EE346 NLO
1/13/21, #5A slide 1
5A_Nonlinear Susceptibilities:
Definitions
EE 346 Nonlinear Optics
M.M. Fejer
[email protected] 01/22/21
EE346 NLO
1/13/21, #5A slide 2
Previously
Lecture 2 Lecture 3 & 4
−5
How to obtain n2ω − nω ~ 10 ?
Phasematching in birefringent media
Z
2
2 sin(∆kz / 2) no ,2ω k
I 2ω ( z ) = κ 2
2ω
2
Eω ,0 ne ,ω (θ PM ) = no ,2ω
∆k / 2 θ PM
ne ,ω
X
2ω ωχ (2)
∆k = ( n2ω − nω ) κ jω ≡
c 2cn jω
η 1
ηPM
CEO model for susceptibility
F = −k q q
0.5 −e
E
Ne 2 1
χ (1)
=
0
-10 -5 0 5 10 ε 0 m [(ω02 − ω 2 ) + i γ ω ]
∆kL / 2
EE346 NLO
1/13/21, #5A slide 3
These Notes
• Anharmonic oscillator model for nonlinear response
– Harris 3.7, Boyd 1.4, Yariv 16.3
• How to formulate systematically in terms of nonlinear susceptibility?
– ch. 3.2 Harris
– ch. 1.3 Boyd briefer, readable
• Return to CEO version of nonlinear susceptibility
– discuss Miller’s rule
– ch. 3.7 Harris
– ch. 1.4 Boyd
EE346 NLO
1/13/21, #5A slide 4
Nonlinear Response: Classical
• Consider an electron moving in noncentrosymmetric potential:
anharmonic
1 1 V (q) 1.2
V = kq 2 + Dq 3
motion 2 3 1
• Eqn of motion (as in #4.14)*: 0.8
E
q + γ q + ω02 q + Dq 2 = −e ω ( eiωt + e − iωt ) 0.6
D ≡ D / m 2m 0.4
1
– ansatz: q = ( q1e + q2 e + c.c.)
iωt i 2ω t
0.2
2
– substitute in eqn of motion 0
iωt
armonics equate coeffs of e : (P.S. 2) -1 -0.5 0 0.5 1
balance
q q
approach −eEω 1
find: q1 = leads to same as #4.15 for χ(1)
m ω0 − ω + iωγ
2 2
discussion follows
i 2ωt Yariv ch. 16.3
equate coeffs of e :
− Dq12 / 2 * Here we take field as
find: q2 = E (t ) =
Eω iωt
( e + c.c.)
−4ω 2 + i 2ωγ + ω02
E 2
− De Eω2 2
insert result for q1: =
2m 2 [ ( ω02 − ω 2 ) + iωγ ] [ ω02 − (2ω )2 + i 2ωγ ]
2
EE346 NLO
1/13/21, #5A slide 5
Nonlinear Response 2
Component of motion at harmonic frequency ∝ DEω :
2
e2 1
q2 = − DEω 2
2m 2
(ω02 − ω 2 ) + iωγ
2
ω02 − (2ω ) 2 + i 2ωγ
∝ asymmetry
and quadratic in field
Second harmonic polarization:
NDe3 Eω2
P2ω = − Neq2 =
(ω ) + iωγ
2
2m 2 2
0 −ω 2
ω02 − (2ω ) 2 + i 2ωγ
Nonlinear susceptibility:
P NDe3
χ = 2 2ω 2 =
(2)
see def’n of χ(2) ε 0 Eω
(ω ) + iωγ
2
#5A.7
ε 0 m 2 2
0 −ω 2
ω02 − (2ω ) 2 + i 2ωγ
explanation
forfactorof 2 later
Discuss further (#5A.11) after looking at formal definition of nonlinear susceptibility
EE346 NLO
1/13/21, #5A slide 6
Nonlinear Susceptibilities
• A confusing subject
– various not entirely compatible conventions exist
– partly historical artifact, partly due to inherent subtlety compared
to linear susceptibility
– care required in reading the literature, using tabular data
• Why is NL susceptibility more confusing than linear?
– linear constitutive relation simple
conventions for definitions of fields and Fourier components drop out
– nonlinear constitutive relations sensitive to field conventions
various factors of two arise
errors in literature were not uncommon
factors
of 2 etc
EE346 NLO
1/13/21, #5A slide 7
Formal Definition of NL Susceptibility
(2) 2
• We would like to have a simple relation like P2ω = ε 0 χ eff Eω
– connects pertinent (given phasematching scheme) scalar magnitudes
– need to go through some formal steps to do so consistently
• Define real quantities. For a field with N discrete frequencies:
N for CW, related to def’ns of #2.7 by:
E(r, t ) = ∑ E(ω )e
n =− N
n
iωnt
where E( −ωn ) = E (ωn ) *
E(r, t ) = Eω (r )cos(ωt + φ )
togetrealfield 1 1
E(ω ) = Eω (r )
– similarly for the NL polarization: N 2
P (r, t ) = ∑ P (ωn )eiωnt 1
= ω (r )e − ikω z
n =− N 2
• Susceptibility easiest to define in the frequency domain
outputw
input ws
Pi (ωn + ωm ) = ε 0 ∑ ∑ χ ijk ( −(ωm + ωn ); ωma, ωn ) E j (ωm ) Ek (ωn )
(2) do
j ,k ( n ,m )
Cartesian component Cartesian components
third rank tensor makes discussion of permutation
tensor contraction symmetries easier. Harris uses (+) in ch.2, (-) in other chapters,
Boyd uses (+)
– “permutation notation”: sign convention
(n, m) ⇒ all distinguishable pairs of field components that yield the
same ωn + ωm
Why is this relevant? Keeps track of cross terms in quadratic form.
examples on following slide
EE346 NLO
1/13/21, #5A slide 8
Why are distinguishable permutations relevant?
Consider three cases:
• SHG with a single input polarization (Type I interaction, #4.5):
E(r, t ) = E X (ω )eiωt + c.c.
⇒ E(r, t ) 2 = E 2X (ω )ei 2ωt + c.c.
• SHG with two input polarizations (Type II interaction, #4.5):
E(r, t ) = E X (ω )eiωt + EY (ω )eiωt + c.c.
⇒ E(r, t ) 2 = E2X (ω )ei 2ωt + EY2 (ω )ei 2ωt + 2E X (ω )EY (ω )ei 2ωt + c.c.
• SFG with a single input polarization:
E(r, t ) = E X (ω1 )eiω1t + E X (ω2 )eiω2t + c.c.
⇒ E(r, t ) 2 = E2X (ω1 )ei 2ω1t + E2X (ω2 )ei 2ω2t + 2E X (ω1 )E X (ω2 )ei (ω1 +ω2 ) t + c.c.
• Cross terms from frequencies or polarizations lead to factors of two
– definition of susceptibility must keep track of these
EE346 NLO
1/13/21, #5A slide 9
Examples: SFG and SHG
Consider a particular tensor component: PNL is Z-polarized, E is X-polarized
PZ (ωn + ωm ) = ε 0 ∑ χ ZXX
(2)
( −(ωm + ωn );ωm , ωn ) E X (ωm ) E X (ωn )
( n ,m )
• Consider application to two examples:
– SFG: inputs ω1 and ω2, output: ω3 = ω1 + ω2
– SHG: input ω1, output ω3 = 2ω1
• SFG: PZ (ω3 = ω2 + ω1 ) = ε 0 χ ZXX
(2)
( −(ω2 + ω1 );ω2 , ω1 ) E X (ω2 ) E X (ω1 )
= by intrinsic permutation symmetry #5B
+χ ( −(ω + ω );ω , ω ) E (ωequal
(2)
ZXX 1 2 1 ) E X (ω2 )
1 2 X
= 2ε 0 χ ZXX
(2)
( −(ω1 + ω2 );ω1 , ω2 ) E X (ω1 ) E X (ω2 )
#5A.7 for CW : E(ω ) = Eω (r ) / 2, P(2ω ) = P2ω (r ) / 2 ⇒ Pω3 ,Z = ε 0 χ ZXX ( −(ω1 + ω2 );ω1 , ω2 ) Eω1 , X Eω2 , X
(2)
• SHG: PZ (ω3 = 2ω1 ) = ε 0 χ ZXX ( −2ω1 ; ω1 , ω1 ) E X (ω1 ) E X (ω1 )
(2)
= ε 0 χ ZXX
(2)
( −2ω1;ω1 , ω1 ) E X2 (ω1 ) indistinguishable
1 (2)
#5A.7 for CW : E(ω ) = Eω (r ) / 2, P(2ω ) = P2ω (r ) / 2 ⇒ P2ω ,Z = ε 0 χ ZXX ( −(2ω1 );ω1 , ω1 ) Eω1 , X Eω1 , X
2
• Note factor of two difference in PNL for SHG vs SFG (consistent with #5A.8)
– note: for same total input power, SFG in limit ω2 → ω1 produces
same P(2ω) as SHG (P.S. 2)
EE346 NLO
1/13/21, #5A slide 10
Nonlinear Coefficient, d
• Historically, another convention was widely used
– the nonlinear coefficient d defined in the time domain
implicitly assumes d is non-dispersive not a bad approxin many cases
Pi (t ) = ε 0 ∑ 2dijk E j (t ) Ek (t ) ⇒ PZ (t ) = ε 0 2d ZXX [ E X (t ) ]
2
if again consider ZXX case:
j ,k
• For two monochromatic input fields at frequencies ω1 , ω2
2
1
PZ (t ) = ε 0 2d ZXX
2
(E ω1 X ) (
eiω1t + Eω*1 X e − iω1t + Eω2 X eiω2t + Eω*2 X e − iω2t )
d ZXX
= ε0 Eω21 X ei 2ω1t + 2 Eω1 X Eω2 X ei (ω1 +ω2 ) t + ...
2
1
PZ (t ) =
2
( P2ω Z ei 2ωt + P2*ω Z e − i 2ωt + ...)
⇒ P2ω1Z = ε 0 d ZXX Eω21 X ; P(ω1 +ω2 ) Z = ε 0 2d ZXX Eω1 X Eω2 X
d I2 ≡ dSHG ≡ dSFG
Tabulations are usually of d = dSHG
some authors incorporated factors of 2 into definition of d: dijk = 2dijk (not usually done)
SFG SHG
EE346 NLO
1/13/21, #5A slide 11 χ(2) vs d
• Nonlinear coefficient d is important for historical reasons
– most tabulated data is in terms of d for SHG
easiest quantity to measure experimentally
– good reference on conventions and measurement of χ(2):
Stewart Kurtz, Ch. 3, in Quantum Electronics Vol. 1, H. Rabin and C.L. Tang eds., A.P. (1975)
(2)
• Relationship to χ(2): χ ijk (−2ω ; ω , ω ) = 2dijk
SHG
– χ(2) is weakly dispersive, so same value often used for other
processes (and see discussion of Kleinman’s symmetry, #5B)
– can use Miller’s ∆ to estimate at other frequencies (see #5A.13)
• Note that Yariv uses yet another form of constitutive relationship
– he absorbs ε0 into d: PNL = d Yariv E 2
– there seems no good reason to do this
dimensions of conventional d = [m/V] while d Yariv= [mess]
be careful when using results for NLO from his books
EE346 NLO
1/13/21, #5A slide 12
Return to CEO version of χ(2)
• With these definitions, the CEO calculation for anharmonic oscillator (#5A.5)
⇒
DNe3
χ (−2ω ; ω , ω ) =
(2)
ε 0 m (ω0 − ω ) + iωγ ω02 − (2ω ) 2 + i 2ωγ
2
2 2 2
note: in Harris Eq. 3.79
e3 / m3 → e3 / (ε 0 m 2 )
• Can rewrite in terms of linear susceptibilities (#4.15):
typo
ε mD
2
χ (2) (−2ω ; ω , ω ) = 0
2 3
χ (1) (2ω ) χ (1) (ω ) 2
N e dispersion largely accounted for
by linear susceptibilities (#4.15)
much less dispersive than χ(2)
q2
hard to measure X for large rangeof w ω resonant
2ω resonant
Usually operate well away from resonances (2ω << ω0)
to avoid absorption
⇒ slow, smooth dispersion of nonlinear response expected
0 0.5 1 1.5
ω / ω0
EE346 NLO
1/13/21, #5A slide 13
Miller’s Delta
• Generalizing to other interactions and anisotropic media (P.S. 2):
χ ijk(2) (−(ω1 + ω2 ≡ ω3 ); ω2 , ω1 ) =
− DNe3
ε 0 m 2 (ω02i − ω32 ) + iω3γ i ω02 j − ω22 + iω2γ j ω02k − ω12 + iω1γ k
• Can rewrite in terms of linear susceptibilities:
ε 02 mD
χ (−(ω1 + ω2 );ω2 , ω1 ) =
(2)
ijk 2 3
χ ii(1) (ω3 ) χ (1)
jj (ω2 ) χ kk (ω1 )
(1)
N e
• Miller’s Delta: χ ijk(2) (−(ω1 + ω2 );ω1 , ω2 )
∆ ijk ≡ (1) less dispersive
χ ii (ω1 + ω2 ) χ (1)jj (ω1 ) χ kk (ω2 )
(1)
found to vary orders of magnitude less than χ(2) between materials
– can use to extrapolate χ(2) measured at one wavelength to another
“constant Miller delta scaling”
using X not'Iririaldifferences
phenomenological
for
pp
