Variational Principles and Lagrange’s Equations
Mar 16, 2016
Variational Principles
and Lagrange’s Equations
Definitions
• Lagrangian density:
• Lagrangian:
• Action:
• How to find the special value for action corresponding to observable ?
nin
nmi
xxxη ,)(L tzyxx
i
n ,,,?...3,2,1,0
dxdydzL L
t
dttrdL i
mi
,)(
dtt
dttrdLI i
mi
,)(
)(tr Joseph Louis
Lagrange/Giuseppe Luigi
Lagrangia (1736 – 1813)
zyxm rrrr ,,
Variational principle
• Maupertuis: Least Action Principle
• Hamilton: Hamilton’s Variational Principle
• Feynman: Quantum-Mechanical Path Integral Approach
Pierre-Louis Moreau de Maupertuis (1698 – 1759)
Sir William Rowan Hamilton
(1805 – 1865)
Richard Phillips Feynman
(1918 – 1988)
Functionals
• Functional: given any function f(x), produces a number S
• Action is a functional:
• Examples of finding special values of functionals using variational approach:
shortest distance between two points on a plane;the brachistochrone problem;minimum surface of revolution; etc.
)]([ xfS
2
1
,)()]([t
ti
mi
dttdt
trdLtrI
Shortest distance between two points on a plane
• An element of length on a plane is
• Total length of any curve going between points 1 and 2 is
• The condition that the curve is the shortest path is that the functional I takes its minimum value
22 dydxds
2
1
dsI
2
1
2
1 dxdxdy
The brachistochrone problem
• Find a curve joining two points, along which a particle falling from rest under the influence of gravity travels from the highest to the lowest point in the least time
• Brachistochrone solution: the value of the functional t [y(x)] takes its minimum value
dtdsv /
2
112 v
dst
2
1
2
2/1
dxgydxdy
22 dydxds
mgymv
2
2
gyv 2
Calculus of variations
• Consider a functional of the following type
• What function y(x) yields a stationary value (minimum, maximum, or saddle) of J ?
2
1
),...,,,,()]([x
x
dxxyyyyfxyJ dxdyy
x
y
0
),( 22 yx
),( 11 yx
Calculus of variations
• Assume that function y0(x) yields a stationary value and consider all possible functions in the form:
x
y
0
),( 22 yx
),( 11 yx
...)()()(),( 12
0 xxxyxy
0)()( 21 xx
Calculus of variations
• In this case our functional becomes a function of α:
• Stationary value condition:
)()(),( 0 xxyxy 0)()( 21 xx
)(),()],([2
1
JdxxfxyJx
x
0)()(
0)(),( 0
ddJ
ddJ
xyxy
Stationary value
2
1
),...,,,()( x
x
dxxyyyfdd
ddJ
2
1
),...,,,(x
x
dxd
xyyydf
2
1
...x
x
dxyyfy
yfy
yf
1
2
3
2
1
.1x
x
dxyyf
2
1
x
x
dxyf
)()(),( 0 xxyxy
Stationary value
2
1
),...,,,()( x
x
dxxyyyfdd
ddJ
2
1
),...,,,(x
x
dxd
xyyydf
2
1
...x
x
dxyyfy
yfy
yf
1
2
3
2
1
.2x
x
dxyyf
2
1
2x
x
dxxy
yf
u
dv
2
1
x
x
yyf
u
v
2
1
x
x
dxyf
dxdy
v
du
)()(),( 0 xxyxy
2
1
x
xyf
2
1
x
x
dxyf
dxd
0)(0)(
2
1
xx
Stationary value
2
1
),...,,,()( x
x
dxxyyyfdd
ddJ
2
1
),...,,,(x
x
dxd
xyyydf
2
1
...x
x
dxyyfy
yfy
yf
1
2
3
2
1
.3x
x
dxyyf
2
1
x
x
yyf
2
1
x
x
dxyf
dxdy
2
1
x
xyf
2
1
x
xyf
dxd
2
1
2
2x
x
dxyf
dxd
Stationary value
2
1
),...,,,()( x
x
dxxyyyfdd
ddJ
2
1
),...,,,(x
x
dxd
xyyydf
2
1
...x
x
dxyyfy
yfy
yf
1
2
3
2
1
x
xyf
2
1
2
2x
x
dxyf
dxd
2
1
x
x
dxyf
2
1
x
x
dxyf
dxd
2
1
...2
2x
x
dxyf
dxd
yf
dxd
yf
2
1
x
xyf
...
...
Stationary value
... ...2
1
2
1
2
2
x
x
x
x yfdx
yf
dxd
yf
dxd
yf
ddJ
),( xyff
2
1
x
x
dxyf
ddJ
0)(
0
ddJ 0
2
1 0 )(),(
x
x xyxy
dxyf
arbitrary0
yf
Trivial …
Stationary value
... ...2
1
2
1
2
2
x
x
x
x yfdx
yf
dxd
yf
dxd
yf
ddJ
),,( xyyff
2
1
x
x
dxyf
dxd
yf
ddJ
0)(
0
ddJ
0 2
1 0
x
x
dxyf
dxd
yf
arbitrary0
yf
dxd
yf
Nontrivial !!!
Shortest distance between two points on a plane
2
1
2
1 dxdxdyI 21 yf
0
yf
dxd
yf
0
10
2
y
ydxd
c
y
y
21
21 c
cy
bxc
cy
21
Straight line!
The brachistochrone problem
2
1
2
12 2/1
dxgydxdy
t gyy
f2
1 2
0
yf
dxd
yf
0
1222
123
2
ygy
ydxd
gy
y
Scary!
Recipe
• 1. Bring together structure and fields
• 2. Relate this togetherness to the entire system
• 3. Make them fit best when the fields have observable dependencies:
Structure
Fields Fields
Structure
Physical Laws
Best F
it
mη mη
Back to trajectories and Lagrangians
• How to find the special values for action corresponding to observable trajectories ?
• We look for a stationary action using variational principle
2
1
,)()]([t
ti
mi
dttdt
trdLtrI
)()(),( 0 ttrtr mmm
0)()( 21 tt mm 0)(
0
ddI
2
1
,),()],([)(t
ti
mi
m dttdttrdLtrII
Recipe
• 1. Bring together structure and fields
• 2. Relate this togetherness to the entire system
• 3. Make them fit best when the fields have observable dependencies:
Structure
Fields Fields
Structure
Physical Laws
Best F
it
mη mη
Back to trajectories and Lagrangians
• For open systems, we cannot apply variational principle in a consistent way, since integration in not well defined for them
• We look for a stationary action using variational principle for closed systems:
2
1
,),()],([)(t
ti
mi
m dttdttrdLtrII
0)(
0
ddI
dxdydzL L dxdydzdtI L
Stationary value
... ...2
1
2
1
2
2
x
x
x
x yfdx
yf
dxd
yf
dxd
yf
ddJ
),,( xyyff
2
1
x
x
dxyf
dxd
yf
ddJ
0)(
0
ddJ
0 2
1 0
x
x
dxyf
dxd
yf
0
yf
dxd
yf
Nontrivial !!!
Simplest non-trivial case
• Let’s start with the simplest non-trivial result of the variational calculus and see if it can yield observable trajectories
2
1
,),()(t
ti
mi
dttdttrdLI
0)(
0
ddI
2
1
,),(),,(t
t
mm dtt
dttdrtrL
zyxmi
,,1,0
2
1
,,t
tmm dttrrL
Stationary value
... ...2
1
2
1
2
2
x
x
x
x yfdx
yf
dxd
yf
dxd
yf
ddJ
),,( xyyff
2
1
x
x
dxyf
dxd
yf
ddJ
0)(
0
ddJ
0 2
1 0
x
x
dxyf
dxd
yf
0
yf
dxd
yf
Nontrivial !!!
Euler- Lagrange equations
• These equations are called the Euler- Lagrange equations
0)(
0
ddI
2
1
,,t
tmm dttrrLI
0
yf
dxd
yf
0
mm rL
dtd
rL
Joseph Louis Lagrange
(1736 – 1813)
Leonhard Euler (1707 – 1783)
Recipe
• 1. Bring together structure and fields
• 2. Relate this togetherness to the entire system
• 3. Make them fit best when the fields have observable dependencies:
Structure
Fields Fields
Structure
Physical Laws
Best F
it
mη mη
How to construct Lagrangians?
• Let us recall some kindergarten stuff
• On our – classical-mechanical – level, we know several types of fundamental interactions:
• Gravitational• Electromagnetic• That’s it
0
mm rL
dtd
rL
Gravitation
• For a particle in a gravitational field, the trajectory is described via 2nd Newton’s Law:
• This system can be approximated as closed
• The structure (symmetry) of the system is described by the gravitational potential
gUdtvmd
)(
gm
),,,( tzyxgg
Sir Isaac Newton(1643 – 1727)
Electromagnetic field
• For a charged particle in an electromagnetic field, the trajectory is described via 2nd Newton’s Law:
• This system can be approximated as closed
• The structure (symmetry) of the system is described by the scalar and vector potentials
),,,(),,,(
tzyxtzyxAA
)( AvqqAqvmdtd
Really???
Electromagnetic field
)( AvqqAqvmdtd
)()( AvqqdtAdq
dtvmd
dtAd
),,,( tzyxAA
zzAy
yAx
xA
tA
AvtA
)(
AvqtAqAvqq
dtvmd
)()()(
AvAvqtAqq
dtvmd
)()()(
Electromagnetic field
AvAvqtAqq
dtvmd
)()()(
FGFGGFGF
)()()(
GF
vAvAAvAv
)()()(
Av
AvqtAqq
dtvmd
)(
Electromagnetic field
• Lorentz force!
AvqtAq
dtvmd
)(
tAE
AB
BvEqdtvmd
)(
Hendrik Lorentz(1853-1928)
Kindergarten
• Thereby:
• In component form
0)( AqvmdtdAvqq
0)()(
dtrmd
rm j
j
g
0)(
dtvmdm g
0)())((
dtqArmd
rArqq jj
j
How to construct Lagrangians?
• Kindergarten stuff:
• The “kindergarten equations” look very similar to the Euler-Lagrange equations! We may be on the right track!
0
jj rL
dtd
rL
0)()(
dtrmd
rm j
j
g
0)())((
dtqArmd
rArqq jj
j
Gravitation
0
jj rL
dtd
rL
0)()(
dtrmd
rm j
j
g
j
g
j rm
rL
)(
j
j
rL
dtd
dtrmd
)(
),,,( trrrTmL zyxg C
rLrmj
j
),,,( trrrS zyx),,,( trrr zyxgg
2)( 222
zyx rrrmL
Gravitation
gzyx mtrrrTL ),,,(
),,,(2
)( 222
trrrSrrrm
L zyxzyx
gzyx mrrrm
L
2
)( 222
Electromagnetism
0
jj rL
dtd
rL
0)())((
dtqArmd
rArqq jj
j
)(2
)( 222
Arqqrrrm
L zyx
Bottom line
• We successfully demonstrated applicability of our recipe
• This approach works not just in classical mechanics only, but in all other fields of physics
Structure
Physical LawsBes
t Fit
Some philosophy
• de Maupertuis on the principle of least action (“Essai de cosmologie”, 1750): “In all the changes that take place in the universe, the sum of the products of each body multiplied by the distance it moves and by the speed with which it moves is the least that is possible.”
• How does an object know in advance what trajectory corresponds to a stationary action???
• Answer: quantum-mechanical pathintegral approach
Pierre-Louis Moreau de Maupertuis (1698 – 1759)
Some philosophy
• Feynman: “Is it true that the particle doesn't just "take the right path" but that it looks at all the other possible trajectories? ... The miracle of it all is, of course, that it does just that. ... It isn't that a particle takes the path of least action but that it smells all the paths in the neighborhood and chooses the one that has the least action ...”
Richard Phillips Feynman
(1918 – 1988)
Some philosophy
• Dyson: “In 1949, Dick Feynman told me about his "sum over histories" version of quantum mechanics. "The electron does anything it likes," he said. "It just goes in any direction at any speed, forward or backward in time, however it likes, and then you add up the amplitudes and it gives you the wave-function." I said to him, "You're crazy." But he wasn't.”
Freeman John Dyson (born 1923)
Some philosophy
• Philosophical meaning of the Lagrangian formalism: structure of a system determines its observable behavior
• So, that's it?
• Why do we need all this?
• In addition to the deep philosophical meaning, Lagrangian formalism offers great many advantages compared to the Newtonian approach
Lagrangian approach: extra goodies
• It is scalar (Newtonian – vectorial)
• Allows introduction of configuration space and efficient description of systems with constrains
• Becomes relatively simpler as the mechanical system becomes more complex
• Applicable outside Newtonian mechanics
• Relates conservation laws with symmetries
• Scale invariance applications
• Gauge invariance applications
Simple example
• Projectile motion gzyx mrrrm
L
2
)( 222
zzyx mgrrrrm
L
2
)( 222
0
yy rL
dtd
rL
0
xx rL
dtd
rL
0
zz rL
dtd
rL
0xrmdtd
0yrmdtd
constrm x
constrm y
zg gr
mgrmdtd
z constgtrz
Another example
• Another Lagrangian
• What is going on?!
xy
zx mgrrm
rrmL 2
2
0
yy rL
dtd
rL
0
xx rL
dtd
rL
0
zz rL
dtd
rL
0 zrmdtd
0yrmdtd
constrm x
constrm y
constgtrz mg
0 0 xrmdtd
Gauge invariance
• For the Lagrangians of the type
• And functions of the type
• Let’s introduce a transformation (gauge transformation):
trrL ii ,,
dt
trdFtrrLtrrL iiiii
,,,,,'
trF i ,
Gauge invariance
dtdFLL ' j
j j
rrF
tF
dtdF
trFF i ,
dtdF
rrL
rL
iii
'
jj jii
rrF
tF
rrL
jj jiii
rrrF
trF
rL
22
Gauge invariance
dtdFLL ' j
j j
rrF
tF
dtdF
trFF i ,
dtdF
rrL
rL
iii '
jj jii
rrF
tF
rrL
ii rF
rL
iii rF
dtd
rL
dtd
rL
dtd
'
ii rF
trL
dtd
jj ij
rrF
r
Gauge invariance
jj ijiii
rrF
rrF
trL
dtd
rL
dtd
'
jj ijii
rrrF
rtF
rL
dtd
22
jj jiiii
rrrF
trF
rL
rL
22'
ii rL
rL
dtd
'' ii r
LrL
dtd
0
Back to the question: How to construct Lagrangians?
• Ambiguity: different Lagrangians result in the same equations of motion
• How to select a Lagrangian appropriately?
• It is a matter of taste and art
• It is a question of symmetries of the physical system one wishes to describe
• Conventionally, and for expediency, for most applications in classical mechanics:
VTL
Cylindrically symmetric potential
• Motion in a potential that depends only on the distance to the z axis
• It is convenient to work in cylindrical coordinates
• Then
22
222
2)(
yxzyx rrVrrrm
L
zrrrrr zyx ;sin ;cos
zr
rrr
rrr
z
y
x
cossin
sincos
Cylindrically symmetric potential
• How to rewrite the equations of motion in cylindrical coordinates?
22
222
2)(
yxzyx rrVrrrm
L
22222
sincos2
rrVzm
2
)cossin( 2 rrm
2)sincos( 2 rrm
)(2
)( 2222
rVzrrm
0
jj rL
dtd
rL
Generalized coordinates
• Instead of re-deriving the Euler-Lagrange equations explicitly for each problem (e.g. cylindrical coordinates), we introduce a concept of generalized coordinates
• Let us consider a set of coordinates
• Assume that the Euler-Lagrange equations hold for these variables
• Consider a new set of (generalized) coordinates
),...,,(: 21 Ni rrrr
0
ii rL
dtd
rL
),,...,,( 21 trrrqq Njj
Generalized coordinates
• We can, in theory, invert these equations:
• Let us do some calculations:
N
i m
i
im qr
rL
qL
1
),,...,,( 21 trrrqq Nmm
),,...,,( 21 tqqqrr Mii
M
mm
m
iii q
qr
trr
1
N
i m
i
im qr
rL
dtd
qL
dtd
1
N
i m
i
i qr
rL
dtd
1
m
iN
i i
N
i m
i
i qr
dtd
rL
qr
rL
dtd
11
m
i
m
i
qr
qr
Generalized coordinates
• The Euler-Lagrange equations are the same in generalized coordinates!!!
M
kk
m
i
k
i
m
i
m
i qqr
qr
qr
tqr
dtd
1
m
iN
i i
N
i m
i
im qr
dtd
rL
qr
rL
dtd
qL
dtd
11
M
kk
k
ii
m
qqr
tr
q 1
M
mm
m
iii q
qr
trr
1
m
i
qr
mqL
dtd
m
iN
i i qr
rL
1 mq
L
N
i m
i
i qr
rL
1
Generalized coordinates
• If the Euler-Lagrange equations are true for one set of coordinates, then they are also true for the other set
ii rL
rL
dtd
),,...,,( 21 trrrqq Nmm
mm qL
qL
dtd
Cylindrically symmetric potential
• Radial force causes a change in radial momentum and a centripetal acceleration
)(2
)( 2222
rVzrrmL
0
jj qL
dtd
qL
),,(: zrqi 0
rL
dtd
rL
rrV
)( 0)(
dtrmd
dtrmdmr
rrV )()( 2
2mr
Cylindrically symmetric potential
• Angular momentum relative to the z axis is a constant
)(2
)( 2222
rVzrrmL
0
jj qL
dtd
qL
),,(: zrqi 0
L
dtdL
0 0)( 2
dtmrd
constmrrmr )(2
0)( 2
dtmrd
Cylindrically symmetric potential
• Axial component of velocity does not change
)(2
)( 2222
rVzrrmL
0
jj qL
dtd
qL
),,(: zrqi 0
zL
dtd
zL
0 0)(
dtzmd
constzm
0)(
dtzmd
Symmetries and conservation laws
• The most beautiful and useful illustration of the “structure vs observed behavior” philosophy is the link between symmetries and conservation laws
• Conjugate momentum for coordinate :
• If Lagrangian does not depend on a certain coordinate, this coordinate is called cyclic (ignorable)
• For cyclic coordinates, conjugate momenta are conserved
)( iqfL
mqL
mq
0
ii qL
dtd
qL
0
iqL
dtd
Symmetries and conservation laws
• For cyclic coordinates, conjugate momenta are conserved
p =
cons
t p ≠ const
Cylindrically symmetric potential
• Cyclic coordinates:
• Rotational symmetry Translational symmetry
• Conjugate momenta:
)(2
)( 2222
rVzrrmL
0
L
dtdL
constmr 2
0
zL
dtd
zL
constzm
z
Electromagnetism
• Conjugate momenta:
)(2
)( 222
Arqqrrrm
L zyx
jrL
jrm jqA jrm
Noether’s theorem
• Relationship between Lagrangian symmetries and conserved quantities was formalized only in 1915 by Emmy Noether:
• “For each symmetry of the Lagrangian, there is a conserved quantity”
• Let the Lagrangian be invariant under the change of coordinates:
• α is a small parameter. This invariance has to hold to the first order in α
),,...,,(~21 tqqqqq Niii
Emmy Noether/Amalie Nöther(1882 – 1935)
Noether’s theorem
• Invariance of the Lagrangian:
• Using the Euler-Lagrange equations
0ddL
N
i
i
i
i
i
qqLq
qL
ddL
1
~~
~~
N
ii
ii
i qL
qL
1~~
N
ii
ii
i qL
qL
dtd
1~~
N
ii
iqL
dtd
1~ 0
constqLN
ii
i
1
),,...,,(~21 tqqqqq Niii
Example
• Motion in an x-y plane of a mass on a spring (zero equilibrium length):
• The Lagrangian is invariant (to the first order in α) under the following change of coordinates:
• Then, from Noether’s theorem it follows that
2)(
2)( 2222
yxyx rrkrrmL
xyyyxx rrrrrr ~ ;~
constrL
rL
yy
xx
yxrrm xyrrm const
Example
• In polar coordinates:
• The conserved quantity:
• Angular momentum in the x-y plane is conserved
constrrmrrm xyyx
sincosrrrr
y
x
cossin
sincos
rrr
rrr
y
x
xyyx rrmrrm sin)sincos( rrrm
cos)cossin( rrrm 2mr const
Example
• For the same problem, we can start with a Lagrangian expressed in polar coordinates:
• The Lagrangian is invariant (to any order in α) under the following change of coordinates:
• The conserved quantity from Noether’s theorem:
constmr 12
2)(
2)( 2222
yxyx rrkrrmL
22)( 2222 krrrm
1~
constL
Back to trajectories and Lagrangians
• How to find the special values for action corresponding to observable trajectories ?
• We look for a stationary action using variational principle
2
1
,)()]([t
ti
mi
dttdt
trdLtrI
)()(),( 0 ttrtr mmm
0)()( 21 tt mm 0)(
0
ddI
2
1
,),()],([)(t
ti
mi
m dttdttrdLtrII
),,...,,(~21 tqqqqq Nmmm
Stationary value
2
1
),...,,,()( x
x
dxxyyyfdd
ddJ
2
1
),...,,,(x
x
dxd
xyyydf
2
1
...x
x
dxyyfy
yfy
yf
1
2
3
2
1
.2x
x
dxyyf
2
1
2x
x
dxxy
yf
u
dv
2
1
x
x
yyf
u
v
2
1
x
x
dxyf
dxdy
v
du
)()(),( 0 xxyxy
2
1
x
xyf
2
1
x
x
dxyf
dxd
0)(0)(
2
1
xx
constqLN
ii
i
1
More on symmetries
• Full time derivative of a Lagrangian:
• From the Euler-Lagrange equations:
• If
dtdL
M
m
M
mm
mm
m
qqLq
qL
tL
1 1
M
m
M
mm
mm
m
qqLq
qL
dtd
tL
1 1
M
mm
m
qqL
dtd
tL
1
LqqL
dtd
tL M
mm
m1
dt
dH
0tL constLq
qLH
M
mm
m
1
What is H?
• Let us expand the Lagrangian in powers of :
• From calculus, for a homogeneous function f of degree n (Euler’s theorem) :
......),,...,,(
),,...,,(),,...,,(
3210,
212
211210
LLLLqqtqqql
qtqqqltqqqLL
jji
iMij
iiMiM
fnxfx
i ii
iq
...210
ii
iii
iii
iii
i
qqLq
qLq
qLq
qL
...320 321 LLL
What is H?
• If the Lagrangian has a form:
• Then
• For electromagnetism:
LqqLH
M
mm
m
1
...32 321 LLL
...)( 3210 LLLL ...2 320 LLL
210 LLLL
02 LLH
)(2/2 ArqqrmL
2L 0L 1L
02 LLH qrm 2/2 EVT
Conservation of energy
• In the field formalism, the conservation of H is a part of Noether’s theorem
210 LLLL
EH
constEtL
0
The brachistochrone problem
• Similarly to the “H-trick”:
2
112 dxft
gyy
f2
1 2 0
xf
0
1222
123
2
ygy
ydxd
gy
y
Scary!
constfyfy
H
gyy
ygy
yy2
1
12
2
2
constygy
212
1
21/ yyC
!!!
The brachistochrone problem
• Change of variables:
• Parametric solution (cycloid)
21/ yyC
2sinCy
dxdy
yCy 1 dy
yCydx
)sin(sin
sin 22
2
CdCC
Cdx
dC sin2 2
BdCx sin2 2
2sin
)2/)2(sin(
Cy
CBx
)2/)2(sin( CB
Scale invariance
• For Lagrangians of the following form:
• And homogeneous L0 of degree k
• Introducing scale and time transformations
• Then
jji
iijM qqlqqqLLLL ,
221020 ),...,,( constl ij 2
ttqq ii
''
),...,,(),...,,(' 2102100 Mk
M qqqLqqqLL
jji
iij qql
,2
2
ii qq
' j
jiiij qqlL '''
,22
Scale invariance
• Therefore, after transformations
• If
• Then
• The Euler-Lagrange equations after transformations
• The same!
2
2
0' LLL k
k
2
LL k'
0''
jj qL
dtd
qL
0)()(
j
k
j
k
qL
dtd
qL
0
jj qL
dtd
qL
Scale invariance
• So, the Euler-Lagrange equations after transformations are the same if
• Free fall
• Let us recall
k
2
2/1 k2/1
2/1 ''k
i
ik
tt
mgzzmL 2
21k
2/1'' k
zz
tt
2/1'
zz
2/12 zgzt
Scale invariance
• So, the Euler-Lagrange equations after transformations are the same if
• Mass on a spring
• Let us recall
k
2
2/1 k2/1
2/1 ''k
i
ik
tt
22
22 KzzmL
2k2/1'' k
zz
tt
0'
zz
02 zKmT
Scale invariance
• So, the Euler-Lagrange equations after transformations are the same if
• Kepler’s problem
• Let us recall 3rd Kepler’s law
k
2
2/1 k2/1
2/1 ''k
i
ik
tt
2
2
2 r
MmGrmL 1k
2/1'' k
zz
tt
2/3'
zz
2/3RT
Johannes Kepler(1571-1630)
How about open systems?
• For some systems we can neglect their interaction with the outside world and formulate their behavior in terms of Lagrangian formalism
• For some systems we can not do it
• Approach: to describe the system without “leaks” and “feeds” and then add them to the description of the system
How about open systems?
• For open systems, we first describe the system without “leaks” and “feeds”
• After that we add “leaks” and “feeds” to the description of the system
• Q: Non-conservative generalized forces
jj qL
qL
dtd
jQ
Generalized forces
• Forces
• 1: Conservative (Potential)
• 2: Non-conservative
j
jj
Tq
Tdtd
UU
jjjjj
Qqqdt
dqT
qT
dtd
UU
...),,...,,( 21 tqqqV NU
12
1
UTL
Generalized forces
• In principle, there is no need to introduce generalized forces for a closed system fully described by a Lagrangian
• Feynman: “…The principle of least actiononly works for conservative systems — where all forces can be gotten from a potential function. … On a microscopic level — on the deepest level of physics — there are no non-conservative forces. Non-conservative forces, like friction, appear only because we neglect microscopic complications — there are just too many particles to analyze.”
• So, introduction of non-conservative forces is a result of the open-system approach
Richard Phillips Feynman
(1918 – 1988)
Degrees of freedom
• The number of degrees of freedom is the number of independent coordinates that must be specified in order to define uniquely the state of the system
• For a system of N free particle there are 3N degrees of freedom (3N coordinates)
N) ..., ,2 ,1(
ˆˆˆ
Ni
rkrjrir ziyixii
Constraints
• We can impose k constraints on the system
• The number of degrees of freedom is reduced to 3N – k = s
• It is convenient to think of the remaining s independent coordinates as the coordinates of a single point in an s-dimensional space: configuration space N
), ..., , ,(...
), ..., , ,(
321
32111
tqqqrr
tqqqrr
kNNN
kN
k
Types of constraints
• Holonomic (integrable) constraints can be expressed in the form:
• Nonholonomic constraints cannot be expressed in this form
• Rheonomous constraints – contain time dependence explicitly
• Scleronomous constraints – do not contain time dependence explicitly
kj
tqqqf nj
,...,2,1
0), ..., , ,( 21
Analysis of systems with holonomic constraints
• Elimination of variables using constraints equations
• Use of independent generalized coordinates
• Lagrange’s multiplier method
Double 2D pendulum
• An example of a holonomic scleronomous constraint
• The trajectories of the system are very complex
• Lagrangian approach produces equations of motion
• We need 2 independent generalized coordinates (N = 2, k = 2 + 2, s = 3 N – k = 2)
0)( 21
21 lr 0)( 2
22
21 lrr
1 2
Double 2D pendulum
• Relative to the pivot, the Cartesian coordinates
• Taking the time derivative, and then squaring
• Lagrangian in Cartesian coordinates:
11,1 sinlr x
11,1 coslr z 2211,2 sinsin llr x
2211,2 coscos llr z
21
21
21 lr
)sinsincos(cos2 212121212
22
22
12
12
2 llllr
)(2 ,22,11
222
211
zz rmrmgrmrmL
Double 2D pendulum
• Lagrangian in new coordinates:
• The equations of motion:
)coscos(cos
2)cos(2
222112111
2121212
22
22
12
122
12
11
llgmglm
llllmlmL
222212
1212
212
12122
22
22
1121212
2212
212
22122
12
121
sin)sin(
)cos(0
sin)()sin(
)cos()(0
glmllm
llmlm
glmmllm
llmlmm
Double 2D pendulum
• Special case
• The equations of motion:
• More fun at:
http://www.mathstat.dal.ca/~selinger/lagrange/doublependulum.html
21 mm 0, 21 lll 21
221
121
0
220
gl
gl
Lagrange’s multiplier method
• Used when constraint reactions are the object of interest
• Instead of considering 3N - k variables and equations, this method deals with 3N + k variables
• As a results, we obtain 3N trajectories and k constraint reactions
• Lagrange’s multiplier method can be applied to some nonholonomic constraints
Lagrange’s multiplier method
• Let us explicitly incorporate constraints into the structure of our system
• For observable trajectories
• So
kjtqqqf nj ,...,2,1 ;0), ..., , ,( 21
k
jnjj tqqqftLL
121 ), ..., , ,()('
0), ..., , ,( 21 tqqqf nj
k
jjj fLL
1
' L
ii qL
qL
dtd
''
k
j i
jj
ii qf
qL
qL
dtd
1
0
0
Lagrange’s multiplier method
• - constraint reactions
• Now we have 3N + k equations for and
kjtqqqf nj ,...,2,1 ;0), ..., , ,( 21
k
j i
jj
ii qf
qL
qL
dtd
1
iQ
iQ
Niqf
qL
qL
dtd k
j i
jj
ii
3,...,2,1 ;1
iq j
Application to a nonholonomic case
• A particle on a smooth hemisphere
• One nonholonomic constraint:
• While the particle remains on the sphere, the constraint is holonomic
• And the reaction from the surface is not zero
02222 arrr zyx
02222 arrr zyx
Application to a nonholonomic case
• Constraint equation in cylindrical coordinates:
• New Lagrangian in cylindrical coordinates:
• Equations of motion
0 ar
)(cos2
)(' 1
2222
armgrzrrmL
rf
rL
dtd
rL
1
1
01cos 12 mgmrrm
Application to a nonholonomic case
• Constraint equation in cylindrical coordinates:
• New Lagrangian in cylindrical coordinates:
• Equations of motion
0 ar
)(cos2
)(' 1
2222
armgrzrrmL
0
L
dtdL
0sin2 mgrmr
Application to a nonholonomic case
• Constraint equation in cylindrical coordinates:
• New Lagrangian in cylindrical coordinates:
• Equations of motion
• Trivial
0 ar
)(cos2
)(' 1
2222
armgrzrrmL
0
zL
dtd
zL
0zm
Application to a nonholonomic case
• Constraint reaction:
ar
0cos 12 mgmrrm
0sin2 mgrmr
amg /cos 12
sinag
sin22ag
cos22
ag
dtd
dtd
0
0
)cos1(22
ag
)2cos3(1 mg
)2cos3(111
1 mgrf
Application to a nonholonomic case
• Constraint reaction:
• Reaction disappears when
• The particle becomes airborne
)2cos3(1 mg
2cos3
32cos 1
ar