Professional Documents
Culture Documents
Creedon, Juliet T. Gopinath, Vincenzo Daneu, Bien Chann, and Robin Huang
solid angle per unit aperture area, and in the limit when
the beam is diffraction limited, it reaches the value P/l2,
where P is the power and l is the wavelength. (See the side- 10 mm
bar entitled Intensity versus Brightness for more on these Position along array
attributes and their role in achieving high local intensity for (b) (c)
10 10
1 1
0.1 0.1
0.01 0.01
10 100 1000 10 100 1000
Number of elements, N Number of elements, N
(a)
10 10
1 1
0.1 0.1
0.01 0.01
10 100 1000 10 100 1000
Number of elements, N Number of elements, N
(b)
FIGURE 5. The beam size D and the transform optic focal length are plotted versus the number of elements in the array for
(a) a semiconductor laser array (Dl = 0.2, Pi = 1 W), and (b) a fiber laser array (Dl = 0.02, Pi = 1 kW). The grating dis-
persion is assumed to be = 2 rad/m. Two values are assumed for the critical optical intensity S (20 kW/cm2 in
blue and 50 kW/cm2 in red). Note the transition from D and f scaling as N (intensity-limited regime) to D and f
scaling as N (bandwidth-limited regime). A design for combining a semiconductor laser array of 1000 elements
would be bandwidth limited, and it would require an output beam D 0.5 cm and f 40 cm.
In designing a combiner for N elements, we must Figure 5 shows the results for two generic examples: a
then select the optics such that the combined beam diam- semiconductor laser array (B/l = 0.2, Pi=1 W) and a fiber
eter is, at least, the larger of DS and DB [D > DS , DB ]. laser array (B/l = 0.02, Pi = 1 kW). Two values for the criti-
The focal length of the transform optic is then given by cal intensity S are assumed: 20 and 50 kW/cm2. (These val-
f = D d /l . If DB > DS (bandwidth-limited design), the ues are intended to be representative of limiting intensities
focal length is for high-performance optical gratings.) As the array size N
f > f B = (Pi /S )1/2N/(gbB). increases, the beam diameter D is initially determined by
the limiting intensity S and scales as N 1/2; at some critical
If DS > DB (intensity-limited design), the focal length is value of N = Nc = (gbB/l)1/2Pi /S, there is a transition to being
limited by the spectral resolution (bandwidth limit), and in
f > fS = N 1/2Pi /(Sl). this regime D scales asN. There is a corresponding behav-
ior for the focal length of the transform optic, f=Dd/l. Dispersive Element the Diffraction Grating
It can be observed in Figure 5 that the focal length for Critical among the various optical elements is the disper-
combining a diode laser array with 300 elements is 10 cm sive element (in our case, a grating). The number of ele-
and it is bandwidth limited. On the other hand, combining ments that can be combined for a given beam diameter
300 fibers would be intensity limited, and it would require D and global bandwidth B is proportional to the mag-
a focal length ranging from 55 m (for S = 50 kW/cm2) to nitude of the grating dispersion b. The choice of grating
90 m (for S = 20 kW/cm2). These long focal lengths would determines the dispersion. The grating needs to efficiently
make the system impractically large if it were to be imple- diffract into the desired order, and it needs to withstand
mented with a single lens or mirror. It is possible to design the high incident optical intensity without distorting the
a compact transform optical system with multiple optical diffracted beam. Fortunately, dielectric gratings have
elements while keeping the intensity on all optical elements been fabricated with 96% diffraction efficiency [15, 16],
below the critical value. Details for such a point design of and with absorption losses of <104, they can accept high
the optics are beyond the scope of this article. intensity with negligible thermo-optical distortions.
An alternate design option makes use of cylindrical
rather than spherical optics and leads to quite smaller val- Grating Geometry
ues of the focal length. A cylindrical lens focuses a beam A beam incident on a grating at an angle relative to
to a line (in one dimension), in contrast to a spherical lens the grating normal generates, in general, multiple dif-
that focuses it to a point (in two dimensions), to a much fracted-order beams, one for each order m. The diffrac-
higher intensity for the same focal length. With cylindri- tion angle m for the mth order beam is governed by the
cal optics, it would be d = Pi1/2/(NS)1/2, fB =(NPi /S)1/2/ grating equation
(bB), and fS = Pi /(Sl). Table 1 lists, for comparison, the
expressions for d, D, and f when using spherical and cylin- d (sin a + sin m) = ml ,
drical optics. Note that the value of the focal length (and
d) scales less dramatically with N when using cylindrical where d is the grating period (see Figure 6). We will
optics; compared to the case of spherical optics, the focal assume that the grating profile is such that most of the
length is reduced by a factor N 1/2. Cylindrical optics are incident power is diffracted into the desired m order. We
generally more challenging to fabricate, so that their use will also assume that at the nominal center wavelength
would appear to be attractive only if they provide signifi- l0, of the band to be combined, the diffracted beam is
cant packaging advantages. at an angle m that is at or near the angle of incidence,
SPHERICAL CYLINDRICAL
DB Nl/(gbB) Nl/(gbB)
The values for DB and DS are the same in both columns. The smaller value, by N1/2, for d in the
cylindrical case carries through to smaller values for fB and fS.
am = tan-1(bl0/2).
Table 2 shows examples for three desired values of FIGURE 6. Grating geometry in the near-Littrow condition.
b. Included are the corresponding values for the grating The diffracted beam satisfies m ~ am, and most of the dif-
period d for m = 1, assuming a nominal l0 = 1 mm. fracted power is in the mth order. The residual zeroth order
All these cases satisfy a > 19.5 degrees (or sin reflection is shown by the faint line.
1 26.5 1.118
2 45 0.7071
4 63.4 0.5590
Multi- beam 1 6
h
1
1... 15 11
D
5
10
e
an
pl
15 15
e
15
ag
Im
Cylindrical
telescope
1 1 y
5
15 6
10 z
11 x
15
FIGURE 7. Concept for two-dimensional wavelength beam combining. The optical layout shows the crossed gratings (the
single vertical grating and the stack of horizontal dispersion gratings). The cylindrical transform lens fy focuses the multi-
wavelength beam (coming from the left) to a series of horizontal lines, one for each wavelength. After one of the beams is
incident on one of the gratings in the stack, the beam pointing in the horizontal plane changes with wavelength. The trans-
form lens fx , together with the cylindrical relay telescope, brings the beam to a focus in the image plane. Each grating in the
stack has an appropriately different dispersion value; as the wavelength changes and the beam transitions for the top grating
to the next in the stack, a new line is initiated in the image plane. Each grating in the stack covers a spectral subrange, and
there is a corresponding line in the image plane. For clarity in visualization, the example in the figure shows three gratings
and five wavelengths per grating.
also be anticipated that the need for a second disperser (see placed at their correct location, a closed-loop configuration
Figure 7) would result in additional combining efficiency with a common output coupler to feed back the combined
loss. (For the mathematics behind Figures 7 and 8, see the beam would automatically determine the wavelength of
appendix of this article.) It is worth highlighting that the each emitter, just as in the one-dimensional closed-loop
two-dimensional grid supporting the location of the emit- case. In the open-loop configuration, the wavelengths of
ters will not be in the form of a series of perfectly straight individual emitters would also need to be precisely con-
lines because of the difference in nonlinear wavelength trolled, just as in the one-dimensional case.
dependence of the diffracted angle in two dimensions In scaling up the number of elements to be combined,
(the dispersion b is just a linear approximation, valid over at some point there is a role for two-dimensional beam
a limited range of wavelengths). The distortion of such a combining. Exactly where that point lies depends on a care-
grid, as small as it is, needs to be anticipated in the design ful comparison of one-dimensional and 2D point designs.
so that the emitters are correctly placed within a small The objective is to disperse a multiwavelength beam into
fraction of their beam diameters in order to maintain good its spectral components, forming a 2D pattern. Used in
beam overlap of the components at the combined beam. reverse, "assembling" the beams provides a method for
Imperfect overlap would lead to a decrease in the far-field 2D WBC. Figure 7 shows a multiwavelength, diffraction-
on-axis intensity of the combined beam, equivalent to a limited beam with diameter D and spectrum spread over a
reduction in beam-combining efficiency. With emitters bandwidth B that is incident from the left on a grating Gy.
1 10
2
0.1 /cm 1
0kW
=2
2
S /cm Cylinder fx
0kW
S =5
0.01 0.01
10 100 1000 10,000 10 100 1000 10,000
Number of elements, N Number of elements, N
(a) (b)
100
FIGURE 8. Projections for scaling a two-dimensional
array of diode lasers. The element laser is diffrac-
Array dimension, Dx' , Dy' (cm)
The design task will specify the optical componentsgrat- sented has been how to design for a large number of ele-
ing and lensesin order to combine N elements that are ments. With emitter beams that are diffraction limited,
placed in a 2D pattern with n rows. Figure8 shows specific the combined beam will also be diffraction limited. As
results for an example in which a 2D array of diode lasers the number N of elements increasesas the bandwidth
is combined. The output beam diameter D values for the allocation B/N per element decreasesthere is a point at
two focal lengths fx and fy , and the dimensions D'x , D'y , of which the spectral linewidth of the emitter element may
the 2D diode array are plotted versus the number N of array no longer be narrow enough and, consequently, its spec-
elements for an array with n=8 rows. tral content gives rise, when dispersed by the grating, to a
far field that is no longer diffraction limited.
Emitter Element Spectrum and Quality Let us consider the example in which the bandwidth
of Combined Beam allocation per element is 10 GHz. This allocation would
So far it has been assumed that the emitter element spec- allow us to combine up to 600 elements within B = 20 nm
trum is very narrow. Each emitter with a wavelength global bandwidth (10 GHz ~ 0.033 nm at 1 m). Fiber laser
within the global band B is assigned a location in the Fou- amplifiers have demonstrated 1 kW within a 10 GHz line-
rier plane. Its emission propagates through the combining width. An array of 600 such elements would produce a 600
optics and exits, filling the output aperture and pointing kW beam; such a beam, however would be approximately
in a direction that is common for all emitters. Also pre- twice diffraction limited (at best, for fill factor g ~ 1). There
Appendix
Two-Dimensional Wavelength Beam Combining of Lasers
Consider a diffraction-limited beam with diameter D The spacing between lines is hk = by , k Dfy so that Ny
and spectrum spread over a bandwidth B that is inci- = B/D lines extend over D'y = by , k B fy where by = <by , k > is
dent on a grating Gy. The beam is dispersed over an angle the average dispersion for grating Gy over the wavelength
by B in the vertical ( yz) plane, where by is the average range B. The total number of combined elements is
dispersion over the wavelength range (bandwidth) B. A
cylindrical lens Ly with focal length fy maps each spectral N = Nx Ny = bx BD/l ,
component at its Fourier plane into a linear footprint of
width w = fy l/D, and the footprint of all the spectral where bx is the average dispersion for the grating stack
components spreads over by B fy. At the Fourier plane of over the wavelength range B. Note that this is the same
Ly, there is a stack with n gratings to disperse each com- expression as for the linear array, the basis for the plots in
ponent beam in the horizontal (xz) plane. Each grating Figure 5. In the 2D concept, the same number N of ele-
has a height h =by D fy so as to intercept components ments will now be distributed in n = Ny rows.
within a bandwidth D = B/n. The period d or the angle of The design task will specify the optical components
incidence a for grating k of the stack (1 < k < n, count- grating and lensesin order to combine N elements in
ing from the top) is such that the diffracted beam for lk a 2D pattern with n rows of elements. The total avail-
= l0 + (k 1/2) D incident on the mid-height level of able bandwidth B is given. The element is specified by its
k-gratingis along the nominal optical axis. For a range power Pi and beam diameters dx and dy, consistent with
of wavelengths D centered at k , that is for k D/2 < not exceeding a critical optical intensity S,
< k + D/2, the diffracted beam is off axis, at angle bx , k (
k) in the horizontal (xz) plane. A lens Lx with focal dx = d/r , dy = d r
length fx , together with a cylindrical relay telescope with and
unity magnification, transforms the beam with wave- r2 = dy / dx,
length to a spot with dimensions
where d = (Pi /S )1/2 and r2 is the element beam aspect
dx = l fx /D ratio. The combined beam diameter D must be large
and enough to satisfy the critical intensity condition
dy = fy/D
D > DS = (NPi /S )1/2,
at location in the image plane given by
and it must also be large enough to provide angular sepa-
x = bx , k ( k) fx ration of adjacent dispersed beams with dispersion bx and
and spatial fill factor g so that
y = by , k ( k) fy .
D > DB = N/( bx B).
As the wavelength changes from 0 to 0 + B, the
beam is mapped to a set of Ny = B/D lines, similar to a It follows that the focal lengths are then given by
raster scan. The number of elements in line k of the scan
is given by fx = dx D/
and
Nx = bx , k DD/l fy = dy D/ .
and are spread over a length D'x = bx , k D fx with a spatial The length D'x of one row is given by
fill factor given by g. D'x = bx D f x ,
the technical staff in the Laser Technology and Applications Group. Bien Chann was previously a technical
Current areas of interest include QCL technology, standoff chemical staff member at Lincoln Laboratory, where
detection, and hyperspectral imaging. He has served as an associ- his research interests covered wavelength
ate editor for the IEEE Journal of Quantum Electronics and has more beam combining (WBC), coherent beam
than 30 publications and four patents. combining, high-power laser systems,
ultrafast lasers, and laser cooling. The
TeraDrive technology is based on his work.
Kevin J. Creedon is an associate staff He has published extensively in the fields
member in the Laser Technology and of atomic physics and high-power lasers, including 16 papers in
Applications Group at Lincoln Labora- peer-reviewed journals. Dr. Chann also holds six patents issued or
tory. Since 2008, he has developed pending in the areas of stabilized diode arrays and stacks, single-
high-brightness laser sources for various frequency diode arrays, WBC of diode stacks, WBC of diode
applications by beam combining. He is stacks independent of imperfections, 2D WBC laser system for
currently working to implement coherent scaling to the MW-power level, and multi-kW beam-combinable
combining architectures using quantum- fiber amplifiers. Dr. Chann holds a doctorate in atomic physics
cascade lasers and pulsed fiber amplifiers. from the University of Wisconsin and a bachelors degree in phys-
ics and mathematics from Stony Brook University. Dr. Chann was
a 2012 R&D 100 award winner for his work on the wavelength
Juliet T. Gopinath received a bachelors beam-combined diode laser. He won the 2012 Berthold Leibinger
degree in electrical engineering in 1998 Innovationspreis (First Prize) for his work with Dense Wavelength
from the University of Minnesota. She was Multiplexing of High-Power Diode Lasers, and was honored with
awarded masters (2000) and doctoral the Museum of Science Invented Here award for his work with
(2005) degrees in electrical engineering wavelength beam combining.
and computer science from the Massachu-
setts Institute of Technology. From 2005
to 2009, she worked as a member of the Robin Huang is a cofounder of TeraDiode
technical staff at Lincoln Laboratory. Her work included wavelength and a member of its board of directors. He
beam combining of eyesafe diode arrays, cryogenic Yb:YAG lasers/ is responsible for business development for
amplifiers, mode-locked semiconductor optical waveguide lasers the government market and contributes to
(SCOWL), high-power eyesafe laser sources, and Raman spectros- internal R&D and new innovations. Prior
copy. She is now an assistant professor at the University of Colo- to joining TeraDiode in October 2009,
radoBoulder in the Electrical, Computer, and Energy Engineering Dr. Huang was a technical staff member at
Department. Her research interests include ultrafast solid-state and Lincoln Laboratory. He investigated and led
fiber lasers, semiconductor lasers, wavelength beam combining, Department of Defense programs on electro-optical materials and
spectroscopy, nonlinear processes in fibers, mid-infrared sources, devices for high-power semiconductor lasers, thermophotovoltaic
and adaptive optics. Dr. Gopinath is the recipient of the National Sci- devices, mid-infrared diode lasers, quantum-cascade lasers, and
ence Foundation Graduate Fellowship (19982001) and an R&D vertical-cavity surface-emitting lasers. He has published over 45
100 Award (2012). She has authored and coauthored 29 journal papers and presented research at more than 35 conferences. Dr.
papers and more than 37 conference proceedings. Huang also has five issued U.S. patents. He holds a doctorate in
applied physics and a masters degree in electrical engineering
from Stanford University, and a bachelors degree in physics with
Vincenzo Daneu was a staff member in electrical engineering from the Massachusetts Institute of Technol-
the Quantum Electronics Group at Lincoln ogy, where he graduated Phi Beta Kappa.
Laboratory until his retirement in 2002. He
worked on solid-state and semiconductor
laser development. He contributed to the
development of laser sources for the genera-
tion of a sodium guide star. He performed
one of the early demonstrations of wave-
length beam combining using a 2 m semiconductor laser array. He
also contributed to the development of sensing technology for early
warning of a biological attack. Since his retirement, he has resumed
work at Lincoln Laboratory as a consultant with DAG Consulting.
His recent activity has mainly been in the area of optical design and
measurements. He holds a doctorate in electrical engineering from
the University of Palermo, Italy.