Intl.

Journal of Microcircuits and Electronic Packaging

Comparison of Paraelectric and Ferroelectric

Materials for Applications as Dielectrics in Thin
Film Integrated Capacitors
Richard Ulrich and Len Schaper
High Density Electronics Center (HiDEC)
University of Arkansas
Fayetteville, Arkansas 72701
Phone: 501-575-5645
Fax: 501-575-2719
e-mail: rulrich@engr.uark.edu
David Nelms and Matt Leftwich
Arkansas Microelectronics Development Corp.
700 Research Center Blvd.
Fayetteville, Arkansas 72701
Phone: 501-575-5614
Fax: 501-575-7446
e-mail: dmn@engr.uark.edu

Abstract
Ferroelectrics such as BaTiO3, PbxZr1-xTiO3 and BaxSr1-xTiO3 can exhibit dielectric constants up to three orders of magnitude higher
than those of paraelectric materials such as SiO2, Al2O3, Ta2O5, and BCB. However, the dielectric properties of ferroelectrics are
typically a stronger function of temperature, frequency, film thickness, and bias resulting in significant nonlinearities in their performance. Also, the dielectric constant of some ferroelectrics degrades with time. As a result, paraelectrics are more suitable for hightolerance applications such as filtering, timing, RF/wireless, and A/D where constant, predicable capacitance is required. Ferroelectrics would be preferred for decoupling due to their higher dielectric constants and due to the fact that less tolerance is necessary; the
capacitor only needs to meet a certain minimum value. Either might be used for termination as long as footprint requirements can be
met. Annealing temperatures for high-k ferroelectrics are currently excessive for applications on organic boards which complicate
their integration into these substrates. Since capacitor values on a single board might span up to six orders of magnitude, it may be
necessary to use more than one dielectric material to restrict the capacitor footprints to an acceptably smaller range. BCB and SiO2
may be more useful for these low values and may be already present on the substrate as an interlayer dielectric.

Key words:

1. Introduction

Integrated Capacitors, Integrated Passives, Dielectric Materials,

and Dielectric Constant.

In the coming years, capacitors will be increasingly fabricated

into microelectronic substrates in some manner since there are
clear benefits in functional density, performance, and reliability
relative to surface mount discretes1-5. The main questions to be
resolved involve the choice of materials, configurations, and fabrication processes to implement integrated capacitors in a cost-

The International Journal of Microcircuits and Electronic Packaging, Volume 23, Number2, Second Quarter 2000 (ISSN 1063-1674)
172

International Microelectronics And Packaging Society

Comparision of Paraelectric and Ferroelectric Materials for Applications as Dielectrics in Thin Film
Integrated Capacitors

effective manner. The purpose of this article is to compare and

contrast, for integrated capacitor applications, the two major
classes of dielectric materials: paraelectrics and ferroelectrics,
sometimes called class 1 and class 2 dielectrics, respectively.
First, the mechanisms of charge storage are briefly reviewed for
the two classes in order to elucidate the differences in fabrication
options, electrical performance, and the effects of operational
conditions such as temperature, bias, frequency, and film thickness. Then, the two types of dielectrics are compared, based on
these physical and electrical properties, for use in the integrated
capacitor applications of decoupling, analog functions, and termination. It will be shown that the fundamental difference in
the way the two materials store charge is very important to their
respective suitability for these applications. It will also be shown
that selecting the dielectric with the largest k is not always the
optimal choice. Broad criteria for the selection of materials for
integrated dielectrics will be given as a function of application,
footprint requirements, and substrate material.

tion of the materials structure and lattice flexibility as well as its

composition. Ionic polarization is similar to atomic polarization
but involves the shifting of ionic species under the influence of
the field which can lead to very high dielectric constants, up to
several thousand.
Both paraelectric and ferroelectric materials exhibit electronic,
atomic, and ionic polarization; the two classes of materials are
not distinguished from one another in that regard. The distinguishing feature is that ferroelectric materials do not lose their
ionic polarization when the field is removed and paraelectrics
do. Due to the lattice hindrances in ferroelectrics, the electric
field can pull the ions into configurations that do not relax back
to the previous state once the field is removed. As a result, ferroelectrics, analogous to ferromagnetic, materials, can possess a
residual polarization after the field is removed. Furthermore,
this residual polarization can be oriented in one direction or the
other depending on the direction of the last field. It is this trait
that gives them utility in non-volatile memories. Paraelectric
materials are those which cannot be left with a residual polarization once the field is removed since they do not have a mobile
charged atom with more than one stable lattice position.
2. Mechanisms of Dielectric Charge
This residual polarizability defines ferroelectric and
paraelectric materials but this feature is irrelevant to any proStorage
jected substrate-level integrated capacitor application. The important difference from the point of view of integrated passives is
that ferroelectrics generally have much greater dielectric conThe ability of a dielectric material to store energy under the
stants than do paraelectrics, sometimes by as much as three orinfluence of an electric field results from the separation and alignders of magnitude, due to the mobile ionic charge. For example,
ment of electric charges brought about by that field. The larger
the common ferroelectric barium titanate is a perovskite crystal
the dipole moment arm, the separation of charges in the direcwith a Ti cation at the center which can shift back and forth
tion of the field, and the larger the number of these dipoles, the
within the confines of this crystal a distance larger than the ions
higher the materials dielectric constant. There are several poscan move in a typical paraelectric such as tantalum oxide. This
sible contributions to this polarizability which, depending on the
motion creates a considerably larger dipole arm than is possible
mechanisms operative in a given dielectric, determine not only
in Ta2O5 resulting in a much larger dielectric constant for BaTiO3.
the value of k, but also how it varies with frequency, temperature,
Incidentally, it is the fact that this Ti ion can remain shifted to
bias, impurity concentration, and crystal structure6, 7. The three
one side or the other after the field is removed that gives BaTiO3
general mechanisms important to candidate materials for inteits ferroelectric residual polarization behavior.
grated capacitors are electronic, atomic, and ionic polarization.
Table 1 shows the dielectric constants for ten paraelectrics
Electronic polarization involves charge symmetry distortion
and four ferroelectrics clearly showing the wide difference beof an atom. Under the influence of an applied field the nucleus
tween them. The ks given for the ferroelectrics are maximum
and the negative charge center of the electrons shift in opposite
amounts since their specific values depend on grain size, crystal
directions creating a small dipole. This induced dipole effect
orientation, electrical bias, frequency of measurement, and film
occurs in all materials, including air, but is usually very small
thickness. Ferroelectrics must possess a crystal structure in orcompared to other polarization mechanisms since the moment
der to exhibit these high dielectric constants otherwise their ks
arms of these dipoles are very short, usually a fraction of the size
are no higher than typical paraelectrics. The dielectric constant
of an atom. Atomic polarization occurs in substances made up
of amorphous barium titanate is comparable with paraelectrics
of more than one type of nonionic atom due to the different eleat a value of about 17 since the polarization contribution of the Ti
ments will not normally share the electron cloud equally; it will
moving in the crystal cage is lost when there is no crystal strucbe shifted towards the more electronegative atoms resulting in a
ture to support it8. The values for ferroelectrics in Table 1 are for
permanent dipole. The electric field will then produce opposite
completely oriented bulk material single crystals at low frequency
forces on various parts of the molecules causing the atoms themand no bias. Therefore, they represent maximum ranges of valselves to move to create the alignment of charges and field. It is
ues for these materials. The dielectric constant for ferroelectrics
the movement of charges aligning with the field which creates
is so highly dependent on processing and measurement condicapacitance, not the presence of permanent dipoles. If these rantions that it is not possible to be more specific but the values
domly-oriented permanent dipoles were somehow not capable of
given for paraelectrics are only weakly dependent on how they
movement and alignment, there would be no dielectric constant
are fabricated or how the ks are measured and represent typifor that material above that of a vacuum. Therefore, k is a funcThe International Journal of Microcircuits and Electronic Packaging, Volume 23, Number2, Second Quarter 2000 (ISSN 1063-1674)
International Microelectronics And Packaging Society

173

Intl. Journal of Microcircuits and Electronic Packaging

cally achieved values. Only a few paraelectrics exist in multiple

crystal forms such as amorphous and hexagonal Ta2O5.
Table 1. Dielectric constants for paraelectric and ferroelectric
dielectrics.

3. Fabrication Effects on Dielectric

Constants
The electrical properties of any dielectric are a function of its
chemical composition and crystal morphology which, in turn,
are influenced by the technique used to form the material. There
are many ways to create dielectric films including spin-coating,
sputtering, CVD, MOCVD, sol-gel, pulse-laser deposition, anodization, dry calcination, hydrothermal, among many techniques.
In general, ferroelectric materials must achieve a crystalline or
polycrystalline state in order to exhibit higher dielectric constants
than paraelectrics. This may require anneal temperatures of 500
- 700C in oxygen which precludes the use of this technology on
some common substrates. Most paraelectric materials are amorphous and their dielectric constants are relatively insensitive to
processing conditions. Some of the fabrication methods germane
to integrated capacitors are outlined next.
BaTiO3 powder can be made by dry calcination of BaCO3 and
TiO2 at temperature >1200C resulting in particles 1000 - 2000
in diameter with dielectric constants in the thousands. These
or other types of ferroelectric high k particles can be mixed with
epoxy up to 70% loading by volume, cast in thin film form, and
the epoxy cured at under 200C. The mixing rules for two materials of different dielectric constants are such that the dielectric
constant of the final composite will be much closer to that of the
low k material which is generally an epoxy with a k of about 3 to
5. The overall dielectric constant will end up being around 30 100. The result is films21-23 as thin as 1 micron which can be
made pinhole-free with some tens of nF/cm2. The advantage of
this method is that much of the processing, and all of the high
temperature steps necessary to obtain high k from ferroelectrics,
can be achieved in advance of application to the substrate. This
method is not suitable for paraelectrics since the mixture of these

much lower-k powders with epoxy would have about the same
dielectric constant of the epoxy alone.
A number of sol-gel methods have been developed for ferroelectrics which result in films of under 1 m and dielectric constants in the low thousands14, 17, 18, 24, 25. These techniques can give
specific capacitances of thousands of nF/cm2 but require a high
temperature cure to achieve the necessary crystal structure. Currently, there are some aqueous routes at temperatures as low at
100C to make high-k BaTiO3 but the solution must be very alkaline, as high25-28 as pH 14.
CVD, laser-ablation29, MOCVD, and sputtering of ferroelectrics and paraelectrics have been practiced for decades. Again,
the ferroelectrics will usually require a high temperature cure to
achieve the very high-k values which result from the crystal structure8. For sputtered dielectrics of any sort, the roughness of the
bottom electrode is critically important since steep slopes might
not be sufficiently covered resulting in unacceptable leakage or
even shorts28, 30, 31. Most investigations of implementing a sputtering procedure for integrated capacitors end up being fixated
on this issue in order to bring the yield up to acceptable levels.
As a result, obtaining acceptable yields for sputtered integrated
capacitors with either dielectric type on Kapton and alumina substrates is more difficult than it is on Si due to the inherently
higher roughness of these materials22, 32, 33.
Anodization avoids this roughness problem since a thick layer
of anodizable metal can be deposited, either Al or Ta, and the
anodization procedure can be controlled to convert only enough
metal into dielectric as is necessary to achieve the desired specific capacitance34. Therefore, the composition or roughness of
the metal underlying the Al or Ta is not an issue, and this approach avoids having to planarize and monitor the state of the
bottom electrode. Thickness control is achieved through the final voltage of the anodization cell. For Ta, the thickness of the
resulting dielectric oxide is 16 per anodizing volt and it is only
a weak function of the processing time or current density35-38.
The resulting Ta2O5 films have a k of about 24, two orders of
magnitude below deposited and annealed ferroelectrics. BaTiO3
has been made by anodization without a high-temperature anneal but, to date, with poorer dielectric properties than sol-gel
films which were subsequently thermally cured39, 40.
Most ferroelectric deposition methods (CVD, MOCVD, sputtering) require an oxidizing atmosphere of 500 - 700C to form
crystalline phases. This restricts the choice for the bottom electrode to nonoxidizable materials such as Pt or Au. Other metals,
such as Al or Cu, react to form their own oxides which, for thin
dielectrics, will dominate the overall dielectric constant of the
two layers through the mixing rule mentioned earlier for powders in epoxy. Pt is popular as the base electrode since, in addition to thermodynamic stability, it has a high Schottky barrier
and, therefore, a lower leakage for thin films11. Figure 1 shows
the measured dielectric constant of sputtered ferroelectric PZT
and BST films deposited on Pt as a function of film thickness
along with paraelectric Ta2O5. Substrate effects such as stress,
defects, grain size, and interdiffusion restrict the ferroelectrics
dielectric constant to values below those of the bulk material until

The International Journal of Microcircuits and Electronic Packaging, Volume 23, Number2, Second Quarter 2000 (ISSN 1063-1674)
174

International Microelectronics And Packaging Society

Comparision of Paraelectric and Ferroelectric Materials for Applications as Dielectrics in Thin Film
Integrated Capacitors

quency and, for that matter, the dielectric constant of the ferroelectrics are highly dependent upon crystal structure and orientation. Amorphous BaTiO3 gives a k of only 17, similar to
paraelectrics, and also has flat frequency response8. SiO2 and
Ta2O5 are amorphous.
1.2

SiO 2
1.0

k / k at 100 Hz

the film is at least a few thousand angstroms thick14. These same

films show a higher k on Pt than on stainless steel due to better
lattice match. There is no film thickness or substrate composition dependencies for paraelectrics in general41.
Regardless of the deposition method, the dielectric constant
of most ferroelectric materials is a function of the film stoichiometry17, 42, crystal structure, substrate characteristics15, interfacial reactions, microstructural heterogeneities, lattice defects,
mechanical stresses, film thickness16, and electrode material11.
Amorphous paraelectrics are largely immune to all of these effects but at the cost of a much lower k.
1,400

PZT

Al 2O3 Ta2O5
PZT

0.8

BST
BaTiO3

0.6

0.4

0.2

1,200

Dielectric Constant

BST

0.0
1.0E+00

1,000

1.0E+03

1.0E+06

1.0E+09

Frequency (Hz)
800

Figure 2. Dielectric constant versus frequency for

paraelectrics and ferroelectrics.

600

400

200

Ta O
2 5

0
0

Film Thickness (m)

Figure 1. Dielectric constant versus film thickness.

4. Operational Effects on Dielectric

Constants

10

Temperature has a greater effect on the dielectric constant of

most ferroelectrics than paraelectrics since the polarization mechanisms are dependent on the mobility of segments of the crystal
lattice as well as on the crystal structure itself and due to the
phase transitions in the crystalline materials. Figure 3 shows the
effects of temperature on the same three ferroelectrics from Figure 2 and demonstrates the widely-varied types of temperaturedriven behavior ferroelectrics can exhibit15, 18, 20. The peaks in
the BaTiO3 data are due to crystal transitions42, such as the tetragonal to cubic conversion at 120C. Some ferroelectrics can
be modified with additives to reduce temperature dependencies19, 42.
Most paraelectrics exhibit moderate positive temperature coefficients of capacitance, generally from +100 to +400, near room
temperature and similar to PZT in Figure 3. The common
paraelectrics, such as SiO2, Ta2O5, Al2O3 and BCB, do not show
phase transition behavior within expected microelectronic operation temperatures of -50 to +150C. Figure 4 shows k versus
bias for Ta2O5 and BST12. Many ferroelectrics exhibit a marked
decrease in dielectric constant with increasing DC bias which
may be used to advantage in fabricating a variable capacitor for
tuning applications.

Temperature, frequency, and bias can affect the dielectric constant of both paraelectric and ferroelectric dielectrics by acting
through the relevant polarization mechanisms. For instance, when
a high frequency field is applied, the dipole must reverse direction at the same rate for the polarization to remain in synchronization with the field. As the frequency increases, it may outrun
the ability of the particular dipole to keep up with the reversals
resulting in the dipole arm being effectively shortened, resulting
in a decrease in dielectric constant. Of the charge storage mechanisms described above, only the Ti motion in ferroelectrics is
affected at frequencies below the infrared range. Figure 2 shows
the ratio of the dielectric constant measured at various frequencies to the value at low frequencies18, 43, 44. While the paraelectrics
show no significant decrease and the ferroelectrics show a sharp
dropoff, it should be remembered that ferroelectrics may start
with such a high k that, even at GHz frequencies, they may still
have higher dielectric constants. These curves for BaTiO3 and
BST are for sintered particles while the PZT is a sol-gel thin
film. The actual degree of this dropoff, where it begins in freThe International Journal of Microcircuits and Electronic Packaging, Volume 23, Number2, Second Quarter 2000 (ISSN 1063-1674)
International Microelectronics And Packaging Society

175

Intl. Journal of Microcircuits and Electronic Packaging

to show cycle fatigue but SrBi2Ta2O9 does not45. Paraelectrics do

not show aging effects to any measurable degree. Figure 5 shows
that, as a general trend, the breakdown field is lower for materials with higher dielectric constants, although even the high-k
ferroelectrics exhibit breakdown fields of over 100 V for a 1 micron film (1 MV/cm)11, 46. Under high field conditions, the log of
time to breakdown is proportional to applied field for ferroelectrics11, 14, as shown in Figure 6, while paraelectrics do not show
this tendency.

1.60

BaTiO3

k / k at 20C

1.40

1.20

PZT
1.00

BST
0.80

0.60

0.40
10,000

0.20

-60

-40

-20

20

40

60

80

100 120 140

Breakdown Field (V/m)

0.00
-100 -80

Temperature (C)

Figure 3. Dielectric constant versus temperature for

ferroelectric dielectrics.

SiO 2

1,000

BiN2

NbO 2
BCB

Al 2O3
PTO

Ta 2O5

BST

TiO2

100

STO

PZT

BST

1.20

Ta O
2 5

1.00

10

k at bias / k at 0 V

10

100

1,000

10,000

Dielectric Constant

0.80

Figure 5. Breakdown field versus dielectric constant.

0.60

BST

1,000

0.40

0.20

0.00
0

50

100

150

200

Bias (V/m)

Time to Breakdown (sec)

BST
100

PZT

10

SrTiO 3

1.0

Figure 4. Effect of bias on dielectric constant.

0.1
0

20

40

60

80

100

120

140

Applied Field (V/m)

5. Dielectric Fatigue
Capacitors made with ferroelectric formulations can display a
decrease of capacitance with time, number of charge cycles11,
and temperature45. This phenomenon, called dielectric fatigue
or aging, occurs due to crystallographic changes related to the
relaxation of lattice strain energy42. The rate of decay is logarithmic, of the form,
k = ko - m log t

Figure 6. Time to breakdown versus applied field.

6. Selection of Dielectrics for

Integrated Capacitors
The general characteristics of these two classes of dielectrics
are summarized in Table 2. The factors that influence the selection of a dielectric for integrated capacitor applications include
the following factors,

The quantity m is sometimes expressed in percent per decade-hour. These effects are reversible with temperature and/or
applied field and not all ferroelectrics exhibit them; PZT seems
The International Journal of Microcircuits and Electronic Packaging, Volume 23, Number2, Second Quarter 2000 (ISSN 1063-1674)
176

International Microelectronics And Packaging Society

Comparision of Paraelectric and Ferroelectric Materials for Applications as Dielectrics in Thin Film
Integrated Capacitors
Table 2. Comparision of paraelectric and ferroelectric
dielectrics.
Property

Paraelectrics

Ferroelectrics

2 - 50

up to 1000s

little dependence,
<500 ppm/C

k vs. film structure

little or no dependence

can be highly dependent due to crystal

phase transitions and ion mobility
decreases significantly,
typically above a few GHz
highly dependent due to
effects on crystal structure
decreases with DC bias
k can decrease significantly
with cycles and time
film must be crystalline

cure requirements

none

may require up to 700C in O2

k vs. T
k vs. frequency

little dependence
no dependence since
amorphous
no dependence

k vs. film thickness

k vs. bias
dielectric fatigue

none

6.1. Performance Stability and Predictability

The dielectric constant of most ferroelectrics are a much stronger function of frequency, bias, temperature, time, film thickness,
and film crystal structure than is the dielectric constant of
paraelectrics.

6.2. Leakage
At equal values of specific capacitance in nF/cm2, the high-k
ferroelectrics will exhibit less leakage current than the lower-k
paraelectrics by virtue of their thicker films.

6.3. Capacitor Size

This should be as small as possible to maximize the number of
components per unit board area but, if the specific capacitance of
the film is too high, it may be difficult to fabricate a low-valued
capacitor, say 1 pF, in a size which is large enough to be easily or
reproducibly defined by standard board-level photolithography.
Figure 7 shows the length required for the side of a square integrated capacitor in mils to provide a given amount of capacitance
for various dielectric materials.
8

Square Plate Width (mils)

1,000

B,
BC

k=

4 2 m

7
2.
fe

100

/ep
r ro

y,
ox

0
=5

1 m 2000
1000
500
5000
2000
area of 1206
area of 0603
area of 0402

10

O5
Ta 2
1
0.001

0.010

k=

24

ec
el
rro
e
F

0.100

c ,k
tri

00
20

10

100

1000

Capacitance (nF)

Figure 7. Sizes required for various integrated capacitor

technologies.

6.4. Film Integration

At present, almost all ferroelectrics require a high-temperature

anneal in oxygen, at least 500C, to attain the crystal structure
necessary for high-k behavior. Paraelectrics do not require this
and are, therefore, easier to integrate on some substrates.
The integration of capacitors into interconnect substrates is
largely driven by three applications; decoupling of ICs from power
supplies, analog functions such as those in RF/wireless devices,
and termination of transmission lines. These capacitor characteristics affect their suitability for specific applications as described
below.

6.5. Decoupling
The ubiquitous need for high-frequency decoupling capacitors to supply the transient current requirements of fast logic devices is well documented35 ,47-50. Since the purpose of decoupling
capacitors is to provide low-impedance for a single clock cycle,
their most important characteristics are high capacitance and low
inductance. The values of these capacitors will typically be among
the largest on the substrate, from perhaps 20 to 1000 nF. Even
higher capacitor values may be required for other energy storage
applications. The inductance of an integrated capacitor is not a
function of the dielectric material.
Tolerance is not important as long as they provide a minimum
amount of capacitance. The resulting capacitance of the structure
as a function of film composition, morphology, and thickness, as
well as the capacitance as a function of frequency, bias, and time is
not an issue as long as the minimum capacitance level is maintained under all operating conditions. Ferroelectric films or ferroelectric powders in epoxy generally exhibit less stable and predictable electrical performance in all of these respects than paraelectrics
but their much higher dielectric constant can enable enough extra
capacitance to be designed in from the start to make up for these
shortcomings. Paraelectric materials are much more stable and
predictable but cannot achieve the high capacitance of ferroelectrics.
Table 3 shows the length in mils required for the side of a
square integrated capacitor to provide 100 nF. Film thicknesses of
2000 were chosen for most materials in this comparison since
this thickness may be practically achieved by deposition (sputtering or CVD), anodization, and sol-gel, although deposited films
may suffer from yield problems when they are this thin28. From
Table 3, low-k spin-on paraelectrics such as BCB or silica glasses
result in unacceptably large structures while the higher-k anodized or deposited paraelectrics can provide 100 nF in areas which
are larger than the footprint of a discrete but will fit in the real
estate under a chip. A 1 micron layer of cured epoxy + ferroelectric
powder on the high end of its k range has about the same specific
capacitance as the highest capacitance thin film paraelectrics. Only
thin film ferroelectrics can produce a decoupling capacitor that is
smaller than a discrete of similar capacitance, but this technology
has yet to overcome the problem of requiring a high temperature
anneal, which precludes it from use directly on anything but ceramic substrates.

The International Journal of Microcircuits and Electronic Packaging, Volume 23, Number2, Second Quarter 2000 (ISSN 1063-1674)
International Microelectronics And Packaging Society

177

Intl. Journal of Microcircuits and Electronic Packaging

Table 3. Square plate size to provide 100 nF for various

integrated capacitor technologies.

lack of stability and predictability with respect to frequency, film

thickness, film morphology, temperature, bias, and time. However,
the decrease in k with frequency and bias for some ferroelectrics
may be used as a tuning feature. Anodized films are very stable in
all these respects but may have specific capacitances too high to
be used in some low-value applications.

6.7. Termination of Transmission Lines

In order to provide 100 nF in a reasonably small area, the highest dielectric constants are called for and, since leakage is not of
major importance, the thinnest films which provide acceptable yield
may be used. A 1000 anodized Ta2O5 film would require 0.47 cm2
to provide 100 nF and, under 5 V bias, would leak approximately 3
A/cm2. The resulting power loss due to leakage is then only
about 7 W which would not overheat the capacitor or overtax the
power supply. A similar calculation for high-k ferroelectrics is
even more favorable since the films can be thicker to give the same
specific capacitance, which will result in less leakage per unit area,
and/or the plates can be made smaller in area.

6.6. Analog Functions

As system bus speeds increase, a significant number of lines

must be terminated to avoid signal reflection. In CMOS systems,
purely resistive termination may consume unacceptable amounts
of DC power. Thus, an AC termination scheme consisting of series R/C is preferred. Over 2000 discretes are used on a Pentium III
motherboard for termination51. Integrating R/C series terminators
such as 50 /100 pF or 100 /50 pF into either passive networks or
the motherboard itself, is highly desirable.
Exact tolerance of these capacitors is not critical, as long as
they have a significantly lower impedance than the series resistor
at the frequencies of interest. 5% tolerance is much tighter than
what is required. Leakage is also not an issue for capacitors in R/
C terminations as the expected leakage currents are very low compared to the drive capabilities of buffer outputs which are required
to drive transmission lines in the 50 range. Therefore, the choice
of dielectrics for termination is driven by the same considerations
for decoupling except that less capacitance is required.

7. Conclusions

In many analog functions such as RF/wireless, A/D conversion, filtering, and timing applications, required capacitor values
range from about 1 to 100 pF, several orders of magnitude below
Ferroelectric thin films have sufficient specific capacitance to
those needed for decoupling. For these values, capacitor dielecbe used as decoupling capacitors with advantageously small foottrics which give high specific capacitance due to either high k or
prints. Since dielectric constant stability is not too important in
thin films should be avoided if the resulting plate size would be so
this case, the ferroelectric thin films would excel in decoupling due
small that precision would be unobtainable with board-scale lithoto their much higher k values. The main problem with implementgraphic resolution. For example, a 1 pF capacitor realized with 3000
ing ferroelectrics on organic substrates is the need for high an of anodized Ta2O5 (about the thickest the film can be anodized)
nealing temperatures to form the high-k phases. Ferroelectrics are
would be only about 1.5 mils on a side and the same capacitor
less suitable for analog applications, since their dielectric conmade from 3 microns BaTiO3 with a k = 2000 would be 0.5 mils
stant is usually a strong function of frequency, film thickness,
across. Achieving 5% tolerance on a 1 mil x 1 mil capacitor plate
crystallography, temperature, voltage, and time which also makes
requires a photolithographic tolerance of 0.025 mils or 0.6 m. Thus,
them less attractive for use in analog functions where high tolerthe use of thick layers of relatively low dielectric constant materiance may be required. Also, many analog capacitors require valals (<10) may be better for these low-valued components so that
ues so low that, if made from high-k ferroelectrics, they may be too
the resulting devices are large enough to be fabricated with suffismall for high-tolerance fabrication.
cient dimensional tolerance to provide acceptable value tolerance.
The high-k paraelectric materials exhibit specific capacitances
BCB and SiO2 dielectrics are feasible dielectrics for this; 5 microns
that are just high enough for use in decoupling while still producof BCB would require 18 mils x 18 mils to give 1 pF which requires
ing capacitors which are as small as a chip or the larger surface
0.44 mils or 11 m for 5% tolerance. Thicker films are also favored
mount capacitors. Also, these dielectrics are currently more easfor most analog applications since leakage is very important and
ily integrated into a wide variety of substrates. Leakage is not
since higher yields will result. It may be possible to use the
very important in decoupling so the lower limit on film thickness
interlayer dielectric for these small capacitors thus saving a maskwill be driven by defect density. Anodized films are more desirable
ing step.
than sputtered films since the former exhibits a lower fatal defect
In addition to problems with capacitor size, ferroelectrics are
density which has a significant positive impact on yield with the
not as suitable as paraelectrics in these applications due to of their
large footprints required for decoupling. For analog functions,
The International Journal of Microcircuits and Electronic Packaging, Volume 23, Number2, Second Quarter 2000 (ISSN 1063-1674)
178

International Microelectronics And Packaging Society

Comparision of Paraelectric and Ferroelectric Materials for Applications as Dielectrics in Thin Film
Integrated Capacitors

thin film anodized paraelectrics are much more electrically stable

than ferroelectrics but, again, may not be easily fabricated in hightolerance values in the range of less than a few pF. BCB and SiO2
capacitors in micron-plus thicknesses would be sufficiently large
to be easily fabricated with precision and have very flat dielectric
constants with regard to frequency, film morphology, bias and
operating environment. For termination capacitors, 50 - 100 pF are
required and can be made at reasonable sizes with anodized
paraelectric dielectrics. They could be made smaller with high
capacitance density ferroelectrics, but the thin film resistor for the
R/C terminator would also have to be scaled down and this places
a lower limit on the achievable total R/C area.
Value requirements for integrated capacitors extend over six
orders of magnitude from about 1 pF to 1 F. For many applications, this entire range must be present on the same substrate with
the upper end serving as decoupling capacitors and the lower end
providing a variety of filtering, timing, coupling and A/D functions. It is difficult, if not impossible, for any single dielectric
material and capacitor configuration to provide this entire value
range without the resulting device being too large to be practical
or too small to be fabricated with sufficient value precision. The
best compromise might be to use either thin film anodized Ta2O5 or
ferroelectric/epoxy composite for decoupling and, perhaps, termination and either BCB or SiO2 for other functions. If either of these
two latter materials are present as an interlayer insulating material,
it might be possible to use it as a capacitor dielectric also without
having a additional deposition/patterning steps.

Acknowledgments
This work was supported by NASAs Jet Propulsion Laboratory (JPL) and the Defense Advanced Research Project Agency
(DARPA).

References

turing Technology, 1995 Japan IEMT Symposium, pg. 365, 1996.

5. J. Rector Jr., J. Dougherty, V. Brown, J. Galvagni, and J. Prymak,
Future Trend Towards Integral Passives, 17th Capacitor
and Resistor Technology Symposium, CARTS 97, pg. 1, 1997.
6. J. Anderson, Dielectrics, Chapman and Hall Ltd, pg. 49, 1964.
7. A. Von Hippel, Editor, Dielectric Materials and Applications,
John Wiley & Sons Inc., pg. 5, 1954.
8. W. Liu et al., Low-Temperature Fabrication of Amorphous
BaTiO3 Thin-Film Bypass Capacitors, IEEE Electron Device
Letters., Vol. 14, No. 7, pg. 320, July 1993.
9. P. Garrou, Polymer Dielectrics for Multichip Module Packaging. Proceedings of the IEEE, Volume 80, No. 12, December
1992.
10. J. Sergent and C. Harper, Editors, Hybrid Microelectronics
Handbook, Second Edition, McGraw-Hill, pp. 1-4, 1995.
11. B. Gnade, S. Summerfelt, and D. Crenshaw, Processing and
Device Issues of High Permittivity Materials for DRAMS,
Science and Technology of Electroceramic Thin Films; NATA
ASI Series, Auciello and Waswer, Editors, Vol. 284, pg. 373,
1995.
12. B. Lai and J. Lee, Leakage Current Mechanism of Metal-Ta2O5Metal Capacitors for Memory Device Applications Journal
of the Electrochemical Society, Vol. 146, No. 1, pg. 266, 1999.
13. C. Chaneliere et al., Dielectric Permittivity of Amorphous and
Hexagonal Electron Cyclotron Resonance Plasma Deposited
Ta2O5 Thin Films, Electrochemical and Solid-State Letters,
Vol. 2, No. 6, p. 291, 1999.
14. J. Scott, High-Dielectric Constant Thin Films for Dynamic
Random Access Memories (DRAM) Annual Review of Materials Science, Vol. 28, pg. 79-100, 1998.
15. A. Gitellson et al., Physical Properties of (Ba,Sr)TiO3 Ferroelectric Thin Films in Weak Electric Fields, Soviet Physics Solid
State, Vol. 19, No. 7, pg. 1121, July 1997.
16. Y. Tu et al., Synthesis and Electrical Characterization of Thin
Films of PT and PZT Made From a Diol-Based Sol-Gel Route,
Journal American Ceramic Society, Vol. 79, No. 2, pg. 441,
1996.
17. N. Tohge, S. Takahashi, and T. Minami, Preparation of PbZrO3PbTiO3 Ferroelectric Thin Films by the Sol-Gel Process, Journal American Chemical Society, Vol. 74, No. 1, pg. 67, 1991.
18. D. Liu et al., Integrated Thin Film Capacitor Arrays, Proceedings of the International Conference and Exhibition on High
Density Packaging and MCM 99, Denver, Colorado, IMAPS,
pg. 431, 1999.
19. U. Syamaprasad et al., A Modified Barium Titanate for Capacitors, Journal American Ceramic Society, Vol. 70, No. 7, pg. C147, 1987.
20. W. Merz, The Electric and Optical Behavior of BaTiO3 SingleDomain Crystals, Physical Review, Vol. 76, No. 8, pg. 1221,
1949.
21. S. Liang et al., Barium Titanate/Epoxy Composite Dielectric
Materials for Integrated Thin Film Capacitors, Proceedings of
the 1998 Electronics and Technology Conference, ECTC 98,
pg. 171, 1998.
22. R. Kambe, R. Imai, T. Takada, M. Arakawa, and M. Kuroda,

1. J. Rector Jr., J. Dougherty, V. Brown, J. Galvagni, and J. Prymak,

Integrated and Integral Passive Components: a Technology
Roadmap, Proceedings of the 1997 47th Electronic Components and Technology Conference, ECTC 97, pg. 713, 1997.
2. E. Logan, H. Clearfield, J. Young, and D. Bolton, Advanced
Packaging of Integrated Passive Devices for RF Applications,
Proceedings RAWCON 98, 1998 IEEE Radio and Wireless
Conference, pg. 289, 1998.
3. R. Frye, Integrated Passive Components: Technologies, Materials and Design, International Journal of Microcircuits and
Electronic Packaging, IJMEP, Vol. 20, No. 4, pg. 578, 1997.
4. K. Paik, Studies on Thin Film Integral Passive Components for
Mixed Mode Multichip Module (MCM) Applications, Proceedings of 1995 Japan International Electronic ManufacThe International Journal of Microcircuits and Electronic Packaging, Volume 23, Number2, Second Quarter 2000 (ISSN 1063-1674)
International Microelectronics And Packaging Society

179

Intl. Journal of Microcircuits and Electronic Packaging

mance and Applications of the Stealth Decoupling CapaciMCM Substrate with High Capacitance, Proceedings of the
tor, International Journal of Microcircuits and Electronic
International Conferences on Multichip Modules, MCM 94
Packaging, Vol. 20, No. 4, pg. 494, 4th Quarter 1998.
Denver, Colorado, pg. 136, 1994.
38. R. Ulrich, D. Nelms, L. Schaper, and D. Parkerson Anodized
23. S. Ogitani, S. Bidstrup-Allen, and P. Kohl, An Investigation of
Integrated Capacitors for High Density Microelectronics, High
Fundamental Factors Influencing the Permittivity of ComposDensity Interconnects, Vol. 1, No. 4, pg. 40, August 1998.
ite for Embedded Capacitors, Proceedings of the 49th Elec39. S. Venigalla, P. Bendale, and J. Adair, Low Temperature Electronic Components and Technology Conference, ECTC 99,
trochemical Synthesis and Dielectric Characterization of Barium
pg. 77, 1999.
Titanate Films Using Nonalkali Electrolytes, Journal Electro24. T. Kutty and P. Padmini, Mechanism of BaTiO3 Formation
chemical Society, Vol. 142, No. 6, pg. 2101, 1995.
through Gel-To-Crystallite Conversions, Materials Chemis40. P. Bendale, S. Venigalla, J. Ambrose, E. Verink Jr. and J. Adair,
try and Physics, Vol. 39, pg. 200, 1995.
Preparation of Barium Titanate Films at 55C by an Electro25. D. Hennings et al., Hydrothermal Preparation of Barium Titanate from Barium-Titanium Acetate Gel Precursors, Journal of
chemical Method, Journal of the American Ceramic Society,
the European Ceramic Society, Vol. 8, pg. 107, 1991.
Vol. 76, No. 10, pg. 2619, 1993.
26. P. Bendale et al., Preparation of Barium Titanate Films at 55C
41. N. Kim, K. Coates, G. Kunze, C. Chien, and M. Tanielian, Deby an Electrochemical Method, Journal American Ceramic
velopment of Multi-Chip Modules with Integrated Thin Film
Society, Vol. 76, No. 10, pg. 2619, October 1993.
Passive Elements, 1997 International Symposium on Micro27. R. Bacsa et al., Electrochemical Synthesis of Barium Titanate
electronics, ISHM 97, Philadelphia, Pennsylvania, 1997.
Thin Films, Journal Materials Research, Vol. 11, No. 1, pg.
42. NOVACAP Technical Brochure from www.novacap.com
194, January 1996.
43. A. Von Hippel, Editor, Dielectric Materials and Applications,
28. G. Morcan, S. Ang, T. Lenihan, L. Schaper, and W. Brown,
John Wiley & Sons, pg. 300, 1954.
Characterization of Thin Film Tantalum Oxide Capacitors on
44. H. Yoshino, T. Ihara, S. Yamanaka, and T. Igarashi, Tantalum
Polyimide Substrates, IEEE Transactions on Advanced PackOxide Thin Film Capacitors Suitable for Being Incorporated
aging, Vol. 22, No. 3, pg. 499, August 1999.
Into an Integrated Circuit Package, IEEE/CHMT 89 Japan
29. Z. Fu and Q. Quin, Pulsed Laser Deposited Ta2O5 Thin Films
IEMT Symposium, pg. 156, 1989.
as an Electrochromic Material, Electrochemical and Solid
45. Arita et al., Ferroelectric Nonvolatile Memory Technology
State Letters, Vol. 2, No. 11, pg. 600, November 1999.
with Bismuth Layer Structured Ferroelectric Materials, 10th
30. S. Byeon, and Y. Tzeng, Improved Oxide Properties by AnodInternational Symposium on Applications of Ferroelectrics,
ization of Aluminum Films with Thin Sputtered Aluminum OxNew Brunswick, New Jersey, IEEE, pg. 13, 1996.
ide Overlays, Journal of the Electrochemical Society, Vol.
46. Dan Scheck, Dow Corporation, private communication.
135, No. 10, pg. 2452, October 1988.
47. B. Sen and R. Wheeler, Performance Comparison of Discrete
31. S. Byeon and Y. Tzeng, High Performance Sputtered/Anodand Buried Capacitors, International Journal of Microcirized Tantalum Oxide Capacitors. Proceedings of the Internacuits and Electronic Packaging, IJMEP, Vol. 19, No. 4, pg. 449,
tional Electron Devices, IEDM 88, San Fransisco, California,
4Q 1996.
pp. 722-725, 1988.
48. J. Cain, The Effects of ESR and ESL in Digital Decoupling
32. K. Chen, M. Nielsen, S. Soss, and J. Rymaszewski, Study of
Applications, AVX Corporation, March 1997.
Tantalum Oxide Thin Film Capacitors on Metallized Polymer
49. T. Roy, L. Smith, and J. Prymak, ESR and ESL of Ceramic
Sheets for Advanced Packaging Applications, IEEE TransCapacitor Applied to Decoupling Applications, Proceedings
actions on Components, Packaging, and Manuracturing Techof the 7th Topical Meeting on Electrical Performance of Elecnology, Part B, Vol. 20, No. 2, pg. 117, May 1997.
tronic Packaging, IEEE, West Point, New York, pg. 213, Octo33. K. Chen, M. Nielsen, S. Soss, J. Rymaszewski, T. Lu, and C.
ber 1998.
Wan, Study of Tantalum Oxide Thin Film Capacitors on Metal50. R. Frye, Passive Components in Electronic Applications: Relized Polymer Sheets for Advanced Packaging Applications,
quirements and Prospects for Integration, International JourIEEE Transactions on Components, Packaging, and Manunal of Microcircuits and Electronics Packaging, Vol. 19, No.
facturing Technology, Part B, Vol. 20, No. 2, pg. 117, May 1997.
4, pg. 483, 4Q 1996.
34. R. Berry, P. Hall, and M. Harris, Thin Film Technology, D.
51. R. Heistand II et al., Advances in Passive Integration for C/RC
Van Nostrand Co., Inc., pg. 271, 1968.
Arrays and Networks with Novel Thin and Thick Film Materi35. L. Schaper, R. Ulrich, D. Nelms, E. Porter, T. Lenihan, and C.
als, Proceedings of the 36th IMAPS Nordic Conference,
Wan, The Stealth Decoupling Capacitor, 47th Proceedings
Helsinki, pg. 41, 1999.
of the Electronics Components and Technology Conference,
ECTC 97, San Jose, California, May 1997.
36. D. Nelms, R. Ulrich, L. Schaper, and B. Bowen Fabrication and
Performance of Integrated Capacitors on Silicon and Flexible
Substates Proceedings of the International Symposium on
Microelectronics, pg. 441, October 28, 1998.
37. L. Schaper, R. Ulrich, and D. Nelms et al., Fabrication, PerforThe International Journal of Microcircuits and Electronic Packaging, Volume 23, Number2, Second Quarter 2000 (ISSN 1063-1674)
180

International Microelectronics And Packaging Society