration, dust and additional harsh environments. Also, the pressures on electronics product developers means that product life cycles are diminishing and rapid assessment and feedback is essential to enable short times-to-market to be met. It is exactly along these lines that ongoing activity at TWI is aimed.

References