Recent Advances in Optically Pumped Semiconductor Lasers

bentgalaxyΗμιαγωγοί

1 Νοε 2013 (πριν από 3 χρόνια και 7 μήνες)

254 εμφανίσεις





Recent Advances in Optically Pumped Semiconductor Lasers

Juan Chilla, Qi-Ze Shu, Hailong Zhou, Eli Weiss, Murray Reed & Luis Spinelli
Coherent Inc. 5100 Patrick Henry Dr. Santa Clara, CA 95054
ABSTRACT
Optically pumped semiconductor lasers offer significant advantages with respect to all traditional diode-pumped solid
state lasers (including fiber lasers) in regards to wavelength flexibility, broad pump tolerance, efficient spectral and
spatial brightness conversion and high power scaling. In this talk we will describe our recent progress in the lab and
applying this technology to commercial systems. Results include diversified wavelengths from 460 to 570nm, power
scaling to >60W of CW 532nm, and the launch of a low cost 5W CW visible source for forensic applications.

1. INTRODUCTION

Semiconductors are in many ways ideal laser materials. They are suitable for both optical pumping and direct electrical
pumping, they are able to produce high optical gain and their quantum efficiency is high. Their emission is not limited to
discrete lines set by atomic levels but can instead be chosen by design. Similarly multi-layered structures with varying
index of refraction can be fabricated with great precision.
The same physical processes that are responsible for such advantages impose limitations on the kind of devices that can
be realized. In order to achieve optical gain in semiconductor, the active material cannot have arbitrary shape: one of the
dimensions must be small (a few microns in the case of optical pumping, around one micron or less for electrical
pumping), thus defining a plane that dictates the geometry of the laser devices.
To date, the typical (and enormously successful) implementation of the semiconductor laser is what is commonly known
as a “laser diode”, variations of which can be found everywhere. Low power devices are used in your CD drives and
carry phone conversations and data through optical communication lines, high power diodes pump solid state laser
materials and are used in material processing applications.
Laser diodes are electrically pumped monolithic devices, contained within a “chip” of semiconductor material with
typical dimensions around one millimeter. In spite of its many applications, this particular implementation of the
semiconductor laser severely limits the performance of semiconductors as laser materials.
If laser emission is in a direction contained in the plane, the most common case known as “edge emitters”, the laser is
optically just a waveguide with gain, the modal characteristics of the laser defined by the waveguide. In order to obtain
single mode operation the mode size has to be small (typically 1 × 3
μ
m) and the output power is limited by optical
damage at the facets. High power can be obtained of course, but at the expense of brightness.
Even single mode edge emitters are astigmatic and highly divergent, in order to obtain round beams with lower
divergence the active region can be placed between two highly reflective mirrors, commonly made out of semiconductor
materials. In that case the light is emitted perpendicular to the plane of the device. The resonator Fresnel number for this
configuration (known as VCSEL, i.e. Vertical Cavity Surface Emitting Laser) is very high, and again there is a severe
limitation to the power generated in a single mode. Contrary to the case for edge emitters, VCSELs cannot be practically
scaled to high power by increasing the emission area. To bring the pump current to the center of the active region highly
doped, or very thick layers would be needed right in the optical path, where they would cause unacceptable absorption
loss.





2. OPTICALLY PUMPED SEMICONDUCTOR LASERS

In order to take full advantage of semiconductors materials, our approach is to use them as we would any other solid-
state laser material, pumping it optically and building it into a conventional laser cavity (Fig. 1). By using optical
pumping we avoid the optical losses associated with doped materials, and we gain the ability to deliver the energy
precisely where it is needed, while at the same time the fabrication process is simplified. By not restricting ourselves to a
monolithic implementation we can use discrete optical components and free space propagation to obtain the desired
characteristics of the output beam. The ability to access the intracavity radiation allows us to include a frequency
doubling crystal and expose it to high circulating power
1
.
Output Coupler
Pump Diode
Pump Delivery Optics
OPS chip on Heatsink
Optical Cavity

Fig. 1. Schematic representation of an optically pumped semiconductor laser

The OPS chip (Fig. 2) consists of two sections, both semiconductor alloys grown by MBE or MOCVD epitaxy. The top
section is the gain medium and the bottom section is a high reflecting mirror that constitutes one of the ends of the
cavity. The gain at the laser wavelength is provided by narrow layers of InGaAs (quantum wells), the composition of the
spacer layers is chosen such as to make them transparent to the laser wavelength, and strongly absorbent to the pump
wavelength (around 800 nm)
2
. With this arrangement, all of the volume of the gain section is available for pump
absorption, and the carriers that are generated decay rapidly to the lower energy levels available at the quantum wells,
where population inversion is achieved.
The quantum wells are located at the antinodes of the standing wave pattern of the linear resonator. This periodic gain
structure makes optimum use of the available gain and eliminates any possibility of spatial hole burning, making it ideal
for single longitudinal mode operation.





Confinement Layer
Quantum Wells
HR Mirror
Pump Absorption
Layers
Heat Sink
Pump
Laser Out

Fig. 2. The OPS chip (optically pumped semiconductor). The different layers and their functions are indicated.

Optical pumping of semiconductors had long been used as one of the early steps in the development of new
semiconductor lasers, but usually as a research tool, not with the objective of creating a practical device but as a means
of learning about material properties, i.e. as a step towards the ultimate goal of creating a monolithic, electrically driven
device, a laser diode. Only recently there has been increased academic interest on the technology as a means to
producing unique practical devices, a review of the theory of operation and recent activity can be found in ref. 3.

3. SEMICONDUCTORS AS SOLID STATE LASER MATERIALS

In our description so far we have concentrated on the advantages of using semiconductors in the OPSL configuration
instead of the more traditional laser diode configurations, in this section we will discuss what are their differences and
advantages respect to traditional solid state laser materials.
3.1 Wavelength Flexibility
In traditional solid state laser materials the energy levels involved in the laser transitions are those of a dopant or
impurity in a crystal or glass. The levels in play are essentially atomic levels, with some limited modification caused by
the solid matrix that surrounds these atoms. As atomic levels are discrete, so are the available transitions. Many
applications for lasers were developed using bulky inefficient gas lasers, or gas laser pumped dye lasers. For some of
those applications the exact wavelength used is not important, and can thus be served by solid state lasers, one example
of such application is pumping of Ti:Sapphire ultrashort pulse lasers, initially carried out with Ar ion lasers at 488 and
514 nm, now almost exclusively done with frequency doubled Nd lasers at 532 nm. Other applications require a
particular wavelength, for reasons related to the chemistry of the application. An example of these applications is
ophthalmology, that has switched to 532 nm due to the advantages of solid state lasers, but is much better served at
577 nm, initially achieved with dye lasers.
For semiconductors on the other hand, the laser transitions are the bands of the solid material itself. Traditional diode
laser technology is based on epitaxial growth of material on top of crystalline substrates. Two III-V compounds have
been developed significantly and there are established processes for their manufacture and use as substrates for epitaxial
growth, they are GaAs, used for near IR and visible lasers, and InP, used in long wavelength communication devices.
The wavelength flexibility of semiconductors comes from the possibility of growing alloys, i.e. mixtures of two or more




III-V compounds. The ones most commonly used are GaAs, InP, AlAs, InAs and GaP. When these material alloys are
grown in crystalline form many of the physical properties of the resulting material are simply weighted averages of the
properties of the constituting compounds. One of those physical properties is the bandgap energy, i.e. the emission
wavelength for a semiconductor laser. This is usually represented in the following diagram, that also shows the lattice
constant of the different materials. Epitaxial growth is only possible when the free standing lattice constant of the desired
alloy is very close to that of the substrate.
5.4
5.6
5.8
6
6.2
6.4
6.6
0
0.5
1
1.5
2
2.5
AlP
AlAs
AlSb
GaP
GaAs
GaSb
InP
InAs
InSb
Lattice Constant (A)
BandgapEnergy (eV)
5.4
5.6
5.8
6
6.2
6.4
6.6
0
0.5
1
1.5
2
2.5
AlP
AlAs
AlSb
GaP
GaAs
GaSb
InP
InAs
InSb
Lattice Constant (A)
BandgapEnergy (eV)

Fig. 3. Representation of possible III-V alloys, the shaded areas roughly represent the alloys that can be grown epitaxially
on GaAs and InP, widely available substrates.
Our interest in the OPS technology was initially sparked by the possibility of creating solid state lasers emitting at
488 nm. In 2001, the first commercial lasers based on the OPSL technology were introduced. The low power
Sapphire™ platform was launched with two versions available: 20 mW at 488 nm and 10 mW at 460 nm. Since then,
Sapphire™ has proven its extraordinary reliability and superior performance compared to the air-cooled argon ion laser.
In Fall 2002 the family was expanded with the 200 mW version: Sapphire™ 488-200. Versions available today range
between 10 mW and 500 mW.
We have concentrated on frequency doubled configurations based on chips grown in the InGaAs material system, we
have demonstrated a variety of wavelengths as listed in the table below.

Doubled
Wavelength (nm)
460 479 488 505 532 570
Demonstrated
Power (W)
7 7 15 8 64 9
Application Display Pump Ar Ion Ar Ion Many Ophtalmic

Table 1. Frequency doubled wavelength of different OPS designs and maximum demonstrated power.




OPS structures in the InGaAs material system can be grown to emit in any wavelength roughly within the range spanned
by the table, the particular wavelengths we explored were chosen because of potential commercial application, and
typically have been previously produced with other laser technologies. The limit on the short wavelength side is the
reduced confinement and consequent loss of efficiency due to low indium concentration, on the long wavelength side the
limit is the ability to grow highly strained material and the incorporation of high indium concentration. We have yet to
convince ourselves that we have reached that limit. The powers listed in the table are only partially related to the
inherent efficiency of each design, there is only a small increase in efficiency as wavelength gets longer, but due to
mostly historic reasons our attempts to push the envelope and achieve new records are usually conducted on 980 nm
material, the most mature, or on 1060 nm material, slightly more efficient.

3.2 High Power Scaling
The ability to use a laser material to generate high powers is to a great extent dictated by the ability to remove excess
heat. This is certainly true for solid state laser materials, for which their highest power expressions are obtained by using
geometries that allow for easy removal of the excess heat, i.e. one dimension extremely longer than the other two, as in
fiber lasers, or one dimension much shorter than the other two, as in disk lasers or slabs. Similar to the case of VCSELs,
when pump power is increased in OPS lasers, we observe a roll over effect, i.e. output power increases up to a maximum
and then falls with increasing pump power. Again the same as VCSELs, this experiment can be repeated several times
with identical results, indicating that the limitation in output power is the effect of temperature on laser efficiency and
not optical damage.

Fig. 4. Some recent single chip high power results on different OPS designs
The OPSL architecture lends itself to a natural way of scaling in power by simply increasing the beam size and pumped
area on the OPS chip. By increasing the area in which heat is generated, we increase the amount of power needed to




reach temperatures that have a detrimental effect on laser efficiency. Here the similarities with VCSELs end, as the
external cavity can be redesigned to match the new spot size, the OPSL can be scaled up in power without sacrifice in
optical performance. This procedure was successfully followed in going from the 20 mW product to the 200 mW
product. The same OPS chip is used in both lasers, but cavity design and pump laser are different.
In addition, the materials and techniques employed in mounting the OPS chip can be optimized for heat extraction
efficiency. This approach was employed to obtain the results presented herein. The OPS structure is grown inverted,
with the DBR as the topmost layer. After processing the wafer is cut into individual chips, which are then soldered on
CVD diamond heat-spreaders. The DBR and solder layer are only a few microns thick, thus ensuring the heat only
travels a short distance through high thermal impedance materials before entering the diamond that has very good
thermal conductivity. The heat spreader in turn is soldered to a copper submount for handling of the device. Figure 4
summarizes some of our higher power results out of single OPS chips. In all cases the heat is ultimately removed by
means of a water cooling system, but that is a matter of convenience more than strict need. The chips are pumped by
fiber coupled arrays of diodes (FAP), in some cases more than one of them is arranged to illuminate a single chip.
An effect that must be taken into account when designing an end-pumped high power solid state laser is thermal lensing.
The excess heat is generated in the same volume that is occupied by the laser beam, the surrounding areas are colder.
This difference induces in turn a difference in index of refraction (higher in the center) producing a considerable
focusing effect. The power of the thermal lens depends on the amount of power being dissipated, and as a consequence
solid state lasers typically operate in optimum configuration only for a narrow range of pump power. In contrast, the
thermal lens is a minor effect for an OPSL because of the small optical thickness of the structure, furthermore, due to the
area scaling procedure it becomes less of a problem as we increase the power level. Higher power requires larger area,
that in turn results in smaller temperature gradients. We performed experiments to verify this.

Fig. 5. Setup employed to measure thermal lensing on OPS chips
The experiment was conducted on 980 nm chips, but the results apply equally well to the other designs. A chip was used
as one of the mirrors of a Mach-Zehnder interferometer, the chip could be pumped and used within a laser cavity as
usual. The interferometer was illuminated with a single mode 980 nm laser (a 500 mW OPSL), in order to detect any
phase disturbance caused by the chip relative to a reference flat mirror. The reflection of the beam on the chip happens in
the DBR after going through the active region, in that way the interferometer detects not only changes in shape but also
dephasing due to change in index of refraction. The setup is able to resolve changes of the order of 20 nm.
Pump Beam
980 nm SM Laser
OPS Chip
Wedge

Wedge

HR mirror
Laser cavity

CCD Camera
Pump Beam
980 nm SM Laser
OPS Chip
Wedge

Wedge

HR mirror
Laser cavity

CCD Camera




50
100
150
200
50
100
150
200
-1
´
10
-7
0
1
´
10
-7
2
´
10
-7
50
100
150
200
50
100
150
50
100
150
-1
´
10
-7
-5
´
10
-8
0
5
´
10
-8
50
100
150

Fig. 6. Measured phase disturbance normalized to surface height (in meters). Left: Laboratory Aluminum mirror pumped
with 0.69W on 420 μm diameter spot. Right: OPS chip pumped with 70 W on 900 μm diameter spot.

The experimental setup was carefully characterized; we verified its calibration by using it to measure a known surface,
previously measured by using a commercial interferometric optical profiler. We also made sure it was able to detect
thermally induced deformation by replacing the chip with a conventional aluminum coated laboratory mirror on glass
substrate. The aluminum coating absorbs part of the incident pump power, and the poor thermal conductivity of the glass
assures the presence of a hot spot at the center of the beam. Non uniform thermal expansion then induces a phase
disturbance that shows as a clearly resolved hump in the pumped spot. When the chip was pumped with up to 70 W on a
spot 900 μm in diameter we clearly saw the fringes move due to overall thermal expansion of the copper submount and
supporting hardware, but the shape of the interferogram did not change, we were unable to detect any lensing effect. In
further efforts pumping tighter with 9 W on a 420 μm diameter spot we were able to detect lensing at the edge of our
resolution, a negligible effect of the order of λ/40.
The fact that thermal lensing is negligible for high power OPS lasers opens an additional avenue for further scaling of
the output power. Scaling up the power without radical changes to the chip structure does not improve the efficiency, so
more output power implies proportionally higher excess heat that has to be removed from the chip. If we continue to
pump harder and harder on a chip, we have large amounts of excess heat that have to be removed from a very small
volume. From purely thermal considerations it would make more sense to separate the heat sources, instead of doubling
(or tripling) the pump power to a chip, use two (or three!) chips. With conventional solid state laser materials thermal
lensing makes the idea of multiple gain elements in a cavity a harder proposition, specially if good mode quality is
required, in the case of OPS it is relatively simple.
0
50
100
150
200
250
0
10
20
30
40
50
60
Pump Power (W)
Power (W)
2 chips
3 chips

0
50
100
150
200
250
300
0
10
20
30
40
50
60
70
Pump Power (W)
Power (W)

Fig. 7. Green (532 nm) output from multi-chip OPS lasers. Left: TEM
00
(M
2
<1.3). Right: Multimode with 3 chips




Figure 7 shows the results of some of our multi-chip experiments, we are able to deliver close to 100 W of pump power
in 900 μm diameter spot on each of the chips. Multiple chips are added as folding mirrors, the cavity design includes
curved mirrors to create a waist on each of the chips. On one end of the cavity a 5 mm long LBO crystal is placed in a
tighter waist within the cavity and provides the frequency doubled output. As seen in the left plot, the increase in power
can be achieved without sacrifice of beam quality, when the mirror distances within the cavity are properly chosen the
laser emits in a single transverse mode, and roll-over powers of 40 and 55 W are obtained with two and three chips
respectively, always with M
2
of less than 1.3. The lack of thermal lensing is evident from the smooth shape of these
curves, the cavity operates the same way at all power levels.
As it is also the case for single chip cavity, our multi-chip laser is capable of providing higher green power with slightly
worse beam quality. The plot on the right shows the green output power off the three chip laser when aligned for
maximum green output and operating slightly multimode. More than 60 W of power are obtained. We did not measure
the M
2
in this configuration, but the increase in power is consistent with what is observed in single chip lasers where the
maximum green power is obtained at M
2
around 4.

3.3 Relaxed Pump Requirements
One of the key advantages of OPS respect to traditional solid state laser materials is the relaxed requirements it imposes
on the pump diodes. First, the absorption spectrum of the semiconductor materials is much broader than traditional solid
state laser materials, essentially any wavelength shorter than the bandgap can be absorbed and used for pump purposes.
There is not, as for traditional laser crystals, the need to accurately control the temperature of the diode to keep the
emission within a narrow absorption line.

0
5
10
15
20
25
30
35
40
-2
0
2
X view
0
5
10
15
20
25
30
35
40
-10
-5
0
5
10
Y view

Fig. 8. Layout and ray tracing simulation of the pump optics




Not only there is great freedom to choose pump wavelength, but also the spatial requirement on the pump beam quality
are significantly reduced. This is an advantage even respect to fiber lasers, that require high brightness pumps, typically
achieved by aggregating several fiber coupled single emitters. The pump radiation is absorbed in just a few microns of
material, therefore the pump beam needs practically no depth of focus, i.e. the acceptance angle is almost unlimited. The
significance of this is that for OPS there is no need to go through the extra step of coupling the pump into a fiber, with
the associated loss of efficiency and increased complexity. Figure 8 shows a design for pump optics used in our concept
demonstration of lasers for rear projection TV
4
. As with other aspects of that prototype we limited ourselves to “off the
shelf” optics, because of that it was necessary to use a total of four lenses to achieve the desired pump spot
characteristics. We have explored the design space and believe the same performance can be achieved using just two
simple custom lenses, making the design simpler and hence more suitable for mass production.

4. APPLICATIONS OF HIGH POWER OPS LASERS

So far we have discussed experiments and laboratory demonstrations of the advantages of OPS lasers. As a commercial
enterprise, Coherent manufactures laser systems, and is now introducing the first products in the high power OPS line.
These laser designs represent an evolution from the design used for the concept TV demonstration, and are targeted at
applications where the advantages of OPS bring value to the user, we will briefly discuss two of those applications.
One of the tools employed in crime scene investigation is the fluorescence of trace evidence. Fingerprints, bodily fluids
and fibers fluoresce under blue-green illumination, either naturally or when sprayed with a dye solution. Lasers have
long been recognized as the preferred source of light for this application, in work pioneered using ion lasers from
Coherent, but the cost, bulk and power requirements of lasers have until now prevented their use outside the laboratory.
Investigators have used portable lamps to take to the field, the evidence found was bagged and photographed in the lab
typically using a laser.

Fig. 9 TracER, battery operated portable laser system for forensic applications.
Coherent recently released the TracER ™ laser system that includes an OPS laser emitting in the green. The system was
designed specifically for forensic application both in the lab and in the field. The laser is fiber delivered to a wand that
provides the user with a variable zoom and the ability to remotely control the laser with up to 5 W out of the hand-piece.




The most unique feature of this system is that it is battery powered. When plugged in, it can simultaneously operate and
charge the batteries, off the electric grid it can run for over 90 minutes before needing a charge. This is made possible by
the high efficiency of the OPS laser. As can be expected that a device used in the field will be subjected to severe
treatment, the system was designed with ruggedness in mind and was subjected to extensive testing. As part of that
testing the OPS laser head survived shocks over 500 G!
Another example where the wavelength flexibility of OPS is a valuable advantage is ophthalmic applications. Studies
conducted by using dye lasers showed that the preferred radiation for the treatment was yellow light, 577 nm in
wavelength. There have been some attempts to address this application with Nd:YAG doubled to 561 nm, but the
difficulties associated with operating the laser material outside of the most efficient green line make it not very practical.
With OPS we can simply dial in the design to produce the desired wavelength.

Fig. 10. Prototype yellow (577 nm) laser for ophthalmic applications

REFERENCES
1. A. Caprara et al, US Patent No. 6167068, US Patent No. 5991318
2. M. Kuznetsov et al, IEEE Photonics Technology Letters, vol. 9 (1997), p. 1063.
3. A. C. Tropper & S. Hoogland, Prog. Quantum Electronics 30 (2006), p. 1
4. J. Chilla et al, Proc. SPIE Vol. 5740 p. 41-47, Projection Displays XI (2005)