See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/344456087

Revisiting the Newton's second law and Lagrange's equations in orthogonal

curvilinear coordinates: some remarks

Preprint · October 2020

CITATIONS

2 authors, including:

J. H. Muñoz
University of Tolima
55 PUBLICATIONS   303 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Nonleptonic charmless two-body B decays involving tensor mesons in the covariant light-front approach View project

All content following this page was uploaded by J. H. Muñoz on 03 October 2020.

The user has requested enhancement of the downloaded file.

Revisiting the Newton’s second law and Lagrange’s equations in orthogonal
curvilinear coordinates: some remarks

J. Guerreroa and J. H. Muñozb

Departamento de Física, Universidad del Tolima, Ibagué, Colombia

Abstract
In this paper it is revisited the derivation of the Lagrange’s equations from Newton’s second law using
curvilinear coordinates. The time derivative of the covariant form of the momentum is our starting point.
Following this way, it is found that the gradients of the Lagrangian and the kinetic energy describe the
effective and fictitious forces, respectively. Also, it is obtained a vector equation of motion in terms of
gradients in coordinates and velocities. This treatment is applied to four illustrative examples involving
polar and spherical frames of reference.

Keywords: Curvilinear coordinates; Newton’s second law; Lagrange’s equations

a
[email protected]
b
[email protected]
I. INTRODUCTION the gradients of the kinetic energy and the La-
grangian, respectively.
It is known that the treatment of the Newton’s
Some important considerations in the mathe-
second law in curvilinear coordinates provides
matical developments are necessary. One of
another way to obtain the Lagrange’s equations.
them refers to the explicit inclusion of the scale
However, some advanced classical mechanics
factors in the mathematical computations, mean-
textbooks [1–9] do not present this approach
ing that the Einstein summation convention [12]
that was developed, explicitly, since 1949 by
is not used here. Another important assumption
Synge and Schild [10]. This omission is prob-
is related to some mathematical tools provided
ably because it requires an extensive and cum-
by differential geometry and tensor calculus, as
bersome knowledge of tensor calculus. Never-
the reciprocal base vectors[13–16], which allow
theless, the absence of a curvilinear handling of
to obtain important information about the phys-
Newton’s second law could diminish the scope
ical properties of motion, once the base vectors
of a complete physical interpretation of the na-
have been normalized.
ture of motion, hiding important information
This paper is organized as follows: section II
about the movement of the particles.
shows some important mathematical tools used
In this work, it is revisited the approach to ma- in curvilinear coordinates. Section III is dedi-
nipulate the Newton’s second law of a particle cated to the Newton0 s second law in curvilinear
of mass m in curvilinear coordinates, motivated coordinates. The Lagrange’s equations are de-
by the interesting introduction of the paper of rived in section IV. A vector equation of motion
Casey [11]. The starting point in our procedure in terms of gradients in coordinates and veloci-
is to consider d~p/dt in a covariant basis. It is a ties is obtained in section V. In section VI, the
differentiator element in our approach. In this developed formalism is applied to some illus-
direction, it is showed that important concepts trative examples. Finally, in section VII some
arise as the gradient of the kinetic energy, which concluding remarks are given.
describes the fictitious forces acting on the ob-
ject, and the gradient of the Lagrangian, which
refers to the effective force in the system.
II. SOME GEOMETRY
One of the purposes of this note is to show that
it is easy, in this treatment, to identify the fic- In this section we provide some important math-
titious (or inertial) and effective forces in many ematical tools which are necessary to develop
problems of classical mechanics, by means of the following sections.
2
A. Base vectors B. Derivative of base vectors

Let us assume that the set {~e1 , ~e2 , ~e3 } denotes If {q k } denotes a set of curvilinear coordinates,
a covariant orthogonal vector basis in some then the partial derivative of the covariant vector
three-dimensional vector space, while the set ~ei with respect to q k is given by [13, 16, 17]
{~e1 , ~e2 , ~e3 } represents the contravariant vector ∂~ei X
= Γlik~el , (6)
basis. It is known that ∂q k l
~el ~el
êl = = , (1) where Γlik is the Christoffel symbol. Mean-
|~el | hl
where ~el is a covariant vector, êl is a vector of while, the derivative of the covariant vector ~ei
unit magnitude in the same direction of ~e , and with respect to the time is
l

hl is the scale factor (magnitude of ~el ). In the d~ei X ∂~ei k

= q̇ ,
dt ∂q k
same way k
~el ~el
êl = = , (2) which leads to
|~el | hl
where ~el is a contravariant vector. It is important d~ei X l k
= Γik q̇ ~el . (7)
dt
to know that êl = êl since êl and êl are vectors kl

in the same direction. Furthermore, hl = 1/hl . On the other hand

Thus, from Eq. (2), it is obtained d~ei X ∂~ei k X
= k
q̇ = − Γilk q̇ k~el , (8)
lêl dt ∂q
~e = . (3) k kl
hl
where [13, 16, 17]
On the other hand, it is known that the covariant
base vectors ~ei are related to the contravariant ∂~ei X
= − Γilk~el . (9)
i
∂q k
base vectors ~e as follows: l

X
~ei = gij ~ej , (4) C. Relation between dq and dq k
k
j

where gij represents the components of the

The differential of the position vector ~r is given
doubly−covariant metric tensor in the frame
by
of reference. Using an orthogonal vector basis, X X
d~r = dqi~ei = dq i~ei . (10)
Eq. (4) equals to i i

But
~ei = (hi )2~ei , (5)
X X X
i i 2
or, equivalently, ~e = (h ) ~ei , which can be writ- dqi~ei = dqi hi êi = dqi hi êi
i i i
ten as ~ei = j g ij ~ej using the components g ij
P
X dqi
= ~e . (11)
2 i
of the doubly-contravariant metric tensor. i
(hi )

3
Comparing the last terms of Eqs. (10) and (11), to
it is obtained the following relation between the f~real − ~Ff ict = ~Freal − f~f ict . (13)
covariant and contravariant differentials:
This expression can be interpreted as New-

dqi = (hi )2 dq i ; (12) ton’s second law in its conventional form (f~ =
d~p/dt), i.e., f~ is given by f~real − ~Ff ict and d~p/dt
This expression can be generalized as Ai = corresponds to ~Freal − f~f ict .
(hi )2 Ai , giving a relation between the covari-
ant Ai and contravariant Ai components of any A. d~p/dt in the covariant basis {~e }
i
~
vector A.
The three-dimensional contravariant linear mo-
mentum of a particle of mass m, written as a
III. NEWTON’S SECOND LAW IN CURVI-
linear combination of covariant base vectors ~ei ,
LINEAR COORDINATES
is given by
X
In this section, Newton’s second law for non- p~ = m q̇ i~ei . (14)
i
conservative systems will be manipulated in a
Applying the derivative with respect to the time,
curvilinear frame of reference, using reciprocal
it is obtained
orthogonal vector basis. In this way, it is ob-
d~p X  dq̇ i 
tained the Lagrangian form of Newton’s law of i d~
ei
=m ~ei + q̇ . (15)
dt i
dt dt
motion (sec. IV). For that, it is necessary to in-
clude the kinetic energy into Newtons’s second Using Eq. (7) to describe the time derivative of
law. the base vector ~ei , Eq. (15) becomes
( )
If the potential energy depends on coordinates d~p X dq̇ i X
=m ~ei + q̇ i Γlik~el q̇ k .
and velocities, Newton’s second law, f~ = d~p/dt, dt i
dt kl

can be reorganized as an effective equation of

Rearranging index, this equation can be written
motion (EEM) that relates the effective force
as
~
fef f acting on the particle, with the effective be- ( )
d~p X dq̇ l X l i k
havior ~Fef f experimented by the particle. It is =m ~el + Γik q̇ q̇ . (16)
dt l
dt ki
f~ef f = ~Fef f , where f~ef f is the sum of the real
and fictitious forces (f~ef f = f~real + f~f ict ). In This is the contravariant form of d~p/dt. The
a similar way, ~Fef f is the sum of the real and components of this vector equation are
fictitious behaviors (~Fef f = ~Freal + ~Ff ic ). Thus,
 l
d~p dq̇ l X
=m +m Γlik q̇ i q̇ k . (17)
~ ~ ~ ~
the EEM is freal + ff ict = Freal + Ff ic , leading dt dt ki

4
It is highlighted that the right-hand side of this Using Eq. (8) to describe the time derivative of
equation agrees with Eq. (5.209), page 150, of the base vector ~ei , Eq. (19) is given by
Ref. [10] (also, with Eq. (11.12a), page 156, of ( )
d~p X
l d l 2
X
2 i i k
Ref. [18] and Eq. (47) of Ref. [19]). =m ~e [q̇ (hl ) ] − (hi ) Γlk q̇ q̇ .
dt l
dt ki
It is worth to note that some authors use indis- (20)
criminately covariant instead of invariant, to re- This is the covariant form of d~p/dt. Using Eq.
fer to the contravariant form of Newton’s sec- (3), this equation can be written in terms of the
ond law as covariant form (see Eq. (47) of Ref. normalized base vector êi as:
[19]), which could leads to misconceptions. d~p X m d l Xm
= êl [q̇ (hl )2 ]− (hi )2 q̇ i q̇ k Γilk êl .
dt l
hl dt ikl
hl
(21)
B. p/dt in the contravariant basis {~ei }
d~
From the right-hand side of Eq. (13), the first
In order to obtain d~p/dt in the contravariant or- term in the right-hand side of Eq. (21) is the real
thogonal basis {~ei }, let us transform the covari- behavior ~Freal experimented by the particle; and
ant base vectors ~ei in Eq. (14) into their equiv- the second term is related to the fictitious force
alent contravariant forms, using the doubly- f~f ict acting on the particle, due to changes in the
contravariant metric tensor. Then direction of motion, which are given by the time
X derivative of the contravariant base vectors ~ei .
p~ = m q̇ i (hi )2~ei , (18)
i Taking the components of Eq. (20), it is ob-
where mq̇ i (hi )2 are the covariant components of tained
p~, i.e., pi = mq̇ i (hi )2 .  
d~p d l 2
X
The clue point of the approach developed in = m [ q̇ (hl ) ] − m (hi )2 Γilk q̇ i q̇ k ,
dt l dt ki
this work is to write the linear momentum p~ in (22)
the covariant form, using the contravariant basis which are the covariant components of d~p/dt.
{~ei } (see Eq. (18)), which leads to the covari-
ant form of d~p/dt. The result obtained is then C. Gradient of the kinetic energy
matched with the covariant form of any gen-
eral force f~, yielding Lagrange’s equations in The kinetic energy function of a particle of mass
its general form, given by Eq. (39). m is given by (1/2)m(~v ·~v ), and it can be written
Taking the time derivative, Eq. (18) becomes in terms of the contravariant velocities as fol-
d~ei lows:
 
d~p X d
=m ~ei [q̇ i (hi )2 ] + q̇ i (hi )2 . 1 X i l
dt i
dt dt T = m q̇ q̇ (~ei · ~el ). (23)
(19) 2 il

5
Computing the scalar product between the or- Interchanging i ⇔ l, and then k ⇔ l, it becomes
thogonal base vectors ~ei and ~el (i.e., ~ei · ~el = X
~ =m
∇T ~el q̇ k q̇ i Γikl (hi )2 .
hi hl δil ), Eq. (23) becomes
ikl

1 X
T = m (hi )2 (q̇ i )2 . (24) Using Γikl = Γilk , it is obtained
2 i
X
Applying the covariant derivative [16, 17, 20– ~ =m
∇T ~el q̇ k q̇ i Γilk (hi )2 , (26)
ikl
23] to the kinetic energy in Eq. (23), it is ob-
tained which allows to write Eq. (20) as

~ = 1m ∂ d~p
X X d
∇T ~ek q̇ i q̇ l k (~ei · ~el ). (25) =m ~
~el [q̇ l (hl )2 ] − ∇T. (27)
2 ikl ∂q dt dt
l

This expression involves a bilinear term in the This result shows that the covariant form of
velocities (i.e. q̇ i q̇ l ). To calculate the partial d~p/dt includes the covariant derivative (∇~ =
derivative of ~ei · ~el , it is necessary to use Eq. P ~el ∂ ) of the kinetic energy, plus a vector
l ∂q l
(6) as follows: term (in fact, as it will be outlined later, this
∂ X
m
X
term is related to the canonical momenta in the
k
(~
e i · ~
e l ) = ~
e i · Γlk ~
e m + ~
e l · Γm
ik~
em ,
∂q m m Lagrangian formalism).
which can be written as On the other hand, the derivative of the kinetic
∂
el ·~em ). energy given by Eq. (24), with respect to veloc-
X X
m
k
(~
e i ·~
e l ) = Γlk (~
e i ·~
e m ) + Γm
ik (~
∂q m m ities q̇ l , leads to the following expression:
Substituting ~ei · ~em = hi hm δim and ~el · ~em =
∂T
hl hm δlm , leads to = mq̇ l (hl )2 , (28)
∂ q̇ l
∂
(~ei · ~el ) = Γilk (hi )2 + Γlik (hl )2 . which allows to rewrite Eq. (27) as
∂q k
d~p X l d ∂T ~
Thus, Eq. (25) takes the form = ~e − ∇T. (29)
dt dt ∂ q̇ l
l
~ = 1m
X
T∇ ~ek q̇ i q̇ l [Γilk (hi )2 + Γlik (hl )2 ]
2 ikl This is the covariant form of d~p/dt, in terms
( )
1 X k X i l i of the kinetic energy. According to the right-
= m ~e q̇ q̇ [Γlk (hi )2 + Γlik (hl )2 ] .
2 k il hand side of Eq. (13), the fictitious force f~f ict
But il q̇ q̇ Γlk (hi )2 = il q̇ i q̇ l Γlik (hl )2 . So it is ~ , while the real behavior ~Freal is
P i l i P
is given by ∇T
P l d ∂T
obtained l~
e dt ∂ q̇l . Eq. (29) allows to obtain the most
X general form of the Lagrange’s equations (see
~ =m
∇T ~ek q̇ i q̇ l Γlik (hl )2 .
ikl Eq. (39)).

6
D. Effective equation of motion ~
where ∇(T − U ) is the effective force. Eq. (32)
is an equation of the form f~ef f = ~Fef f . The
In order to derive the EEM, let us consider a par-
right-hand side of Eq. (32), as it will be dis-
ticle of mass m that experiences a force gener-
cussed below, is related to the time variations
ated by a potential energy function that depends
of canonical momenta in Lagrangian mechanics
on coordinates and velocities. It is proposed that
for non-conservative systems.
this force is given by
The fictitious force f~f ict = ∇T
~ depends on co-
X d ∂U
f~ = −∇U
~ + ~el , (30) ordinates and velocities. If the kinetic energy
l
dt ∂ q̇ l
function does not contain any specific coordi-
Eq. (30) is written in the covariant form. Based
nate explicitly, it means that there is not an in-
on the left-hand side of Eq. (13), the first term in
ertial force related with that specific coordinate.
the right-hand side of Eq. (30) is the real force
Furthermore, if the kinetic energy does not con-
f~real , while the second term is the fictitious be-
tain coordinates, it means that the system does
havior ~Ff ict . If the potential energy does not de-
not have fictitious forces, leading to the New-
pend on velocities, ~Ff ict disappears, and the ob-
ton’s second law in rectilinear coordinates. On
tained force corresponds to a conservative sys-
the other hand, the fictitious behavior ~Ff ict =
tem.
− l ~el dtd ∂∂Uq̇l depends on coordinates and veloc-
P
From Eqs. (29) and (30), Newton0 s second law
ities too. If the potential energy does no con-
takes the form
tain velocities, the system does not have ficti-
X d ∂U X d ∂T
−∇U~ + ~el ~ +
= −∇T ~el . tious behavior ~Ff ict , yielding Newton’s second
dt ∂ q̇ l dt ∂ q̇ l
l l
(31) law for conservative systems.
Eq. (31) is the covariant form of Newton’s sec- If the particle is constrained to describe a certain
ond law in terms of the potential and kinetic en- line or surface in space, it is necessary to intro-
ergy functions. Newton’s second law for conser- duce the constraint forces into the left-hand side
vative systems is obtained if ∂U/∂ q̇ l = 0. Ac- of Eq. (32), due to holonomic constraints. [1–
cording to Eq. (13), the real force f~real is −∇U ~ ; 9]. Then

the fictitious behavior ~Ff ict is − l ~el dtd ∂∂Uq̇l ; the

P
X X d ∂
~ ~ ~ −U)+
∇(T λ ~ i=
∇Ω ~el (T − U ),
fictitious force ff ict is ∇T , and the real behavior i
dt ∂ q̇ l
i l
~Freal is P ~el d ∂Tl . (33)
l dt ∂ q̇

Eq. (31) can be reorganized as where Ωi = Ωi (q 1 , q 2 , q 3 ) represents the con-

X d ∂ straint equation in coordinates. The i index runs
~
∇(T − U) = ~el (T − U ), (32)
l
dt ∂ q̇ l over the constraint conditions in the system.

7
The coefficients λi are the Lagrange multipliers IV. LAGRANGE’S EQUATIONS
corresponding to each constraint force. Assum-
First, it is pointed out a brief remark on the gen-
ing the existence of a unique constraint equation
eralized forces Ql and the conjugate momenta
in space (i.e, for motions on a surface), Eq. (33)
℘l (the notation ℘l is used to differentiate it from
becomes
the covariant components pl of the linear mo-
~
X
ld ∂
∇(T − U + λΩ) = ~e (T − U ). (34) mentum p~).
l
dt ∂ q̇ l
It is well known that the generalized forces Ql in
On the other hand, supposing that there is not a
Lagrangian formalism, which are scalar quanti-
real force acting on the particle (i.e. U = 0), the
ties, are related to the force f~ in the Newtonian
Newton0 s second law takes the form
mechanics as follows: [1, 4, 5, 8]
X d ∂T
~
∇(T + λΩ) = ~el
. (35) ∂~r
l
dt ∂ q̇ l
Ql = f~ · l = f~ · ~el . (36)
∂q
Eq. (35) can not describe an accelerated motion This definition means that the generalized forces
via the real force. There would be an acceler- Ql are the projection of the force f~ in the ~el
ated motion via the fictitious force. However, base vectors. Assuming that f~ is written as a
the constraint condition is necessary to counter linear combination of the non−normalized base
the motion in a certain coordinate. In that case, vectors {~el } (i.e., f~ written in its contravariant
Eq. (35) allows to determine the constraint force form), Eq. (36) yields
acting on the object (see example B. in Sec. V). "
X
#
X
Ql = f j ~ej ·~el = f j (~ej ·~el ) = (hl )2 f l .
It is known that the introduction of holonomic
j j
constraint conditions into the kinetic energy (37)
function eliminates some coordinates from its From Eq. (12), it is seen that (hl )2 f l are the co-
expression. For this reason, the constraint con- variant components of the force f~, i.e., the gen-
dition must be introduced once the gradients of eralized forces Ql are the covariant components
kinetic energy and potential energy have been of f~ (Ql = fl ).
calculated. i.e., it is very important to apply the In a similar way, the conjugate momenta ℘l can
constraint conditions at the end of all these com- be defined as scalar quantities by the following
putations, which enable us to obtain important relation:
information related to the motion of the object. ∂~r
℘l = p~ · = p~ · ~el ,
∂q l
Otherwise, no information about fictitious and
or, equivalently
real forces could be obtained through the gradi-
ent of kinetic and potential energies. ℘l = (hl )2 pl ,

8
for conservative systems. Since the covariant where Uef f = U − T = Ureal + Uf ict . From
components pl of p~ are (hl )2 pl , the conjugate this definition, the fictitious potential Uf ict is
momenta ℘l for conservative systems are the co- −T ; while the real potential energy Ureal is, of
variant components of p~ (i.e., ℘l = pl ). course, the original potential energy U .
The components of Eq. (29) are given by On the other hand, Eq. (40) can be written as
 
d~p d ∂T ∂T X ∂L X d ∂L
= l
− l, (38) ~el l = ~el , (42)
dt l dt ∂ q̇ ∂q ∂q dt ∂ q̇ l
l l
which are the covariant components of d~p/dt. If
~ = P ~el ∂ l . Thus
since ∇
fl is the covariant component of a general force l ∂q

f~, the covariant form of Newton’s second law ∂L d ∂L

l
= . (43)
leads to fl = (d~p/dt)l , or, equivalently ∂q dt ∂ q̇ l

d ∂T ∂T These expressions are the Lagrange’s equations,

Ql = l
− l, (39)
dt ∂ q̇ ∂q derived from Newton’s second law in the covari-
since fl = Ql . It is the most general form of ant form, given by Eq. (31).
Lagrange’s equations, derived from the covari- Eq. (43) is written as EEMs that relates the
ant form of f~ and d~p/dt. In this way, the La- effective generalized forces (Ql )ef f in the left-
grange’s equations for point particles have been hand side, given by the sum of real and fictitious
derived via an alternative differential procedure generalized forces ((Ql )real = −∂U/∂q l and
that does not require to use the D’Alembert’s (Ql )f ict = ∂T /∂q l ), with the effective gener-
principle. alized behaviors (Ql )ef f in the right-hand side,
given by the sum of real and fictitious general-
d ∂T
A. Effective equation of motion ized behaviors ((Ql )ef f = dt ∂ q̇ l
and (Ql )f ict =
− dtd ∂∂Uq̇l ).
Using the definition of the Lagrangian function,
If Eq. (31) does not contain the second term
given by L = T − U , it is possible to rewrite Eq.
of the left-hand side (i.e., the term related with
(32) as follows:
the fictitious behavior ~Ff ict ), Eq. (43) will be
X d ∂L
~ =
∇L ~el , (40) Lagrange’s equations for conservative systems.
l
dt ∂ q̇ l
Since Ql are the covariant components of the
~ is the effective force f~ef f . If it is force f~, then (Ql )ef f are the covariant compo-
where ∇L
defined Uef f = −L, Eq. (40) becomes nents of the effective force f~ef f = ∇L,~ i.e.,
X d ∂L (Ql )ef f = (fl )ef f .
~ ef f =
−∇U ~el , (41)
l
dt ∂ q̇ l In the Lagrangian formalism, the conjugate

9
momenta[1–9] are defined as ities. Furthermore, ∂T /∂ q̇ i depends on coordi-

∂L ∂ nates and velocities too. Thus

℘l = l
= l (T − U ), (44)
∂ q̇ ∂ q̇ d ∂T j ∂ ∂T ∂ ∂T ∂ ∂T
X X
i
= q̇ j i
+ q̈ j j i + i
.
which allows to write Eq. (43) as follows: dt ∂ q̇ j
∂q ∂ q̇ j
∂ q̇ ∂ q̇ ∂t ∂ q̇
(47)
∂L d℘l ∂ ∂T ∂ ∂T ∂ ∂T ∂ ∂T
l
= . (45) However, ∂q j ∂ q̇ i
= ∂ q̇ i ∂q j
and ∂t ∂ q̇ i
= ∂ q̇ i ∂t
,
∂q dt
using the Schwarz’s theorem[24]. Then, Eq.
Thus, the effective generalized forces (Ql )ef f
(47) becomes
equals to the time derivative of canonical mo-
d ∂T X ∂ ∂T
menta ℘l . Eq. (45) allows to interpret the canon- = q̇ j i j
dt ∂ q̇ i j
∂ q̇ ∂q
ical momenta ℘l as the effective generalized mo- X ∂ ∂T ∂ ∂T
+ q̈ j j i + i . (48)
menta in Lagrangian mechanics, which has no ∂ q̇ ∂ q̇ ∂ q̇ ∂t
j
equivalent momenta in Newtonian mechanics,
But
due to its vector nature.
∂ q̇ j ∂T
 
∂ j ∂T ∂ ∂T
Using Eq. (28), the conjugate momenta in terms q̇ = q̇ j + ,
∂ q̇ i ∂q j ∂ q̇ i ∂q j ∂ q̇ i ∂q j
of the scale factors are given by
yielding
∂U
℘l = mq̇ l (hl )2 −
 
. (46) j∂ ∂T ∂ j ∂T ∂T
∂ q̇ l q̇ i j
= i q̇ j
− δ ji j , (49)
∂ q̇ ∂q ∂ q̇ ∂q ∂q
It is highlighted that the superscript notation in
since ∂ q̇ j /∂ q̇ i = δ ji . On the other hand
l l
coordinates q and velocities q̇ does not have
∂ q̈ j ∂T
 
∂ j ∂T j ∂ ∂T
mathematical implications in the Lagrange’s q̈ = q̈ + ,
∂ q̇ i ∂ q̇ j ∂ q̇ i ∂ q̇ j ∂ q̇ i ∂ q̇ j
equations, since it obeys the index convention
leading to
from Newton0 s second law in Eq. (32), while  
∂ j ∂T ∂ ∂T ∂ ∂T
Lagrangian formalism involves scalar quantities i
q̈ j
= q̈ j i j = q̈ j j i ,
∂ q̇ ∂ q̇ ∂ q̇ ∂ q̇ ∂ q̇ ∂ q̇
only, so it is possible to simplify the notations (50)
q̇ i → q̇i and q i → qi without loss of generality, since ∂ q̈ j /∂ q̇ i = 0. The Schwarz’s theorem has
yielding the notation used in some textbooks[1– been used again in the last step.
9]. Introducing Eqs. (49) and (50) into (48), it is
obtained

V. THE EQUATION OF GRADIENTS d ∂T X ∂  ∂T 

i
= i
q̇ j j +
dt ∂ q̇ j
∂ q̇ ∂q
 
It is known that the kinetic energy in curvilinear X ∂ ∂T ∂ ∂T ∂T
i
q̈ j j + i − i,
coordinates depends on coordinates and veloc- j
∂ q̇ ∂ q̇ ∂ q̇ ∂t ∂q

10
which can be written as Including constraint forces in Eq. (55) due to
constraint conditions in space, it is obtained
!
d ∂T ∂ X ∂T X ∂T ∂T
= i q̇ j j + q̈ j j + n
~ q̇ dL = ∇
dt ∂ q̇ i ∂ q̇ ∂q ∂ q̇ ∂t X
j j ∇ ~ q (2L) + ~ q Ωj ,
λj ∇
∂T dt j=1
− . (51)
∂q i where the sum includes the total constraint

The quantity inside the parenthesis is dT /dt. forces involved in the system. Assuming the
Thus, Eq. (51) becomes existence of a unique constraint equation (i.e.,
for motions on a surface), this becomes
d ∂T ∂ dT ∂T
= i − i. (52)
dt ∂ q̇ i ∂ q̇ dt ∂q ~ q̇ M = ∇
∇ ~ q N, (56)

In a similar way, it is obtained with N = 2L + λΩ.

d ∂U ∂ dU ∂U
i
= i − i, (53)
dt ∂ q̇ ∂ q̇ dt ∂q VI. ILLUSTRATIVE EXAMPLES

for the potential energy that depends on the co-

In this section, the foregoing presentation will
ordinates and velocities. Based on these expres-
be applied to some two-dimensional and three-
sions, the time derivative of Eq. (44) is
dimensional conservative systems of one parti-
d ∂ ∂ d ∂ 1 2
i
(T − U ) = i (T − U ) − i (T − U ). cle, written in the polar (q , q ) = (r, θ) and
dt ∂ q̇ ∂ q̇ dt ∂q
spherical (q 1 , q 2 , q 3 ) = (r, θ, φ) coordinates.
Substituting this result into Eq. (42), it is ob-
tained A. The two-body central force problem
X êi ∂ dL X êi ∂
= (2L). (54) It is well known that the two−body central force
i
hi ∂ q̇ i dt i
hi ∂q i
problem is reduced to a one-body system [1–
Defining the function N = 2L, Eq. (54) takes
3, 5, 7–9], with reduced mass µ. Using polar co-
the form
ordinates (q 1 , q 2 ) = (r, θ), and assuming a po-
~ q̇ M = ∇
∇ ~ q N, (55)
tential energy U = U (r), the Lagrangian func-
where M denotes the time derivative of the La- tion of the reduced system is
grangian function. ∇ ~ q̇ and ∇
~ q are P êi ∂ i and 1
i hi ∂ q̇ L = µ[ṙ2 + r2 θ̇2 ] − U.
P êi ∂ 2
i hi ∂q i , respectively.
According to previous results, the effective po-
Eq. (55) is formed by a compact gradient de-
tential energy is given by −L. Thus
scription, so it could be seen as the equation of
1
gradients in the Lagrangian mechanics. Uef f = − µ[ṙ2 + r2 θ̇2 ] + U. (57)
2

11
On the other hand, the effective force has the Reminding that l = µr2 θ̇, this equation finally
form f~ef f = ∇L
~ = −∇U~ ef f . It is becomes

f~ef f = (µrθ̇2 )eˆr − ∇U

~ ; (58)
f~ef f = (µrθ̇2 )êr − ∇U.
~
while the gradient of the kinetic energy is
This equation is precisely the same expression
~ = (µrθ̇ )êr + (0)êθ .
∇T 2
obtained in Eq. (58), so it was found the ficti-
tious force by applying the gradient operator to
This equation shows that the fictitious force
the Lagrangian function in the two-body central
does not have component in θ coordinate (in
force problem.
contrast to the radial coordinate).
In this problem, the angular momentum (l =
µr2 θ̇) is a conserved quantity in time. This fact B. Circular motion
allows to write the effective potential in terms of
Consider a particle of mass m that describes
that specific constant of motion [1–3, 5, 7–9] as
a circular motion of radio R. Its Lagrangian
l2
Uef f = + U. (59) function in polar coordinates (q 1 , q 2 ) = (r, θ) is
2µr2
given by
Eqs. (57) and (59) do not agree once the relation
1
2 L = m[ṙ2 + r2 θ̇2 ], (60)
l = µr θ̇ is introduced in Eq. (59), even when 2
they generate the same effective force through with the constraint condition r = R. The dy-
~ e f f , as it will be outlined below. This namics of the system is determined by Eq. (35).
−∇U
is because the effective potential energy given The right-hand side of Eq. (35) has the form
by (57), defined in this article, includes one ad- X êl d ∂T
l
= (mr̈)êr + (mrθ̈ + 2mṙθ̇)êθ ,
ditional term that depends on ṙ. On the other l
h l dt ∂ q̇

hand, the inclusion of l in the kinetic energy while the left-hand side is
given by Eq. (57) generates a change in the sign
~
∇(T + λΩ) = (mrθ̇2 + λ)êr + (0)êθ .
of the exponent of r (i.e., r2 → r−2 ). This fact
implies that this term has to change its sign to So, Eq. (35) becomes
preserve the force generated by minus the gra-
dient of U . (mrθ̇2 +λ)êr +(0)êθ = (mr̈)êr +(mrθ̈+2mṙθ̇)êθ .
ef f

Applying the expression f~ef f = −∇U~ ef f to Eq.

After including the constraint condition r = R,
(59), it is obtained it is obtained
 2 
l
f~ef f = ~
êr − ∇U.
µr3 (mRθ̇2 + λ)êr + (0)êθ = (0)êr + (mRθ̈)êθ .

12
Thus, the constraint force in r coordinate is right-hand side represents the fictitious force
given by λ = −mRθ̇2 , while the motion in θ acting on the particle. This is
coordinate is a conserved quantity in the time,
~ = mg(cos θ)êr − mg(sin θ)êθ ,
−∇U (61)
since mRθ̈ = 0.
and

C. The spherical pendulum ~ = mR(θ̇2 + φ̇2 sin2 θ)êr +

∇T (62)
mR(φ̇2 sin θ cos θ)êθ +
Consider a spherical pendulum of mass m and
length R. Using spherical coordinates (q 1 , q 2 , q 3 ) = (0)êφ .

(r, θ, φ), the Lagrangian function has the form Due to the constraint equation Ω = r − R, the
1 effective force in r coordinate is countered by
L = m[ṙ2 + r2 θ̇2 + (r sin θ)2 φ̇2 ] + mgr cos θ,
2 the constraint force (i.e., f~ = λ∇Ω),
~ where the

with the constraint condition r = R. Lagrange multiplier λ will be calculated below.

Based on the ideas previously developed, the ef- Obviously, there is no motion in the radial coor-
fective potential energy is given by U = −L; dinate.
ef f

thus, the effective force has the form f~ef f = If the components of Eq. (62) are multiplied by
~ and the fictitious force is f~f ict = ∇T
∇L, ~ . Of the scale factor hi in each coordinate, the results

course, it is very important to introduce the con- obtained coincide with the values obtained by
straint condition r = R once all these expres- Lagrangian formalism. This is because of the
sions have been calculated. Therefore, the ef- form of the generalized force in Lagrangian me-
fective force in the spherical pendulum system chanics (see Eq. (36)), which is calculated by
is given by the projection of the force in the non-normalized
covariant base vectors ~ei .
f~ef f = m(g cos θ)êr − m(g sin θ)êθ + Returning to the expression (62), there is not in-
mR(θ̇2 + φ̇2 sin2 θ)êr + ertial force associated to the azimuth coordinate,

mR(φ̇2 sin θ cos θ)êθ + because the φ variable does not appear in the ki-
netic energy function. Furthermore, the radial
(0)êφ .
component contains two inertial terms; each of
The vector formed by the two first terms on the them depends on some of the two velocities θ̇
right-hand side describes the real force gener- or φ̇, showing that there exist inertial forces due
ated by the potential energy function; while the to motions in direction of θ and φ coordinates.
vector formed by the last three terms on the The coefficient in the êθ base vector has a unique

13
term, which depends on the φ coordinate, show- However, there exist other terms in Eq. (63)
ing that there is a fictitious force in êθ direction that are not included in Eq. (62). They can
due to a motion in the direction of φ coordinate. be gotten from the time derivative of conjugate
The zero coefficient in êφ means that there is not momenta in Lagrangian mechanics (Eq. (44)),
inertial or real force acting in êφ direction, yield- which must be divided by the scale factors to
ing a conserved linear momentum component. obtain the real effect in Newtonian mechanics.
As seen before, the effective potential energy in According to Eq. (44), the equations of motion
the spherical pendulum is given by the negative for each coordinate can be obtained. Once the
of its Lagrangian function (Uef f = −L). So, constraint condition r = R has been included,
the left-hand side of this equation yields
Uef f = − mgrcosθ −
X êi d℘i
1 h 2 i
= mR(θ̈)êθ + (64)
m ṙ + r2 θ̇2 + (r sin θ)2 φ̇2 . h i dt
2 i

Introducing the constraint condition r = R, it mR(sin θφ̈ + 2φ̇θ̇ cos θ)êφ ,

becomes while the right-hand side is given by

1 h 2 2 2 2
i ~
∇(L + λΩ) = (65)
Uef f = −mgRcosθ− m R θ̇ + (R sin θ) φ̇ .
2
m[g cos θ + R(θ̇2 + φ̇2 sin2 θ)]êr −
On the other hand, if the constraint condition
mR(g sin θ + φ̇2 sin θ cos θ)êθ + λêr ,
r = R is introduced in the time derivative of the
linear momentum, written in spherical coordi- where Ω = r − R. Based on Eqs. (64) and (65),
nates, it is obtained the Lagrange multiplier in r coordinate has the

d~p form
= −mR(θ̇2 + φ̇2 sin2 θ)êr + (63)
dt λ = −mRθ̇2 − mRφ̇2 sin2 θ − mg cos θ.
mR(θ̈ − φ̇2 sin θ cos θ)êθ +
The first term of right-hand side of this equation
mR(2φ̇θ̇ cos θ + φ̈ sin θ)êφ .
refers to the fictitious force due to the motion in
Comparing the components of Eq. (62) with the θ coordinate; while the second one describes the
components of this last vector, the time deriva- fictitious force caused by the motion in φ coor-
tive of the linear momentum contains the ficti- dinate. The third therm refers to the projection
tious force, given by the gradient of the kinetic of the gravitational force on r coordinate direc-
energy function, except by a minus sign, so it tion.
can be written Meanwhile, the equation of motion for θ coor-
d~p ~ + mR(θ̈)êθ +
= −∇T dinate has the form
dt
mR(2φ̇θ̇ cos θ + φ̈ sin θ)êφ . θ̈ = −(g + φ̇2 cos θ) sin θ,

14
and the equation of motion for φ coordinate is According to Eq. (56), it is obtained

mR(sin θφ̈ + 2φ̇θ̇ cos θ) = 0. (66) N = m[ṙ2 + r2 θ̇2 + (r sin θ)2 φ̇2 ] −

However, the conjugate momentum associated 2mgr cos θ + λ(θ − α);

to φ coordinate has the form ℘φ = mφ̇(r sin θ)2 .
while M is given by the derivative of the La-
The component of the time derivative of the lin-
grangian function with respect to the time. It is
ear momentum in φ coordinate, after including
the constraint condition r = R and dividing by M = mṙr̈ + m(rṙθ̇2 + r2 θ̇θ̈) +
the scale factor hφ = R sin θ, is mφ̇2 r sin θ(ṙ sin θ + rθ̇ cos θ) +
1 d℘φ mφ̇φ̈(r sin θ)2 .
= mR(sin θφ̈ + 2φ̇θ̇ cos θ);
hφ dt
thus, the Eq. (66) becomes Thus
1 d℘φ
= 0. ~ q̇ M = m(r̈ + φ̇2 r sin2 α)êr +
∇ (68)
hφ dt
This result implies that the conjugate momen- m(rφ̇2 sin α cos α)êθ +
tum ℘φ is a conserved quantity in the time, as m(φ̈r sin α + 2φ̇ṙ sin α)êφ ,
might be expected.
where the constraint θ = α has been included.
On the other hand, the gradient in coordinates
D. Particle moving on a conic surface
of N function including the constraint θ = α, is
Consider a particle of mass m subject to a gravi- given by
tational force, constrained to move on the inside ~ q N = (−2mg cos α + 2mrφ̇2 sin2 α)êr +
∇
surface of a smooth cone of half angle α. Its
(2mrφ̇2 sin α cos α + 2mg sin α + λr−1 )êθ
motion is described by the equation of gradients
+(0)êφ . (69)
(Eq. (56)).
In spherical coordinates, (q 1 , q 2 , q 3 ) = (r, θ, φ), Based on Eqs. (68) and (69), Eq. (56) finally
the Lagrangian function, without the inclusion yields the equation of motion for each coordi-
of the constraint θ = α, is given by nate. For r, it is obtained
1 h i
L = m ṙ2 + r2 θ̇2 + (r sin θ)2 φ̇2 −mgr cos θ.
2 r̈ + r(φ̇ sin α)2 = −2g cos α + 2r(φ̇ sin α)2 ,
(67)
As mentioned previously, the constraint condi- while the equation of motion for θ does not de-
tion must be introduced once the gradients have scribe any motion, due to the constraint condi-
been computed. tion θ = α. The equation obtained allows to

15
determinate the Lagrange multiplier associated and
to θ coordinate. It is
~ q T = 2mrφ̇2 sin α[(sin α)êr + (cos α)êθ ].
∇
λ = −2(mr2 φ̇2 cos α + mgr) sin α. (70)
This last equation means that there are fictitious
Finally, the equation of motion for φ is given by
forces in r and θ coordinates, in contrast with φ
m(φ̈r + 2φ̇ṙ) sin α = 0. (71) coordinate.

However, the conjugate momentum associated

VII. CONCLUDING REMARKS
to φ coordinate has the form ℘φ = mφ̇(r sin θ)2 .
The component of the time derivative of the lin-
The manipulation of Newton’s second law in
ear momentum in φ coordinate, after including
curvilinear coordinates allows to obtain La-
the constraint condition θ = α and dividing by
grange’s equations without using D’Alembert’s
the scale factor hφ = r sin θ, is
and Hamilton’s principles. This approach is re-
1 d℘φ visited in this work, taking as starting point the
= m(φ̈r + 2φ̇ṙ) sin α.
hφ dt
covariant expression for the linear momentum
Thus, Eq. (71) becomes
and then the time derivative of it in a contravari-
1 d℘φ
= 0. ant vector basis {~ei }. This procedure following
hφ dt
this way has not been published in the literature.
This result implies that the conjugate momen-
In this direction, the more important obtained
tum ℘φ is a conserved quantity, as might be ex-
remarks are:
pected.
On the other hand, introducing the constraint
(a) The gradient of the kinetic energy is asso-
condition θ = α into the gradient in coordinates
ciated with the fictitious forces (see Eq. (26)).
of Eq. (67), it is obtained
Thus, it can be related to the geometric proper-
~ q L = (−mg cos α + 2mrφ̇2 sin2 α)êr +
∇ ties of the system, via the Christoffel symbols.
(2mrφ̇2 sin α cos α + mg sin α)êθ + (b) The gradient of the Lagrangian is the effec-

(0)êφ . tive force in the system (see Eq. (40)).

(c) A vector equation of motion in terms of
It is the effective force acting on the particle.
gradients in coordinates and velocities, called
Based on this equation, the real and fictitious
equation of gradients, has been obtained (see
forces are, respectively
Eq. (55)).
~ q U = mg[−(cos α)êr + (sin α)êθ ],
−∇ (d) The treatment facilitates a physical inter-

16
pretation of the terms that conform Newton’s The authors acknowledgment to professors Daniel
~ and ∇L,
second law, as ∇T ~ which represent Esteban Jaramillo (Universidad de Antioquia)
the fictitious and effective forces in the system, and Ramiro Uribe (Universidad del Tolima) for
respectively. reading the manuscript and make important sug-
These outstanding facts can be included in some gestions. J. H. Muñoz is also grateful to Comité
advanced classical mechanics textbooks. Central de Investigaciones - Universidad del
ACKNOWLEDGMENTS Tolima for financial support.

[1] H. Goldstein and C. Poole and J. Safko, Classical Mechanics (Addison-Wesley, MA, 1980).
[2] S. T. Thornton and J. B. Marion, Classical Mechanics of Particles and Systems (Thomson, 1980).
[3] L. D. Landau and E. M. Lifshitz, Mechanics (Butterworth-Heinemann, 1976).
[4] J. B. Griffiths, The Theory of Classical Dynamics (Cambridge University Press, 2008).
[5] W. Greiner, Classical Mechanics (Springer, 2003).
[6] V. Arnold, Mathematical Methods Of Classical Mechanics (Springer, 1989).
[7] H. Iro, A Modern Approach to Classical Mechanics (World Scientific, 2002).
[8] L. N. Hand and J. D. Finch, Analytical Mechanics (Cambridge University Press, 1988).
[9] T. W. B. Kibble and F. H. Berkshire, Classical Mechanics (Imperial College Press, 2004).
[10] J. L. Synge and A. Schild, Tensor Calculus (Dover print, 1949).
[11] J. Casey, Am. J. Phys. 62, 836 (1994).
[12] A. Einstein, Annalen der Physik 6, 284-339 (1916).
[13] L. P. Lebedev and M. J. Cloud and V. A. Eremeyev, Tensor Analysis with Applications in Mechanics
(World Scientific, 2010).
[14] G. B. Arfken and H. J. Weber and F. E. Harris, Mathematical Methods for Physicists (Elsevier, 2010).
[15] I. A. Borisenko and I. E. Tarapov, Vector And Tensor Analysis with Applications (Dover publications,
1968).
[16] E. C. Young, Vector and Tensor Analysis (CRC Press, 1978).
[17] P. Grinfeld, Introduction to Tensor Analysis and the Calculus of Moving Surfaces (Springer, 2013).
[18] D. C. Cay, Schaum’s Outline of Theory and Problems of Tensor Calculus (McGraw-Hill, 1949).
[19] L. de la Peña and A. M. Cetto and A. Valdés, Rev. Mex. Fis. E 17 (1) 47, 2020.
[20] S. Kobayashi and K. Nomizu, Foundations of Differential Geometry (Volumen II, Wiley Interscience,

17
 1996).
[21] M. P. Carmo, Differential Geometry of Curves and Surfaces (Prentice Hall, 1976).
[22] M. Berger and B. Gostiaux, Differential Geometry: Manifolds, Curves, and Surfaces (Springer, 1987).
[23] W. L. Burke, Applied Differential Geometry (Cambridge University Press, 1985)
[24] H. A. Schwarz, Archives des Sciences Physiques et Naturelles 48, 38–44 (1873).

18
View publication stats