Searching for appropriate humidity accelerated ... - Semantic Scholar
Microelectronics Reliability 42 (2002) 1213–1218 www.elsevier.com/locate/microrel
Searching for appropriate humidity accelerated migration reliability tests methods q P. Bojta, P. Nemeth, G. Hars anyi
*
Department of Electronics Technology, Budapest University of Technology and Economics, H-1521 Budapest, Hungary Received 31 January 2002; received in revised form 4 April 2002
Abstract Lifetime estimation is one of today’s frequently used reliability testing tool. There are standardized test methods, mathematical apparatus and failure acceleration models for predicting failure rate of electronic components, circuit modules or equipment. Do the widespread used models give a precise description for all kind of failure mechanisms? Can they define the acceleration factor of any test, and the accelerated lifetime of any test vehicles? In connection with open surface migration tests (when the test circuit samples are not covered by any protective or packaging material) two climatic test methods (Thermal Humidity Bias––THB test methods with different parameter settings) have been compared: 40 C/95%RH suggested by ‘‘a well known IEC standard’’ and 85 C/85%RH required by the newer JEDEC standard. The novelty of the paper is the comparison between the two climatic test methods. The main conclusion is a suggestion to keep on with the old method in the mentioned particular case, which may be shocking for people who prefer the new standards. All conclusions are strengthened both with theoretical and experimental test results. 2002 Elsevier Science Ltd. All rights reserved.
1. Introduction 1.1. Electrochemical migration: one of the most frequent failure modes in THB testing Recently, in connection with the production of highdensity interconnection systems in integrated circuits and multichip modules (MCMs), the demand for conductor-systems with very high resolution and high reliability has emerged. The possibilities of integration are determined not only by the technological bases but also by those physical and chemical processes that can cause resistive shorts between adjacent metallization stripes during the operation. One of these phenomena is the electrochemical migration. This can be defined as a transport of ions between two metallization stripes
q
Supported by the following projects in Hungary: OTKA T030574, OM FKFP 0300/1999. * Corresponding author. Tel.: +36-1-463-3634; fax: +36-1463-4118. E-mail address: [email protected] (G. Harsanyi).
under bias through an aqueous electrolyte: the metal ion deposition at the cathode forms dendrites or dendritelike deposits [1]. Ultimately, such a deposit can lead to a short circuit in the device and can cause catastrophic failure. The preliminary conditions are a film of polar liquid (usually water) to form an electrolyte, bias, and operating time. It must strongly emphasized at this point that migration can only occur when a continuous moisture film had preliminary condensed onto the substrate surface or into the dielectric film between the metallization stripes under bias. Moisture condensation happens when the humidity exceeds a critical threshold value which is characteristic for the surface; it is near 100% RH (dew formation conditions) at clean, non-hydrophilic surfaces, but may be at much lower humidity levels for example at contaminated surfaces (75% with NaCl, 69% with CuCl2 contamination) [2]. Silver is well known for its inclination to form migratory shorts. Studies on silver migration have been made by many researchers among whom Kohman et al. [1] investigated it very intensively and gave the first explanation of the mechanism. They have defined it as a process by which silver is anodically dissolved from its
0026-2714/02/$ - see front matter 2002 Elsevier Science Ltd. All rights reserved. PII: S 0 0 2 6 - 2 7 1 4 ( 0 2 ) 0 0 0 8 3 - 5
1214
P. Bojta et al. / Microelectronics Reliability 42 (2002) 1213–1218
initial location and is redeposited as metal at a cathodic site. Hþ ions collect around the cathodic site where their discharge maintains the electrical balance: 2Hþ þ 2e ! H2
ð1Þ
At the anode, metal dissolves forming Menþ ions: Me ! Menþ þ ne
ð2Þ
or compare the lifetime and reliability behavior of the electronic systems. Almost all processes of nature follow the Arrheniustype exponential thermal acceleration; thus, an elevated temperature will result in a faster analysis and lifetime data for normal operating temperatures can be estimated by applying the Arrhenius equation [17]: ACCFðT Þ ¼ exp½ðEa =kÞð1=Tu 1=Ta Þ
ð4Þ
where Me represents the metal, Menþ the metal ion with n valence, as well as e : the electron. Metal ions then may combine with H2 O or OH ions to form oxides or hydroxides. These compounds may remove ions for the solution by forming insoluble precipitates or form a passivating layer that stops the anodic dissolution. If the latter is not prevented, metal ions migrate toward the cathode, where their electrochemical deposition occurs:
where ACCF is the acceleration factor, Ea the activation energy (Ea ¼ 0:5–1 eV at migration type phenomena), k the Boltzmann-constant, Tu the temperature at normal use conditions and Ta the temperature at accelerated conditions. The situation is much more complicated with humidity. Great efforts have already been made to find humidity acceleration factors. Today the Peck’s model seems to be the most appropriate [13,14,22]:
Menþ þ ne ! Me
ACCFðRHÞ ¼ ½ðRHa Þ=ðRHu Þx
ð3Þ
The reason why silver is more susceptible to migration than other metals is that it is anodically very soluble and its precipitates have also good solubility. The anodic dissolution behavior can be reduced by alloying the silver with palladium (Pd) and platinum (Pt) [3–6]. At special ternary Pd–Pt–Ag compositions, the migration can practically be eliminated [6]. The description and comparison of the various metals can be found in more details in the literature [7–12]. Returning to the theoretical model, the generalized process consists of the following steps: 1. Moisture film formation: water molecules penetrate the package materials and/or condense onto the circuit surfaces. 2. Metal-ion formation by an anodic corrosion, which may either, be direct electrochemical dissolution or a multistep electrochemical–chemical process resulting in precipitates as well. 3. Metal-ion migration through the electrolyte under electrical field toward the cathode. 4. Electrochemical metal deposition at the cathode forming dendrites or dendrite-like metallic deposits, which are growing toward the anode and may result in short circuit bridges when reaching it. 1.2. Failure rate acceleration model of thermal humidity bias testing Environment simulating climatic tests, which means exposing the circuits to high temperature/high humidity conditions, plays an essential role in reliability predictions. The main purpose is to accelerate the spontaneous processes with forced environments in order to estimate
ð5Þ
RH is the relative humidity (RH) in %; x is an empirical extension between 2.5 and 5. (RHu the RH at normal use conditions, RHa the RH at accelerated conditions.) Available models assume silently that during thermal humidity bias (THB) test failure processes induced by temperature and by RH are fully independent and therefore ACCFðT ; RHÞ ¼ ACCFðT Þ ACCFðRHÞ. Although some arguments may be emphasized for a more integrated mathematical model, in which the functions are not totally separated, the main goal of this paper is not to make any discussion over the mentioned assumption, which is used widespread in the literature [13–19]. These type humidity acceleration factors can generally describe rather the moisture penetration processes, but not the moisture condensation. The empirical extension is intended to describe the casing materials’ parameters and behaviors, such as surface roughness, porosity, cleanness, permeability, etc. Permeability can be calculated using Fick’s diffusion equations, but the model of non-linear diffusion gives more precise description of this phenomenon [20]. There is another possibility to make more precise model including nonsteady-state calculations, because reaching the steadystate moisture distribution can take as long time as the test itself. Special limits are available in connection with corrosion–migration type processes that always start when moisture condensation also takes place. Under the threshold humidity level, no migration occurs even for a very long time [2]; this type of behavior cannot be described by the above equations. Some statements, however, are well known from the practice: the likelihood of moisture condensation depends on the following parameters:
P. Bojta et al. / Microelectronics Reliability 42 (2002) 1213–1218
• The RH in the environment. • The temperature difference between the condensation surface and environment. • The pressure. • The surface roughness, porosity. Climatic tests are generally more suited for estimating lifetime data of appliances and subassemblies. With open surface devices, the effect of the surface quality in the water adsorption/absorption/condensation processes and humidity fluctuations may influence the migration test results significantly. There are various metallization types, pure metals as well as alloys, showing very different ability for migration. Their comparison is available mainly through empirical way, for example by performing climatic tests. The results are generally uncertain showing large spreading and they can only be interpreted with difficulties. One of the major problems is that circuits can survive the 85 C/85%RH test and operate many hundreds of hours, but after a short exposure to high humidity conditions, they fail. THB test conditions may also be different. While IEC 68-2-3 conditions (40 2 C and 90–95%RH) had been used for a long time during 1980s, today the application of JEDEC Test Method A101-B become widespread, which requires other conditions (85 2 C and 85
5%RH). Conventional THB tests were performed at 40 C/95%RH conditions and a number of potential failures have been revealed that may hidden in 85 C/ 85%RH tests that are preferred in today’s’ quality standards. The reasons are rather obvious: • The increase of the temperature will accelerate the thermally activated Arrhenius-type processes. • The decrease of the RH will slow down the humidity related processes in the same time. • Solid-state surfaces remain generally clean under 90%RH, since moisture deposition will occur above this level. • At 40 C/95%RH, the dew point on the surfaces may occur even in a temperature fluctuation of one degree, while this needs a much larger range at 85 C/ 85%RH. • The most dangerous realistic natural environmental conditions result in low temperatures and high humidity levels (heavy summer rain, fast autumn cooldown, etc.). Normal ambient conditions were defined as 25 C/ 50%RH. Using this, we can calculate the acceleration factors of the tests according to the Eqs. 4 and 5. The parameters of RH, water damp partial pressure (p), saturated water damp partial pressure (ps ), and dew point temperature (Td ) have the following relationships [21]:
1215
Table 1 Comparison of the properties of the two THB test methods Properties
ACCF(T) related to 25 C, (Ea ¼ 1 eV) ACCF(T) related to 25 C (Ea ¼ 0:5 eV) ACCF(RH) related to 50%RH, (x ¼ 5) ACCF(RH) related to 50%RH, (x ¼ 2:5) Humidity fluctuation necessary for condensation (%) Temperature fluctuation necessary for condensation (C)
Test methods 40 C/ 95%RH
85 C/ 85%RH
6.5
683
2.5
26.1
24.8
14.2
5
3.8
þ5
þ15
0.9
4.1
RH ¼ 100p=ps
ð6Þ
pðTd Þ=ps ¼ 1
ð7Þ
ps ¼ A expðB=T Þ
ð8aÞ
In the case of water damp (between 40 and 85 C): ps ¼ ð9:6 1010 PaÞ expð5130=T Þ
ð8bÞ
A comparison between the two THB methods can be seen in Table 1. The most important conclusions are as follows: • The 85/85 method has higher thermal acceleration factor while its humidity acceleration is somewhat lower. • At 40/95, only a one-degree surface temperature fluctuation (which is within the tolerance range allowed by the IEC 68-2-3 standard: 40 2 C) is enough for moisture condensation. At 85/85, there is a need for 4 C fluctuation, which is out of the allowed range given by the JEDEC Test Method A101-B: 85 2 C. • Thus, when testing free surface migration samples, the preliminary conditions for migration, i. e. the moisture condensation has high probability at randomly located sites during 40/95 tests. This probability is practically very low at 85/85, for which an improper temperature control must be supposed.
2. Experimental In order to have a realistic comparison of the THB test methods in connection with their acceleration effect on the migration process, experimental procedures have
1216
P. Bojta et al. / Microelectronics Reliability 42 (2002) 1213–1218
Fig. 1. Layout structure of the samples used in the climatic tests.
been carried out on Ag–Pt and Ag–Pd thick film metallization systems fired on alumina substrates having a geometry of 4 mm length, line width 0.5 mm, spacing 0.25 mm (see Fig. 1). The reason why this conductor material family was chosen is because it shows the largest migration ability. THB tests with the two abovementioned settings (40 C/95%RH/10VDC as well as 85 C/85%RH/10VDC) were performed on the samples. Contaminated thick film samples were also included, since hygroscopic contaminants can strongly influence the moisture condensation processes. The presence of water-soluble compound impurities may cause the condensation of water at
Searching for appropriate humidity accelerated migration reliability tests methods q P. Bojta, P. Nemeth, G. Hars anyi
*
Department of Electronics Technology, Budapest University of Technology and Economics, H-1521 Budapest, Hungary Received 31 January 2002; received in revised form 4 April 2002
Abstract Lifetime estimation is one of today’s frequently used reliability testing tool. There are standardized test methods, mathematical apparatus and failure acceleration models for predicting failure rate of electronic components, circuit modules or equipment. Do the widespread used models give a precise description for all kind of failure mechanisms? Can they define the acceleration factor of any test, and the accelerated lifetime of any test vehicles? In connection with open surface migration tests (when the test circuit samples are not covered by any protective or packaging material) two climatic test methods (Thermal Humidity Bias––THB test methods with different parameter settings) have been compared: 40 C/95%RH suggested by ‘‘a well known IEC standard’’ and 85 C/85%RH required by the newer JEDEC standard. The novelty of the paper is the comparison between the two climatic test methods. The main conclusion is a suggestion to keep on with the old method in the mentioned particular case, which may be shocking for people who prefer the new standards. All conclusions are strengthened both with theoretical and experimental test results. 2002 Elsevier Science Ltd. All rights reserved.
1. Introduction 1.1. Electrochemical migration: one of the most frequent failure modes in THB testing Recently, in connection with the production of highdensity interconnection systems in integrated circuits and multichip modules (MCMs), the demand for conductor-systems with very high resolution and high reliability has emerged. The possibilities of integration are determined not only by the technological bases but also by those physical and chemical processes that can cause resistive shorts between adjacent metallization stripes during the operation. One of these phenomena is the electrochemical migration. This can be defined as a transport of ions between two metallization stripes
q
Supported by the following projects in Hungary: OTKA T030574, OM FKFP 0300/1999. * Corresponding author. Tel.: +36-1-463-3634; fax: +36-1463-4118. E-mail address: [email protected] (G. Harsanyi).
under bias through an aqueous electrolyte: the metal ion deposition at the cathode forms dendrites or dendritelike deposits [1]. Ultimately, such a deposit can lead to a short circuit in the device and can cause catastrophic failure. The preliminary conditions are a film of polar liquid (usually water) to form an electrolyte, bias, and operating time. It must strongly emphasized at this point that migration can only occur when a continuous moisture film had preliminary condensed onto the substrate surface or into the dielectric film between the metallization stripes under bias. Moisture condensation happens when the humidity exceeds a critical threshold value which is characteristic for the surface; it is near 100% RH (dew formation conditions) at clean, non-hydrophilic surfaces, but may be at much lower humidity levels for example at contaminated surfaces (75% with NaCl, 69% with CuCl2 contamination) [2]. Silver is well known for its inclination to form migratory shorts. Studies on silver migration have been made by many researchers among whom Kohman et al. [1] investigated it very intensively and gave the first explanation of the mechanism. They have defined it as a process by which silver is anodically dissolved from its
0026-2714/02/$ - see front matter 2002 Elsevier Science Ltd. All rights reserved. PII: S 0 0 2 6 - 2 7 1 4 ( 0 2 ) 0 0 0 8 3 - 5
1214
P. Bojta et al. / Microelectronics Reliability 42 (2002) 1213–1218
initial location and is redeposited as metal at a cathodic site. Hþ ions collect around the cathodic site where their discharge maintains the electrical balance: 2Hþ þ 2e ! H2
ð1Þ
At the anode, metal dissolves forming Menþ ions: Me ! Menþ þ ne
ð2Þ
or compare the lifetime and reliability behavior of the electronic systems. Almost all processes of nature follow the Arrheniustype exponential thermal acceleration; thus, an elevated temperature will result in a faster analysis and lifetime data for normal operating temperatures can be estimated by applying the Arrhenius equation [17]: ACCFðT Þ ¼ exp½ðEa =kÞð1=Tu 1=Ta Þ
ð4Þ
where Me represents the metal, Menþ the metal ion with n valence, as well as e : the electron. Metal ions then may combine with H2 O or OH ions to form oxides or hydroxides. These compounds may remove ions for the solution by forming insoluble precipitates or form a passivating layer that stops the anodic dissolution. If the latter is not prevented, metal ions migrate toward the cathode, where their electrochemical deposition occurs:
where ACCF is the acceleration factor, Ea the activation energy (Ea ¼ 0:5–1 eV at migration type phenomena), k the Boltzmann-constant, Tu the temperature at normal use conditions and Ta the temperature at accelerated conditions. The situation is much more complicated with humidity. Great efforts have already been made to find humidity acceleration factors. Today the Peck’s model seems to be the most appropriate [13,14,22]:
Menþ þ ne ! Me
ACCFðRHÞ ¼ ½ðRHa Þ=ðRHu Þx
ð3Þ
The reason why silver is more susceptible to migration than other metals is that it is anodically very soluble and its precipitates have also good solubility. The anodic dissolution behavior can be reduced by alloying the silver with palladium (Pd) and platinum (Pt) [3–6]. At special ternary Pd–Pt–Ag compositions, the migration can practically be eliminated [6]. The description and comparison of the various metals can be found in more details in the literature [7–12]. Returning to the theoretical model, the generalized process consists of the following steps: 1. Moisture film formation: water molecules penetrate the package materials and/or condense onto the circuit surfaces. 2. Metal-ion formation by an anodic corrosion, which may either, be direct electrochemical dissolution or a multistep electrochemical–chemical process resulting in precipitates as well. 3. Metal-ion migration through the electrolyte under electrical field toward the cathode. 4. Electrochemical metal deposition at the cathode forming dendrites or dendrite-like metallic deposits, which are growing toward the anode and may result in short circuit bridges when reaching it. 1.2. Failure rate acceleration model of thermal humidity bias testing Environment simulating climatic tests, which means exposing the circuits to high temperature/high humidity conditions, plays an essential role in reliability predictions. The main purpose is to accelerate the spontaneous processes with forced environments in order to estimate
ð5Þ
RH is the relative humidity (RH) in %; x is an empirical extension between 2.5 and 5. (RHu the RH at normal use conditions, RHa the RH at accelerated conditions.) Available models assume silently that during thermal humidity bias (THB) test failure processes induced by temperature and by RH are fully independent and therefore ACCFðT ; RHÞ ¼ ACCFðT Þ ACCFðRHÞ. Although some arguments may be emphasized for a more integrated mathematical model, in which the functions are not totally separated, the main goal of this paper is not to make any discussion over the mentioned assumption, which is used widespread in the literature [13–19]. These type humidity acceleration factors can generally describe rather the moisture penetration processes, but not the moisture condensation. The empirical extension is intended to describe the casing materials’ parameters and behaviors, such as surface roughness, porosity, cleanness, permeability, etc. Permeability can be calculated using Fick’s diffusion equations, but the model of non-linear diffusion gives more precise description of this phenomenon [20]. There is another possibility to make more precise model including nonsteady-state calculations, because reaching the steadystate moisture distribution can take as long time as the test itself. Special limits are available in connection with corrosion–migration type processes that always start when moisture condensation also takes place. Under the threshold humidity level, no migration occurs even for a very long time [2]; this type of behavior cannot be described by the above equations. Some statements, however, are well known from the practice: the likelihood of moisture condensation depends on the following parameters:
P. Bojta et al. / Microelectronics Reliability 42 (2002) 1213–1218
• The RH in the environment. • The temperature difference between the condensation surface and environment. • The pressure. • The surface roughness, porosity. Climatic tests are generally more suited for estimating lifetime data of appliances and subassemblies. With open surface devices, the effect of the surface quality in the water adsorption/absorption/condensation processes and humidity fluctuations may influence the migration test results significantly. There are various metallization types, pure metals as well as alloys, showing very different ability for migration. Their comparison is available mainly through empirical way, for example by performing climatic tests. The results are generally uncertain showing large spreading and they can only be interpreted with difficulties. One of the major problems is that circuits can survive the 85 C/85%RH test and operate many hundreds of hours, but after a short exposure to high humidity conditions, they fail. THB test conditions may also be different. While IEC 68-2-3 conditions (40 2 C and 90–95%RH) had been used for a long time during 1980s, today the application of JEDEC Test Method A101-B become widespread, which requires other conditions (85 2 C and 85
5%RH). Conventional THB tests were performed at 40 C/95%RH conditions and a number of potential failures have been revealed that may hidden in 85 C/ 85%RH tests that are preferred in today’s’ quality standards. The reasons are rather obvious: • The increase of the temperature will accelerate the thermally activated Arrhenius-type processes. • The decrease of the RH will slow down the humidity related processes in the same time. • Solid-state surfaces remain generally clean under 90%RH, since moisture deposition will occur above this level. • At 40 C/95%RH, the dew point on the surfaces may occur even in a temperature fluctuation of one degree, while this needs a much larger range at 85 C/ 85%RH. • The most dangerous realistic natural environmental conditions result in low temperatures and high humidity levels (heavy summer rain, fast autumn cooldown, etc.). Normal ambient conditions were defined as 25 C/ 50%RH. Using this, we can calculate the acceleration factors of the tests according to the Eqs. 4 and 5. The parameters of RH, water damp partial pressure (p), saturated water damp partial pressure (ps ), and dew point temperature (Td ) have the following relationships [21]:
1215
Table 1 Comparison of the properties of the two THB test methods Properties
ACCF(T) related to 25 C, (Ea ¼ 1 eV) ACCF(T) related to 25 C (Ea ¼ 0:5 eV) ACCF(RH) related to 50%RH, (x ¼ 5) ACCF(RH) related to 50%RH, (x ¼ 2:5) Humidity fluctuation necessary for condensation (%) Temperature fluctuation necessary for condensation (C)
Test methods 40 C/ 95%RH
85 C/ 85%RH
6.5
683
2.5
26.1
24.8
14.2
5
3.8
þ5
þ15
0.9
4.1
RH ¼ 100p=ps
ð6Þ
pðTd Þ=ps ¼ 1
ð7Þ
ps ¼ A expðB=T Þ
ð8aÞ
In the case of water damp (between 40 and 85 C): ps ¼ ð9:6 1010 PaÞ expð5130=T Þ
ð8bÞ
A comparison between the two THB methods can be seen in Table 1. The most important conclusions are as follows: • The 85/85 method has higher thermal acceleration factor while its humidity acceleration is somewhat lower. • At 40/95, only a one-degree surface temperature fluctuation (which is within the tolerance range allowed by the IEC 68-2-3 standard: 40 2 C) is enough for moisture condensation. At 85/85, there is a need for 4 C fluctuation, which is out of the allowed range given by the JEDEC Test Method A101-B: 85 2 C. • Thus, when testing free surface migration samples, the preliminary conditions for migration, i. e. the moisture condensation has high probability at randomly located sites during 40/95 tests. This probability is practically very low at 85/85, for which an improper temperature control must be supposed.
2. Experimental In order to have a realistic comparison of the THB test methods in connection with their acceleration effect on the migration process, experimental procedures have
1216
P. Bojta et al. / Microelectronics Reliability 42 (2002) 1213–1218
Fig. 1. Layout structure of the samples used in the climatic tests.
been carried out on Ag–Pt and Ag–Pd thick film metallization systems fired on alumina substrates having a geometry of 4 mm length, line width 0.5 mm, spacing 0.25 mm (see Fig. 1). The reason why this conductor material family was chosen is because it shows the largest migration ability. THB tests with the two abovementioned settings (40 C/95%RH/10VDC as well as 85 C/85%RH/10VDC) were performed on the samples. Contaminated thick film samples were also included, since hygroscopic contaminants can strongly influence the moisture condensation processes. The presence of water-soluble compound impurities may cause the condensation of water at
