journal J Am Cefam SOC, 73 [S) 323-28 (1990)

Energy Storage in Ceramic Dielectrics

Gordon R. Love*,*
Sprague Electric Company, North Adams, Massachusetts 01267

Historically, multilayer ceramic capacitors (MLC’s) have Recent improvements in the manufacturing process for
not been considered for energy storage applications for two MLC’s have permitted dramatic increases in the short-time
primary reasons. First, physically large ceramic capacitors ultimate breakdown voltage stress of commercial devices.
were very expensive and, second, total energy density ob- While not the subject of this paper, these improvements
tainable was not nearly so high as in electrolytic capacitor incorporate significantly better control of dielectric stoi-
types. More recently, the fabrication technology for MLC’s chiometry, higher and more reproducible fired density, and
has improved significantly, permitting both significantly significantly “smoother” buried electrode layers. The latter
higher energy density and significantly lower costs. Simul- two differences act to make the peak voltage stress on the
taneously, in many applications, total energy storage has be- dielectric a smaller multiple of the average or design voltage
come smaller, and the secondary requirements of very low stress and are thought to be particularly important even to
effective series resistance and effective series inductance short-time, room-temperature, results. Further, a number of
(which, together, determine how efficiently the energy may requirements for energy storage devices exist in which rela-
be stored and recovered) have become more important. It is tively high voltages are desirable, and the highest possible en-
therefore desirable to reexamine energy storage in ceramics ergy density (per unit weight or per unit volume) is essential.
for contemporary commercial and near-commercial dielec- High voltage (300 V and higher) means that existing state-of-
trics. Stored energy is proportional to voltage squared only the-art fabrication technology for ceramics is appropriate and
in the case of paraelectric insulators, because only they have that conventional electrolytic devices (based on A1 or Ta) are
capacitance that is independent of bias voltage. High dielec- not well suited. We therefore have reexamined an array of
tric constant materials, however, are ferroics (that is ferro- commercial and near-commercial dielectric formulations to
electric and/or antiferroelectric) and display significant see what energy densities might practically be achieved.
variation of effective dielectric constant with bias voltage.
The common ferroelectric materials, whether based upon 11. Samples
barium titanate or lead manganese niobate (PMN), in the
high-field limit, exhibit an energy storage which increases All test devices were MLC’s having design dielectric thick-
linearly with bias voltage. Mixed phase, ferroelectric plus ness (electrode spacing) from 8 to 200 pm. Thin dielectrics
antiferroelectric, dielectrics from the lead lanthanum zir- are attractive experimentally because smaller power supplies
conate titanate (PLZT) system, as predicted theoretically, and less elaborate test fixturing are suitable. They are also
show the best energy density at low to moderate fields. Sur- useful because they are somewhat more likely to be defect-
prisingly, maximum energy storage is not obtained in high free and hence to survive to higher voltage stresses. Thicker
dielectric constant materials but in those materials which dielectrics were examined in less detail to verify that the ma-
display intermediate dielectric constant and the highest ulti- terials parameters were truly independent of the fabricated
mate breakdown voltages. [Key words: capacitors, dielectric structures. In the thin dielectric devices in particular, the
materials, multilayer, ferroelectrics, storage.] electrode thickness was a not-trivial fraction of the electrode
separation; electrode thickness and uniformity was estimated
I. Introduction using quantitative microscopy,+but precision measurement of
1.5- to 3.0-pm layers deposited as thick films and cofired with

D IELECTRICS for multilayer ceramic capacitor (MLC) de-

vices have historically been extremely conservatively
rated. Whereas the ultimate dielectric breakdown strength of
the ceramic is extremely difficult. The average dielectric
thickness is believed to be known to +7% in the worst case
(see the Discussion below), but no attempt was made to esti-
fully dense ceramic (single crystals and melt-grown polycrys- mate the minimum dielectric thickness, because no satisfac-
tals) has been several tens of volts/micrometer (MV/m), typi- tory sampling method could be devised. A small number of
cal design voltage stress has been 2 V/pm.l One explanation samples obtained commercially were tested to attempt to de-
for this is practical: manufacturing defects have been suffi- termine the generality of the measurements; the effort was in
ciently common and sufficiently severe that the actual ob- no sense a competitive product evaluation, so no systematic
servable breakdown voltage was extrinsic at a relatively small sampling of commercial sources was contemplated, and no
fraction of ultimate (say 5 V/pm). The second explanation is distinction will be made between proprietary and commercial
commercial: the vast majority of applications for MLC‘s are at sources of the dielectrics examined.
such low voltages that the manufacturing process itself limits
the voltage stresses on the parts. That is, parts for use in 5-V 111. Experimental Procedure
or less circuits must nevertheless be manufactured to 50-V or
greater design rules. Both of the measurement procedures described by Burn and
Smyth2 were used. At least for ferroics, there is a significant
difference between the two measurement methods, resulting
R. McSweeney- contributing edit01 from the fact that ferroic materials exhibit relatively long
time-constant “relaxation.” The method based upon super-
posing a 1 V ac “ripple” on the dc bias provides an essentially
“static” estimate of stored energy, because the relatively slow
Manuscript No. 199484. Received June 16, 1987; approved August 20,
1989.
*Member, American Ceramic Society.
*Now with Alcoa Technical Center, Alcoa Center, PA. ‘Quantimet, Cambridge Instruments, Boston, MA.

323
324 Journal of the American Ceramic Society - Love Vol. 73, No. 2

relaxation of effective dielectric constant with bias voltage is decreasing field), but the physical significance of the derived
allowed to approach completion. The method of pulse dis- parameters from this experiment is not clear. On the other
charge: by contrast, is “dynamic” in that measurement time hand, the “dynamic” measurement of stored energy is highly
constants are short relative to relaxation time constants. “topical”; the numbers are sensitive to the rate of charging,
The two measurement techniques impose different require- the dwell time at bias before discharge, and, perhaps, the rate
ments on the test devices. The “static” measurement in fact of discharge.
measures two capacitors in series, the test device of interest The three physical parameters which are most frequently
and a large capacitance value part used to protect the bridge used to characterize dielectric materials are the permittivity
from the dc bias voltage. When the test capacitor is small rela- in F/cm (or the dielectric constant which is dimensionless),
tive to the blocking capacitor, the bridge realdings are trivially the polarization in C/cm2, and the energy density in J/cm3.
different from the values of the test capaci1,or. The blocking These may be defined in more than one way, depending on
capacitor used throughout these measurements was a 10-pF, the primary data. We stress that each of these “constants” is
2.5-kV, paper/oil capacitor, and the largest capacitance value useful for intercomparing results from samples of widely dif-
device tested in this apparatus was 0.33 pF. ‘Theseparameters ferent dimensions and designs, but they are not constant and
would introduce a measuring error of 2.2545, and are consid- their physical significance is not clear.
ered “worst case.” The “static” test yields capacitance as a function of bias
On the other hand, accuracy in the “dynamic” measure- voltage stress. We adopt the notation C , as a reminder that
ment is greatly facilitated by actually storing and recovering the measured capacitance is voltage-dependent. Then, a de-
significant amounts of energy. In other words, large capaci- fining equation for effective dielectric constant, K,, as a func-
tance values, large bias voltages, and large series resistances tion of bias voltage stress, v, is
(long discharge time constants) all combine to improve the
test results. For this reason, although single dielectrics could
be examined in either test method, single samples, in general,
could not. Typical samples for these tests were 1.5 to 17 pF at
in which 1 is the electrode spacing, A the electrode area,
voltages of 400 to 1200 V.
Two methods were used to estimate the ultimate dc break- the permittivity of vacuum, and K Oand C o are the dielectric
down voltage for all devices. When sufficient parts were avail- constant and capacitance, respectively, at zero bias.
able from a single manufacturing lot, 50 pieces were tested to From the definition C = dD/dv, the apparent polarization is
catastrophic failure, a straight line was fitted to the cumula-
tive failure curve (in Weibull coordinates): and the 5% inter- D, = K”eo C,/Codv
cept of this line was the reported “breakdown.” For smaller
sample sizes, testing of capacitance versus, bias voltage was Finally, the stored energy is given by
done first and was carried out until the parts failed or the dc
bias reached 750 V. Then the remaining palrts were tested to
failure. The reported value in this case too is the estimated
5% failure point, but it is much less well-known because of
E , = Koso
I
C,/CoVdV

In the “dynamic” measurement,’ the primary data are volt-

(3)

the smaller sample size.

All testing was done at uncontrolled room temperature age, v, and current, i, as a function of time. In this case the
(23” ? 3°C); all ac was 1 kHz, 1 V rms. Priimary data acquisi- same three derived parameters have the following, quite dif-
tion in the pulse discharge test was through a storage oscillo- ferent, form. The dielectric constant is given by
scope.*Data reduction and graph generation used commercial
software’ supplemented by numerical integration (Simpson’s (4)
rule) and differentiation (with simple statistical “smoothing”)
in BASIC. where A is the active electrode area, R the external load re-
sistance, and t the time; and the derivative must be evaluated
IV. Discussion at the appropriate values of v.
The polarization is more transparently given by
There is an extensive literature on energ:y storage in dielec-
tric materials’ which need not be reviewed Ihere. The reader is
reminded of two points which are of particular importance
for these experiments, or indeed any experiments on materials
upon which many commercial MLC capacitors are based. and the stored energy by
First, there are multiple mechanisms having significantly dif-
ferent time and temperature dependence for energy storage in
ferroics. The measured stored energy will therefore depend
strongly on the “time constant” (in the broadest sense) of the The most severe ambiguity in these expressions is in the
measurement. Second, the appropriate samples for the two low voltage limit, both because measurement errors in the
kinds of measurements we report are quite different. Inter- i and v become significant, and because the longer time-
comparison of the results of the two kinds of measurements constant phenomena intrude. These give very high apparent
involves, therefore, a fairly significant extrapolation of either K , near v = 0 but do not greatly distort the apparent D and E
the physical or the electrical parameters. results, since i is very small.
The so-called “static” test was concentrated on physically
small capacitors with electrode spacing as small as 8 pm. This
allowed voltage stresses in excess of 80 V/rim despite a mod- V. Results
est power supply. Further, the dc bias voltage is imposed for Preliminary experiments were carried out to establish that
“long” times (5 min, typical) before the apparent capacitance the derived parameters were indeed “fundamental” in the
is recorded. This eliminates hysteresis from the data (points limited sense that they were device-independent and there-
taken in an increasing field superimpose on points taken in a fore could be used to predict performance across a broad
range of design parameters. Capacitors were built from a sin-
‘Nicolet Explorer 111, Nicolet Instrument Corp., Madison, WI. gle dielectric lot, with electrode spacings varying from
‘Number Crunching Statistical System, J. Hintze, Kaysville, UT. 8 to 44 pm and nominal capacitance values of 90 to 330 nF.
February 1990 Energy Storage in Ceramic Dielectrics 325

FA. Thickness, pm

0 8
0 10 3 1
.0
. 0

.
2 A A 27
A 44 0
7 0 A
5 %AA
X
2 0

.
’B Thickness, pm

1- 0 0 8
10
O4 0
.+o 0,
0
’t A 27
A 44

I I I I
I
10 20 30 40
Bias, Vlprn 0 10 20 30 40
Bias, Vipm
Fig. 1. Effective dielectric constant as a function of bias
voltage stress for C08 capacitors of the indicated electrode Fig. 3. Energy storage as a function of bias voltage stress
separation. for the devices of Fig. 1.

They were tested with the “static” procedure and compared of breakdown voltage exceeding 50 V/pm, the total stored en-
as shown in Figs. 1to 3. It is apparent that the derived param- ergy approaches 1 J/cm3. Because of its high breakdown volt-
eters are indeed device-independent. The few high points in age, an example of this dielectric type ((216) is illustrated with
the K, plot, Fig. 1, near 4 V/pm may be attributed to experi- dissimilar dielectrics later.
mental error. This dielectric contains both ferroelectric and Dielectrics of the Z5U class (Figs. 4 to 6) are relatively
antiferroelectric phases (see below), and, in the preliminary “pure” ferroelectrics. That is, they are relatively coarse-
tests, insufficient relaxation time under dc bias was permitted grained, single-phase, solid solutions of barium titanate
before recording apparent capacitance. with just enough substitutional additives to shift the Curie
There was great variability among the dielectrics in break- temperature to 10” to 20°C. They have relatively high small-
down voltage stress, so much, indeed, that graphic compari- signal dielectric constants, which vary strongly with tempera-
sons among the dielectrics are very difficult. In the remaining ture, time and, most importantly for this study, bias voltage.
figures the dielectrics are “clustered” by breakdown voltage Despite the high initial dielectric constant, KO,of this Z5U
for graphing convenience. dielectric, the maximum stored energy is quite modest. The
Dielectrics of the COG (NPO) family are not ferroic (or, if rapid loss of effective dielectric constant, K,, with bias voltage
they contain any ferroics, are far removed from any non- leads to the effective polarization, D,, being little higher than
linearities). They display relatively low small-signal dielectric that of dielectrics with far lower KO.Equally important, the
constant which is essentially independent of temperature, fre- ultimate breakdown voltage stress is modest (10 V/pm), so the
quency, storage time, and bias voltage stress. The behavior is integrals defined in Eqs. (2) and (3) must be evaluated at low
quite like the introductory textbooks: K independent of bias,
D linear in bias, and E quadratic in bias. However, by virtue

U
.A
0

.-
-z
0
a
100 -
.
Thickness, pm
0 8
10
A 27
A 44
-

I I I I I I
I I I
2 4 6 8 10 12
0 Bias, Vlpm
0 10 20 30 40
Bias, Vlpm Fig. 4. Effective dielectric constant as a function of bias
voltage stress for a variety of ZSU dielectrics. C11 is a re re
Fig. 2. Dielectric polarization as a function of bias voltage sentative BaTiOs material, and PMNl and PMN2 are Lad
stress for the devices of Fig. 1. magnesium niobates.
326 Jouiwal of the American Ceramic Society -Love Vol. 73, No. 2

300 Table I. Optimum Designs in Various Dielectrics

Breakdown Clem’ Energy at
Dielectric stress r unbiased I kV
type (Vw) K (rm) (uF) (J)
C16 (COG) 65 75 18.5 1.75 0.88
N
E
c11 (ZSU) 13.2 7500 95 5.7 0.02
5 200 - PMN-a (ZSU) 10 11600 120 5.8 0.02
5.8
sE PMN-b (Z5U)
PLZT(X7T)
11
16
9700 109
2800 70 3.8
0.02
0.71
C
C08 (X7R) 30 2000 38 10.9 1.40
.-0 C20 (X7R) 40 2000 30 17.4 1.34
.-w
c)
EXD(X7R) 90 1800 20 35.9 4.82
k -
a
n
100
A C11 in the figures and Table I) of intermediate “ K in which the
A PMNl antiferroelectric phase is selected to enhance the low-field
0 PMN2 stability of capacitance with bias voltage. The second (PLZT)
is higher in K, with less temperature stability (roughly Y7T).
Indeed, each of these dielectrics displays enhanced polariza-
I I I I I tion and energy density attributed to the presence of the anti-
0
2 4 6 8 10 12 ferroelectric phase. Together they display the highest energy
Bias, V/pm densities observed in the range of bias stresses from 10 to
Fig. 5. Dielectric polarization as a function of bias voltage
30 V/pm. Once again, the higher-K dielectric actually stores
stress for the dielectrics of Fig. 4. less maximum energy, because it breaks down at lower stress.
The two PLZT-based dielectrics displayed the highest
value of polarization observed in this study and the greatest
voltage stresses. It is possible that both these results trace to
gain of energy with incremental bias. This appeared to con-
the relatively large grain size in this material. It is frequently
firm the utility of the antiferroelectric phase in improving en-
observed that large-grained ferroelectrics, although of high
ergy storage. However, each also is a relatively fine-grained
dielectric constant, display rapid saturation of polarization
dielectric with good sintered density. It is possible that sig-
with bias voltage and low breakdown voltages.
nificantly improved energy storage can be achieved in any
Certainly these observations were echoed in the other
fine-grained, high sinter density, ferroic system. To clarify
very high “ K dielectrics studied. Two dielectric compositions
this postulate, two additional X7R dielectrics were examined.
meeting the Z5U temperature characteristic but based upon
The first of these is C20, a relatively conventional barium
the lead-magnesium-niobate (PMN) system were also exam-
titanate which achieves X7R performance by a combination
ined (Figs. 4 to 6). Again, despite their high dielectric con-
of fine grain size and controlled chemical heterogeneity.’ It
stants at low bias, they have a low slope of stored energy
does not have as high a D , as the PLZT-based X7R but, by
versus bias, and they break down at relatively low stress and
virtue of its excellent breakdown voltage, achieves ultimate
therefore have relatively low maximum energy density.
energy storage comparable to the two PLZT-based dielectrics.
It has been suggested that energy storage in ceramic dielec-
The second of the barium titanate-based dielectrics, Exp,
trics could be greatly increased by incorporating a phase dis-
achieves X7R performance with a much greater reliance on
playing antiferroelectricity.6 This suggestion follows from the
grain size control (its fired grain size is less than 1 pm).
observation that an antiferroelectric may store significant po-
Although its saturation polarization is less than that of the
larization charge at nonzero bias; equivalently, it has a dielec-
PLZT-based material, its ultimate breakdown is so high that
tric constant which, over some voltage bias range, increases
it displays easily the best ultimate energy density observed in
with bias. There are two commercial dielectrics which exploit
this study. These observations are summarized in Figs. 7 to 9.
this phenomenon. One is an X7R characteristic (called C08
Also included in these figures, for reference only, is a typical
NPO dielectric, C16. Samples were not available in all dielec-
trics for the “dynamic” measurement of stored energy. In the
1.o I 1 I I 1 NPO dielectric, for example, a 1-pF capacitor would be very
A C11 large. In the various Z5U dielectrics, although large capaci-
A PMNl
0.8 0 PMN2
4

.
0

E,
3
-a 0.6 3

0 ?
0 C16
7
2
x 2
i
0.4 Y
C
w 1

0.2
0
0 10 20 30 40 50 60 70 80 90 100
Bias, V/pm
0.0 I- Fig. 7. Effective dielectric constant as a function of bias voltage
0 2 4 6 8 10 12 stress for a variety of higher-breakdown-voltage dielectrics. C16
Bias, Vlpm is a typical COG capacitor dielectric; C08 and PLZT are commer-
cial and experimental lead-lanthanum-zirconate-titanate dielec-
Fig. 6. Energy storage as a function of bias voltage stress trics res ectively; C20 and Exp are commercial and experimental
for the dielectrics of Fig. 4. BaTi03-gased dielectrics, respectively.
February 1990 Energy Storage in Ceramic Dielectrics 327

400 I I I I I I I I I I 1

300 - A “Stat“
0 “dyn“

200 -
%
% 200
100
0 C16
- 8
E

0
0 10 20 30 40 50 60 70 80 90 100
Bias, Vlpm
Fig. 8. Dielectric polarization as a function of bias voltage stress
for the dielectrics of Fig. 7.

6 I I I I I I I I I

0 10 20 30
Bias, V/pm

Fig. 11. Comparison of the apparent dielectric polarization

i
of C20 dielectric using “static” and “dynamic” measurement
g2 techniques.
C
w
1 compared in Figs. 10 and 13. As one would expect, samples in
the “dynamic” test return less energy than predicted by the
1 I I I
0
0 10 20 30 40 50 60 70 80 90 100
“static” test, and this leads to a lower estimate of effective
Bias, Vlpm
dielectric constant at high bias. On the other hand, the appar-
ent dielectric constant at low bias voltage is abnormally high
Fig. 9. Energy storage as a function of bias voltage stress for the in the “dynamic” measurement. It appears that this is a con-
dielectrics of Fig. 7. sequence of slow discharge at low bias voltage-one of the
long time-constant energy storage modes.
tance values are easily obtained, the very poor energy storage In Table I, major observations regarding the several dielec-
results discouraged building the necessary test devices. In trics are summarized. There are three general and striking
the dielectrics identified as C20 and CO8, representative of observations contained in that summary. First, high initial
the better materials, samples suitable for “dynamic” measure- dielectric constant is not at all a good indicator of energy stor-
ments were available. For these samples, a series of charge/ age capability; the lowest “ K dielectric in the study equalled
discharge cycles to progressively higher voltages yielded esti- or exceeded the three highest “K” dielectrics in energy stor-
mates of stored energy for comparison with the earlier data. age capability. Second, ultimate dielectric breakdown voltage
Examples of these data are presented as Figs. 10 to 15. The stress is a pretty good indicator of energy storage capability,
largest discrepancy between the methods arises in the appar- particularly among dielectrics of even roughly comparable
ent dielectric constant as a function of bias voltage, which is “K’: Third, improvements in processing are yielding dielec-

c20
3.0

A “Stat“
0 “dyn“

2.0 L

i
1.o E
W l

0.0
0 10 20 38
Bias, Vlpm Bias, Vlpm
Fig. 10. Comparison of the apparent dielectric constant of Fig. 12. Comparison of the apparent energy storage in
C20 dielectric using “static” and “dynamic” measurement C20 dielectric using “static” and “dynamic” measure-
techniques. (See text for details.) ment techniques.
328 Jownal of the American Ceramic Society - Love Vol. 73, No. 2

COB C08
3 I

LI ”stat“ A - “Stat“
0 “dyn” 0 -“dyn”

2
p?
z
Y

0
0 10 20 30
Bias, V/pm 0
0 10 20 30
Bias, Vlpm
Fig. 13. Comparison of the apparent dielectric constant
of C08 dielectric, using “static” and “dynamic” measure- Fig. 15. Comparison of the apparent energy storage in C08
ment techniques. (See text for details.) dielectric, using “static” and “dynamic” measurement
techniques.

trics with excellent ultimate breakdown voltage stress and

hence energy storage capacity. density for a solid tantalum capacitor (based upon catalog
values and dimensions) is 0.1 J/cm3 and for an aluminum
electrolytic capacitor is 0.12 J/cm3.
VI. Summary This is of course a very aggressive rating for the ceramic
The investigation reported here has dernonstrated that di- dielectrics and might or might not provide adequate reliability
electric constant, saturation polarization, and ultimate break- in a particular application. With the single exception of the
down voltage stress all play an important parametric role in NPO dielectric, any other choice of ratio of rated voltage to
energy storage in ferroic dielectric materials. The final breakdown voltage will not change the relative ranking of the
column of Table I is an attempt to compare these dielectrics dielectrics, since all have stored energy which is linear in bias
in “real” energy storage devices. For the salkeof the compari- voltage stress.
son, the maximum energy density displayeld in these tests was Another, and perhaps more important, observation comes
reduced by reducing the maximum voltage stress to 80% of out of this investigation. Modern manufacturing technology
breakdown stress, and normal design rules were then fol- is more and more able to produce “ideal” MLC‘s with high
lowed to postulate a Lcm3 total volume capacitor. To put sintered density, fine-grained dielectrics of controlled chem-
these numbers in engineering perspective, comparable energy istry, and electrodes which much more nearly approximate
planar “foils.” It may no longer be necessary to design devices
for 2 V/pm bias voltage stresses. In fact, if physically larger
devices can be fabricated at acceptable yields to perform reli-
C08 ably in the 20 V/pm range of bias voltages, entire new areas
of application become available to electronic ceramic dielec-
tric materials.
A more detailed study of these dielectrics in terms of
their reliability in high-stress service is both highly desirable
and beyond the scope of this report. It is worth noting that
all these data support the assumption that failures result
from extrinsic (manufactured structure) rather than intrinsic
(materials-limited) defects; there is reason to expect excellent
demonstrated reliability in these kinds of designs.

References
IW. B. Pennebacker, “Radio Frequency Sputtered Strontium Titanate
Films,” IBMJ. Res. Deu, 13, 686-95 (1969).

4
zI. Burn and D.M. Smyth, “Energy Storage in Ceramic Dielectrics,”
A “stat” J. Mafer. Sci., 7, 339-43 (1972).
mi “dyn” 31. Burn, “Field-Enforced Ferroelectricity in Glass-Bonded Lead Zir-
conate,”Am. Ceram. Soc. Bull., 50 [5] 501-505 (1Y71).
4W. Weibull, ‘A Statistical Distribution Function of Wide Applicability,”
J. Appl. Mech., 18, 293 (1951).
SN. Klein, Electrical Breakdown in Solids, Ad% Elecfron. Elecfron Phys.,
10 20 26, 309-420 (1969).
Bias, Vlpm 6B. Jaffe, ‘Antiferroelectric Ceramics with Field-Enforced Transitions:
New Nonlinear Circuit Elements,” Proc. IRE, 49, 1264-67 (1961).
Fig. 14. Comparison of the apparent dielectric,polarization ’G. Maher and S.E. Stefanik, “Method for Manufacturing a Ceramic
of C08 dielectric, using “static” and “dynamic” measure- Body Having Multiple BaTiO3 Phases,” U.S. Patent application No. 885509,
ment techniques. July 14, 1986. 0