PHYSICAL REVIEWA VOLUME 52, NUMBER 6 DECEMBER 1995
Cavity enhancement
of
resonant frequencies
in semiconductor
lasers
subject
to optical injection
T.
B.
SimpsonJAJ'COR,
P
0
Box.85154,San Diego, California 92186-5154J.
M.
LiuDepartment
of
Electrical Engineering, Universityof
California, LosAngeles, LosAngeles, California 90095-159410K. F.
Huang andK.
TaiDepartment
of
Electro Physic-s, National Chiao Tung University, Hsinchu, TaiwanC. M.
Clayton, A. Gavrielides, and V.KovanisNonlinear Optics Center, Phillips Laboratory, Kirtland AFB,New Mexico 87117-5776 (Received 21July 1995)
The injection of an optical signal into a semiconductor laser biased near or above the lasing threshold modifies the coupling between the free carriers and the intracavity field. The detuning between the frequency of the injected signal and the free-running oscillation frequency and the ratio of the photon lifetime to the carrier lifetime are key parameters in determining the enhancement of the carrier-field resonant coupling frequency and the stability ofthe output field. Experimental results using a vertical cavity surface emitting laser biased near threshold are in agreement with calculations using alumped-element oscillator model.
PACS number(s): 42.55.Px, 42.65.Hw, 42.65.Ky, 42.70.Nq
Semiconductor lasers subject to external optical injection are being studied asamodel nonlinear dynamical system and for their potential in optical communication and processing applications. External optical signals can induce stable and unstable injection locking
[1],
chaotic dynamics and multi-wave mixing[2,
3],
and mode hopping[4],
depending on the amplitudesof
the injected and oscillating fields and the fre-quency offset between them. These various characteristics have all been recovered in lumped-circuit oscillator modelsof
the semiconductor laser subject to external injection. Re-cently, an amplifier model was used to describe some novel features in the optical spectraof
a vertical cavity surface emitting laser(VCSEL)
under strong optical injection[5].
The model predicted that stimulated emission and absorption due to the coherent transfer
of
energy significantly enhanced the semiconductor response and produced new resonances in the optical spectrum that are distinct from the relaxation resonances typically observed in semiconductor lasers. Here, we show that the laser cavity plays a major role in the gen-erationof
the new resonances and in the enhancementof
the modulation bandwidth, and that the new resonances can be related back to the relaxation resonances observed in free-running semiconductor lasers.In our analysis we will consider a single-mode semicon-ductor laser subject to external optical injection
[3,
6,
7].
We will consider two external fields, a strong injection field that satisfies the conditions for injection locking, and aweak field that can be used as a linear probe. The interaction can be described by two coupled equations:dA
y,
dt
2 A+
t(Co—
to )AdN
J
2epn(2)
coL cop
—
—
b 6/2—
bU+
V.Here, cop is the optical frequency
of
the free-running laser,U=
r/IAt/Az~cos@z andV=
r/IA, /Az~sin@z, where Az is the steady-state amplitudeof
the injection-locked circulating field. The gain defect6=
y,
—
I
gp, where gp is the steady-state gainof
the free-running laser, can be important when the free-running laser is biased below threshold. The free-running coherent field amplitude, Ap, and AL are relatedthrough
where A is the total complex intracavity field amplitude at the locked oscillating frequency
col,
y,
is the cavity decay rate, co, is the longitudinal mode frequencyof
the cold laser cavity, j. is the confinement factor, b is the linewidth en-hancement factor, g is the gain coefficient, A& and A; are theamplitudes
of
the strong injection signal at the locking fre-quency and the weak probe signal at cuL+A, respectively,y
isthe coupling rate, N is the carrier density,J
isthe injec-tion current density, eis the electronic charge, d is the active layer thickness,y,
is the spontaneous carrier decay rate, and n is the refractive indexof
the semiconductor medium. The gain is assumed to obey alinear dependence ondeviationsof
both the carrier density and the circulating photon density about steady-state values
[7].
Steady-state values for the circulating field are obtained by setting the derivative terms and A; equal to zero. The locked phase
of
the oscillating field,@z,
relative to the phaseof
the injection field, is obtained through the relationI
+
—
(1
—
ib)gA+
r/(At+A;e
'0'),
l
—
Xy,
/yI
iAO/Azi
=
(4)
'~~ ' ll ~IIII]~I 1 ~ If'llll ~
I IIII ii
52 CAVITY ENHANCEMENT OF RESONANT FREQUENCIES
IN.
. . R4349where the factor,
X,
is given by30
Here,
(2U
—
8)
y„L,X=
(y,
—
2 U) y 1.+
y ylL.20
lO C10
2n2Ep2'
2Epr.
L=
&
IAil'g.
andr,
L=
—
& IALI'I
g,
, (6)AcoL flCOL
where
g„and
g~ are the differential gain and the nonlinear gain parameters defined as the derivativesof
g with respect to the carrier density and the photon density, respectively, evaluated at the operating point[7].
When the free-running laser is biased below threshold, lAo/A~l=0.
With these definitions, we can find the perturbations in-duced by the weak injected field. The weak field induces a regeneratively amplified (RA) sideband,
A„,
at the injection frequency,cul+II,
and a four-wave-mixing (FWM) side-band,Af,
at ~L—
A.
In the perturbation limit:30
I
20
'U 05 C10
-6 -2 Offset Frequency (GHz}(a)
A„
—
=
—
((y, +
y„~
—
if')[U
—i(II —v)]
+
—,'(1+ib)(y,
y„L+
y,
y I,—
2Uy„l
—
iAy„L)),
-6
-2 Offset Frequency (GHz)(b}
Af
A;(7)
D, {2(I
1—
ib)(y,
y.
l.+
y, y„l.
2Uy.
L+i&—y„L))/,(8)
FIG.
l.
(a) Optical spectrum ofthe free-running VCSELbiased well above threshold and (b) the regeneratively amplified spectrum due tothe injection ofaweak optical probe. The experimental data are compared with the calculated spectra using the parameters listedin the text.
D=(y,
+
y„L—
iA)[(U
—
iA)(U+
y„L—
iA)
+
v(v+by„L)]+(U+bv
En)(r,
2U
— y„i)
r.
—
i.
—
(9)
Measured optical sidebands are proportional to the squared magnitudeof A„and Af
. In addition, the spontaneous emis-sion noise spectrum is proportional to the sumof
the squared magnitudesof A„and
A&[8].
This is because thespontane-ous emission effectively acts as a weak external optical source. Therefore, this modeling also gives the sideband spectra due to amplified spontaneous emission and, for sim-plicity, we analyze the problem from this viewpoint. These equations simplify to the appropriate expressions for the free-running semiconductor laser when U
=
V=
0
andAL
=Ap,
y„L= y„,
and y~L=y~,
the respective free-running values[7].
Under optical injection,D
can be0
and unstable and chaotic dynamics can ensue[2,3,
9].
Here, we are primarily interested in the offset frequenciesof
the reso-nances and defer considerationof
their stability.When the free-running laser isbiased below threshold and is subjected to a near-resonant locking field, the limit lAo/A~l(&1, and
y,
&)2U andy,
, thenU=8/2+
y,
y„L,/2
y,
. Inthis limit,Eqs. (7)
and(8)
can be solved to show that the RA term dominates the FWM term.If
the gain defect,6', is large compared to
y,
and V, the central peakof
the RA spectrum is shifted from the free-running spectrum. The dif-ference between the two in the weak-locking, large-6' limit has maxima and minima shifted from the free-running fre-quency by—
y, [1/b
~
( I+
I/b )'],
the same as predictedby the amplifier analysis
[5].
However,Eq. (7)
predicts that the amplifier analysis is inadequate when the free-running laser is near or above threshold and/or when V becomes comparable to6.
As the injecting field isincreased sothat U and V dominate the other rates, except fory,
, the RA termcontains a strong resonance at
A
=
—
V=
~p—
~L
—
by,
y„L/2y, . The corresponding resonance at positivefre-quencies ismuch weaker due to a canceling term,
A
—
V, in the numerator. The magnitudeof
the shift is proportional to the circulating locked power, due to the coherent field-induced carrier decay rate, y I . Unlike the amplifier analy-sis, there is a multiplicative factorof
approximatelyby,
/2y, for the resonance shift. This factor is typically on the orderof
100—1000
in semiconductor lasers. Wheny,
&&y,
, alarge shift in the resonance can be observed evenif y.
&y.
I.
To verify these predictions, we have investigated a
VCSEL.
Manyof
the characteristicsof
thisVCSEL
havebeen described previously
[10,
11].
All measurements re-ported here were made at low output power, under operating conditions where theVCSEL
displayed single-mode opera-tion and no significant transverse profile variations from ef-fects such as thermal lensing or spatial hole burning. Figure 1shows the output spectrum and the RA spectrum when the laser isbiased at4.
8mA, well above the threshold valueof
4.
2 mA. Total output power at this injection current is=0.
35
mW and the coherent output power is=0.
3 mW. Here, we can determine the key dynamic parametersof
the laser from its spectra. Figure 1shows the good agreement between data and model that isachieved using the determined parameters,R4350 T,
B.
SIMPSON etal. 0.015 0.012 o 0.008 0$ C Ul 0.004III«I rr v- w ««r«re .. «na «rrr«e-~a«raw'«r.en
-10 -5
Offset Frequency (GHz)
b=6,
y=5.
5&&10"s',
y=5&10
s',
y=3.
5PX10
s',
and y=
5.3P
X10
s'.
P
in the formulas refers to the coherent output power in milliwatts, and the uncertainty for the parameters is~20%.
The enhancement factor for the frequency shifts, relative to the amplifier case, is approxi-mately330.
For
the injection-locking measurements, the bias current was set to3.9
mA and the output from a tunable, narrow linewidth, low noise, external cavity laser (New Focus Model6126)
was injected into theVCSEL.
Output power from the free-runningVCSEL
was=0.
02
mW. Below threshold operation insured that lAo/AL &&1.Here, we con-centrate on optical injection at the free-running frequency. Consistent results were obtained for detuned injection. Fig-ure 2shows free-running and injection-locked optical spectra taken with ahigh finesse optical spectrum analyzer (Newport ModelSR-240C)
with=60
MHz resolution. The resonance feature shifts to lower frequencies as the injection power increases.Features from these spectra can be compared with the predictions
of
the model. The injection-locked output power0.08 o 005 CL. CL 0.04
0
0.02 o 12 N K CQ 8 O 4 0 0.1 1 10 100Injected Power (arb.units)
0 1000
FIG.3.The dependence ofthe injection-locked power, ( )data
and (
—
)model, and the shift ofthe resonance from the free-running frequency,(6)
data and (——
) model, for the VCSELbiased just below threshold when subject to an injected signal at thefree-running oscillation frequency.
FIG. 2. Representative optical spectra of the free-running VCSEL biased just below threshold (a), and under increasing in-jected power from a narrow-band laser tuned to the free-running oscillation frequency (b)—(e). The injected power ratio of (b):(c):(d):(e)is
1:10:100:1000.
L
r,
r.
l.+
r,
r„l.
+
(r,
—
r.
L)U,(10)
YrL
Ys+
YnL+ YpL,+
U.Both A„Land Y,
I
show the direct modification due to the enhancementof
the circulating coherent power under injection-locked operation. In addition, there are the terms proportional to U. These terms dominate the changes over a wide rangeof
injection levels. More generally, when$1
40,
there is a complicated dependence on both U and V which can lead to the destabilizationof
the laser as well as the enhancementof
the modulation bandwidth. It is this complex coupling, involving both phase and amplitude modifications due to the strong injection-locking field, that dominates the shift in the resonance frequencies and their damping characteristics, not a simple enhancementof
stimu-lated emission and absorption rates.The lumped-element analysis
of
laser oscillation assumes that spatial effects can be averaged over the mode profiles. It has given excellent quantitative agreement with awide vari-etyof
single-mode semiconductor lasers, including Fabry-Perot edge emitting lasers with large output coupling[3,
7].
The linearized treatment, like that given here, fails to accu-rately reproduce the central linewidth
of
the free-running la-ser and, more generally, the dynamics whenever aresonance can be determined from the spectra and the measured output power. Figure 3 shows the dependenceof
the injection-locked coherent power and the frequency shiftof
the reso-nance on the power injected into theVCSEL.
Because we are unable to independently determine the coupling param-eter, rg, we can only make a relative measurementof
theinjection power, and the relative uncertainty is
50%.
Using the experimentally determined parameters, model calcula-tions for both the injection-locked coherent power and the frequency shift are in good agreement with the data. At high injection power levels, the offset frequency scales linearly with the injection-locked coherent power, and both scale with the cube rootof
the injected power. Even at the highest injected power measured, where the injection-locked coher-ent power is=
0.
07
mW, the field-induced enhancementsof
the decay rate are
Y„L=0.
05Y,
andY„I=0.
075y, .
These values coincide with the expected values at the lower oper-ating power.The modulation and noise spectra
of
the injection-locked laser aremore complicated than the free-running spectra due, in part, to the dependenceof
the field amplitude on the locked phase.If
the locking frequency is detuned from the free-running frequency so that@L=0,
the amplitude be-comes decoupled from the phaseof
the locking field, as is the case in the free-running laser. The correspondence be-tween the new resonance peaks and the resonance peaksof
an above threshold, free-running laser is more direct. The resonance peaks
of
the free-running laser are determined by the relaxation resonance frequency,A„=(y,
y„+
y,
y„)",
and the damping rate,y„=
y,
+ y„+
y„[7].
For the$1
=
0
locked laser,52 CAVITY ENHANCEMENT OF RESONANT FREQUENCIES
IN.
. . R4351becomes unstable, but it shows good accuracy in the predic-tion
of
the positions and shiftsof
the resonance frequencies. The laser cavity strongly enhances the frequency shifts in-duced by the injection field beyond what is expected from the increased stimulated emission due to the stronger oscil-lating field. Phase and amplitude characteristics must beana-lyzed for adetailed quantitative understanding
of
the spectral features.The authors would like to thank Dr. Tim Day forthe loan
of
the New Focus tunable laser. The workof
T.
B.
S.
andJ.
M.
L.
was supported by theU.
S.
AirForce's
Phillips Labo-ratory under Contract No.F29601-94-C-0166.
[1]
I.Petitbon, P. Gallion, G.Debarge, and C.Chabran, IEEEJ.
Quantum Electron. 24, 148(1988).
[2] J.
Sacher, D.Baums, P. Panknin, W. Elsasser, andE.
O.Gobel, Phys. Rev.A45, 1893(1992).[3]
T.B.
Simpson,J.
M.Liu, A. Gavrielides, V.Kovanis, and P.M. Alsing, Phys. Rev.A51, 4181(1995).
[4]
P.Schanne, H.J.
Heinrich, W.Elsasser, and E.O.Gobel, Appl. Phys. Lett. 61, 2135 (1992).[5]
C.W. Lowry, F. Brown de Colstoun, A.E.Paul, G, Khitrova, H.M. Gibbs,J.
W. Grantham, R. Jin, D. Boggavarapu,S.
W. Koch, M. Sargent III, T.M. Brennan, andB.
E.
Hammons,Phys. Rev. Lett.
71,
1534(1993).
[6]
T.B.
Simpson andJ.
M.Liu,J.
Appl. Phys. 73,2587(1993).
[7] J.
M. Liu and T.B.
Simpson, IEEEJ.
Quantum Electron. 30,957 (1994).
[8]
T.B.
Simpson andJ.
M.Liu, Opt. Commun. 112,43(1994).
[9] J.
R.Tredicce, F.T. Arecchi, G.L.Lippi, and G.P. Puccioni,J.
Opt. Soc.Am. B2, 173(1985).