Polymers for Microelectronics - American Chemical Society

---

Chapter 18

Out-of-Plane Expansion Measurements in Polyimide Films

Downloaded by UNIV OF PITTSBURGH on February 17, 2016 | http://pubs.acs.org Publication Date: November 23, 1993 | doi: 10.1021/bk-1994-0537.ch018

Michael T. Pottiger and John C. Coburn Experimental Station, DuPont Electronics, P.O. Box 80336, Wilmington, DE 19880-0336

Out-of-plane linear CTEs (α ) were calculated from the difference between the volumetric CTE and the sum of the in-plane linear CTEs (α and α ). Volumetric CTEs were obtained from a pressure-volume-temperature (PVT) technique based on Bridgeman bellows. Although the linear CTEs vary significantly with processing, the volumetric CTE is essentially constant, inde­ pendent of molecular orientation. For all of the polyimide films studied, the out-of-plane linear CTEs (α ) were higher than the in-plane linear CTEs (α and α ). z

x

z

y

x

y

Polyimide films are used in a variety of interconnect and packaging applications including passivation layers and stress buffers on integrated circuits and interlayer dielectrics in high density thin film interconnects on multi-chip modules and in flexible printed circuit boards. Performance differences between polyimides are often discussed solely in terms of differences in chemistry, without reference to the anisotropic nature of these films. Many of the polyimide properties important to the microelectronics industry are influenced not only by the polymer chemistry but also by the orientation and structure. Properties such as the linear coefficient of thermal expansion (CTE), dielectric constant, modu­ lus, strength, elongation, stress and thermal conductivity are affected by molecu­ lar orientation. To a lesser extent, these properties as well as properties such as density and volumetric CTE are also influenced by crystallinity (molecular ordering). A typical microelectronics device construction consists of multiple layers of different materials, e.g., metals, ceramics and polymers, in contact with one another. The materials are exposed to repeated thermal cycling during device fabrication leading to the development of thermally induced stresses. The thermally induced stresses result from the mismatch in linear CTEs between the various materials. Polyimides with low in-plane linear CTEs, such as those based on B P D A / P P D , were developed to address this problem. The low in-plane linear CTE of this polymer arises from the high degree of planar molecular orientation.

0097-6156y94/0537-0282$06.00/0 © 1994 American Chemical Society In Polymers for Microelectronics; Thompson, Larry F., et al.; ACS Symposium Series; American Chemical Society: Washington, DC, 1993.

Downloaded by UNIV OF PITTSBURGH on February 17, 2016 | http://pubs.acs.org Publication Date: November 23, 1993 | doi: 10.1021/bk-1994-0537.ch018

18.

POTTTGER A N D COBURN

OuUof-Plane Expansion Measurements

283

While considerable attention has been focused on the relationship of stress to in-plane properties such as the linear CTE and Young's modulus, there is little data on the relationship of stress to out-of-plane properties, in part due to the difficulty of measuring these properties. A high out-of-plane linear CTE of the polymer can result in a high CTE mismatch leading to interlayer delamination and/or cracking. For example, cracking of copper plated-through holes during thermal cycling of printed circuit boards is related to the mismatch between the out-of-plane linear CTEs of the dielectric and the copper (1-2). Processing plays a critical role in affecting molecular orientation and structure and in turn the properties of the resulting film. Molecular orientation in spin coated polyimides develops from a competition between the planar conformation induced by the volume collapse during processing and the tendency of the polymer molecules to assume a random equilibrium conformation (3-16). During processing, the evaporation of solvent and/or reaction by-products leads to shrinkage forces. A spin coated polyimide film is constrained in the plane by the substrate, therefore, the bulk of the shrinkage occurs in the thickness direction. The residual stresses that develop during this volume collapse induce anisotropy in the film resulting in in-plane molecular orientation. The development of in-plane orientation is complicated by the conversion of the relatively flexible and soluble poly(amic acid) precursor into the relatively inflexible and insoluble polyimide. During the conversion, solvent decomplexes from the poly(amic acid) and subsequently evaporates, and water is released as a reaction by-product (17-20). In addition, the molecular weight of the poly(amic acid) is believed to initially decrease, and then slowly build during the later stages of conversion (21). The loss of solvent, the conversion of the relatively flexible and soluble poly(amic acid) precursor into the relatively inflexible and insoluble polyimide, and the increase in molecular weight during cure severely restricts the ability of the molecule to relax to its equilibrium conformation. The formation of the relatively inflexible polyimide and the molecular ordering that develops during processing, lock in the orientation induced by the volume collapse during solvent evaporation. The degree of molecular ordering developed is affected by the heating rate during cure and the final cure temperature, relative to the glass transition temperature of the film (11). The effect of heating rate during cure on the in-plane orientation and molecular ordering can be explained in terms of an effective glass transition of the film. For slow heating rates, the effective glass transition temperature increases faster than the film temperature. Molecular mobility is restricted and significant relaxation does not occur. In contrast, during rapid heating, the film temperature increases faster than the effective glass transition temperature. Above the effective glass transition, significant molecular relaxation can occur. Increasing the heating rate may also increase the depolymerization reaction (21) leading to lower molecular weight and increased molecular mobility. The increased mobility as a result of rapid heating leads to a loss of molecular orientation and an increase in crystalline order as evidenced by both x-ray and dynamic mechanical data. Lower in-plane orientation results in a higher in-plane linear CTE, leading to an increase in the in-plane residual stress (11, 14, 22-23).

In Polymers for Microelectronics; Thompson, Larry F., et al.; ACS Symposium Series; American Chemical Society: Washington, DC, 1993.

Downloaded by UNIV OF PITTSBURGH on February 17, 2016 | http://pubs.acs.org Publication Date: November 23, 1993 | doi: 10.1021/bk-1994-0537.ch018

284

POLYMERS FOR MICROELECTRONICS

The degree of molecular orientation in many polyimides has been shown to be a function of film thickness (10-16). The decrease in orientation with increasing film thickness observed in PMDA-ODA and BPDA-PPD polyimides has been attributed to the rate of solvent loss during the curing process (3-4, 16). A skin is believed to form at the free surface of the film during cure (5-6). The skin slows the evaporation rate of the solvent from the film increasing the time available for relaxation, which leads to a loss in orientation. With increasing film thickness, increasing time, due to longer solvent resident time in the curing film, is available for relaxation to occur. The increase in birefringence with increasing final cure temperature observed in PMDA-ODA and BPDA-PPD films cured using a slow heating rate (2 °C/min) is attributed to increased in-plane orientation and subsequently to ordering at final cure temperatures above the glass transition of the fully cured film. The slow heating rate during cure assures that the curing process takes place predominantly in the glassy (vitreous) state. Molecular mobility is limited in the glassy state and the molecular orientation present in the film cannot relax. Curing above the glass transition provides sufficient mobility to allow for molecular ordering. The influence of processing, molecular orientation and molecular ordering on the linear CTEs (a , a and a ) and volumetric CTE (p) will be discussed. The determination of the out-of-plane linear CTEs (a ) of PMDA-ODA and BPDA-PPD films prepared by two different processes (spin coating and casting) will be described. A comparison of our results with the findings of others will also be presented. x

y

z

z

EXPERIMENTAL Materials Free films of poly(4,4'-oxydianiline-pyromellitic dianhydride) (PMDA-ODA) and poly(p-phenylenediamine-3,3',4,4'-biphenyltetracarboxylic dianhydride) (BPDA-PPD) were prepared from Pyralin PI-2540 and PI-2611 poljKamic acid) in NMP solutions, respectively, by spin coating the precursor solutions onto 5 inch silicon wafers containing a 1000 A thermally grown oxide layer. The films were dried in a V W R Clean Room convection oven at 135 °C in air for 30 minutes. The dried films were cured in a Blue M AGC-160F programmable oven under nitrogen purge by heating the wafer at 2 °C/minute to 200 °C, holding at 200 °C for 30 minutes, heating at 2 °C/minute to the final cure temperature and holding for one hour. The fully cured films were removed from the wafers by dissolving the oxide layer in a 6:1 buffered H F solution. 50 fim Kapton H N (PMDA-ODA) and 25 fim Upilex S (BPDA-PPD) films prepared by commercial casting processes were obtained from DuPont and Ube, respectively. Measurement Techniques The in-plane linear CTEs (a and a ) were measured on fully cured films using a Perkin-Elmer TMA-7 thermomechanical analyzer. A constant force of x

y

In Polymers for Microelectronics; Thompson, Larry F., et al.; ACS Symposium Series; American Chemical Society: Washington, DC, 1993.

18.

POTTIGER AND COBURN

Out-of-Plane Expansion Measurements

285

Downloaded by UNIV OF PITTSBURGH on February 17, 2016 | http://pubs.acs.org Publication Date: November 23, 1993 | doi: 10.1021/bk-1994-0537.ch018

0.030 N was applied to film samples 15 mm long and 2 mm wide. The samples were initially heated at 20 °C/min to 220 °C in the T M A and held for 60 minutes to erase previous thermal history. The samples were then cooled to 20 °C at a rate of 2 0 °C/min and held for 10 minutes. The samples were reheated at 5 °C/min to 220 °C. A n average linear CTE was determined between 50 and 200 °C from the second heat cycle using the following equation:

where L is the initial length of the sample between the grips and A L is the change in the length of the specimen over the temperature interval AT. Spin coated polyimide films are isotropic in the plane of the film, therefore the in-plane linear CTEs are identical, i.e. a = a =» a . i . In contrast, the commercial films, prepared by a casting process, exhibit some in-plane anisotropy, i.e. a # a . Therefore, two in-plane CTE measurements are needed to describe the in-plane CTE behavior. For Kapton H N and Upilex S, a and a were measured parallel and normal to the optical axis, respectively. The pressure-volume-temperature (PVT) data were acquired on predried (to remove moisture) film samples using a Gnomix Research PVT Apparatus in isothermal mode (24-26). The PVT apparatus consists of a sample cell with a flexible bellows on one end containing approximately 1 gram of film and mercury as a confining fluid. The cell is placed inside a pressure vessel. The deflection of the bellows as a result of temperature and/or pressure changes is measured by a linear variable differential transducer (LVDT) located outside the pressure vessel. The deflections are converted to volume changes of the sample using the known PVT properties of mercury. The accuracy of the PVT apparatus is ± 0 . 0 0 2 cm /g up to 250 °C and ± 0 . 0 0 4 cm /g at higher temperatures, with a sensitivity of better than 0.0005 cm /g. The data are acquired at constant temperature and at pressures ranging from 10 MPa up to 200 MPa in 10 MPa increments. Once a measurement has been made at a pressure of 200 MPa, the pressure is reduced back to 10 MPa and the temperature is increased by 10 °C prior to the next series of measurements. This cycle is repeated until a final temperature of approximately 400 °C is reached. TTie data at 0 MPa is found by extrapolation using the Tait equation. The film density at room temperature and atmospheric pressure was obtained using a Micromeritics autopycnometer. 0

x

x

y

in

p

ane

y

x

3

y

3

3

DATA ANALYSIS The PVT data is fitted to the Tait equation, which describes the volume dependence along isotherms as follows (25):

V(P,T) - V(0,T) 1 - 0.0894In 1 +

P B(T)

)

In Polymers for Microelectronics; Thompson, Larry F., et al.; ACS Symposium Series; American Chemical Society: Washington, DC, 1993.

(2)

286

POLYMERS FOR MICROELECTRONICS

where B(T) is the temperature dependent Tait parameter, often given by B(T)=B exp(-B T) x

(3)

2

and V(0,T) is the temperature dependent volume at zero pressure. The expression for the compressibility K(P,T) is

Downloaded by UNIV OF PITTSBURGH on February 17, 2016 | http://pubs.acs.org Publication Date: November 23, 1993 | doi: 10.1021/bk-1994-0537.ch018

1

tdV\ -1

1

= \[P + B(T))

0.0894

-In 1 +

B(T)

1)

(4)

and the expression for the volumetric coefficient of thermal expansion /3(P,T) is 1 tdV

= /3 -PB 0

2 K

(P,r)

(5)

where p is the zero pressure thermal expansivity, taken from the expression for V(0,T). An average volumetric CTE was determined between 50 and 200 °C from the expression for V(0,T) as follows: 0

Hi

(6)

o\Ar)

where V is the volume of the sample at 25 °C and AV is the change in the volume of the specimen over the temperature interval AT. The relationship between the linear CTEs and the volumetric CTE is given by 0

13

~

a + a + a x

V\dTJ

y

2

(7)

P

The out-of-plane linear CTE (a ) is calculated by subtracting the sum of the in-plane linear CTEs from the volumetric CTE. z

RESULTS

Isobaric specific volume versus temperature curves are shown in Figure 1 for Kapton H N , PI-2540, Upilex S and PI-2611. The temperature dependent volume at zero pressure V(0,T) was obtained from an extrapolation of the Tait equation fit to the 10 to 50 MPa data. The V(0,T) data were fit with a second order polynomial. The coefficients for the zero pressure volume, V(0,T), and the Tait parameter, B(T), for each film are listed in Table 1.

In Polymers for Microelectronics; Thompson, Larry F., et al.; ACS Symposium Series; American Chemical Society: Washington, DC, 1993.

In Polymers for Microelectronics; Thompson, Larry F., et al.; ACS Symposium Series; American Chemical Society: Washington, DC, 1993.

200

300

300

Temperature (°C)

200

Temperature (°C)

400

180 MPa

400

h

7

0.65

0.67

0.69

o 0.71

0.73

r

h

o> 0.75 o

0.77

0.73

100

PI-2611

100

Upilex S

200

300

300

Temperature (°C)

200

Temperature (°C)

Figure 1. Isobaric specific volume vs. temperature curves from 20-180 MPa (20 MPa increments).

100

100

7

o> 0.75 |o

0.77

Downloaded by UNIV OF PITTSBURGH on February 17, 2016 | http://pubs.acs.org Publication Date: November 23, 1993 | doi: 10.1021/bk-1994-0537.ch018

400

180 MPa

20 MPa

400

180 MPa

20 MPa

-I

288

POLYMERS FOR MICROELECTRONICS

Table 1. Coefficients for Tait Equation Kapton H N 0.6934 9.157 X 10" 1.337 X 10" 504 2.092 X 10"

V VI V2 Bl B2

Downloaded by UNIV OF PITTSBURGH on February 17, 2016 | http://pubs.acs.org Publication Date: November 23, 1993 | doi: 10.1021/bk-1994-0537.ch018

0

Upilex S

PI-2540 5 7

3

0.7004 9.851 X 101.686 X 10" 425 2.442 X 10"

5 7

3

0.6720 6.230 X 10" 1.470 X 10" 778 2.554 X 10"

PI-2611 0.6710 5.197 X 10" 2.250 X 10" 704 3.116 X 10"

5 7

3

5 7

3

The average volumetric and linear CTEs from 50-200 °C, and the compressibility (#c) and density (p) data at 25 °C are reported in Table 2. The average out-of-plane linear CTE (a ) is calculated from the difference between the average volumetric CTE (j8) and the sum of the two average in-plane linear C T E s ( a and a ). PI-2540 processed in the manner described in this paper has relatively little crystallinity ( i i ) . In contrast, Kapton H N has appreciable crystalline order (27). The higher volumetric CTE for PI-2540 is due to the lower crystallinity in this sample. This is consistent with a lower density and higher compressibility compared with Kapton HN. The lower average in-plane CTEs (