Soft aperture spatial filtering: 1.

5W in a single spatial mode from a

highly multi-mode laser diode in an external cavity

Mallachi-Elia Meller, Idan Parshani, Leon Bello, David Goldovsky, Amir Kahana, Avi Pe’er∗
Department of Physics and BINA Institute of Nanotechnology,
Bar-Ilan University, Ramat-Gan 52900, Israel
(Dated: August 29, 2022)
arXiv:2208.12303v1 [physics.optics] 25 Aug 2022

Abstract
Broad area laser diodes are attractive for the high optical power they can produce. Unfortunately,
this high power normally comes at the cost of severely reduced spatial coherence since the wide area of
the semiconductor wave-guide is inherently spatially multi-mode (in the slow axis). We demonstrate a
method to majorly improve the spatial coherence of a high-power broad-area diode laser by placing it
in an external cavity that is mode selective. We design the cavity, such that the diode aperture acts as
its own spatial filter, obviating the need for an intra-cavity slit-filter, and optimally utilizing the entire
gain medium. We demonstrate this soft filtering method using wide diodes of 200µm and 300µm widths
and compare its power-efficiency to the standard approach of hard-filtering with a slit. We obtain high-
gain operation in a pure single-mode, demonstrating up to 1.5W CW power at 1064nm with high beam
quality.

Keywords: Laser Diode, External Cavity, Continuous Wave, Single Mode

∗
Correspondence email address: [email protected]

1
I. INTRODUCTION

Semiconductor lasers are advantageous over other types of gain media: in their high electrical-
to-optical efficiency, simple and robust construction, high optical power, tunability, wide range
of available wavelengths that cover the entire VIS-NIR range, and low cost. Consequently, high-
power semiconductor lasers with a good beam quality (brightness) are desirable for many applica-
tions, such as pump sources for solid-state lasers and fiber lasers [1, 2], frequency conversion [3, 4],
direct material processing [5, 6] and medical applications [7]. However, these two requirements,
brightness and power, do not normally go hand-in-hand [8]. In order to increase the output power,
while avoiding nonlinearities and saturation effects, it is necessary to increase the gain-medium
area. This requires wide wave-guides, which are inherently multi-mode. Consequently, the num-
ber of spatial modes increases, which severely degrades the brightness of the diode and its spatial
coherence.

The standard method for overcoming this problem in broad aread amplifier (BAA) diodes is
to try to enforce a single spatial-mode on the diode laser by placing it in an external cavity [9–11]
that stimulates the BAA with spatially selective feedback. The spatial selectivity is normally
achieved by introducing a slit at the Fourier-plane of the diode facet [12, 13], which acts as a
spatial filter to determine the number of allowed spatial modes and their width. Unfortunately,
this hard-aperture method often endures unnecessary losses that hamper its power-efficiency due
to the inherent mismatch between the lowest spatial mode of the diode wave-guide and that of
the slit.

We suggest to mitigate this limitation using a “soft filtering” design of the cavity, where the
diode itself acts as the spatial filter, which eliminates any possible mismatch from the source.
While some "soft filtering" cavity designs were demonstrated in the past in an anamorphic external
cavity [14] , the obtained power was relatively low (120mW in single mode). We demonstrate
soft filtering at much higher power levels with much wider diodes, obtaining up to 1500mW in a
single mode beam.

2
II. MODE-SELECTIVE CAVITY DESIGN - SOFT FILTER VS HARD SLIT

To introduce mode-selectivity that prefers single-mode operation in the laser cavity over multi-
mode, while preserving most of the multi-mode power, we implemented both the standard hard
slit-filtering and our modified soft-filtering, as schematically shown in figure 1. To compensate for
the highly astigmatic beam at the diode output facet, we employ a common anamorphic design
of the cavity, based on two lenses - a spherical lens with a short focus (fast lens) that operates
on both the fast and the slow axes of the diode, followed by a cylindrical lens of a longer focus
(slow lens) that operates only on the slow axis. The fast axis, which is single-mode by definition,
is simply imaged by the fast lens onto the end mirror (figure 1a). The configuration of the slow
axis for standard hard-filtering with a slit is shown in figure 1b, and our soft filtering approach
is shown in figure 1c.

To implement a hard slit filter, the fast lens generates a Fourier-transform of the diode aperture
at its back focal plane. This Fourier plane is then imaged by the slow lens to the end mirror,
where a slit is positioned to filter a single spatial frequency (corresponding to a single plane wave
at the diode) to provide feedback to the cavity. The multi-mode operation can be described as
an incoherent sum of plane waves that propagate at different directions out of the diode, and are
focused onto different points on the end mirror. The slit then selects a range of plane waves at
the end mirror, where the width of the slit selects the number of allowed modes. This enforces a
wide single-mode at the diode facet that covers most of the diode’s aperture. The magnification
ratio of the two images on the end mirror (fast and slow) allows to compensate for the spatial
astigmatism of the beam. As mentioned, the main problem with slit-filtering is the mismatch
between the mode determined by the slit, and the mode determined by the diode, as will be
demonstrated in the results section.

In contrast, the soft filtering approach moves the slow lens away from the imaging condition,
towards the fast lens, which gradually reduces the accepted beam divergence at the diode plane,
until finally only the fundamental mode with the lowest divergence can survive (a single plane-
wave). This condition is reached when the slow-lens and the fast-lens form a telescope on the
slow-axis, where the image of the diode acts as an effective entrance pupil for the beam in the
return path (see figure 1c), which indicates that only a beam with sufficiently low divergence can
safely pass the aperture in the backwards direction. When the separation of those two "aperture

3
Figure 1. Cavity beam shaping mechanism (a) Fast-axis: The fast axis is inherently single-mode due to
the diode structure, in order stabilize the fast axis inside the cavity, a spherical fast lens is located near
the diode at a distance slightly larger than its focus. This forms a magnified, low divergence image of the
diode’s facet in the fast axis on the end mirror. (b) Slit-filtering: The fast spherical lens forms a Fourier
transform of the diode’s facet plane at its back focal plane. This Fourier plane is then imaged by the
cylindrical slow lens onto the end mirror, where the slit is placed. The two colors visualize two different
modes of the cavity, showing the slit’s operation to filter only the desired mode. (c) Soft-filtering: In
the soft-filtering approach, we operate in a plane configuration, employing the diode’s aperture as its
own spatial filter. We shift the position of the slow lens near the telescope position (together with the
fast lens), where the diode’s magnified image acts as an entrance-pupil for the returning beam from the
high-reflector. Single mode operation is enforced when the two-way distance from the effective pupil to
itself exceeds the Rayleigh range of a single mode beam with the pupil’s width.

locations" approaches, or exceeds the Rayleigh range of a single-mode beam with a waist that
matches the aperture width, only this lowest spatial mode can survive. Our filtering mechanism
operates therefore between two spatial stability limits - the hard slit-filtering in figure 1b is

4
equivalent to a concentric cavity and the soft filtering configuration of figure 1c is equivalent to
a planar cavity.

III. RESULTS

To demonstrate the advantage of soft-filtering over hard filtering we used the experimental
cavity configuration of figure 2, which is comprised of two arms around the diode - the right arm,
which implements the spatial filtering (on both approaches) and a left arm, which supports all
modes, and includes a variable output coupler, implemented by a polarizing beam-splitter and
a rotating λ/4 wave-plate that optimizes the output power (see caption of figure 2 for details).
The left side is operated in the concentric configuration where the diode plane (near field) is
Fourier-transformed onto the end mirror (far field). We therefore expect that at the transition
from multi-mode to single-mode the beam profile at the left end mirror will converge into a narrow
profile of minimum width accompanied by a maximal brightness, as illustrated in Fig. 3(a).

Figure 2. Experimental setup: The construction of the cavity is based on a standard linear cavity
where the diode (gain bar) is located in the center. In order to stabilize the fast and slow axis, the
diode facet is located roughly at the focal plane of the two spherical fast lenses (f3). The right cavity
arm contains the spatial shaping mechanism - for hard-filtering a slow cylindrical lens (f1) is placed at a
one-to-one imaging distance of 2f1 from the end mirror and from the Fourier-plane of the diode (at the
back focus of the fast lens). The slit is applied near the right end mirror. For the soft-filtering the slit is
removed, and the slow lens is shifted towards the fast lens. The left cavity arm is multi-mode by design
and contains a variable output coupler in the form of a polarized beam splitter and a λ/4 wave plate. A
CCD camera behind the left HR (R ≈ 99.5%) is used to analyse the laser beam

5
 We employed in our experiments two broad-area diodes, to study low-power and high-power
regimes: One diode of 200µm width and 4mm length, of relatively low power (up to 220mW
multi-mode). Both diodes were AR coated to suppress self-lasing. The low power cavity was
comprised of slow lenses with focal length: f1 = 50mm (right arm, filtering) and f2 = 80mm
(left arm, output coupling) and the fast lenses focal length was f3 = 7.5mm. The high-power
configuration (up to 2.6W multi-mode) used a larger diode with dimensions 300µm × 5mm. In
addition to the AR-coating, the 300µm diode was also placed at a small angle to the optical axis
to suppress even further self-lasing from reflections of the end-facets. The focal lengths of the
lenses in the high-power cavity were f1 = 50mm, f2 = 100mm, and f3 = 9mm.

A. Hard (slit) filtering

To explore the performance of hard-slit filtering in detail we first employed the lower power,
200µm diode in the configuration of figure 2 and measured the output power and the beam
properties for varying slit widths, aiming to quantify the beam purity and the filtering power-
efficiency. The diode was operated at a working current of 2.2A, which is sufficiently above lasing
threshold of the external cavity (≈ 1.7A) to generate a highly multi-mode beam, but safely below
the threshold for self-lasing of the gain chip alone (≈ 2.6A) due to residual reflections from the
diode facets.
In the experiment we measured the output-power and the beam profile at the left end-mirror
(using a CCD camera) for a varying slit width, as shown in figure 3. For each slit width we
optimized the output coupler, to assure the maximal achievable power-efficiency. Single mode
and multi-mode can be distinguished by examining the dependence of the beam irradiance (peak
intensity) and the total power on the slit width. The goal of the method is to reduce the number
of lasing modes. In spite of the slight reduction of the output power as the slit is closed, the
irradiance rises considerably towards the single-mode condition, since the power is spread over
less modes, as shown in Fig. 3(a).
Looking more closely at the beam properties, we measured the beam profile along the slow axis
near the single-mode transition, as shown in figure 3(b). We observe the crossover point where
the slit was sufficient to enforce single mode at 310µm (solid black). This is the optimal point,
where the slit eliminates the additional modes (Gaussian content = 0.9 [15]), while maintaining

6
Figure 3. (a) Normalized output power and the normalized irradiance as a function of the slit size, the
dashed line is the output power and the solid line is the irradiance measured on the cavity end mirror.
For a fully open slit (850µm) the output power is 160mW and the peak irradiance is 40 mm
W
2 at 310µm.

(b) Beam profiles for three different slit sizes around the transition to single-mode operation.

reasonable mode matching with the diode, such that the output power at single-mode is > 160mW,
which is ≈ 75% of the multi-mode power at 2.2A. Closing the slit further (dashed red) remains
single mode (Gaussian content = 0.9), but reduces the power due to increased losses from the
slit also for the single-mode. With a wider slit, an additional mode arises, forming a dual-peaked
beam profile (solid blue), reducing the Gaussian content to < 0.6.

B. Soft-filtering

To demonstrate the soft-filtering method we shift the position of the slow (cylindrical) lens
away from the imaging condition. Figure 4a shows the output power and output irradiance and
figure 4b shows a line scan of the beam profile along the slow axis. both as a function of the

7
Figure 4. Laser preferments as a function of the cylindrical lenses’ (F1) position (xf ). (a) The left axis
(blue dashed curve) is the normalized power, the value at the peak (xf = 92.5mm) is 166mW. The
right axis (solid orange line) is the normalized irradiance, with peak value (xf = 54mm) of 68 mm
W
2 . (b)

Line section of the spatial beam profile as a function of the cylindrical lenses’ position (f1) position, the
2f marker is the concentric configuration – this supports multi-mode operation as can be seen from the
spread of the beam at the point. The f marker is the plane configuration and as can be seen from this
point the beam profile collapses to a single high-brightness point, in the top right corner a measurement
of the Gaussian content across the single mode zone.

position of the slow lens (relative to the back focal plane of the fast lens). The vertical dashed
black lines mark the 2f imaging location, where the beam is highly multi-mode and the telescope
location f , where the beam forms a nice single-mode with high Gaussian content (≈ 0.9), with
power that is nearly the same as the maximum multi-mode power (96%). This is a clear advantage
over the hard filtering that showed only 75% power efficiency at maximum.

The superiority of the soft-filtering is also reflected when examining the beam fill-factor inside
the diode, as shown in Fig. 5 that presents the beam profile on the diode facet (near field) and

8
at the Fourier plane (far field) on the left end-mirror for both hard filtering (top three lines for
different slit widths) and soft filtering (bottom fourth line). The physical borders of the diode
wave-guide are marked on the near-field profiles by vertical red lines. The best possible fill-factor
is represented by the top row, which shows multi-mode operation with no filtering restrictions on
the laser operation. Ideally, the single-mode operation should reach a similar fill-factor. Clearly,
the hard-filtering is not optimal in this regard, with a fill factor of roughly 50%, whereas for the
soft-filtering, the fill factor is nearly ideal. This difference is also reflected in the far field beam,
where the soft-filtering profile is narrower than that of the hard-filtering, which highlights the
fact that the slit mode for hard filtering does not perfectly match the fundamental mode of the
diode wave-guide.

Figure 5. Line section of the spatial beam profile across the slow axis as measured at the far field on the
end mirror (left column) and at the near field on the diode facet (right column). The first three rows
represent hard-filtering of different slit widths – (i) completely open, (ii) partially open and (iii) optimum
for single mode operation. Row (iv) shows single-mode operation using the soft-filtering configuration.

C. High-power results

All the results presented thus far were obtained with the narrower diode of 200µm × 4mm
dimensions, which could be driven only up to ≈ 2.6A before self-lasing occurred, generating
220mW of output power at 2.2A. To explore spatial filtering at higher power levels we introduced
a wider and longer diode (300µm × 5mm at 1064nm) into the experimental configuration. This
wide diode can be driven by much higher currents (> 10A), however, to operate at such high

9
Figure 6. Laser performance in soft and hard filtering. (a) Power-Current curves at single mode operation,
the black line is the power obtained using soft-filtering and the red is the power obtained using hard-
filtering, the blue line is the output power in multi-mode operation at the confocal configuration. (b)
Estimated coupling power to single mode fiber, black line the power received by soft-filtering and the red
is the power received by hard-filtering.

currents, annihilation of self-lasing is absolutely necessary, and only the AR coating is not sufficient
for this. To further suppress self-lasing, the wide diode was placed at an angle to the optical axis,
which directs the residual reflection out of the diode’s wave-guide and prevents it from amplifying.
We then applied both the hard and soft-filtering approaches to the high-power wide diodes and
examined their performance and limitations.
Figure 6a shows power-current (PI) curves of the wide diode in multi-mode (blue), as well as
the optimal single mode configurations for both hard (black) and soft (red) filtering. Figure 6b
presents the estimated power into a single-mode fiber based on the calculated Gaussian content
of the output beam. Clearly, the soft filtering performed better than hard filtering in two aspects:
First, the overall power efficiency was near 60% for soft filtering, noticeably improved compared

10
to hard-filtering. Second, with soft filtering the diode could be driven to higher currents, while
maintaining single-mode operation. Specifically, the maximum current with soft filtering was
8.5A, which generated 1.5W of output power at single-mode, compared to < 1W at 7A for hard
filtering.

Finally we measured the beam quality of the single-mode beam for both soft and hard filtering
using two parameters - M 2 and Gaussian content. The Gaussian content for all drive currents
was between 0.7 − 0.8, and M 2 < 1.5 was obtained when low intensity side lobes and image noise
(contain less than 10% percent of the total power) were omitted (M 2 < 2 for the raw beam image)
as demonstrated in figure 7 for 8.5A.

Figure 7. Beam intensity profile normalized by the peak intensity. Near field - (left): the beam
intensity profile as seen on the end-mirror. (Far field - (right): beam intensity profile, as imaged
outside the cavity. Red dashed lines indicate the raw data, solid blue lines indicate the data with
the side-lobes filtered out in the near field using a slit. M 2 value for the unfiltered beam was 1.9, and
1.3 for the filtered beam.

11
IV. CONCLUSIONS

We demonstrated soft spatial filtering in broad-area, high-power semiconductor lasers, where

the diode aperture itself acts as the spatial filter in an external cavity configuration. We extracted
up to 1.5W of high-brightness power, with 1W power that can be efficiently coupled to a single
mode fiber. We demonstrate an improvement in the single mode power and irradiance compared
to state-of-the-art methods of hard slit-filtering by roughly 50%. Another advantage of our soft-
filtering method is the ability to operate at higher working currents, without being negatively
affected by self-lasing. Shaping the beam inside the cavity efficiently enforced single-mode oper-
ation in a very high-gain regime, which is highly desirable for many applications in the field of
laser engineering.

[1] K. Price, S. Karlsen, P. Leisher, and R. Martinsen, in High-Power Diode Laser Technology and
Applications VIII , Vol. 7583, edited by M. S. Zediker, International Society for Optics and Photonics
(SPIE, 2010) p. 758308.
[2] L. Bao, M. Kanskar, M. DeVito, M. Hemenway, W. Urbanek, M. Grimshaw, Z. Chen, W. Dong,
X. Guan, S. Zhang, D. Dawson, and R. Martinsen, in High-Power Diode Laser Technology and
Applications XIII , Vol. 9348, edited by M. S. Zediker, International Society for Optics and Photonics
(SPIE, 2015) p. 93480C.
[3] K.-J. Boller, B. Beier, and R. Wallenstein, “Properties and frequency conversion of high-brightness
diode-laser systems,” in High-Power Diode Lasers: Fundamentals, Technology, Applications: With
Contributions by Numerous Experts, edited by R. Diehl (Springer Berlin Heidelberg, Berlin, Heidel-
berg, 2000) pp. 225–263.
[4] M. Klein, D.-H. Lee, J.-P. Meyn, B. Beier, K.-J. Boller, and R. Wallenstein, Optics letters 23, 831
(1998).
[5] H. Fritsche, F. Ferrario, R. Koch, B. Kruschke, U. Pahl, S. Pflueger, A. Grohe, W. Gries, F. Eibl,
S. Kohl, and M. Dobler, in High-Power Laser Materials Processing: Lasers, Beam Delivery, Diag-
nostics, and Applications IV , Vol. 9356, edited by F. Dorsch, International Society for Optics and
Photonics (SPIE, 2015) p. 93560I.

12
 [6] B. O. Faircloth, in High-Power Diode Laser Technology and Applications, Vol. 4973, edited by M. S.
Zediker, International Society for Optics and Photonics (SPIE, 2003) pp. 34 – 41.
[7] M. Michalik, J. Szymańczyk, M. Stajnke, T. Ochrymiuk, and A. Cenian, Micromachines 12 (2021),
10.3390/mi12060710.
[8] M. Fallahi and R. Bedford, in Semiconductor Lasers, Woodhead Publishing Series in Electronic and
Optical Materials, edited by A. Baranov and E. Tournié (Woodhead Publishing, 2013) pp. 81–120.
[9] M. Chi and P. M. Petersen, “15.1 edge emitters with external cavities: Datasheet from landolt-
börnstein - group viii advanced materials and technologies · volume 1b3: “laser systems” in springer-
materials (https://doi.org/10.1007/978-3-642-14177-5_15),” Copyright 2011 Springer-Verlag Berlin
Heidelberg.
[10] V. Raab and R. Menzel, Opt. Lett. 27, 167 (2002).
[11] B. Mroziewicz, Opto-Electronics Review 16, 347 (2008).
[12] S. Wolff, A. Rodionov, V. Sherstobitov, and H. Fouckhardt, IEEE Journal of Quantum Electronics
39, 448 (2003).
[13] M. Chi, N.-S. Bøgh, B. Thestrup, and P. M. Petersen, Applied Physics Letters 85, 1107 (2004),
https://doi.org/10.1063/1.1783017.
[14] S. Mailhot, Y. Champagne, and N. McCarthy, Appl. Opt. 39, 6806 (2000).
[15] S. Ruschin, E. Yaakobi, and E. Shekel, Applied optics 50, 4376 (2011).

13