Hong-Ou-Mandel Interference

Agata M. Brańczyk
Perimeter Institute for Theoretical Physics, Waterloo, Ontario, N2L 2Y5, Canada
[email protected]

This article is a detailed introduction to Hong-Ou-Mandel (HOM) interference, in which two

photons interfere on a beamsplitter in a way that depends on the photons’ distinguishability.
arXiv:1711.00080v1 [quant-ph] 31 Oct 2017

We begin by considering distinguishability in the polarization degree of freedom. We then con-

sider spectral distinguishability, and show explicitly how to calculate the HOM dip for three
interesting cases: 1) photons with arbitrary spectral distributions, 2) spectrally entangled
photons, and 3) spectrally mixed photons.

1 Introduction
When two indistinguishable photons interfere on a beam splitter, they behave in an interesting way.
This e↵ect is known as Hong-Ou-Mandel (HOM) interference, named after Chung Ki Hong, Zhe
Yu Ou and Leonard Mandel, who experimentally verified the e↵ect in 1987 [1]. HOM interference
shows up in many places, both in fundamental studies of quantum mechanics and in practical
implementations of quantum technologies. At its heart, HOM interference is quite simple to
understand. But it can also be very rich once di↵erent aspects of the incoming light are considered.
This document contains a step-by-step account of how these more interesting e↵ects can be
modelled. We begin, in Section 2, with a basic model of two photons interfering on a beam
splitter. We then consider the photons’ polarization degree of freedom in Section 3, and show that
the output state (after the beam splitter) is fundamentally di↵erent depending on the photons’
relative polarizations. We then extend the model to account for the photons’ spectro-temporal
properties in Section 4, and show how to calculate the famous HOM dip. As concrete examples,
we examine three interesting cases: 1) photons with arbitrary spectral distributions, 2) spectrally
entangled photons, and 3) spectrally mixed photons.
This document is intended to serve as a pedagogical guide; we therefore go into much more
detail than in a typical research paper.

2 A basic model of two-photon interference

Consider two photons incident on a beam splitter, as shown in Figure 1. The combined two-photon
state before arriving at the beam splitter, i.e. the input state, is:

| in iab = â†j b̂†k |0iab = |1; jia |1; kib , (1)

where â†j and b̂†k are bosonic creation operators in beam splitter modes, a and b, respectively. In
addition to being identified by their respective beam splitter modes, the photons can have other
properties, labeled by j and k, that determine how distinguishable they are.

1
 2 A BASIC MODEL OF TWO-PHOTON INTERFERENCE

Some examples of such additional properties are the

b)
photon’s polarization, spectral mode [2, 3], temporal
mode [4], arrival time, or transverse spatial mode [5].
For the time being, we make no assumptions about
the photons’ level of distinguishability.
The evolution of a state as it interferes on a beam
splitter with reflectivity ⌘ can be modelled with a
unitary Ûbs [6]. The unitary acts on the creation
operators as follows:

Û p p
↠!
bs
1 ⌘↠+ ⌘ b̂† (2a)
Figure 1: Two photons interfere on a beam
Ûbs p p
b̂† ! ⌘↠1
⌘ b̂ . (2b) splitter of reflectivity ⌘. Upon exiting the
†

beam splitter, the photons are detected.

The combined two-photon state after exiting the
beam splitter, i.e. the output state, is then:
out
| iab = Ûbs | in iab (3)
⇣ ⌘
† †
= Ûbs âj b̂k |0iab (4)
⇣p p ⌘ ⇣p † p ⌘
= 1 ⌘â†j + ⌘ b̂†j ⌘âk 1 ⌘ b̂†k |0iab (5)
⇣p p ⌘
= ⌘(1 ⌘)â†j â†k + ⌘â†k b̂†j (1 ⌘)â†j b̂†k + ⌘(1 ⌘)b̂†j b̂†k |0iab . (6)

In the case where ⌘ = 1/2, the output state is

1⇣ † † ⌘
| out
iab = âj âk + â†k b̂†j â†j b̂†k b̂†j b̂†k |0iab . (7)
2
Figure 2 shows a diagram representing the four terms in Eq. (7).

+ - -

Figure 2: Diagram showing four di↵erent ways for two photons to interact on a beam splitter. The
signs correspond to signs in front of terms in Eq. (7)

In HOM interference, we are often interested in the coincidence probability, that is, the proba-
bility of detecting one photon in each output port of the beam splitter. To compute this, we must
take into account the distinguishability of the input photons. In the next section, we consider
distinguishability in the polarization degree of freedom.

2
 3 POLARIZATION DISTINGUISHABILITY

3 Polarization distinguishability
Just like classical light, an individual photon can be described as having horizontal (H) or vertical
(V ) polarization, or a superposition of the two (↵H + V , where ↵ and are complex numbers
satisfying |↵|2 + | |2 = 1). In this section, we consider how the polarization of the input photons
influences the HOM coincidence probability.

3.1 Distinguishable photons

First consider two photons with orthogonal polarizations, H and V . Also assume that all other
properties of the photons (spectrum, arrival time, transverse spatial mode, etc.) are identical.
Two photons with orthogonal polarizations are said to be distinguishable. In this scenario, where
j = H and k = V , the output state is
1⇣ † † ⌘
| out iab = âH âV + â†V b̂†H â†H b̂†V b̂†H b̂†V |0iab (8)
2
1
= (|1; Hia |1; V ia + |1; V ia |1; Hib |1; Hia |1; V ib |1; Hib |1; V ib ) . (9)
2
The first term contains both photons in mode a, the second and third terms contain only one
photon in each mode a and b, and the fourth term contains both photons in mode b.
We can compute the coincidence probability from the probability amplitudes in front of terms
with only one photon in each output port, i.e., the two middle terms in Eq. (9). The coincidence
probability is therefore p = |1/2|2 + | 1/2|2 = 1/2.

3.2 Indistinguishable photons

Now consider that the two photons have the same polarization, j = k = H. All other things
being equal (spectrum, arrival time, transverse spatial mode, etc.), two photons with the same
polarization are said to be indistinguishable. In this scenario, the output state is

⇣ ⌘
| out
iab = â†H â†H + â†H b̂†H â†H b̂†H b̂†H b̂†H |0iab (10)
⇣ ⌘
= â†H â†H b̂†H b̂†H |0iab (11)
1
= p (|2; Hia |2; Hib ) . (12)
2
Notice that the twop middle terms in Eq. (10) cancel, but the state still comes out normalized
because (↠)n |0i = n!|ni. The first term in Eq. (12) has both photons in mode a and the second
term has both photons in mode b, but there are no terms corresponding to one photon in each mode
a and b. Compare this with the output state for distinguishable photons in Eq. (9). Incidentally,
the state in Eq. (12) is sometimes called a two-photon N00N state, i.e., |N i|0i + |0i|N i where
N = 2 [7].
Here, the coincidence probability of detecting one photon in each output mode is p = 0. So we
see that when two indistinguishable photons interfere on a beam splitter of reflectivity ⌘ = 1/2,
the amplitudes for “both transmitted” and “both reflected” perfectly cancel out.

3
 4 TEMPORAL DISTINGUISHABILITY

4 Temporal distinguishability
Until now, we implicitly considered photons with the same spectral and temporal properties (their
details were thus not relevant to our analysis). We now extend our analysis to include the spectral
profile of the photons, characterized by the spectral amplitude function (!), and the relative
arrival times of the photons, parametrized by the time delay ⌧ . By controlling the time delays
between two such photons, it is possible to tune their level of distinguishability. This is shown
schematically in Figure 3.

a) b)
indistinguishable

partially distinguishable

distinguishable

direction of propagation

Figure 3: a) Changing the time delay ⌧ between two photons with finite bandwidths changes how
distinguishable they are. b) A time delay in mode b might be introduced by sending the photon
in that mode through a prism that introduces a phase shift.

4.1 Photons with arbitrary spectra

The quantum state for a photon with spectral amplitude function (!), in beam splitter mode a,
is
Z
|1; ia = d! (!)↠(!)|0ia , (13)

where ↠(!) represents

R a creation operator acting on a single frequency mode !. The state is
2
normalized such that d!| (!)| = 1.

Now consider two input photons with arbitrary spectral amplitude functions and '. The
two-photon input state is
in
| iab = |1; ia |1; 'ib (14)
Z Z
= d!1 (!1 )â (!1 ) d!2 '(!2 )b̂† (!2 )|0iab .
†
(15)

We are interested in how the coincidence probability changes as a function of the overlap
between the photons. We thus introduce a time delay in, say, mode b. In practice, this might be
done by sending the photon in mode b through a prism that introduces a phase shift (see Figure

4
4.1 Photons with arbitrary spectra 4 TEMPORAL DISTINGUISHABILITY

3 b)), or perhaps by forcing it to take a longer path. The time delay has the following action on
the creation operator:

b̂† (!) ! b̂† (!)e i!⌧

. (16)

The time-delayed state is

Z Z
†
| td
iab = d!1 (!1 )â (!1 ) d!2 '(!2 )b̂† (!2 )e i!2 ⌧
|0iab . (17)

The beam splitter acts on each frequency mode independently, and we’ll assume that the reflec-
tivity is not frequency-dependent. The beam splitter unitary thus acts on the creation operators
as follows:
p
Û p
↠(!) !
bs
1 ⌘↠(!) + ⌘ b̂† (!) (18)
Ûbs p p
b̂† (!) ! ⌘↠(!) 1 ⌘ b̂† (!) . (19)

After passing through a beam splitter with ⌘ = 1/2, the output state of the two photons is
out
| iab = Ûbs | td iab (20)
Z ⇣ ⌘Z ⇣ ⌘
1 † † † †
= d!1 (!1 ) â (!1 ) + b̂ (!1 ) d!2 '(!2 ) â (!2 ) b̂ (!2 ) e i!2 ⌧ |0iab (21)
2
Z Z
1
= d!1 (!1 ) d!2 '(!2 )e i!2 ⌧
2 (22)
⇣ ⌘
⇥ ↠(!1 )↠(!2 ) + ↠(!2 )b̂† (!1 ) ↠(!1 )b̂† (!2 ) b̂† (!1 )b̂† (!2 ) |0iab .

Earlier, when we considered photons of a single frequency, it was simple to read o↵ the coinci-
dence probability from the state. Here, it is a bit more tricky so we should calculate it explicitly.
We’ll model each detector as having a flat frequency response. The projector describing detection
in mode a is given by
Z
P̂a = d!↠(!)|0ia h0|a â(!) , (23)

and the projector describing detection in mode b is given by

Z
P̂b = d! b̂† (!)|0ib h0|b b̂(!) . (24)

The coincidence probability of detecting one photon in each mode is

out out out out
p = Tr[| iab h |ab P̂a ⌦ P̂b ] = h |ab P̂a ⌦ P̂b | iab . (25)

For two photons with arbitrary spectral amplitude functions and ', the coincidence probability

5
4.1 Photons with arbitrary spectra 4 TEMPORAL DISTINGUISHABILITY

is
" Z Z
1 ⇤
parb = d!1 d!2 (!1 )'⇤ (!2 )ei!2 ⌧
2
#
⇣ ⌘
⇥ h0| â(!1 )â(!2 ) + â(!2 )b̂(!1 ) â(!1 )b̂(!2 ) b̂(!1 )b̂(!2 )
"Z Z #
⇥ d!a ↠(!a )|0ia h0|a â(!a ) d!b b̂† (!b )|0ib h0|b b̂(!b ) (26)
" Z Z
1 i!20 ⌧
⇥ d!1 d!20 (!10 )'(!20 )e
0
2
#
⇣ ⌘
⇥ ↠(!10 )↠(!20 ) + ↠(!20 )b̂† (!10 ) ↠(!10 )b̂† (!20 ) b̂† (!10 )b̂† (!20 ) |0iab ,

where all we did so far was insert Eqs. (22), (23), and (24) into Eq. (25). Reshu✏ing some parts,
we can write
Z Z Z Z Z Z
1 0
parb = d!a d!b d!1 d!2 d!1 d!20 ⇤ (!1 )'⇤ (!2 ) (!10 )'(!20 )ei(!2 !2 )⌧
0
4
⇣ ⌘
⇥ h0|ab â(!1 )â(!2 ) + â(!2 )b̂(!1 ) â(!1 )b̂(!2 ) b̂(!1 )b̂(!2 ) ↠(!a )b̂† (!b )|0iab (27)
⇣ ⌘
⇥ h0|ab â(!a )b̂(!b ) ↠(!10 )↠(!20 ) + ↠(!20 )b̂† (!10 ) ↠(!10 )b̂† (!20 ) b̂† (!10 )b̂† (!20 ) |0iab .

Terms with an odd number of operators in one mode, e.g. h0|ab ââ↠b̂† |0iab and h0|ab b̂b̂↠b̂† |0iab , go
to zero, while terms such as h0|ab âb̂↠b̂† |0iab give delta functions:
Z Z Z Z Z Z
1 0
parb = d!a d!b d!1 d!2 d!1 d!20 ⇤ (!1 )'⇤ (!2 ) (!10 )'(!20 )ei(!2 !2 )⌧
0
4
(28)
⇥ ( (!2 !a ) (!1 !b ) (!1 !a ) (!2 !b ))
⇥ ( (!20 !a ) (!10 !b ) (!10 !a ) (!20 !b )) .

Using the delta functions to evaluate the integrals over !a and !b gives an expression with two
terms:
Z Z Z Z
1 0
parb = d!1 d!2 d!1 d!20 ⇤ (!1 )'⇤ (!2 ) (!10 )'(!20 )ei(!2 !2 )⌧
0
2 (29)
0 0 0 0
⇥ ( (!2 !2 ) (!1 !1 ) (!1 !2 ) (!2 !1 )) .

Using the remaining delta functions

R to evaluate the integrals over !10 and !20 , and taking advantage
of the normalization condition d!| (!)|2 = 1, gives
Z Z
1 1 ⇤
parb = d!1 (!1 )'(!1 )e i!1 ⌧
d!2 '⇤ (!2 ) (!2 )ei!2 ⌧ . (30)
2 2
If (!) = '(!), this expression simplifies to
Z Z
1 1
parb = d!1 | (!1 )|2 e i!1 ⌧
d!2 | (!2 )|2 ei!2 ⌧ . (31)
2 2

6
4.1 Photons with arbitrary spectra 4 TEMPORAL DISTINGUISHABILITY

4.1.1 Example: Gaussian photons

Consider two photons with Gaussian spectral amplitude functions,
(! !¯ i )2
1 2 2
i (!) = p e i ; (i = a, b) , (32)
(⇡)1/4 i

where !¯ i is the centralR frequency of photon i, i defines its spectral width, and the normalization
was chosen such that d!| i (!)|2 = 1. From Eq. (30), the coincidence probability is
Z ! Z !
¯ a )2 (! 1 !¯ b )2 ¯ a )2 (! 2 !¯ b )2
1 1 (!1 !
2
(!2 !
2
parb,gauss = d!1 e 2
2 a e 2
b e i!1 ⌧ d!2 e 2
2 a e 2
b ei!2 ⌧ .
2 2⇡ a b
(33)

The Fourier Transforms can be evaluated using your favourite method (mine is Mathematica). The
coincidence probability simplifies to
2 2 2 ¯ b )2
a b ⌧ +(!
¯a !
1 a b 2+ 2
parb,gauss = 2
e a b . (34)
2 ( 2 +
a b)

If the Gaussians are equal, that is a = b, we have

1 1 2 2
a⌧
parb,gauss = e 2 . (35)
2 2

4.1.2 Example: Sinc-shaped photons

Consider two photons with a sinc spectral amplitude function,
r
Ai
'i (!) = sinc (Ai (! !¯ i )) ; (i = a, b) , (36)
⇡

¯ i is the centralRfrequency of photon i, Ai 1 defines its spectral width, and the normalization
where !
was chosen such that d!|'i (!)|2 = 1. From Eq. (30), the coincidence probability is
✓Z ◆
1 Aa Ab i!1 ⌧
parb,sinc = d!1 sinc (Aa (!1 ! ¯ 1 )) sinc (Ab (!1 ! ¯ 1 )) e
2 2⇡ 2
✓Z ◆ (37)
i!2 ⌧
⇥ d!2 sinc (Aa (!2 ! ¯ 2 )) sinc (Ab (!2 ! ¯ 2 )) e .

Computing the Fourier Transforms, in the case where 'a = 'b , the coincidence probability simpli-
fies to
1 1 ⇣ ⌧ ⌧ ⌘2
parb,sinc = |⌧ | Aa + A a . (38)
2 8A2a 2 2

Figure 4 compares the coincidence probabilities, Eqs. (35) and (38), for photons with Gaussian
and sinc spectral amplitude functions respectively.

7
4.2 Spectrally entangled photons 4 TEMPORAL DISTINGUISHABILITY

SEPARABLE PHOTONS
0.5

0.4

coincidence probability
0.3

0.2

0.1 gaussian
sinc
0.0
-4 -2 0 2 4
τ (10-12 s)

Figure 4: The coincidence probability for separable photons, as a function of time delay ⌧ . The
solid line is parb,gauss definedpin Eq. (35) for a = 10 12 rad/s, and the dashed line is parb,sinc defined
in Eq. (38) for Aa = 1/ a 2 where a = 10 12 rad/s and = 0.193. This value of ensures
that the Gaussian and sinc spectral amplitude functions have the same widths, for comparison.

4.2 Spectrally entangled photons

In the previous section, we considered two photons with arbitrary, but separable, spectral ampli-
tude profiles. More generally, however, the spectral amplitudes of two photons can be correlated—
that is, the photons can be entangled in the spectral degree of freedom. The nature of this spectral
entanglement is captured by the joint spectral amplitude (JSA) function f (!1 , !2 ). The quantum
state of two spectrally entangled photons is
Z Z
in
| iab = d!1 d!2 f (!1 , !2 )↠(!1 )b̂† (!2 )|0iab . (39)

Notice that for f (!1 , !2 ) = (!1 )'(!2 ), this state reduces to the separable state in Eq. (15).
Entangled photons can be sent onto a beam splitter in exactly the same way as separable photons,
but the way they interfere will also depend on the nature of their entanglement.
As before, to calculate the HOM dip, we begin by introducing a time delay ⌧ in mode b, by
applying the transformation in Eq. (16). We then model how the photons interact via the beam
splitter by applying the beam splitter unitary in Eq. (18) to the time-delayed state. We finally
calculate the coincidence probability, as defined in Eq. (25), using projectors, defined in Eqs. (23)
and (24), that describe detection in modes a and b. The steps are identical to those in Section 4.1,
and yield the coincidence probability:
Z Z
1 1
pent = d!1 d!2 f ⇤ (!1 , !2 )f (!2 , !1 )ei(!2 !1 )⌧ . (40)
2 2
In fact, this is equivalent to the coincidence probability in Eq. (30) when f (!1 , !2 ) = (!1 )'(!2 ).
In general, however, f (!1 , !2 ) will not take such a nice form, and the integrals in Eq. (40) will
need to be evaluated numerically.
It can sometimes be useful to express the JSA in terms of its’ Schmidt decomposition,
X
f (!1 , !2 ) = uk k (!1 )'k (!2 ) , (41)
k

8
4.2 Spectrally entangled photons 4 TEMPORAL DISTINGUISHABILITY

where the RSchmidt modes k (!) R and⇤ 'k (!) each form a discrete basis of complex orthonormal
⇤
functions ( Pd! k (!) k0 (!) = d!'k (!)'k0 (!) = kk0 ), and the Schmidt coefficients uk are real
and satisfy k u2k = 1 if f (!1 , !2 ) is normalized. The coincidence probability can then be expressed
in terms of the Schmidt coefficients and Schmidt modes as
Z Z
1 1X ⇤
pent = uk uk0 d!1 k (!1 )'k0 (!1 )e i!1 ⌧
d!2 '⇤k (!2 ) k0 (!2 )ei!2 ⌧ . (42)
2 2 k,k0

4.2.1 Example: SPDC pumped by a pulsed pump laser

The joint spectral amplitude for photons generated via spontaneous parametric downconversion is

f (!1 , !2 ) / (!1 , !2 )↵(!1 + !2 ) , (43)

R
such that d!1 d!2 |f (!1 , !2 )|2 = 1, where (!1 , !2 ) is known as the phase-matching function and
↵(!1 + !2 ) is the pump amplitude function [8].
A typical SPDC crystal of length L generates a phase-matching function with a sinc profile:
✓ ◆
k(!1 , !2 )L
sinc (!1 , !2 ) = sinc , (44)
2

where k(!1 , !2 ) = kp (!1 + !2 ) k1 (!1 ) + k2 (!2 ), and ki (!) are the wave numbers associated with
the respective fields.
To simplify calculations, we can Taylor expand k(!1 , !2 ) to first order:

k(!1 , !2 ) = k1,0 + k2,0 kp,0 + k10 (!1 ¯ ) + k20 (!2

! !
¯) kp0 (!1 + !2 2¯
!) , (45)

where kp,0 = kp (2¯ ! ), kp0 = @p (!)/@!|!=2¯! , and k1/2

! ), k1/2,0 = k1/2 (¯ 0
= @1/2 (!)/@!|!=¯! . This
approximation is valid in many regimes. We can then write

sinc (!1 , !2 ) = sinc (A!1 + B!2 C) , (46)

where
L 0
A= (k kp0 ) (47)
2 1
L
B = (k20 kp0 ) (48)
2
L
C = (k1,0 + k2,0 kp,0 + (k10 + k20 2kp0 )¯
!) . (49)
2
For comparison, it is useful to define a Gaussian phase-matching function of the same width:
(A!1 +B!2 C)2
gauss (!1 , !2 ) =e , (50)

where the parameter = 0.193 ensures that the Gaussian and sinc functions have the same widths.
Gaussian phase-matching functions were originally used in the literature to simplify calculations.
But, more recently, methods have been developed to generate them in practice [9, 10, 11, 12, 13]. In
combination with the right set of parameters (A, B, C, and ), Gaussian phase-matching functions
make it possible to generate separable joint spectral amplitudes via SPDC (see Fig. 5 a))

9
4.2 Spectrally entangled photons 4 TEMPORAL DISTINGUISHABILITY

994 1000 1006 994 1000 1006

a) 1006 1006 b) 1006 1006

ω2 (1012 rad/s)

ω2 (⨯1012 )
1000 1000 1000 1000

994 994 994 994

994 1000 1006 994 1000 1006
12 12
ω1 (10 rad/s) ω1 (⨯10 )

p
Figure 5: Joint spectral amplitudes: a) |fgauss (!1 , !2 )|, and b) |fsinc (!1 , !2 )|, with A = 1/ 2 ,
B = A, C = 0, = 0.193, = 1012 rad/s, and ! ¯ = 1015 rad/s.

For a pulsed pump laser, it is common to assume a Gaussian pump amplitude function:
¯ 2
(! !)
↵(!) = e 2 2 , (51)

where !
¯ is the central frequency of the pump, and defines the spectral width.
We define the corresponding joint spectral amplitudes:

fgauss (!1 , !2 ) / gauss (!1 , !2 )↵(!1 + !2 ) (52)

fsinc (!1 , !2 ) / sinc (!1 , !2 )↵(!1 + !2 ) , (53)

which are plotted in Figure 5.

Figure 6 compares the coincidence probabilities for photons from an SPDC source pumped by
a pulsed laser, with Gaussian and sinc phase matching functions functions respectively.

SPDC PAIR (PULSED)

0.5

0.4
coincidence probability

0.3

0.2

0.1 gaussian
sinc
0.0
-4 -2 0 2 4
τ (10-12 s)

Figure 6: The coincidence probability for photons from an SPDC source pumped by a pulsed laser,
Eq. (40), as a function of time delay ⌧ . The solid line is for fgauss (!1 , !2 ) defined in Eq. (52), and
the dashed line is for fsinc (!1 , !2 ) defined in Eq. (53). All parameters are the same as for Figure
5.

10
4.3 Spectrally mixed states 4 TEMPORAL DISTINGUISHABILITY

4.2.2 Example: SPDC pumped by a CW pump laser

For a nonlinear source pumped by a continuous wave (CW) laser at frequency 2¯
! , the joint spectral
amplitude takes the form

f (!1 , !2 ) / (!1 , !2 ) (!1 + !2 2¯

!) , (54)
R
such that d!1 d!2 |f (!1 , !2 )|2 = 1. The coincidence probability in Eq. (40) then simplifies to
Z
1 1
pcw = d!g ⇤ ( !)g(!)ei2!⌧ , (55)
2 2
R
where g(!) / (¯ ! !, ! ¯ + !), such that d!|g(!)|2 = 1.
Figure 7 compares the coincidence probabilities for photons from an SPDC source pumped by
a CW laser, with Gaussian and sinc phase matching functions functions respectively.

SPDC PAIR (CW)

0.5

0.4
coincidence probability

0.3

0.2

0.1 gaussian
sinc
0.0
-4 -2 0 2 4
τ (10-12 s)

Figure 7: The coincidence probability for entangled photons from an SPDC source pumped by a
CW laser, Eq. (55), as a function of time delay ⌧ . The solid line is for gauss (!1 , !2 ) defined in
Eq. (50), and the dashed line is for sinc (!1 , !2 ) defined in Eq. (46). All parameters are the same
as for Figure 5.

4.3 Spectrally mixed states

In the previous sections, we considered single photons and photon pairs in spectrally pure states.
It is possible, however, for a photon to be in a mixture of di↵erent spectral amplitude functions
k (!). This can be represented by the density matrix
X
⇢ = qk |1; k ia h1; k |a , (56)
k

where
Z
|1; ia = d! (!)↠(!)|0ia , (57)

11
4.3 Spectrally mixed states 4 TEMPORAL DISTINGUISHABILITY

is the quantum state for a single photon in a spectral mode defined by the spectral amplitude
function (!). The density matrix ⇢ can be realized in one of two ways: 1) as a probabilistic
preparation on the pure state |1; k i with probability qk , or 2) as the reduced density matrix of a
spectrally entangled two-photon state, such as Eq. (39) (see Appendix A for details).
Spectrally mixed photons can be sent onto a beam splitter in exactly the same way as separable
or entangled photons. The density operator for a two-photon input state can be written as
⇢in = ⇢ ⌦ ⇢' (58)
X X
= qk |1; k ia h1; k |a ⌦ qk0 0 |1; 'k0 ib h1; 'k0 |b (59)
k k
X
= qk qk0 0 (|1; k ia |1; 'k0 ib ) (h1; k |a h1; 'k0 |b ) . (60)
kk0

As before, the next step is to introduce a time delay ⌧ in mode b, by applying the transformation
in Eq. (16), and then model how the photons interact via the beam splitter by applying the beam
splitter unitary in Eq. (18) to the time-delayed state. But notice that the state |1; k ia |1; 'k0 ib
is just the state of two photons with arbitrary spectral amplitude functions k and 'k0 that we
saw in Eq. (14) in Section 4.1. Due to the linearity of quantum mechanics, we can simply use the
result from Section 4.1 to write the output density operator
X
⇢out = qk qk0 0 | kk
out out
0 iab h kk 0 |ab , (61)
kk0

where
Z Z
out 1 i!2 ⌧
| kk0 iab = d!1 k (!1 ) d!2 'k0 (!2 )e
2 (62)
⇣ ⌘
† † † † † † † †
⇥ â (!1 )â (!2 ) + â (!2 )b̂ (!1 ) â (!1 )b̂ (!2 ) b̂ (!1 )b̂ (!2 ) |0iab .

This is equivalent to Eq. (22) for (!) = k (!) and '(!) = 'k0 (!). The coincidence probability
of getting one photon in each mode is
pmix = Tr[⇢out P̂a ⌦ P̂b ] (63)
X
= qk qk0 0 h kk
out
0 |ab P̂a ⌦ P̂b |
out
kk0 iab , (64)
kk0

out out
where h kk 0 |ab P̂a ⌦ P̂b | kk 0 iab is the coincidence probability, defined in Eq. (25), for two photons
out
with arbitrary spectral amplitude functions k (!) and 'k0 (!). We can therefore replace h kk 0 |ab P̂a ⌦
out
P̂b | kk0 iab with Eq. (30), for (!) = k (!) and '(!) = 'k0 (!), to get
Z Z
1 1X 0 ⇤
pmix = qk qk0 d!1 k (!1 )'k0 (!1 )e i!1 ⌧
d!2 '⇤k0 (!2 ) k (!2 )ei!2 ⌧ . (65)
2 2 kk0

4.3.1 Example: Independent SPDC sources

Consider two independent SPDC sources that generate the entangled states:
Z Z
| 1 iab = d!1 d!2 f (!1 , !2 )↠(!1 )b̂† (!2 )|0iab (66)
Z Z
| 2 icd = d!1 d!2 h(!1 , !2 )ĉ† (!1 )dˆ† (!2 )|0icd . (67)

12
4.3 Spectrally mixed states 4 TEMPORAL DISTINGUISHABILITY

To model a HOM experiment between photons in modes a and c, we first compute the reduced
density operators for those modes (see Appendix A):
X
⇢ = trb [| 1 iab h 1 |ab ] = u21 |1; k ia h1; k |a (68)
k
X
⇢' = trd [| 2 icd h 2 |cd ] = v12 |1; 'k ic h1; 'k |c , (69)
k

where k and 'k are defined in terms of the Schmidt decompositions of the joint spectral ampli-
tudes:
X
0
f (!1 , !2 ) = uk k (!1 ) k (!2 ) (70)
k
X
h(!1 , !2 ) = vk 'k (!1 )'0k (!2 ) . (71)
k

The coincidence probability, Eq. (65), becomes

Z Z
1 1X 2 2
pmix = u v 0 d!1 ⇤k (!1 )'k0 (!1 )e i!1 ⌧
d!2 '⇤k0 (!2 ) k (!2 )e
i!2 ⌧
. (72)
2 2 kk0 k k

Figure 8 compares the coincidence probabilities for photons from two independent SPDC sources
pumped by pulsed lasers, with Gaussian and sinc phase matching functions functions respectively.
MIXED STATES FROM INDEPENDENT SPDC SOURCES
0.5

0.4
coincidence probability

0.3

0.2

0.1 gaussian
sinc
0.0
-4 -2 0 2 4
τ (10-12 s)

Figure 8: The coincidence probability for photons from two independent SPDC sources pumped
by pulsed lasers, Eq. (72), as a function of time delay ⌧ . The solid line is for f (!1 , !2 ) =
h(!1 , !2 ) = fgauss (!1 , !2 ), where fgauss (!1 , !2 ) is defined in Eq. (52), and the dashed line is for
f (!1 , !2 ) = h(!1 , !2 ) = fsinc (!1 , !2 ), where fsinc (!1 , !2 ) is defined in Eq. (53). All parameters are
the same as for Figure 5.

4.3.2 Purity and Visibility

In the special case of mixed, identical, and separable photons, there is a nice relationship between
the visibility of the HOM dip and the purity of the input photons.

13
 5 (NOT SO) FINAL WORDS

The visibility of the HOM dip is given by

pmax pmin
V = , (73)
pmax
where
1
pmax = lim pmix = (74)
⌧ !1 2 Z Z
1 1X 0 ⇤
pmin = lim pmix = qk qk 0 d!1 k (!1 )'k0 (!1 ) d!2 '⇤k0 (!2 ) k (!2 ) . (75)
⌧ !0 2 2 kk0

Given two photons with the same mixed density matrix ('k = k ),
Z Z
1 1X ⇤ ⇤
pmin = qk qk0 d!1 k (!1 ) k0 (!1 ) d!2 k0 (!2 ) k (!2 ) (76)
2 2 kk0
Z
1 1X
= qk qk0 d!1 ⇤k (!1 ) k0 (!1 ) kk0 (77)
2 2 kk0
1 1X 2
= q . (78)
2 2 k k

In this case, the visibility is

1
P
2
( 12 1
2
2
k qk )
V = 1 (79)
2
X
= qk2 . (80)
k

The visibility is then equal to the purity of each photon

⇥ ⇤
purity = tr ⇢2 (81)
" #
X X
= tr qk |1; k ia h1; k |a qk0 |1; k0 ia h1; k 0 |a (82)
k k0
X
= qk qk 0 kk0 (83)
kk0
X
= qk2 . (84)
k

5 (Not so) Final words

In this document, we introduced Hong-Ou-Mandel interference in terms of the photons’ distin-
guishability in the polarization degree of freedom. We then examined distinguishability as a
function of the temporal overlap between the photons, and introduced the HOM dip. We saw
how the HOM dip depends on the spectral properties of the input photons; in particular, how the
HOM dip depends on the photons’ spectral amplitude functions, as well as their entanglement with
each other and with other photons. We also saw some examples relevant to photons generated via
spontaneous parametric downconversion (SPDC).

14
 A REDUCED DENSITY MATRIX

The observations in these notes are not new, and similar results are scattered throughout the
literature (e.g. [14, 15, 16, 17, 18]). It was my aim, however, to provide a self-contained pedagogical
resource for students and researchers who want to see how these calculations are done explicitly.
I intend for these notes to be a work-in-progress. In future versions, I’d like to include the
e↵ects of spectral filtering [19], multi-photon states [16, 18], and interference on multi-port beam
splitters [20, 21]. I would also like to include examples relevant to quantum-dot sources [22, 23, 24],
which not only have di↵erent spectral amplitude functions, but also unique features such as time
jitter in the emitted photon. If there are other examples that you would like included in further
versions of these notes, please contact me.

6 Acknowledgements
An earlier informal version of these notes has been floating around the internet since 2012. I’d like
to thank Hubert de Guise for motivating me to write them in the first place. Over the years, I’ve
been pleasantly surprised by the number of people that read them and contacted me to say that
they found them useful. This made me think it would be worthwhile to make a more formal version
for the arXiv, with (hopefully) fewer typos and slip-ups. I’d also like to thank Francesco Graffitti
and Morgan Mastrovich for making useful suggestions that I’ve incorporated in this version.

A Reduced density matrix

The density matrix
X
⇢ = qk |1; k ih1; k| , (85)
k

where
Z
|1; i = d! (!)↠(!)|0i , (86)

can be realized as the reduced density matrix of a spectrally entangled two-photon state. In other
words, by preparing a spectrally entangled state such as the one introduced in Eq. (39),
Z Z
in
| iab = d!1 d!2 f (!1 , !2 )↠(!1 )b̂† (!2 )|0iab , (87)

and discarding one of the photons. Mathematically, this is represented by “tracing out” the
discarded mode using the partial trace operation. To perform the partial trace, we first make use
of the Schmidt decomposition to write
X
| in iab = uk |1; k ia |1; 'k ib . (88)
k

15
REFERENCES REFERENCES

The reduced density matrix for system a is

⇥ ⇤
⇢ = trb | in iab h in | (89)
X
= uk uk0 trb [|1; k ia |1; 'k ib h1; k0 |a h1; 'k0 |b ] (90)
kk0
X
= uk uk0 |1; k ia h1; k0 |a tr [|1; 'k ib h1; 'k0 |b ] (91)
kk0
X
= uk uk0 |1; k ia h1; k0 |a kk0 (92)
kk0
X
= u2 |1; k ia h1; k |a , (93)
k

which has the same form as Eq. (85) for qk = u2k .

References
[1] C. K. Hong, Z. Y. Ou, and L. Mandel, “Measurement of subpicosecond time intervals between
two photons by interference,” Phys. Rev. Lett. 59, 2044 (1987).

[2] A. Eckstein et al., “Highly Efficient Single-Pass Source of Pulsed Single-Mode Twin Beams
of Light,” Phys. Rev. Lett. 106, 013603 (2011).

[3] P. Sharapova et al., “Schmidt modes in the angular spectrum of bright squeezed vacuum,”
Phys. Rev. A 91, 043816 (2015).

[4] V. Ansari et al., “Temporal-mode measurement tomography of a quantum pulse gate,” ArXiv
e-prints (2017).

[5] N. K. Langford et al., “Measuring Entangled Qutrits and Their Use for Quantum Bit Com-
mitment,” Phys. Rev. Lett. 93, 053601 (2004).

[6] H.-A. Bachor and T. C. Ralph, A Guide to Experiments in Quantum Optics, 2 ed. (Wiley-
VCH, 2004).

[7] H. Lee, P. Kok, and J. P. Dowling, “A quantum Rosetta stone for interferometry,” Journal of
Modern Optics 49, 2325 (2002).

[8] W. P. Grice and I. A. Walmsley, “Spectral information and distinguishability in type-II down-
conversion with a broadband pump,” Phys. Rev. A 56, 1627 (1997).

[9] A. M. Brańczyk et al., “Engineered optical nonlinearity for quantum light sources,” Opt.
Express 19, 55 (2011).

[10] P. B. Dixon, J. H. Shapiro, and F. N. C. Wong, “Spectral engineering by Gaussian phase-

matching for quantum photonics,” Opt. Express 21, 5879 (2013).

[11] A. Dosseva, L. Cincio, and A. M. Brańczyk, “Shaping the joint spectrum of down-converted
photons through optimized custom poling,” Phys. Rev. A 93, 013801 (2016).

16
REFERENCES REFERENCES

[12] J.-L. Tambasco et al., “Domain engineering algorithm for practical and e↵ective photon
sources,” Opt. Express 24, 19616 (2016).

[13] F. Graffitti et al., “Pure down-conversion photons through sub-coherence length domain en-
gineering,” ArXiv e-prints (2017).

[14] H. Fearn and R. Loudon, “Theory of two-photon interference,” Journal of the Optical Society
of America B: Optical Physics 6, 917 (1989).

[15] T. Legero et al., “Time-resolved two-photon quantum interference,” Applied Physics B 77,
797 (2003).

[16] Z. Y. Ou, “Temporal distinguishability of an N -photon state and its characterization by

quantum interference,” Phys. Rev. A 74, 063808 (2006).

[17] Z. Y. OU, “Multi-photon Interference and Temporal Distinguishability of Photons,” Interna-

tional Journal of Modern Physics B 21, 5033 (2007).

[18] O. Cosme et al., “Hong-Ou-Mandel interferometer with one and two photon pairs,” Phys.
Rev. A 77, 053822 (2008).

[19] A. M. Brańczyk et al., “Optimized generation of heralded Fock states using parametric down-
conversion,” New Journal of Physics 12, 063001 (2010).

[20] M. C. Tichy et al., “Zero-Transmission Law for Multiport Beam Splitters,” Phys. Rev. Lett.
104, 220405 (2010).

[21] S.-H. Tan et al., “SU(3) Quantum Interferometry with Single-Photon Input Pulses,” Phys.
Rev. Lett. 110, 113603 (2013).

[22] A. Dousse et al., “Ultrabright source of entangled photon pairs,” Nature 466, 217 (2010).

[23] M. E. Reimer et al., “Bright single-photon sources in bottom-up tailored nanowires,” Nature
Communications 3, 737 EP (2012).

[24] M. A. M. Versteegh et al., “Observation of strongly entangled photon pairs from a nanowire
quantum dot,” Nature Communications 5, 5298 EP (2014).

17
REFERENCES REFERENCES

[12] J.-L. Tambasco et al., “Domain engineering algorithm for practical and e↵ective photon
sources,” Opt. Express 24, 19616 (2016).

[13] F. Graffitti et al., “Pure down-conversion photons through sub-coherence length domain en-
gineering,” ArXiv e-prints (2017).

[14] H. Fearn and R. Loudon, “Theory of two-photon interference,” Journal of the Optical Society
of America B: Optical Physics 6, 917 (1989).

[15] T. Legero et al., “Time-resolved two-photon quantum interference,” Applied Physics B 77,
797 (2003).

[16] Z. Y. Ou, “Temporal distinguishability of an N -photon state and its characterization by

quantum interference,” Phys. Rev. A 74, 063808 (2006).

[17] Z. Y. OU, “Multi-photon Interference and Temporal Distinguishability of Photons,” Interna-

tional Journal of Modern Physics B 21, 5033 (2007).

[18] O. Cosme et al., “Hong-Ou-Mandel interferometer with one and two photon pairs,” Phys.
Rev. A 77, 053822 (2008).

[19] A. M. Brańczyk et al., “Optimized generation of heralded Fock states using parametric down-
conversion,” New Journal of Physics 12, 063001 (2010).

[20] M. C. Tichy et al., “Zero-Transmission Law for Multiport Beam Splitters,” Phys. Rev. Lett.
104, 220405 (2010).

[21] S.-H. Tan et al., “SU(3) Quantum Interferometry with Single-Photon Input Pulses,” Phys.
Rev. Lett. 110, 113603 (2013).

[22] A. Dousse et al., “Ultrabright source of entangled photon pairs,” Nature 466, 217 (2010).

[23] M. E. Reimer et al., “Bright single-photon sources in bottom-up tailored nanowires,” Nature
Communications 3, 737 EP (2012).

[24] M. A. M. Versteegh et al., “Observation of strongly entangled photon pairs from a nanowire
quantum dot,” Nature Communications 5, 5298 EP (2014).

17