IEEE JOURNAL OF SELECTED TOPICS IN QUANTUM ELECTRONICS, VOL. 1, NO.

2, JUNE 1995 741

11-VI Blue-Green Laser Diodes
Akira Ishibashi Member, IEEE
Abstnzct- ZnMgSSe, forming a type I heterostructure with
Zn(Cd)Se, fully lattice-matched to GaAs with an energy gap
tunable up to -4.5 eV, has made possible continuous-wave (CW)
operation of both a green laser diode (LD) and a blue LD
at m m temperature. The device characteristics of the 11-VI
wide-gap LDs are becoming as good as those of established
111-V LDs, except for device lifetime. Remaining key issues are
p-doping in the wide-gap ZnMgSSe and reliability of 11-VI wide-
gap LDs. Valence-band engineering via superlattice (SL) use is
proposed, based on the amphoteric defect model, for removing
the doping limit in the p-type ZnMgSSe. This will lead to CW
operation of a blue-emitting laser diode with a wavelength of
450 - 460 nm. For reliability, employing a GaAs-buffer layer
has made possible room-temperature (RT) CW operation with
a lifetime of 1 hour. Analysis of the degradation process shows
that no catastrophically fast degradation occurs when 11-VI LDs
degrade. The reliability of the ZnMgSSe-based LDs would soon
be established, and the 11-VI wide-gap LDs will likely blossom
colorfully in the near future.
I. INTRODUCTION
EMICONDUCTOR laser diodes (LD s) have been play-
S ing a central role in many systems, such as a high
transmission-rate communication system, a high power energy
source system and an optical or magneto-optical memory
system, e.g., a consumer audio R O W disk system with
AlGaAs infra-red LDs, or a high definition ROM disk sys-
temwith AlGaInP red-emitting LDs. The evolution of the
semiconductor LD will be two-fold. One is to integrate the
conventional LDs, which have already been established as a
discrete chip, into monolithic highly-functional devices. The
AlGaAs-GaAs-based LD has been in this phase. The other is
to establish a discrete chip that achieves challenging targets,
such as a super high power, an extremely rapid response, and
a short wavelength. For the short wavelength, green and blue
LDs are needed as the next generation for higher density disk
systems. The shorter the wavelength becomes the higher the
recording density can be, leading to either a long time or a high
definition recording. Furthermore, with the green and the blue
LDs a new paradigm is approaching where RGB full-color,
all solid-state light sources are available, in an inexpensive,
mass productive, and highly compact manner, replacing vac-
uumtubes (light bulbs) just as transistors did 40 years ago.
Among blue-emitting light sources, chalcopyrites, I-III-VI2
compounds, are still in an early stage of development. As for
column IV semiconductors, there is Si c which is, however,
unsuitable for an active layer of LDs, having an indirect
band structure. For 111-V compounds, GaN-based materials are
highly promising but only LEDs are available so far. Finally,
Manuscript received November 1, 1994; revised January 28, 1995.
Theauthor is with the Sony Corporation Research Center, 174 Fujitsuka-
IEEE Log Number 941 1159.
cho, Hodogaya-ku, Yokohama 240 Japan.
ZnSe-based materials of 11-VI compounds are the only system
that has ever provided lasing, both in green and blue range of
the wavelength.
Until Haase et al. reported in 1991 the first semicon-
ductor pulsed blue-green LD operation at 77 K based on
ZnCdSe-Zn(S)Se [ 11, 11-VI wide-gap semiconductors had
lased only in photopumped or electron-beam-excited mode.
The epoch-making first 11-VI LD was based on the novel
p-type doping technique that uses nitrogen excited by an RF-
plasma source [2], [3]. Much progress has been made since
then-3M Company [4] and Purdue and Brown Universities
[5] reported 77 K CW and RT pulsed operation. Phillips
demonstrated pulsed operation up to 394 K [6]. In 1993, 3M
reported a very low threshold current value of 2.5 mA [7].
Sony demonstrated CW operation of both green [8] and blue
[9] lasers in 1993. PurduelBrown achieved - 20-sec-long CW
operation [lo] in 1993. Before going into the latest results, we
will discuss first the major problems with the early 11-VI LDs,
i.e., the high operation voltage and the lack of a wide-gap
cladding layer.
11. LOWERING THE OPERATION VOLTAGE
Since valence-band-maximum (VBM) measured in energy
from the vacuumlevel is larger than the work function of
virtually all metals available, the valence band of ZnSe is
located deeper than the Fermi level of any metals. In order to
have good p-type contact, some materials have to be inserted
between the metal (preferably with a large work function)
and the p-type ZnSe, as the metal-semiconductor interfacial
depletion layer in the ZnSe is too thick for holes to tunnel
through due to the relatively low doping level in the p-ZnSe
(- 10 ~m- ~) . This renders the conventional hole-injection
scheme impossible, i.e., the scheme in which the hole tunnels
from the metal to the semiconductor through the p+-doping-
induced thin depletion layer, which is typically resorted to in
the case of In-V compound semiconductors such as GaAs.
The poor ohmic contact to p-ZnSe leads to too high an
operation voltage of blue-green LDs, as well as a parasitic
heating. Efforts to obtain a good contact to the p-ZnSe is
two-fold, as shown below.
A. Au-Pt-Pd Multilayered Metal to p-ZnTe
HgSe was found to form ohmic contacts to p-type ZnSe by
reducing the interfacial energy barrier [l l ]. ZnTe is another
candidate for a hole injection layer for p-contacts among 11-VI
compounds. The ZnTe is promising, because it can be doped
with a p-concentration above l OI 9 cmP3 in marked contrast
to ZnSe that can only be doped as high as - l0ls ~m- ~.
Since the work functions of Pt and Au are large (5.65 eV and
1077-260X/95$04.00 0 1995 IEEE
742 EEE JOURNAL OF SELECTED TOPlCS IN QUANTUM ELECTRONICS, VOL. 1, NO. 2, JUNE 1995
a,
0
a
CI
-
5 s
0
0 10 20 30 40 50
Current (mA)
Fig. 1. I-V and L-I characteristics of a ZnMgSSe-based LD with the
resonant tunneling ZnSe-ZnTe MQW. Applied voltage for the onset of the
laser operation is 4.7 (4.9) V for pulsed (CW) operation. The inset shows the
resonant tunneling ZnSe-ZnTe MQW structure.
5.1 eV, respectively [12]), with reactions being expected at
the interface between a transition metal such as Pd or Ni and
the semiconductor, Au-Pt-Pd were tried [13] as the ohmic
contacts to p-doped ZnTe. The specific contact resistance is as
low as 5 x lop6 Ohm cm2. The good ohmic contact metal
to p-ZnTe is obtained.
B. ZnTe-ZnSe Resonant Tunneling MQW
The p-type ohmic contact problem is now reduced to what
to put between the p-ZnTe and the p-ZnSe. One candidate
is a ZnTe-ZnSe pseudo-graded layer [14], which has been
shown to be feasible to have a low operation voltage [IO].
In this case, the superlattice is a substitute for a graded
ZnSeTe ternary alloy whose energy gap is difficult to design
due to Te-clustering and large band-bowing. Using a similar
structure but based on a different concept, a ZnTe-ZnSe
multiquantum-well (MQW) structure was proposed whose
resonating subbands enables a quasi-ohmic contact to p-type
ZnSe to occur through sequential resonant tunneling of holes
[15]. The lowest threshold voltage so far is brought by the
resonant tunneling ZnTe-ZnSe MQW, as shown in Fig. 1.
The applied voltage is 4.5 (4.7) V [16] for the onset of room
temperature pulsed (CW) operation with a ZnMgSSe-based
laser diode. Operating voltage was reduced even further down
to 3.7 V [17]. The threshold voltage is about one fifth of that
observed in early 11-VI LDs, and the p-type contact is not
anymore a major problem, although the reproducibility and
reliability issues are yet to be established.
111. ZnMgSSe QUATERNARY ALLOY
Other remaining problems are in obtaining wider energy gap
materials which are suitable for serving as cladding layers for
Zincblende, 77K
n
a
v
0 1 2 3 4 5
Energy Gap (eV)
Fig. 2. Plot of the lattice constant versus energy gap for ZnMgSSe qua-
ternary alloy. The inset shows the band line-up of the constituent binary
semiconductors.
blue LDs. Theearly 11-VI wide-gap semiconductor LDs [l],
based on ZnCdSe-Zn(S)Se, have an intrinsic problemof poor
carrier confinement that makes the operation difficult at high
temperature or at short wavelength. For shorter wavelength
laser diodes, with an active layer made of ZnCdSe with small
Cd mole fraction, or even ZnSe, use of conventional 11-VI
binary compounds such as ZnSe, ZnS, ZnTe, and CdS is not
sufficient to fulfill the requirement of carrier confinement. In
seeking a wide energy-gap material lattice-matched to GaAs,
ZnSSe is as a good starting material. ZnS has an energy gap
of 3.80 eV (at 77 K) with a lattice constant of 5.41 A. What is
needed is a column I1 element that bonds with S or Se to form
a wide-gap 11-VI compound with a lattice constant, larger than
the lattice constant of GaAs. Then the column I1element is to
be mixed with ZnSSe to form a quaternary alloy. A column I1
element is sought because the conduction band of a compound
semiconductor is made mainly from s-orbitals of the cation
(the column I1 element ) and its valence band from p-orbitals of
the anion (the column VI element). Since holes are ready to be
confined in Zn(Cd)Se due to valence-band modulation brought
by S in ZnSSe, a column I1 element is needed for confining
electrons in Zn(Cd)Se active layer through modulation of the
conduction band with respect to Zn(S)Se.
Compound semiconductors consisting of elements sitting in
the upper rows of the Periodic Table have a wide energy gap
because of wide spacing in energy of their outermost orbitals.
As both Zn and Se occupy a position in the fourth row of
the Periodic Table, the column I1 element needed to make the
energy gap wider will be in between the first and the third rows
(Postulate a). On the other hand, atomic radii become longer
with the row number in the Periodic Table, due to the increase
in the number of orbitals around a nucleus. In each row of
the Periodic Table, the atomic radius becomes shorter with
the column number, due to the increase in positive charges of
the nucleus for large column-number elements in that row. In
this context, the column I1 element needed to make the lattice
constant larger will be a small column-number element in the
fifth row or lower of the Periodic Table (Postulate b). Since
ISHIBASHI 11-VI BLUE-GREEN LASER DIODES 743
Postulate b sharply contradicts Postulate a, it seems impossible
to have such a column I1 element that enables both a large
lattice constant and a wide energy gap, concurrently. However,
because the column-number dependence of the covalent radius
becomes weaker for lower-row elements, due to the screening
by inner orbital electrons, a dramatic reversal occurs in the
row-number dependence of the radius; upper-row elements do
have longer covalent radii than lower-row ones for specific
elements [18]. Mg, in the third row, has a larger tetrahedral co-
valent radius than Zn, in the fourth row. Note that for the third
and the fourth rows the column Ill elements, A1and Ga, are on
the verge of such a reversal, i.e., they have an identical tetra-
hedral covalent radius, which has made (will make) a wide-
range of heterostructure physics blossomin the AlAs-GaAs
(AlN-GaN) system. Exploiting the reversal for the column I1
elements, it has been proven that Mg in Zn(S)Se makes the
energy gap wider and the lattice constant longer than Zn(S)Se,
as desired. In the lattice-constant-energy-gap plane (Fig. 2),
MgSe (in zinc-blende type structures) and ZnS are located
at roughly symmetrical positions with respect to ZnSe (and
GaAs). Incorporating Mg into ZnSSe allows us to cover a wide
range of energy from 2.78 eV to - 4.5 eV (at 77 K), remaining
fully lattice-matched to GaAs, as one can see in Fig. 2 [ 191. Al-
though MgS and MgSe naturally have the rock-salt structure,
ZnMgSSe grown on GaAs has the zinc-blend structure.
A. Band Line-up and Refractive Index
ZnSe and MgSe (ZnS and MgS) have the anion of Se (S)
in common, while ZnSe and ZnS (MgSe and MgS) have the
cation of Zn (Mg) in common. From the common aniodcation
rule as the 0th approximation, the valence (conduction) bands
are aligned for ZnSe and MgSe, or ZnS and MgS ( for
ZnSe and ZnS, or MgSe and MgS). As shown in the inset
of Fig. 2, the conduction-band (valence-band) offset being
- 0.8 eV (0.9 eV) for ZnSe and MgSe (ZnS and ZnSe),
the ZnSe-ZnMgSSe system is expected to have a type I
heterostructure [ 191, [20] with a valence band discontinuity of
N 50%. In addition, from a reflection-spectrum measurement
[21], it has been shown that ZnMgSSe has a refractive
index smaller than that of ZnSe. Those two facts indicate
that both photons and carriers can be well confined in the
ZnSe-ZnMgSSe double heterostructure (DH). Photo-pumped
lasing experiments directly followed using undoped
Zn(S)Se-ZnMgSSe DH structures [22]. The photo-pumped
lasing was observed at temperatures up to 500 K. This proves
that the ZnMgSSe forms a type I heterostructure with ZnSe
and has a refractive index smaller than that of ZnSe. It also
suggests that ZnMgSSe has good thermal stability.
B. Doping in ZnMgSSe
For current-injected laser operation, both n-type and p-type
doping are needed. It is fortunate that p-type and n-type
doping techniques developed for ZnSe are extendable to
ZnMgSSe. ZnC12 makes it possible to dope ZnMgSSe with
an electron concentration of up to 3 x 10'' cmW3, with
good optical quality. Using nitrogen atoms excited by electron
cyclotron resonance or RF plasma source, ZnMgSSe can be
xperiment
~ ~ ~ = 4 5 % . mh=l .4m,
AEV=33%. mh=0.6m0
.................................
15 ZnMgSSe
i n
.-
2.7 2.8 2.9 3.0 3.1 3.2
Energy Gap (eV)
Fig. 3. Net acceptor concentration in ZnMgSSe as a function of the energy
gap (77 K). The solid and dashed lines are the theoretical calculation based
on amphoteric defect model. The inset shows the schematic band structure of
GaAs, ZnSe, and Znh4gSSe.
doped with a hole concentration of up to - 1 x 1017 cmW3
(for Eg N 2.95 eV at 77 K), maintaining high epitaxial-
film quality [23]. However, in this matter a big problem
occurs. As shown in Fig. 3, the net acceptor concentration
NA-ND rapidly decreases as the energy gap of ZnMgSSe
increases. Nitrogen-doping in ZnMgSSe practically determines
the shortest wavelength of ZnSe-based blue laser diodes.
The amphoteric native defect model [24]-[26] extended to
11-VI materials system explains experimental results very
well [27]. The inset of Fig. 3 shows schematic band struc-
tures of heterointerfaces among GaAs, ZnSe, and ZnMgSSe,
where EFS is the Fermi-level stabilization energy, ESI the
energy-reference of the defect-creation, and AE, ( AE, , ) is the
conduction (valence) band offset between ZnSe and ZnMgSSe.
The highest p-type doping is limited by the position of the
ESI. The ESI determined to be located 1.895 eV [24] bellow
the EFS based on the experimental result of p-doping in ZnSe.
Then ESI , on the other hand, reproduces not only the p-doping
in ZnTe but also the p-doping in ZnMgSSe for a wide range of
energy gap, as shown in Fig. 3. Recently, based on the result
of persistent photoconductivity, so called AX-center model is
proposed [28]. The analysis based on the amphoteric native
defect model is in accordance with the interpretation based on
the AX-center model, if the site that kills one hole is identified
to be a donor level caused by structural transformation instead
of impurity interstitials.
Under limited but successful p-doping in ZnMgSSe, 77-K
CW operation of the blue LD has been demonstrated with
a blue-light-emitting ZnSe-based MQW active layer with a
threshold current density of 225 Ncm2 at a wavelength of 447
nm[29]. Thus, the feasibility of ZnMgSSe for the cladding
layer of blue LD's has been established. Removing the p-
doping limit is discussed in Section VI-A.
IV. PULSED OPERATION
Under the constraint of the poor p-doping for the wider-gap
ZnMgSSe, there are only two options. One is to aimat blue-
emitting operation at the expense of discarding CW operation.
The other is to aim at CW operation at the expense of
discarding blue-emitting operation. Merging the two will lead
744
2nTe:N
ZnSe/ZnTe
MQW
Pd/Pt/Auelectrode
2nSe:N
2nSSe:N
In electrode
Fig. 4. The schematic structure of ZnMgSSe-based SCH LD.
IEEE JOURNAL. OF SELECTED TOPICS IN QUANTUM ELECTRONICS, VOL. 1, NO. 2, JUNE 1995
to the blue-emitting CW operation. The separate confinement
heterostructure (SCH) of the LD is schematically shown in
Fig. 4. The epitaxial layers for the laser diodes are grown on
Si-doped GaAs (100) substrates by molecular beamepitaxy
(MBE) at a substrate temperature of 280 "C. The epitaxial
layers consist of an n-type ZnSe-ZnSSe layer, an n-ZnMgSSe
cladding layer, an n-Zn(S)Se optical guiding layer, a ZnCdSe
single quantum well active layer, a p-Zn(S)Se optical guiding
layer, a p-ZnMgSSe cladding layer, a p-ZnSSe-ZnSe layer,
a p-type ZnSe-ZnTe resonant tunneling MQW 1151, and a p-
ZnTe contact layer. The doping concentration in the cladding
layers is from low 1017 cm-3 near the active layer to high
1017 ~m- ~, away from the active layer. The p-ZnSe top
contact layer is chemically etched off leaving a 10-pm-
wide mesa stripe. An insulating layer is deposited on the
exposed p-ZnSSe layer to reduce the current path. Pd-R-Au
multilayered metal [ 131 is then evaporated onto the p-ZnSe
as a p-electrode. Indium metal serves as an n-electrode to
the n-GaAs substrate. Mounting the LD with the p-side down
on a heatsink, room-temperature pulsed operation is obtained
for ZnCdSe-ZnSe-ZnMgSSe SCH lasers with uncoated facets.
Blue-emitting pulsed LD operation has been demonstrated at
wavelengths from 498 nm[30] or 481 nm1311with the SCH
structure and down to 471 nm 1321, [33] with pure ZnSe
as active layer in a double heterostructure (DH) and finally
down to 462.7 nmwith a ZnSe-ZnMgSSe-ZnMgSSe SCH
LD 1341. The emission spectrum of the 471 nmLD is in the
inset of Fig. 5. Based on those results of pulsed operation,
the threshold-current density, Jth, is plotted as a function of
AE,, the energy-gap difference between the active and the
cladding layers. The solid line is the theoretical calculation
developed for conventional 111-V LD's [35]. Shaded areas
show the overflow and nonradiative recombination current.
Experimental data agree well with the solid line. Fromthe
result in Section 111, it is shown that the current overflow is
almost negligible for
AE, 2 0.35 (1)
leading to a low threshold current.
V. CW OPERATION
Facet coatings are an effective way to reduce the threshold-
current density further, since the internal loss a of the LD's
' 0
10 :
1 :
.
v I
Overflow 470 480
L
cn +
l- 0.20 0.30 0.40 0.50
a Eg (eV)
Fig. 5. Threshold current density of RT pulseoperation as afunction of the
energy-gap difference between the active and cladding layers. Solid circles
areof SCH LD's, and theblank circle is of asimple DH LD. The inset shows
the spectrumof the471-nmblue-emitting ZnSe-ZnMgSSe simple DH LD.
being small (2-4 cm-' [36]), the threshold-current density is
predominantly determined by mirror loss.
A. CW Green Operation
The band-gap energies of the ZnCdSe and the ZnMgSSe
are set to be 2.54 eV and 2.94 eV at 77 K, respectively
and the energy-gap difference is, thus, 0.4 eV, satisfying the
formula (1). The LD structure is essentially the same as the
one shown in Fig. 4. The wafer is cleaved into 640-pm-wide
bars, whose facets are coated with reflectivities of 70% for
the front and 95% for the rear. The strips are then cleaved
into 400-pm-wide pellets and each pellet is mounted with the
p-side down onto a copper heat sink. The emission spectrum
taken at RT (296 K) is shown in Fig. 6. The achieved green
emission wavelength is 523.5 nm [SI. The inset shows the
light-output versus injection-current (L-I) characteristics of
the laser diode at RT (296 K) measured under continuous
and pulsed-current conditions. The small difference in the
threshold-current densities for CW and pulsed operation is
ascribed to the fact that the SCH LD has a high characteristic
temperature (To) of 216 K because of the large energy-gap
difference between the active and the cladding layers.
B. CW Blue Operation
CW operation can be achieved by merging the above two
paths by setting AE, - 0.38 eV and Ertive =2.65 eV,
low p-doping in ZnMgSSe as discussed in Section 111-B, but
this problem can basically be solved as shown in Section
VI-A. The LD structure is, again, essentially the same as
Fig. 4, except the larger band-gap energies of the active and
the cladding (3.03 eV) layers. The LD chip has 10-pm-wide
mesa stripes and a 610-pm-long cavity. The reflectivities are
again 70% for the front-facet coating and 95% for the rear.
where EaCtiVe , is . the active-layer's energy-gap. This gives
ISHIBASHI: &VI BLUE-GREEN LASER DIODES
RT
Pulsed 834mW
745
60
h
a
E
v
5 40-
s
L L
20
0 :
, " " I ' " ' I " . ' I " ' ' I " " , ' " ' I " ' '
296K
L=640 p rn
70-95% Reflective
Coated
- 4
ao c. IrnW
h =509 nm
v
-
- 3 g
a
3
c
- 2 ,a
5
- 1 g
7 -
2
0
Fig. 6.
tion at RT. The inset shows the L-I characteristics.
Emission spectra of a green-emitting LD for pulsed and CW opera-
I , .
I
I I
//'
0
0 - - ' " ' . " '
0 50 100
Current (mA)
Fig. 7.
at RT. The inset shows the emission spectra.
L-I characteristics of a blue-emitting LD for pulsed and CW operation
The LD is mounted with the p-side down onto a copper
heatsink. G I and V-I characteristics of the LD measured
at room temperature (296 K) are shown in Fig. 7. The solid
(dashed) line shows characteristics in CW (pulsed) operation
[9]. The threshold current is 93 mA for the CW operation. The
threshold-current density (operation voltage) is 1.5 kA/cm2
(6.3 V). The emission spectra taken at room temperature under
pulsed and CW operation are shown in the inset. Stimulated
emission is seen at the wavelength of 489.9 nm. This is the
shortest wavelength of RT CW operation ever reported. The
characteristic temperature (To) is as high as 126 K.
VI. HIGH-POWER OPERATION AND
HIGH-TEMPERATURE OPERATION
One thing that should be aimed at now is CW operation
under high temperature and at high optical output. Fig. 8
shows the highest pulsed output power of 0.834 W from an
uncoated facet of a ZnMgSSe-based SCH LD with 50-pm-
wide stripe at a wavelength of 517 nm. The threshold-current
"
0 500 1000 1500 2000
Current (mA)
Fig. 8. L-I characteristics of RT high-power pulsed operation of a Zn-
MgSSe-based SCH LD with uncoated facet. The inset shows the L-I char-
acteristics of RT high-power CW operation of another ZnMgSSe-based SCH
LD with uncoated front facet and coated rear facet.
80 I I
(density) is 230 mA (460 A/cm2). Even when the stripe
width is 10 pm, the output power exceeds 420 mW. Detailed
analysis is not yet completed, however catastrophic optical
damage (COD) is not the cause of the failure. The inset shows
L-I characteristics of CW operation for another LD whose
threshold current, voltage, and wavelength are 48 mA, 9 V, and
508 nm, respectively [37]. The optical output power reaches
up to 30 mW, which is high enough for rewritable RAM disk
systems. Fig. 9 shows the aging experiment of CW operation
at 80 OC under automatic power control (APC) at 1 mW [16].
Consumer use requires LD operation under 60 "C ambient,
and this result shows that high-temperature operation of 11-VI
wide-gap LD is basically of no concern.
VII. REMAINING KEY ISSUES
The remaining key issues are-how to achieve high p-
doping in wide-gap ZnMgSSe, and whether it is possible to
establish long-lived 11-VI wide-gap laser diodes.
146 IEEE JOURNAL OF SELECTED TOPICS IN QUANTUM ELECTRONICS, VOL. 1, NO. 2, JUNE 1995
A. High p-Type Doping
As shown in Section 111-D, the highest p-type doping is
limited by the position of the E ~ I , with respect to VBM.
For obtaining higher p-type doping we need to make the
spacing in energy between ESI and the VBM as narrow as
possible. A ZnSe-ZnMgSSe superlattice (SL), as a substitute
to a bulk ZnMgSSe alloy, is considered with the constraint that
the SL structure is so designed that the energy gap between
the electron's and the hole's lowest states is set equal to the
energy gap of the alloy. Even under that constraint, due to
large differences in effective masses for electrons and holes,
we can make the energetic spacing between ESI and the lowest
hole-state in ZnSe-ZnMgSSe superlattice (SL) smaller (by the
amount of A) than that in an alloy of identical energy gap
[38], as shown in the inset of Fig. 10. A is the reduction
brought by the ZnSe-ZnMgSSe SL, compared to the case
with the alloy, in terms of the energy spacing between ESI
and the effective valence-band maxima. In Fig. 10, thick solid
lines show A, and the thin solid lines show the energy
gap of ZnSe-Znl-,Mg,S,Sel-, SL as a function of the
slab thickness in the monolayer (ML). In the calculation,
we assume: a hole effective mass ( mh) of 1.4 mo, a band
discontinuity of 50% [27] (see also Section 111-A) and an
energy-gap for ZnMgSSe of 3.4 eV (at 77 K). As shown in
Fig. 10, A can be as high as 63 meV for (ZnSe)4(ZnMgSSe)lS
SL. ESI being wide apart in energy from VBM, Fermi-Dirac
Distribution is now well approximated by Boltzmann distribu-
tion, and the enhancement in p-doping in the SL compared
to the alloy is roughly exp(A/kT) - cxp(2.5) -12 at
RT. For a more quantitative estimation of the enhancement,
precise determination of AE, and mh is needed, but generally
speaking an enhancement of a factor of 5 (if mh =0.6
mo, we have A =39 meV [38]) to 10 is expected for
(ZnSe),(ZnMgSSe), SL with m =4 - 6 and n =10 - 15.
At the same p-type concentration, on the other hand, carrier-
confinement potential can be set larger by ( l +E, / Eh) A in the
case of the SL cladding layers than in the case of conventional
alloy cladding layers, where &(Eh) is the electron (hole)
subband energy in the SL. Thus, currently available 463-nm
RT pulse mode LD can be scaled up into a CW LD with a
wavelength of 450-460 nm, or even less.
B. Reliability
The other key issue is the reliability of 11-VI wide-gap LD.
A study is made on whether the bonds in wide-gap 11-VI
materials are strong enough to make long-term CW-operation
possible.
studies
showed that the degradation is due to preexisting defects,
i.e., stacking faults and related threading dislocations.
Arrow-head-like triangular dark regions were observed in
degrading LD's and LED's, and these dark region have
been identified by transmission electron microscopy (TEM) to
the dislocation networks that are nucleated in quantum-well
region at threading dislocation originating from stacking faults
[391-[41]. (100) dark line defects [42], [43] are also seen in
degraded 11-VI blue-green LD's, particularly in better samples
GaAs-Buffer as a Countermeasure: Degradation
2
r
ln
10
Y
._
E
a,
N
'?
56meV
2 " " " " " ' " ' 1
2 5 10 15
Znl-xMg,SySel-y Thickness (ML)
Fig. 10. Thick solid lines show A, the difference in Esl-VBM spacing
between the SL and the alloy with an identical energy gap Et""" =EsL
Thin solid lines show EgL. Both A and E:L are plotted as a function of
the ZnSe- and ZnMgSSe-slab thickness in monolayer (ML). The inset shows
the schematic band structure of the ZnSe, Znl -zMgzS, Sel -t, alloy and
ZnSe-Znl-,-Mg,S,Sel_, SL. In the calculation, hole effective mass of 1.4
7 n 0 and the band discontinuity of 50% are assumed [27].
9 '
in terms of device lifetime. Point defects are indicadet to be
responsible for this process [42].
Improving the substrate/epi-layer interface is important to
eliminate stacking faults and related misfit dislocations. The
problemof our early LD's [8], [9] was that the interface is
degenerately singular. The interface is an epi-layedsubstrate
interface and also a 11-VI/III-V interface concurrently. To
remove the degeneracy, one can insert a GaAs buffer [441.
With the GaAs-buffer layer, the epi-layer/substrate interface is
now just a well-established homogeneous growth of GaAs and
the 11-VI/III-V interface is cleanly made under the ultra-high
vacuumof MBE.
The LD structure is the same as the one shown in Fig. 4,
except that there is an n-type GaAs buffer layer inserted
between the substrate and the n-ZnSe and the active-layer
is a ZnCdSe-ZnSSe MQW. The LD has a 5-pm-wide stripe
and a 610-pm-long cavity with facet coatings of 70%/95%
reflectivity. Stimulated emission is observed at a wavelength of
507 nmunder CW operation. The threshold-current (density)
is about 60 mA (- 1.3 kA/cm2). Even though the current
density is relatively high, one of the LD's had operated for
0.67 hour and another LD for 1 hour at room temperature
under optical output power of 1 mW, as shown in Fig. 11.
This improvement is attributed mainly to the reduction of the
preexisting defect density down to lop5 cmp2 [16]. The inset
of Fig. 11 shows the LD lifetime thus obtained (solid circles)
as a function of the preexisting defect density together with
experimental data obtained for LED's operated with current
density of -5 A/cm2 (blank circles) [45]. Both for LED's and
for LD's, the device lifetime T is inversely proportional to
the defect-density squared. The enhancement of the lifetime
for the LED's compared with the LD's is roughly given by
the reciprocal ratio of their operation current densities raised
up to the lst to 2nd power, which is just like the case in
conventional 111-V LD's.
ISHIBASHI: n-VI BLUE-GREEN LASER DIODES 747
Analysis of Degradation Process: Wenow make a macro-
scopic discussion about the degradation process. Letting N( t )
be the time-dependent defect density, we expand N( t ) into
power series of time. Then, taking the leading term, we have
for t >>t o [46]
where No is the initial defect density and E, is estimated
to be -0.5 --0.7 eV from the temperature dependence of
the device lifetime [42]. Contributions of other parameters,
such as current density, are included in t o that determines
the time-scale. If in the worst case the bonds in 11-VI wide-
gap semiconductors, being more ionic and less covalent than
those 111-V, are too weak, the defects would self proliferate
under a recombination enhanced defect reaction (REDR) and
the defect density should diverge. In this case,
y >l(or y >>1)
(3)
in (1). This means N( t ) increases superlinearly in the wide-
gap 11-VI compounds, and the degradation proceeds at a
catastrophic speed. This occurs if the defects incorporated (or
generated) in the active-layer at one moment in time serve
as the defect-generation sites for the next moment in time,
completing a positive feedback loop.
We assume LD dies when N( t ) reaches at a critical defect
density, N, at t =r, r being the device lifetime. Then (2)
gives r by the following equation
(4)
By definition, we can take NO as the preexisting defect
density, and then (4) reproduces quite well the defect-density
dependence of the device lifetime in the inset of Fig. 11.
Thus, the slope obtained in the inset, or the defect-density
dependence of the lifetime, gives
log(r/to) =const. - (I /?) log(N0).
l / Y - 2,
or
y - 1/2 <1
( 5)
which is in marked contrast to (3), indicating that no cata-
strophic degradation is taking place in degrading active-layers
of 11-VI LD's. Furthermore, the fact that the identical y applies
to both LED and LD and that the enhancement in the lifetime
seen for the LED's compared to the LD's is simply governed
by ratio of their operation current densities, suggest that the
optical field never accelerates the defect generation. Thus, the
supposedly critical issue of bond energy <energy gap in 11-VI
wide-gap materials will be never crucial in practice. Equation
(2) thus leads to
(6)
N( t ) increases just sublinearly as a function of time. Equation
(6) indicates that the degradation is due to a diffusion-like
process. This result supports the speculation made by Guha et
al., concerning the progressive darkening of degrading active
region in LD's and LED's [42]. Identifying such point defects
~ ( t ) =const. . t l / ' .
350, . . I ' I - 35
2 300
E
250
g 200
,.
4 ' h
LD SONY
OLEDNCSU I 1
U)
15 . ;
10 g
0
5
0 10 20 30 40 50 60"
Aging Time (min.)
Fig. 11. 0.67-hour and I-hour-long RT CW operation of ZnMgSSe-based
SCH LD's with a GaAs buffer layer. The inset shows the preexisting
defect-density dependence of the device lifetime for LED's [35] and LD's.
The LED was operated at current density of -5 A/cm2, and the LD's at more
than two orders of magnitude larger current density.
and reducing themare currently of increasing importance for
further improving the device lifetime. Indeed, there is a scatter
in lifetime for the defect density of -lo5 cm-' presumably
due to some extrinsic damages made during processing or even
in growth sequence, but the scatter never changes the slope
in the inset substantially, keeping invariant the main result
obtained above, i.e, the 11-VI LD can intrinsically live long
in a time scale where a minute or a hour can be regarded as
negligible. This is expected, because for the defect density of
-3 x lo4 cm-' or less, we have no preexisting defect in
average in a typical LD stripe of -10 pm x 250 pm, where
the device lifetime is no longer limited by rapid degradation,
but by gradual degradation due to the point defects.
VIII. SUMMARY
ZnMgSSe, being fully lattice-matched to GaAs (001) sub-
strates with an energy gap tunable up to - 4.5 eV, forms a
type I heterostructure with ZnSe to serve as optimum cladding
layers for blue and green LD's. The ZnMgSSe-based system
has made possible the continuous-wave (CW) operation of
both a green LD with a wavelength of 523 nmand a blue
LD with a wavelength of 489.9 nm at room temperature.
Also demonstrated-are CW operation up to 353 K at 1 mW,
room-temperature pulsed operation up to 0.834 W output
power, CW operation with 30 mW output power, a threshold
voltage of 4.7 V and a room-temperature pulsed operation for
a blue 471-nm LD. The device characteristics of the 11-VI
wide-gap LD's are becoming as good as those of established
111-V materials-based LD's, except for the device lifetime.
Remaining key issues are high p-doping in wide-gap ZnMgSSe
and good reliability of 11-VI wide-gap LD's. For the first
issue, valence-band engineering through using a superlattice
based on the amphoteric native defect model, is proposed
for removing the doping limit in alloy semiconductors, lead-
ing to high-doping densities that are otherwise impossible
if one uses alloy semiconductors. The doping concentration
could be increased about one order of magnitude larger for
(ZnSe),(ZnMgSSe), SL with m =4 - 6 and n =10
- 15 compared to conventional structures. This will lead
748 IEEE J OURNAL OF SELECTED TOPICS IN QUANTUM ELECTRONICS. VOL. 1. NO. 2, JUNE 1995
to a deep-blue-emitting laser diode. As for the reliability
issue, room-temperature (RT) continuous-wave (CW) opera-
tion with a lifetime of one hour has been achieved using
GaAs-buffer layer for ZnCdSe-ZnSSe-ZnMgSSe separate-
confinement-heterostcture (SCH) LDs. The analysis of the
degradation process shows that no catastrophically fast degra-
dation occurs when 11-VI LDs degrade. We believe that we
can establish the reliability of the ZnMgSSe-based LDs, by
preventing the start of rapid degradation by eliminating the
preexisting defects, and slowing the gradual degradation by
reducing point defects. It is anticipated that the blue and green
LDs will blossom colorfully in the near future.
ACKNOWLEDGMENT
The author would like to thank Dr. T. Ohata for discussions
concerning the degradation mechanism. Thanks are extended
to Drs. N. Nakayama, S. Ito, H. Okuyama, and K. Nakano
for stimulating discussions. The author is also grateful to
Drs. Y. Mori, J. Seto, and Director T. Yamada for continual
encouragement.
REFERENCES
[l ] M. A. Haase, J . Qiu, J. M. Depuydt, and H. Cheng, Appl. Phys. Lett,
vol. 59, p. 1272, 1991.
[2] R. M. Park, M. B. Troffer, C. M. Rouleau, J . M. DePuydt, and M. A.
Haase, Appl. Phys. Lett., vol. 57, p. 2127, 1990.
[3] K. Ohkawa, T. Karasawa, and T. Mitsuyu, Jpn. J. Appl. Phys., vol. 30,
p. L152, 1991.
[4] H. Cheng, J. M. DePuydt, M. A. Haase, and J. Qiu, LEOS topical
merting on epitaxial materials and in situ processing for optoelectronics
devices, Newport Beach, CA, 1991.
[5] A. V. Nurmikko, R. L. Gunshor, N. Otsuka, and M. Kobayashi, Int.
Con$ Solid-State Dev. Mater., Tsukuba, J apan, p. 342, 1992.
[6] J . M. Gaines, R. R. Drenten, K. W. Haberern, T. Marshall, P. Mensz,
and J . Petrozzello, Appl. Phys. Lett., vol. 62, p. 2462, 1993.
[7] M. A. Haase, P. F. Baude, M. S. Hagedorn, J . Qiu, J . M. DePuydt, and
M. Cheng, Appl. Phys. Len., vol. 63, p. 2315, 1993.
[SI N. Nakayama, S. Itoh, T. Ohata, K. Nakano, H. Okuyama, M. Ozawa, A.
Ishibashi, M. Ikeda, and Y. Mori, Electron. Lett., vol. 29, p. 1488, 1993.
[9] N. Nakayama, S. Itoh, H. Okuyama, M. Ozawa, T. Ohata, K. Nakano,
M. Ozawa, M. Ikeda, A. Ishibashi, and Y. Mori, Electron. Lett., vol.
29, p. 2194, 1993.
[lo] A. Salokatve, H. J eon, J . Ding, M. Hovinen, A. V. Nurmikko, D. C,
Grillo, Li He, J . Han, Y. Fan, M. Ringle, R. L. Gunshor, G. C. Hua,
and N. Otsuka, Electron. Lett., vol. 29, p. 2192, 1993.
[ 1 I] Y. Lansari, J. Ren, B. Sneed, K. A. Bowers, J . W. Cook, J r., and J . F.
Schetzina, Appl. Phys. Lett., vol. 61, p. 2554, 1992.
[I21 H. B. Michaelson, J. Appl. Phys., vol. 48, p. 729, 1977.
[13] M. Ozawa, F. Hiei, A. Ishibashi, and K. Akimoto, Electron. Lett., vol,
29, p. 503, 1993.
[14] Y. Fan, J . Han, L. He, J . Saraie, R. L. Gunshor, M. Hagerott, H. J eon,
V. Nurmikko, G. C. Hua, and N. Otsuka, Appl. Phys. Lett., vol. 61, p.
3160, 1992.
[15] F. Hiei, M. Ikeda, M. Ozawa, T. Miyajima, A. Ishibashi, and K.
Akimoto, Electron. Len., vol. 29, p. 878, 1993.
[16] S. Itoh, N. Nakayama, S. Matsumoto, M. Nagai, K. Nakano, M. Ozawa,
H. Okuyama, S. Tomiya, T. Ohata, M. Ikeda, A. Ishibashi, and Y. Mori,
Jpn. J. Appl. Phys., vol. 33, p. L938, 1994.
[17] T. Marshall, J . Gaines, J . Petruzzello, R. Drenten, P. Mensz, and K.
Haberren, in 7th IEEE Lasers and Electro-Opt. Society Annu. Meet.,
VS4.1, Boston, 1994.
[18] L. Pauling, The Nature of the Chemical Bond. Ithaca, NY: Cornell
Univ. Press, 1960, p. 246.
[19] H. Okuyama, K. Nakano, T. Miyajima, and K. Akimoto, Jpn. J. Appl.
Phys., vol. 30, p. L1620, 1991.
[20] -, J. Crys. Growth, vol. 117, p. 139, 1992.
[21] M. Ukita, H. Okuyama, M. Ozawa, A. Ishibashi, K. Akirnoto, and Y.
Mori, Appl. Phys. Lett, vol. 63, p. 2082, 1993.
[22] H. Okuyama, Y. Morinaga, and K. Akimoto, High temperature blue
lasing in photopumped ZnSSe-ZnMgSSe double heterostructure, in
7th Int. Con$ Molecular Beam Epitaxy, Schwabisch Gmund, Germany,
1992, TUP 2.
[23] Y. Morinaga, H. Okuyama, and K. Akimoto, , Jpn. J. Appl. Phys., vol.
32, p. 678, 1993.
[24] W. Walukiewicz, Phys. Rev., vol. B 37, p. 4760, 1988.
[25] -, Mater. Res. Soc. Symp. Proc., vol. 300, p. 421, 1993.
[26] -, Mater. Sci. Forum (Trans. Tech., 1993). vol. 143-147, p. 519.
[27] K. Kondo, H. Okuyama, and A. Ishibashi, Appl. Phys. Lett., vol. 64, p.
3434, 1994.
[28] R. Gunshor, J . Han, and A. V. Nurmikko, in 8th Int. Con$ Molecular
Beam Epitaxy, Bll-1, Osaka, J apan, 1994.
[29] H. Okuyama, T. Miyajima, Y. Morinaga, F. Hiei, M. Ozawa, and K.
Akimoto, Elec. Len., vol. 28, p. 1798, 1992.
[30] S. Itoh, H. Okuyama, S. Matsumoto, N. Nakayama, T. Ohata, T.
Miyajima, A. Ishibashi, and K. Akimoto, Electro. Lett., vol. 29, p. 766,
1993.
[31] H. Okuyama, S. Itoh, E. Kato, M. Ozawa, N. Nakayama, K. Nakano, M.
Ikeda, A. Ishibashi, and Y. Mori, Electron. Lett., vol. 30, p. 415, 1994.
[32] T Ohata, S. Itoh, N. Nakayama, S. Matsumoto, K. Nakano, M. Ozawa,
H. Okuyama, S. Tomiya, M. Ikeda, and A. Ishibashi, Device structures
and characteristics of laser diodes with AzMgSSe-cladding layers,
in Proc. Int. Workshop on ZnSe-Based Blue-Green Laser Structures,
Wurzburg, 1994, p. 13.
[33] H. Okuyama, E. Kato, S. Itoh, N. Nakayama, T. Ohata, and A Ishibashi,
Device structures and characteristics of laser diodes with AzMgSSe-
cladding layers, Appl. Phys. Lett., vol. 66, p.656, 1995.
[34] D. C. Grillo, J . Han, M. Ringle, G. Hua, R. L. Gunshor, P. Kelkar, V.
Kozlov, H. J eon, and A. V. Nurmikko, Electron. Len., vol. 30, p. 2131
1994.
[35] S. Itoh, N. Nakayama, T. Ohata, M. Ozawa, H. Okuyama, S. Matsumoto,
K. Nakano, M. J keda, A. Ishibashi, and Y. Mori, Jpn. J. Appl. Phys.,
vol. 33, p. L639 1994.
[36] K. Kondo, M. Uluta, H. Yoshida, Y. Kishita, H. Okuyama, S. Itoh, T.
Ohata, K. Nakano, and A. Ishibashi, J. Appl. Phys., vol. 76, p. 2621,
1994.
[37] N. Nakayama, H. Okuyama, E. Kato, S. Itoh, M. Ozawa, T. Ohata, N.
Nakano, M. Ikeda, A. Ishibashi, and Y. Mori, Electron. Lett., vol. 30,
p. 568, 1994.
[38] A. Ishibashi, unpublished.
[39] S. Guha, J . M. DePuydt, M. A. Hasse, J . Qiu, and H. Cheng, Appl.
Phys. Lett., vol. 63, p. 3107, 1993.
[40] G. C. Hua, N. Otsuka, D. C. Grillo, Y. Fan, J . Han, M. D. Ringle, R.
L. Gunsher, M. Hovinen, and A. V. Nurmikko, Appl. Phys. Lett., vol.
65, p. 1331, 1994.
[41] S. Tomiya, E. Morita, M. Ukita, H. Okuyama, S. Itoh, K. Nakano, and
A. Ishibashi, to bepublished in Appl. Phys. Lett., vol. 66, p.1208, 1995.
[42] S. Guha, H. Cheng, J . M. DePuydt, M. A. Haase, J . Qiu, B. J . Wu, and
G. E. Hofler, Appl. Phys. Lett., vol. 65, p. 801, 1994.
[43] L. H. Kuo, L. Salamanca-Riba, G. M. Haugen, J . M. DePuydt, H. Cheng
and B. J . Wu, to be published in Appl. Phys. Lett.
[44] R. L. Gunshor, L. A. Kolodziejski, M. R. Melloch, M. Vaziri, C. Choi,
and N. Otsuka, Appl. Phys. Len., vol. 50, p. 200, 1987.
[45] J . F. Schetzina, R. P. Vaudo, and J . W. Cook, J r., Integrated heterostruc-
ture devices composed of 11-VI materials with Hg-based contact layers,
presented at 6th. Int. Con$ 11-VI Compounds and Related Optoelectron.
Mater., ThB-1, Newport, RI, 1993.
[46] A. Ishibashi, S. Itoh, and T. Ohata, unpublished.
Akira Ishibashi (M91). was born in Saga, J apan in
1958. He received the Ph.D. degree fromUniversity
of Tokyo, in 1990.
After spending years at KEK and Lawrence
Berkeley Laboratory he joined Sony, in 1983 .
He studied GaAs JFET, AIAs-GaAs superlattices,
and nanostructures at Sony. He was a visiting
faculty member at Loomis Laboratory, University
of Illinois, in 199C1991. At Sony, having finished
the development of AlGaInP red-emitting low noise
laser diodes, he has been engaged in the physics of
ZnMgSSe blue-green laser diodes.