Classical Mechanics
Classical Mechanics
Classical Mechanics
Abstract This chapter is intended to remind the basic notions of the Lagrangian
and Hamiltonian formalisms as well as Noether’s theorem. We shall first start with
a discrete system with N degrees of freedom, state and prove Noether’s theorem.
Afterwards we shall generalize all the previously introduced notions to continuous
systems and prove the generic formulation of Noether’s Theorem. Finally we will
reproduce a few well known results in Quantum Field Theory.
As it is irrelevant for this first part (the discrete case), we shall drop the super-index
notation for coordinates or vectors that we have introduced in the previous chapter.
The action associated to a discrete system with N degrees of freedom (i =
1, . . . , N ) reads:
t2
S(qi ) = dt L(qi , q˙i , t), (2.1)
t1
where L = L(qi , q˙i , t) is the Lagrangian of the system and where {qi }i=1
N are the
generalized coordinates and q˙i ≡ dqi /dt the generalized velocities. In order to
obtain the Euler-Lagrange equations of motion we consider small variations of the
generalized coordinates qi keeping the extremes fixed:
∂L ∂L
L(qi + δqi , q˙i + δ q˙i , t) = L(qi , q˙i , t) + δqi + δ q˙i
∂qi ∂ q˙i
d
δqi (t) = qi (t) − qi (t) ⇒ (δqi ) = q˙i (t) − q˙i (t) = δ q˙i (t). (2.4)
dt
∂L ∂L
δL = δqi + δ q˙i
∂qi ∂ q˙i
∂L ∂L d
= δqi + (δqi )
∂qi ∂ q˙i dt
∂L d ∂L d ∂L
= − δqi + δqi . (2.5)
∂qi dt ∂ q˙i dt ∂ q˙i
In order to obtain the equations of motion we apply the Stationary Action Principle:
For the physical paths, the action must be a maximum, a minimum or an inflexion
point. This translates mathematically into:
t2 t2
δS = δ dt L = dt δL = 0 . (2.6)
t1 t1
Expanding δL we get:
t2 t2
∂L d ∂L d ∂L
δS = dt − δqi + dt δqi = 0. (2.7)
t1 ∂qi dt ∂ q˙i t1 dt ∂ q˙i
From (2.8) we can also deduce an important aspect of Lagrangians, that they are
not uniquely defined:
d F(qi , t)
L(qi , q˙i , t) and
L(qi , q˙i , t) = L(qi , q˙i , t) + (2.11)
dt
generate the same equations of motion. We have an alternative way to directly check
that adding a function of the form d F(qi , t)/dt to the Lagrangian, doesn’t alter the
equations of motion. Applying (2.10) to d F(qi , t)/dt we obtain:
∂ d F(qi , t) d ∂ d F(qi , t)
− = 0. (2.12)
∂qi dt dt ∂ q˙i dt
Next we will present one of the most important theorems of analytical mechanics,
a powerful tool that allows us to relate the symmetries of a system with conserved
quantities.
so that the variation of the Lagrangian can be written as the exact differential of some
function F:
d F(qi , q˙i , t) d F(qi , q˙i , t)
L(qi , q˙i , t) = L(qi , q˙i , t) + ⇒ δL = . (2.14)
dt dt
Note that here we allow F to also depend on q˙i (that was not the case for (2.11)). On
the other hand, we know that we can write δL as:
∂L d ∂L d ∂L d ∂L
δL = − δqi + δqi = δqi . (2.15)
∂qi dt ∂ q˙i dt ∂ q˙i dt ∂ q˙i
To get to the last equality we used the equations of motion. Let’s now write δqi as
an infinitesimal variation of the form
with || 1 a constant, and f a smooth, well behaved function. Obviously, in the
limit → 0 we obtain
and
Thus, necessarily F must be of the form F = F,
i , q̇i , t)
d ∂L d F(q
( f i ) = . (2.18)
dt ∂ q˙i dt
Integrating in t we obtain
∂L i , q̇i , t) + C,
f i = F(q (2.19)
∂ q˙i
∂L i , q̇i , t) .
C= f i − F(q (2.20)
∂ q˙i
2.3 Examples
Next, we are going to apply this simple formula to a few interesting cases and
reproduce some typical results such as energy and momentum conservation, angular
momentum conservation, etc.
Let’s consider an infinitesimal time shift: t → t + . The first order Taylor expansion
of qi and q̇i is given by:
If the Lagrangian does not exhibit an explicit time dependence (∂ L/∂t = 0) then
∂L ∂L dL = L.
δL = q˙i + q¨i = ⇒ F (2.22)
∂qi ∂ q˙i dt
2.3 Examples 33
∂L
q˙i − L = E , (2.23)
∂ q˙i
obeys pi = ∂T /∂ q˙i . Due to the fact that the potential is central and T = T (qi ) the
Lagrangian obeys
with rα the coordinates of the particle α and n an arbitrary spatial direction with
|n| = 1. We conclude that δL = 0. Under this spatial translation the coordinates of
the particle α transform the following way:
rα → r α = rα + n , (2.26)
that is
rα j → rα j = rα j + n j , (2.27)
∂L
C= nj = pα j n j = pα n = P n , (2.28)
α
∂ q̇α j α α
for an arbitrary n. Thus, the constant associated to this transformations is the total
momentum P of the system.
34 2 Lagrangians, Hamiltonians and Noether’s Theorem
2.3.3 Rotations
Again, let’s consider a Lagrangian with the same properties as in the previous exam-
ple. Under an infinitesimal rotation we have
∂L
C= jkm n m rαk = pα j jkm n m rαk = (pα × rα )n = −Ln . (2.30)
∂ q̇α j
Again, this holds for an arbitrary n, thus, the conserved quantity is the total angular
momentum L of the system.
For this last example we shall consider the same type of Lagrangian as in the previous
cases. A Galileo transformation reads
rα → r α = rα + vt, (2.31)
1 1
T = m α (ṙα + v)2 = T + m α ṙα v + m α v2
2 α
2
1 d 1 d
=T+ Mv2 + (m α rα v) = T + Mv2 + (MR v). (2.33)
2 dt 2 dt
Considering an infinitesimal transformation v = n with || 1 and ignoring terms
of O(2 ) we have
d
δL = δT = (MR n). (2.34)
dt
2.3 Examples 35
∂L
pi ≡ , (2.37)
∂ q˙i
The Hamiltonian equations of motion are obtained just as before by applying the
principle of the stationary action:
t2
δS = dt δL
t1
t2
= dt δ( pi q˙i − H )
t1
t2 ∂H ∂H
= dt δ pi q˙i + pi δ q˙i − δqi − δ pi
t1 ∂qi ∂ pi
t2 d ∂H ∂H
= dt δ pi q˙i + ( pi δqi ) − ṗi δqi − δqi − δ pi
t1 dt ∂qi ∂ pi
t2
t2 ∂H ∂H
= dt δ pi q˙i − + δqi − ṗi − + d( pi δqi )
t1 ∂ pi ∂qi t1
36 2 Lagrangians, Hamiltonians and Noether’s Theorem
t2 ∂H ∂H
= dt δ pi q˙i − + δqi − ṗi − = 0. (2.39)
t1 ∂ pi ∂qi
This must hold for arbitrary δ pi y δqi , therefore, the Hamiltonian equations of motion
are simply given by:
∂H ∂H
q˙i = , ṗi = − . (2.40)
∂ pi ∂qi
dH d
= ( pi q˙i − L)
dt dt
∂H ∂H ∂H
= ṗi + q˙i +
∂ pi ∂qi ∂t
∂H
= q˙i ṗi − ṗi q˙i +
∂t
∂L ∂L ∂L
= ṗi q˙i + pi q¨i − q˙i − q¨i − . (2.41)
∂qi ∂ q˙i ∂t
∂H ∂L
=− . (2.42)
∂t ∂t
Because now we also have spatial dependence besides time dependence, the follow-
ing replacement is also justified:
2.5 Continuous Systems 37
Notice that we have introduced the compact relativistic notation (from Chap. 1) and
we have supposed that the partial derivatives of the fields are a Lorentz (or Poincaré)
covariant quantity (of the form ∂μ φ(x)),1 with
∂
∂μ ≡ = ( ∂t , ∇) ≡ (∂t , ∂k ) . (2.45)
∂x μ
∂ μ ≡ g μν ∂ν = ( ∂t , −∇ ) ≡ ( ∂t , −∂k ) . (2.46)
Also, we are only interested in Lagrangians that are invariant under space-time
translations besides Lorentz transformations (Poincaré group), therefore they cannot
depend explicitly on x μ . The most generic Lagrangian that exhibits all the properties
we have just described can be written as:
Thus, the action can simply be written as an integral of the Lagrangian density
t2
S= dt L
t1
t2
= dt d 3 x L φi (x), ∂μ φi (x)
t1 V
x2
1 Thisis not the most general case, of course, but as we are interested in applying field theory to
Special Relativity we shall only restrict our study to this case.
38 2 Lagrangians, Hamiltonians and Noether’s Theorem
∂L ∂L
δL = δφi (x) + δ [∂μ φi (x)]
∂φi (x) ∂[∂μ φi (x)]
∂L ∂L
= δφi (x) + ∂μ [δφi (x)]
∂φi (x) ∂[∂μ φi (x)]
∂L ∂L ∂L
= − ∂μ δφi (x) + ∂μ δφi (x) , (2.51)
∂φi (x) ∂[∂μ φi (x)] ∂[∂μ φi (x)]
where summation over all repeated indices is understood. Similar to (2.4), the
variation and derivation operators commute. Applying the principle of the Sta-
tionary Action we obtain the Euler-Lagrange equations for continuous systems as
follows:
x2 x2
∂L ∂L ∂L
δS = − ∂μ δφi (x) + ∂μ δφi (x)
x1 ∂φi (x) ∂[∂μ φi (x)] x1 ∂[∂μ φi (x)]
x2 ∂L ∂L
= − ∂μ δφi (x) = 0, (2.52)
x1 ∂φi (x) ∂[∂μ φi (x)]
for arbitrary δφi (x), therefore, the equations we are looking for take the form
∂L ∂L
− ∂μ =0, (2.53)
∂φi (x) ∂[∂μ φi (x)]
x2 x2 ∂bμ
δ ∂μ bμ φi (x) = ∂μ δ [φi (x)] = 0. (2.55)
x1 x1 ∂φi (x)
2.5 Continuous Systems 39
μ
case, one can always add a functional of the form ∂μ b φi (x) without altering the
equations of motion. Therefore
where φ̇i (x) ≡ ∂t φi (x) and πi (x) is the canonical momentum associated to the field
φi (x):
∂L
πi (x) ≡ . (2.58)
∂ φ̇i (x)
for arbitrary δπi and δφi . Thus the equations of motion simply read
∂H ∂H ∂H
φ̇(x) = , π̇(x) = − + ∂k . (2.61)
∂π(x) ∂φ(x) ∂(∂k φ(x))
Until now we have only introduced a global variation of a field, which is defined as
the variation of the shape of the field without changing the space-time coordinates
x μ:
Besides this, we can define another type of variation which is closely related, a local
variation. It is defined as the difference between the fields evaluated in the same
space-time point but in two different coordinates systems:
x μ → x μ = x μ + x μ , (2.64)
We therefore, have found the following relation between δφ(x) and δ̄φ(x) for an
infinitesimal transformation of the type (2.64):
We can draw the following conclusion. If φi (x ) = φi (x) (which is in general the case
for a scalar field; it is also the case for spinor fields under space-time translations)
then
thus, in this case, an equivalent way of making a transformation of the type (2.64),
which acts on the coordinates, is by making an opposite transformation on the field:
Let us now deduce how the Lagrangian transforms under these type of variations.
In order to keep the notation short, we shall introduce the following short-hand
notations:
∂
where ∂μ ≡ . Keeping only terms up to O(x) we can calculate δL(x) under
∂x μ
(2.64):
δL(x) = L (x ) − L(x)
∂L(x) ∂L(x)
≈ L(x) + δφi (x) + δ [∂μ φi (x)] − L(x)
∂φi (x) ∂[∂μ φi (x)]
∂L(x) ∂L(x)
∂μ L(x) ≡ ∂μ φi (x) + ∂μ ∂ν φi (x). (2.71)
∂φi (x) ∂[∂ν φi (x)]
Let’s now consider the transformation of the action. A transformation that leaves
the equations of motion invariant is a symmetry of the system. Under such a symmetry
the action S will mostly transform as S → S with S given by
S = d 4 x L (x )
= d 4 x L(x) + d 4 x ∂μ bμ (x)
=S+ d 4 x ∂μ bμ (x), (2.74)
Inserting the explicit form of δL from (2.51) in the last expression, we obtain
∂L(x) ∂L(x)
− ∂μ δφi (x) +
∂φi (x) ∂[∂μ φi (x)]
∂L(x)
+ ∂μ δφi (x) + ∂μ L(x)x μ − ∂μ bμ (x) = 0. (2.79)
∂[∂μ φi (x)]
Using the Euler-Lagrange equations of motion we finally get to the conservation law
we were looking for
∂L(x)
∂μ j μ (x) = 0, j μ (x) = δφi (x) + L(x)x μ − bμ (x) , (2.80)
∂[∂μ φi (x)]
2.7 Noether’s Theorem (The General Formulation) 43
with j μ (x) the conserved Noether current. Note that our result is completely general,
in the sense that it holds for continuous space-time transformations of the type (2.64)
and also for transformations that only imply a field variation without modifying the
space-time configuration. In this last case we would simply set x μ = 0 in (2.80).
As it is usual, we can also define a conserved charge Q associated to the conserved
current j μ as:
dQ
Q= d 3 x j 0, = d 3 x ∂0 j 0 = − d3x ∇ j = 0 . (2.81)
dt
2.8 Examples
x μ → x μ = x μ − μ , (2.82)
with μ real constants. For scalar or spinor fields we have φi (x) = φi (x ) thus
δφi (x) = 0. Under this type of transformation our Lagrangians remain unchanged
so L (x ) = L(x), therefore, by taking a look at (2.77) we conclude that ∂μ bμ = 0.
We can thus, eliminate the bμ term from (2.80) and the conserved current is simply
given by:
∂L ∂L
jμ = ∂ν φi ν − Lμ = ∂ ν φi − Lg μν ν . (2.83)
∂(∂μ φi ) ∂(∂μ φi )
∂L
∂μ T μν = 0, T μν = ∂ ν φi − Lg μν , (2.84)
∂(∂μ φi )
with T μν the four-momentum tensor. The conserved Noether charges are then given
by
Pν ≡ d 3 x T 0ν
∂L ν
= d3x ∂ φi − g 0ν L
∂ φ̇i
44 2 Lagrangians, Hamiltonians and Noether’s Theorem
= d 3 x (πi ∂ ν φi − g 0ν L)
= d 3 x (πi φ̇i − g 00 L, −πi ∇φi )
= d 3 x (H, P )
= ( H, P ), (2.85)
where we have used (2.46). As we can see, the conserved charges are the Hamiltonian
and three-momentum operators.
with θ 1 (and where global means that the phase does not depend on the space-time
coordinates θ = θ(x)), then we find:
As this transformation doesn’t involve the space-time coordinates we can already set
x μ = 0 in (2.80). Therefore δL = δ̄L = ∂μ bμ . Again, if we only consider the free
Dirac or Klein-Gordon Lagrangians then δL = 0 = ∂μ bμ , so we can also eliminate
bμ from (2.80). The conserved current is then given by
∂L ∂L ∂L ∂L
jμ = δφ + δφ† = (−iφ)θ + iφ† θ, (2.88)
∂(∂μ φ) ∂(∂μ φ† ) ∂(∂μ φ) ∂(∂μ φ† )
for an arbitrary θ. Thus, redefining the current without the θ multiplying term we
find
∂L ∂L
∂μ j μ = 0, j μ (x) = −i φ+i φ† . (2.89)
∂(∂μ φ) ∂(∂μ φ† )
∂μ ψ̄γ μ ψ = 0. (2.90)
2.8 Examples 45
x μ → x μ = x μ + ω μν x ν , (2.91)
μ
with ω ν = −ω νμ real constants. Defining ωμν ≡ gμα ω αν , it is easy to show
that the field transformation reads
1 μν
φi (x ) = φi (x) + (i) ωμν φi (x), (2.92)
2
μν i 1 μν
with (i) = − σ μν = [γ μ , γ ν ] for spinorial4 fields and, (i) = 0 for scalar fields
2 4
(no summation over the “i” index must be understood in (2.92) nor in the following
expression). Using (2.66) we easily find:
1 αβ
δφi (x) = ωαβ φi (x) − ∂ α φi (x) ωαβ x β
2 (i)
1 αβ 1 α
On the other hand, our Lagrangians are all Lorentz invariant, thus δ̄L = 0 and so
again, we can eliminate bμ in (2.80) just as in the previous examples. We obtain that
the expression for our conserved current reads:
∂L αβ
1
μ
jμ = (i) + x α ∂ β − x β ∂ α φi ωαβ + Lω β x β
∂[∂μ φi ] 2
∂L αβ
1
= (i) + x α ∂ β − x β ∂ α φi ωαβ +
∂[∂μ φi ] 2
1
+ L(g μα x β − g μβ x α )ωαβ , (2.94)
2
for arbitrary ωαβ . Thus, we obtain
∂L αβ
∂μ J μ,αβ = 0, J μ,αβ = x α T μβ − x β T μα + φi , (2.95)
∂[∂μ φi ] (i)
which is the conservation law of the angular momentum pseudo5 tensor J μ,αβ (obvi-
ously, for the previous expression, summation over all repeated indices must be
understood).
Further Reading