Role of Cavity Dispersion in Femtosecond Mode-Locked Lasers;

Phase Velocity, Group Velocity, and Group Delay Dispersion:

Let us consider a linearly polarized, monochromatic em wave at frequency ω, propagating in the

z-direction in a transparent medium. The electric field E(z,t) of the wave at a time t can be
written as,
E (z, t ) = A0 exp[ j (ωt − βz )] , where
A0 a constant
β propagation constant which is a function of angular frequency ω, i.e.,
β(ω) characteristics of the medium
is the dispersion relation of the medium.

Since the total phase of the wave

φt = ωt − βz = const.,
The temporal and spatial coordinates must satisfy the condition,
dφt = ωdt − βdz = 0
dz ω
⇒ v ph = = (8.26.24) ,
dt β
is the phase velocity of the wave.

Fig. 8.25 Phase velocity and group

velocity in a dispersive medium

Next we consider a light pulse traveling in the medium and let ωL and ∆ωL be, respectively, the
center frequency and the width of the corresponding spectrum. We then assume that the
dispersion relation over the bandwidth ∆ωL can be approximated by a linear law,

 dβ 
β = βL +   (ω − ωL ), where
 dω ω =ω L

βL propagation constant corresponding to the frequency ωL

Substituting the value of β in the Eqn. of E(z,t) and considering the Fourier expansion of the
wave,
 ∞    dβ   
E (z, t ) = exp[ j (ωLt − β L z )] ∫ Aω (∆ω ) exp  j∆ω t −   z  d∆ω
−∞    dω ω L  
where, ∆ω = ω − ω L , and
Aω = Aω (ω − ω L ) complex amplitude
β = β (ω − ω L ) describes the dispersion relation of the medium for the pulse.

  z  ,
Integrating over ∆ω, the above Eqn. leads to a function of single variable t −  dβ
  d ω ωL 

 
E (z, t ) = exp[ j (ωLt − β L z )] × At −  dβ  z 
  dω  ω L 
  z 
⇒ E (z, t ) = At −   exp[ j (ωLt − β L z )] (8.6.25) ,
v 
  g 
Where, A is the pulse amplitude, exp[ j (ω L t − β L z )] is the carrier wave and group vg is given by,
 dω 
v g =   (8.6.26)
 dβ  β = β L

Therefore, after traveling a length l of the medium, the pulse is subjected to a time delay,

 dβ 
τg = l = l  = φ ′(ωL ) (8.6.27 )
 dω   d ω ω =ωL
 
 dβ ω =ωL

Here, we have defined a phase φ, dependent on the frequency ω, such that

φ (ω − ωL ) = β (ω − ωL )l (8.6.28)
and we have set,
 dφ (ω − ωL )
φ ′(ωL ) =   .
 dω ωL

The quantity, φ′(ωL) = τg is referred to as the group delay or the inverse of the group velocity at
frequency ωL of the medium.

(d)

Fig. 8.25 (b) Dispersion in time delay for two pulses of carrier frequencies ω1 and ω2 and (c) group velocity
dispersion for a pulse of large oscillation bandwidth ∆ωL, (d) visual picture of pulse propagation in a medium.

where we have set,

  d 2φ 
φ ′′(ω1 ) =  2  .
 dω ω1

Equation (8.6.29) holds exactly if the relation between φ and ω, in the frequency range between
ω1 and ω2, can be approximated by a parabolic law (Taylor series expansion), viz.;

From Eqn. (8.6.30),

dφ  dφ   d 2φ 
=  +  2  (ω − ωL ) + 
d ω  d  ω L  dω  ω L
ω

and then comparing with Eqn. (8.6.29) we obtain,

∆τ d ≅ φ ′′(ωL ) ∆ωL (8.6.31)

Using Eq. (8.6.28) into (8.6.31), ∆τd also can be written as,

 d 2β 
∆τ d ≅ l  2  ∆ωL = φ ′′(ωL )∆ωL = GDD (8.6.32 )
 dω ω L
Again,

  1 
 d  vg  
 d 2β 
 2  =    = 1  ∆τ g  = GVD (8.6.33)
 dω  ω L  dω  l  ∆ωL 
 
ωL

From Eqs. (8.6.32) and (8.6.33), we can get,

GDD
GVD = (8.6.34) .
l

Limitation on Pulse Duration Due to Group-Delay Dispersion:

When a dispersive medium is present within a mode-locked laser cavity, an approximate value
for the steady state pulse duration can be obtained from the condition that relative time
shortening due to mode locking (gain time window as shown in Figs. 8.22/8.24), must be equal
to the pulse broadening due to both the gain medium (δτp/τp)g and the dispersive medium
(δτp/τp)d.

ML (Saturable Absorber)

Gain Medium
(δτ )
p g Absolute value of pulse broadening due to gain medium

 δτ p 
 
τ  Relative value of pulse broadening due to gain medium
 p g
(δτ )
p d Absolute value of pulse broadening due to dispersive medium

 δτ p 
 
τ  Relative value of pulse broadening due to dispersive medium
 p d
(δτ )
p s Pulse shortening due to mode-locking (e.g. saturable absorber)

 δτ p 
 
τ  Relative value of pulse shortening due to mode-locking
 p s

Under steady state condition,

 δτ p   δτ p   δτ p 
  =  + 
τ  τ  τ 
 p  s  p  g  p d
 δτ p   2 ln 2   φ ′′2
g + (8 ln 2 )
  =  1 
 τ   π 2  ∆τ 2 ∆ν 2  0
2

 p s  p 0  ∆τ p
4

 δτ p   g   φ ′′2 
  = 0.14 0  + 3.84  (8.6.36)
τ   2   4 
 ∆τ p ∆ν 0   ∆τ p 
2
 p s
∆ω0
Where, ∆ν 0 =
2π and ∆ τ p = 2 (ln 2 ) 1
2τ
p . Here, φ ′′ = GDD is calculated

at the central frequency ωL of the laser pulse.

  g0   φ ′′2 
0.14   = 3.84 
 ∆τ 2 ∆ν 2   ∆τ 4 
 p 0   p 
1
 27.4  2
∆τ p =   φ ′′ ∆ν 0 (8.6.37 )
 0 
g

For Example:
∆ν 0 ≅ 100 THz approximate bandwidth for a Ti:Sapphire gain medium

φ ′′ = 100 fs 2 equivalent @ λ ≅ 800 nm to the presence of ~ 2 mm of quartz

material in the cavity.
g0 = 0.1 (say) then we obtain from Eq. (37),
∆τ p ≅ 162 fs .

To get pulses shorter than 150 fs, down to the inverse of the gain line width ∆ν0
1
∆τ ′p = ≅ 4 fs , we must reduce the GDD by about an order of magnitude.
∆ν 0

To obtain shortest pulses, in addition to the compensation of second order GDD, we must also
compensate higher order dispersion terms in the power expansion of Eq.(30). The next term to be
considered in Eq. (30) is third-order dispersion (TOD), defined as TOD = φ ′′′ = β ′′′l ,
where third order derivatives are taken at laser’s central frequency ωL. We do not consider the
effects of TOD on an incident pulse and discussion of this topic. However, we must point out
that in case of a Ti:Sapphire laser (with ∆ν 0 ≅ 100 THz ), TOD begins to play a pulse limiting
role for pulses shorter than ~30 fs.
Dispersion Compensation:

Now to obtain a negative GDD for ω2 > ω1 , one must have, ∆τ g < 0
i.e., from Eq. (29), ∆τ g = φ ′(ω2 ) − φ ′(ω1 ) < 0
τ g (ω2 ) − τ g (ω1 ) < 0
τ g (ω2 ) − τ g (ω1 ) < 0
τ g (ω2 ) < τ g (ω1 )
This is exactly what the four prism sequence does. In fact, the angular dispersion of the prisms is
such that two pulses at ω2 and ω1 entering the prism sequence at the same time and in the same
direction but propagate at two different paths and exit at two different times as shown in the Fig.
The path length for the pulse at frequency at ω2 is thus shorter than that for the pulse at
frequency ω1, which actually means GDD = φ ′′ < 0 .
Example:
For quartz prisms, and a length l = 250 mm gives a negative dispersion, which can

compensate for the positive GDD at λ ≅ 800 nm , of a quartz element with a thickness of 6.6
mm i.e., φ ′′ ≅ −360 fs 2 .

The four-prism sequence in Fig . 26 is a convenient way of introducing a negative GDD in a

laser cavity for the following reasons:
(1) Since all faces are at Brewster’s angle to the beam paths, losses by the system
are low.
(2) The negative value of GDD can be coarsely changed by changing the
separation l of the two prism pairs.
(3) By translating any one of the prisms along an axis normal to its base (suppose
prism II), we change the total length of the optical medium traversed by the
medium. This motion introduces, in a finely and controlled way, a positive
(material) dispersion of adjustable size without altering the ray directions and
hence the negative dispersion due to the geometry of the ray paths.
(4) The transmitted beam is collinear with the incident beam; this facilitates
inserting the four prism sequence in an already-aligned cavity. Note that, since
MM′ is a symmetry plane, we can use just the first two prisms in a two-mirror
resonator if one cavity mirror is plane and located at the position MM′. In this case
the GDD per pass is of course half that of the four-prism sequence.

Prism-pair arrangement
for negative dispersion
Un-chirped

Up Chirp
 Un-chirped

Down Chirp

Fig. Arrangements of down chirp using multi layer mirror pair