Maxwell – Boltzmann distribution for molecular velocities

James Clerk Maxwell Ludwig Boltzmann

(1831-1879) (1844 – 1906)

The velocity distribution is derived for an ideal gas based on very simple assumptions:
• The distribution of velocity is isotropic in all directions (spatial homogeneity)
• The velocity distributions in the x, y, and z directions are independent and identical

• External factors enter into the distribution function for the velocity
• We later derive the distribution function using statistical mechanics
1
 Kinetic Theory of Gases: A Mechanistic Viewpoint of Gas Properties
Ideal Gas Model:
• Atoms are small hard spheres which only interact upon collision
• The volume of the atoms is much smaller than the volume of the gas

v : speed of the gas molecules

vx, vy, vz: components of velocity
of the gas molecules
v 2  v x2  v 2y  v z2

• We assume there is no bias in the

distribution of the velocity
components vx, vy, vz
• The choice of coordinate system is
arbitrary

The average of the square of the velocity components is identical in all directions
v x2  v 2y  v z2  v 2 3 Why not state the average of the velocity2
components?
Preamble: Relating velocity distribution to macroscopic pressure of an ideal gas
• The collision of molecule i with the
Lx wall results in a momentum change and
a force exerted on the wall.
≈
Change in momentum
vx,i px ,i
f x ,i  
2 mvx ,i
i
t 2 L vx ,i
Wx2 Wx1
Time between
consecutive collisions
with the wall
• Contribution to pressure from molecule i
with velocity vx,i
2 2
f i mv x ,i mv x ,i
Px ,i   
≈

A AL V
x
N N mv 2 Nm v x2 Nm v 2
Px   Px,i  
Total gas pressure from all x ,i
 
molecules in x-direction V V 3V
i 1 i 1 3
 Mechanical expression for pressure of an ideal gas - 2
• From microscopic mechanical analysis
Nm v 2 2 N EK
Pmicro  
3V 3V

• From experimental macroscopic study of ideal gases

nRT NkT
Pexpt  
V V
Pexpt  Pmicro
m 2 2 Result: Temperature is related to the
kT  v  EK average kinetic energy of molecules
3 3 in an ideal gas!

• A macroscopic thermodynamic variable T is related to the average microscopic

kinetic energy of the molecules;
• The average squared speed of molecules in the system is proportional to the
temperature;
• Can we find the probability distribution of speed and kinetic energy? 4
Assumptions used to derive the Maxwell – Boltzmann distribution for velocities

1) The probability distributions in x, y, z directions have identical mathematical form

and are independent of each other.
P(vx )  P(v y )  P(vz ) P (v x , v y , v z )  P (v x ) P (v y ) P (v z )

2) The probability distribution of the molecule only depends on the speed and not
individual Cartesian components of the velocity, which are arbitrary (why?)

P (v x , v y , v z )  P (v ) v 2  v x2  v 2y  v z2

Combining 1) and 2)  P (v )  P (v x ) P (v y ) P (v z )

Taking the derivative of both sides with respect to vx (using the chain rule),
dP(v) dP(v x ) 1 dP(v) dv 1 dP(v x )
 P (v y ) P (v z )  
dvx dvx P (v) dv dvx P (v x ) dvx

Cvx2
1 dP(v)

1 dP(v x )
 const. P( v x )  Ae
vP(v) dv v x P (v x ) dvx
A Gaussian distribution! 5
 Maxwell – Boltzmann distribution for molecular velocities
The probability distribution has two unknown factors, A and C:
Cvx2
P( v x )  Ae
• The A factor is determined by the normalization condition of the probability:

 
1/ 2 1/ 2
 C  Cvx2

I 0   e x
2
dx   
 P( v x )dvx  1  P( v x )    e
 

 
Note that the range of the velocity components is from -∞ to + ∞

• The C factor is determined from the kinetic theory of gases result:


1 mv 2  1 kT
x
1
2 m  vx2 P(vx )dvx  12 kT
2 2 

1 3  5    (n  1)   
1/ 2

In   x en  ax 2
dx    n even
 2 n / 2  

1/ 2
The probability distribution for  m  mvx2
Cartesian components P (v x )    e
2 kT

 2kT  6
 Maxwell – Boltzmann distribution for molecular velocity components
1/ 2 • Distribution is symmetric w.r.t.
 m  mv x2
P1 ( v x )    e
2 kT positive and negative directions;
 2kT 
• Average velocity component is
0;
vx  0

• Most probable value for the

component of velocity is 0,

• The standard deviation is:

1/ 2
 kT 
vx  vx      
2
m
• Distribution is broad!

• Distribution becomes broader

at higher temperature
https://phet.colorado.edu/en/simulation/gas-properties
7
 Maxwell – Boltzmann distribution for molecular speed
Combining the three components of the velocity distribution
3/2
 m   m ( vx2  v 2y  vz2 ) 2 kT
P1 (v)dvx dv y dvz  P1 (vx ) P1 (v y ) P1 (vz )dvx dv y dvz    e dvx dv y dvz
 2 kT 
Change the velocities from Cartesian coordinates to spherical polar coordinates:
v 2 sin dvdd  dv x dv y dv z  dv
The probability distribution does not depend on angles θ and 
3/ 2
 m   2
P1 (v)dv    e  mv 2 2 kT
v dv  sin  d 
2
d
 2 kT  0 0

Giving a distribution function for speed:

3/ 2
 m  2  mv 2 kT
2
P1( v )dv  4   ve dv
 2kT 

The speed distribution is not Gaussian!

8
 The Maxwell – Boltzmann distribution for speed of one molecule
3/ 2
 m   mv 2 2 kT
The probability distribution for speed: P1 ( v )  4  
2
v e
 2 kT 
Most probable speed:
dP1 (v) 2 2 kT
 mvP m 3  mvP2 2kT 2kT
 0  2v P e  vP e  0  vP 
dv kT m
Average speed:
3/ 2
 m    3  mv 2 / 2 kT 8kT
v  4   0 v e dv 
 2kT  m

 n!
In   x 2 n 1  ax 2
e dx  n odd
0 2 n 1

• The probability distribution is broad!

• Distribution becomes broader at
higher temperature
9
The Maxwell – Boltzmann distribution for speed of one molecule

3/ 2
 m   mv 2 2 kT
P1 ( v )  4  
2
v e
 2kT 

Molecular dynamics simulation of ideal

gas and determining the velocity
distribution one molecule

10
 Maxwell – Boltzmann distribution for single molecule energy
We can determine an energy distribution for a single ideal gas molecule by changing
the variable from v to ε   mv 2 2  d  mvdv
3/ 2
• Distribution function for the 2  1  1/ 2  kT
P1 ( )d     e d
energy of one molecule:   kT 
• Most probable energy
dP1 ( )
 0   P1/2e P
1 1 1/2  P kT kT
kT
 P e  0  P 
d 2 kT 2
• Average energy

     P1 ( )d  32 kT
0

 n(n  1)   12  

n x
• Variance (spread) of the energy distribution? x e dx 
3/ 2
0  n 1
 1 
   
2

2
  0   

2  1 / 2e / kT d
  kT 
3/ 2
2  1   15  (kT )
7/2
18  (kT ) 7 / 2 9  (kT ) 7 / 2  3
      (kT ) 2
  kT   8 8 8  2
  11
Maxwell – Boltzmann distribution for a single molecule energy

3/ 2
2  1 
 P  12 kT P1 ( )d     e
1/ 2  kT
d
  kT 

σ = (3/2)1/2kT=1.22kT

  32 kT
• The distribution of energy for a
single ideal gas molecule is broad

 3 / 2kT 2
 
 3 / 2kT 3

12
Convolution: Maxwell – Boltzmann distribution for the energy of 2 molecules
In an ideal gas, the total energy is the sum of one-molecule energies and the
probabilities for energies in different molecules are uncorrelated
E   I   II P2 ( I ,  II )  P1( I ) P1( II )

Convolution of two one-molecule energy probability distributions

 P1 ( I ) P1 ( II )d I d II  0 P1 ( I ) P1 ( E   I )d I
E
P2 ( E ) 
 I  II  E
4  1/ 2 1 / 2  E / kT
P2 ( E )    ( E   ) e d
 kT  3 0

The integral can be evaluated (see Chapter 5) to give

1
P2 ( E )  E 2 e  E / kT
2( kT )3
13
 Maxwell – Boltzmann distribution for energy of two molecules
1
P2 ( E )  E 2 e  E / kT
2( kT )3

dP2 ( E )
Most probable two-
molecule energy:
 0  2 EP e EP / kT  1 EP2 e  EP / kT  0  EP  2kT
dE kT

1  3  E / kT 1 6
Average: E   E e dE   3kT
3 0 3 4
2(kT ) 2(kT ) (1 / kT )

Variance: 2 
1 
3 0

E  E 
2 2  E / kT
E e dE
2(kT )
1  5 4 3 2
 24(kT )  12 E (kT )  2(kT ) E  3(kT ) 2
2(kT )3  

 3kT 1
 
E 3kT 3 14
 Maxwell – Boltzmann distribution for energy of N-molecules
We can continue the convolution process to get probability for energies of N molecules:
1
E   I   II     N PN ( E )  3 N E [( 3 N / 2 ) 1]
exp(  E kT )
 2  1 !(kT ) 3N / 2

Most probable: dPN ( E )  0  E   3N 2  1 kT

P
dE
1 
Average: E 
 3N
2  1 !(kT )3 N /2  0
E (3 N /2) exp( E kT )dE  3N
2 kT

Variance:

E   1 
 E 2  3NkTE   3 NkT 2  E (3 N /2) 1 exp( E kT )dE 

2
E  3N
(kT ) 2
 32N  1 !(kT )3 N /2 0  2  2

   2
E 3N

• Change of variables to energy relative to the maximum E/EP = E*

PN ( E )dE 
3 N 2  13 N / 2 [(3 N / 2) 1]
E* exp 3 N 2  1E *dE*  PN ( E*)dE *
3N 2  1 ! 15
 Maxwell – Boltzmann distribution for energy of N-molecules
E   I   II     N PN ( E*) 
3 N 2  13 N / 2 [(3 N / 2) 1]
E* exp 3 N 2  1E *
3N 2  1 !
“Reduced energy” E*=E/EP

Probability distributions for 1 to 10 molecule energies

EP*  1
• Probabilities get “sharper” as the
1
E*  number of molecules increases
1  32N
• The shape of the distribution becomes
PN(E*)

Gaussian (central limit theorem)!

E* 16
 Energy distributions for large collections of molecules

PN ( E*) 
3 N 2  13 N / 2 [(3 N / 2) 1]
E* exp 3 N 2  1E *
3N 2  1 !
Probability distributions for 10 to 1000 molecule energies

Stirling’s approximation can be used to

evaluate factorials of large numbers:
ln N ! N ln N  N  N ! e N ln N  N

• Probabilities get sharp and narrow for large N

PN(E*)

• The most probable energy Emax overwhelms

the distribution

N 1

X N

E*
17
 Central limit theorem applies for the energy distribution of for large
collections of molecules
1 [(3 N /2) 1]
PN ( E N )  exp(  E N kT )
 32N  1!(kT )
E
3 N /2 N

Simplify the notation using n = 3N/2 – 1 and β = 1/kT to define:

x = (EN – n/β)/(n/β)

n n 1e n
PN ( E N )dE N  P ( x)dx  (1  x) n e  nx dx
n!
n nx nx 2 /2
Using the logarithm approximation for small x: lim (1  x)  e e
x 0

n n 1e n nx 2 / 2 2 nn n e  n
= 1 from the relation
n nx 2 / 2
P( x)  e  e for the factorial of
n! n! 2 large numbers

The energy distribution for the N molecule system becomes Gaussian!

n  nx 2 / 2 1  ( E N  E N , P )2 / 2 N
2
P ( x)dx  e dx  PN ( E N )dE N  e dE N
2 2 N
2 18