Marcus P. Corrosion mechanisms in theory and practice

Подождите немного. Документ загружается.

Al(OH)

]. Most of the metals used in microelectronics are susceptible to electrolytic

dissolution.

Cathodic Alkalization (Cathodic Corrosion) This degradation mechanism

is specific to active/passive, amphoteric metals such as aluminum. In the absence

of chloride or other aggressive contamination, the leakage current through the

conductive path causes anodization (oxide buildup) at the anode instead of

localized dissolution. The potential drop across the oxide will consume part of the

applied potential and tend to reduce the leakage current. However, while current

is flowing, proton and water reduction reactions will take place at the cathodically

biased line, causing an increase in local pH by the following reactions:

+ 2e

–

→ H

or 2H

O + 2e

–

→ 2OH

–

. Because the metal is amphoteric,

the increase in pH results in dissolution of the protective oxide on the surface of the

cathodic line. The metal can then dissolve rapidly, even with a net negative current

at the surface. Analogous to the case of electrolytic dissolution at anodic lines, devices

fail by an open circuit, except now on the cathodic line (Fig. 4c). Similarly, a

corrosion product such as Al(OH)

will precipitate near the dissolution region.

Electrolytic Migration In this degradation mechanism, metallic ions dissolved

from the anode (electrolytic dissolution) are transported through the surface

electrolyte film and are electrolytically reduced at the cathode [45]. Metal

dendrites grow from the cathode toward the anode and lead to failure via a short

of the adjacent conductors (Fig. 4d). Important factors that affect dendritic

growth include (1) properties of the metal (some metals such as Ag and solder

readily electroplate with a dendritic morphology), (2) local current densities that,

if high enough, deplete cations in the surface electrolyte film in the vicinity of the

cathodic lines, and (3) mass transport–limited conditions that promote the

stability and propagation of dendrites. Notably, metals such as aluminum that

have a deposition potential that is outside the stability limit of water are not

vulnerable to this mechanism. Steppan et al. [45] listed the order of susceptibility

to electrolytic migration for many commonly used metals: Ag > Mo > Pb >

solder > Cu > Zn > bronze. A small volume of metal dissolved from the anode

can result in comparatively large dendrites, so the lines may short long before

an open would have formed in the anodic line [45]. As dimensions of micro-

electronics shrink, this failure mechanism becomes more important. Warren,

Wynblatt, and co-workers performed a series of studies on electrolytic migra-

tion of Cu [47–51]. Their work, which developed a good understanding of the

phenomenon, showed that contamination greatly accelerates electrolytic migra-

tion. The phenomenon of electrolytic migration should be distinguished from

another failure mechanism with a similar name, electromigration, which is not

electrochemical in nature [45].

Open Circuit (without Electrical Bias)

Although deleterious environmental interactions normally occur in microelectronics

under electrical bias, more traditional atmospheric corrosion mechanisms

certainly do take place in unbiased regions during the powered-off state and/or

during storage. As noted previously in this section, the power-off condition for ICs

may actually be more aggressive in some circumstances because a greater amount

of water may exist at the metal surface. Relevant mechanisms, with examples given in

654 Frankel and Braithwaite

parentheses, include uniform attack (Cu contacts), pitting (Al lines and bondpads),

crevice corrosion (Al under wirebonds and under passivation layers), and intergranular

corrosion (Al bondpads). Under these unbiased conditions, chemical incompatibility

is the primary driver. Given the wide variety of metals that are present in intimate

contact with each other (e.g., Au/Al wirebonds), all of these mechanisms can be

enhanced in specific configurations by galvanic interactions. To help characterize the

potential galvanic effects, Griffin and co-workers [52] developed a galvanic series

specifically for the common thin-film metallization and barrier layers used in

microelectronic devices. The corrosion potential in 2000 ppm NH

C1 solution

decreased in the following order: Au-40%Pd, Au, Ag, Cu, Al-0.5%Cu-l%Si, CVD W,

W-10%Ti, Al-2%Cu-l%Si, Al-2%Cu, Al-0.4%Pd-l%Si, Si, Al-l%Si on CVD W, Al

on CVD W, Al on W-10%Ti, Al, Mg.

These traditional corrosion mechanisms are described in detail in standard

corrosion textbooks [53] and elsewhere in this book. However, care must be

exercised during the application of this conventional understanding to

microelectronics because of their very small dimensions. This caution can be

demonstrated by considering the atmospheric pitting of aluminum. A standard

technique for characterizing the atmospheric corrosion susceptibility of Al

involves a salt fog test (ASTM B117-90). An associated military specifications

(mil-spec) (Mil-C-5541E) states that no more than five pits can exist on a panel

and normal practice is not to count a pit that is less than 125 μm in diameter. As

noted previously, many IC feature are < 1 μm. Clearly, even small metastable pits in

the wrong place can be disastrous.

Device-Specific Corrosion Behavior and Concerns

Integrated Circuits

Aluminum Metallization Because of the predominant use of aluminum alloys

in integrated circuits and their vulnerability to atmospheric corrosion, the majority of

the environmentally induced problems that have historically occurred have involved

aluminum. As introduced in the section on contamination, corrosion has been

observed even during manufacture due to exposure to aggressive processing

chemicals. During the early phases of the microelectronic industry, yield loss from

in-process corrosion was a costly problem. Chemical and RIE processes have been

the most damaging. The preferred use of Al-Cu alloys exacerbates the problem

because Cu is enriched in the subsurface region during RIE due to the lower volatility

of Cu chlorides compared with Al chlorides [36]. At the high etching

temperatures, Al

Cu θ-phase particles precipitate and accelerate the subsequent

corrosion of the Al matrix. Several postetch process steps have been developed

to reduce corrosion, including CF

plasma cleaning, O

plasma cleaning, DI

water rinsing, wet etching, and heat treatments [30–36]. Brusic et al. [30] evaluated

the efficacy of several of these cleaning steps. By simply rinsing in water before

stripping the photoresist, the impurity level and associated corrosion rate were

reduced by more than two orders of magnitude compared with parts cleaned only

with an O

plasma. Elimination of CHCl

from the RIE gas in the last step resulted

in still lower corrosion rates and, unlike those with etching in CHCl

, the

properties did not degrade during subsequent storage in air. Any in-process

Corrosion of Data Storage Devices 655

corrosion or staining that does take place on bondpads during manufacture can

adversely affect long-term device reliability because the integrity of the subsequent

wirebond can be reduced [54].

During service, biased Al metallization features most commonly fail at

defects or breaks in the passivation layer (upper right photograph in Fig. 4). If

chloride contamination is present, electrolytic dissolution occurs in the positive

(or anodically) biased lines [55–57]. As implied in this photograph, the production of

corrosion products causes more passivation layer defects and, in turn, more

corrosion. As with pitting corrosion of aluminum, the presence of a halide is very

important because it induces breakdown of the Al oxide film and permits rapid

dissolution. Biased Al metallization that is not exposed to chloride-contaminated

environments may still exhibit extensive corrosion and fail by cathodic corrosion

[58–64]. This phenomenon is promoted for structures having a top passivation

layer of phosphosilicate glass (PSG) with a high P content. The addition of P to

SiO

improves coverage by lowering the stress in the oxide and reducing cracks.

However, cracks along the sidewall and lack of coverage at steps and pinholes can

still occur. When PSG containing more than about 5% P is exposed to moisture,

phosphate dissolves to form a highly conductive surface layer. Cathodic corrosion is

promoted because of the high surface conductivity and the absence of chloride ions.

The anodic aluminum anodizes rather than dissolves. The activation energy for

cathodic corrosion has been found to be similar to that for dissolution of Al in an

NaOH solution [62]. Other ions besides phosphate can cause this form of corrosion,

even those that do not form a strong base [62]. The effect of P in PSG was identified

in the early 1980s and has been addressed by keeping the P content low or by using

silicon nitride passivation, which has been found to be more protective [65–67].

Even in the absence of an applied bias, galvanic-induced electrochemical

driving forces for corrosion can exist because unpassivated wirebonds often contain

couples of Al and Au [57,68]. In the presence of moisture and chloride contamination,

the galvanic driving force due to the Au can be sufficient to cause pitting

and inergranular attack of Al and eventually produce device failure. Several

researchers have studied the corrosion of Au/Al wirebonds [69,70]. Recent work

at Sandia National Laboratories has focused on galvanic corrosion under dormant,

unpowered storage conditions and has specifically been exploring the role of Al-Au

intermetallic compounds that form during encapsulant curing operations. This study

has shown that, under mildly accelerating conditions, unbiased Al bondpad corrosion

occurs only in the presence of an Au wire and that it initiates under the wirebond

and propagates along an Al/intermetallic interface (Fig. 5). Also, considerable

variability exists in the distribution and structure of the intermetallic phases (Fig. 5).

This latter finding suggests that the often-observed stochastic nature of corrosion

may actually be the result of manufacturing process variability. If the galvanic

influence of Au is not present, θ-phase particles in Al-Cu alloys can still initiate

pitting or intergranular corrosion on an unbiased surface via local galvanic action

[38,71,72]. Pits associated with θ phase may be limited in size but are large

enough to cause failure of micrometer-sized features.

Gold Metallization Anodically biased gold metallization can fail in the

absence of a contaminating salt by electrolytic dissolution if the ionic pathway

contains an effective gold complexing agent. The dissolution is accompanied by

656 Frankel and Braithwaite

Corrosion of Data Storage Devices 657

Figure 5 Corroded aluminum-bondpad metallization layer taken under back-lit conditions. A photograph of the entire test device prior to any

corrosion is shown on the right. The photograph of the corroded structure shows the importance of galvanic interactions and the existence of

crevice corrosion (under the passivation layer) and was taken after underlying silicon and glass insulation layer were etched away.

the formation of a voluminous Au(OH)

. However, examples of this type of failure

have not been widely reported [73]. In the presence of a halide contaminant, failure

can also result from electrolytic migration because gold metallization can form

dendrites [73]. Der Marderosian [74] found a humidity threshold below which no

migration of Au occurred. The threshold value was a function of the type and

amount of purpose applied halide salt contamination but ranged from about 30 to

50% RH.

Plastic Encapsulated Microelectronics Because the plastic encapsulant

materials are relatively permeable to water, chip metallization in PEM devices is

vulnerable to corrosion when exposed to or operated in humid environments. Nguyen

and colleagues [75–77] used an in situ capacitance monitor to study moisture

permeation in plastic packages [5]. The uptake and transport of moisture in epoxy

were Fickian in nature and exhibited a diffusivity of about 2×10

/s, meaning that

water can penetrate a polymeric package and reach the metals on the die relatively

quickly. Moisture can also migrate along defects in the lead frame–polymer interface

and condense in undesirable voids near metallization surfaces. Because the bulk

permeability of the plastic to typical external contaminants is very low, these lead

frame defects appear to be a primary contaminant transport path. Historically, the

plastic encapsulants themselves have been a source of significant contamination, with

liquid/glob materials having higher corrosion-enhancing levels than the injection

molded compounds. Once sufficient moisture and contamination reach the IC

metallization and a void for adsorption/condensation exists, corrosion can occur and

degradation may ensue by any of the mechanisms described previously, including

electrolytic, pitting, galvanic, and crevice corrosion. It should be noted again that

top-level conductor lines are normally under passivation layers that, if impermeable,

would prevent corrosion [78,79]. Historically, defects such as cracks, pinholes, and

inadequate edge coverage were commonly present [55]. During the past 10–l5years,

processes and materials have improved to the point that passivation defects are no

longer a primary concern. Now, the unpassivated bondpads themselves are the most

susceptible.

In practice, the key factors that influence PEM corrosion vulnerability are

probably the defects in the plastic encapsulant, the level of leachable contaminant in

the encapsulant, and the degree of drying during power-on cycles. The first two

factors have been adequately addressed in modern, best commercial practice

devices to the point that passivation detects and residual contamination are

essentially inconsequential. Pecht and co-workers [9,29] have now concluded that

modern, properly fabricated PEM devices are very reliable under a variety of service

conditions.

Ceramic Hermetic-Packaged Microelectronics The meaningful differences

between CHP and PEM devices in the context of corrosion are the following: (a) Al

wires are used instead of Au so that the prime corrosion vulnerability is the

ultrafine 25-μm wire instead of the bondpad, (b) the internal environment can

theoretically be controlled such that a benign environment will always exist, and

as sites for water condensation. To achieve the environmental control advantage,

three factors must be satisfied: (a) a hermetic seal, (b) a clean and dry assealed internal

658 Frankel and Braithwaite

environment, and (c) control of outgassing from internal materials. If they are

properly designed and manufactured, the reliability of CHP devices is not a concern.

However, in practice, all three of these aspects have produced problems as evidenced

by field-returned “hermetic” packages in which corrosion has occurred [80–82].

The specifications on hermetic seals are not always adequate to ensure total

isolation from the environment for extended periods of time. For example, as noted by

Pecht and Ko [83], hermeticity is defined by a maximum leak rate (e.g., 10

–7

atm cc/s).

If the device leaks at this particular maximum rate, the critical moisturecontent (three

monolayers of water) can diffuse through the seal in as little as 2500 h. Supporting this

finding is a study in which 20% of field failures of equipment sited in a tropical envi-

ronment were due to corrosion of interconnects in packages that met the hermeticity

specification [17]. Also, the seals and/or ceramic bodies can crack and create larger

leaks during handling, soldering, or qualification thermal cycling, which permits both

water and contamination to enter easily. If care is not exercised during fabrication,

moisture adsorbed on the inside walls of the package can desorb during subsequent

processing steps or during use. To address this possibility, mil-spec CHP parts are nor-

mally sealed in an atmosphere containing no more than 5000 ppm of water, ensuring

that no liquid phase will exist at temperatures down to 0°C, where ice forms. The seal-

ing glass used in some packages or a popular organic die-attach material can actually

be sources of moisture. Devitrifying glass is one type of sealant with a high moisture

content that is evolved during closure, whereas vitreous glass contains little moisture

and can result in a ceramic package with less than 500 ppm moisture in the cavity [84].

Moisture trapped inside a sealed cavity can leach ions from the sealing glass or other

sources to form a conductive electrolyte. Once an ionic path exists between conduc-

tors, corrosion and failure by any of the mechanisms described previously can ensue.

Finally, one researcher found that residual stresses in Al-containing wirebonds could

be an important factor in failure (e.g., by stress-induced corrosion) [85].

Macro Interconnects

Solder, copper conductors, and plated-copper connectors constitute the major metallic

components of second-level packaging for functional electronic circuits. These

metals are all susceptible to corrosion. Typically, interconnects are less protected and

more exposed to ambient environments. However, their dimensions are much larger

than those of chip metallization and more corrosion can therefore be tolerated before

failure is produced.

Solder PbSn alloys, ranging from Pb-rich to Sn-rich compositions, are the

most solder materials used in electronic applications. Pure Sn forms a protective

oxide film, but Pb forms an oxide that is not stable and can easily react with

chlorides, borates, and sulfates [86]. Frankenthal and Siconolfi [87,88] found that

both Sn- and Pb-rich PbSn alloys form an Sn oxide, most likely SnO, during the

initial exposure of oxide-free surfaces to oxygen. Lead is oxidized on the surface

of PbSn and a mixed oxide results only after all the metallic Sn is totally depleted

from the surface. Similarly, the corrosion product formed on Pb-50In solder during

exposure to an aggressive gaseous environment was found to be rich in In [89].

Not surprisingly, the corrosion resistance of PbSn in various aqueous solutions

and gaseous environments depends strongly on the alloy composition, improving

greatly as the Sn content increases above 2wt% [86].

Corrosion of Data Storage Devices 659

In the presence of moisture and contamination, the most common degradation

mechanism of solder is electrolytic migration (shown previously in lower right

photograph in Fig. 4). Thus, the cleanliness of fabrication steps and the effective

removal of flux residues are critical factors. Some IC package types (e.g., surface

mount) are very hard to clean effectively. Manko [43] summarized the corrosion

behavior of PbSn solders in various chlorides that are typically used as activators in

fluxes and noted their deleterious effect in destroying the native passivating oxide

layers. He also noted that corrosion related to flux use might result from flux or flux

residue that is trapped in inaccessible locations or from fumes liberated during

soldering and subsequent condensation in uncleaned regions. Specifically, water-

white rosin flux has been found to leach Sn from the solder and therefore decrease its

corrosion resistance [90]. Soldering traditionally uses nonaqueous fluxes that require

cleaning with Freon-based solvents. In an effort to reduce chlorofluorocarbon

usage, water-soluble fluxes have been substituted in some applications. Cleaning of

water-soluble fluxes can normally be accomplished in deionized water. Sn-Pb-In

solders corrode in water-soluble fluxes containing chloride-based activators but are

not attacked in fluxes containing phosphate-based activators [91]. Finally, in-process

corrosion can result in poor quality solder joints or decreased thermal fatigue life.

This latter result is especially significant for C4 solder connections [90].

Copper Conductors Although precautions are taken to protect the copper

conducting lines in printed wiring boards from environmental exposure, defects exist

and corrosion does occur under many atmospheric conditions. For most

environments that are encountered, copper is not thermodynamically stable.

However, its native cuprous oxide surface film does offer some limited protection.

Atmospheric corrosion of copper is briefly described in the chapter on atmospheric

corrosion. Examples of typical industrial atmospheric pollutants that are harmful to

copper include SO

, H

S, COS, NO

, Cl

, and CO

Contamination of copper surfaces with atmospheric particulate matter can

accelerate the corrosion process. For example, the corrosion of copper in 100°C

air in the presence of submicrometer ammonium sulfate particles has been found

to depend strongly on RH [40]. Below the critical RH, Cu

O formed uniformly on

the Cu surface. At the critical RH of 75%, Cu

(SO

)(OH)

formed in the region

where the (NH

)

particles had been deposited. At 85% RH, a thick corrosion

product covered the entire surface. Particulate contamination such as this is typically

believed to be an issue during manufacturing. However, it can also cause field

failures in electronic installations, such as telecommunications centers [92].

The formation of conductive anodic filaments (CAFs) has been observed

during accelerated aging testing to cause failure in epoxy-glass printed circuit

boards containing copper conductors, a process clearly related to electrolytic

dissolution [93]. Here, a conductive ionic path consisting of precipitated corrosion

products presumably forms between the oppositely biased conductors. These fibres

grow from the corrosion-producing anodic line to the cathodic line. The degradation

was found to be most severe between two nearby through-holes. Because CAF has not

been observed in field-returned parts, it may be only an artifact of the test conditions.

Au-Plated Connectors and Contacts As noted earlier, the subtrates of

connectors and contacts are normally fabricated from copper or copper alloys and

660 Frankel and Braithwaite

then plated with a noble metal. The presence of the thin plating that almost always

contains some level of physical defects (e.g., cracks and pores) enhances the

corrosion susceptibility of the substrate and enables other processes to take

place. Slade [22] and Abbott [23] recently published a comprehensive review of

connector/contact corrosion. One form of observed degradation occurs when

corrosion products creep across a noble metal surface. When Au-plated Cu is

exposed to a sulfide-containing atmosphere, a tarnish film composed of copper

oxide and copper sulfide forms in the plating pores and cracks where the Cu layer

is exposed [94,95]. With time, a predominately copper sulfide tarnish film migrates

over the Au surface, resulting in increased contact resistance. This phenomenon,

termed creep corrosion by Tierney [94], is becoming increasingly important as the

thickness of the gold plating is reduced for economic reasons and as the edges of

exposed Cu become closer to contact points as a result of miniaturization. A second

relevant degradation mode is often referred to as pore corrosion and simply consists

of accelerated corrosion of locally exposed regions of the substrate. Two important

processes are probably responsible: (a) capillary condensation of water in cracks,

pores, and between mated surfaces and (b) galvanic interactions with a potentially

large cathode-to-anode area. Ming et al. [14] showed that contact force and

electrical loading can also affect corrosion. For example, contact force can

improve performance in stationary contacts and applied voltage can accelerate

growth of surface films and decrease contact service life. Finally, in recent years, a

few special contact lubricants have been developed that can also inhibit corrosion [23].

Product Qualification/Reliability Testing and Analysis

A significant effort has been made over the past quarter century to develop effective

techniques to test and assess the effect of corrosion on the performance and life of

electronic devices. Such techniques are desired for two specific uses: (a) as a means

of rapidly characterizing product quality during manufacturing (acceptance testing)

and (b) to provide customers and users with an accurate estimate of expected service

life. In this context, corrosion degradation is primarily characterized in terms of

reliability, which is the probability that a device will not perform its intended

function (i.e., electrically fail). Such device-level studies are required because the

direct measurement of corrosion is very difficult or even impossible in actual

operating environments using standard corrosion characterization techniques (e.g.,

electrochemical and gravimetric). A device manufacturer wants to be certain that a

product will have at least a certain functional lifetime in service with a minimum

number of failures. A typical historical requirement is 100 failures in 10

device hours

(100 FIT) or 0.01% failure in 1000 hours [96]. In modern practice, most electronic

devices are expected to function reliably for at least 10 years [17], although

obsolescence can certainly set in much sooner. Therefore, reliability engineers are

forced to perform accelerated aging exposures on large numbers of parts. For general

reference, Nelson [97] has published a comprehensive overview of accelerated

aging, and Osenbach [17] describes factors and issues associated with accelerated

aging of microelectronics. A clearly recognized deficiency of the reliability

approach is that such tests do not permit a fundamental understanding of the corrosion

mechanisms and processes to be easily identified. Truly predictive reliability

Corrosion of Data Storage Devices 661

assessments must have this type of physical basis. Nevetheless, accelerated reliability

testing is of tremendous practical importance because susceptible materials, design

flaws, and processing problems can be identified in a timely manner and a qualitative

understanding of the degradation processes can be generated.

The underlying principle behind the use of accelerated aging is that the

functional behavior of real devices or the response of test structures with similar

materials and dimensions can be characterized using various combinations

of environmental stress factors that include temperature, humidity, bias, and

contamination. Each of these variables is typically used at levels more severe than

actual operating conditions. Osenbach also notes that all of the accelerated aging

strategies include two significant assumptions: (a) the parameters that characterize

failure under high-stress conditions (where failure occurs in short time periods of

days to months) can be extrapolated to actual use conditions where failure occurs

after years of exposure, and (b) the test population is representative of the entire

population. This section contains brief descriptions of (a) the aging techniques

commonly being used, (b) resultant models of acceleration factors, and (c) the

concerns and potential deficiencies associated with this subject.

Techniques

Microelectronics Peck and Zierdt [98] led the way for adopting elevated

humidity and temperature as the primary stress factors. The most common

accelerated test used over the past 25 years is referred to as THB (temperature,

humidity, and bias). In this test, parts are exposed for extended periods (> 1000 h)

at 85°C, 85% RH, wih the conductor lines biased at, or sometimes above, the

operating voltage [68,99]. As PEM device reliability improved during the 1980s

due primarily to a reduction in residual chloride contamination and molding

compound cleanliness, exposure times under THB conditions required to generate

failures started approaching 10,000 h. In order to formulate predictive relationships,

failure during accelerated testing must be observed. The extremely long times

needed in conventional THB tests led to the development of tests with more severe

conditions to reduce needed test time. A pressure-cooker test at 100% RH and

temperatures above 100°C has been used for this purpose [99–102]. Its efficiency

can be demonstrated by the observation that for one specific device, results

equivalent to those obtained in 1000 h at 85°C, 85% RH could be produced in 20 h

at 140°C, 100% RH [99]. However, artifacts associated with condensation and

difficulties in applying voltage in saturated atmospheres complicated widespread

adoption of this procedure. Because of attractive equipment cost and availability,

many manufacturers still use this type of testing and the standard conditions are

121°C, 100% RH and no bias. More recently, the highly accelerated stress test

(HAST) has become the standard. HAST tests typically use temperatures up to

150°C and humidity levels less than 100% [63,103,104]. Several other variations

of the standard THB aging techniques have been developed that involve additions

of internal and external contamination and/or cyclic application of the environmental

parameters to better simulate actual use [28,105–109]. Failure analysis techniques

to supplement the reliability information have also been reported [110].

Although most accelerated aging studies are performed on actual functional

devices and the time to failure is directly measured, some studies have used test

662 Frankel and Braithwaite

structures designed specifically to better address particular degradation modes.

For corrosion under electrical bias, common configurations include two interdigitated

combs, a triple track with three parallel but meandering lines, and combinations of

tracks and bondpads/wirebonds [8,65,66,103,111,112]. Example of triple-track

structures are shown in Figure 6 [8]. The two outer track conductors are normally

biased positively and the center one is negative or grounded. The portion of the

triple-track structure in Figure 6b is contained within the test device previously

shown in Figure 5. To assess wirebond degradation, the design permits conventional

four-point resistance measurements to be made. Such test devices can be tested

bare or as packaged.

Macro Interconnects Because connectors and contacts are usually exposed

without protection to the environment during operation, testing typically has been

performed in the context of atmospheric corrosion [113]. Various forms of accelerated

tests for indoor office environments have been developed for this purpose

[26,114–117]. For example, procedures have been developed at IBM [the G1(T)

test] [26,114,117], Battelle Institute (Flowing Mixed Gas Class II or FMG II)

[23,115,116], and in Europe by the International Electrotechnical Commission

(IEC Test 68-2-60) [118] for simulating an indoor office environment. These

techniques vary slightly but involve exposure to an atmosphere containing a

combination of dilute pollutants such as H

S, SO

, NO

, and Cl

and humidity

(Table 1) and are usually performed at near-ambient temperatures because of

complications arising from reactions between the various components at higher

temperatures. Acceleration factors from 10 to 1000 are possible depending on the

chosen conditions [115]. These aging environments are now used to test and qualify

many electronic parts other than connectors and contacts, including printed circuit

boards. The reliability of PCBs is also routinely tested using THB testing as discussed

in the previous paragraph.

Mathematical Relationships for PEM Aging

When plastic-encapsulated microelectronic devices are exposed to aggressive

environments, there is often a bimodal distribution in the time to fail resulting from

Corrosion of Data Storage Devices 663

Figure 6 Schematic diagram (left) and an SEM photograph (right) of two triple-track test

structures that are used to study electrolytic corrosion mechanisms along with the

effectiveness of passivation layers. The structure shown on the right was encapsulated and

then exposed to HAST conditions until failure. This particular structure is part of the integrated

test device shown previously in Figure 5. These test devices contain eight triple-track sections

(the left set with windows in the passivation layer) and exposed wirebond pads.