JOURNAL
OF MATERIALS
SCIENCE:
MATERIALS
IN ELECTRONICS
5 (1994) 3 3 9 - - 3 4 3
Thermal shock resistance of miniaturized multilayer ceramic capacitors F. Y E U N G , Y. C. C H A N *
Department of Electronic Engineering, City Polytechnic of Hong Kong, Kowloon, Hong Kong Y. W A N G , Z. L. G U I , L. T. LI
Department of Materials Science and Engineering, Tsinghua University, Beijing, People's Republic of China The thermal shock resistance of miniaturized multilayer ceramic capacitors (MLCs), of sizes "0402", "0603", " 0 8 0 5 " and "1206", was investigated by comparing the leakage currents before and after thermal shock. It was generally found that smaller capacitors have a higher thermal shock resistance than larger ones. The " 0 4 0 2 " M LC possesses a thermal shock resistance in excess of 420 ~ The linear interdependence of thermal shock resistance and reciprocal of half thickness, as predicted by conventional thermal shock analysis, was not observed. Instead, the thermal shock resistance of an MLC was found to be inversely proportional to the total area of its ceramic surface. This confirms that pre-existing flaws on the ceramic surface dominate the crack initiation process and are therefore primarily responsible for determining the thermal shock resistance of an MLC.
1. I n t r o d u c t i o n The recent trend towards miniaturized electronics raises serious concern on the reliability of printed circuit board assemblies (PCBAs) that form the backbone of any electronic system. Miniaturized electronic components, being the basic units of a PCBA, are often the focus of attention in relation to its long-term reliability. During surface mount assembly processes, the PCBAs are subjected to a number of thermal and mechanical stresses. In particular, the environmental stress screening, such as thermal cycling and random vibration which are frequently used in a stringent quality-control procedure, can also induce severe thermal and mechanical stresses on the PCBAs. Multilayer ceramic capacitors are thought to be more susceptib]e to these problems due to their multilayer metal/dielectric structure. For example, the thermal mismatch between the metal electrodes, the ceramic dielectric and the termination material of an M L C can give rise to cracks as a result of applied thermal stresses. Only recently, miniaturized MLCs as small as the "0402" (1.00 x 0.50 mm) and "0603" (1.60 x 0.80 mm) are being used in large-scale manufacturing. It is therefore important and timely to examine the reliability of these electronic components. The aim of this paper is to report the results of thermal shock resistance measurements on BaTiO3-based X7R-type MLCs and offer an explanation for their thermal shock behaviours.
2. Theory Fig. 1 shows the construction of a multilayer ceramic capacitor (MLC). It consists of alternating layers of
ceramic dielectric and electrode material, printed consecutively and co-fired at a temperature of around 1000-1400 ~ The main constituent of the dielectric is usually barium titanate, and the electrodes are palladium-silver thick film. Alternate electrodes are connected to opposite end terminations to form a set of parallel-plate capacitors. Due to the brittle nature of a ceramic dielectric and the thermal mismatch between the constituent materials, an M L C is susceptible to cracks when subjected to mechanical stresses or thermal shock. To gain better insight into the thermal shock behaviour of an MLC, it is necessary to understand the relationship between thermal stress, fracture mechanics and thermal shock. Thermal stress is one kind of inner stress that is caused by thermal expansion or thermal contraction of materials. When a ceramic material is subjected t o a thermal shock (a rapid change in ambient temperature), mechanical stresses result because the surface reaches the new temperature instantly whereas the bulk of the material remains at the initial temperature. In practice, there is a temperature differential between the bulk and the surface of the material. Fig. 2 illustrates typical distributions of temperature and thermal/ mechanical stress for a plate ceramic of infinite length which is thermally shocked at the surface [1]. If the temperature difference between the bulk and the surface is large enough, the thermal/mechanical stress that has developed can exceed the fracture strength of the sample. Consequently, fracture will occur. Generically, the thermal shock resistance of an M L C is defined as its resistance to fracture when subjected to a rapid change in ambient temperature. Generally, the
* A u t h o r t o w h o m c o r r e s p o n d e n c e s h o u l d be a d d r e s s e d .
0957-4522 9 1994 Chapman & Hall
339
modulus 13 = r~,h/~: (where r m is the half thickness of sample and h is the heat transfer coefficient) is greater than 20 [1] and the sample thickness (r m • 2) is larger than 0.16cm (h is taken to be 1 c a l s - l c m - 2 ~ from [1] and ~ is taken as 0 . 0 1 6 9 W c m - l ~ -1 as given in [4]). This simulation model has limited validity even on MLCs of sizes "1808" and "1815" whose thicknesses are usually larger than 0.160 cm [6-8]. In the case of miniaturized MLCs, the heat transfer condition is not very severe because the sample thickness can be as small as 0.050 cm (say, "0402" MLC). Therefore, a more accurate expression [1] describing the temperature differential (thermal shock resistance) of an M L C resisting to fracture is
Termination Bandwidth
Internal Electrode
Figure 1 Construction of a multilayer ceramic capacitor.
ATr =
/
I Ts
AT
t Surface temperature
I Ta
(or shock temperature)
-~
7~: Initial temperature
Ta: Average temperature AT." Temperature difference Tension ~ I ~ C o m p r e s s i o n Figure 2 Temperature and stress distributions of a plate ceramic o f infinite length which is thermally shocked at the surface. A T is the temperature difference between the surface and the bulk, and is equal to the maximum thermal shock temperature applied to the sample.
temperature differential (thermal shock resistance) of a homogeneous material resisting to fracture, A T r, when subjected to a thermal shock, is expressed as
Arf
-
KO'f(1 - - It)
E~
(0
where of is the fracture stress, ~ is the thermal expansion coefficient, g is Poisson's ratio, E is Young's modulus, • is thermal conductivity [1]. However, the simple expression in Equation 1 cannot be readily applied to estimate the thermal shock resistance of an M L C because it is a highly inhomogeneous body due to its multilayer structure. Its thermal characteristics are strongly dependent upon a number of physical parameters including ductility, porosity and pre-existing flaws. During the last decade, the influence of different parameters, such as material properties, component dimensions, electrode configuration and composition, on the thermal shock resistance of M L C s has been investigated [2-5]. The discrepancies between theoretical predictions and experimental results are not yet well understood. Equation 1 has been used to evaluate the thermal shock resistance of X7R-type barium titanate materials I-4]. F r o m such a calculation, crack initiation is shown to occur when the material is subjected to a thermal shock of 93 ~ under a severe heat transfer condition (such as water quenching). It should be noted that this simulation model is only applicable to a simple homogeneous sample without inner electrodes and under a severe heat transfer condition. Further, this equation is valid only when the Biot
340
K~f(1 - ~) 1 E~ 0.31rmh
(2)
(here again, the M L C is treated as if it were a homogeneous body). Meanings of the various symbols have been.defined in the text. However, the effect of the internal metal electrodes of an M L C on its thermal shock behaviour cannot be ignored. It has been reported that the metal electrodes may degrade the fracture toughness and the metalceramic interface is a preferred fracture path for crack propagation parallel to the electrodes [5] due to the mismatch of thermal expansion coefficients of the dielectric and electrode materials. For an M L C with an interelectrode separation of about 10 20 gm layer dielectric thickness, the dimension of a microcrack is usually comparable to the thickness, thus making it vulnerable to electrical short-circuiting between adjacent electrodes. On the other hand, the presence of metallic electrodes in a M L C has a particular effect on its thermal shock resistance. Since the electrode materials, such as silver or silver/palladium, have very good thermal conductivity, it can be approximated that all the electrodes are at the shock temperature T s (or surface temperature, which is taken as the ambient temperature to which the M L C is thermally shocked). Thus this makes it possible to apply Equation 2, at least to a first order, to estimate the thermal shock resistance of each individual dielectric layer provided that other determining factors, such as the presence of pre-existing flaws on the M L C free surface, are ignored. Further, each layer can be regarded as one of infinite length as far as the thermal shock calculation is concerned because the length-to-thickness ratio is well over 400. The thermal shock resistance of an M L C is taken as when one of the N (N is the number of dielectric layers in an MLC) identical thin ceramic capacitors of an M L C has developed cracks, since all of them are connected in parallel. The thermal shock resistance of one such thin ceramic capacitor is therefore assumed to be the thermal shock resistance of an MLC. With this assumption in mind, there is a good possibility of applying the current models (based on a homogenous body) to a complex multilayer structure where theoretical analyses can be extremely tedious. This paper attempts, for the first time, to apply this simplified model to fit the experimental data and offer an explanation for the discrepancies.
T A B L E I Details of M L C samples Sample group
Size code/EIA spec C(gF)/rated voltage
Main composition
Dimension L • W • H (mm)
B (mm)
Dielectric layer thickness (p.m)
N
A
"0402"/X7R 0.01 [xF/16 V
BaTiO 3
1.00 • 0.50 • 0.50
0.20
11
21
B
"0603"/X7R 0.01 gF/50 V
BaTiO s
1.60 • 0.80 • 0.80
0.30
17
18
C
"0805"/X7R 0.01 gF/50 V
BaTiO 3
2.00 • 1.20 x 0.85
0.45
18
28
D
"1206"/X7R 0.1 gF/50 V
BaTiO 3
3.20 • 1.60 x 0.90
0.64
16
36
B: termination bandwidth. N: dielectric layer number.
3. Experimental procedure The characteristics of the test samples are summarized in Table I. Thermal shock measurements were carried out by preheating an M L C sample to a defined temperature ranging from 50~ to 500 ~ and the incremental temperature for each measurement was 20 ~ It was then quenched in ice water, cleaned in pure alcohol and dried at r o o m temperature in an ambient humidity of 50% relative humidity. Before and after thermal shock, electrical parameters of the sample (leakage current, capacitance and dissipationfactor) were measured using a PC-based data analysis system consisting of an HP4284A L C R meter, a HP6624 D C power supply and an HP3458 digital multimeter. The criterion of a failed M L C capacitor was taken as one where the leakage current, measured after thermal shock, was larger than 1 0 - S A upon application of the rated working D C voltage. Typically, this leakage current was at least three to four orders of magnitude larger than that of the leakage current detected before thermal shock. After electrical measurements, all samples were examined using optical microscopy and destructive physical analysis.
4. Results and discussion Fig. 3 summarizes the leakage current characteristics of MLCs of different sizes which were subjected to different degrees of thermal shock by ice-water quenching. It can be seen that the leakage current generally remains fairly constant with increasing thermal shock until its magnitude is greater than the thermal shock resistance of an MLC. Once the thermal shock temperature exceeds the thermal shock resistance of an MLC, microcracks would initiate from the concentration points of thermal stress and propagate to connect adjacent internal electrodes. If moisture or polarized organic medium is present in these cracks, an ionic current path would be made and the resultant leakage current would be significant [9]. There must be a strong correlation between the density of cracks in an M L C al~d the magnitude of thermal shock. It is generally true that when the magnitude of thermal shock exceeds the thermal shock resistance of an MLC, further increase in thermal shock will bring about mpre cracks, thus possibly causing a corres-
100 10-2 "1206" MLC 10.4 10 -8
/
~
"0603" MLC
g lO-8 ._1
10-1o 10-12 0
o4o ? j 50 100 150 200 250 300 350 400 450 500 Temperaturedifferential,(~
Figure3 Leakage current characteristics of different X7R-type miniaturized M L C s subjected to different degrees of thermal shock by ice-water quenching
ponding increase in leakage current. This was indeed observed in Fig. 3. By applying the criterion for a failed MLC, the thermal shock resistance can be estimated from the leakage current characteristic [10]. F r o m the curves in Fig. 3, the thermal shock resistance of"0402", "0603", "0805" and "1206" X7R-type M LCs were estimated to be 440~ 380~ 220~ and 100~ respectively. It should be emphasized that these data were taken from thermal shock measurements done under ice-water quenching. Such thermal shock treatments are different from those in the usual environmental stress screening or surface mount assembly processes. Since the M L C test samples were selected to have the same material composition, all the variables in Equation 2 can be regarded as constants except the half thickness rm. According to this equation, and assuming a homogeneous body, the critical temperature differential (thermal shock resistance) is inversely proportional to the half thickness (rm) of an MLC. Thus a plot of ATe versus 1/rm should be a straight line. However, the plot in Fig. 4, does not seem to show a linear relationship between ATf and r m. It should be noted that in this analysis r m is taken as the half thickness of an MLC. Since it is a multi341
600
T,I I
U
Ts: Surface temperature
(or shock temperature)
*---500
roii
o
7]: Initial temperature
400
Ta: Average temperature (between inner electrodes)
"0603" ~300
AT."Temperature difference
200
Figure 5 Temperature distribution of a multilayer ceramic capa-
citor.
~100 "1206" 0 2.0
i
2.5
r
3.0 3.5 1 / rm, (ram -1)
i
4.0
500
4.5
Figure 4 Thermal shock resistance versus reciprocal of half thick-
"~400 H
u
ness of MLC. '~ 300
0200
layer structure and cannot be treated as a homogeneous one, the deviation from the linear plot behaviour is within expectations. To simplify the thermal analysis of an MLC subjected to thermal shock, it is assumed that all the metallic electrodes, such as silver or silver/palladium, are at the shock temperature due to their very good thermal conductivity. The M L C can then be thought of comprising N thin ceramic capacitors stacked together in parallel, where N is the number of dielectric layers. The temperature distribution of an MLC subjected to thermal shock is shown in Fig. 5. Further, each dielectric layer can be regarded as one of infinite length, because the length-to-thickness ratio is well over 400. The thermal shock applied to an M L C is therefore simplified to consider the effect of thermal shock on each dielectric layer separately. The thermal shock resistance of one such thin ceramic capacitor is therefore assumed to be the thermal shock resistance of a n MLC. It is clear that the advantage of this approach is the great simplification in handling a complex multilayer structure, where theoretical analysis can be very tedious. The temperature differential (thermal shock resistance) resisting to fracture can be expressed as AZf
--
•of(1 - ~t) 1 E~ 0.31tmh
(3)
where tm is the half thickness of a dielectric layer and other symbols have the usual meanings. A plot of ATf v e r s u s 1 / t m is shown in Fig. 6. Again, the thermal shock resistance does not have a good linear relationship with the reciprocal of half thickness of a dielectric layer. This tends to suggest that the conventional thermal shock analysis for a homogeneous layer of infinite length [1] does not yield predictions compatible with experimental data. With the aid of destructive physical analysis (DPA) and optical microscopy, the vast majority of microcracks was found either near the ceramic surface or emanating from the interface between internal electrode and dielectric layer of an M L C sample as illustrated in Fig. 7. This kind of crack distribution can be attributed to the tensile force on the ceramic surface 342
~"0805"
~ 100
o
"1206" 0
0.10
r
0.12
p
r
0.14 0.16 1 / tm, (t~m-1)
i
0.18
0.20
Figure 6 Thermal shock resistance versus reciprocal of half thickness of single-layer dielectric.
Figure 7 Cross-section of a fractured "1206" size X7R-type MLC subjected to thermal shock by ice-water quenching, showing cracks near surface of sample ( x 100).
and on the interface when the MLC is cooled suddenly by ice-water quenching, as shown in Fig. 5. It is a reasonable assumption that the probability of microcrack initiation has a strong relationship with the total ceramic surface area (external and internal). As the total ceramic surface area of an M L C decreases, the total defect count on the sample surface and dielectric layer also decreases, thus reducing the probability of crack initiation during thermal shock. As illustrated in Fig. 1, an MLC body has six outer faces in which two termination faces are presumed not to undergo thermal shock. The total external ceramic surface remains as four faces of dielectric surface. The total area of
600 50O
"0402"
.~ 400 _~ 300 O
_~ 200 r
U4~,
a
== loo F-
2'
~ ~ ' 4~ 6~ 8' 10 12~ 14 16 Total surface area of MLC (rnm 2)
1'8
20
Figure 8 Thermal shock resistance versus total area of ceramic surface of MLC.
external ceramic surface ( S e x t . . . . 1), total area of internal ceramic surface (Sint. . . . 1) and total area of ceramic surface (Stotal) can be defined as Sext. . . . 1 =
2 [ ( L - - 2 B ) ( W + H)]
(4)
Sint
=
2N x L x W
(5)
=
S e x t . . . . 1 -~- S i n t . . . . 1
(6)
....
Stotal
1
where L, W and H are the length, width and height of the MLC, respectively, B is the termination bandwidth and N is the dielectric layer number. From Table I, the total areas of ceramic surface for the "0402", "0603", "0805" and "1206" MLCs are 22.2, 49.28, 138.91 and 378.24 mm 2, respectively. A plot of thermal shock resistance versus total external ceramic surface area of an M L C does indeed yield approximately a linear relationship (Fig. 8). In fact, the thermal shock resistance of an M L C is inversely proportional to its total ceramic surface area. The interpretation is that the preexisting flaws at the ceramic surface are mainly responsible for the crack initiation process when the M L C is subjected to sufficient thermal shock. Therefore, these surface flaws are dominant factors in determining the thermal shock resistance of an MLC. F r o m the above i it is also assumed that the free sample surface reaches the shock temperature quicker than the inner electrodes, despite its very good thermal conductivity. This explains why there were more microcracks distributed near the surface of the MLC.
5. C o n c l u s i o n s The thermal shock resistance of multilayer ceramic capacitors were studied by comparing the leakage
current behaviours before and after thermal shock. It is generally found that smaller capacitors have a higher thermal s h o c k resistance than larger ones. "0402" X7R-type MLCs possess a thermal shock resistance in excess of 440 ~ The linear dependence between thermal shock resistance and reciprocal of half thickness, as predicted by conventional thermal shock analysis, was not observed. Instead, the thermal shock resistance of an M L C is shown to be inversely proportional to the total area of ceramic surface. This confirms that pre-existing flaws on the ceramic surface dominate the crack initiation process and are therefore primarily responsible for determining the thermal shock resistance of an MLC.
Acknowledgements The authors thank the City Polytechnic of Hong Kong for financial support ( C P H K Strategic Research Grant No.: 700329). One of the authors (F. Y.) is grateful for a C P H K research studentship in support of his study for an MPhil degree in Electronic Engineering.
References 1.
2. 3.
4. 5. 6.
7.
8.
9. 10.
W.D. KINGERY, H. K. BOWEN and D. R. UHLMANN, in "Introduction to Ceramics" (eds E. Burke, B. Chalmers and A. Krumhansl) 2nd Edn (Wiley, New York, 1976) Ch. 16. S.W. FREIMAN and R. C. POHANKA, J. Amer. Ceram. Soc. 72 (1989) 2258. R.S. RAWAL, R. LADEW and R. GARCIA, in Proceedings of the 37th Electronic Components and Technology Conference, Boston, MA, 1987 (IEEE CHMT) Piscatanky, NJ, 1987, p. 145. H.V. DEMATOS and C. R. KORIPELLA, in Proceedings of the 8th CART Symposium, 1988, p. 25. K . R . MCKINNEY, R. W. RICE and C. C. WU, J. Amer. Ceram. Soc. 69 (1986) 228. MURATA Chip Monolithic Ceramic Capacitor Application Manual (obtainable from MURATA MFG Co, Ltd., Nagaokakyo-shi, Kyoto 617, Japan). KEMET Mulfilayer Ceramic Chip Capacitors Application Manual (obtainable from KEMET Electronics Corp., P.O. Box 5928, Greenville, S.C. 29606, USA). TDK Multilayer Ceramic Chip Capacitors Application Manual (obtainable from TDK Corp., 13-1, Nihonbashi 1 Chome, Chuo-Ku, Toyko 103, Japan). R.C. CHITTICKA and E. GRAY, IEEE Trans. Components, Hybrids M a n u f Technol. 12 (1983) 510. Y.C. CHAN, F. YEUNG and T. S. MOK, J. Mater. Sci. Mater. Elec. 5 (1994) 25.
Received 11 January and accepted 26 April 1994
343