Snowball/Log Rolling down a Snowy Slope

Kirk T. McDonald
Joseph Henry Laboratories, Princeton University, Princeton, NJ 08544
(March 12, 2019; updated May 25, 2019)

1 Problem
An ostensibly simple problem is the motion of a snowball (or better, a cylinder/log) that rolls
without slipping down a snowy slope, accumulating mass as it moves. A naı̈ve approximation
is that the cross section of the ball/log remains circular at all times (which implies that
snow moves from the slope to be instantaneously distributed over the entire surface of the
rolling object, thereby instantaneously acquiring kinetic energy, momentum and angular
momentum).1 Show, that this (unphysical) assumption leads to different equations of motion
via a force/torque analyses about the center or mass of the log and about its line of contact
with the slope, as well as different ones based on energy conservation and a Lagrangian.
Does the motion have a simple, asymptotic (terminal) character?
Consider also the motion during the first and subsequent full turns of rolling, for which
a more accurate analysis can be given.2

2 Solution
This problem is a variant on the many examples discussed by the author in [5].
We consider a cylindrical log that rolls down a slope of angle α to the horizontal which is
covered by a layer of snow of depth b (normal to the slope), as sketched in the figure below.

1
This assumption is tacitly made, for example, in [1] and sec. V.D of [2], and was explicit in Art. 203,
p. 261, of [3] (which may have been the first discussion of this problem).
2
Another (simpler) problem in which the analysis can be partitioned into segments of fixed azimuthal
rotation is a hexagonal pencil that rolls without slipping on an incline [4].

1
 The rolling log (of mass density ρ, the same as that of the snow) accumulates all of the
snow that it encounters, without loss of energy to possible compaction of the snow. As such,
the shape of the cylinder is not quite circular, but an appealing approximation is that the
cylinder remains circular at all time, with radius r(t).
In this approximation, the mass of the log (of length l) is m = πρr2 l, and the rate of
accumulation of mass is,
dm dr mbv dr bv ωb
= 2πρrl = ρblv = , = = , (1)
dt dt πr2 dt 2πr 2π
where v = dx/dt is the speed of the center of the log down the slope, and ω = v/r is the
angular velocity of the log.

2.1 Force/Torque Analysis

The force (component) Fx parallel to the slope (and uphill) is related to the x-component
of the momentum, px = mv, of the log by,3

dpx dv dm dv mbv 2
mg sin α − Fx = =m +v =m + , (2)
dt dt dt dt πr2
where g is the acceleration due to gravity.

2.1.1 Torque Analysis about the Center of Mass of the Log

In addition, the torque equation with respect to the center of mass of the log is,

dL d(Iω) d(mrv/2) rv dm mv dr mr dv
rFx = τ = = = = + +
dt dt dt 2 dt 2 dt 2 dt
2 2
mbv mbv mr dv
= + + , (3)
2πr 4πr 2 dt
in the approximation that the moment of inertia of the (cylindrical) log about its axis is
I = mr2 /2. Combining eqs. (2) and (3), we obtain an equation of motion,4

3m dv 7mbv 2 dv 2g 7bv 2
= mg sin α − , a= = sin α − . (4)
2 dt 4πr2 dt 3 6πr2
3
The normal force Fy is related by Fy − mg cos θ = dpy /dt = d(m dr/dt)/dt, but we don’t need to pursue
this, as Fy exerts no torque about the center of mass of the log (or about its line of contact with the slope).
4
On p. 262 of [3], Loney found the equation of motion (4), and noted that the accumulation of snow
over distance x increases the radius of the log from r0 to r according to πr 2 = πr02 + b x. On changing
variables, u = v2 = ẋ2 , then 2a = 2ẍ = u̇/ẋ = du/dx = (4g/3) sin α − 7b u/3(πr02 + b x). This can be
integrated to give u = ẋ2 = 2g sin α(πr02 + b x)/5b + C/(πr02 + b x)7/3 , whose time derivative tells us that
a = (g/5) sin α − 7b C/3(πr02 + b x)10/3, which goes to the constant value (g/5) sin α for large x.

2
2.1.2 Torque Analysis about the Line of Contact of the Log and Slope
We could also consider the torque equation with respect to the line of contact of the log with
the slope, which line is instantaneously at rest.
dLC d(IC ω) d(3mrv/2) 3rv dm 3mv dr 3mr dv
rmg sin α = τ C = = = = + +
dt dt dt 2 dt 2 dt 2 dt
2 2
3mbv 3mbv 3mr dv
= + + , (5)
2πr 4πr 2 dt
noting that the moment of inertia about the point of contact is IC = I + mr2 = 3mr2 /2. The
resulting equation of motion is (without need to consider the force at the line of contact),

dv 2g 3bv 2
a= = sin α − . (6)
dt 3 2πr2

2.1.3 Comments
The two equations of motion, (4) and (6), differ, which alerts us to the possibility that the
preceding analysis is not sufficiently accurate.
In the limit of no snow on the slope, b → 0, both torque analyses yield the well known
result (reviewed in the Appendix) that the acceleration of a solid cylinder which rolls without
slip down a slope of angle α is 23 g sin α. And, if the correction to the acceleration in case
of a slope with thickness b is proportional to that thickness, but independent of g, then
dimensional analysis tells us that the correction is proportional to bv 2/r2 . The task of a
successful analysis of the motion is to identify the numerical coefficient of this term, which
the torque analyses apparently fail to do in a convincing manner.

2.2 Energy Analysis

A different analysis can be based on the approximation that no energy is dissipated by the
accumulation of snow on the rolling log, or by air resistance. Then, the mechanical energy,
E = T + V is constant.
The kinetic energy T is related by,
  2   2
m 2 dr Iω2 3mv 2 m dr 3πρlr2 v 2 ρlb2v 2
T = v + + = + = + . (7)
2 dt 2 4 2 dt 4 8π

For the potential energy, we suppose that the log started from rest with a radius r0 and
mass m0 = πρr02 l and rolled distance x down the slope to its present position. During this
time, it accumulated snow of volume blx, such that the present radius r and mass m are
related by,
∂r b
πr2 = πr02 + bx, = , m = m0 + ρblx, (8)
∂x 2πr
in the approximation that the log is always circular. Then, relative to the origin, the initial
potential energy V0 , and the present potential energy V  , of the log plus accumulated snow

3
are,
x sin α + b cos α
V0 = m0gr0 cos α − ρblxg , V  = mg(r cos α − x sin α). (9)
2
Redefining the initial potential to be zero, the present potential energy V is,
x sin α + b cos α
V = mg(r cos α − x sin α) + ρblxg − m0gr0 cos α
2
(2πr3 + b2x) cos α + (bx2 − 2πr2 x) sin α
= ρlg − m0gr0 cos α. (10)
2
In the approximation of conservation of mechanical energy E = T + V , we have that,
3πρlr2 v 2 ρlb2 v 2 (2πr3 + b2 x) cos α + (bx2 − 2πr2 x) sin α
0= + + ρlg − m0 gr0 cos α. (11)
4 8π 2
Taking the time derivative of the energy (and dividing by ρl), we obtain the equation of
motion,
3πr2 v dv 3bv 3 b2 v dv gv cos α
0= + + + (3br + b2 ) + g sin α(bvx − πr2 v − bvx), (12)
2 dt 4 4π dt 2
3πr2 dv b2 3bv 2 g cos α
1 + 2 2 = πr2 g sin α − − (3br + b2 ), (13)
2 dt 6π r 4 2
 
dv b2 2g bv 2 g cos α
1+ 2 2 = sin α − 2
− 2
(3br + b2). (14)
dt 6π r 3 2πr 3πr

If we neglect the small terms in b2, the equation of motion is,

dv 2g bv 2 gb
= sin α − 2
− cos α, (15)
dt 3 2πr πr
which disagrees with both eqs. (4) and (6), except in the limit of no snow, b → 0. Note the
appearance in eq. (15) of the term proportional to gb/r, which did not arise in the torque
analyses.

2.3 Lagrangian Method

For completeness, we recall that the equation of motion can also be deduced from the La-
grangian, L = T − V , although strictly this method is for the motion of a rigid body. Here,
we take x as the single, independent coordinate, and note that ẋ = v. Then, recalling eqs. (1)
and (8),
d ∂L ∂L
= , (16)
dt ∂ ẋ ∂x
∂L ∂T 3πρlr2v ρlb2v
= = + , (17)
∂ ẋ ∂v
 2  4π
d ∂L 3ρlπr2 dv b2 3ρlbv 2
= 1+ 2 2 + , (18)
dt ∂ ẋ 2 dt 6π r 2

4
 ∂L ρlg
=− [(3rb + b2 ) cos α + (2bx − 2πr2 − 2bx) sin α], (19)
∂x 2 
3πr2 dv b2 3bv 2 g cos α
1 + 2 2 = πr2 g sin α − − (3rb + b2 ), (20)
2 dt 6π r 2 2
 
dv b2 2g bv 2 g cos α
1+ 2 2 = sin α − 2 − (3rb + b2). (21)
dt 6π r 3 πr 3πr2
The equation of motion (21) differs slightly from eq. (14), as well as from eqs. (4) and (6),
although all four equations agree in the limit that b → 0.

It appears that the approximation of the log as circular at all times as it rolls down a
snowy slope does not lead to a consistent equation of motion.

2.4 Terminal Acceleration of the Rolling Log on the Snowy Slope

It is claimed in [1] that although the acceleration of the rolling log is not constant, it ap-
proaches a constant (terminal) value, such that the motion of the log is eventually similar
to that of rolling on a slope without snow.5
The equation of motion used in [1] is based on a torque analysis about the center of mass,
as in our sec. 2.1.1 above. Because this equation of motion was derived using the unphysical
assumption that the ball/log had a circular cross section at all times, the inference of a
nonzero terminal acceleration is doubtful.

We now present an approximate analysis (that will support the existence of a terminal
acceleration) which avoids use of the assumption that the log has an exactly circular cross
section, instead taking it to be only approximately circular.
The total kinetic energy T of the rolling log can be written as,6
2
mvcm Icm ω 2
T = + , (22)
2 2
where all of the parameters vary with time. We define rcm as the distance from the center of
mass of the log to the line of contact of the log (which rolls without slipping) on the snowy
slope. The mass and moment of inertia of the snow-covered log are approximately related
by,
2
mrcm
m ≈ πρrcm
2
l, and Icm ≈ , (23)
2
where ρ is the mass density of the log, and l is its length. In addition, we approximate the
center-of-mass velocity vcm as,
vcm ≈ ωrcm , (24)
5
This result was obtained in an analysis that (tacitly) assumed conservation of mechanical energy. In
practice, energy is not conserved in the rolling process, which is subject to various forms of energy dissipation
that, in general, lead to a terminal velocity (zero terminal acceleration) of the motion. Analysis of these
processes is beyond the scope of this note (and of [1]).
6
This form does not hold in the approximation that the log is always circular in cross section, as this
requires instantaneous motion of snow from the line of contact to the entire surface of the log.

5
which holds exactly only if vectors vcm and rcm are perpendicular. Then, the kinetic energy
is approximately,
2
3mvcm
T ≈ , (25)
4
and its time derivative is approximately,
2
3ṁvcm 3mvcm v̇cm
Ṫ ≈ + . (26)
4 2
The rolling log accumulates mass from the snowy slope at rate,
mbvcm  
ṁ = ρblvcontact ≈ ρblvcm ≈ 2
and 2
πrcm ≈ πr02 + bxcm , (27)
πrcm
noting that the velocity of the line of contact of the log with the slope is approximately the
same as the velocity of the center of mass of the log. We can also relate the rate of change
ṁ of mass of the log to the rate of change ṙcm of its radius as approximately,

ṁ = 2πρlrcm ṙcm , (28)

which together with eq. (27) implies that,

bvcm
ṙcm ≈ . (29)
2πrcm
As the log rolls down the slope its gravitational potential energy V decreases at a rate
approximately given by,

V̇ ≈ −mgvcm sin α, (30)

where we have neglected the small rate of change of the potential energy associated with the
changing mass of the log.
In the approximation of conservation of mechanical energy, we have that
2
 2

3ṁvcm 3mvcm v̇cm 3mvcm 2g sin α bvcm
Ṫ + V̇ = 0 ≈ + − mgvcm sin α ≈ v̇cm − + 2
. (31)
4 2 2 3 2πrcm
Thus, we obtain an approximate equation of motion,
2
2g sin α bvcm
v̇cm ≈ − 2
, (32)
3 2πrcm
which is similar to (but not the same as) the four equations of motion previously deduced
using the unphysical assumption of a circular cross section of the log at all times. This
gives some confidence that the form of these equations of motion has (approximate) physical
relevance.
A clever suggestion in [1] is to take the time derivative of the equation of motion (32),
 2 
bvcm ṙcm bv̇cm b bvcm
v̈cm ≈ 3
− 2
≈ 2 3
− v̇cm , (33)
2πrcm 2πrcm 2πrcm 2πrcm

6
recalling eq. (29), which indicates that the acceleration v̇cm takes on a constant (terminal)
value,
2 

bvcm 
aterm ≈ 3 
. (34)
2πrcm term

Using this value in the equation of motion (32), we learn that the terminal acceleration is,
g sin α
aterm ≈ , (35)
3
which is 1/2 the acceleration of a solid cylinder that rolls without slipping on a slope without
snow. That is, the log starts from rest on the snowy slope with acceleration 2g sin α/3, but
decelerates (while its velocity increases) until the acceleration is only g sin α/3, after which
the velocity increases linearly with time.
2 2
From eq. (32), we see that v̇cm would be zero if bvcm /2πrcm = 2aterm . For large times,
we have that vcm ≈ atermt, xcm ≈ acm t /2, and from eq. (31), πrcm
2 2
≈ bxcm, such that
bvcm/2πrcm ≈ aterm, so v̇cm remains at aterm at large times, and never drops to zero (i.e.,
2 2

there is no terminal velocity in the present approximations).

The approximate character of the derivation of the equation of motion (32) does not
exclude the existence of an additional small term, such as
gb cos α
− , (36)
πr
that was found in secs. 2.2-3 above. As the radius r of the rolling log grows with time,
the effect of such a term becomes negligible, and the acceleration approaches a constant,
terminal value, albeit this approach is slower than in the absence of the term (36).

3 Single Turn of a Rolling Log

We can make a more physical analysis, without the approximation that the rolling log is
circular at all times, for the first full turn of rolling, as sketched below.

7
 When the “center” of the log, at (x, r0), has moved distance x down the slope, an arc of
angle θ = x/r0 of snow has accumulated over a portion of the surface of the log, where r0 is
the initial radius of the (initially circular) log of length l
We make the (unphysical) assumption that the snow within the circular arc of thickness
b has angular velocity ω about the line of contact at (x, 0), which implies that the snow at
(x, −b) instantaneously takes on velocity ωb. While this is a small (unphysical) effect, it
limits the accuracy of the analysis for large times. Of course, at the end of the first full turn
of rolling, the “step” in the thickness of the snow on the rolling log encounters the snowy
slope, and the analysis is not readily continued.
The initial mass of the log is,7

m0 = πρ l r02 , (37)

and the mass of accumulated snow is, neglecting terms of order b2/r02 ,
θ bθ
ms = πρ l[(r0 + b)2 − r02 ] ≈ ρ b l r0 θ = m 0 , (38)
2π πr0
taking the radial thickness of the layer accumulated on the log to be b, and the density ρ of
the accumulated snow to be the same as that of the initial log. We also suppose that the
thickness of the snow on the slope is b, which implies that the density of the snow on the
slope is slightly greater than ρ.
Once the log has accumulated snow, its center of mass is not at the nominal center (x, r0)
of the log. To analyze this, we first compute the center of mass (xs , ys ) coordinates of the
accumulated snow, neglecting terms of order b2 /r02 ,
r0 +b θ
(r0 + b)3 − r03
ms xs = ρ l r dr dθ  (x − r sin θ) = ms x − ρ l (1 − cos θ)
r0 0 3
b
≈ ms x − ρ l r02 b (1 − cos θ) = ms x − m0 (1 − cos θ) , (39)
π
r0 +b θ
(r0 + b)3 − r03
m s ys = ρl r dr dθ (r − r cos θ ) = ms r − ρl sin θ
r0 0 3
b
≈ ms r0 − ρ l r02 b sin θ = ms r0 − m0 sin θ. (40)
π
The center of mass (cm) coordinates of the rolling log are then related by, to order b/r0 ,
b
(m0 + ms )xcm = m0 x + ms xs = (m0 + ms )x − m0 (1 − cos θ) , (41)
π
b
(m0 + ms )ycm = m0 r0 + ms ys = (m0 + ms )r0 − m0 sin θ , (42)
π
and, noting from eq. (38) that m0/(m0 + ms ) ≈ 1 − b θ/πr0 , the cm coordinates are,
b b
xcm = x − (1 − cos θ) , ycm = r0 − sin θ, (43)
π π
7
If the initial log had very low mass, it would not roll down the slope, as the snow must be lifted as it
sticks to the log.

8
 b θ̇ sin θ b θ̇ cos θ
ẋcm = r0 θ̇ − , ẏcm = − , (44)
π π
b
2
 b
2

ẍcm = r0 θ̈ − θ̈ sin θ + θ̇ cos θ , ÿcm =− θ̈ cos θ − θ̇ sin θ . (45)
π π
The velocity of the nominal center of the log is again v = ẋ, and the angular velocity of
the log is again ω = θ̇ = v/r0 = ẋ/r0 , while now the nominal radius r0 is constant during
the first turn of rolling.
The angular momentum with respect to the center of mass is,

Lcm = dm (x − xcm ) × (v − vcm ) = dm (x − xcm ) × v, (46)

since dm (x−xcm ) = 0. To carry out the integral, we note that for a mass element centered
on (r , θ, z  ) in cylindrical coordinates about the axis of the log, its rectangular coordinates
(x, y, z ) are,

x = x − r sin θ , y  = r − r cos θ , ẋ = ẋ − r θ̇ cos θ = (r − r cos θ )θ̇, ẏ  = r θ̇ sin θ ,(47)

considering that as the log rolls, r of a mass element stays constant while its angular velocity
 
θ̇ is that of the rigid log, θ̇ = θ̇. Then, the angular momentum (46) about the center of
mass has only a z-component,
r0 2π r0 +b θ 
   
Lcm = ρ l r dr dθ + ρ l r dr dθ (x − xcm) × v
 
0 0 r0 0
r0 2π r0 +b θ 
     
= ρ l r dr dθ + ρ l r dr dθ
r0
 0 0
  0

 b  b
x̂ −r0 sin θ + (1 − cos θ) + ŷ −r0 cos θ + sin θ × [x̂ (r0 − r cos θ) + ŷ r sin θ] θ̇
π π
r0 2π r0 +b θ 
     
= −ρ l θ̇ ẑ r dr dθ + r dr dθ
r0
0 0 0

  b r0 b r  
r0 r (1 − cos θ ) + sin θ + [cos θ sin θ + sin θ (1 − cos θ)] (48)
π π
 4   
πr0 b πr03 sin θ b πr03 m0 r02 m0 b r0 θ
≈ −ρ l + + (θ − sin θ) θ̇ ẑ = − + θ̇ ẑ,
2 π π 2 π

to order b/r0 . This is the same as the angular momentum about the axis of the cylinder, to
order b/r0 .

3.1 Torque Analysis

A torque analysis about the center of mass would include the torque due to the contact force
FC on the line of contact of the initial log with the snowy slope. Instead, we perform a
torque analysis in the lab frame about the point xC = (x, 0, 0) of contact of the rolling log
with the snowy slope, as the torque τ C does not involve the contact force FC . However, the

9
rolling log as considered in this section does not have a symmetry axis, in contrast to the
assumption of sec. 2 above, which complicates the torque analysis, as reviewed in [6].
In particular, while the torque analysis involves the time derivative dLC /dt, this does
not equal ∂LC /∂t, such that one should not first compute the angular momentum LC about
point C and then take its (partial) time derivative. Furthermore, there are two possible
meanings of the angular momentum about point C, which can be termed the absolute angular
momentum,

LC = dm (x − xC ) × v, (49)

and the relative angular momentum,


LC = dm (x − xC ) × (v − vC ). (50)

And, there are at least three possible interpretations of point C, as fixed in the lab frame, as
fixed in the rolling log, and as the moving point of contact along the snowy slope. As such,
there are (at least) five variants of torque analyses based on the point C of contact [6].
Here, we consider use of the absolute angular momentum LC about the point of contact,
taking this to be that point C in the rolling log that happens to be the point of contact at
the time of interest. In this convention, the point C of contact is instantaneously at rest,
vC = 0, but it has nonzero acceleration perpendicular to the snowy slope. Then, according
to eq. (13) of [6], the torque equation of motion is,

dLC dLcm
= τC = + (xcm − xC ) × m acm . (51)
dt dt
The torque τ C about the line of contact is due to the force of gravity, (m0 + ms ) g =
(m0 + ms )(x̂ g sin α − ŷ g cos α), which acts at the center of mass, whose position relative to
the line of contact is x̂ (xcm − x) + ŷ ycm ,

τ C = [x̂ (xcm − x) + ŷ ycm ] × (m0 + ms )(x̂ g sin α − ŷ g cos α)

= (m0 + ms ) g [(xcm − x)(− cos α) − ycm sin α] ẑ
    
bθ b b
≈ m0 g 1 + (1 − cos θ) cos α − r0 − sin θ sin α ẑ
πr0 π π
  
bθ b b
≈ −m0 g r0 + − sin θ sin α − (1 − cos θ) cos α ẑ. (52)
π π π

From eq. (48), we have,

 
dLcm m0 r02 m0 b r0 θ m 0 b r0 2
=− + θ̈ ẑ − θ̇ ẑ. (53)
dt 2 π π

and also,

10
   
b b
(xcm − xC ) × m acm = −x̂ (1 − cos θ) + ŷ r0 − sin θ
π π
  
  

bθ b 2 b 2
×m0 1 + x̂ r0 θ̈ − θ̈ sin θ + θ̇ cos θ − ŷ θ̈ cos θ − θ̇ sin θ
πr0 π π
  
b r
2
 b r θ̈ sin θ b r θ θ̈
0 0 0
≈ m0 − r02 θ̈ − θ̈ sin θ + θ̇ cos θ + − ẑ
π π π
 
br
2
= −m0 r0 θ̈ +
2
θ θ̈ − 2θ̈ sin θ − θ̇ cos θ ẑ. (54)
π

With eqs. (52)-(54) in eq. (51) we have the torque equation of motion,
   
dLC m0 r02 m0 b r0 θ m 0 b r0 2 b r0
2
− = + θ̈ + 2
θ̇ + m0 r0 θ̈ + θ θ̈ − 2θ̈ sin θ − θ̇ cos θ
dt 2 π π π
 2 
3r0 b r0 m 0 b r0 2
= m0 + (2θ − 2 sin θ) θ̈ + (1 − cos θ) θ̇
2 π π
  
bθ b b
= −τ C ≈ m0 g r0 + − sin θ sin α − (1 − cos θ) cos α .(55)
π π π

For later comparisons we note that (to order b/r0) the moment of inertia IC of the log
(plus accumulated snow) about the line of contact is,
r0 +b θ
3m0 r02
IC = + ρ l r dr dθ (r02 + r2 − 2r0 r cos θ )
2 r0
 0 
2 (r0 + b) − r0 (r0 + b)4 − r04 (r0 + b)3 − r03
2 2 2
3m0 r0
= + ρ l θ r0 + − 2ρ l r0 sin θ
2 2 4 3
3m0 r02 2m0 b r0 2m0 b r0
≈ + θ− sin θ. (56)
2 π π

It is noteworthy that although LC = −IC θ̇, the torque equation (55) can be written as,

dLC θ̇ dIC
− = IC θ̈ + = −τ C , (57)
dt 2 dt

rather than d(IC θ̇)/dt = −τ C , as first deduced by Loney (1909), Art. 214, p. 287 of [3].
We also make the substitutions v = r0 θ̇ and dv/dt = r0 θ̈ to write the torque equation
(55) as,
  
dv m0 b v 2 bθ b b
IC + (1 − cos θ) ≈ m g r0 r0 + − sin θ sin α − (1 − cos θ) cos α . (58)
dt π π π π

From this equation, which holds only for the first turn of the rolling motion, it is not very
apparent that the acceleration dv/dt approaches a constant value. See also sec. 3.4 below.

11
3.2 Energy Analysis
The rolling log plus accumulated snow is instantaneously rotating with angular velocity ω
about the line of contact, so its kinetic energy is,8

IC ω 2 IC v 2 dT IC v dv v 2 dIC IC v dv m0 b v 3
T = = , = 2 + = 2 + (1 − cos θ). (59)
2 2r02 dt r dt 2r02 dt r0 dt πr02

Relative to the origin, the initial gravitational potential energy V0 , and the present po-
tential energy V  , of the log plus accumulated snow are,
x sin α + b cos α
V0 = m0 g r0 cos α − ms g , (60)
2
V  = m0 g(r cos α − x sin α) + ms g(−xs sin α + ys cos α). (61)

Redefining the initial potential to be zero, the present potential energy V is, to order b/r0,

x sin α + b cos α
V = −m0 g x sin α + ms g(−xs sin α + ys cos α) + ms g
  2
b
≈ −m0 g x sin α − ms x − m0 (1 − cos θ) g sin α
π
 
b x sin α + b cos α
+ ms r0 − m0 sin θ g cos α + ms g
π 2
 2 
m0 bθ b
≈ −m0 g r0 θ sin α − − m0 (1 − cos θ) g sin α
2π π
 
bθ b
+ m0 − m0 sin θ g cos α. (62)
π π
 
dV v v θ b v b
− = m0 g r0 sin α + m0 − sin θ g sin α − m0 g (1 − cos θ) cos α. (63)
dt r0 r0 π π r0 π

Assuming that mechanical energy is conserved during the rolling, dT /dt = −dV/dt, and we
arrive at the equation of motion,
  
dv bθ b b m0 b v 2(1 − cos θ)
IC = m 0 g r0 r 0 + − sin θ sin α − (1 − cos θ) cos α − , (64)
dt π π π π

in agreement with eq. (58).

3.3 Lagrangian Analysis

We now take the independent coordinate to be θ, with θ̇ = ω = v/r0. Then,

d ∂L d ∂L ∂L
= = , (65)
dt ∂ θ̇ dt ∂ω ∂θ
8
We could not use eq. (59) when we assumed that the rolling log is always circular, as this implies that
it is not a rigid body, but has instantaneous motion of snow over its entire surface.

12
 ∂L ∂T IC v
= = IC ω = , (66)
∂ω ∂ω r
d ∂L IC dv 2m0 b v 2
= + (1 − cos θ), (67)
dt ∂ω r0 dt πr
 
∂L m0 b v 2 b b b
= (1 − cos θ) + m0 g sin α r0 + − sin θ − m0 g cos α (1 − cos θ) , (68)
∂θ πr π π π
  
dv b b b m0 b v 2
IC = m0 g r0 r0 + − sin θ sin α − (1 − cos θ) cos α − (1 − cos θ), (69)
dt π π π π
which agrees with eqs. (58) and (64).

3.4 Motion during the Second (and Later) Full Turn

At (or slightly before) the end of the first full turn of rolling, the new layer of snow on the
log encounters the layer of snow still on the slope, and a kind of collision occurs. Here, we
make the idealization that just after this collision the snow log is circular with radius and
mass given by
   
(r0 + b)2 − r02 2b
y1 = r1 = r0 + b, 2
m1 = 2πρ l r1 = m0 1 + ≈ m0 1 + , (70)
r02 r0
and that its center is at (x1 , y1) = (2πr0, r0 + b) = (2πr0 , r1). The configurations of the
snowy log just before and after the collision are illustrated below.

After the collision, the snowy log rolls and accumulates more snow in the manner of its
behavior during the first full turn, except that the initial velocity v1 is not zero, and the
roles of r0 and m0 during the first full turn are played by r1 and m1 during the second full
turn. Again, assuming that energy is conserved during the rolling motion of the second full
turn, the equation of motion follows from eq. (64) (or eq. (58) or (69)) as,
  
dv bθ b b m1 b v 2(1 − cos θ)
IC1 = m 1 g r1 r1 + − sin θ sin α − (1 − cos θ) cos α − , (71)
dt π π π π
where,
3m1 r12 2m1 b r1 2m1 b r1
IC1 ≈ + θ− sin θ, (72)
2 π π

13
to order b/r1 .
If we accept the above model of the collision at the end of a full turn, we can extrapolate
that during the nth full turn, the equation of motion is,
  
dv bθ b b mn b v 2(1 − cos θ)
ICn = m n g rn rn + − sin θ sin α − (1 − cos θ) cos α − ,(73)
dt π π π π

where,

yn = rn = rn−1 + b = r0 + nb, xn = xn−1 + 2πrn−1 = 2πnr0 + n(n − 1)πb, (74)

 2
2 nb
mn = πρ l rn = m0 1 + , (75)
r0

and,

3mn rn2 2mn b rn 2mn b rn

ICn ≈ + θ− sin θ, (76)
2 π π
to order b/rn .
For large n, ICn → 3mn rn2 /2, and the equation of motion (73) becomes,

dv 2g sin α 2v 2(1 − cos θ)

≈ − , (77)
dt 3 3πn2 b
Now, by conservation of energy, the velocity during the nth turn is roughly given by,

3mn v 2 mn g xn sin α πg n2 b sin α

≈ , v2 ≈ , (78)
2 2 3
for large n, such that,
dv 2g sin α 2g sin α(1 − cos θ) 4g sin α
≈ − ≈ , (79)
dt 3 9 9
where the last form is the average acceleration. However, the term 1 − cos θ in eq. (77) oscil-
lates between 0 and 2, and indicates that the snowy log never achieves a steady acceleration,
but rather oscillates about the value (79).9,10
For completeness, we note that the velocity vn,beg and vn,end of the center of the log at the beginning
and end of the nth full turn follows from the assumption of conservation of energy except during the collision
at the end of a turn (and that there is no “hopping”).
At the beginning of the nth turn, the snowy log is circular, with mass mn−1 , radius rn−1, angular velocity,
vn,beg
θ̇n,beg = , (80)
rn−1
9
The results of sec. 2 above, which suggest that the acceleration of the snowly log takes on a terminal
velocity, is an artifact of the assumption that the log is perfectly circular in cross section at all times. In
reality, the accumulation of snow is not azimuthally symmetric, and therefore the acceleration of the log is
never steady.
10
Conceivably, this oscillatory acceleration could be associated with “hopping”, in which the snowball
briefly leaves the snowy surface once each turn. Compare with the discussion in Appendix A.4 of [7].

14
moment of inertia about the point C of contact with the snowy slope,
2
3mn−1 rn−1
IC,n,beg = , (81)
2
kinetic energy,
2 2
IC,n,beg θ̇n,beg 3mn−1 vn,beg
Tn,beg = = (82)
2 4
and potential energy defined to be zero,

Vn,beg = 0. (83)

At the end of the nth turn, the snowy log is again circular, with mass mn and radius rn , while the
distance from the point C of contact with the snowy slope is still rn−1 , so the various kinematic parameters
are now,
vn,end
θ̇ n,end = , (84)
rn−1
2 mn rn2 2
IC,n,end = Icm,n,end + mn rn−1 = + mn rn−1 (85)
2
2   2
IC,n,end θ̇ n,end r2 vn,end
Tn,end = = mn 1 + 2n , (86)
2 2rn−1 2
 
ρ l b xn
Vn,end = (mn − mn−1 )g cos α − mn − g xn sin α. (87)
2

Conservation of energy during the nth turn then tells us that,

Tn,beg = Tn,end + Vn,end , (88)

2 2
which gives a somewhat lengthy expression for vn,end in terms of vn,beg and other, known kinematic param-
eters.
All that remains is to relate the velocities just before and after the collision at the end of a full turn.
For this, we note that to a good approximation, the angular momentum about the point C of contact is
conserved during the collision,
IC,n,end vn,end IC,n+1,beg vn+1,beg
LC,n,end = LC,n+1,beg , = . (89)
rn−1 rn

A Appendix: Cylinder Rolling on a Snowless Slope

As in the figure on p. 1, the coordinate x of the line of contact of the cylinder with the slope
has nonzero acceleration a = dv/dt = d2 x/dt2 . This is also the x-coordinate of the center of
mass of the rolling log.
Force/torque analyses about points P that depend on the coordinate x(t) of the line of
contact can be done either in the lab frame or in an accelerated frame based on point P .
In the latter case, the analysis must include the “fictitious” (coordinate) force −m aP which
appears to act on the center of mass of the cylinder in the accelerated frame.

We first note that the moments of inertia for a solid cylinder of mass m and radius r are,
mr2 3mr2
Icm = , Icontact = . (90)
2 2

15
 Also, as in the figure on p. x = distance along the slope to the point of contact
We consider rolling without slipping, such that the angular velocity ω of the rolling
cylinder is related by,
v
ω= . (91)
r

A.1 Energy Analysis

The cylinder is in instantaneous rotation about the point of contact, so the kinetic energy T
(in the lab frame) is related by,

Icontact ω 2 3mr2 ω 2 3mv 2

T = = = . (92)
2 4 4
The gravitational potential energy V is related by,

V = −mgx sin α, (93)

taking V to be zero when x = 0.

The total mechanical energy E = T + V is constant (neglecting possible dissipative
interactions), so,

dE
= 0 = 3mva/2 − mgv sin α, (94)
dt
and the acceleration a of the cylinder down the slope is,
2g
a= sin α. (95)
3

A.2 Lagrangian Analysis

The equation of motion can also be deduced from the Lagrangian, L = T − V . Here, we
take x as the single, independent coordinate, and note that ẋ = v. Then,
d ∂L ∂L
= , (96)
dt ∂ ẋ ∂x
∂L ∂T 3mv
= = , (97)
∂ ẋ ∂v 2
d ∂L 3ma
= , (98)
dt ∂ ẋ 2
∂L
= mg sin α, (99)
∂x
2g
a= sin α. (100)
3

16
A.3 Torque Analysis about the Line of Contact
A.3.1 Analysis in the Lab Frame
The lab-frame torque equation about the line of contact is simply,
 
d d d 3mr2 v 3mra
τ contact = rmg sin α = Lcontact = (Icontact ω) = = , (101)
dt dt dt 2 r 2
and hence,
2g
a= sin α. (102)
3

A.3.2 Analysis in the Accelerated Frame

When considering a torque analysis about a point xP (t) that is accelerated in the lab frame,
it is often tacitly assumed that the origin of the accelerated coordinate system at time t is
at the point P . However, we do not make this assumption here.
The x-y coordinates of the line of contact are xP = (x, 0), so the lab-frame acceleration
of the reference point is aP = a x̂. In the accelerated frame, the center of mass of the rolling
cylinder is at rest, so its acceleration acm in the accelerated frame is zero. That is, we can
rewrite the lab-frame relation Fx + mg sin α = macm as

Fx + mg sin α − ma = macm = 0, (103)

where “fictitious” force −m aP = −ma x̂ acts on the center of mass of the cylinder.
A torque analysis about a general point P in the accelerated frame can be written in the
form,
dLP
τ P = , (104)
dt
where the torque τ P is the sum of the torque about P in the (inertial) lab frame and the
“fictitious” torque associated with the “fictitious” force that appears to act on the center of
mass of the cylinder,

τ P = τ P + (xcm − xP ) × (−m aP ), (105)

and the angular momentum LP relative to the moving point P is related lab-frame quantities
by (see, for example, sec. 3 of [8], and the Appendix of [9]),


LP = dm[(x − xP ) × (v − vP )] = L − xP × mvcm − (xcm − xP ) × mvP
= Lcm + m(xcm − xP ) × (vcm − vP ) = LP − (xcm − xP ) × mvP , (106)

where,

L= dm x × v = Lcm + xcm × mvcm , (107)

17
is the (lab-frame) angular momentum about the origin, and,

LP = dm (x − xP ) × v = L − xP × mvcm = Lcm + (xcm − xP ) × mvcm , (108)

is the angular momentum about point P regarding this point as fixed in the lab frame.11
For the present case where P = (x(t), 0) is on the line of contact, vP = vcm = v, so,
mr2 v mrv
LP = Lcm , LP = Lcm = Icm ω = = , (111)
2 r 2
and,
τ P = rmg sin α − rma. (112)
The torque equation (104) implies,
dLP mra
τ P = rmg sin α − rma = = , (113)
dt 2
and hence,
2g
a= sin α, (114)
3
as found previously.

A.4 Torque Analysis about the Center of Mass

A.4.1 Analysis in the Lab Frame
The force F that acts on the cylinder at the line of contact has components related by,
Fx + mg sin α = ma, Fy − mg = 0, (115)
The lab-frame torque equation about the center of mass is,
d d d mrv mra
τ cm = −rFx = −rma + rmg sin α = Lcm = (Icm ω) = = , (116)
dt dt dt 2 2
3mra
= mgr sin α , (117)
2
and again,
2g
a= sin α. (118)
3
11
In general, the torque equation for LP of eq. (108) is not dLP /dt = τ P , but rather,
dLP
= τ̃ P = τ P + mvcm × vP . (109)
dt
as discussed, for example, in sec. 11.8, p. 256 of [10] and sec. 2 of [8]. See also [11].
For the present case where P = (x(t), 0) is on the line of contact, vP = vcm = v, and xP is parallel to
vcm , so τ̃ P = τ P = rmg sin α. Then, from eq. (108), LP = LC M + mrv = 3mrv/2, and the torque equation
(109) implies,
dLP d 3mrv 3mra 2g
= = τ̃ P = rmg sin α, and hence, a= sin α, as found previously. (110)
dt dt 2 2 3

18
A.4.2 Analysis in the Accelerated Frame
In the accelerated frame defined by the position of the center of mass of the cylinder at point
P = xcm = (x, r), we have that xP = xcm and vP = vcm . Then, according to eqs. (105) and
(116),

τ P = τ P = τ cm , τ P = −rma + rmg sin α, (119)

and according to eq. (106),

LP = Lcm. (120)

The torque equation τ cm = dLcm /dt in the accelerated frame implies,

dLcm dLcm mra

τ cm = −rma + rmg sin α = = = , (121)
dt dt 2
and again the equation of motion is,
2g
a= sin α. (122)
3

A.5 Torque Analysis about the Point (x, 2r)

A.5.1 Analysis in the Lab Frame
The moment of inertia Ip about this point is the same, 3mr2 /2, as about the line of contact.
The force F that acts on the cylinder a the line of contact has components related by
eq. (115), and the lab-frame torque equation about P is,

τ P = −2rFx − rmg sin α = −2mar + mgr sin α. (123)

The lab-frame angular momentum about point P is,

mrv mrv
LP = Lcm + (xcm − xP ) × mvcm, LP = − rmv = − (124)
2 2
so the lab-frame torque equation is
dLP mra
τ P = −2mar + mgr sin α = =− , (125)
dt 2
and again,
2g
a= sin α. (126)
3

19