Determining dielectric constants using a parallel plate capacitor

T. T. Grove, M. F. Masters, and R. E. Miers

Department of Physics, Indiana University-Purdue University Fort Wayne, Fort Wayne, Indiana 46805
共Received 27 January 2004; accepted 22 July 2004兲
The availability of low cost digital multimeters capable of measuring capacitance has made parallel
plate capacitor investigations common in the introductory laboratory. Typically, students add
multiple thin sheets of dielectric material between conducting plates. The capacitance is measured
and then plotted versus the reciprocal of the dielectric thickness 共the nominal plate separation兲. We
explain why the experiment fails for small dielectric thickness and discuss an improved version of
the experiment using dielectric sheets of various thicknesses rather than multiple thin sheets of a
dielectric. © 2005 American Association of Physics Teachers.
关DOI: 10.1119/1.1794757兴

I. INTRODUCTION has a dielectric constant that can be as low as ␬ ⫽2. In short,

we cannot rely on a specific dielectric constant for paper.
Because the relation between capacitance and plate sepa- Because of these concerns, we selected Teflon rather than
ration for an ideal parallel plate capacitor is very simple, it is paper for our dielectric material. Teflon is available in thick-
desirable to do experiments to examine this relation. There nesses 共and tolerances兲 that are similar to paper. The dielec-
have been many articles relating to capacitors and dielectric tric constant for Teflon at 22 °C for frequencies between 100
materials.1–10 A frequent purpose of these experiments is to Hz to 1 MHz is ␬ ⫽2.1.16 The manufacturer’s product de-
verify the relation between the capacitance and the plate scription lists ␬ ⫽2.0 at 25 °C for frequencies between 100
separation based on the parallel plate capacitance equation:
Hz and 1 MHz.17 We are uncertain why Refs. 16 and 17
␬ ␧ 0A differ, because other listed values from each reference indi-
C⫽ , 共1兲 cate that ␬ does not vary greatly with temperature.
d
where C is the capacitance, ␬ is the dielectric constant, ␧ 0 is
II. EXPERIMENT 1
the permittivity of free space, A is the area of the plates, and
d is the plate separation. A common approximation in ca- The parallel plate capacitors we construct are simple and
pacitors is that the dielectric material fills the void between inexpensive. The construction is diagramed in Fig. 1. A
the two conducting plates, which implies that the dielectric single sheet of ordinary household aluminum foil
thickness, D, is equal to d, the plate separation. A common (⬃30.3 cm width兲 is placed on a piece of particle board 共ap-
student task is to determine ␬ and verify the linear relation proximately 1.6 cm thick, 29 cm long, and 35 cm wide兲. The
between C and the inverse dielectric thickness, 1/D. Typi- foil is rolled flat with paint rollers and any foil that is
cally, the dielectric thickness is varied by using multiple thin crinkled or folded is replaced by fresh foil. A number of
sheets 共of the same nominal thickness兲 of dielectric material. Teflon sheets 共larger in area than the foil兲 are placed on top
However, this experiment consistently results in a nonlinear of the foil. The Teflon film sheets18 are 0.050⫾0.013 mm
relation between capacitance and inverse dielectric thickness thick 共nominal 0.002⬙ thickness with a 0.0005⬙ tolerance17兲.
and yields dielectric constants that are much too low pre- The thickness of each individual sheet and the total dielectric
dominantly because of air trapped between the layers of the thickness of all the sheets of dielectric are measured using a
dielectric. micrometer accurate to 0.005 mm. The results of these mea-
Many instructors choose paper as the dielectric material. surements are consistent with the specifications given by the
Paper has the advantage of being thin 共small plate separa- manufacturer. A second sheet of aluminum foil 共identical to
tions result in large capacitances兲 with a reasonably uniform the first foil sheet兲 is placed over the Teflon for the top plate.
thickness. Most introductory texts give a value of roughly 3 To flatten the capacitor, a second piece of particle board is
for the dielectric constant ␬ of paper. However, the values placed on top of the foil and sheets. To increase the flatten-
given for ␬ in various textbooks range from 1.7 to 4.0,11–15 ing, a load mass can be distributed across the top of the
and none of the textbooks provide a reference to an original particle board. The top foil is folded back over the particle
source. board 共see Fig. 1, side view兲 so that the top and bottom foils
There are multiple factors that can cause a variance in the are nearest each other under the particle board. Portions of
dielectric constant of paper. Paper can be wood-based, rice- the foil that are not between the two boards are far enough
based, or rag-based, for example. Different inks and bleach- apart so their contribution to the total capacitance is negli-
ing processes are used in its production as well as various gible. Care is taken that the top piece of foil is directly over
surface finishes. Wood-based paper 共presently the most com- and aligned with the bottom piece. By this arrangement of
mon paper兲 undergoes a drying process and chemical process the two foils, the area of the capacitor is determined by the
with aging. Residual acids left on the surface from the manu- width of the foil and the length of the particle board.
facturing process cause the paper to yellow and become drier We then measure the capacitance by connecting the foil
over time. In addition, paper is hygroscopic, and the dryness sheets to a digital multimeter19 by means of two 30 cm long,
of paper is very important for the value of the dielectric single conductor leads with alligator clips. By varying the
constant. Water has a dielectric constant of ␬ ⫽83 and wood number of Teflon sheets we observe a changing capacitance.

52 Am. J. Phys. 73 共1兲, January 2005 http://aapt.org/ajp © 2005 American Association of Physics Teachers 52
Fig. 1. Two sheets of ordinary household aluminum foil (⬃30.3 cm width兲
are rolled flat on the table. We used 0.05 mm thick Teflon film sheets that are
wider than the foil. The dielectric sheets are sandwiched between the two
sheets of aluminum foil and a particle board is placed on top to help smooth
out any remaining wrinkles. We vary the nominal plate separation by vary-
ing the number of dielectric sheets and measure the capacitance by connect- Fig. 2. 共a兲 A plot of capacitance versus inverse Teflon thickness for three
ing the foil sheets to a digital multimeter via 30 cm long alligator clips. applied pressures. The triangles 共䉱兲, squares 共䊏兲, and circles 共䊉兲 represent
data acquired with applied pressures of 2855 Pa, 1503 Pa, and 150 Pa,
respectively. Each capacitor has an area of 0.0892 m2 . 共b兲 A plot of only the
data shown in Fig. 2共a兲 for five through ten sheets of Teflon. This graph
appears linear, but the slope for any of the lines does not yield an accurate
The obvious method of analyzing the data using Eq. 共1兲 is dielectric constant for Teflon. If we examine the data values in this plot and
to plot the capacitance versus the reciprocal of the dielectric calculate dielectric constants based on the slopes 共ignoring the nonzero in-
thickness, 1/D 共see Fig. 2共a兲兲. The dielectric constant can be tercepts兲, we find dielectric constants of 1.31⫾0.12, 1.21⫾0.10, and 0.78
found from the slope and Eq. 共1兲, assuming that D is equiva- ⫾0.09 for pressures 2855 Pa, 1503 Pa, and 150 Pa, respectively. Teflon has
lent to the plate separation. However, the data in Fig. 2共a兲 a dielectric constant of 2.0. The expected capacitance intercept is zero, but
our measured intercepts are 0.42⫾0.07 nF, 0.48⫾0.09 nF, and 0.46
becomes increasingly nonlinear as 1/D becomes larger. Be- ⫾0.03 nF, respectively.
cause of this nonlinearity, many investigations do not mea-
sure capacitance for small dielectric thickness. If the mea-
surements for one to four sheets of dielectric material are
neglected, a linear plot can be produced as shown in Fig. five brands of meters and compared their measurements with
2共b兲. However, a comparison of the results from this plot those of a commercial capacitance bridge. Four of the five
with Eq. 共1兲 raises several questions. The data has a nonzero multimeters use an AC method for measuring capacitance.
intercept 共which varies with load pressure兲. The determina- This AC method applies the same frequency across the ca-
tion of ␬ from the slopes results in values ranging from pacitor independent of its capacitance. The fifth multimeter
0.78⫾0.09 to 1.31⫾0.12 depending on the pressure applied uses an RC timing technique to measure capacitance. The
to the capacitor. These values are outside of the experimental only difference in capacitance readings between the different
uncertainty from the expected value of ␬ ⫽2.0.17 Obviously, multimeters is that the meter using the RC timing technique
this simplified approach has problems. gave a large zero capacitance offset with no load capacitance
The nonzero intercepts correspond to a stray capacitance. (⬃10 nF). If this ‘‘zero’’ capacitance is subtracted, this mul-
Stray parallel capacitance could come from the leads or from timeter’s readings were virtually identical with the others as
the meter. However, the largest capacitance we can produce well as with the commercial capacitance bridge.
using the 30 cm single conductor leads is approximately 10 Because the nonlinearity of Fig. 2共a兲 cannot be attributed
pF. We also considered and rejected effects due to resistance to measurement flaws, we considered surface irregularities of
in the leads and current through the dielectric material. Each the plates of the capacitor as the possible cause. The foil has
of these effects is much too small to cause any measurable some irregularities as does any surface, and these irregulari-
difference given our materials. We have no other physical ties would be a reason to ignore the data for the first four
explanation for the reproducible stray capacitance that we dielectric sheets. Consider a parallel plate capacitor with a
observe. single flaw in one plate that penetrates into the dielectric
To make certain that the stray capacitance and low dielec- material. The electric field will be stronger in the region of
tric constants were not due to the multimeters, we examined the flaw leading to an excess of charge near the flaw. This

53 Am. J. Phys., Vol. 73, No. 1, January 2005 T. T. Grove, M. F. Masters, and R. E. Miers 53
 approach does not require us to arbitrarily discard the data
for small numbers of dielectric sheets.
We rewrite Eq. 共1兲 in terms of the series capacitance
1 D 1
⫽ ⫹ , 共4兲
C measured ␬ ␧ 0A C 0
where D is the dielectric thickness, and 1/C 0 is due to the
stray series capacitance. The intercept of the plot of
1/C measured versus D correspond to a stray series capacitance
共Fig. 3兲. These capacitance values range from C 0 ⫽13
⫾1 nF to 30⫾3 nF with increasing applied pressure. From
the slope of each line, we find dielectric constants ranging
from ␬ ⫽1.22⫾0.08 to 1.71⫾0.05 with increasing load pres-
sure. Even though we found a way of producing straight
Fig. 3. A plot of inverse capacitance versus dielectric thickness for the same lines using all the data, the measured dielectric constant of
data shown in Fig. 2共a兲. The triangles 共䉱兲, squares 共䊏兲, and circles 共䊉兲 Teflon, which is load pressure dependant, is still not within
represent the same applied pressures as in Fig. 2. The error bars are approxi- experimental uncertainty of the accepted value.
mately the same size as each marker. All of the data can be fit to a straight
line, but the slopes still do not yield accurate dielectric constants. The ef-
The stray series capacitance calculated from the intercept
fective dielectric constants determined from the slopes are 1.71⫾0.05, shown in Fig. 3 increases with increased load pressure,
1.65⫾0.06, and 1.22⫾0.08 for the same applied pressures. The stray series which leads us to suspect that there is air trapped between
capacitance based on the 1/C intercepts are 30.0⫾3.0 nF, 24.7⫾2.4 nF, the plates. As pressure is applied to the top plate of the ca-
12.7⫾1.2 nF for the same load pressures, respectively. pacitor, some air leaks out of the capacitor reducing the ef-
fective plate separation. If these air pockets are important,
they must be incorporated into our model for the capacitor.
We can imagine random pockets of air filling variations in
excess would result in a larger capacitance measurement for the material surfaces 共dielectric and foil兲. However, if we
a given plate separation than we would predict for a parallel examine the derivation of the effect of dielectrics on
plate capacitor by Eq. 共1兲. The effects of the flaw would be capacitance,20 the dielectric constant arises from an averag-
most significant when the plate separation is small. As the ing process of unevenly distributed electric fields at an
plate separation becomes larger, the capacitance would ap- atomic level. Thus, the process of determining the capaci-
proach the ideal parallel plate capacitor value. This situation tance is always an averaging process of nonuniform fields.
would result in a concave upward curve on a C versus 1/D Therefore, it is reasonable to use a simple model that accu-
plot for small values of D, rather than a straight line. rately describes the experiment by replacing the randomly
Alternatively, a single flaw of the dielectric material jut- placed and sized air pockets with uniform layers of air, di-
ting into one plate would decrease the charge in that region electric material, and plates 共see Fig. 4共a兲兲. Although we do
and result in a lower overall capacitance. The measured ca- not expect the air pockets to be uniform, the capacitor be-
pacitances for small d would be smaller than those predicted haves as if the air pockets were uniform. As mentioned in
by Eq. 共1兲. In this case a plot of C versus 1/D would result in many introductory texts, different dielectric materials sand-
concave down curve. This case agrees with our data and is wiched between parallel plates may be treated as multiple
repeatable. However, there is no reason to believe that the capacitors in series. If we rearrange the order of these sand-
surface flaws are always directed into the conducting plates. wiched dielectric layers 共see Fig. 4共b兲兲, we can replace the
Because the dielectric materials we use are rigid, we cannot multiple layers of dielectric and air by a single solid dielec-
expect the dielectric to fill all voids in the parallel plates. tric sheet and a single layer of air. Hence, our measured
Hence, an explanation of the nonlinearity and nonzero inter- value of capacitance may be thought of as two capacitors in
cepts of Fig. 2共a兲 using a model based on surface irregulari- series, one a perfect parallel capacitor filled with the dielec-
ties or one based on stray parallel capacitances is question- tric material,
able. ␬ ␧ 0A
Consider what would happen if there was an extra capaci- C d⫽ , 共5兲
tance in series with our parallel plate capacitor. The mea- D
sured capacitance becomes
where C d is the capacitance due to the dielectric material.
CC series The other capacitor is a perfect parallel capacitor filled with
C measured⫽ . 共2兲 air,
C⫹C series
With the use of Eq. 共1兲, we obtain ␧ 0A
C a⫽ , 共6兲
da
␬ ␧ 0 AC series
C measured⫽ . 共3兲 where C a is the capacitance due to air and d a is the effective
共 ␬ ␧ 0 A⫹C seriesD 兲
air thickness between the plates. The measured capacitance
This situation would imply a concave downward curve for a is the series combination of the dielectric capacitor and the
plot of C measured versus 1/D. If we replot the data shown in air capacitor,
Fig. 2共a兲 as the reciprocal of the capacitance versus the di-
electric thickness 共see Fig. 3兲, the data becomes linear sup- 1 D da
⫽ ⫹ . 共7兲
porting the idea that there is a stray series capacitance. This C measured ␬ ␧ oA ␧ oA

54 Am. J. Phys., Vol. 73, No. 1, January 2005 T. T. Grove, M. F. Masters, and R. E. Miers 54
 共10兲兲. The dielectric constants are for material-air composites
and the intercept corresponds to an air capacitor with a plate
separation of 2d f ⫺d s .
If we want to determine the dielectric constant of the ma-
terial without air gaps, we either have to remove all the air
gaps, or we have to measure d s /D 1 . We can obtain a rough
measurement of the latter by comparing the measurements of
a large number of stacked dielectric sheets and the average
thicknesses of these same sheets. However, because the av-
erage air gap thickness is roughly the same as the uncertainty
of the micrometer, large relative uncertainties result. This
technique produces a measurement of d s /D 1 ⫽0.06⫾0.04
for the Teflon sheets. If we use ␬ eff⫽1.71 共from the slope of
the Teflon data in Fig. 3 with the largest load pressure兲, the
ratio of the air gap distance to the dielectric thickness, and
Eq. 共10兲, we find a dielectric constant for Teflon of ␬
⫽1.91⫾0.19.
Because this result is based on a term with large relative
uncertainty, d s /D 1 , we may ask if there is a more accurate
method? We could consider adding larger forces to the top of
the capacitor. Increased load pressure should reduce the mag-
nitude of d s /D 1 making its inclusion negligible. However, to
squeeze the air from between the plates would require pres-
sures as large as atmospheric pressure.

III. EXPERIMENT 2
Because we are uncomfortable with the results of the in-
vestigations in Sec. II, we investigated an alternative ap-
Fig. 4. 共a兲 A cross section of a Teflon-air-foil parallel plate capacitor. 共b兲 The proach with no air layers between the sheets. The simplest
equivalent Teflon-air-foil parallel plate capacitor with the paper and air in a method of removing the air gaps between sheets is to replace
single grouping. the multiple thin sheets of dielectric material by single sheets
of varying thicknesses. Teflon film sheets come in a variety
of nominal thicknesses 0.002⬙ , 0.003⬙ , 0.005⬙ , 0.010⬙ ,
0.015⬙ , 0.020⬙ , 1/32⬙ , and 3/64⬙ . Because Teflon film sheets
If we consider N sheets of dielectric material in the ca- tend to roll up, which would add unwanted extra air gaps
pacitor, then we have (N⫺1) layers of trapped air between between the dielectric and the plates, we still use a load
the dielectric sheets. If we express Eq. 共7兲 in terms of the mass. However, the pressures caused by these load masses
nominal thickness of an individual sheet of dielectric, we are so low 共we used a load pressure of 2855 kPa兲 that the
find compression of the Teflon is negligible. A plot of the inverse
capacitance versus the dielectric thickness produces the line
1 ND 1 2d f ⫹ 共 N⫺1 兲 d s represented by triangles in Fig. 5. The dielectric constant
⫽ ⫹ , 共8兲
C measured ␬ ␧ o A ␧ oA calculated from the slope of this line and Eq. 共4兲 is 1.96
⫾0.06, which is in good agreement with the manufacturer’s
where D 1 is the nominal thickness of a single sheet of di- value of 2.0.
electric material, d f is the average air layer thickness be- For further reassurance that stray or irregular electric fields
tween the dielectric and a foil plate, and d s is the average air due to the surface irregularities were not responsible for the
layer thickness between two dielectric sheets. We rearrange results, we performed several more experiments. We once
Eq. 共8兲 and arrive at more constructed capacitors using Teflon sheets of varying
thicknesses, except this time we intentionally introduced
1 ND 1 2d f ⫺d s large surface imperfections and defects in the capacitor’s
⫽ ⫹ , 共9兲
C measured ␬ eff␧ o A ␧ oA conducting plates. The curve marked by the squares in Fig. 5
used aluminum screens 共the same type used in household
where ␬ eff is an effective dielectric constant given by screen doors兲 for the plates. The data marked by circles used
aluminum foil with a #6 washer (⬃0.97 mm thick兲 between
␬ the foil and the board. As shown in Fig. 5, the aluminum
␬ eff⫽ . 共10兲
ds screen data does not fit a straight line for small thicknesses 共a
1⫹ ␬ single layer of 0.05 mm thick Teflon film兲. This deviation is
D1
not due to stray fields, but arises because the screen is made
The effective dielectric constants determined using this of woven aluminum wire (⬃0.25 mm diameter threads with
analysis will be lower than the dielectric constants of a pure ⬃5.5 threads per cm兲 which causes peaks and valleys. The
material. The dielectric constants we determined using the thin dielectric sheets cannot support much weight and when
slopes in Fig. 3 are effective dielectric constants 共see Eq. the top screen is placed on top, peaks from the top plate push

55 Am. J. Phys., Vol. 73, No. 1, January 2005 T. T. Grove, M. F. Masters, and R. E. Miers 55
 and still cause air gaps. We could remove all the air gaps
using a resin 共or oil兲 between the dielectric sheets. However,
finding the dielectric constant of the material would require
knowledge of the dielectric constant of the resin 共oil兲 and the
exact thicknesses of each resin 共oil兲 layer.
Ultimately, we found that the simplest way to do accurate
measurements is to remove the air gaps between the indi-
vidual sheets of the dielectric. This necessity increases the
cost of an individual experimental set-up. However, the ac-
curacy using multiple sheets of dielectric is acceptable as
long as the inherent flaws in the experiment are recognized.
In this case, the parallel plate capacitor experiment changes
from a simple cookbook experiment to a true experiment
with unexpected results. Students must consider their obser-
vations and modify a simple model to explain their observa-
Fig. 5. A plot of the inverse capacitance versus the dielectric thickness with tions.
different thickness Teflon sheets. The different data sets represent different
1
plate surfaces. The triangle 共䉱兲 data is from aluminum foil plates 共same G. W. Parker, ‘‘Electric field outside a parallel plate capacitor,’’ Am. J.
plates as used to acquire the data shown in Figs. 2 and 3兲; the square 共䊏兲 Phys. 70, 502–507 共2002兲.
2
data used woven aluminum screen 共typically used in household screen Yaakov Kraftmakher, ‘‘Measurement of dielectric constant of gases,’’ Am.
doors兲 for the conducting plates; the circle 共䊲兲 data used foil plates with a J. Phys. 64, 1209–1210 共1996兲.
3
standard 6-32 washer 共0.970 mm thick兲 inserted under the bottom foil to John J. Meyer and Anthony F. Behof, ‘‘Experiment on the motion of a
form an irregularity. The error bars are approximately the same size as each dielectric in a parallel plate capacitor,’’ Am. J. Phys. 62, 931 共1994兲.
4
marker. All these capacitors have an area of 0.0892 m2 . The slopes for the G. T. Carlson and B. L. Illman, ‘‘The circular disk parallel plate capaci-
graphs are nearly identical regardless of the plate material or plate flaw tor,’’ Am. J. Phys. 62, 1099–1105 共1994兲.
5
introduced. Each data set results in a dielectric constant within the experi- F. N. Yan and H. K. Wong, ‘‘Force between the plates of a parallel-plate
mental uncertainty of the accepted value for Teflon of 2.0. capacitor,’’ Am. J. Phys. 61, 1153 共1993兲.
6
David E. Wilson, ‘‘A direct laboratory approach to the study of capaci-
tors,’’ Am. J. Phys. 57, 630– 632 共1989兲.
7
Constantino A. Utreras-Diaz, ‘‘Dielectric slab in a parallel plate capaci-
the thin dielectric sheets into valleys of the bottom plate 共and
tor,’’ Am. J. Phys. 56, 700–701 共1988兲.
vice-versa兲 resulting in a smaller plate separation for thin 8
Ricky Joe Harwood and Donald E. Kinkaid, ‘‘Use of a Mettler balance and
dielectric sheets than for thicker dielectric sheets. From the a parallel plate capacitor to measure the permittivity of free space,’’ Am. J.
data in Fig. 5, it is apparent that defects in the plates affect Phys. 43, 924 –925 共1975兲.
the intercept values, but not the slope 共which is used to find 9
Edwin A. Karlow, ‘‘Let’s measure the dielectric constant of a piece of
the dielectric constant兲. The data presented in Fig. 5 is con- paper!,’’ Phys. Teach. 29, 23–25 共1991兲.
10
sistent with our model because these defects only create a S. K. Foong and C. H. Lim, ‘‘On the capacitance of a rolled capacitor,’’
Phys. Educ. 37, 429– 430 共2002兲.
much larger air layer thickness between the dielectric mate- 11
John D. Cutnell and Kenneth W. Johnson, Physics 共Wiley, New York,
rial and the plates. The value of the dielectric constant for 2001兲, 5th ed.
Teflon from each of these experiments is (1.96⫾0.06) for 12
Jerry D. Wilson and Anthony J. Buffa, College Physics 共Prentice Hall,
the foil, (1.95⫾0.10) for the screen, and (1.93⫾0.15) for Upper Saddle River, NJ, 1990兲, 3rd ed.
13
the foil and washer. These results indicate that the surface David Halliday, Robert Resnick, and Jearl Walker, Fundamentals of Phys-
ics 共Wiley, New York, 2001兲, 6th ed.
and stray electric fields are not a problem in the determina- 14
Raymond Serway and Jerry S. Faughn, College Physics 共Saunders College
tion of the dielectric constant. Publishing, New York, 2000兲, 5th ed.
15
Sheldon H. Radin and Robert T. Folk, Physics for Scientists and Engineers
IV. CONCLUSION 共Prentice Hall, Englewood Cliffs, NJ, 1982兲.
16
Robert C. Weast, Melvin J. Astle, and William H. Beyer, CRC Handbook
Small plate separations are required for appreciable ca- of Chemistry and Physics 共Chemical Rubber Company Press, Boca Raton,
pacitances which require the use of thin dielectric sheets. FL, 1986兲, 67th ed.
However, the use of multiple thin dielectric sheets leads to
17
Teflon is a registered trademark of Dupont. Information on 0.002⬙ films
may be found at 具http://www.dupont.com/teflon/films/H-04321-2.pdf典.
proportionally larger air gaps. When dealing with thin dielec- 18
The Teflon was purchased from McMaster–Carr.
tric materials, the material is likely to be very flexible 共such 19
We used a KRM AEEC-2890 digital multimeter for these measurements.
as paper or Teflon兲 or extremely brittle 共like glass兲. Flexible 20
For example, see Edward M. Purcell, Electricity and Magnetism
material will cause air gaps. Brittle material can easily break 共McGraw-Hill, New York, NY, 1985兲, 2nd ed.

56 Am. J. Phys., Vol. 73, No. 1, January 2005 T. T. Grove, M. F. Masters, and R. E. Miers 56