Lecture notes
9710
Intro
 Office hours: Monday, 45pm, Shapiro CEPSR 810
 TA office hours: Fri 23pm
 40% ps’s
 25% midterm: Thu 102110
 35% final: 121610
 Exams are closed book, open notes
 Assignments due 1 week after assigned…1112 assignments in total
Lecture 1
 Ray optics
 Principles
 Rectilinear propagation
 Law of reflection
 Snell’s law
 Unity in Fermat’s principle
 Photonic components
_______________________________________________________________________________________________________
Photonics overview
 Lasers are essential to photonics in their control of the properties of light
 Crucial that light can be described by photons
 Electrooptics: light goes into electric field, which modifies the light’s polarization
 Nanooptics: devices that are smaller in size than wavelength of light, used to manipulate light
 Quantum optics (QED, quantum electrodynamics): what does it cover that’s not in classical optics?
 Quantum nature of light
 particle and wave nature of probability distributions
 entanglement  fundamental aspects of quantum theory: principle of nonlocality
Optics overview
 Scope of optics (in chronological order), and increasing in scope
 Ray (geometric) optics
 Wave optics (polarization not included)…in this light properties is described by the scalar wave equation
 Electromagnetic optics: Maxwell’s equations (polarization important)
 Quantum optics: quantized EM fields
 This describes everything previously too (highest scope)
Ray optics
 Valid when wavelength is small compared to dimensions of optical devices, and also when photon energy is small relative
to energy sensitivity of measurements
 Postulates (hierarchy?):
 Light is described by lines and density of lines
 Speed of light is v =
 Important: optical pathlength: λ = nd…formally:
 Fermat’s principle: path taken by light ray is that which minimizes the time it takes to travel that path
 Physics: want constructive interference along only one path
 Reflection and refraction
 There’s always some reflection and refraction
 Ray optics doesn’t tell us the proportions
 Refraction law: Snell’s law: n_{i} sinθ_{i} = n_{r} sinθ_{r}
 Proved by minimizing time with repsect to distance and substituting trig (sine) definitions
 Alternative to Fermat’s principle: Huygen’s principle:
 Light considered as a a spreading, circular wave front from current position
 Can also derive Snell’s law
 Evanescent waves carry no energy
 Optical components:
 Mirrors (flat, spherical, elliptical)
 Paraxial approximation: light rays are close to axis of spherical mirror
9910
Intro
 Will add hint for #5 on homework
 Subscribe to Google course calendar for uptodate information on the class
Lecture 2
 Optical components:
 Spherical mirror: imaging
 Prisms
 Lenses
 Graded index optics
 Matrix optics
_______________________________________________________________________________________________________
Review
Matrix optics
 Prescription to trace optical rays through complicated optical systems while always in paraxial approximation
 Ray transfer matrix if the principal object in this
 General prescription to see if optical system is bounded:
91410
Review
 GRIN fiber…a fiber with a parabolically decaying refractive index
Wave equations
91610
Review
 If waveform stays same at different times we have a nondispersive wave v in the wave equation is constant
 To solve any PDE you can perhaps plug in plane wave form: u = u_{0}e^{i(ωtkt)}…gets you an algebraic equation after
plugging in trial form
 For wave equation; ω = vk; for Schrodinger equation: ℏω = + V
 Group velocity for Schrod does depend on k (dispersive then I believe), but not so for wave equation
 Optical waves described by u(r,t)
 Spherical waves (also) solve the Hemholtz equation
 All phases (in a wavefront) are equal to multiples of 2π: ik ⋅ r = 2πq
New material
 He’s plotting the real part (of a Gaussian?) in the slide
 Wavelength is much smaller than decrease of envelope in the slowlyvarying envelope approximation
 Phasefront = wavefront
 Now plug general solution into Hemholtz equation and then apply slowlyvarying approx
 Paraxial = slowly varying envelope…this is a typical situation
 ∇^{2} = ∇_{T}^{2} +
 Now we have diffractionlimited spot size
Interference
 Interference only works if two sources are phaselocked?…I think this probably means they have the same frequency…it’s
the difference in phase that causes interference
92110
Prof’s advice
 Average on HW1 was 88%
 In for instance focusing onto a lense, the image is independent of the incident angle, so you can solve the matrix
multiplication more quickly…shortcuts
Review
 Gaussian wave comes directly from paraxial approx to Hemholtz eqn
 visibility = …maximum visiblity you can get is one
 Formula for constructive interference in m slits: mλ = sinθd
New material
Waves, ETALON
 Finesse basically tells you how many times the field bounces
 The “sharper” the peaks, the high the finesse
Polychromatic waves
 Wave packets I think
 Constructive interference with all different frequencies
Beam optics
 Gaussian waveform maximizes energy transfer between two points
 Parabolic solution blows up (cuz of ) at origin, which is unphysical and can’t be our solution
 Argues that shifting z by complex constant ξ, and as long as it has complex component denominator does not go to
zero
92310
Came in a couple minutes late…things written on the board…see his slides…I think this is a calculation of Etalon properties.
This is something like a filter. (Etalons and cavities). It increases intensity inside the cavity by a lot…constructive interference I
think
 These are “resonators” I think.
 If you expand intensity about small θ, you see the line width is proportional to , just as before it was .
 Cavity amplifies electric field by factor F.
 Effectively in the cavity we’re slowing light down, so it must be getting amplified?
Now polychromatic waves looking at same cavity phenomena
_______________________________________________________________________________________________________
 Gaussian modes are used everywhere (applications)
 Gaussian modes remove divergence issue
 Depth of focus: what distance in z is my minimum width (approximately) maintained?
 Minimum radius ends at z_{0}
 Two measurements are needed to fully characterize the beam (origin and z_{0} or W_{0})
 Standard measure of beam quality: M^{2} factor
 For a Gaussian the product is always
Transmission of Gaussian beams (optical components)
 Thin lense adds an extra spherical phase dependence
 Divergence and lense? have similar effects…Gaussian beam in and Gaussian beam out!
 Convenient to define magnification factor
 Spherical mirror very similar to effect of thin lense
 Problem with Etalons…light not really confined laterally, so cavities are actually concave mirrors not plane ones…R_{1} =
R?? (p91 of text, bottom right image)
 Stopped at Hermite Gaussian modes, higher order modes, the last solutions to Helmholtz equation
92810
Homework 3 problem 4 notes
 See notebook notes
 Finesse is just like a grating but with more bounces.
 Can solve for the 3 db point, but no need to be too exact.
_______________________________________________________________________________________________________
 You really need 3 parameters to characterize a Gaussian beam…one of his slides should say 3 not 2, in the transmission
through a lense review slide
 Gaussian waves can be focused to a minimum spot size, and have a minimum divergence for a given something (width
or radius of curvature?)?…so perhaps at the moon a Gaussian beam area is the smallest
93010
Came in ~30 min late…
 r is reflectivity I think…look into the appropriate equation for finesse in terms of r or R
 Finesse refers directly to number of round trips in a resonator…Q refers to the time, the response time I think, time
light remains in the cavity perhaps…they are both measures of resonator quality
 Δν_{FWHM} ⋅ Δτ = 1
Spherical mirror resonators
 What does “stability” really refer to?…see my outline on chapter 10
 Symmetrical confocal resonator is actually the most stable resonator I think
 To find resonance modes: after one round trip, phase must equal integer multiple of 2π
 Resonance in Gaussian beams has an extra phase (“extra” over that of plane wave)
 He did first 2/3 of ch 10
10510
 Note B is not the magnetic field but rather the combination:
 H is the magnetic field
 All six fields are interchangeable
 At a boundary:
 Normal component of electric fields are discontinuous by ~ σ
 Normal component of magnetic fields are continuous
 Tangential components of electric fields are continuous
 Tangential components of magnetic fields are discontinuous by ~ K
 I should try to derive these boundary conditions by myself from first principles…done
 Look into possible error in Poynting theorem
10710
 Perhaps the negative sign in the possible error in homework 5 is due to a cross product switching order perhaps
 P = ε_{0}χ_{e}E
 D = εE
 Nondispersive: P changes instantaneously to a change in E
 Isotropy: In general: P_{i} = ε_{0}χ_{ij}E_{j}
Reflection
 There are two possible configurations: B parallel to surface or E parallel to surface
 Because the magnitudes of k are equal, i = i′
 Jackson ch 7 explains the details of the reflection at an interface slide
 The transmission and reflection coeffients are ratios of electric fields
 “Bother Jake”
101210
 Midterm on 10/21
 Review session next week (instead of class I think)
 Photonic crystals is first topic not on exam…so we’re tested on everything up to that
_______________________________________________________________________________________________________
 Linear relationship between P and E only if electromagnetic field is weak…I think so
 I should probably derive the reflection and transmission coefficients myself…?? (for both TE and TM modes)…done
 We should just always make the coefficients positive and just add the phase change ourselves manually I think…correct
 The wave equation is in fact valid if the spatial derivative of ϵ is small (from when we assumed this in chapter 2)
 We’re talking about electron frequencies in the optics of metals section??…no
 E&M wave incident on metal is what we’re talking about…here the “penetration depth” is introduced
 Using the complex Maxwell’s equations for the rest of this lecture I think
 Use real quantities of E, D, B, H for the EM plane wave energy derivation (not resulting formula, which only includes
E_{0})
 Energy density in an EM field can be calculated from only E or only H
 Through today’s class is what we’re responsible for on the midterm
101410
 Problem at hand: describe E&M phenomena in periodic structure
 There are different band gaps for electrons, photons, phonons, etc…perhaps study these overall
 Band gap means light is reflected…so otherwise light is trapped!…depends on where it starts…can it just go through? (of
course…transparence)
101910
 Understand, given the ABCD problem, what is the image and what is the object, and understand the order of things,
why, for the sphere example we can’t just set A = 0.
 FP interferometer extends effective path length by the finesse
 Coherence?
 Understand how LIGO works
 Understand interferometers
 Have definitions listed perhaps!
 A modelocked system is essentially an interference effect
 For a given divergence angle, the Gaussian beam has the smallest width (or minimum width?)
 For a given beam width, the divergence angle is minimized in a Gaussian wave
 For example problem we want to determine = 0
 Check his equations if I use them…there are typos!
 Race through the beginning (truefalselike) questions on the exam
102610
 Two ways to confine light: TIR and DBR (distributed Bragg reflection)
 Periodic dielectric constant…look into this and how it occurs, etc.…got it I think
 Bloch theorem applies to electric field?…yes I believe
 Photonic band gap emits only evanescent waves
 Index contrast is what matters, e.g., silicon/air…if it’s large enough there is a band gap
 “Dielectric index of beads [materials]”??
 CMOS?
 At some point the wavevector is such that the light is no longer confined in the crystal…at this point I think in the
vertical direction the wavevectors are imaginary in order for ω equation to hold…I think (yes I think) this is due to TIR
no longer applying
 To normalize coordinates multiply by a∕π, so you get a normalized frequency a∕λ
 3 confinement mechanisms:
 Everything below light line is confined by TIR
 Band gap shows us confinement in inplane direction
 To do this we can either lower sizes of (one of the) holes in the band gap, which pushes down band structure…this
causes air band to be pushed into band gap
 Or we can enlarge holes locally, pushing dielectric band up
 For a waveguide we do this for a whole line
 Third mechanism?
 FDTD just keep iterating through Maxwell’s equations solving for the fields successively in time
 Idea is to excite all modes at once and all but resonant ones will decay in time…or just take FT and you won’t have to wait in
time!…then you have the eigenmodes of the structure
 Advantage of FDTD is that one simulation covers a very broad spectrum
 TIR is always associated with evanescent field in transmitted side
 For m large enough, BC is that evanescent field goes immediately to zero…this is the whispering gallery mode
approximation
102810
 TIR in the vertical direction, DBR otherwise in planar photonic crystals
 Q = = 10^{6}…solve for Δω then the lifetime (ω ≈ 10^{15})
 The higher the index contrast, the broader the band gap
 Spacing inside a cube or something: k_{n} = n, so energy spacing between modes is ΔE = cℏ
 Zeropoint energy is responsible for spontaneous emission
 Think of zeropoint energy as oscillating electric field
11410
 Whenever we have binomial distrib with no memory, we must derive Poisson statistics
 If there’s memory, I think the observations are no longer independent, which results in noncoherence (to some
degree), from which I believe the Boltzmann distribution results??
 Coherent vs. thermal fit? field? fsomething (types of randomness?)
11910
 Amplitude noise vs. phase noise, good for quantum communication I think
 Most interaction between light and matter occurs with electrons
 GaAs has a direct band gap, that’s why it’s used to much
 By making the quantum well smaller (via quantum confinement methods), we change the energy levels and can tune
where the transition will occur
 Understand this dipole interaction with light…accelerating dipole emits photons
 Couple ways to look at absorption
 z = z(τ,t), mess with partial derivatives to do transformation into photon’s frame of reference (where there’s a retarded
time), and we end up with a simplifying equation to specify the process
 The slowlyvarying approx has to do with a radiating dipole slowly decaying in its radiation (but staying at the same
frequency)
 If incident light is not exactly at resonance but rather off by Δ, multiply polarization (I think) by
 Out of phase radiation because of the i times i…there are two π∕2 outofphases occurring (I belive this is for a dipole
emitting radiation…the dipole and photon are out of phase)
111610
 Interesting note: particle in a box can be applied to band gaps and stuff…electrons stuck in lower band or something
 ℏω( + n)…the 1/2 is spontaneous, the n is stimulated emission
 Δω = ω_{01}…sometimes he write ω_{10}
 Classical spectrum blackbody radiation gives us the uv energy catastrophe I think.
112310
113010
 Gain equation is only valid in the nonsaturation regime
 No population inversion possible in twolevel system…can only get N = 0 from negative start
 Gain will saturate when input power is too high
 Gain drops to zero with increasing distance (along z)
 The absorption of saturable abosrbers saturates and absorbs no more as input is increased
 Average speed comes from 3∕2kT = 1∕2mv^{2}
12710
 Gain saturates with distance
 For inhomogeneous broadening the gain is different for each of the populations
 Q = τω
 loss, gain, reflectivity given, you can determine transmission for the laser
 N_{T} ∝
 Qswitching makes super strong pulses
 Now rate equations are observing photon number inside cavity (per unit volume): n
 n_{s} = τ_{p}, where τ_{p} is the length of time they’re in the cavity, τ_{s} is stimulated emission
 τ_{p} is related to Q value
12910
 Know Gaussian beams
 Divergence of Poynting vector gives us the flow of energy into/out of a volume
 Isotropic media means polarizability tensor is diagonal only?
 Spherical waves won’t be on exam
 Dipole radiation is like a donut
 Don’t worry about derivation of periodic coupling of Bloch equations…
 A particular k point describes the real space direction of the wave
 Boundary in the 2D photonic crystals slide is that between TIR and DBR
 Complex k vector I believe describes decaying wave (evanescent?)
 TIR leads to an evanescent field
 Understand WGM mode of 16,2, conceptually
 Effective index: n_{eff} =
 Vibrational transitions are of much lower energy than electronic transitions
 S ∝
 Coherence has to do with phase
 Be familiar with the rate equations!!!…study orders of magnitude times τ??
 Know population inversion.…it starts to drop if there’s saturation.
_______________________________________________________________________________________________________
Ray and matrix optics (chapter 1)
Intro
 Light = electromagnetic optics
 “Wave optics” is a scalar treatment describing only the wavevector, not the E or B fields
 Ray optics = wave optics in which wavelength of light is infinitesimally small…like photons
 Light follows geometrical rules
Postulates of ray optics
Propagation in a homogeneous medium
In this the path of minimum time is also the path of minimum distance (the latter of which is Hero’s principle), since c is
constant.
Reflection from a mirror
Law of reflection:
 The reflected ray lies in the plane of incidence
 Angle of reflection equals the angle of incidence
Reflection and refraction at the boundary between two media
 At the boundary between two media of refractive indexes n_{1} and n_{2}, an incident ray is split into a reflected ray and a
refracted (or transmitted) ray.
 The reflected ray obeys the law of reflection.
 The refracted ray obeys the law of refraction:
 The refracted ray lies in the plane of incidence
 Snell’s law:
 Ray optics does not describe the proportions reflected and refracted
Simple optical components
Mirrors
 Planar mirror: causes rays to appear to come from an image, what light rays appear to come from upon reflection
 Paraboloidal mirror:
 All incident light rays parallel to its axis are reflected to its focus
 Focal length: distance between “minimum” point and the focus
 Elliptical mirror: all rays emitted from one focus are “imaged” onto the other focus
 From Hero’s principle, all these total distances along any of the paths are equal
 Spherical mirror: no cool properties, but parallel rays close to the axis approximately reflect to the axis at the same
point
 The focal length F = (the radius of curvature R is correspondingly positive for convex mirrors)
 Paraxial approximation: Consider only paraxial rays, which are rays that make small angles with the axis, i.e.,
sinθ ≈ θ, i.e., θ_{max} ≪ 1
 In this approximation, the spherical mirror has approximately both the properties of the paraboloidal and
elliptical mirrors (I don’t see the elliptical mirror similarities)
 This approximation defines paraxial optics (also: firstorder optics; Gaussian optics)
 Imaging equation for paraxial rays (both incident and reflected rays are paraxial):
 In a spherical mirror, rays from the same point (y_{1},z_{1}) meet at the same point (y_{2},z_{2}), so that essentially we have an
object plane z = z_{1} and an image plane z = z_{2}; the mirror serves to magnify the object by . (Negative magnification
means that the image is inverted.)
Planar boundaries
 Prism deflection angle:
 When α and θ are very small, this becomes
Spherical boundaries and lenses
 In paraxial approximation at a spherical boundary:
 For lenses, if we further assume a thin lense:
Light guides
 For an optical fiber:
 θ_{a} is called the acceptance angle
Gradedindex (GRIN) optics
 Since here n = n(r), we use Fermat’s principle δ ∫
_{A}^{B}n(r)ds = 0, which results in the ray equation:
 If we’re in the paraxial approximation so that ds ≈ dz, the ray equation becomes:
 A gradedindex fiber is a glass cylinder with a refractive index that varies as a function of the radial distance from its
axis.
 Numerical aperture of it: NA ≡ sinθ_{a} ≈ n_{0}aα, where a is the fiber radius and θ_{a} is the maximum acceptance
angle for which the ray trajectory is confined within the fiber
 The scalar function S(r), called the eikonal, is like electric potential V …its gradient ∇S describes the ray directions, just like the
gradient ∇V describes the electric field directions.
Matrix optics
Matrices of simple optical components
Very cool fact from homework solutions (and that he mentioned in class)
Wave optics (chapter 2)
Introduction
 Optical wavelengths:
 Infrared: 760 to 300,000 nm
 Visible: 390 to 760 nm
 UV: 10 to 390 nm
 In our daily lives, light to us is like a particle (a ray…the limit of zero wavelength), since the wavelength of light is much smaller
than the scale of objects we normally encounter. It’s when we look closely using more precise (smaller) instruments that the ray
limit breaks down and we thus observe the wave nature of light.
 New phenomena: diffraction (opticalwave transmission through apertures), interference
 Light is described by a scalar function, called the wavefunction, which obeys the wave equation.
Postulates of wave optics
Monochromatic waves
Solutions to Hemholtz equation
Simple optical components
 Reflection of wavevectors of plane waves behaves like that of optical rays.
 Reflection from a planar mirror:
 BC’s: phases of the waves are equal at interface
 k_{1} ⋅ r = k_{2} ⋅ r
 The total wavefunction must satisfy the Hemholtz equation, not just the w.f. of each part
 Transmittances:
 of a transparent plate:
 of a variablethickness plate:
 of a thin lens:
 of a gradedindex thin plate:
 Transmittances multiply.
 A diffraction grating periodically modulates the phase or amplitude of an incident wave.
 To have an effect on incident waves, you can vary either the refractive index or the thickness (or both)
Interference
 Interference equation:
 Interference of two plane waves propagating at an angle θ to each other:
 Coherent light: light with constant phases (no random fluctuations in phase). Here we limit ourselves to the study of
coherent light.
 An interferometer is an optical instrument that splits a wave into two waves using a beamsplitter, delays them by
unequal distances, redirects them using mirrors, recombines them using another (or the same) beamsplitter, and detects
the intensity of their superposition.
 Energy conservation in an interferometer requires that the phases of the waves reflected and transmitted at a
beamsplitter differ by .
 Interference of M waves of the same frequency and amplitude each separated by a phase difference of ϕ:
 For the superposition of waves of constant phase differences and decreasing amplitudes, we have
where F is the finesse.
 The point of all this seems to be that we can filter out different frequencies of light primarily by manipulating the
distance traveled between the plantes.
Polychromatic and pulsed light
 Waves whose time dependence is arbitrary are polychromatic.
 Nothing special, just expand them using monochromatic waves and use superposition and things we already know.
 A quasimonochromatic wave has a range Δν of frequencies about ν_{0}, where Δν ≪ ν_{0}.
 The optical intensity of a quasimonochromatic wave:
Pulsed plane wave
_______________________________________________________________________________________________________
 If you have M waves of equal intensity and phase and constant frequency difference, in their superposition their
frequency differences are similar to constant phase differences, so we now have pulses (in intensity) of light in time,
whereas before the pulses were in phase difference.
 The FabryPerot interferometer is also called an “etalon,”…this is the device that consists of two parallel plane
mirrors between which light bounces, resulting in constant phase differences and decreasing amplitudes. The intensity
(interference) pattern is described by the finesse F.
Beam optics (chapter 3)
The Gaussian beam
 A paraxial wave is a plane wave traveling along the z direction, modulated by a complex envelope that is a slowly
varying function of position…here is the complex amplitude:
 The amplitude must satisfy the Helmholtz equation, so the complex envelope must satisfy the paraxial Helmholtz
equation.
 A solution to the paraxial Helmholtz equation is a parabaloidal wave, which via a transformation can be expressed as
a Gaussian wave (of a single wavelength), whose complex amplitude is
where
The two parameters, in addition to the wavelength λ, are A_{0} and z_{0} (Rayleigh range), which are determined from the
boundary conditions.
 Intensity:where P is the total optical power carried by the beam.
 The Gaussian beam decays in all spatial directions: radially like a Gaussian curve, axially similar to an inverse square
dropoff.
 The beam width is defined as the radius at which the beam intensity decreases by , which might be twice a standard deviation
σ.
 The beam waist is where the smallest beam width occurs, i.e., at z = 0, where W = W_{0}.
 The spot size is defined as the waist diameter, 2W_{0}.
 Divergence angle of cone of propagating Gaussian beam: 2θ_{0} = . Note the tradeoff between wavelength and waist
size.
 In any crosssection of this cone, there is a constant amount of beam power.
 Depthoffocus (or confocal parameter), total length of “cone” whose crosssectional area is within twice that at the z = 0
plane:
 A measure of the quality of an optical beam is its deviation of its profile from Gaussian form.
which is always ≥ 1.
Transmission through optical components
 As long as the component is cirularly symmetric and aligned with the beam axis, the result is itself a Gaussian beam,
though the waist and curvature are altered.
 Transmission through a thin lense:
 Beam radius is positive if the beam is diverging, and negative if converging
 Beam parameters through components depending on the magnification factor M:
 Reflection from a spherical mirror:
Transmission through an arbitrary optical system
 ABCD law:
 This is a general rule for paraxial Gaussian beams through an optical system
 Use previous M matrices of chapter 1’s ray optics
 In free space, q = z + jz_{0}
 The ABCD law applies even to inhomogeneous media, since they can be thought of as series of optical components
HermiteGaussian beams
 HG beams are other solutions of the paraxial Helmholtz equation
 They have nonGaussian intensity distributions, but share the wavefronts (and angular divergence) of the Gaussian
beam, whose wavefronts are nice because they can match the curvatures of spherical mirrors of large radii (think optical
resonators)
 HG beams have the same phase as long as the excess phase Z(z) varies slowly with respect to z
 We obtain the solutions by plugging the above (not explicitly stated here) general form of the solutions (Gaussian times a couple
other functions) into the paraxial Helmholtz equation and then applying separation of variables. The solutions to
the first two resulting ordinary differential equations are the Hermite polynomials; the solution to the last is
 This result shows that Z(z) indeed varies slowly for any z in
 Plugging the solutions back into the general form (that we initially put into the paraxial Helmholtz eqn), we obtain the complex
amplitude of a HermiteGaussian beam of order (l,m):
where
is known as the HermiteGaussian fuction of order l, where H_{l}(u) (where l = 0,1,2,…) is the Hermite polynomial of order
l
 G_{l}(u), where l is even, is an even function, whereas if l is odd, G_{l}(u) is an odd function of u.
 The HG beam of order (0,0) is the simple Gaussian beam.
 Of course, the intensity distributions of the HG beams look like the probability distributions of a particle in a parabolic
well.
Resonator optics (chapter 10)
Math notes
 If a differential equation is linear, the solution satisfying it and the initial conditions is a unique solution.
Intro
 Optical resonator parameters:
 Degree of temporal light confinement: quality factor Q, which is proportional to the storage time of the resonator
in units of optical period
 Degree of spatial light confinement: modal volume V , which is the volume occupied by the confined optical mode
 Resonator uses:
 Confinement and storage of light at resonance frequencies–laser light is generated and built up
 Optical transmission system–filters, or spectrum analyzers
 Pulsed lasers, via their confinement properties
From lecture slides
Planarmirror resonators
How things change when there are losses
 The precise frequencies allowed now have a finite width about those frequencies.
 The plot of I vs. ν goes from equallyseparated delta functions to equallyseparated peaks.
 The general case is when we’re not in a mode (q≠integer) and there are losses, in which the amplitude steadily decreases
upon the reflections. The analysis is then the same as that in chapter 2; here are the results:
Note: r is the magnitude of the roundtrip attenuation factor.
 Sources of loss:
 Imperfect reflection
 Light escaping the mirrors I believe in a vertical direction…this can be expressed as an effective imperfect reflection
though
 So these first two sources can be expressed by reflectances R_{1} = r_{1}^{2} and R_{2} = r_{2}^{2}
 Losses in the medium between the mirrors…this source of loss can be expressed by exp(2α_{s}d), where α_{s} is the loss
coefficient of the medium associated with absorption and scattering
 So the total loss can be expressed as a roundtrip intensity attenuation factor:where
where α_{r} is called the loss coefficient.
 The finesse becomes when R_{1} ≈ R_{2} ≈ 1:
 The loss per unit length is α_{r}, so cα_{r} is the loss per unit time, so we have a characteristic decay time:
known as the resonator lifetime or photon lifetime.
 We thus have an uncertainty relation between time interval and frequency interval:
 Defining the quality factor Q as 2π times the stored energy E over the energy loss per cycle cα_{r}E∕ν, we have:which we see is understood to be the storage time of the resonator in units of the optical period T = 1∕ν. Note the dependence
on frequency…the higher the frequency normally the better the quality.
Sphericalmirror resonators
Gaussian modes
 A Gaussian beam is a mode if its radius of curvature equals the radius of curvature of each mirror, and if the phase
also retraces itself
 Solutions of the above condition (ignoring the phase condition for now):
Everything about the beam is now known, since we’ve determined the beam center at the depth of focus 2z_{0}.
 z_{0} must be real for this all to represent a Gaussian beam; otherwise, it represents a paraboloidal wave, which is an
unconfined solution of the paraxial Helmholtz equation.
 It can be shown that the above satisfies the confinement condition 0 ≤ g_{1}g_{2} ≤ 1.
 For symmetric spherical resonators, if the quantity d∕R is zero, we have plane waves of course as the type of resonant
wave, and of course the resonator is planar. If the quantity is 1, the resonator is confocal, and if the quantity is 2, the
resonator is concentric and we have a spherical wave as the resonant wave.
 Since the wavefronts match the curvature of the mirrors, the phase is the same everywhere at the mirror.
 The resonance frequencies of the Gaussian modes are
where ν_{F} is still c∕2d, and Δζ = ζ(z_{2})  ζ(z_{1})
 Since the wavefronts of HermiteGaussian waves are the same as those of Gaussian waves, they too represent modes of
a sphericalmirror resonator. For them, the resonance frequencies are
 Longitudinal (or axial) HG modes have the same (l,m) but different q; transverse HG modes have the same q (perhaps)
but different (l,m). In other words, the mode number for longitudinal HG waves is q, whereas the mode numbers for
the transverse HG waves are l and m.
Two and threedimensional resonators
Microresonators
 These have one or more spatial dimensions of size on of a few wavelengths of light or smaller.
 Microcavity resonator, or microcavity, is one in all spatial directions.
 Examples of microresonators:
 Micropillar  Bragggrating reflectors  DBR axially and TIR perpendicularly (around)
 Microdisk and microsphere  ligh reflects near the surface in whisperinggallery modes  TIR
 Microspheres:
 Light travels in great circles around the inside
 Formed by surface tension in a molten state, so very high reflectivities and thus very high quality factors
 Microtoroid  resembles small fiber rings formed by surface tension in a molten state  TIR
 2D photonic crystals containing lighttrapping defects that function as microcavities
 The goal of microresonators: (1) small modal volume V , (2) high quality factor Q
 To solve for the modes of dielectric microresonators: use the Helmholtz equation in the appropriate coordinate system, fit
appropriate BCs, for all components of the EM fields
 A FabryPerot resonator confines light via distributed Bragggrating reflectors (DBRs). I believe these just represent materials
whose reflectivity is high.
 Think of DBR as periodic lattices that have bandgaps!
Electromagnetic optics (chapter 5)
Electromagnetic theory of light
 H is the magnetic field
 Maxwell’s equations:
 So, compared to what I’m used to: B → μ_{0}H
 All six components of E and H, which are externally applied, must satisfy the wave equation:
 The speed of light in a medium:
 Derivation of the wave equation (for the electric field) from Maxwell’s equations, in free space:
 Since the wave equation is linear, we can apply superposition to the fields.
 If μ or ϵ are not equal to their free space values, we must be in a medium, so we must also describe the induced fields in the
medium; hence we have something of effective electric and magnetic fields:
 D = ϵ_{0}E + P is the electric flux density, or displacement.
 B = μ_{0}H + μ_{0}M is the magnetic flux density.
 The pattern here is, for fields/(charges and currents), <applied> + <induced> = <effective/free>.
 Basically, “free” current and charge refers to the amount of current and charge effectively being applied, since
“incuded” inherently subtracts from the initial “applied.”
 P is the polarization density, while M is the magnetization density.
 Thus, Maxwell’s equations become, doing what we need to do to get rid of the constants:
 So there’s a difference between “free space” and “sourcefree.”
_______________________________________________________________________________________________________
Boundary conditions between two dielectrics
In general
 Paralleltosurface E is continuous
 Paralleltosurface H is not continuous, by K_{f} ×
 Normaltosurface D is not continuous, by σ_{f}
 Normaltosurface B is continuous
 Extra rule to remember: for parallel components consider the applied fields E and H; for normal components consider
the effective fields D and B. Observing Maxwell’s equations, we relate “parallel” to the curl, and “normal” to the
divergence…it’s “backwards.” The RHSs of Maxwell’s equations are all “effective” and “free.”
No free charges/currents
 Paralleltosurface E is continuous
 Paralleltosurface H is continuous
 Normaltosurface D is continuous
 Normaltosurface B is continuous
_______________________________________________________________________________________________________
 A perfect mirror is a perfect conductor.
 The paralleltosurface component of the electric field of a perfect conductor must be zero, so by the boundary conditions
above the total E&M wave normal to the surface on the incident side must have an electric field equal to zero. Since
<incident> + <reflected> = <transmitted>. Since all of the incident light is reflected, the magnitudes of the incident
and reflected light must be equal. But since the waves must now add to zero, there must be a phase shift of π.
 To go from orig ME’s to ME’s in a dielectric:
 Need equations for D and B
 Need ρ = ρ_{f} + ρ_{b} = ρ_{f} ∇⋅ P
 Need J = J_{f} + J_{b} + J_{p} = J_{f} + J_{b} + = J_{f} + ∇× M +
 To get S&T’s form of curl of electric field equation, just memorize it; can’t find derivation, though haven’t looked
in Jackson: ∇× E =  (for no sources)
_______________________________________________________________________________________________________
Fundamental theorems of vector calculus
 Gradient theorem:
 2D divergence theorem:
 3D divergence theorem (Gauss’s theorem):
 Stokes’ theorem:
_______________________________________________________________________________________________________
Electromagnetic waves in dielectric media
 Nondispersive: instantaneous response of the system to applied field (no time lag)
 Spatially nondispersive: same but with space…no spatial “lag”
Linear, nondispersive, homogeneous, isotropic media
 P = ϵ_{0}χE
 Electric susceptibility: χ
 D = ϵE (true for homogeneous and isotropic media, period [not nec linear])
 Electric permittivity: ϵ = ϵ_{0}(1 + χ)
 Relative permittivity (aka the dielectric constant): = 1 + χ
 B = μH (true for homogeneous and isotropic media, period [not nec linear])
 Magnetic permeability: μ = μ_{0}(1 + χ_{m})
 μ_{0} =
 Maxwell’s equations:
 Identical to Maxwell’s equations in free space, except with the substitutions ϵ_{0} → ϵ and μ_{0} → μ
 For nonmagnetic material, μ = μ_{0} and n = =
Relaxing these conditions (only listed are the changes from the previous state)…
Inhomogeneous media (e.g., GRIN materials)
 χ → χ(r)
 ϵ → ϵ(r)
 μ stays the same
Anisotropic media (but homogeneous!)
 χ → χ_{ij}
 ϵ → ϵ_{ij}
 Ditto for χ_{m} and μ
Dispersive media
 Now there’s a linear differential equation relating P and E with respect to time
 Point I think is that χ → χ(t)
 Can describe system using a linearsystems approach, using impulse response functions, convolutions, transfer functions
(FT of impulse response function)
Nonlinear media
 In these the relation between P and E is nonlinear.
 Wave equation for all homogeneous and isotropic dielectric media, regardless of linearity or dispersivity:
 For nonlinear and nondispersive, this becomes (with P being some nonlinear function Ψ(E)):
 Most dielectric media are approximately linear unless the optical intensity is substantial.
_______________________________________________________________________________________________________
Derivation of particular equations from Maxwell’s equations
Maxwell’s equations in matter (general):
Assuming no currents or charges (sourcefree):
This becomes (sourcefree):Starting with the second equation, we have (sourcefree all the way through):Our grand starting point for the electric field (assuming all is similar for the magnetic field) is (sourcefree):
Dirk said that the permeability is typically constant throughout the material, so I’m pretty sure that implies ∇× M = 0:
 It’s okay to always assume the magnetic properties throughout a material do not change (unless we’re talking about
rather cuttingedge research).
_______________________________________________________________________________________________________
 Useful vector identities:
_______________________________________________________________________________________________________
Monochromatic electromagnetic waves
 Assume waves of the form
 Assuming Maxwell’s equations are satisfied by the full, complex expressionsMaxwell’s equations become
 From now on, we will write the complex amplitudes as regular letters. i.e., (r) → D(r) → D. Until this section when we’re
written D we’ve referred to the real, r and tdependent field. So now, it is complex, and only dependent on r. So now, the true
field (r,t) = ℜ.
 D = ϵ_{0}E + P
 B = μ_{0}H + μ_{0}M
 = ℜ, where
is the complex Poynting vector
 Optical intensity:
 We know the full complex field vectors must obey the wave equation
For waves that have timedependence X(r,t) = X(r)e^{iωt}, this becomes the Helmholtz equation:
where c = , the dispersion relation for light in a common (linear, nondispersive, homogeneous, isotropic)
medium.
 This Helmholtz equation holds in inhomogeneous media if ϵ = ϵ(r) varies slowly compared to the wavelength (just with the
appropriate substitution k → k(r).
 In dispersive media the constants χ, ϵ, k all pick up νdependence (and are generally complex quantities in dispersive
media):
Elementary electromagnetic waves
 Since we’re going to talk about monochromatic waves, the complex amplitudes of the fields satisfy the Helmholtz
equation
 We’re also dealing with common (linear, nondispersive, homogeneous, isotropic) media
The transverse electromagnetic (TEM) plane wave
The spherical wave
 My definitions:
 Derivation from the vector potential:
 Then,and from Maxwell’s equation, the electric field iswhere the approximations are for where r ≫ λ. Further,
 Note that though of course the wavefronts are spherical, the magnitudes vary as sinθ, due to the representation of the spherical
wave as radiation far away from a dipole.
Paraxial rays (points near the zaxis)
 We have
 These imply that
 If z ≫ x, this becomesSince U(r) →e^{ikz}, we approach a TEM plane wave.
The Gaussian beam
 As before, we obtain this by making the substitution z → z + iz_{0} in the paraboloidal wave equations:
Optics of conductive media
Polarization optics (chapter 6)
Intro
 Polarization of an E&M wave is determined by the time course of its electric field
 TEM means transverse electromagnetic, in which the electricfield vectors lie approximately in planes transverse to
propagation
Reflection and refraction
 Again, at a boundary the tangential components of E and H are continuous, as are the normal components of D and
B.
 Impedance η is a characteristic of the medium.
 If the electric field is perpendicular to the plane of incidence, the mode is called the transverse electric (TE) polarization,
the orthogonal polarization, or the s polarization.
 If the magnetic field is perpendicular to the plane of incidence, the mode is called transverse magnetic (TM) polarization,
the parallel polarization, or the p polarization.


 Record more equations from my notebook!!!
TE Fresnel equations
TM Fresnel equations
 Of course, a phase shift of π corresponds to a multiplication by 1 via Euler’s formula: e^{iπ} = 1.
 The phase shift of π upon reflection by a higherindex material is simply a result of the argument (the phase) of the
reflection coefficient being π, and since E′ = rE, this phase shift shifts to the new electric field E′.
 My confusion with the step function change of the phase could possibly be resolved as: Consider a smooth function
that crosses the independentvariable axis. Instead of it doing that, plot the absolute value of it, so we still have
a smooth function, except the phase changes suddenly by π instead of the absolute value going negative. In other
words, we have z = ze^{iϕ}, where perhaps what we really care about is z…??
 In general r and t are ratios of complex amplitudes and are thus themselves complex.
 Brewster angle:
 Power reflectance: R = r^{2}
 Power transmittance: T = 1  R≠t^{2} in general
 Power reflectance at normal incidence:
Photoniccrystal optics (chapter 7)
Intro
 Study various interferometers for the exam
 For monochromatic waves in inhomogeneous materials the Helmholtz equations become generalized:
where η(r) ≡ ϵ_{0}∕ϵ(r).
 While waves in a homogeneous medium are plane waves, modes in a periodic medium are Bloch modes.
 Bloch modes are traveling waves modulated by standing waves.
From lecture slides
 First use “generalized Helmholtz equations” (i.e., “master equations”), which are the Maxwell curl equations, to solve
for E or H, and then use the divergence equations to obtain the other.
 Reciprocal and direct lattice vector relations: b_{i} ⋅ a_{j} = 2πδ_{ij}
 Mode (Bloch mode or Bloch state) of field E: E_{K}(r) = e^{iK⋅r}u_{K}(r), where u_{K}(r) has the periodicity of the lattice:
u_{K}(r + a) = u_{K}(r):
 Plug this into the master equations to obtain the A (amplitudes) and the ω(K) (frequencies = dispersion relations)
 Thus, a general field E can be expressed as an LC of these modes.
 Only a 3D photonic crystal can exhibit a full band gap.
 Index contrast is what matters, e.g., silicon/air…if it’s large enough there is a band gap
 Note that in a region a traveling wave can written as a sum of forward and backwardtraveling waves.
 The “n” terms in periodic lattices usually refer to the offset (i.e., multiply n by the spacing), whereas the actual variable is the
total distance, so that something like z  (n  1)a simply represents position at the point z…it’s such because
this is part of the argument of a periodic function, e.g., sin, and it’s the distance from the
beginning of the period location that matters. You can always evaluate sin37π but you may as well evaluate
sin = sinπ = 0.
Onedimensional photonic crystals
Two and threedimensional photonic crystals
 1D means a 3D crystal with 1D symmetry. It’s not actually simply one dimension. Ditto for 2D and 3D of course.
Photon optics (chapter 12)
Intro
 The optical part of quantum electrodynamics (QED) is called quantum optics.
 Photon optics is a set of rules from QED that permits us to deal with optical phenomena that lie beyond the reach of
classical theory, but cannot provide an explanation for all the effects that can be explained by quantum optics.
From lecture slides
 Uncertainty:
 As a wave travels forward (whether it’s a lefttraveling wave or a righttraveling wave), its phase decreases by the
amplitude of its wavevector times the distance traveled.
The photon
 The spin of a photon is associated with its polarization. This angular momentum of the photon is intrinsic (no reference
point needed).
 Photons in a resonator are like phonons in a crystal, in which properties of the mode are assigned to the quanta.
 For photons, these properties are for example frequency, spatial distribution, direction of propagation, and
polarization.
 A photon of wavelength 1 μm has energy 1.24 eV, so we have
 Energy of a photon is proportional to its frequency and thus inversely proportional to its wavelength:
 Associated with each photon of frequency ν is a wave described by the complex wavefunction U(r)exp of the
mode.
 Photon position; The probability of observing a photon at r within an incremental area dA, at any time, is proportional to the
local optical intensity I(r) ∝^{2}, so that:
 So an optical photon behaves as part wave and part particle.
 Through a beamsplitter a photon must choose the path and cannot go through both. It’s statistical.
 Energy formula to remember:
 So momentum of a photon in a plane wave mode is
 Radiation pressure is just the force (per area) emitted by a photon…think of it in terms of the fact that a change in momentum
must be caused by an impulse (units: force * time = momentum)
 Photon position and time:
 The Fourier transform indeed is what basically implies the quantum uncertainty principles.
 Timeenergy uncertainty, in which σ_{t} is the duration of the photon function I(t):
 So a monochromatic photon has an eternal duration. This still satisfies the uncertainty principle.
 A wavepacket photon (polychromatic…LC of monochromatic waves) passes a point in a finite time σ_{t} and thus has a finite
frequency spread.
 All radiation can be described as a sum of modes (a basis), and if our modes are monochromatic uniform plane waves, we have
various occupations of the different modes.
Photon streams
 The number of photons occupying any mode is generally random.
 Photon streams often contain numerous propagating modes, each carrying a random number of photons.
 Mean photonflux density (analog to optical intensity) is the average number of photons incident per unit area per unit
time:
 Mean photonflux (analog to optical power) is the average number of photons incident on area A per unit time:
where hν is the average photon energy and the optical power P = ∫
_{A}I(r)dA
 Mean photon number (analog to optical energy) is the average number of photons incident on area A in time T:
where the optical energy E = PT
 In general, i.e., for polychromatic light, we subscript any quantity above with frequency so we have spectral density.
E.g., P_{ν}dν represents the optical power in the frequency range ν to ν + dν
 The properties of the light source determine the flucuations in ϕ(r,t).
 Coherent light results from a deterministic power source, as opposed to partially coherent light, which is generated by
a timevarying random power source. If the power source is not completely coherent, fluctuations in photon detection
are not independent, altering the statistics of the process. I believe for our purposes we’re considering coherent light as
resulting from constant optical power P.
 In a time T, the expected number of photons is thus n = , and for coherent light the probability distribution is the
Poisson distribution:
 PT above in general is:
so that in general the Poisson expectation number (of photons) is
 This is the result of the binomial distribution when dividing our interval T into N subintervals such that the probability
of observing a photon in time T∕N is pretty much 0 or 1…then we take the limit as N →∞.
 Mean and variance of a random number:
 The variance of a Poisson distribution is equal to its mean:
 Signal to noise:
 For Poisson distrib, SNR = n
 Laser light follows Poisson
 For an optical resonator whose walls are held at a temperature T in thermal equilibrium, so that the photons are emitted
into the modes (each of energy E_{n}) of the resonator, this is thermal light and it follows the Boltzmann probability
distribution:
 So in thermal equilibrium, the energy associated with each mode is random and follows the Boltzmann distribution
above.
 For photons, in which E_{n} = (n + )hν, we haveCalculating the mean of this distribution we find
Eliminating exp we end up with the BoseEinstein distribution:
 So the emission of photons from atoms in thermal equilibrium follows the BoseEinstein distribution!
 This “thermal” case was not “coherent.”
 SNR for the BoseEinstein distribution is
which is poor due to the randomness of the amplitude and phase of thermal light.
Photons and atoms (chapter 13)
Intro
 If a photon is absorbed by an atom, it’s said to be annihilated; stimulated emission, a photon is said to be created
Energy levels
Occupation of energy levels
 Temperature is the principal determinant of both the average and fluctuations of energylevel occupancy
 Distinguishable particles follow the Boltzmann distribution:
 Indistinguishable fermions (halfinteger spin) follow the FermiDirac distribution:
 Indistiguishable bosons (integer spin), like photons emitted from a blackbody, follow the BoseEinstein distribution:
 Calculating occupancy:
 Normalize probability: ∑
_{i}P(E_{i}) = 1
 (Occupancy of state) = (probability of state being occupied) × (degeneracy g(E_{i}) of state)
 Check: If there are N particles in the system we must have:
 Note: In the degeneracy (density of states) g(E_{i}), don’t forget to include the contribution from the Pauli exclusion
principle (e.g., for electrons there’s a factor of 2 in it)
 Pauli exclusion principle: For two identical fermions, the total wave function is antisymmetric
 Note that if E ≫ μ, the quantum distributions become the Boltzmann distribution…this is usually the case when optical
transitions occur and thus for such transitions we apply the Boltzmann distribution.
Interactions of photons with atoms
 Photons can interact with:
 Electrons in atoms
 Electrons or vibrations or rotations in molecules
 Electrons and holes in semiconductors
 Phonons in a solid
 More?
 These results come from quantum electrodynamics…these are laws that govern photonatom interactions
Spontaneous emission
Absorption
Stimulated emission
 Like absorption, only occurs when the mode already contains a photon
 Is inverse of absorption
 Produces a photon identical in every way to the incident photon
 Law:
_______________________________________________________________________________________________________
 So total probability densities are
 Transition strength or oscillator strength…describes strength of the interaction:
 Normalized transition cross section is the lineshape function:
 Linewidth is Δν, which is the FWHM of g(ν)
 Also, Δν ∝ 1∕g(ν_{0})
 σ_{0} = σ(ν_{0})
Spontaneous emission into all modes
Induced (stimulated) transitions: absorption and stimulated emission
 Now switch from considering an atom in a resonator to a stream of photons (monochromatic of frequency ν and intensity
I) incident on an atom, where this photon flux is described by
 If we construct a cylinder of volume V , length c (the distance light travels in a second), and base area A, parallel to
the stream, at any given time in the cylinder is n = ϕA = ϕ, so
 We thus have
 So ϕ is like the total photon flux (which is basically probability density per unit area), σ is like the effective
area that captures the photons for a transition, and W_{i} is the resulting probability density for the induced
transition.
 Note that whereas spontaneous emission is independent of n(ν) of the mode, absorption and stimulated
emission can be greatly enhanced by the presence of photons n(ν) already in the mode.
 Polychromatic (broadband) light:
 Einstein coefficients:
 Ratio of basically rates of spontaneous to stimulated transitions:
Homogeneous line broadening
This is when all of the atoms of a medium are taken to be identical and to have identical lineshape functions.
Lifetime broadening
Collision broadening
Inhomogeneous line broadening
Thermal light
Laser amplifiers (chapter 14)
Tip
 In a laser amplifier system never forget that ∑
_{i}N_{i} = N_{a}, where N_{i} is the population density of level i and N_{a} is the
total population density of the system.
 Assume N_{a} is a given, a parameter, of the system.
Intro
 A coherent optical amplifier is a device that increases the amplitude of an optical field while maintaining its phase.
 An incoherent optical amplifier increases the intensity of an optical wave without preserving its phase.
 A laser is basically an optical oscillator; LASER refers to the underlying principle for achieving coherent amplification
of light and stands for Light Amplification by Stimulated Emission of Radiation.
 Electronic amplifiers use cavities to select the principal frequency; lasers use such resonators only for auxiliary frequency
tuning and simply rely on the energy levels of the material to select the principal frequency.
 Whereas light transmitted through matter in (thermal) equilibrium is attenuated due to more absorption than
stimulated emission occurring, in a laser instead the higher energy levels are already excited (via pumping) and thus
more stimulated emission occurs…there is instead amplification rather than attenuation.
 The amplifier gain is the increase of amplitude of an input signal.
Theory of laser amplification
Amplifier pumping
Pumping schemes
 Again, the purpose is to increase the population in the higherenergy level and to decrease that in the lowerenergy
level.
 W is the pumping transition probability
 N_{0} and τ_{s} saturate as W increases in both schemes below.
 Fourlevel laser amplifiers typically have larger gain than threelevel amplifiers
Fourlevel system
Threelevel system
Amplifier nonlinearity
 The gain coefficient γ(ν) depends on the population difference N:
so γ = γ(ν,N)
 N depends on the pumping rate R
 e.g., in a fourlevel system, using approximations, we have
so N = N(R)
 N also depends on the transition rate W_{i}:
so N = N(R,W_{i})
 W_{i} depends on the radiation photonflux density ϕ:
so W_{i} = W_{i}(ϕ)
 Thus, γ = γ(ν,R,W_{i}) = γ(ν,R,ϕ). However, R can be written in terms of other variables, so I believe we have γ = γ(ν,ϕ): The
gain coefficient of a laser medium is dependent on the photonflux density that is to be amplified. This is the origin of gain
saturation and laser amplifier nonlinearity:wherewhere again, λ = c∕ν_{0} is the wavelength of the light at the transition frequency ν_{0}. (It would be more clear if λ was written as
λ_{0}.) ϕ_{s}(ν) is known as the saturation photonflux density.
 The point is that ϕ depends on N (photon flux depends on population difference) and N depends on ϕ, so there’s a
nonlinearity.
 Finally,whereγ(ν) is the most general gain coefficient, known as the saturated gain coefficient (for homogeneously broadened media). γ_{0}(ν) is
the smallsignal gain coefficient.
 Again, the point is that the gain coefficient, which is responsible for amplifying ϕ, in turn depends on ϕ.
 ϕ_{s}(ν) represents the photonflux density at which the gain coefficient decreases to half its maximum value γ_{0}(ν).
 Unsaturated gain coefficient γ_{0}(ν) has linewidth Δν, whereas saturated linewidth has linewidth:
 So gain coefficient, as saturation nears (ϕ ~ ϕ_{s}) gets short and fat w.r.t. ν. The fattening corresponds to reduced frequency
selectivity, of course.
 Remember, gain coefficient γ is gain per unit length.
 Gain:
 Small signal approximation is when γ ≈ γ_{0} (not near saturation).
 Solving
when γ depends on z (implying saturation) this time yields
where X ≡ ϕ(0)∕ϕ_{s} and Y ≡ ϕ(d)∕ϕ_{s} are the input and output photonflux densities normalized to the saturation photonflux
density and the gain G = Y∕X.
 Examining limiting cases:
 X and Y both much smaller than 1:
This is the solution we obtained way above for γ ≈ γ_{0}.
 X and Y both much greater than 1:
which implies
This is the situation of heavy saturation in which the atoms of the medium are “busy,” emitting a constant
photonflux density , which is independent of the input ϕ(0).
 If the gain coefficient is negative (N_{0} < 0), the absorption coefficient α = γ is positive, and instead of amplification we have
attenuation, and instead of gain we have transmittance. The equation
still applies, except now γ_{0}d is a negative quantity. As the input transmits through the medium, N → 0 (but might not reach
zero since the medium has a finite size). As the input is increased, eventually N will actually approach 0 in the medium, so that
the transmittance eventually equals 1.
 This relation holds:
 Such a medium is called a saturable absorber.
 So now transmittance = Y∕X.
Saturated gain in inhomogeneously broadened media
 Average smallsignal gain coefficient:
 Saturated gain coefficient (be careful when doing this one!):
where
where
Dopplerbroadened medium
 Here g_{β}(ν) = g(ν  ν_{β}), where
and we have
where
 I believe the subset of atoms β is homogeneously broadened.
 NOTE THAT ν_{β} IS CENTERED AROUND ZERO, SO BE MINDFUL OF THIS WHEN CALCULATING
THE INTEGRAL (THE AVERAGE γ)!!!!
 For normally distributed (σ_{D}) velocities of the atoms and for this distribution being much broader than the gains
γ_{β}(ν):
where
Since the average gain coefficient γ at ν = ν_{0} is and that for a homogeneously broadened (due to saturation) medium
is
we see that with an increase in the photon flux at ν = ν_{0} the gain coefficient increases more slowly in inhomogeneously
broadened media.
 Note that *only here* ϕ is for ν = ν_{0}. ν_{0} is the transition frequency corresponding to the subset of atoms that is
not moving at all.
 I believe homogeneous broadening can and does result from saturation.
 If there is a normally distributed inhomogeneously broadened medium, the average gain curve at an arbitrary frequency γ(ν)
will also be normal, I believe with a standard deviation that’s the same as that of the broadened medium. We just showed that
the gain coefficient for photons of frequency ν_{0} is
so if a monochromatic beam is incident on the medium we can have saturation at only the frequency ν_{1} of the beam, which
means there will be a drop in the average gain curve at ν_{1}. This is called spectral hole burning.
 Note the actual ν dependence of ϕ:
 So if the photon flux were the same for all populations (with difference speeds and different νs), Δν_{s} would be
the same for all populations, but if the flux at ν_{1} is larger than at other frequencies, there will be a hole at ν_{1} of
width Δν_{s}(ν_{1}), so that by increasing the flux ϕ(ν_{1}) the hole will become both deeper and wider.
 Understand when these holes are symmetric in the curve.
Lasers (chapter 15)
Intro
 The output of a laser is fed back to the input with matching phase (multiple of 2π) to do more stimulated emission.
 The laser initially bursts I believe and then the steadystate is the actual observed laser state.
 Steady state occurs because the gain saturates.
 Both the gain and phase shift are functions of frequency.
Theory of laser oscillation
 Coherent = at the same phase
 Phases of a laser:
 Increase the pumping rate R (increasing N_{0} [not N]) until the pumping threshold is exceeded (gain condition)
and thus so is the minimum population difference needed for laser oscillation (or “lasing”). ϕ and I will then go
from not increasing to increasing exponentially.
 No amplifier radiation (no stimulated emission):
 As radiative processes get stronger, N → 0, as the radiative processes balance each other and push the system
this way…independent of the initial sign of N.
 In the presence of amplifier radiation the steadystate population difference becomeswhere
 Gain coefficient (gain per unit length): γ(ν)
 Governs the rate at which the photonflux density ϕ or the optical intensity I = hνϕ increases via
 For small ϕ the (smallsignal) gain coefficient is:
where:
 N_{0} is the equilibrium density difference (the subscript stands for smallsignal, not equilibrium [but equilibrium
is implied])…equilibrium refers to, I believe, steadystate + no stimulated emission
 σ(ν) is the transition cross section
 t_{sp} is the effective spontaneous lifetime (for stimulated emission?)
 g(ν) is the transition lineshape
 λ = λ_{0}∕n is the wavelength in the medium
 ϕ will increase and at some point the amplifier enters a region of nonlinear operation in which it saturates and its gain
decreases.
 Then its population difference becomes, for a homogeneously broadened medium,
where
 ϕ_{s}(ν) = is the saturation photonflux density
 τ_{s}is the saturation time constant, which depends on the decay times of the energy levels involved (for a
fourlevel pump, τ_{s} ≈ t_{sp}; for a threelevel pump, τ_{s} = 2t_{sp})
 Thus, a saturated amplifier has a gain coefficient (for homogeneous broadening):
 The amplifier phase shift per unit length that is due to the laser amplification, when the lineshape is Lorentzian with linewidth
Δν and
is
 This phase shift is in addition to that introduced by the medium hosting the laser atoms.
 Wavenumber (wavevector) is like phase shift per unit length
 α_{r} is overall loss per unit length
 Photon lifetime: τ_{p} = 1∕α_{r}c
 When resonator losses are small and the finesse is thus large
 Gain condition is that the smallsignal gain coefficient be greater than the loss coefficient:
 Thus,
where
is the threshold population difference. This last equation implies that the threshold the lowest at the frequency of the peak
of the lineshape function (ν_{0}).
 Phase condition for coherence is
 Perhaps typically the atomic transition linewidth is much wider than the resonator mode linewidths.
Characteristics of the laser output
Power
 Steady state occurs when N = N_{t} or when
 This condition leads to the steadystate internal photonflux density:
 The middle of the inverted “S” curve occurs when N_{0} = 2N_{t} (seen from the above equation).
 The facts that the steady state N value can never increase above N_{t} and that the steady state gain γ(ν) can never
increase above α_{r} is called gain clamping.
 Note that the photonflux density (steady state) for photons traveling in a single direction is ϕ∕2.
 Thus, the output photon flux density iswhere T is the intensity transmittance of the transmitting mirror.
 The optimization of the transmittance is based off the principle that to increase the output flux you can increase the
transmittance, but this will add to the loss inside the resonator and thus decrease the steady state flux inside the oscillator.
That’s where the tradeoff comes in.
 We can relate ϕ_{O} and T using equations from this subsection to obtain:
 Assuming T ≪ 1 (which is fine as can be seen by plotting this for common parameters and seeing that the maximum occurs at
T ≪ 1) we have ln(1  T) ≈T and minimizing ϕ_{O} with respect to T we find:
Spectral distribution
Homogeneously broadened medium
 Only the mode or two modes (in the symmetric case) closest to ν_{0} will remain in the end, since the entire gain curve
will reduce in size because the most dominant modes will deplete (to a point) the population difference:
 UNLESS there’s spatial hole burning, because the different modes occupy different spatial portions of the active
medium. The bottom line is that spatial hole burning allows another mode, whose peak fields are located near the
energy nulls of the central mode, the opportunity to lase as well.
Inhomogeneously broadened medium
Spectral hole burning in a Dopplerbroadened medium
 The interaction radiation of a group of an atom moving with velocity v toward the direction of propagation interacts
with radiation of frequency ν_{0}(1 + v∕c).
 Since radiation of frequency ν_{q} interacts with populations of atoms ν_{q} = ν_{0}(1 ± v∕c)v = ±(ν_{q}  ν_{0}), symmetric
hole burning occurs for the frequencies ν_{q}  ν_{0} = ±.
 Note that this symmetricity is due to the fact that the photons are in a resonator and thus are bouncing in two different
directions.
 Of course, holes have widths Δν_{s}.
Spatial distribution and polarization
 A laser with two planar mirrors outputs a plane wave propagating along the axis of the resonator.
 A spherical mirror resonator supports HermiteGaussian beams.
 Because of their different spatial distributions, different transverse modes undergo different gains and losses (less so for
longitudinal modes unless there’s spatial hole burning).
 Higherorder modes can generate larger optical power, while the Gaussian mode has the smallest beam diameter and
can be focused to the smallest spot size.
 Of course, each mode (l,m,q) has two polarizations (two independent modes). If everything else is equal, the laser will
oscillate on each of the two polarizations simultaneously, independently, and with the same intensity. The laser output
is then unpolarized.
 Advantages of unstable resonators:
 A greater portion of the gain medium contributing to the laser output power as a result of the availability of a
larger model volume
 Higher output powers attained from operation on the lowestorder mode, rather than on higherorder transverse
modes as in the case of stable resonators
 High output power with minimal optical damage to the resonator mirrors, as a result of the use of purely reflective
optics that permits the laser light to spill out around the mirror edges (this configuration also permits the optics
to be watercooled and thereby to tolerate high optical powers without damage).
Mode selection
 The point is to use an element within the resonator to provide loss sufficient to prevent oscillation of the undesired
modes.
 Selection of a laser line (certain frequency): You can use a prism to deflect lines of undesired modes (which won’t be
reflected from the other mirror and hence not participate in feedback).
 Selection of a transverse mode: Since they have different spatial distributions, you can physically use an aperture to
prevent some modes, or even design the mirrors to favor a particular transverse mode.
 Selection of a polarization: Use a polarizer, optimally inside the resonator. Use of Brewster windows aids in this.
 Selection of a longitudinal mode:
 Increase loss so that only the largest remains. Bad because the surviving mode would itself be weak.
 Increase the spacing ν_{F} = c∕2d by making the resonator length shorter. Not too good either, since this reduces
the volume of the active medium. Better ways to increase the spacing:
 Put an etalon inside the resonator that filters out the undesired modes. Usually tilted for finetuning.
 Using multiplemirror resonators, i.e., coupled resonators (both active or just one active) of different lengths
or coupled resonator/interferometers
Common lasers (skipping details)
Solidstate lasers
Examples are ruby, alexandrite, Nd^{3+}:YAG, Nd^{3+}:glass, Er^{3+}:silica fiber, Nd^{3+}:YVO_{4}, Yb^{3+}:YAG, Ti^{3+}:sapphire, fiber lasers, Raman
fiber lasers, random lasers (scattering itself provides feedback)
Gas lasers
Examples are HeNe, Ar^{+}, Kr^{+} (atomic and ionic lasers), CO_{2}, methanol, water (molecular gas lasers), excimer (“excited dimer”)
lasers, chemical lasers such as HF
Other lasers
Organic dye lasers, solidstate dye lasers, extreme UV and xray lasers, freeelectron lasers
Pulsed lasers
Why pulse lasers? Doing it greatly increases optical power (total energy is the same of course, it’s just stored in the “off
time”).
Methods of pulsing lasers
Obviously we should do this internally, not externally, to not waste energy.
Gain switching (periodically increasing gain, constant loss)
Turn the laser pump on and off. These is easy to do using electrical pulses.
Qswitching (constant gain, periodically decreasing loss)
Reduce the loss via a modulated absorber inside the resonator. Energy is stored in the atoms.
Cavity dumping (constant gain, periodically increasing loss)
Increase the loss via altering the mirror transmittance. Energy is stored in the photons. Like periodically dumping a bucket of water
that’s constantly being filled.
Mode locking
Couple together the modes of a laser and lock their phases to each other. Like Fourier components.