You are on page 1of 35

Laser optics

Mike Tarbutt, MSc Optics 2010.

1. Introduction

ü Aims

By the end of this course, you should:

æ Understand the properties of laser beams,

æ Understand how laser beams are transformed by optical elements,

æ Understand the properties of optical resonators,

æ Be able to apply what you've learnt to practical problems.

ü Course outline

æ The fundamental Gaussian beam - its derivation and its properties

æ Higher-order Gaussian beams

æ Transformation of Gaussian beams by optical elements

æ Optical resonators - transverse and longitudinal modes, stability, mode frequencies

æ Unstable resonators

æ Laguerre-Gaussian beams

æ Applications

ü Sources of information

There are many text books that cover this material. The following ones contain most of the material, at the right level:

æ "Lasers" - Milonni & Eberly

æ "Lasers" - Siegman

æ "Modern classical optics" - Brooker

æ "Lasers and Electro-Optics" - Davis

æ "Introduction to Optical Electronics" - Yariv


Laser optics, Mike Tarbutt, 2010 2

2. Gaussian beams

ü 2.1 From the wave equation to the paraxial wave equation

Start from the scalar wave equation

1 ∂2 E Hr, tL
∇2 E Hr, tL = . (2.1)
c2 ∂ t2
The equation is linear, so we'll be able to apply the superposition principle. We'll be able to build general solutions from a
complete set of simpler solutions, so we'll look for monochromatic solutions (of course, many laser beams are very nearly
monochromatic):

E Hr, tL =  HrL e− ω t

Substituting this into Eq (2.1), and defining k = w êc, we get the Helmholtz equation:

∇2  HrL + k2  HrL = 0 . (2.2)

We want to find a solution that "looks like" a laser beam. The solution should be almost unidirectional, i.e. it should propa-
gate along z with a small, but non-zero, divergence. The solution should also have a limited spatial extent in the transverse
direction. Think of some solutions we already know about - plane waves and spherical waves.

A
Plane waves, HrL=A ‰Â k z Spherical waves, HrL = ‰Â k r
r

Neither looks much like a laser beam. The plane wave has no localization and no divergence. The spherical wave has
maximal localization (a point), and maximal divergence. We'll look for a solution that has the general form

 HrL = U HrL  k z , (2.3)

where UHrL is a slowly-varying function of position, in the sense that it does not change much on the scale of the wavelength
l.

We'll substitute (2.3) into (2.2). Let's evaluate the z-derivatives first:
∂ ∂U
U HrL  k z = +  k U  k z
∂z ∂z
∂2 ∂2 U ∂U
U HrL  k z = +2  k − k2 U  k z
∂ z2 ∂ z2 ∂z
∑2 ∑2
Substituting into (2.2), and using a shorthand notation “T 2 = + (sometimes called the "transverse Laplacian"), we
∑x2 ∑y2

reach

∂2 U ∂U
∇T 2 U + +2k = 0. (2.4)
∂ z2 ∂z
Laser optics, Mike Tarbutt, 2010 3

So far, everything has been exact. At this point, we'll use our assumption that UHrL is a slowly-varying function of position to
make an approximation. The value of UHx0 , y0 , z0 + lL is related to UHx0 , y0 , z0 L by the Taylor expansion,
∂U
U Hx0 , y0 , z0 + λL = U Hx0 , y0 , z0 L + λ + ...
∂z
∑U
Since the change in U over the distance l is supposed to be very slight, we must have l ∑z
<< U. A similar argument
holds for the gradient of U:

∂U ∂U ∂2 U
Hx0 , y0 , z0 + λL = Hx0 , y0 , z0 L + λ + ...
∂z ∂z ∂ z2
∑2 U ∑U
Since the gradient is not supposed to change much over the distance l, we must have l << ∑z
, or equivalently
∑z2
∑2 U ∑U
<< k ∑z
. Therefore, we can neglect the second term in Eq (2.4) because it is so much smaller than the the third
∑z2
term. Then, we're left with the paraxial wave equation:
∂U
∇T 2 U + 2  k = 0. (2.5)
∂z
Laser optics, Mike Tarbutt, 2010 4

ü 2.2 The fundamental Gaussian beam

We're going to use a trial solution, and show by substitution that it is indeed a solution of the paraxial wave equation. Let's
first try to motivate the trial solution a bit, so that it seems like a reasonable guess from the outset. We have the idea that the
wavefronts of a laser beam, diverging away from some focal point might, very schematically, look a bit like this:

i.e. there's some radius of curvature, the beam has a limited transverse extent, and it is growing as it propagates. If we
consider the wavefronts in the plane z = R to be a small part of some much larger sphere of radius R (i.e. x, y << R, for all
regions of interest), we can make the approximation
1ê2
x2 + y2 x2 + y2
r = Ix2 + y2 + R2 M
1ê2
= R 1+ ≈ R+ + ...
R2 2R

As such, in the vicinity of x = y = 0, that small portion of the spherical wave can be approximated by the form
A A
 HrL =  k R  k Hx +y2 Lë2 R
2
 k r ≈ .
r R
Inspired by this, we choose a trial solution of the form
1
 k Hx
2 +y2 Lë2 q HzL
U HrL = A (2.6)
q HzL

This looks similar to the above, except that we have replaced the fixed value R with a function of z, qHzL, recognizing that the
'radius of curvature' is probably going to depend on z (at the very least, the beam has to converge towards a focal point, and
diverge away from it!). In addition, we'll allow for the possibility that q is complex.

We'll need the following derivatives of (2.6) ...

∂U 1 q k Ix2 + y2 M q
= − − U
∂z q z 2 q2 z
∂U kx
= U
∂x q
∂2 U k k2 x2
= − U
∂ x2 q q2

Substituting into (2.5) we get

2 k k2 Ix2 + y2 M 2 k q Ix2 + y2 M
− − + k2 =0
q q2 q z q2

We see from this that our trial solution is indeed a solution of the paraxial wave equation provided that
q
= 1. (2.7)
z
Thus, we see that qHzL is just a linear function of z. Letting q0 ª qH0L, we have

q HzL = q0 + z. (2.8)

We said that qHzL can be complex. So let's separate it out into a real and an imaginary part as follows:
Laser optics, Mike Tarbutt, 2010 5

1 1 λ
= + (2.9)
q HzL R HzL π w HzL2

Here, RHzL and wHzL are both real, and the reason for inserting the l and the p will soon become clear. Let's find the functions
RHzL and wHzL, using what we know about qHzL. We define the (arbitrary) position z = 0 such that q is purely imaginary here,
i.e. z = 0 is at some position where the wavefronts are planar (R = ¶). We'll also define w0 ª wH0L. Then, we have
1 λ 
= = ,
q0 π w0 2 z0

where

π w0 2
z0 = (2.10)
λ
Then,

1 1 1 ê q0  ê z0  ê z0 H1 − z ê z0 L z ë z0 2 +  ê z0 1  w0 2
= = = = = = + ,
q q0 + z 1 + z ê q0 1 +  z ê z0 1 + z2 ë z0 2 1 + z2 ë z0 2 R z0 w2

where we used (2.9) and (2.10) in the last step. Equating the real and imaginary parts we obtain

z0 2
R HzL = z + (2.11)
z
z0 w2
= z0 I1 + z2 ë z0 2 M
w0 2

⇒ w HzL = w0 1 + z2 ë z0 2 (2.12)

Finally, to deal with the complex denominator of Eq (2.6), let's write

1 z ë z0 2 +  ê z0
=  a HzL e−φ HzL = ,
q HzL 1 + z2 ë z0 2

where a, f are real. We find a and f in the usual way, with the results:
1 1
a HzL = (2.13)
1 + z2 ë z0 2
z0

z
φ HzL = tan−1 (2.14)
z0

This completes the solution - what we have found is known as the fundamental Gaussian beam.
Laser optics, Mike Tarbutt, 2010 6

ü 2.3 A summary

Because of the cylindrical symmetry of the solution, it will often be helpful to use the radial coordinate defined by
r2 = x2 + y2 . To summarize, the solution is:

A  Hkz−ωtL ë2 q HzL
E Hr, tL =
2
 kρ (2.15)
q HzL

where the z-dependence of q is

q HzL = q0 + z . (2.16)

The above form is the most memorable one. It is often useful, and adds extra insight, to separate the real and imaginary parts
of q by writing:

1 1 λ
= + (2.17)
q HzL R HzL π w HzL2

Then, we get

A
 Hkz−ωtL −ρ
2 ëw2 2 ë2
E Hr, tL =  k ρ R
−φ , (2.18)
1 + z2 ë z0 2

where the spot size is

w HzL = w0 1 + z2 ë z0 2 , (2.19)

the radius of curvature is

z0 2
R HzL = z + , (2.20)
z

and the Gouy phase is

z
φ HzL = tan−1 . (2.21)
z0

The Rayleigh range z0 is related to the spot size at the waist w0 by

π w0 2
z0 = . (2.22)
λ
Laser optics, Mike Tarbutt, 2010 7

ü 2.4 Properties of Gaussian beams

ü Intensity distribution and spot size

The intensity profile of the Gaussian beam is, unsurprisingly, a Gaussian:


I H0, 0L 2 ëw2
I Hρ, zL = E 2
= −2 ρ
1 + z2 ë z0 2

For any z, the intensity falls to 1ëe2 = 0.135 of its central value when r = wHzL. It's easy to show that the full width at half
maximum (FWHM) of the intensity profile is proportional to w:

FWHM = 2 ln 2 w = 1.177 w

The spot size, wHzL, has its smallest value (w0 ) at z = 0 (the beam 'waist'), increasing on either side. The figure below shows
how the spot size w varies with z for a Gaussian beam that has a wavelength of l = 500 nm and a waist of w0 = 100 mm. The
black lines show the 1ë e2 extent of the beam.

0.4

0.2
wHmmL

0.0

-0.2

-0.4

-200 -100 0 100 200


zHmmL

The intensity profiles for this beam at the positions z = 0, 50 mm, 100 mm (indicated by the dashed lines above) are plotted
below.
1.0

0.8
Normalized intensity

0.6

0.4

0.2

0.0
-0.3 -0.2 -0.1 0.0 0.1 0.2 0.3
r HmmL

ü Rayleigh range

The Rayleigh range is the distance from the waist at which the spot size has grown by 2 , i.e. wHz0 L = 2 w0 . This is
easily seen from Eq (2.19). The Rayleigh range scales as the square of the waist size (Eq (2.22)) - so you can only have a
small spot size over a small distance, whereas a large beam can travel a long way without expanding much (note the connec-
tion with diffraction). The Rayleigh range for a l = 500 nm beam is plotted as a function of the waist size below (N.B
logarithmic scale). The Rayleigh range is only 6mm when the waist is 1mm, but is 6m when the waist is 1mm.
Laser optics, Mike Tarbutt, 2010 8

1000

100
z0 HmmL

10

0.1

0.01
0.001 0.005 0.010 0.050 0.100 0.500 1.000
w0 HmmL

Some more terminology: while z0 is called the Rayleigh range, 2 z0 is often called the confocal parameter or the depth of
focus.

ü Radius of curvature

The radius of curvature is given by (2.20), and is plotted below as a function of z for a beam with l = 500 nm, w0 = 100 mm.
At the beam waist (z = 0) the wavefronts are planar HR = ¶). As z increases the radius of curvature decreases, reaches a
minimum and then increases again.
At large z, the radius of curvature is equal to the distance from the waist, R º z (shown by dashed line in figure). In this limit,
the wavefronts are the same as those of a spherical wave centred at the origin. (Note - this does not mean that the beam looks
just like a spherical wave - it doesn't at all. The wavefronts are the surfaces of constant phase. They do not tell you anything
about the intensity distribution on these surfaces, which are very different for the two cases - uniform for a spherical wave,
localized near the beam axis for the Gaussian beam).

600

400

200
R HmmL

-200

-400

-600
-200 -100 0 100 200
z HmmL

The figure below shows what the wavefronts look like for our l = 500 nm, w0 = 100 mm beam. Note that, in order to make
the curvature of the wavefronts visible I have chosen an x-range that is much larger than the spot size of the beam.
30
20
x HmmL

10
0
-10
-20
-30
-200 -100 0 100 200
z HmmL

ü Beam divergence

The far-field (z >> z0 ) divergence angle is

w HzL w0 λ
θ= ≈ ≈
z z0 π w0
Laser optics, Mike Tarbutt, 2010 9

0.4

0.2
wHzL
q
wHmmL

0.0

-0.2

-0.4

-200 -100 0 100 200


zHmmL

This may be interpreted as the spread due to diffraction. Recall that the first zero of intensity in the far-field diffraction
l
pattern of a slit of width a is at the angle q = a . For a circular aperture, there is a factor of 1.22 out the front. The general rule
l
is that the divergence angle due to diffraction is Hfactor ~ 1L a . The result for the Gaussian beam conforms to this.

ü Gouy phase

The phase fHzL = tan-1 Hz ê z0 L is usually called the Gouy phase. Below, it is plotted as a function of z, for our
l = 500 nm, w0 = 100 mm beam. The phase changes from -p ê2 far to the left of the waist to zero at the waist to p ê2 far to
the right of the waist. At z = ≤ z0 , f = ≤ p ê 4, so the value of z0 sets the scale for how quickly f varies as we pass through the
beam waist. The total phase on the beam axis (r = 0) is FHzL = k z - fHzL. Even near the waist, where it has its maximum rate
of change, the Gouy phase changes very slowly compared to the first term (which changes by 2 p per wavelength).

0.4

0.2
HradsL

0.0
f

-0.2

-0.4

-400 -200 0 200 400


z HmmL

Note how the phase of the Gaussian beam changes relative to the phase of a plane wave. On the beam axis, and to the left of
the focus, the Gaussian beam wavefront of a given phase lags behind the same wavefront of the plane wave. To the right,
they are ahead. At the waist, the wavefronts of the Gaussian beam are the same as those of the plane wave.

z = -100 mm z=0 z = 100 mm

ü 2.5 Higher order Gaussian beams

The Gaussian beam that we looked at so far is just one of an entire family of solutions to the paraxial wave equation - it is
usually called the fundamental Gaussian beam. In this section, we look for the more general solutions. We take the following
trial solution:
x y
f HzL  k Hx
2 +y2 Lë2 q HzL
U HrL = A g h (2.23)
w HzL w HzL
Laser optics, Mike Tarbutt, 2010 10

Here, qHzL is exactly the same as before - it is qHzL = q0 + z, with q0 = -Â z0 . It follows that the spot size and the radius of
curvature of our more general solutions will be exactly the same as for the fundamental Gaussian beam (i.e. as in (2.19) and
(2.20)). Note that if gHx êwL = hHx êwL = 1 and f HzL = 1 êqHzL, we recover the fundamental Gaussian beam. By writing g and h
as functions of x ê wHzL and y ê wHzL, we ensure that the intensity distributions scale with the spot size wHzL. Our task now is to
find the functions f , g, h such that (2.20) is a solution of the paraxial wave equation.

Define
x y
X= , Y=
w HzL w HzL

The derivatives of g can then be written


∂g g ∂X 1 g
= = ;
∂x X ∂x w X
∂2 g 1 2 g
= ;
∂ x2 w2 X2
∂g g ∂X x w g X w g
= =− =−
∂z X ∂z w2 z X w z X

Using the expression for w (2.19) we see that

1 w z 1  w0 2
= = − ,
w z z2 + z0 2 q z0 w2

1 1 Â w0 2 ∑g
where in the last step we used the fact that q
= R
+ z0 w2
, and equation (2.20). We use the above in the expression for ∑z
...

∂g  w0 2 1 g
= − X .
∂z z0 w2 q X

The derivatives for h can be written down by analogy. Now work out the relevant derivatives of U:

∂U 1 ∂g 1 ∂h 1 f  k Ix2 + y2 M q
= + + − U
∂z g ∂z h ∂z f z 2 q2 z

1  w0 2 1 g 1  w0 2 1 h 1 f  k w2 IX2 + Y2 M
= : − X + − Y + − > U.
g z0 w2 q X h z0 w2 q Y f z 2 q2
„q
In the last step above, we used the fact that „z
= 1. Continuing...

∂U 1 ∂g kx
= + U
∂x g ∂x q
∂2 U 1 ∂2 g 1 ∂g 2 k 1 ∂g 2 k2 x2 2 kx ∂ g
= − + + − + U
∂ x2 g ∂ x2 g2 ∂x q g2 ∂x q2 qg ∂x
1 1 2 g k k2 w2 X2 2 k X g
= + − + U,
g w2 X2 q q2 qg X

and similarly for the derivatives with respect to y. Now we put all of this into the paraxial wave equation, Eq (2.5)...
Laser optics, Mike Tarbutt, 2010 11

∂2 U ∂2 U ∂U
+ +2k =0
∂ x2 ∂ y2 ∂z

1 1 2 g 1 1 2 h 2 k k2 w2 IX2 + Y2 M 2 k X g 2 k Y h
∴ + + − + + +
g w2 X2 h w2 Y2 q q2 qg X qh Y

2k 1 w0 2  g 2k 1 w0 2  h 2  k f 2 k2 w2 IX2 + Y2 M
− − X + − − Y + + =0
g z0 w2 q X h z0 w2 q Y f z 2 q2

Cancelling the terms that cancel, and tidying things up...

1 1 2 g 1 1 2 h 2 k 2 k w0 2 g 2 k w0 2 h 2  k f
+ + − X − Y + = 0.
g w2 X2 h w2 Y2 q g z0 w2 X h z0 w2 Y f z

Multiplying through by w2 , applying z0 = 1 ê 2 k w0 2 , and collecting terms together in an appropriate fashion...

1 2 g g 1 2 h h 2 k 2  k f
: −4X >+: −4 Y >+: + w2 > = 0.
g X2 X h Y2 Y q f z

This is the equation that must be satisfied if the form (2.23) is to be a solution of the paraxial wave equation. Notice that
everything in the first curly bracket is a function only of X , everything in the second curly bracket is a function only of Y ,
and everything in the third curly bracket is a function only of z. It follows that the equation can only be satisfied for all values
of the independent variables X , Y , z, if each of the curly brackets is a constant. So we put

1 2 g g
−4X = − 4 m, (2.24)
g X2 X

1 2 h h
−4Y = − 4 n, (2.25)
h Y2 Y
2 k 2  k f
+ w2 = 4 Hn + mL, (2.26)
q f z

where n, m are constants (and the reason for putting the 4 will soon become clear). We'll make one final change of variable
so that (2.24) and (2.25) are transformed into a standard form. The change is simply

u= 2 X, v = 2 Y.
Using

g g 2 g 2 g
= 2 , =2 ,
X u X2 u2

equation (2.24) is transformed into

2 g g
−2 u + 2 m g = 0, (2.27)
u2 u

with an analogous equation for h. This is a very well known differential equation that appears in many branches of physics
(e.g. harmonic oscillator in quantum mechanics). It's solutions only remain finite as u Ø ¶ if the constant m is a non-negative
integer. In that case, the solutions are the Hermite polynomials Hm . So we have found the allowed solutions for g and h:
x y
g = Hm 2 , h = Hn 2 , m = 0, 1, 2 ..., n = 0, 1, 2 ...
w w
We still have (19) left to deal with. It reads

1 f 2 1
=− Hn + mL − .
f z k w2 q

Using 1 ê q = Hz + Â z0 L ëIz2 + z0 2 M and 2 k ëw2 = z0 ëIz2 + z0 2 M this becomes


Laser optics, Mike Tarbutt, 2010 12

1 f z0 Hz +  z0 L z z0
= −  Hn + mL − =− − Hn + m + 1L
f z z2 + z0 2 z2 + z0 2 z2 + z0 2 z2 + z0 2

Integrating this we get


1 z
ln f = − ln Iz2 + z0 2 M −  Hn + m + 1L tan−1 .
2 z0

Taking the exponential of both sides, we finally end up with


1
f HzL = − Hn+m+1L φ HzL , (2.28)
z2 + z0 2

where fHzL is the Gouy phase encountered before, Eq (2.21). Putting everything together we have the general Hermite-
Gaussian beam:

E Hr, tL =
A x y
 Hkz−ωtL −ρ
2 ëw2 2 ë2
Hm 2 Hn 2  k ρ R
− Hn+m+1L φ (2.29)
ë z0
w w
1 + z2 2

Note well - z0 , wHzL, RHzL, fHzL have exactly the same definitions as before. Also note that since H0 HxL = 1, the form given
above reduces to the fundamental Gaussian beam when n = m = 0 - it is just one of an entire family of solutions of the
paraxial wave equation (though by far the most important). The Hm are just regular polynomials [e.g
H1 HxL = 2 x, H2 HxL = 4 x2 - 2 etc]. You can look them up in many text books (don't try to learn them!). The plots below
show the intensity distributions in an x y plane for a few of the lowest-order Hermite-Gaussian beam modes.
Laser optics, Mike Tarbutt, 2010 13

The shape of a Hermite-Gaussian beam does not change as the beam propagates. We can build up completely general
solutions on the basis of the Hermite-Gaussian solutions - so we have a general formalism that we can use to describe any
laser beam. The notation TEMmn is often used to label these Hermite-Gaussian beam modes - it stands for "transverse
electromagnetic of order (m, n)".

ü 2.6 Laguerre-Gaussian beams

We can also look for solutions to the paraxial wave equation that have cylindrical symmetry. Then we end up with another
family of modes called the Laguerre-Gaussian modes. Just as for the Hermite-Gaussians, the Laguerre-Gaussians are labelled
by two mode indices, p and l. They are described by:

C r l
E Hr, θ, z, tL = 2 ×
ë z0
w
1 + z2 2
(2.30)
2 r2
Lp l  Hkz−ωtL  −ρ2 ëw2
 k ρ2 ë2 R
− l θ − H2 p+l+1L φ
w2
Laser optics, Mike Tarbutt, 2010 14

Again, note that z0 , wHzL, RHzL, fHzL all have exactly the same definitions as before. The L p l are the associated Laguerre
polynomials [L0 0 HxL = 1, L1 0 HxL = 1 - x, L1 1 HxL = 2 - x etc]. You can look them up, or ask Mathematica what they are. Note
that the radial coordinate always appears scaled by w, meaning that, just as for the HG modes, the LG modes do not change
shape as they propagate. Also note the appearance of ‰-Â l q in the above expression, meaning that the phase of the wave is a
function of the azimuthal coordinate q. Consequently, the wavefronts of the Laguerre-Gaussian beams are helical, and there
is an orbital angular momentum of l Ñ per photon associated with the beam. Here are the intensity distributions of a few LG
beams:
Laser optics, Mike Tarbutt, 2010 15

3. Optics with Gaussian beams

ü 3.1 A reminder about ray matrices

To propagate rays through optical systems, we often use the ray matrix approach. Each optical element transforms the ray
according to
x2 x
J N=J N J 1 N,
A B
θ2 C D θ1

where the matrices for particular elements are easy to derive (or look up!). By writing it this way, it is obvious how to do the
transformation through any number of elements - the appropriate matrix is just the product of the matrices for the individual
elements (taking care to do the multiplication in the correct order, i.e. Mn Mn-1 ....M2 M1 ).

For free propagation through a length L,


x2 = x1 + L θ1 ,
θ2 = θ1 .
x x
∴J 2N=J N J 1 N.
1 L
θ2 0 1 θ1

For a lens of focal length f (using the Cartesian sign convention with origin in the lens, positive x-axis upwards, positive z-
axis to the right):

x1
q1 -q2

-z0 zi
x1 = x2 ,
x1 x2
θ1 = and − θ2 =
− z0 zi
1 1 1 x1
Since − = , θ1 − θ2 = .
zi z0 f f

x2 x
∴J N=J N J 1 N.
1 0
θ2 −1 ê f 1 θ1

For a curved mirror of radius of curvature R:

a a x1
b b c
R
Laser optics, Mike Tarbutt, 2010 16

a+b−β = a+β−c = 0
⇒ b − 2 β = − c.

b = θ1 and c = − θ2 and β > tanβ = x1 ê R

x2 x
∴ J N=J N J 1 N.
1 0
θ2 −2 ê R 1 θ1

ü 3.2 The ABCD rule for Gaussian beams

Our Gaussian beams are not rays. How will they be transformed by lenses, mirrors etc? The crucial thing is to find out what
happens to the q parameter, since we know everything about the beam if we know q (we just have to know it somewhere -
anywhere will do). Let's start with free space. That's easy - the answer is already given by Eq (2.16). If we have q = q1 in one
plane, then after travelling a distance L, the final q is

q2 = q1 + L HTransformation in free spaceL (3.1)

Now for a thin lens. The wavefront has a radius of curvature R1 at the input side of the lens. This is the same wavefront as
that of a spherical wave centred a distance -z0 = R1 to the left. We want to find the radius of curvature, R2 , of the wavefront
exiting the lens. This will be the same as that of a spherical wave converging to a point a distance v = -R2 to the right where
1 1 1
zi
- z0
= f . (Note the negative sign in front of R2 because the wave is converging from left to right).

R1 R2

-z0 zi

Therefore we have
1 1 1
= −
R2 R1 f
1 λ 1 λ 1
∴ − = − − ,
q2 π w2 2 q1 π w1 2 f
1 1 l
where we have used the relationship between q, R, and w - = +Â (Eq (2.17)). The beam size does not change across
q R p w2
the (thin) lens and so w2 = w1 . This leaves us with
1 1 1
= − .
q2 q1 f
q1
q2 = HTransformation by a thin lensL (3.2)
H− 1 ê fL q1 + 1

Next comes a spherical mirror. It's straightforward to show that the transformation of the q parameter is exactly the same as
(3.2) with f = R ê2. This will not be a surprise - the curved mirror has a focal length R ê 2.

The transformations (3.1) and (3.2) along with the similar one for the curved mirror, can be written in the general form
A q1 + B
q2 = (3.3)
C q1 + D

The values of A, B, C, D can be sumarized as follows:


Laser optics, Mike Tarbutt, 2010 17

J N=J N Hfree space, length LL


A B 1 L
C D 0 1
A B 1 0
J N= 1 Hthin lens, focal length fL
C D −f 1
1 0
J N= Hcurved mirror, radius of curvature RL
A B
2
C D − R
1

But these are exactly the same as the matrices for propagating rays! What's more, it's easy to show that the appropriate matrix
for a series of optical elements is just the product of the individual matrices (taken in 'reverse' order), exactly as it would be
for propagating rays. We can prove this explicitly. Take two optical elements, the first transforming q1 to q2 , and the second
transforming q2 to q3 , according to
A1 q1 + B1
q2 = ,
C1 q1 + D1

A2 J C1 q1 +D1 N + B2
A q +B
A2 q2 + B2 1 1 1 HA1 A2 + C1 B2 L q1 + HB1 A2 + D1 B2 L A3 q1 + B3
q3 = = = = .
C2 q2 + D2 C2 J C1 q1 +D1 N + D2
A q +B HA1 C2 + C1 D2 L q1 + HB1 C2 + D1 D2 L C3 q1 + D3
1 1 1

Thus the relationship between q3 and q1 has exactly the same form as (3.3), with (multiply it out yourself to check):
A3 B3
K O = J 2 2 N.J 1 1 N
A B A B
C3 D3 C2 D2 C1 D1

Another way to see the same thing is to recognize that (3.3) can be written in the form
q2 q
J N = HfactorL J N J 1 N,
A B
1 C D 1

where the factor out front is not interesting to us, since we'll always end up dividing it out. This form makes it obvious that a
sequence of optical elements is handled by matrix multiplication.

An extremely useful conclusion: To propagate a Gaussian beam through a set of optical elements, form the ABCD matrix as
you would in ray optics, and then apply Eq (3.3). -ote that we have only ever used the q-parameter, and that this is common
to all the Hermite-Gaussian beams. So this procedure applies to Hermite-Gaussian beams of any order.
Laser optics, Mike Tarbutt, 2010 18

ü Worked example 1 - focussing a Gaussian beam

Problem: A Gaussian beam, wavelength l, has a waist w01 in the plane of a lens of focal length f . Find the position and size
of the new waist to the right of the lens.

w01 w02

Solution:

Let the waist on the RHS of the lens be a distance d from the lens. We'll use w01 , w02 , z01 , z02 to represent the waists and
Rayleigh parameters on either side of the lens. The matrix for propagating the beam through the lens, and then to the waist is
d
1 0 1− d
J N=J N
A B 1 d f
1 =
C D 0 1 − f
1 1
−f 1

The incoming beam is at a waist, and so too is the final beam:


q1 = −  z01
I1 − f M H− z01 L + d
d

∴ q2 = −  z02 =
I− f M H− z01 L + 1
1

z01 z02 d
⇒ −  z02 = d −  z01 1 −
f f
Equating the real and imaginary parts we get
z01 z02
d=
f
z01
z02 = Hf − dL
f
Eliminating z02 from the first of these, using the second, we arrive at
1
d=f (3.5)
1 + f2 ë z01 2

In ray optics a collimated beam is focussed at a distance f from a lens. Equation (3.5) shows us that for a Gaussian beam the
focus is shifted towards the lens. The amount of shift depends on the ratio f ê z01 . When the input beam is large, z01 will be
very large and the shift relative to the ray optic picture will be very small. For small input beams however, the shift is
significant. The graph below shows the focus shift versus the size of the input beam for l = 500nm and a lens with f = 20cm.
Laser optics, Mike Tarbutt, 2010 19

80

60
Focus shift HmmL

40

20

0
0.2 0.4 0.6 0.8 1.0
w0 HmmL

Continuing with the above analysis we see that the new Rayleigh range is related to the old one as follows:

d 1 1
z02 = z01 1 − = z01 1 − = z01
f 1 + f2 ë z01 2 1 + z01 2 ë f2

Finally, using the above and the relation z0 = p w0 2 ël the new waist size is
w01 w01
w02 = =
1 + z01 2 ë f2
1 + IIπ w01 2 M ë f λM
2

λf 1
w02 =
π w01 (3.6)
1+Jπw 2 N
λf 2
01

When the beam striking the lens is large, p w01 2 >> l f and the size of the focal spot is just l f êHp w01 L. This is the result we
would expect for the diffraction limited spot size in focussing a plane wave of transverse extent w01 . The graph below shows
how the size of the focussed spot varies with the size of the beam striking the lens, for three different focal lengths (and
l = 500 nm). Note that spot sizes of just a few microns can be formed using very reasonable focal lengths and initial waists.

Size of focussed spot versus size of initial beam


60

50 f=200 mm

40
w02 HmmL

30
f=100 mm
20

f=50 mm
10

0
1 2 3 4 5
w01 HmmL
Laser optics, Mike Tarbutt, 2010 20

ü Worked example 2 - 1:1 imaging

Problem:

You want to image the waist, w0 , of a Gaussian beam with unity magnification, by placing a lens of focal length f a distance
z1 away from the waist.
(a) How far from the lens will the image be formed? Is this what you would expect?
(b) What is the required relationship between z1 and f ? Is this what you would expect?
(c) Show that for a given lens, it is impossible to make a 1:1 image of the waist if the waist is too large. What is the largest
size if l=400nm and f=30cm?
(d) If the lens is replaced by a curved mirror that does the same job (except with image and object on same side), what does
the radius of curvature need to be? Does the answer make sense?

Solution:

(a) Let's suppose the image is formed a distance z2 from the lens. Find the ABCD matrix that takes us from the original waist
to the new waist. It is space@z2 D.lens@ f D.space@z1 D:
z2 z1 z2
1 0 1 z1 1− z1 − + z2
J N=J N J N=J N
A B 1 z2 1 z1 1 z2 f f
1 1 z1 =
C D 0 1 −f 1 0 1 0 1 −f − f
+1 −
1

z1
+1
f f

From this we find the relationship between q2 and q1 . Since q1 is at a waist, we have q1 = -Â z0 , where z0 is the Rayleigh
range, z0 = p w0 2 ël. q2 is also at a waist, and the waist size must be the same as before for unity magnification, so q2 = q1 .
Applying these observations in the ABCD rule, we have:

I1 − M H−  z0 L + z1 + z2 −
z2 z1 z2
f f
= − z0
− f H−  z0 L + 1 −
1 z1
f

z2 z1 z2 z1 z0 2
⇒ z0 − 1 + z1 + z2 − = z0 −1 +
f f f f
Equating the imaginary parts gives us z2 = z1 . i.e. For a 1:1 image, the distance from lens to image is the same as the distance
from object to lens.

(b) Equating the real parts (and setting z2 = z1 ):

z1 2 z0 2
2 z1 − =
f f
z0 2
⇒ 2 f = z1 +
z1
i.e. z1 − 2 f z1 + z0 2 = 0
2

So we find that the lens need to be placed at

z1 = f + f2 − z0 2 .

Note - in ray optics you get a 1:1 image of an object when z1 = 2 f , so our result only reduces to the ray optics result when
z0 << f , i.e. for small waist sizes.

(c) It is only possible to form the image if z1 is real, and that requires z0 § f :

π w0 2
≤f
λ


i.e. w0 ≤
π
Laser optics, Mike Tarbutt, 2010 21

When f = 30cm and l = 400nm, the waist size must be smaller than 200mm in order to be able to form a 1:1 image.

(d) The focal length of a mirror is f = R ê 2, where R is the radius of curvature. So for the mirror, the result is
R = z1 + z0 2 ë z1 . But this is just the radius of curvature of the wavefront at z1 . So, to form an image of the waist back at the
same place as the original, we need to match the radius of the curvature of the mirror to that of the wavefront. This makes
good sense, since if the curvatures match the reflected beam will be the exact reverse of the incident beam.
Laser optics, Mike Tarbutt, 2010 22

4. Optical Cavities

ü 4.1 Cavity modes

R1 R2

We're going to show that a Hermite-Gaussian beam (of any order) can be a mode of an optical cavity, and find the conditions
that need to be satisfied to ensure this. By "mode", we mean an unchanging distribution of electromagnetic field - an eigen-
function of the wave equation in the presence of the boundaries formed by the mirrors.
We already have some clues that suggest a Gaussian beam for a mode:
- The Gaussian beam has the properties required of a laser beam, and a laser is a gain medium inside an optical cavity. If a
Gaussian beam comes out of a laser cavity, it seems reasonable that there is a Gaussian beam inside the cavity!
- A Gaussian beam propagates without changing its shape, which is what we need for a cavity mode.
- In the problem discussed above, we already saw that an incoming Gaussian beam is exactly reversed by a mirror whose
radius of curvature matches that of the incoming wavefront. So we get the idea that two such mirrors may support a mode.
This not only suggests that a Gaussian beam can be a mode, but also indicates what we have to do to make it so.

If a Gaussian beam is to be a mode, then it's q parameter must be unchanged after one round trip of the cavity. Let's find the
ABCD matrix for a round-trip. We can start anywhere we like (since qHzL is known for all z once you know it in one plane),
and it will be convenient to start at the left-hand mirror. The matrix we need is mirror@R1 D.space@LD.mirror@R2 D.space@LD.
[Note: it's easy to end up in a minus-sign nightmare when dealing with beams that are sometimes propagating to the right and
sometimes to the left. My advice is always to 'unfold' the diagram, replacing every mirror by an equivalent lens with
f = R ê2, so that the beam always propagates from left to right. That should keep things simple].
We have:
1 0 1 0
J N= J N J N
A B 1 L 1 L
2 2
C D − R1
1 0 1 − R2
1 0 1
1 L 1 L
= 2 2L 2 2L
− R1
1− R1
− R2
1− R2

2L 2 L2
1− R2
2L− R2
=
2 2 4L 2L 4L 4 L2
−R − R2
+ R1 R2
1− R2
− R1
+ R1 R2
1

For a mode, we require


A q1 + B
q2 = = q1
C q1 + D
⇒ C q1 2 − HA − DL q1 − B = 0,

from which we find

2 C q1 = HA − DL ± HA − DL2 + 4 B C = HA − DL ± HA + DL2 − 4 HAD − BCL

Now, A D - B C is the determinant of the ABCD matrix. Each of our elementary matrices has unity determinant, and since
the determinant of a product of square matrices is the product of their determinants, we must have A D - B C = 1. So the
above equation reduces to
Laser optics, Mike Tarbutt, 2010 23

2 C q1 = HA − DL ± HA + DL2 − 4

Recall that the q parameter always has an imaginary part, unless the spot size w is infinite (which we don't want). To ensure
this, the expression under the square root has to be negative. So we find a condition for the cavity to support a Gaussian
beam:

HA + DL2 < 4
1
⇒ HA + DL2 < 1
4
1
⇒ −1 < HA + DL < 1
2
Adding 1, and then dividing through by 2, this becomes
1
0< HA + D + 2L < 1
4
From the ABCD matrix above, we find that this condition is:
L L
0 < 1− 1− < 1.
R1 R2

It is usual to introduce the g parameters of the cavity, gi = 1 - Lê Ri . Then, the criterion for a mode is simply

0 < g1 g2 < 1. (4.1)

The plot below shows the "stable" regions (in green) as a function of g1 and g2 .

1
g2

-1

-2
-2 -1 0 1 2
g1

Now let's find the properties of the mode. The cavity is specified by the values of R1 , R2 and L. We know that the mode is a
Gaussian beam, so we need to find out the size of the waist and its location relative to one of the mirrors. Let's locate the
waist at z = 0 and the left-hand mirror at z = z1 . Our task is to find z1 and w0 ,which together will specify the properties of the
mode.

We already have an expression for q1 above, and we know that q1 = qHz1 L = q0 + z1 = -Â z0 + z1 , where z0 is related to w0 in
the usual way. Thus...

HA − DL 4 − HA + DL2
z1 − z0 = ± ,
2C 2C
Laser optics, Mike Tarbutt, 2010 24

where the expression under the square-root is now known to be positive if the mode exists. We can get z0 and z1 by equating
real and imaginary parts. The algebra gets a bit tedious here. We'll make use of the relation
g1 g2 = 1 - L ê R1 - Lê R2 + L2 ë R1 R2 . From the real part...

A−D I4 L − 4 L2 ë R1 R2 M L H4 L ê R1 − 4 g1 g2 − 4 L ê R1 − 4 L ê R2 + 4L L
z1 = = =
2C 8 L2 ë R1 R2 − 4 L ê R1 − 4 L ê R2 8 g1 g2 + 4 L ê R1 + 4 L ê R2 − 8
L g2 H1 − g1 L
⇒ z1 = − .
g1 + g2 − 2 g1 g2

And from the imaginary part...

1 − I 2 HA + DLM
2
1 − I2 − 2 L ê R1 − 2 L ê R2 + 2 L2 ë R1 R2 − 1M
1 2
2
z0 = = L2 =
C2 I4 L2 ë R1 R2 − 2 L ê R1 − 2 L ê R2 M
2

1 − H2 g1 g2 − 1L2 4 g1 g2 − 4 g1 2 g2 2
L2 = L2
H4 g1 g2 + 2 L ê R1 + 2 L ê R2 − 4L2 H4 g1 g2 − 2 g1 − 2 g2 L2
g1 g2 H1 − g1 g2 L
⇒ z0 2 = L2
H2 g1 g2 − g1 − g2 L2

From z0 , we get w0 :

λL g1 g2 H1 − g1 g2 L 1ê4
w0 =
π H2 g1 g2 − g1 − g2 L2

Condition (4.1) is a necessary condition for there to be a Gaussian mode...but we still have more to do. Consider the phase of
the field on the optical axis. It is
z
θ HzL = k z − Hn + m + 1L tan−1 .
z0

The condition for a mode is that the field does not change in one round trip, meaning that this phase can only change by some
integer multiple of 2p in a round trip. This means that in going from one mirror to another the phase must change by an
integer multiple of p:
2π z2 z1
Hz2 − z1 L − Hn + m + 1LBtan−1 − tan−1 F = p π, p = 0, 1, 2, ...
λ z0 z0

Writing the frequency of the mode as n p,n,m = c êl,

2 L νp,n,m Hn + m + 1L z2 z1
− Btan−1 − tan−1 F=p
c π z0 z0

Using the expressions already derived for z0 , z1 , z2 , it can be shown (with some considerable tedious algebra) that the
expression in square brackets is cos-1 J≤ g1 g2 N. Using this, we obtain an expression for the frequencies of the allowed
modes:
c Hn + m + 1L
νp,n,m = p+ cos−1 J± g1 g2 N (4.2)
2L π
Which sign to use in front of the square root? It turns out that you need the positive sign for all points in the positive quadrant
of the stability diagram, and the negative sign for all points in the negative quadrant.

Special case 1: Planar cavity. A cavity with plane mirrors (R1 = R2 = ¶) has g1 = g2 = 1. So cos-1 g1 g2 = 0. All the
transverse modes with the same p are degenerate with one another, irrespective of n and m. The spacing between the modes
is c ê2 L. As we go away from planar the transverse mode degeneracy is lifted with the higher-order modes extending out
towards the high frequency side.
Laser optics, Mike Tarbutt, 2010 25

Special case 2: Symmetric confocal cavity. This cavity has R1 = R2 = L, which means that g1 = g2 = 0. So

g1 g2 = p ê2. The frequencies of the modes are n p,n,m = Ip + Hn + m + 1LM. All modes having
c 1
cos-1 2L 2

Hn + mL = const. have the same frequency. For example, H p, 0, 0L is degenerate with H p - 1, 0, 2L, H p - 1, 1, 1L,
1
p+ 2

H p - 1, 2, 0L, H p - 2, 0, 4L, H p - 2, 1, 3L etc. Since the expression in the bracket - p + Hn + m + 1L - can either be an integer
1
2
or a half integer, the mode-spacing is c ê 4 L.

Special case 3: Symmetric concentric cavity. This cavity has R1 = R2 = Lê2, which means that g1 = g2 = -1. We need

cos-1 J- g1 g2 N = p. The frequencies of the modes are n p,n,m = H p + n + m + 1L. The expression in brackets is always an
c
2L
integer and so the mode spacing is again c ê2 L. The cavity frequencies are again degenerate but now the degenerate modes
do not have the same values of p. As we go away from the exact concentric configuration the degeneracy is lifted with the
higher-order modes extending out towards the low frequency side.

The diagram below illustrates how the mode shape and the cavity spectrum changes as we go from being nearly planar to
exactly symmetric-confocal to nearly symmetric-concentric. Modes with the same value of p have the same colour. The
index above each mode is the value of n + m. Convince yourself that this diagram is correct.

0 0 0 0 0 0
1 1 1 1 1 1
2 2 2 2 2 2
3 3 3 3 3 3
4 planar
Nearly 4 4 4 4 4

0 1 2 3 4
Frequency

0 0 0 0 0 0
1 1 1 1 1 1
2 2 2 2 2
3 3 3 3 3
Symmetric confocal 4 4 4 4

0 1 2 3 4
Frequency

0 0 0 0 0 0
1 1 1 1 1
2 2 2 2
3 3 3
Nearly symmetric concentric 4 4

0 1 2 3 4
Frequency
Laser optics, Mike Tarbutt, 2010 26

To summarize:

Definition of the g parameter:

L
gi = 1 − (4.3)
Ri

The stability criterion:

0 < g1 g2 < 1 (4.4)

Position of mirrors relative to the waist:

L g2 H1 − g1 L
z1 = − , z2 = z1 + L (4.5)
g1 + g2 − 2 g1 g2

Size of the waist:

λL g1 g2 H1 − g1 g2 L 1ê4
w0 = (4.6)
π H2 g1 g2 − g1 − g2 L2

Resonance frequencies:

c Hn + m + 1L
ν p,n,m = p+ cos−1 ± g1 g2 (4.7)
2L π
Laser optics, Mike Tarbutt, 2010 27

ü Worked example 1: Cavity design

You want to make a 30cm long symmetrical cavity for a 532nm laser, to give a waist size of 100mm. What should the radius
of curvature of the mirrors be?

Solution:
A symmetrical cavity has its waist at the centre. The radius of curvature of the mirrors must match that of the wavefronts. So
the problem is simply that of finding the radius of curvature of a Gaussian beam of waist 100mm at a distance 0.15m from the
waist:

z0 2 z0 2
R = z+ = 0.15 + .
z 0.15
π w0 2
z0 = = 0.059 m
λ
⇒ R = 0.173 m

ü Worked example 2: Another cavity design

Helium-Neon lasers (l = 633 nm) often have cavities that are very nearly hemispherical. One mirror is flat (R1 = ¶) and the
other has a radius of curvature R2 = L + DL, where DL << L.
(a) What are the spot sizes on the two mirrors?
(b) If the laser has L = 30 cm, and a micrometer to adjust DL between 0.1mm and 1mm, by how much can the spot size on
each mirror be changed?
(c) What is the frequency difference between the pth TEM00 mode and the H p - 1Lth TEM11 mode, when L = 30 cm and
DL = 1 mm?

Solution:
(a) Since the first mirror is flat, the waist must be located there. At the other mirror, the radius of curvature of the wavefront
has to be R2 . So,

z0 2
R2 = L + ∆L = L +
L
⇒ z0 2 = L ∆L

On the flat mirror, the waist size is

1ê4
λ λ2 L ∆L
w0 = z0 = .
π π2

On the curved mirror, the spot size is

L2 L
w2 = w0 1+ > w0 .
L ∆L ∆L

The spot on the curved mirror is bigger than that on the flat mirror by the ratio L êDL (which is large).

(b) Putting in the numbers to the above formulae we obtain w0 = 33 mm, w2 = 1.82 mm when DL = 0.1 mm, and
w0 = 59 mm, w2 = 1.02 mm when DL = 1 mm

(c) Using R1 = ¶ and R2 = L + DL we get g1 = 1 and g2 > DLê L. We need to know the value of

cos-1 g1 g2 = cos-1 DL ê L . Since, for small d (do a Taylor expansion around the point p/2),

π
cos − δ > cos Hπ ê 2L − sin Hπ ê 2L H− δL = δ,
2

we see that cos-1 DLê L > p ê2 - DLê L . Using this in the equation for the frequency, we obtain
Laser optics, Mike Tarbutt, 2010 28

c 2 π ∆L c 2 ∆L
νp,0,0 − νp−1,1,1 = 1− − = .
2L π 2 L 2L π L

DL 1
For the case where L
= 300
and L = 30 cm, we get a frequency difference of 18MHz.

ü Worked example 3: How to ensure single transverse mode operation of a laser

Consider the intensity distribution along the x-axis (y = 0) of TEM00 and TEM10 modes. Find the location of the intensity
maximum for the TEM10 mode. At this position, what is the intensity of the TEM00 mode as a fraction of its intensity on
axis? Make a sketch of the intensity distribution of the two modes as a function of x ê w. Suggest a mechanism for ensuring
that a laser can operate only on a single transverse mode.

Solution:
The intensity distribution in any Hermite Gaussian mode has the form

Imn Hx, y, zL = Amn Hm 2 I 2 M Hn 2 I 2 M −2 Hx


w0 2 x y 2 +y2 Lëw2

w2 w w
,

where the Amn is just a constant that ensures the normalization. The intensity distribution of the TEM10 mode along the x-
axis, for any value of z is

8 w0 2 2 ëw2
I10 = A10 x2 −2 x .
w4
To find the position, xm , of maximum intensity we set

I10 8 w0 2 4 xm 3 2 ëw2
= A10 2 xm − −2 xm =0
x w4 w2
i.e. xm = ± w ë 2.

Now consider the intensity of the TEM00 mode at this position...

w
I00 , 0, z = I00 H0, 0, 0L −1 = 0.368 I00 H0, 0, 0L
2
So, the sketch of the intensity of the two modes as a function of x ê w looks like this...
Intensity

0.0 0.5 1.0 1.5 2.0


x/w

A laser can only lase if the round-trip gain is greater than the round trip loss. The TEM00 mode is the smallest mode of the
cavity. If we insert an aperture of a suitable size into the cavity, there will be too much round-trip loss for lasing in every
mode except the TEM00 mode.
Laser optics, Mike Tarbutt, 2010 29

ü 4.2 More cavity considerations

ü Width of transmission peaks, finesse

Obviously, the cavity transmission peaks are not really infinitely narrow. To find out how narrow they really are, we can
follow the usual text-book derivation for the transmission of a Fabry-Perot etalon. The result will apply to the more general
cavities too. The mirrors transmit a fraction t of the field amplitude, and reflect a fraction r.

E0 r6 t2

E0 r4 t2

C E0 r2 t2

B E0 t 2
A
E0

With reference to this diagram, the transmitted electric field is


∞ E0 t2  δ0 E0 T  δ0
Et = E0 t2  δ0 ‚ Ir2  δ M =
n
=
n=0 1 − r2  δ 1 − R  δ

We've introduced the reflection and transmission coefficiencts for the intensity, R = r2 , T = t2 . d0 is the phase shift from A to
B, and d is the phase shift from B to C. We'll not bother with the explicit form of d here since we already know the cavity
resonance condition - Eq. (4.7). The transmitted intensity is It ∂ Et * Et . If the incident intensity is I0 ∂ E0 * E0 , we have

It T2 T2
= = ,
I0 1 + R2 − 2 R cos δ H1 − RL2 + 4 R sin2 Hδ ê 2L

where we used the relation cos d = 1 - 2 sin2 Hdê 2L. If the mirrors absorb a fraction A of the intensity, then by energy conserva-
tion we have R + T + A = 1. So we obtain

It H1 − R − AL 2 1
=B F .
I0 H1 − RL 1+
4R
sin2 Hδ ê 2L
H1−RL2

If the mirror absorption is much smaller than the mirror transmission (A << 1-R), the factor in square brackets will be very
close to 1. Then, on resonance, all the light is transmitted through the cavity. Mirror absorption can be made very small
indeed, though it becomes increasingly difficult to satisfy the A << 1-R condition as the reflectivity increases. In practice
then, it is possible to obtain nearly full transmission of the incident beam, unless the reflectivity is extremely high.

It is usual to rewrite the above formula in the form

It H1 − R − AL 2 1
=B F ,
I0 H1 − RL 1 + I4  2 ë π2 M sin2 Hδ ê 2L

which defines the finesse,  , of the cavity:

π R
= .
1−R
What's the meaning of the finesse? We'll find that it's related to the width of the cavity transmission peaks. Suppose we're on
resonance (for the sake argument, we'll take d=0, though any integer multiple of 2p would do). How much does d have to
change in order that the transmission falls to half of its on-resonance value. For this, the required value of d (call it d1ê2 ) is
given by
Laser optics, Mike Tarbutt, 2010 30

1 1
= .
1 + I4  2 ë π2 M sin2 Hδ1ê2 ê 2L 2
δ1ê2 π
⇒ sin = .
2 2
If we assume that the mirror reflectivity is high so that the peaks are narrow, then d1ê2 will be small and we can use
sinHd1ê2 ê2L > d1ê2 ê2. Then we find the full width at half maximum to be dFWHM = 2 d1ê2 = 2 p ê . The longitudinal mode
spacing (or free spectral range) of the cavity is simply dfsr = 2 p. So we find the following important result which tells us the
meaning of the finesse:
peak separation
= .
FWHM of peaks

ü Power build-up, photon lifetime, quality factor

What is the intensity circulating inside the cavity, Icirc? We can use the above result for the cavity transmission to find this -
the transmitted intensity must be T times the intensity inside the cavity. So we have
Icirc H1 − R − AL 1
= .
I0 H1 − RL2 1+
4R
sin2 Hδ ê 2L
H1−RL2

If the absorption is very small, and the reflectivity very close to unity, the circulating intensity is very much larger than the
incident intensity when the cavity is on resonance. For example, an on-resonance cavity with 99% reflecting mirrors of
negligibe absorption (easily obtained) has Icirc = 100 I0 . The ratio Icirc ê I0 is sometimes called the "build-up factor". State-of-
the-art mirrors may have R = 99.999% and and absorption of about 0.5ppm. In that case the circulating intensity is about
500,000 times the incoming intensity!

How long, on average, does an individual photon spend inside the cavity? When the absorption can be neglected, and the
>  ê p (since
1
reflectivity is very high, the build-up factor is B = 1-R
R is very close to 1). This is also the mean number of
L 
bounces per photon. So the lifetime of a photon in the cavity is t = c p
. The quality factor of a cavity is often used as a
2pc
figure of merit - it is the cavity resonance angular frequency multiplied by this photon lifetime, i.e. Q = l
t = 2 L  êl. In
the optical domain, Q-factors as high as 108 have been achieved. With the use of superconducting mirrors in the microwave
domain, a finesse of 5µ 109 , a Q factor exceeding 1010 and a photon lifetime exceeding 100ms have been obtained.

ü Impedance matching

H1-R-AL T „Icirc
As shown above, when the cavity is on resonance the circulating intensity is Icirc = I0 = I0 . By setting = 0,
H1-RL2 HT+AL2 „T

it's easy to show that the maximum circulating intensity in the cavity is obtained when T = A, i.e. when the transmission of
the input mirror is equal to the total losses in the system. This is called the impedance matching condition. The graph below
shows an example of this - it plots the build-up factor versus the transmission of the input mirror when the absorption loss is
0.2%. The circulating intensity peaks when the transmission is also 0.2%.

Build-up factor vs input-coupler transmission when A=0.2%


120

100
Build-up factor

80

60

40

20

0
0.0 0.2 0.4 0.6 0.8 1.0
Transmission H%L
Laser optics, Mike Tarbutt, 2010 31

We've looked at the case of a two mirror cavity where all the losses are in the mirrors, but the condition we found is actually
a general one. For maximum intracavity power, the transmission of the input coupler should equal the total losses in the
cavity, wherever these losses come from (e.g. reflection and absorption loss in a doubling crystal).

ü Choosing a cavity for a laser

Of course, the choice depends very much on the gain medium and the intended application of the laser. It's usually wise to
avoid cavities where different transverse modes are degenerate since the lasing mode may then be a sum of different trans-
verse modes and the output may be unpredictable. A confocal resonator, for example, would not usually be a good choice.
The near-hempispherical resonator discussed in the worked example above is often a useful design. He-Ne lasers usually
have cavities of this kind. As shown in the example, the spot size at the curved mirror can easily be made to fit the gain
medium by making small adjustments to the cavity length.

ü The confocal resonator and the optical spectrum analyzer

How do you find out what your laser is doing? Is it lasing on a single longitudinal mode of the laser cavity, or on many modes
simultaneously? What is the spread of frequencies in the laser light? To measure what the laser is doing, you need to use an
optical spectrum analyzer. Its function is simple - it tells you the frequency spectrum of your laser light.
At the heart of the spectrum analyzer is a cavity. One of the cavity mirrors is mounted on a piezo so that the cavity length can
be scanned. As the cavity is scanned the various frequencies present in the incident light come into resonance with the cavity.
So, by measuring the cavity transmission as a function of the cavity length you can obtain the frequency spectrum of the
light.

If the various transverse modes of the cavity are not degenerate, this instrument will not work well. Unless the incident light
is mode matched to the cavity (see below - difficult and tedious to ensure) the various transverse modes of the cavity will be
excited, and since their resonance frequencies are all different you will see multiple cavity resonances even if the incident
light is perfectly monochromatic. In other words, you'll end up measuring the resonance spectrum of the cavity rather than
the spectrum of the incident light. That's not what you want. The solution is to use a cavity where the transverse modes are
degenerate. A plane mirror cavity is one possibility, but it is very diffficult to align - the slightest misalignment results in an
unstable cavity that has no resonant modes at all. Much better is the confocal cavity - it is degenerate and easy to align. A
misaligned confocal cavity is stable, and its spectrum is almost identical to the perfectly-aligned case (try to prove these
statements). For these reasons, optical spectrum analyzers almost always use confocal cavities.

ü Mode matching

If the incident beam is not "matched" to the cavity mode, as in the illustration below, most of the light will be reflected (even
when the cavity is "on resonance").

The solution is to "mode match" the incident beam to the cavity mode, using some optics. Once you know the properties of
the mode you're trying to match to (i.e. the location and the size of the waist as given by equations (4.5) and (4.6)), you can
use the ABCD matrix formalism for gaussian beams to calculate what lens you need and where you should put it for proper
mode matching. In doing this, there's a complication that the left hand mirror of the cavity also acts like a lens and needs to
be included in the analysis. However, it often happens that this mirror is flat or has a very large radius of curvature, in which
case it can be neglected in the analysis.
Laser optics, Mike Tarbutt, 2010 32

ü Cavities within cavities

Think about a typical kind of laser with a gain medium in a cavity whose length is, say, 0.5m. The longitudinal mode spacing
of the cavity is 300 MHz. That part of the gain curve that lies above the loss line is often very much larger than this. In this
case the laser may be able to lase on multiple longitudinal modes of the cavity. That's usually undesirable. A good way to
ensure single-mode operation is to put a small cavity inside the big cavity! The idea is sketched below - a thin etalon is
placed inside the much larger laser cavity.

The figure below shows how this works. There are 11 longitudinal modes of the cavity within the region of the gain curve
that lies above the loss line. Without the etalon, the laser could operate on any (or all) of these. The transmission of the etalon
is sketched in blue - the mode spacing is much larger because it is thin. The etalon makes the system too lossy for all cavity
modes except one so that the laser is forced to operate on just this mode. Note that the etalon is usually slightly tilted as
shown in the figure to avoid problems with back reflections off the etalon. There is usually a feedback loop to ensure that the
peak of the etalon transmission coincides exactly with one of the cavity resonances.

Gain curve

This mode lases Etalon modes

Cavity modes

Loss line

Frequency
-4 -2 0 2 4

Without the etalon, the laser could operate multimode, though it might not. If the gain medium is homogenously broadened
the gain will be uniformly saturated down to the loss line and only one mode will oscillate. When the gain medium is inho-
mogenoeusly broadened, gain saturation is local to each mode and multi-mode oscillation is possible. In fact, even with
homogeneous broadening, multimode oscillation is possible due to spatial hole burning (except for a ring cavity). These
complications are an aside - the important point is that the etalon can be used to ensure single-mode oscillation in a laser
where multimode oscillation might otherwise occur.
Laser optics, Mike Tarbutt, 2010 33

ü 4.3 High-loss cavities

So far we've concentrated on low-loss modes of optical cavities - i.e. cavities satisfying the stability criterion. These modes
have low loss because the size of the mode on the mirrors is a lot smaller than the size of the mirrors themselves. So, hardly
any of the light leaks out around the edges of the mirror. Whatever light leaks out of the cavity does so through one of the
mirrors (in a laser cavity one of the mirrors - the output coupler - is less reflective than the others and so almost all the light
leaks out of this mirror to form the laser beam). These low-loss modes are essential for lasers where the round-trip gain is
small since the laser will not work unless the round-trip loss is smaller than the round-trip gain. However, in many lasers, the
round-trip gain is large and then the requirement for a low-loss mode is less important. In this case, it may be more suitable
to use an "unstable" cavity, since these offer several advantages:

- For stable cavities the mode tends to be very small - as shown by Eq. (4.42) above, the mode size is of order l L -
typically about 1mm. Thus, the mode is usally a lot smaller than the volume of the gain medium and can only make use of a
small fraction of the available gain. The mode volume in an unstable cavity can be made much larger, and so can make use
of the entire gain available.

- The output beam of an unstable laser resonator tends to have a large size, and so a highly-collimated beam is easily formed.

- The laser beam is coupled out around the edges of the mirrors rather than through the mirror. This makes it easy to adjust
the fraction of power coupled out (which you cannot do in a stable resonator without replacing the mirror). Also, the back of
the mirror can be used for something else (e.g. cooling water).

- Unstable resonators have very good discrimination against higher-order transverse modes.

We'll follow the treatment of unstable resonators originally used by Siegman, and refer to the diagram shown below. The two
mirrors that form the cavity have radius of curvature R1 and R2 , and are of size (radius) a1 and a2 . Consider the wave leaving
the left-hand mirror - indicated by the wavefronts shown in red - we suppose it to be a spherical wave with virtual centre at
P1 . At the left hand mirror the radius of curvature of the wavefront is p L (not necessarily equal to R1 ). The wave travels to
the right hand mirror where some portion is reflected (the rest carries on and forms the output beam). Upon reflection off the
right hand mirror, the new radius of curvature is q L so that the returning wave is a spherical wave with virtual centre at P2 .
This wave (shown in blue) then travels back to the left-hand mirror, where some portion is again reflected. In order to obtain
a self-consistent solution, the relected wave should again have its centre at P1 . This self-consistent approach leads to a
unique determination of the positions P1 , P2 .

a1 a2
P1 P2

pL L qL
Laser optics, Mike Tarbutt, 2010 34

The self-consistency condition requires that P1 and P2 be images of one another in the respective mirrors. From geometrical
optics this requires
1 1 2 1 1 2
− + =− , and − = .
pL+L qL R2 L+q L pL R1

Using the g parameters, these can be rewritten as


1 1 1 1
− = 2 Hg2 − 1L, and − = 2 Hg1 − 1L.
q p+1 p q+1

Solving for p and q, we get

± g1 g2 Hg1 g2 − 1L − g1 g2 + g2 ± g1 g2 Hg1 g2 − 1L − g1 g2 + g1
p= , and q= . (4.8)
2 g1 g2 − g1 − g2 2 g1 g2 − g1 − g2

(It turns out that only the positive sign in front of the square-root gives self-consistent solutions that are stable with respect to
small perturbations of the points P1 and P2 ).

Now let's work out what fraction of the power is coupled out around the edges of the mirror. For a laser, we usually want all
the light to come out in one direction. We can do this by making sure that one of the mirrors (say the left-hand one) is larger
than the size of the beam. So the geometrical optics picture looks like this:

2M a2

P2
2a2
2

Mirror 2 is smaller than the beam incident on it, so the beam reflected from mirror 2 is of size 2 a2 . This beam expands as it
heads towards mirror 1 where it is all reflected and continues to expand. When the beam has completed a round trip, it's new
size is 2 M a2 where M is the round-trip magnification. The value of M is an important property of the unstable resonator.
Let's calculate it. With reference to the above two figures, we see that the size of the beam on mirror 1 is 2 a2 Hq + 1L ê q.
Similarly the size of the beam in the plane of mirror 2 is equal to the size on mirror 1 times H p + 1L ê p. i.e.
q+1 p+1 Hp + 1L Hq + 1L
2 M a2 = 2 a2 . ⇒ M= .
q p pq

Using Eq. (4.8) for p and q, this can (after a bit of algebra) be written as

M = 2 g1 g2 − 1 + 2 g1 g2 Hg1 g2 − 1L . (4.9)

Note that the magnification depends only in the two g-parameters, and not on the sizes of the mirrors.

If we now assume that the distribution of intensity in the beam is uniform, we get an expression for the fraction of power
coupled out of the cavity in each round trip:

π a2 1
Fraction out = 1 − Fraction hitting mirror 2 = 1 − = 1− . (4.10)
π HM aL2 M2

Notice that this result does not depend on the size of the mirrors. This seems very surprising at first since you may well
expect that the losses would be larger for smaller mirrors. To understand the counter-intuitive result think about making the
right-hand mirror progressively smaller. This reduces the cone-angle of the waves travelling in both directions so that the
relative transverse sizes stay the same.
Laser optics, Mike Tarbutt, 2010 35

Our assumption of uniform illumination can only be justified in the light of more sophisticated numerical calculations of the
cavity modes which shows that the mode with the lowest loss does have an approximately uniform intensity distribution over
the mode volume.

ü Confocal resonators

An important class of unstable resonators are the confocal resonators. Confocal means that the focal points of the two cavity
mirrors coincide. Since the focal length of a curved mirror is R ê 2, the confocal condition is HR1 + R2 L = L. We looked at
1
2
the symmetric confocal resonator before - it lies on the boundary between stable and unstable resonators. But there is an
entire family of non-symmetric confocal cavities. In terms of the g-parameters, the confocal condition translates into
H2 g1 - 1L H2 g2 - 1L = 1. In the g1 g2 -plane, this equation is shown by the dashed lines in the figure below. Referring to Eq.
(4.8), and recognizing that the confocal condition can be written as 2 g1 g2 - g1 - g2 = 0, we see that the output of a confocal
resonator is a plane wave.

2 Confocal
Negative-branch confocal
H+ve branchL

1
F
g2

Positive-branch confocal
-1

Confocal
H-ve branchL
F
-2

-2 -1 0 1 2
g1

The family of confocal resonators separates into two branches, depending on whether the common focal point is inside or
outside the cavity. The negative branch has the focal point inside, while the positive branch has it outside, as shown in the
sketches above. The sketches show single-ended confocal resonators with the expected plane wave outputs (it's easy to see
from the ray diagram, as well as from the arguments given above, why the outputs are plane waves). From Eq. (4.9) , and the
above confocal condition, it's straightforward to show that the magnification of a confocal resonator is M = R2 ê R1 (for
R2 > R1 ). Note that the positive-branch makes a more useful laser resonator because the beam does not come to a focus
inside. The output-coupling fraction for a positive-branch confocal resonator is plotted below.

Output coupling of a positive-branch confocal resonator


1.0

0.8
Fraction coupled out

0.6

0.4

0.2

0.0
0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0
g1

You might also like