Here’s a draft of my upcoming column for CircuiTree magazine,
probably to run in March. I like to show these around to my
colleagues in the business to make sure they are as accurate as
possible, so please look it over in the next few days and let me know
if there is anything you think needs changing. I'll be in the EPUG
conference call tomorrow, too.
- Rick Ulrich
for “Embedded Passive Update” column, CircuiTree, March 2006
Ferro-Filled Polymer Dielectrics: Promises and Problems
Rick Ulrich
It would be the perfect board-level embeddable dielectric. A screen-
printable paste, storable for months, processable by standard methods
familiar to all board shops, curable in minutes at mild temperatures,
low waste, no degrading effects on other layers, and high
capacitance. And it exits for you to use today. Well, all except
for that high capacitance part.
The idea is simple: mix a high-k ferroelectric powder into a curable
polymer binder, screen print, and cure in place. For example, BaTiO3
can be produced in bulk as submicron particles with dielectric
constants in the thousands. For the binder phase, I know of no
polymer with a k higher than about 12, so the selection of the binder
phase is driven by usability considerations such as printability,
cure conditions, and stability. The natural choice is either an
epoxy or a polyimide, both with k’s in the range of 3 to 4. So let’s
mix barium titanate particles with k = 10,000 and an epoxy with k =
3. Naturally, you want as much high-k filler as possible, so how
much can you put in? The densest possible spherical packing is
hexagonal close-pack at 74% by volume which, as every teenager that
works in the grocery store knows, is how you stack oranges so they
won’t roll away. Since barium titanate has a density near 5.9 and
epoxy is usually a little less than one, the mixture ends up about
95% by weight of the high-k phase.
The advantage of this approach is that much of the processing, and
all of the high temperature steps necessary to get high k from the
ferroelectric phase, can be done in advance of application to the
organic substrate. Processing is additive so there is no patterning
and little waste. No vacuum equipment is required and cure
temperatures are comfortably low for the rest of the board and
components. Because the films are thicker than sputtered, sol-gel or
CVD, the working voltages are higher, on the order of 100’s of volts,
and leakage at common board voltages is almost too low to be
measured. Pinholes can be eliminated through multiple printings.
But, despite the overwhelming preponderance of k = 10,000 phase over
k = 3 phase, the overall dielectric constant of the random mixture
will end up being only about 10 - 40, much closer to that of the low-
k phase. You can understand why the mixing rules are not kind to
this approach if you imagine drilling a molecule-wide hole down
through this compound dielectric from one plate to the other. You
would alternately pass through regions of high-k and low-k materials,
and that’s exactly what the electric field between the plates sees:
dielectrics in series. As capacitors are placed in series, the
overall value drops, and the same effect causes the overall k of a
randomly-dispersed composite material to be close to the lower-k
phase. It doesn’t matter which one is dispersed and which is
continuous, each field line sees this as alternating stacked
dielectrics. A printed 10 micron thick film would deliver about 1 to
4 nF/cm2, and that’s about the most you can get from this approach.
Attempts have been made to increase the high-k loading by using a
multi-disperse set of filler sizes and shapes, with the idea that the
small particles will nestle in between the big ones, and this can
give up to 85% by volume or about 98% by weight. But this still does
not increase the overall capacitance density very much and also
creates problems of printability, adhesion and mechanical stability
at such high solids loadings. Another problem is that screen
printing or stenciling is not amenable to tight tolerances and there
is no technology currently available for continuous trimming of
embedded caps. Trim tabs can always be used to decrease the value in
a stepwise fashion by removing capacitor area, but this method
requires extra area of its own, further decreasing the effective
capacitance density.
But it’s still a good idea. It does provide an order of magnitude
more capacitance than unfilled polymers, but two orders of magnitude
less than the more-expensive thin-film paraelectrics (up to about 300
nF/cm2) and pure ferroelectric films (1000’s of nF/cm2). The trick
to increasing capacitance is to somehow stack the high-k phase so
that it is vertically continuous over as wide an area as possible.
Some kind of self-assembly or self-orientation would be necessary,
then we’ll have a technology that can significantly advance the cause
of embedded passives.
******************************************************
Dr. Richard Ulrich
Professor
Dept. of Chemical Engineering
3202 Bell Center
Univ. of Arkansas
Fayetteville, AR 72701
[log in to unmask]
(479) 575-5645
******************************************************
|