Phase-Locked Arrays of Semiconductor Diode Lasers

Phase-Locked Arrays of Semiconductor
Diode Lasers
Dan Botez and Donald E. Ackley
Abstract
Semiconductor diode lasers can be combined into monolithic phase-locked
arrays that operate to continuous-wave (cw) powers in excess oflW
and
exhibit well-defined output beams to hundreds of mW. Applications of
arrays include high-speed optical recording, high-speed printing, free-space
communications, and efficient pumping of solid-state lasers. We review the
current state of phase-locked array design with emphasis on the means for
achieving fundamental mode operation. Models for understanding and
predicting the behavior of phase-locked arrays are discussed, and the poten­
tial of various array structures to operate in a single,
diffraction-limited
output beam is evaluated.
Introduction
In the quest for higher and higher powers from
diode lasers, one major approach has been the use of
one-dimensional arrays of mutually coupled diode la­
sers, which are generally referred to as phase-locked
arrays. Such arrays can operate as a single coherent
source of one to two orders of magnitude wider than
the standard single-element devices. Since in semi­
conductor lasers, failure rate is directly related to the
optical flux density, an array provides significantly
greater reliability at a given output power than a
single-element device. However, making the array
operate in phase so as to produce a single, narrow
beam is no trivial task. In this paper, we present a
review of phase-locked-array work to date, as well as
discuss prospects for the future.
There is substantial interest in phase-locked arrays
for applications that require high optical power from
a semiconductor laser source. While single-element
(fundamental-mode) diode lasers are limited in power
to outputs of the order of 30 mW, arrays of diode
lasers can operate with outputs of hundreds of milli­
watts. These high output powers will be useful for
optical recording at high data rates, and high-speed
laser printers. Arrays have also been proposed for use
in free-space communications links, both in terrestrial
and outer space applications. For all these applica­
tions, it is necessary that the operating modes of the
array be controlled so that the optical output is a
single-lobed, nearly diffraction-limited beam. Arrays
will also be useful to replace flashlamps for pumping
solid-state lasers (such as Nd:YAG) in industrial, mili­
tary, and medical applications because of their high
efficiency and improved reliability. In pumping solidstate lasers, the beam quality of the arrays will not be
important, but the narrow spectral width and high
radiance of the arrays should lead to extremely effi­
cient pumping. One other potential use of phaselocked arrays is in short-haul local area networks
(LANs) where a large amount of power coupled into
8
8755-39%/86/OK
the fiber is an important consideration. In general,
because of the need for high-power, compact and effi­
cient, fundamental-mode diode laser sources, the de­
velopment effort on phase-locked arrays will continue
to expand over the next few vears.
Coupling of a pair of lasers was studied experi­
mentally [1] as early as 1970 and theoretically [2] as
early as 1972. The first operation of a phase-locked
array of diode lasers (five elements) was reported
in 1978 by D. Scifres et al. [3] of Xerox Palo Alto Re­
search Center. The early work involved so-called
gain-guided lasers, that is, devices for which lateral
optical-mode confinement is solely provided by the
injected-carrier profile [4]. As a consequence, the
optical-mode confinement is very weak and thus
vulnerable to changes in drive current and/or changes
in drive conditions [5] (i.e., low-duty-cycle, pulsed
operation versus cw operation). By contrast, devices
with "built-in" dielectric waveguides, so-called indexguided lasers [4], can have stable operation under all
drive conditions. In fact, index-guided laser struc­
tures are solely responsible for the realization of
single-mode lasers, that is, diodes operating in a sta­
ble, fundamental spatial mode over wide ranges of
output power.
However, gain-guided devices are relatively easier
to make than index-guided devices. Thus, not until
1981 were there reports [6], [7] of phase-locked arrays
of index-guided lasers, and then only in low-dutycycle, pulsed, room-temperature operation [6] or in
cw operation at 77 Κ [7]. Meanwhile, advances in
m e t a l - o r g a n i c v a p o r - p h a s e epitaxial ( M O V P E )
growth had allowed the fabrication of good-quality
quantum-well heterostructures, which by virtue of
their inherently low threshold-current densities, per­
mitted the research team at Xerox Laboratories to
achieve hundreds of milliwatts of cw power from
gain-guided arrays [8]. Yet, being gain-guided, these
powerful arrays had poor beam quality, and, thus, it
was quite difficult to make correlations with theory
[9]. The realization in late 1982 of the first roomtemperature cw-operating array of index-guided la­
sers [10] marked two important milestones: (1) the
identification of the out-of-phase mode of operation
[4] as the favored one for conventional phase-locked
arrays, and (2) the first report of single-longitudinalmode operation from phase-locked arrays. Ever since,
most efforts for obtaining arrays operating in a stable,
narrow beam have been concentrated on indexguided array structures [10]-[21]. Below, we describe
the major types of arrays, the coupled-mode fora.00 © 1 9 8 6 I E E E
IEEE CIRCUITS A N D DEVICES M A G A Z I N E
malism applied to arrays, and the best results to date
in achieving stable operation in a single narrow beam,
the so-called fundamental array-mode operation.
Major Array Types
In Fig. 1, we schematically show various array
types. The structures can be made either by using the
AlGaAs/GaAs material system (emitting wavelength,
λ = 0.73 - 0.88 μτη) or the InGaAsP/InP material
system (λ = 1.20 - 1.60 μτη). For simplicity, we in­
dicate only the layers' conductivity type and their
respective function. First displayed is the simplest
device, that is, the gain-guided array (Fig. l a ) , for
which the current is confined to a group of parallel
contact stripes, and for which the optical modes are
controlled laterally by the injected carrier profile. In
most gain-guided arrays, the current is confined via
high-resistivity proton-implanted regions [8], [22] or
Schottky-barrier defined stripes [5]. Coupling be­
Fig. 1
Schematic representation
of phase-locked arrays composed of
gain-guided lasers (a), and index-guided lasers [(b)-(f)] fabricated
by: (a) preferential proton implantation; (b) chemical etching of
ridgetype waveguides; (c) preferential p-type-dopant
diffusion;
(d) metal-organic vapor-phase epitaxial (MOVPE) growth over
channeled substrates; (e) liquid-phase epitaxial (LPE) growth over
channeled substrates; and (f) quantum-well-structure
disorder­
diffusion. With the
ing induced by preferential n-type-dopant
exception of the last one, which has been demonstrated only for
AlGaAs/GaAs
devices, the structure can be either of the
AlGaAs I GaAs type (i.e., GaAs active layer and AlGaAs cladding
type (i.e., InGaAsP active layer
layers) or of the InGaAsP/InP
and InP cladding layers).
JANUARY 1986
tween lasers occurs when the evanescent fields of
adjacent devices substantially overlap (see Fig. 2).
The rest of the arrays are of the index-guided type,
that is, structures with built-in dielectric waveguides
for lateral optical confinement. Chronologically, the
first array structure of evanescently coupled indexguided lasers is of the ridge-guide type (Fig. l b ) , a
configuration that is generally obtained by chemical
etching into planar DH material, and provides mode
confinement mainly due to the variations in claddinglayer thickness [7], [15], [16], [18]-[20]. Next was the
Zn-diffused device [10], obtained by Zn diffusion into
the channeled top of a planar DH structure. The dop­
ant reaching the active layer locally lowers the index
of refraction such that each nondiffused area repre­
sents the high-index region of a lateral dielectric
waveguide. Two other index-guided array types de­
pend on the crystal growth properties of MOVPE and
liquid-phase epitaxy (LPE) over channeled substrates
(Figs. Id and l e , respectively). When using MOVPE,
the substrate channels are identically replicated and a
constant-thickness active layer with periodic bends is
obtained [23]. The bends in the active layer create
lateral mode control [24]. When using LPE, channels
readily fill up such that a nonuniform-thickness
cladding layer is achieved [4]. Waveguides are created
in the vicinity of the channel by the proximity of the
substrate on both sides of the channel [11], [13], [17],
[20], [21].
A very recent approach to making index-guided
arrays exploits the property of diffusion of dopants
such as Si to induce lattice disordering in laser struc­
tures containing alternate thin layers (<200 A each) of
GaAs and AlGaAs material [25], so-called quantumwell lasers. Upon diffusion, the multilayered stack is
transformed into a single layer of material with band-
Fig. 2
Schematic representation of near-field amplitude profiles for an
index-guided
array with elements
of the same
peak-field
amplitude. The interelement spacing is S, and each individual
field-intensity profile of waist w
emitter has a Gaussian-shaped
(Ref. [36]). The two operational modes shown correspond to
adjacent devices being in phase and out of phase,
respectively.
0
9
gap and index of refraction intermediate between
those of its original constituent layers [25]. Just as in
the case of the Zn-diffused array, areas not subjected
to diffusion represent the high-index regions of lateral
dielectric waveguides, and thus by preferential Si dif­
fusion, an array of index-guided devices is formed
[26] (see Fig. If). For simplicity, we refer to such de­
vices as impurity-disordered arrays.
A simple representation of how coupling occurs in
phase-locked arrays of diode lasers is shown in Fig. 2.
We consider five elements of the array fabricated by
LPE growth over channeled substrates (Fig. l e ) . Each
element defines laterally a fundamental mode, and,
for simplicity, the individual modes are assumed to
have the same peak-field amplitude. The fields from
adjacent devices couple when they overlap (in re­
gions where there are index-of-refraction variations),
and thus provide coherent excitation of each element
in the array. Strong coupling occurs if the array
elements have the same resonant frequency. One
coupling configuration is when all the elements are in
phase, so-called 0°-phase-shift operation because
there is no phase difference between adjacent ele­
ments. Then, the optical-fields sum in phase in the
space between devices and the near-field intensity
profile is nonzero across the array. Another possible
configuration is when adjacent devices couple out of
phase, so-called 180°-phase-shift operation, because
of the π phase difference between adjacent devices.
In this case, the fields destructively interfere between
the stripes and the array near-field display nulls. As
shown below> other coupling configurations may
exist, but the two basic ones are 0°-phase-shift and
180°-phase-shift.
Typically, arrays have 8 to 10 e l e m e n t s . Tenelement gain-guided arrays have demonstrated cw
powers as high as 850 mW [22]. Notable exceptions
are 40-unit arrays of gain-guided devices, which have
achieved record-high cw powers: 1.6 W for con­
ventional half-wave coated devices [22] and 2.6 W for
devices with optimized facet coatings [27], although
with poor beam quality. Index-guided arrays have
been the only devices providing well-defined and
drive-condition independent beams; with the best
results to date being single, forward-looking beams
of 200 mW peak-pulsed power [11], and 100 mW
cw power [20].
Coupled-Mode Formalism:
The "Array Modes"
The first analysis of phase-locked arrays was pub­
lished by Scifres et al. [9]. They interpreted their ex­
perimental data by considering the diffraction pattern
from a uniformly illuminated grating with equally
spaced slits corresponding to individual laser-array
elements. This is the so-called simple diffraction the­
ory. While simple diffraction theory has proved use­
ful in interpreting some experimental results [5], [6],
10
[9], it provides no method for describing the allowed
oscillating modes of an array of coupled emitters.
The first application of the coupled-mode theory to
phase-locked arrays was done by Otsuka [28]. How­
ever, Otsuka analyzed an array with an infinite num­
ber of emitters and of infinite aperture and found
only two modes of operation: 0°-phase-shift and
180°-phase-shift, which is no different from what had
been claimed by proponents of the simple-diffraction
theory. Butler et al. have reported [29] and published
[30] the first coupled-mode analysis for an array of Ν
coupled, identical elements. They found that an array
of Ν emitters has Ν normal modes or eigenmodes,
which they chose to call array modes. Each array
mode has a field-amplitude configuration that is non­
uniform across the array. All previous work had as­
sumed that the element amplitudes are uniform
across the array, a case that was shown recently [31]
to be correct only for arrays of two elements or an
infinite number of elements. By using the coupledmode formalism, Butler et al. were able to interpret
many published experimental results. Shortly after,
other research groups published independent work
[32], [33] on coupled-mode formalisms applied to
arrays. Recent detailed experiments by Paoli et al.
[34], Epler et al. [35], and Temkin et al. [19] have
confirmed that coupled-mode analysis best describes
phase-locked-array behavior.
The main findings of the coupled-mode analysis are
schematically depicted in Fig. 3 for an array of nine
elements. By solving an eigenvalue equation, it is
found that for an array of Ν emitters, there are Ν
allowed modes, the so-called array modes. For each
array mode, the near-field element amplitudes vary
across the array. There is a succession from the funda­
mental array mode, L = 1, the 0°-phase-shift mode
defined in Fig. 2, to the last high-order array mode,
L = N, the 180°-phase-shift mode defined in Fig. 2.
Shown with a dashed line are the envelope functions
of the array-mode near-field amplitude profiles,
which correspond to the modes of an infinite poten­
tial well of width (N + 1)S. These box modes are sym­
metric or antisymmetric with the mth (high-order)
mode having m - 1 field nulls. Since the individual
array elements support only fundamental modes, the
maximum possible number of field nulls occurs when
all adjacent elements are out of phase; that is, Ν - 1
nulls for a N-element array. This sets an upper limit of
Ν for the number of box modes of the (N + 1)S wide
(array) aperture, and explains why Ν is the maximum
number of allowed array modes.
The far-field beam patterns corresponding to each
array mode are shown in Fig. 3b. For an array of Ν
coupled, identical elements, the far-field intensity dis­
tribution is given by [9], [30]
F(0) = |E(0)| i(0)
2
(1)
where θ is the angle with respect to the normal to the
IEEE CIRCUITS A N D DEVICES M A G A Z I N E
Fig. 3
Near-field amplitude profiles (a) and far-field patterns (b) for three of the nine operational modes, so-called array modes, of a nineelement array. L is the array-mode number and λ is the free-space wavelength. The dashed curves in (a) are the envelope functions of the
near-field amplitude profiles; and the dashed curves in (b) correspond to the individual-element far-field
pattern.
facet, E(0) is the far-field amplitude distribution for
one of the laser-array elements, and Ι(θ) is a function
characterizing the array and representing the inter­
ference effect of the coupled emitters in the array, the
so-called grating function. For the coupled-mode the­
ory, the array-grating function originally published by
Butler et al. [30] has been recently reduced to a signifi­
cantly simpler expression [31]:
. J ( N + l)u
7(0) = I (u)
oc
L
s i n
[ — 1 —
_sin*(|) -
\E(x)\
2
Tj
+
= E e x p ( - 2 x /wl)
20
where w is the Gaussian waist. (The functional re­
lationship between w and the dielectric guide width
and index step can be found in Ref. [36].) Then, the
condition that the nearest off-normal lobes of the
L = 1 far-field pattern peak below the 1/e point in
intensity of the main lobe is
0
2
Ν
(1)
with u = k S sin 0, where k is the free-space wavenumber (k = 2π/λ), and S is the spacing between
two array elements. 7(0) is shown in Fig. 3b with a
solid curve and represents the array pattern for the
case when the array elements are point sources. In
reality, the array elements have a finite size that pro­
vides an individual far-field intensity distribution
|E(0)| , and thus the array pattern is a product of
|E(0)| and 7(0). From a practical point of view, the
desired beam pattern is that for which the energy is
2
S
0
2
2
JANUARY 1986
(2)
2
0
sin (^^)J
0
2
Lii
L = 1,2
0
primarily contained within a single, forward-looking
beam, that is, the L = 1 pattern with most of the light
concentrated in the central lobe. That imposes re­
strictions on the individual-element spot size. The
restrictions can be easily found by considering that
the individual-element near-field intensity profile
\E(x)\ is Gaussian, a very good assumption for most
waveguides of practical interest [36].
iv ^ —
(3)
0
7Γ
that is, the element near-field full width, 2w , has to
be equal to or more than 64 percent of the interelement spacing S. Technologically, this is a nontrivial
condition to fulfill, which explains why only a few
arrays to date have been operated in a single, forwardlooking beam [11], [13], [16]-[18], [20], [21].
0
11
Even the few arrays that have operated in a single
beam generally have beamwidths wider than the
diffraction-limited beamwidth. That is because arrays
can operate simultaneously in several array modes
(e.g., L = 1 plus L = 2) and thus provide beams two
to three times wider than the "diffraction limit." The
diffraction-limited beamwidth (full width at half
power) is loosely defined to be the ratio of the wave­
length over the emitting aperture, i.e., λ / N S . A close
look shows that the diffraction-limited beamwidth is
a function of the element number, Ν [31]. For arrays
of few elements (2-4), the lobe beamwidth is virtually
the same as λ / N S . Arrays of 10 to 40 diodes have
beamwidth values in the 1.09-1.16 λ / N S range. In
the limit Ν —» °°, for a finite aperture, the beamwidth
tends to 1.19 λ / N S , as expected for a source with a
raised-cosine amplitude profile.
Gain-Loss Considerations
in Uniform Arrays
The coupled-mode analysis presented above con­
siders only arrays of coupled dielectric waveguides.
Thus, there is no concern for optical gain or losses
in the structure. For lasers, however, one has to
take into account gain-loss spatial distributions, and
how they affect which array mode reaches first lasing
threshold. Two typical cases are shown in Fig. 4. The
most common case is that for which there is loss be­
tween adjacent elements (Fig. 4a). The loss may be
due to absorption in unpumped regions (gain-guided
arrays) [3], [5], [22]; absorption to the substrate (arrays
grown by LPE over channeled substrates) [11], [13],
[17], [21]; or absorption in heavily doped interelement
regions (the Zn-diffused [10] and impurity-disordered
[26] arrays). Then, the 180°-phase-shift mode, by vir­
tue of less field overlap with the lossy interelement
regions (see Fig. 2) should have lower lasing thresh­
old than the 0°-phase-shift mode [10]. Theoretical cal­
culations for index-guided arrays [33], [37], [38] have
indeed shown that in structures with net loss between
elements, the 180°-phase-shift mode has a higher net
single-pass gain than the 0°-phase-shift mode. These
analyses explain why most arrays tend to operate in
the 180°-phase-shift mode. However, if the field over­
lap between adjacent devices is poor, the discrimi­
nation between the in-phase and out-of-phase modes
is weak [33], and then single-beam operation is some­
times obtained [11], [13]. Very recently, G. P. Agrawal
has published [55] a self-consistent analysis for both
gain- and index-guided arrays; an analysis that incor­
porates the effect of the injected carriers on array be­
havior. Agrawal's analysis shows that even when the
gain is uniform across the array, index-guided devices
can lase in the 0°-phase-shift mode and, thus, may
well explain reported single-beam operations from
certain types of uniformly pumped index-guided
arrays [11], [13], [20].
The other case considered is when there is more
gain between elements than in the elements (Fig. 4b).
12
Now the 0°-phase-shift mode will be favored to lase
since it has more field in the interelement regions (see
Fig. 2) than the 180°-phase-shift mode. Katz has dem­
onstrated that effect experimentally in gain-guided
arrays with separate stripe contacts [39]; Twu et al.
[15] and Morrison et al. [17] have shown that 0°phase-shift operation is obtained by preferentially
pumping the interelement regions of index-guided
arrays. Mukai et al. [16] and Kapon et al. [18] were
able to achieve 0°-phase-shift operation from lossless,
uniformly pumped ridge-guidetype array structures
(Fig. l b ) , which, according to a recent analysis by
Streifer et al. [40] on arrays made of large-opticalcavity (LOC) -type devices [4], should have larger
gains in between elements than in the elements.
An interesting case occurs when the individual ele­
ments are wide enough to support first-order spatial
modes (e.g., the impurity-disordered array [26]).
When elements operate in first-order modes and with
no phase shift between adjacent devices, the field
goes through nulls in the interelement regions. Thus,
in first-order-mode structures with loss between
elements, the 0°-phase-shift array mode is the one
favored to lase. The corresponding beam pattern is
still two-lobed [26] since the individual-element beam
pattern is two-lobed. The field distribution can be
thought of as the 180°-phase-shift-mode pattern of an
array with twice as many elements.
Means of Achieving Fundamental
Array-Mode Operation
Several approaches have been demonstrated for
achieving fundamental array-mode operation: net
gain in between the elements [15]—[18] (Fig. 5a);
Fig. 4
Schematic representation of gain-loss profiles in index-guided
arrays. For a given configuration, the array mode that "sees" most
gain lases first.
IEEE CIRCUITS A N D DEVICES M A G A Z I N E
Fig. 5
The various schemes employed, to date, for achieving
fundamental
array-mode operation: (a) larger gain between elements than in
the elements [15]-[18]; (b) "chirped" arrays [41]; (c) diffractioncoupled arrays [42], [43]; (d) offset-stripe
arrays [44]; and
(e) Ύ-junction arrays [21].
"chirped" arrays [41] (Fig. 5b); diffraction-coupled
arrays [42], [43] (Fig. 5c); offset-stripe arrays [44]
(Fig. 5d); and Y-junction arrays [21] (Fig. 5e). The
Table s u m m a r i z e s the best-published results to
date for methods used to obtain 0°-phase-shift-mode
operation.
The net-gain-between-elements approach is shown
schematically in Fig. 5a. By providing enhanced gain
in the regions between the lasing elements, the ap­
proach appears rather inefficient. With the exception
of the two-element device [17], all other such arrays
have high thresholds and low-differential quantum
efficiencies. The best result to date is a 1.8°-wide beam
(only 1.2 times the diffraction limit) up to 70 mW
peak-pulsed power [16]. "Chirped" arrays contain
elements that vary in size across the array. In a sym­
metrically chirped array (Fig. 5b), the 0°-phase-shiftmode energy is concentrated in the central elements
[41], while the 180°-phase-shift energy is concentrated
in the outer elements. By providing an injected carrier
profile (i.e., gain) peaked in the center of the struc­
ture, the 0°-phase-shift mode is favored to lase. The
best result to date is a 3°-wide beam (—twice the dif­
fraction limit) to a peak-pulsed power of 40 mW, in a
gain-guided array. The main drawback of this ap­
proach is that the emitting area is relatively small
compared to the array [41], thus defeating the main
purpose for making arrays: the achievement of a
large emitting area. Recent theoretical work [45] con­
firms that w h e n array elements are of different
sizes, the optical field will be localized to only one or
a few elements.
When the array elements are interrupted longi­
tudinally, mode-mixing regions are created, which al­
low for strong coupling, so-called diffraction coupling
[42] (Fig. 5c). It is also a means of providing regions of
high gain, which will generally favor 0°-phase-shift
operations. Such devices have been made with both
gain-guided [42] as well as index-guided [43] lasers.
The best result to date is operation in a 1.2°-wide
diffraction-limited b e a m to 40 m W peak-pulsed
power [43] (at — 5° off the normal to the facet) from
an array of index-guided devices.
The most impressive result reported to date, al­
though still in low-duty cycle, pulsed operation, is one
using offset gain-guided stripes [44], as shown in
Fig. 5e. The authors achieved 575 mW peak-pulsed
power in a 1.9°-wide beam (—3 times the diffraction
limit) at —5° off the normal to the facet. The beam
being off-normal is explained as a result of a mild
array-element chirping across the array [46]. The pref­
erence for 0°-phase-shift operation is, most probably,
a combination of destructive interference and sub­
sequent loss for the 180°-phase-shift mode [47]
and the provision of excess gain in the offset-stripe
(central) region.
Finally, the Y-junction structure [21] (Fig. 5f) relies
on the fact that the 0°-phase-shift mode of the twostripe section is efficiently coupled into the one-stripe
section, while the 180°-phase-shift mode will not
couple but will be lost as a radiation mode (i.e., de­
structive interference). CW operation in a diffractionlimited, forward-looking b e a m was achieved to
65 mW output power. The operating principle is basi­
cally the same as the one originally put forth by Chen
and Wang [47] for an N-unit array. With the success of
the offset-stripe [44] and. the Y-junction [21] schemes
this "interferometric" approach to discriminating
Table
Method
Best Results to Date*
Beamwidth/Diffraction Limit
Net gain between elements
1.8° to 70 mW, pulsed [16]
1.3
"Chirped" structure
3.5° to 40 mW, pulsed [41]
-2.0
Interferometric & chirped structure
1.9° to 575 mW, pulsed [44]
-3.0
Interferometric (Y-junction) structure
4.4° to 65 mW, cw [21]
*Single-beam
JANUARY 1986
+
+
1.1
5° off-normal
13
against a 180°-phase-shift-mode operation appears
most promising.
Other approaches to obtain 0°-phase-shift oper­
ation have been proposed: variable-spacing arrays
[13], and placing π phase-shifters on alternate array
elements [18]. Variable-spacing arrays have demon­
strated some capability for single-beam operation in
both AlGaAs/GaAs [13] as well as InGaAsP/InP [20]
devices, π phase-shifters on alternate elements could
be realized by integrated phase-shifters [48] or prefer­
ential facet coatings [48]. This would force an array
that internally operates out of phase to emit in an
in-phase beam pattern. The π-phase-shifter method
would also be a very efficient one, since, unlike the
net-gain-between-elements approach (Fig. 5a), gain
can be provided preferentially where it is needed
most: the lasing regions.
It is interesting to note that cw operation in a single,
narrow beam has been obtained only from two-unit
arrays [17], [21], which, on the one hand, have rela­
tively low threshold currents, and, on the other hand,
allow easy discrimination against the 180°-phase-shift
mode. It should also be pointed out that index-guided
arrays, when phase-locked, generally operate in a sin­
gle longitudinal mode cw. Single-mode cw powers as
high as 80 mW have been reported [11], [19]. How­
ever, at the present time, the single-mode operation
of these arrays is not very well understood.
Dynamics of Phase-Locked Operation
There is only a limited amount of data available on
the behavior of phase-locked arrays under high-speed
modulation. Early work by the group at Xerox indi­
cated that it was possible to modulate arrays at rates
in excess of 1.8 GHz at a 40-mW output for applica­
tions in local area networks [49]. Rise times as short as
220 ps were reported. However, only recently have
there been studies of the dynamics of the phase lock­
ing between stripes under fast-pulse excitation. An
important question is how quickly the individual ele­
ments of an array can lock together in transient oper­
ation. Work by Van der Ziel and Temkin on the active
mode locking of arrays showed that it is possible to
produce s h o r t p u l s e s ( ~ 6 0 p s ) from arrays of
Schottky-barrier emitters [50]. Based on their data,
they estimated that the time required for phase lock­
ing was less than 20 ps. More recent work has utilized
a streak camera to look at the evolution of the nearand far-field distributions of a gain-guided array with
picosecond resolution [51]. The data indicated that
the arrays would make transitions between various
array modes during the rise of the drive current pulse
on a nanosecond time scale. Although the experi­
ments were done with gain-guided arrays that may
be somewhat less stable than arrays with built-in
index guides, the results suggest that the stability
of array operation under transient conditions needs
further investigation.
14
Reliability of Phase-Locked Arrays
There have been relatively few published results on
the reliability of phase-locked arrays. Life tests of
quantum-well arrays (not phase-locked in a given
mode) fabricated at Xerox has resulted in the predic­
tion of median lives, at 30°C ambient temperature,
greater than 2 2 , 0 0 0 h o u r s at output p o w e r s of
100 mW [52] and 4900 hours at 200 mW [53]. Since the
output of a phase-locked array is distributed over a
number of emitters (usually ten), it is possible to have
a relatively large integrated output power from the
array while still maintaining a low power density for
the individual emitters. In the Xerox life tests, for
example, the individual emitters were operating at a
linear power density of only —1.6 mW//xm, which
can be compared to a reliable power density of
7-10 m W / μ π ι for single-element devices. T h e s e
power densities should be sufficiently low to lead to
a reliable operation over long times. However, many
technical issues concerning the reliability of arrays
remain. In particular, there is the issue of whether the
array mode is stable in time. For example, the ques­
tion of what happens to the operating modes of the
array should one of the individual elements degrade
is a critical one. It is likely that a significant change in
the gain profile would occur if an element failed and
that this failure would change the operating mode of
the array. Another significant question is the effect of
bonding and thermal considerations on the array as it
ages. Recent theoretical work has shown that for ar­
rays with large numbers of elements, it is necessary to
have a highly uniform junction temperature profile
[54] to ensure phase locking. Deterioration in the
bonding, during long-term operation, can therefore
not only affect the overall thermal resistance and
power output of the device, but nonuniformities in
bonding can severely affect the phase locking. These
issues will require additional investigation in order to
ensure the reliable operation of phase-locked arrays.
In any event, array reliability appears a much more
complex issue than that of reliability of individualelement devices.
Conclusions
In the last year, several methods for obtaining
single-beam operation from phase-locked arrays have
been demonstrated. However, published results of
cw operation in a single beam are only for two-unit
arrays, thus pointing to the need for lowering the
threshold currents of many-element devices. Now
that array operation in a single beam appears to be a
problem on the way to being solved, the next chal­
lenge is to controllably obtain operation in a diffrac­
tion-limited beam.
It is apparent that the beam patterns of indexguided devices are significantly more stable and well
defined than the b e a m patterns of gain-guided
devices. In addition, index-guided arrays generally
IEEE CIRCUITS A N D DEVICES M A G A Z I N E
operate in a single longitudinal mode under cw drive
conditions, a property desired in many applications.
However, gain-guided devices have demonstrated
much higher overall cw powers (1.5-2.5 W) and re­
liable operation at outputs in the 100-200 mW range,
although with poor beam quality. The combination of
the recently demonstrated methods for single-beam
operation, index-guided devices, quantum-well
structures for minimal power dissipation, and im­
proved die bonding and heat-sinking techniques has
the potential to allow for reliable array operation in a
stable, single beam to powers as high as 0.5 W.
Acknowledgment
The authors gratefully acknowledge valuable tech­
nical discussions with Eugene I. Gordon and Ronald
J. Nelson of Lytel Incorporated.
References
[1] J. E. Ripper and T. L. Paoli, 'Optical Coupling of Adjacent
Stripe Geometry Junction L a s e r s / ' Appl. Phys. Lett., vol. 17,
pp. 371-373, 1970.
[2] Μ. B. Spencer and W. E. Lamb, Jr., 'Theory of Two Coupled
Lasers," Phys. Rev. A, vol. 5, no. 2, pp. 8 9 3 - 8 9 8 , Feb. 1972.
[3] D. R. Scifres, R. D. Burnham, and W. Streifer, "Phase Locked
Semiconductor L a s e r A r r a y , " Appl. Phys. Lett., vol. 3 3 ,
pp. 1015-1017, Dec. 1978.
[4] D. Botez, "Laser Diodes Are Power-Packed," IEEE Spectrum,
vol. 22, no. 6, pp. 4 3 - 5 3 , June 1985.
[5] J. P. Van der Ziel, R. M. Mikulyak, H. Temkin, R. A. Logan,
and R. D. Dupuis, "Optical Beam Characteristics of Schottky
Barrier Confined Arrays of Phase-Coupled Multiquantum Well
GaAs Lasers," IEEE J. Quantum Electron., vol. QE-20, no. 11,
pp. 1259-1266, Nov. 1984.
[6] D. E. Ackley and R. W. H. Engelmann, "High Power Leaky
Mode Multiple Stripe Laser," Appl. Phys. Lett., vol. 3 9 ,
pp. 2 7 - 2 9 , July 1981.
[7] V. I. Malakhova, Y. A. Tambiev, and S. D. Yakubovich,
"Regular Integrated Arrays of Stripe Injection Lasers," Sov. J.
Quantum Electron., vol. 11, pp. 1351-1352, Oct. 1981.
[8] D. R. Scifres, R. D. Burnham, and W. Streifer, "High Power
Coupled Multiple Stripe Quantum Well Injection Lasers,"
Appl. Phys. Lett., vol. 41, pp. 118-120, July 1982.
[9] D. R. Scifres, W. Streifer, and R. D. Burnham, "Experimental
and Analytic Studies of Coupled Multiple Stripe Diode La­
sers," IEEE J. Quantum Electron.,
vol. Q E - 1 5 , pp. 917-922,
Sept. 1979.
[10] D. E. Ackley, "Single Longitudinal Mode Operation of High
Power Multiple Stripe Injection Lasers," Eighth IEEE Inter­
national Semiconductor Laser Conference, Postdeadline Paper
PD5, Ottawa-Hull, Canada, Sept. 13-15,1982; and Appl. Phys.
Lett., vol. 42, pp. 152-154, Jan. 1983.
[11] D. Botez and J. C. Connolly, "High Power Phase Locked
Arrays of Index Guided Diode Lasers," Appl. Phys. Lett.,
vol. 43, pp. 1096-1098, Dec. 1983.
[12] J. Ohsawa, K. Ikeda, K. Takahashi, and W. Susaki, "PhaseLocked Oscillation of Dual SBH Lasers," Tech. Digest of the
Fourth IOOC Conference, paper 2 8 B 3 - 1 , pp. 172-173, Tokyo,
Japan, June 2 7 - 3 0 , 1983.
[13] D. E. Ackley, "Phase-Locked Injection Laser Arrays with Non­
uniform Stripe Spacing," Electron.
Lett., vol. 2 0 , no. 17,
pp. 695-697, Aug. 1984.
[14] I. Suemune, T. Terashige, and M. Yamanishi, "Phase-Locked,
Index-Guided Multiple-Stripe Lasers with Large Refractive
Index Differences," Appl. Phys. Lett., vol. 45, no. 10, pp. 10111013, Nov. 15, 1984.
JANUARY 1986
[15] Y. Twu, A. Dienes, S. Wang, and J. R. Whinnery, "High Power
Coupled Ridge Waveguide Semiconductor Laser Arrays,"
Appl. Phys. Lett., vol. 45, no. 7, pp. 709-711, Oct. 1, 1984.
[16] S. Mukai, C. Lindsey, J . Katz, E. Kapon, Z. Rav-Noy, S.
Margalit, and A. Yariv, "Fundamental Mode Oscillation of a
Buried Ridge Waveguide Laser Array," Appl. Phys. Lett.,
vol. 45, no. 8, pp. 8 3 4 - 8 3 5 , Oct. 15, 1984.
[17] C. B. Morrison, L. M. Zinkiewicz, A. Burghard, and L.
Figueroa, "Improved High-Power Twin-Channel Laser with
Blocking L a y e r , " Tech. Digest of the IEDM
Conference,
paper 19.3, pp. 517-519, San Francisco, CA, Dec. 6 - 9 , 1984;
and Electron. Lett., vol. 21, pp. 3 3 7 - 3 3 8 , Apr. 1985.
[18] E. Kapon, L. T. Lu, Z. Rav-Noy, M. Yi, S. Margalit, and A.
Yariv, "Phased Arrays of Buried-Ridge InP/InGaAsP Diode
Lasers," Appl. Phys. Lett., vol. 46, no. 2, pp. 136-138, Jan. 15,
1985.
[19] H. Temkin, R. A. Logan, J. P. Van der Ziel, C. L. Reynolds, Jr.,
and S . M . Tharaldsen, "Index-Guided Arrays of Schottky
Barrier Confined Lasers," Appl. Phys. Lett., vol. 46, no. 5,
pp. 465-467, Mar. 1985.
[20] Ν. K. Dutta, L. A. Koszi, S. G. Napholtz, B. P. Segner, and
T. Cella, "InGaAsP (λ = 1.3 μπή Laser Array," Tech. Digest
of CLEO Conference, Paper TuF5, Baltimore, MD, May 2 1 - 2 4 ,
1985.
[21] M. Taneya, M. M a t s u m o t o , S. Matsui, S. Yano, and Τ
Hijikata, "0° Phase Mode Operation in Phased-Array Laser
Diode with Symmetrically Branching Waveguide," Appl. Phys.
Lett., vol. 47, no. 4, pp. 3 4 1 - 3 4 3 , Aug. 15, 1985.
[22] F. Kappeler, H. Westermeier, R. Gessner, and M. Druminski,
"High cw Power Arrays of Optically Coupled (Ga, Al)As Oxide
Stripe Lasers with dc-to-Light Conversion Efficiencies of Up to
36%," Ninth IEEE International Semiconductor Laser Conference,
Paper G-3, Rio de Janeiro, Brazil, Aug. 7-10, 1984.
[23] D. R. Scifres, R. D. Burnham, and W Streifer, "Lateral Grating
Array High Power cw Visible Semiconductor Laser," Electron.
Lett., vol. 18, pp. 5 4 9 - 5 5 0 , June 1982.
[24] J. J. Yang, R. D. Dupuis, and P. D. Dapkus, "Theoretical Analy­
sis of Single-Mode AlGaAs-GaAs Double Heterostructure La­
sers with Channel-Guide Structure," /. Appl. Phys., vol. 53,
pp. 7218-7223, Nov. 1982.
[25] K. M e e h a n , P. Gavrilovic, J . E . Epler, K. C. Hsieh, N.
Holonyak, Jr., R. D. Burnham, R. L. Thornton, and W. Streifer,
"Donor-Induced Disorder-Defined Buried-Heterostructure
A l G a i - A s - G a A s Quantum-Well L a s e r s , " /. Appl.
Phys.,
vol. 57, no. 12, pp. 5345-5348, June 15, 1982.
[26] P. Gavrilovic, K. Meehan, J. E. Epler, N. Holonyak, Jr., R. D.
B u r n h a m , R. L . T h o r n t o n , a n d W. Streifer, "ImpurityDisordered, Coupled-Stripe Al Gai_ As-GaAs Quantum Well
Laser," Appl. Phys. Lett., vol. 46, no. 9, pp. 857-859, May 1985.
[27] D. R. Scifres, C. Lindstrom, R. D. Burnham, W. Streifer, and
T.L. Paoli, "Phase-Locked (GaAl)As Laser Diode Emitting
2.6 W c w from a Single Mirror," Electron. Lett., vol. 19,
pp. 169-171, 1983.
[28] K. Otsuka, "Coupled-Wave Theory Regarding Phase-LockedArray Lasers," Electron. Lett., vol. 19, pp. 723-725, Sept. 1983.
[29] J. K. Butler, D. E. Ackley, and D. Botez, "Coupled-Mode Anal­
ysis of Phase-Locked Injection Laser Arrays," Technical Digest
of the Optical Fiber Communication Conference,
Paper TuF2,
pp. 4 4 - 4 6 , New Orleans, LA, Jan. 2 3 - 2 5 , 1984.
[30] J. K. Butler, D. E. Ackley, and D. Botez, "Coupled Mode Anal­
ysis of Phase-Locked Injection Laser Arrays," Appl. Phys. Lett.,
vol. 44, pp. 2 9 3 - 2 9 5 , Feb. 1984; see also Appl. Phys. Lett.,
vol. 44, p. 935, May 1984.
[31] D. Botez, "Array-Mode Far-Field Patterns for Phase-Locked
Diode-Laser Arrays: Coupled-Mode Theory vs. Simple Dif­
fraction Theory," IEEE J. Quantum Electron.,
vol. QE-21,
no. 11, pp. 1752-1755, Nov. 1985.
[32] E. Kapon, J . Katz, and A. Yariv, "Supermode Analysis of
Phase-Locked Arrays of Semiconductor Lasers," Opt. Lett.,
vol. 10, pp. 125-127, Apr. 1984.
x
x
x
x
15
[33] S. R. Chinn and R. J. Spiers, "Modal Gain in Coupled-Stripe
L a s e r s , " IEEE J. Quantum
Electron.,
vol. Q E - 2 0 , no. 4 ,
pp. 358-363, Apr. 1984.
[34] T. L. Paoli, W. Streifer, and R. D. Burnham, "Observation of
Supermodes in a Phase-Locked Diode Laser Array," Appl.
Phys. Lett., vol. 45, pp. 217-219, Aug. 1984.
[35] J . E . Epler, N. Holonyak, Jr., R. D. Burnham, T. L. Paoli, and
W. Streifer, "Far-Field Supermode Patterns of a Multiple-Stripe
Quantum Well Heterostructure Laser Operated (7330 A, 300 K)
in an External Grating Cavity," Appl. Phys. Lett., vol. 45,
pp. 406-408, Aug. 1984.
[36] D. Botez, "Near and Far-Field Analytical Approximations for
the Fundamental Mode in Symmetric Waveguide DH Lasers,"
RCA Review, vol. 39, pp. 577-603, Dec. 1978.
[53] G . L . Harnagel, D . R . Scifres, P.S. Cross, H . H . Kung, R . C .
Quenelle, A. A. Rozycki, D. P. Worland, R. D. Burnham, T. L.
Paoli, D. L. Smith, and R. L. Thornton, "Lifetime of Diode
Laser Arrays Operating at 200 m W CW," Tech. Digest of CLEO
Conference, Paper ThZl, Baltimore, MD, May 1985.
[54] Μ. T. Tavis and Ε. M. Garmire, "Improved Heat-Sink De­
sign for Coherent Laser A r r a y s , " Electron. Lett., vol. 20,
pp. 689-690, Aug. 1984.
[55] G. P. Agrawal, "Lateral-Mode Analysis of Gain-Guided and
Index-Guided Semiconductor-Laser Arrays," /. Appl. Phys.,
vol. 58, pp. 2922-2931, Oct. 15, 1985.
[37] W. Streifer, A. Hardy, R. D. Burnham, and D. R. Scifres,
"Single-Lobe Phased-Array Diode Lasers," Electron.
Lett.,
vol. 21, no. 3, pp. 118-120, Jan. 1985.
[38] J. K. Butler, D. E. Ackley, and M. Ettenburg, "Coupled-Mode
Analysis of Gain and Wavelength Oscillation Characteristics of
Diode Laser Phased A r r a y s , " IEEE J. Quantum
Electron.,
vol. QE-21, no. 5, pp. 4 5 8 - 4 6 3 , May 1985.
[39] J. Katz, E. Kapon, C. Lindsey, S. Margalit, and A. Yariv, "Far
Field Distribution of Semiconductor Phase Locked Arrays
with Multiple Contacts," Electron. Lett., vol. 19, pp. 660-662,
Aug. 1983.
[40] W. Streifer, A. Hardy, R. D. Burnham, R. L. Thornton, and
D. R. Scifres, "Criteria for Design of Single-Lobe Phased-Array
Diode Lasers," Electron. Lett., vol. 21, no. 11, pp. 505-506,
May 1985.
[41] E. Kapon, C. P. Lindsey, J. S. Smith, S. Margalit, and A. Yariv,
" I n v e r t e d - V C h i r p e d P h a s e d A r r a y s of G a i n - G u i d e d
GaAs/GaAlAs Diode L a s e r s , " Appl. Phys. Lett., vol. 4 5 ,
pp. 1257-1259, Dec. 15, 1984.
[42] T. R. Chen, K. L. Yu, B. Chang, A. Hasson, S. Margalit, and A.
Yariv, "Phase-Locked InGaAsP Laser Array with Diffraction
Coupling," Appl. Phys. Lett., vol. 43, pp. 136-139, July 1983.
[43] D. E. Ackley, "Phase-Locked Arrays with Mode-Mixing Sec­
tions," RCA Corporation, private communication.
[44] D. F. Welch, D. Scifres, P. Cross, H. Kung, W. Streifer, R. D.
Burnham, and J. Yaeli, "High-Power (575 m W ) Single-Lobed
Emission from a Phased-Array Laser," Electron. Lett., vol. 21,
no. 14, pp. 603-605, July 1985.
[45] K. Nishi and R. Lang, "Lateral Mode Stabilization in Multistripe Laser Diode," Japanese J. Appl. Phys., vol. 24, no. 5,
pp. L349-L351, May 1985.
[46] C. P. Lindsey, E. Kapon, J. Katz, S. Margalit, and A. Yariv,
"Single Contact Tailored Gain Phased Array of Semiconductor
Lasers," Appl. Phys. Lett., vol. 45, pp. 722-724, Oct. 1984.
[47] K. -L. Chen and S. Wang, "Single-Lobe Symmetric Coupled
Laser Arrays," Electron. Lett., vol. 21, no. 8, pp. 347-349,
Apr. 1985.
[48] D. E. Ackley, D. Botez, and B. Bogner, "Phase-Locked Injec­
tion Laser Arrays with Integrated Phase Shifters," RCA Review,
vol. 44, pp. 625-633, Dec. 1983.
[49] D. R. Scifres, M. D. Bailey, R. E. Norton, R. D. Burnham, and
E. G. Rawson, "Diode Laser Array Excited Star Coupler Suit­
able for Fibernet II," Electron. Lett., vol. 18, pp. 1028-1029,
Nov. 1982.
[50] J. P. Van der Ziel, H. Temkin, R. D. Dupuis, and R. M.
Mikulyak, "Mode-Locked Picosecond Pulse Generation from
High-Power Phase-Locked GaAs Laser Arrays," Appl. Phys.
Lett., vol. 44, pp. 3 5 7 - 3 5 9 , Feb. 1984.
[51] R. A. Elliot, R. K. DeFreez, T. L. Paoli, R. D. Burnham, and W.
Streifer, "Dynamic Characteristics of Phase-Locked Multiple
Electron.,
Quantum Well Injection Lasers," IEEE J. Quantum
vol. QE-21, pp. 5 9 8 - 6 0 3 , June 1985.
[52] G. L. Harnagel, T. L. Paoli, R. L. Thornton, R. D. Burnham,
and D. L. Smith, "Accelerated Aging of 100 mW cw MultipleStripe GaAlAs Lasers Grown by Metalorganic Chemical Vapor
Deposition," Appl. Phys. Lett., vol. 46, pp. 118-120, Jan. 1985.
16
Dan Botez
Dan Botez was born in Bucharest, Romania, on May 22, 1948. He
received the B.S. degree with "highest honors," the M.S. and Ph.D.
degrees in electrical engineering from the University of California,
Berkeley, in 1971, 1972, and 1976, respectively.
As part of his graduate work, he designed and built the liquidphase epitaxy system used in Berkeley for the growth of AlGaAs
lasers, and studied the characteristics of layers grown over chan­
neled substrates, as well as the novel optical devices made possible
by this method (e.g., waveguide lasers, ring lasers, etc.). For one
year following graduation, he continued his research in the semi­
conductor laser field as Postdoctoral Fellow at the IBM T. J. Watson
Research Center, Yorktown Heights, New York. In 1977, he joined
the technical staff at RCA Laboratories, Princeton, New Jersey,
where he initially worked on high-radiance surface-emitting LEDs;
he then developed two novel types of single-mode cw diode
lasers: the constricted double-heterostructure (CDH) laser, which
has demonstrated cw and pulsed operation to the highest ambient
temperatures ever reported; and the constricted double-heterojunction large-optical-cavity (CDH-LOC) laser, a high-power de­
vice that has achieved the highest cw power conversion efficiency
for AlGaAs diode lasers. Recent work has resulted in a novel type
of phase-locked array, that has demonstrated the highest singlelongitudinal-mode power of any type of semiconductor laser. From
1982 to 1984, he worked as Research Leader in the Optoelectronic
Devices and Systems Research Group at RCA. In 1984, Dr. Botez
joined Lytel Incorporated, as Director of Device Development.
He is the author or coauthor of over 65 scientific papers, has
coauthored a book, and holds 14 patents.
Dr. Botez is a member of Phi Beta Kappa, the Optical Society of
America, a Fellow of the IEEE, and a member in the IEEE Tech­
nical Committee on Semiconductor Lasers. In 1979, he received an
RCA Outstanding Achievement Award for contributions to the
development of a high-density optical recording system employing
an injection laser. As part of the events associated with the IEEE
Centennial Year (1984), he was chosen the Outstanding Young
Engineer of the Quantum Electronics and Applications Society and
was awarded an IEEE Centennial Key to the Future.
IEEE CIRCUITS A N D DEVICES M A G A Z I N E
Donald Ε. Ackley received Sc.Β., Μ.Sc., and Ph.D. degrees in
electrical sciences in 1975,1976, and 1979, respectively, from Brown
University in Providence, Rhode Island. His doctoral work was
concerned with the study of picosecond relaxation processes in
amorphous semiconductors.
Donald E. Ackley
JANUARY 1986
Upon receiving his doctoral degree in 1979, Dr. Ackley joined
Hewlett-Packard Laboratories in Palo Alto, California. At HewlettPackard, he was responsible for the development of new highpower AlGaAs injection lasers, which included phase-locked
arrays and index-guided MOCVD lasers. In 1983, he joined RCA
Laboratories in Princeton, New Jersey, where he continued to pur­
sue his interests in high-power AlGaAs lasers and MOCVD. When
Lytel Incorporated was formed in 1984, Dr. Ackley joined the new
organization with responsibilities that included the design and
characterization of optoelectronic devices. Currently, Dr. Ackley is
the Director of Device Engineering at Lytel.
Dr. Ackley has published more than 20 scientific papers and
holds two patents. He is a member of the American Physical
Society.
17