7KHTXDGUDWLFDOO\GDPSHGRVFLOODWRU$FDVHVWXG\RIDQRQOLQHDUHTXDWLRQ

RIPRWLRQ
%56PLWK

&LWDWLRQ$P-3K\V  GRL

9LHZRQOLQHKWWSG[GRLRUJ
9LHZ7DEOHRI&RQWHQWVKWWSDMSDDSWRUJUHVRXUFH$-3,$6YL
3XEOLVKHGE\WKH$PHULFDQ$VVRFLDWLRQRI3K\VLFV7HDFKHUV

5HODWHG$UWLFOHV
:DYHOHWV$&RQFLVH*XLGH
$P-3K\V 
0DWUL[1XPHURYPHWKRGIRUVROYLQJ6FKU|GLQJHU¶VHTXDWLRQ
$P-3K\V 
3RLQW\LFHGURSV+RZZDWHUIUHH]HVLQWRDVLQJXODUVKDSH
$P-3K\V 
:K\QRVKHDULQ³'LYJUDGFXUODQGDOOWKDW´"
$P-3K\V 
-DFRELHOOLSWLFIXQFWLRQVDQGWKHFRPSOHWHVROXWLRQWRWKHEHDGRQWKHKRRSSUREOHP
$P-3K\V 

$GGLWLRQDOLQIRUPDWLRQRQ$P-3K\V
-RXUQDO+RPHSDJHKWWSDMSDDSWRUJ
-RXUQDO,QIRUPDWLRQKWWSDMSDDSWRUJDERXWDERXWBWKHBMRXUQDO
7RSGRZQORDGVKWWSDMSDDSWRUJPRVWBGRZQORDGHG
,QIRUPDWLRQIRU$XWKRUVKWWSDMSGLFNLQVRQHGX&RQWULEXWRUVFRQW*HQ,QIRKWPO

Downloaded 22 Jan 2013 to 194.254.61.42. Redistribution subject to AAPT license or copyright; see http://ajp.aapt.org/authors/copyright_permission
The quadratically damped oscillator: A case study of a non-linear
equation of motion
B. R. Smith, Jr.a)
United States Naval Academy, Department of Physics, Annapolis, Maryland 21402
(Received 21 November 2011; accepted 31 May 2012)
The equation of motion for a quadratically damped oscillator, where the damping is proportional to
the square of the velocity, is a non-linear second-order differential equation. Non-linear equations
of motion such as this are seldom addressed in intermediate instruction in classical dynamics;
this one is problematic because it cannot be solved in terms of elementary functions. Like all
second-order ordinary differential equations, it has a corresponding first-order partial differential
equation, whose independent solutions constitute the constants of the motion. These constants
readily provide an approximate solution correct to first order in the damping constant. They also
reveal that the quadratically damped oscillator is never critically damped or overdamped, and that
to first order in the damping constant the oscillation frequency is identical to the natural frequency.
The technique described has close ties to standard tools such as integral curves in phase space and
phase portraits. VC 2012 American Association of Physics Teachers.
[http://dx.doi.org/10.1119/1.4729440]

@S X @S dqi
I. INTRODUCTION 0¼ þ ; 1 # i # n; (2)
@t i
@qi dt
Newton’s second law reduces the science of mechanics to
the solution of second-order ordinary differential equations. where the qi ðtÞ include the original and all of the newly iden-
Constants of the motion, where they can be found, provide tified variables, and n is the total number of first-order ordi-
important insight into the solution of the equations of motion nary differential equations in the expanded system. There are
of a physical system. They also elevate the science of an infinite number of solutions, but only n independent solu-
mechanics above a purely mathematical exercise by facilitat- tions. Once n independent solutions Si ðqj ; tÞ have been
ing (if not actually constituting) our physical understanding found, they may be inverted to obtain qi ðSj ; tÞ. Among the
of the system. Kepler’s law of areas was perhaps the first qi ðSj ; tÞ are the sought-after solutions to the original differen-
constant of the motion to be discovered, later to be recog- tial equations, complete with integration constants.
nized as an example of the conservation of angular momen- Newton’s second law for a single particle provides the
tum. Huygens implicitly used conservation of linear equations of motion
momentum in his debate with Leibniz over the true measure
of the “quantity of motion.” The work-energy theorem x i ¼ f i ðxj ; x_ j ; tÞ;
m€ i; j # 3: (3)
applied to conservative forces evolved into the conservation
of energy, a law that reaches far beyond its purely mechani- The expanded system of first-order ordinary differential
cal origins. Techniques for finding constants of the motion equations is
for a physical system are valuable tools in theoretical
physics: think of the importance of cyclic or ignorable coor- vi ¼ x_ i ; (4a)
dinates in Lagrangian and Hamiltonian fomulations, of
Noether’s theorem linking conserved quantities to symmetry m_v i ¼ mai ¼ f i ðxj ; vj ; tÞ: (4b)
and invariance properties, and of the link between conserved
quantities and eigenvalues in quantum mechanics. The velocities are now considered independent variables on
These seminal constants and conservation principles are an equal par with the coordinates. Equation (2) becomes2
core to the discipline of theoretical physics. A student just
acquiring an appreciation for their importance and utility @S X3
@S i X 3
@S i
0¼ þ v þ a: (5)
may unconsciously conclude that they are rare and excep- @t i¼1 @xi @v i
i¼1
tional. In actuality, any system of ordinary differential equa-
tions of any order can be expanded into a larger system of Solutions of this first-order partial differential equation are
first-order differential equations by identifying each deriva- constants of the motion.
tive as a new variable; for example, Reference 2 gives four examples, all important classroom
problems, where the solution of Eq. (5) can be obtained by el-
dxi ementary means. In many other cases, finding a complete set
¼ yi ; (1a)
dt of independent solutions to Eq. (5) is harder than solving the
equations of motion themselves. However, even one solution
d2 xi dyi provides significant insight into the problem. Consider, for
¼ : (1b)
dt2 dt example, that if all the forces are conservative, so that

The expanded system has an associated first-order partial dif- 1 @/

ai ¼ & (6)
ferential equation1 m @xi

816 Am. J. Phys. 80 (9), September 2012 http://aapt.org/ajp C 2012 American Association of Physics Teachers
V 816

Downloaded 22 Jan 2013 to 194.254.61.42. Redistribution subject to AAPT license or copyright; see http://ajp.aapt.org/authors/copyright_permission
for some /ðxi Þ, then Eq. (5) yields Finding T is, in principle, straightforward for one-
dimensional systems. Equation (5) is separable in t with a
@S X3
@S i 1 X 3
@S @/ solution of the form
0¼ þ i
v & ; (7)
@t i¼1 @x m i¼1 @vi @xi
Tðx; v; tÞ ¼ xt þ T^ðx; vÞ; (17)
which has the immediate time-independent solution where
1 X3
@T
S¼ m vi vi þ /: (8) ¼ x; (18a)
2 i¼1 @t
@ T^ f @ T^
This is the well-known result that the motion proceeds in v þ ¼ &x: (18b)
such a way that the total mechanical energy is constant. In @x m @v
this example, S is a time-independent scalar. Equation (10) can be solved to yield
The nature of S varies. It may be a scalar, as in the exam- ðx
ple of constant mechanical energy, or it may be a vector as dx
in the case of total angular momentum Li for central force t & t0 ¼ : (19)
x0 v
motion. If a scalar, it need not be energy or even something
resembling energy. Indeed, it may be something for which a While the constants of the motion treat x and v as independ-
meaningful physical interpretation is hard to find. For the ent variables, for one-dimensional motion they are linked
strictly utilitarian purposes of solving the equations of through Eq. (14). With v considered a function of x, Eq. (19)
motion, the simplest possible expressions for S, of whatever can be integrated by parts to yield
type, are preferred. $ ðx
One-dimensional systems possess only two independent x $$x x dv
t & t0 ¼ þ dx; (20a)
constants of the motion. Restricting the discussion to forces vðxÞ $x0 x0 v 2 ðxÞ dx
that are not explicitly a function of time, the equation of motion $x ðx
x $$ x dv=dt
m_v ¼ f ðx; vÞ (9) ¼ þ dx; (20b)
vðxÞ $x0 x0 v ðxÞ dx=dt
2

can be multiplied by the identity $ ðx

x $$x x a
¼ $ þ dx; (20c)
v ¼ x_ (10) vðxÞ x0 x0 v ðxÞ v
2

$ ðx
to yield the Pfaffian form x $$x xf ðx; vÞ
¼ þ dx: (20d)
vðxÞ $x0 3
x0 mv ðxÞ
mv dv ¼ f ðx; vÞ dx: (11)
Repeatedly integrating by parts yields an infinite series on
All two-variable Pfaffian forms can be transformed into total the right-hand side,
integrals by the use of an integrating factor3 l such that
X1 % & n $x
1 $$
0 ¼ dS ¼ mlv dv & lf ðx; vÞ dx: (12) t & t0 ¼ Fn ðx; vÞ ; (21)
n¼0
v $x0
The integrating factor can be found from the integrability
condition which, if questions of convergence can be adequately dealt
^ vÞ & T^ðx0 ; v0 Þ that brings Eq.
with, represents a function Tðx;
@ðlvÞ @ðlf Þ (19) into the form
m ¼& : (13)
@x @v
Tðx; v; tÞ ¼ xt & T^ðx; vÞ ¼ xt0 & T^ðx0 ; v0 Þ: (22)
Once the integrating factor is known, Eq. (12) can be inte-
grated to yield the first constant of the motion, Alternatively, x can be regarded as a function of v, yielding
a second expression for T that can be found by writing
Sðx; vÞ ¼ Sðx0 ; v0 Þ: (14) Eq. (19) as
ðv ðv
The second independent constant of the motion must be ex- dx dv m dv
t & t0 ¼ ¼ : (23)
plicitly a function of time v0 dv v v0 ðx; vÞ
f
Tðx; v; tÞ ¼ Tðx0 ; v0 ; t0 Þ; (15) Repeated integration by parts again yields an infinite series,
$v ðv
in order for the equation of motion to be found. Equations mv $$ mv df
t & t0 ¼ þ dv; (24a)
(14) and (15) constitute two equations in the three independ- f ðx; vÞ $v0 v0 f 2 ðx; vÞ dv
ent variables x, v, and t. One equation may be used to elimi- $v ðv % &
nate v in the other, yielding the equation of motion mv $$ mv @f dx @f
¼ þ þ dv;
! " f ðx; vÞ $v0 v0 f ðx; vÞ @x dv
2 @v
x ¼ x Sðx0 ; v0 Þ; Tðx0 ; v0 ; t0 Þ; t : (16) (24b)

817 Am. J. Phys., Vol. 80, No. 9, September 2012 B. R. Smith, Jr. 817

Downloaded 22 Jan 2013 to 194.254.61.42. Redistribution subject to AAPT license or copyright; see http://ajp.aapt.org/authors/copyright_permission
 $ % &$
mv $$v mv2 1 @f mv @f $$v It is left as an exercise for the reader to solve Eq. (26) for
¼ þ þ (24c) v(x) and substitute this into either form of T(x,v;t) to obtain
f ðx; vÞ $v0 2 f 2 ðx; vÞ @x f @v $v0
the standard equation of motion for the simple harmonic
ðv ' % &( oscillator.
mv2 d 1 @f mv @f
& þ dv: (24d) A similar analysis can be done for the linearly (viscously)
v0 2 dv f ðx; vÞ @x f
2 @v damped harmonic oscillator. The time-independent con-
stant4,5 is
If this infinite series convergences properly, then a second
representation for T(x,v;t) exists. ' % &(
1 2 2 2 2c &1 xx
S ¼ m½ðv þ cxÞ þ x x (exp tan :
2 x v þ cx
II. THE HARMONIC OSCILLATOR (32)
The simple harmonic oscillator will be used to illustrate The time-dependent constant is
the process. As the restoring force for the simple harmonic % &
oscillator, &1 xx
Tðx; v; tÞ ¼ xt & tan ; (33)
v þ cx
m_v ¼ &mx2 x; (25)
where
is conservative, Eq. (12) immediately yields the first constant, qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 1 x¼ x20 þ c2 (34)
S ¼ mðv2 þ x2 x2 Þ ¼ mðv20 þ x2 x20 Þ: (26)
2 2
is the damped frequency and x0 is the undamped (natural)
Equation (19) becomes frequency of the oscillator.
$ ðx
x $$x x &x2 x III. THE QUADRATICALLY DAMPED HARMONIC
t & t0 ¼ $ þ dx; (27a)
v x0 x0 v ðxÞ
2 v OSCILLATOR
$x $x ðx 4 4 Both the simple and the viscously damped harmonic
x $$ 1 x2 x3 $$ xx
¼ $ & þ dx; (27b) oscillators can be adequately solved with conventional
v x0 3 v3 $x0 x0 v
5
approaches. The value to the student of the approach offered
$ $ $ here for those simple problems is more pedagogical than util-
xx $$x 1 x3 x3 $$x 1 x5 x5 $$x
xðt & t0 Þ ¼ & þ & …: itarian. The balance begins to shift toward the utilitarian for
v $x0 3 v3 $x0 5 v5 $x0 the quadratically damped harmonic oscillator. For damping
(27c) proportional to v2 , Newton’s second law takes the form
*
The sequence converges for ðxx=vÞ2 < 1 to yield &mx2 x & mcx_ 2 ; x_ > 0;
m€x¼ (35)
% & &mx2 x þ mcx_ 2 ; x_ < 0;
!xx" xx0
Tðx; v; tÞ ¼ xt & tan&1 ¼ xt0 & tan&1 :
v v0 and the student is faced with two different non-linear differ-
(28) ential equations. The damping coefficient must change sign
whenever v ¼ 0 since the damping force always opposes the
This expression for T(x,v;t) can be used to immediately cal- velocity; for oscillatory motion, this happens every half-
culate the period s, and therefore the frequency, of the oscil- cycle. As a consequence, each equation must be solved sepa-
lator. At the two turning points of the motion, v ¼ 0, and the rately and the solutions mated at the turning points to yield a
interval between them is continuous solution valid for all t. Assuming that x has its
maximum displacement x0 at t ¼ 0, the first turning point
!s" p ! p" xðT1 Þ must be calculated from the v < 0 solution to Eq. (35)
x ¼ xðtþ & t& Þ ¼ & & ¼ p; (29)
2 2 2 and used as a boundary condition for the v > 0 solution.
Similarly, xðT2 Þ must be calculated from the v > 0 solution
so that the period of the motion is 2p=x, as expected for and used as a boundary condition for the next v < 0 solution,
the simple harmonic oscillator. In exactly the same way, Eq. and so on. The two differential equations are related through
(24a) yields a parity transformation: x€v<0 ¼ &€ x v>0 . Once the solution is
! " % & known for v < 0, the solution for v > 0 is obtained by simply
&1 v &1 v0 changing sign.
Tðx; v; tÞ ¼ xt þ tan ¼ xt0 þ tan :
xx xx0 The brute-forceX approach to Eq. (35) is a power series
1
(30) expansion xðtÞ ¼ a tn . The expressions for an get
n¼0 n
very complex very fast. While formally correct, this
It bears repeating that the constants of the motion are not approach fails to provide any significant physical insight into
unique; the two expressions here for T(x,v;t) are related the behavior of the quadratically damped oscillator. It does
through the trigonometric identity not easily provide an expression for the damped frequency
% & of the oscillator in terms of the natural frequency, for exam-
&1 p &1 1 ple, nor does it provide any convenient method of calculating
tan z ¼ & & tan : (31)
2 z the displacements at the turning points.

818 Am. J. Phys., Vol. 80, No. 9, September 2012 B. R. Smith, Jr. 818

Downloaded 22 Jan 2013 to 194.254.61.42. Redistribution subject to AAPT license or copyright; see http://ajp.aapt.org/authors/copyright_permission
 Instead, we approach the solution through the constants of The plot of constant S^ lays out the&trajectory of the damped
the motion. First, write Eq. (35) as two-variable Pfaffian forms oscillator in phase space. Since S^ is a constant, its value
* can be conveniently found by setting v ¼ 0 in Eq. (40). The
v dv þ ðx2 x & cv2 Þ dx; v < 0; condition ^& < 0 holds for all positive values of u0 and
S
0 ¼ dA ¼ (36) & &
v dv þ ðx2 x þ cv2 Þ dx; v > 0: S^ ! 0 from below as u0 ! 1, limiting S^ to the range

These are not quadratures, as they do not obey the integrabil- & 1
0 > S^ * & : (41)
ity condition. The required integrating factors are found by 4
solving
Starting from maximum displacement u0 where w ¼ 0, the
@l @l 2 trajectory will follow a path set by
v¼ ðx x ) cv2 Þ ) 2cvl: (37)
@x @v % &
& u0 1
The most general solutions are not necessary; the simplest S^ ¼ &e&2u0 þ : (42)
2 4
solutions5–7 l ¼ e)2cx suffice. Equation (36) becomes
The turning point at the end of this half-cycle is found by
l dA ¼ dS) ¼ ðe)2cx v dv ) v2 ce)2cx dxÞ þ e)2cx x2 x dx; iteratively solving
(38a)
' 2 % & ( e&2u0 ð2u0 þ 1Þ ¼ e&2u1 ð2u1 þ 1Þ: (43)
v x2 1
S) ¼ e)2cx ) x6 : (38b)
2 2c 2c A turning point implies that the velocity is zero, so that a
turning point at the origin u1 ¼ 0 means that the oscillator is
Note that S& and Sþ are related by a parity transformation: critically damped or overdamped. Equation (43) then
S& ð&x; &vÞ ¼ Sþ ðx; vÞ. becomes
The quantities S6 , the first constants of the motion for this
problem, are each graphically illustrated over one-half cycle e&2u0 ð2u0 þ 1Þ ¼ 1; (44)
in Fig. 1. The phase space axes are expressed in terms of the
unitless parameters which has no solution for u0 > 0; therefore, the quadrati-
cally damped oscillator is never critically damped or
cv
w¼ ; (39a) overdamped.
x The velocity at the first passage through u ¼ 0 is
u ¼ cx: (39b) rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 & e&2u0 ð2u0 þ 1Þ
w¼& : (45)
The quantities S) normalize to 2
% 2 &
^) c2 ) )2u w u 1 The point of maximum velocity xf , i.e., the point where the
S ¼ 2S ¼e ) & : (40) acceleration is zero, is found from Eq. (38b)
x 2 2 4
x2
2cS& e2cx ¼ cv2 & x2 x & ; (46a)
2c
x2
¼ x€ & ; (46b)
2c
lnð2u0 þ 1Þ
uf ¼ u0 & : (46c)
2

At the first turning point, the expression for S^ changes from

& þ & þ
S^ to S^ . It is clear from Fig. 1 that if S^ ¼ S^ ; S^ will not
be continuous across the u axis. This happens because the
solutions given in Eq. (38b) are only determined up to an
integration constant S^n . For the two expressions to match
smoothly and form a single function,
% &
^& &2u0 u0 1 þ
S ¼ &e þ ¼ S^ þ S^1 ; (47)
2 4
þ
where the value of the constant S^ must be
% &
^þ 2u1 u1 1
S ¼ &e & : (48)
2 4
& þ & þ
Fig. 1. The functions S^ and S^ when S^ ¼ S^ . The axes have been nor-
malized to u ¼ cx and w ¼ cv=x. Therefore, S^1 is

819 Am. J. Phys., Vol. 80, No. 9, September 2012 B. R. Smith, Jr. 819

Downloaded 22 Jan 2013 to 194.254.61.42. Redistribution subject to AAPT license or copyright; see http://ajp.aapt.org/authors/copyright_permission
 % & % &
u1 1 u0 1 The next turning point is found by interatively solving
S^1 ¼ &e2u1 & & e&2u0 & : (49)
2 4 2 4
S^ & S^1 ¼ e2u2 ½2u2 & 1(: (51)
The result is shown in Fig. 2. Measuring the angle h clockwise
from the positive u axis, the function S^ for the first full cycle is
% & The new turning
&
point u2 is fed back into Eq. (42) to
^ &2u0 u0 1 determine S^ for the second cycle, and the process is
S ¼ &e þ repeated. This process generates a sequence of integration
2 4
8 % 2 & constants labeled S^n , where S^0 ¼ 0, and n numbers the
> e&2u w & u & 1 ;
>
< 0 # h < p; half-cycles.
¼ % 22 2 & 4
(50)
This kind of functional patch needs to be made at every
>
> w u 1 turning point. The general equations are
: e2u þ & ^
þ S 1 ; p # h < 2p:
2 2 4

S^ & S^2n ¼ e&2u2nþ1 ð2u2nþ1 þ 1Þ ¼ e&2n2n ð2n2n þ 1Þ; (52a)

S^ & S^2nþ1 ¼ e2u2nþ2 ð2u2nþ2 & 1Þ ¼ e2n2nþ1 ð2n2nþ1 & 1Þ; (52b)
8 % 2 &
> w u 1
% & ><e &2u
& & þ S^2n ; 2np # h < ð2n þ 1Þp;
u0 1 2 2 4
S^ ¼ &e &2u 0
þ ¼ % 2 & (52c)
2 4 >
> w u 1
: e2u þ & þ S^2nþ1 ; ð2n þ 1Þp # h < ð2n þ 2Þp:
2 2 4

The results for three full cycles are displayed in Table I. constant has been determined. Returning to the configuration
Figure 3 shows three full cycles of the quadratically damped variables v and x, using
oscillator in phase space. See Ref. 7 for a parallel, more
mathematically sophisticated, development of the material in x2 ^ ^
En ¼ ðS & S n Þ; (53)
this section. c2

then solving for v(x) and separating the variables, yields the
IV. THE SOLUTION OF THE EQUATION quadrature
OF MOTION ð ð
dx
Equation (52c) leads directly to the equation of motion for dt ¼ sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi%
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi&
ffiffiffi: (54)
2
a given half-cycle, once the value of the corresponding x 1
2e2cx En þ xþ
2c 2c

The duration of the half-cycle is the transit time from one

turning point to the next,
ð xmax
s dx
¼ sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi%
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi&
ffiffiffi: (55)
2 xmin x2 1
2e En þ 2cx xþ
2c 2c

For a single half-cycle, x(t) is one-half of the periodic func-

tion y(t) obtained by performing the integration of Eq. (54).
The phase space picture of y(t) is shown in Fig. 4.

Table I. S^ ¼ &0:101501.

nth Half-cycle S^n&1 un&1

First 0 1
Second 0.098765 &0.41072
Third 0.123863 0.263568
Fourth 0.133834 &0.194566
Fifth 0.138780 0.154318
Sixth 0.141589 &0.127908
Fig. 2. One full cycle of the quadratically damped oscillator.

820 Am. J. Phys., Vol. 80, No. 9, September 2012 B. R. Smith, Jr. 820

Downloaded 22 Jan 2013 to 194.254.61.42. Redistribution subject to AAPT license or copyright; see http://ajp.aapt.org/authors/copyright_permission
 $v $v
xv $
$ 2 !cv" x2 v2 $
$
¼ 2 þ
cv & x2 x $v0 3 x ðcv2 & x2 xÞ2 $v0
$v
1 x3 v3 $
& $
3 ðcv2 & x2 xÞ3 $v0
ð v ! "3 % 4 &
4 cv x 2cv dv & x2 dx
þ
v0 3 x c2 ðcv2 & x2 xÞ3
ðv 3 3
x v ð2cv dv & x2 dxÞ
& : (56d)
v0 ðcv2 & x2 xÞ4

This suggests that T(x,v;t) may be written as

X1 !cv"% xv &n
Tðx; v; tÞ ¼ xt & fn : (57)
n¼1
x cv2 & x2 x

Substituting this into Eq. (5) yields

X 1 % &n
xv
0 ¼ x & x ðn & 1Þfn&1
n¼2
cv2 & x2 x
X1 ! " % &n
cv 0 xv
Fig. 3. Three full cycles of the quadratically damped oscillator. &x f
n¼0
x nþ1 cv2 & x2 x
X1 % &n
The equation of the trajectory x(t) is not easily found; Eq. (54) xv
& x ðn þ 1Þfnþ1
is not found in standard tables of integrals. Using Eq. (24a), n¼0
cv2 & x2 x
!cv"X 1 % &n
ðv xv
x dv þ 2x nfn : (58)
xðt & t0 Þ ¼ ; (56a) x n¼1 cv2 & x2 x
v0 cv 2 & x2 x
$v ð v
xv $$ xv½2cvdv & x2 dx(
¼ 2 þ ; (56b) Setting the coefficients of each order equal to zero yields this
cv & x2 x $v0 v0 ðcv2 & x2 xÞ2 set of equations
$v ðv
xv $ 2cxv2 dv cv
¼ 2 $ þ
cv & x2 x $v0 2 2 w¼ ; (59a)
v0 ðcv & x xÞ
2
x
ðv
x3 v2 dv 0 ¼ 1 & f1 & wf10 ; (59b)
& 3
; (56c)
v0 ðcv2 & x2 xÞ
0 ¼ wf20 þ 2f2 & 2wf1 ; (59c)

0 ¼ ðn & 2Þfn&2 þ wfn0 þ nfn & 2wðn & 1Þfn&1 ;

0 ¼ ðn & 2Þwn&1 fn&2 þ ðwn fn Þ0 & 2ðn & 1Þwn fn&1 :
(59d)

These equations generate the following solutions:

f1 ¼ 1; (60a)
2 !cv"
f2 ¼ ; (60b)
3 x
8 !cv"2 1
f3 ¼ & ; (60c)
15 x 3
16 !cv"3 2 !cv"
f4 ¼ & ; (60d)
35 x 3 x
128 !cv"4 104 !cv"2 1
f5 ¼ & þ ; (60e)
315 x 105 x 5
256 cv 5 176 cv 3 2 !cv"
! " ! "
f6 ¼ & þ ; (60f)
Fig. 4. The phase space plot of y(t). 693 x 135 x 3 x

821 Am. J. Phys., Vol. 80, No. 9, September 2012 B. R. Smith, Jr. 821

Downloaded 22 Jan 2013 to 194.254.61.42. Redistribution subject to AAPT license or copyright; see http://ajp.aapt.org/authors/copyright_permission
 1024 !cv"6 1856 !cv"4 272 !cv"2 1 n&iþ1
X
f7 ¼ & þ & ; 2
2ð&1Þk ðnþi&2kÞ!!ðn&2Þ!!ðnþ1&2kÞ nþ1&2k
3003 x 1155 x 189 x 7 ¼ Ci&1 ;
ðnþiÞ!!ðn&2kÞ!!
(60g) k¼1
(62b)
X
n&1 !cv"i
fn ¼ Cni : (60h) where n and i have opposite parity. It follows that for i ¼ 0,
i¼0
x
8
< n & 2 n&2 n & 4 n&4 ð&1Þðn&1Þ=2
Equation (57) can be now be re-written as C02mþ1 ¼ & n C0 ¼ n C0 ¼ n
; n ¼ odd;
% &n :
X 1 ! "i X 0; n ¼ even:
cv 1 n xv
Tðx; v; tÞ ¼ xt & Ci : (63)
i¼0
x n¼iþ1 cv2 & x2 x
(61) The corresponding summation in Eq. (61) is
% &
From Eq. (59d), &1 xv
g0 ¼ tan : (64)
cv2 & x2 x
2ðn & 1Þ n&1 n & 2 n&2
Cni ¼ Ci&1 & C ; (62a)
nþi nþi i For i ¼ 1,

8
< 2ðn & 1Þ n & 2 n&2 ð&1Þðn&2Þ=2 n & 2 n&2
Cn1 ¼ C0n&1 & C1 ¼ 2 & C ; n ¼ even; (65)
: nþ1 nþ1 nþ1 nþ1 1
0; n ¼ odd:

It is easy to see that if keep xt continuous. The transit time is also the half-period
so that
2
C1n&2 ¼ ð&1Þðn&4Þ=2 + ; (66) 2p
3 s¼ : (70)
x
for n even, then8
To first order in c, the frequency of the quadratically damped
ðn&2Þ=2 2 oscillator is identical to the natural frequency.
Cn1 ¼ ð&1Þ + : (67)
3

Since it is true for n ¼ 2, then by mathematical induction it is

true for all even n. The corresponding summation in Eq. (61) is

2 cxv3
g1 ¼ : (68)
3 ½x2 v2 þ ðcv2 & x2 xÞ2 (

From i ¼ 2 on, the summations do not reduce to readily iden-

tifiable functions.
In the limit c ! 0, the equations should reduce to those
for the simple harmonic oscillator. To first order in cv=x,
Eq. (61) becomes
% &
xv 2 xcv3
0 ¼ xt & tan&1 & ;
cv & x x
2 2 3 ½x2 v2 þ ðcv2 & x2 xÞ2 (
(69)

which agrees with the harmonic oscillator result when the

damping vanishes. The transit time from one turning point
to the other is xðtþ & t& Þ. The last term on the right in Eq.
(69) vanishes since v ¼ 0 at the turning points; the argument
of the tan&1 also vanishes at each turning point, but since
the argument changes sign and diverges at the intermediate Fig. 5. Three full cycles of the quadratically damped oscillator to first-order
point where the denominator is zero, p must be added to in cv=x.

822 Am. J. Phys., Vol. 80, No. 9, September 2012 B. R. Smith, Jr. 822

Downloaded 22 Jan 2013 to 194.254.61.42. Redistribution subject to AAPT license or copyright; see http://ajp.aapt.org/authors/copyright_permission
 tial amplitude decay seen in the linearly damped oscillator.
The differences can be brought out by plotting a similar ex-
ponential fit as shown in Fig. 6. The exponential decay has
been arbitrarily forced to fit the turning points at the begin-
ning and end of the first cycle. Clearly, the decay of the
quadratically damped oscillator is much slower than expo-
nential. The equivalent comparison in phase space, as shown
in Fig. 7, is equally illuminating.
The analytical expression corresponding to the curve in
Fig. 5 is
0 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi2ffiffiffiffiffiffiffiffiffiffiffiffi2ffi1
x 2e2cx En þ xc x þ 2c x
2
0 ¼ xtn & tan&1 @ 2cx x2
A
2ce En þ 2c
0 h i3=2 1
2cx x2 x x2
2B cx 2e E n þ c þ 2c2 C
& @ 2 2 4cx 4 A;
3 4c En e þ 3x En e þ xc4 x þ 3x2 2cx
4c 2

ns ðn þ 1Þs
< tn # : (72)
2 2

Fig. 6. The exponential envelope for Fig. 5 has been set to intersect u(t) at
the beginning and at the end of the first period.
V. CONCLUSION

A graph of x(t) for any given half-cycle is constructed by Not surprisingly, the method used here lies in extremely
inserting vn ðxÞ from Eq. (54) into Eq. (69) and calculating close analogy to other, more standard, methods of solving
tn ðxÞ, then plotting the results as xn ðtÞ. At the end of each this problem. It is, after all, a one-dimensional problem, for
half-cycle, the calculation starts anew with vnþ1 ðxÞ. The which every approach is at heart a variation on a single
graphs are mated at the end of the half-cycle: theme. But choice of approach is suggested by point of view.
The “second-order differential equation” point of view sug-
xn ð0Þ ¼ xn&1 ðp=xÞ ¼ xð½n & 1(p=xÞ: (71) gests the straightforward “brute force” approach of the
power series expansion. By contrast, the use of an integrating
The result for three full cycles is shown in Fig. 5. The decay- factor to find the first integral of Eq. (14) suggests the stand-
ing oscillation in Fig. 5 may be compared with the exponen- ard approach of analyzing the integral curves and the phase
space portraiture9 for insight into the dynamics; inverting the
first integral to get v(x;S), and separating to obtain the quad-
rature of Eq. (19), follows. Focusing on v(x) and proceeding
with the integration over x is very natural at this point.
This paper presents a third approach that starts from the
point of view that Eq. (15) provides a second constant of the
motion, defined in such a way that x and v have equal stand-
ing as independent variables. This approach suggests the
alternate quadrature of Eq. (23) on an equal footing with Eq.
(19).
As demonstrated here, this third point of view can be
rewarding.

ACKNOWLEDGMENTS
Special thanks to the anonymous reviewers whose sharp
eyes, attention to detail, and intelligent suggestions contrib-
uted immensely to the quality of this paper. All of the fig-
ures, the values in Table I, and the coefficients in Eqs. (60a)
were obtained with MathematicaTM 8.

a)
Electronic mail: brsmith@usna.edu; Also at Anne Arundel Community
College, Physics Department, Arnold, Maryland 21012.
1
Fig. 7. Three full cycles in phase space of the quadratically damped oscilla- C. Carathéodory, Calculus of Variations and Partial Differential Equa-
tor (thin solid line) superimposed on the viscously (linearly) damped oscilla- tions of the First Order, 2nd (revised) English ed. (AMS Chelsea Publish-
tor (dotted line). As in Fig. 6, the two have been set to intersect at the ing Co., Providence, Rhode Island, 1999) pp. 24–26.
2
beginning and end of the first cycle. The outer circle represents the B. R. Smith, Jr., “First order partial differential equations in classical
undamped oscillator. dynamics,” Am. J. Phys. 77, 1147–1153 (2009).

823 Am. J. Phys., Vol. 80, No. 9, September 2012 B. R. Smith, Jr. 823

Downloaded 22 Jan 2013 to 194.254.61.42. Redistribution subject to AAPT license or copyright; see http://ajp.aapt.org/authors/copyright_permission
3 7
E. L. Ince, Ordinary Differential Equations (Dover Publications, Inc., Livija Cvetićanin, “Oscillator with strong quadratic damping force,” Publ.
New York, 1956), p. 27. Inst. Math, Nouv. Sér. 85(99), 119–130 (2009).
4
Reference 2, p. 1153. Xm ð2k þ 1Þ!! ð2mþ3Þ!!
5 8
H. H. Denman, “Time-translational invariance for certain dissipative clas- This proves that ¼ 3ð2m Þm! .
sical systems,” Am. J. Phys. 36, 516–519 (1968).
k¼0 2k k!
6 9
J. J. Stoker, Nonlinear Vibrations (Interscience Publishers, Inc., New Michael Tabor, Chaos and Integrability in Non-linear Dynamics: An
York, 1950), p. 60. Introduction (John Wiley and Sons, New York, 1989), Chap. 1.

824 Am. J. Phys., Vol. 80, No. 9, September 2012 B. R. Smith, Jr. 824

Downloaded 22 Jan 2013 to 194.254.61.42. Redistribution subject to AAPT license or copyright; see http://ajp.aapt.org/authors/copyright_permission