PHY 411 / 412 / 436

Quantum Optics Prof. Mark Fox

Part II: Photon statistics Lectures 5 - 8

Overview
Topics covered Photon statistics sub-Poissonian light Hanbury Brown-Twiss experiments Photon anti-bunching Reading Quantum Optics Chapters 5 - 6

What is quantum optics ?

ATOM-LIGHT interaction classical optics Semi-classical optics ATOMS LIGHT

Hertzian dipoles Electromagnetic waves Quantized Electromagnetic waves Photons

Quantum optics Quantized

Almost all of undergraduate physics is well described by classical or semi-classical optics !

Photon counting
low intensity beam power P output t counter

PMT / APD
= P

integration time setting, T

R = N = RT

Photon flux (photons/s) Count rate (counts/s) Number of counts in time T

= detector quantum efficiency = counts out / photons in

Photon streams
30 cm

= 633 nm
= 3.1 109 photons / s

1 nW

average of 3 photons in 30 cm of beam timing random on very short time scales Poissonian statistics

Poissonian statistics
nn P ( n) = exp( n ) n! n = n n = mean n = standard deviation

Random events with discrete outcomes. [cf normal (Gaussian) distribution for continuous variables.] Average well-defined, but individual events random Examples: number of rain drops falling in time T number of radioactive decays in time T number of photons from starlight detected in time T

Poisson distributions
1.0 P(n) 0.5 0.0 0.4 P(n) 0.2 0.0 0.2 P(n) 0.1 0.0 0.15 P(n) 0.10 0.05 0.00 0 5 10 15 20 n

nn exp( n ) P ( n) = n! n =5

n = 0.1

10 15 20 n

10 15 20 n

n =1

n = 10

10 15 20 n

Classification by statistics
0.08 P (n) 0.04 Poisson sub-Poissonian n = 100 super-Poissonian 0.00 50 100 150 n

n > n n = n n < n

super-Poissonian Poissonian sub-Poissonian

CLASSICAL NON-CLASSICAL

Thermal light (black body radiation)

P (n) 0.12 0.06 0.00 0 5 10 15
n

thermal

n = 10 Poissonian

1 n= exp( / kBT ) 1

20

1 n P ( n) = n + 1 n + 1

Bose-Einstein distribution super-Poissonian

n = n + n

Photon number states

equally-spaced photons
ct P (n) 1

0 n

Regular photon flow identical value of n in time T for all measurements

P(n ) = 1 P(n n ) = 0 n = 0

Losses
lossy medium T beam splitter T (1T ) incoming photons output detector

losses can be modelled as beam-splitters random sampling theorem: losses degrade photon statistics Examples of losses: inefficient detectors absorption scattering

Theory of photo-detection
single electron light

Photomultiplier tube (PMT)

output pulse counter

photocathode electron multiplier

Semi-classical theory: quantized atoms + EM waves

N N

N = number of counts

Poissonian or super-Poissonian statistics only sub-Poissonian statistics impossible

Quantum detector theory

both atoms and light quantized photon distribution: mean = n , standard deviation n

( N ) ( N )

( n ) + (1 ) n
2

N = n
2

for = 1 for 0

= detector efficiency
N = number of counts

= n = N

faithfully reproduces photon statistics for = 1 Poissonian for low sub-Poissonian N possible: proof of quantum nature of light

sub-Poissonian light
sub-Poissonian current source PMT photon counter Hg atoms space charge high efficiency emitter sub-Poissonian light Teich & Saleh J. Opt. Soc. Am B 2, 275 (1985)

( N )

= 0.9984 N

convert sub-Poissonian electron stream into subPoissonian photon stream with high efficiency light emitter

Detector noise power

Detection circuit
oscilloscope / spectrum analyser high intensity beam i(t)

i(t)

+V0 PD A V(t) RL C

PD amplifier

i (t ) = e = e

i + i (t )

Pnoise (t ) = ( i (t ) ) RL

Shot noise
i(t) i Pnoise( f ) classical noise shot noise level f

Fourier i(t) transform t of Pnoise(t)

2

Pnoise (t ) = ( i (t ) ) RL

1/D

D = detector response time

( N ) ( i )
2

=N

Poissonian photo-electron statistics shot noise f = detection bandwidth i = average photocurrent

( i ) i
2

= 2e f i

Pnoise ( f ) = 2e RL f i

Shot noise limit

Nd:YAG, 1064 nm, 66 mW
40 50 60 70 80 90 Noise power (dBm)
Ottaway et al, Appl. Phys. B 71, 163 (2000)

Ti: sapphire, 930 nm

Noise power (pW) 50 40 30 20 electrical 10 noise 0 0 10 20 Optical power (mW)

shot noise limit

5 10 15 20 Frequency (MHz)

25

Unavoidable noise arising from photon statistics Independent of frequency (white noise) proportional to average power

PD

You cant beat shot noise easily!

Noise eater
beam splitter laser power supply output detector

Balanced detection
50:50 laser S i2 D2 D1 i 1 i1i2

Noise eaters & balanced detectors cancel classical noise in the laser. Photons split randomly at the beam-splitter: impossible to cancel the photon noise, ie the shot noise

Beating the shot noise limit

sub-Poissonian current source R LED p n amp PD spectrum Analyser high efficiency emitter sub-Poissonian light

Drive current limited by resistor Johnson (thermal) 2 rather than shot noise: ( i ) 2e f i High efficiency LED converts sub-Poissonian current flow to sub-Poissonian photon stream

Experimental data
R LED V p n PD amp HP 4220, 875 nm Noise Power (dBm) -75 1.1dB -80 -85 0 shot noise level LED photocurrent noise electronic noise 5 10 15 Frequency (MHz) 20 i = 4.7 mA spectrum Analyser

Current limited by resistor: Johnson rather than shot noise LED much more efficient than discharge tube Efficiency ~ 22%

( N )

= 0.78 N

Wolfl et al J. Mod Opt. 45, 1147 (1998)

Photon antibunching
Fox, Quantum Optics, Chapter 6 Supplementary reading: Hecht, Optics (4th Edition, Addison Wesley, 2002). Chapter 12 covers light fluctuations and the classical theory of Hanbury Brown-Twiss experiments. Smith and King, Optics and Photonics (Wiley, 2000). Chapters 13 and 14 are more or less equivalent to Chapter 12 in Hecht.

Spatial coherence
Point source produces plane waves with a unique angle Extended source produces waves with angular spread s Fringes washed out when s = / d

s
d

Michelson stellar interferometer

M1 starlight telescope d interference pattern M2 Measures spatial coherence: Light from different parts of extended sources is incoherent Resolution = 1.22 / d ~ 107 radians for d = 6m @ 500nm Resolved size of Betelgeuse ( = 2.2107 radians) Requires interferometric stability between M1 and M2

Mount Wilson, 1919

Mount Wilson Observatory

Intensity interferometers
PMT2 electronic multiplier i1 d i2

starlight

output i1 i2

Jodrell Bank 1955-6 R. Hanbury Brown & R. Twiss

PMT1 Resolution = / d ~ 108 radians for d 10m @ 400nm Resolved size of Sirius (S = 3.3108 radians) Does not require interferometric stability How does it work ?

Narrabri stellar intensity interferometer

Robert Hanbury Brown 1916-2002

Hanbury Brown-Twiss experiments

half-silvered mirror 435.8 nm light from filtered Hg lamp d PMT1 i1(t) I1(t) PMT2 I2(t) AC-coupled amplifier

delay generator i2(t + )

multiplier + integrator i1(t)i2(t + )

Measures second-order correlation function I (t ) I (t + ) (2) g ( ) = I (t ) I (t + )

Chaotic light
Bain & Squire, Opt. Comm. 135, 157 (1997)

I(t) / I

Intensity of chaotic source fluctuates wildly on timescales coherence time c Atomic discharge lamps emit chaotic light

Second-order correlation functions

2 g(2) () 1 coherent 0 0 1 2 chaotic

g
/c

(2)

I (t ) I (t + ) ( ) = I (t ) I (t + )

Coherent light: Gaussian chaotic light: All classical light:

g(2)() = 1 for all g(2)() = 1 + exp[(/c)2] g(2)(0) 1 g(2)(0) g(2)() for all

HBT experiments with photons

D2 number of events 0.0 50:50 beam splitter photons

measures g(2)()

D1

stop start counter/timer

1.0

2.0

time interval (arb. units)

Many events near = 0 if photons come in bunches No events at = 0 if photons come one by one

Photon bunching
coherent light: no intensity fluctuations, random time intervalsall times equally likely (Poissonian statistics) bunched light I(t) Iav g(2) () photon bunches fewer photons t 2 1 coherent light 0 0 1 2 chaotic light

Photon bunching

/c

Bunched light generated by chaotic source Agrees with classical results: g(2)(0) 1 g(2)(0) g(2)() for all

Photon antibunching
antibunched light coherent light: random time intervals g(2) ()

Photons emitted at regular time intervals g(2)(0) < 1

1 0 0

Only possible with photon interpretation of light

Making antibunched light

excitation emission
excitation time exc emission time ~ R

single atom

t=0

time

Use single atom + filter Only one photon per colour per cascade average photon spacing ~ (exc+R) negligible probability for 2nd photon at
small times after 1st photon Not observed in discharge lamps due to large number of atoms

Observation of antibunching
50:50 beam splitter PMT1 microscope objective lens
Kimble et al, Phys. Rev. Lett. 39, 692 (1977)

laser low density Na beam 3 g(2)() 2 1 0 0 25 50 75 100 Time delay (ns)

start

PMT2 stop TIMER

<1 g(2)(0) not equal to zero due to finite probability of having two atoms in view at a time

g(2)(0)

Antibunching from quantum dots

laser InAs quantum dot GaAs 1.5 g(2)() 1.0 0.5 0.0 0 10 10 Time delay (ns) HBT

~ 950 nm

Quantum dot behaves like a single atom g(2)(0) < 0

Michler et al Science 290, 2282 (2000)

Single photon sources

2

pulsed excitation
~ R

1
Filter

1 / ftrigger

time

One photon emitted per trigger pulse

Single photon LED

pulsed voltage source ~ P I N quantum dot LED Number of events 50:50 spectral beamfilter splitter
Yuan et al, Science 295, 102 (2002)

SPAD2 TIMER stop start

aperture SPAD1

0 -40 -20 0 20 40 Time between start and stop pulse (ns)

Select single dot with aperture Voltage pulse every 12 ns No events at t = 0

Photon wave-particle duality

Michelson interferometer
Zwiller et al, Phys. Rev B 69, 165307 (2004)

quantum dot single photon source

50:50

HBT experiment

Interference fringes observed in Michelson interferometer: wavelike behaviour Antibunching observed in HBT apparatus: particle like behaviour