Anda di halaman 1dari 16

Vol. 26, No.

16 | 6 Aug 2018 | OPTICS EXPRESS 21092

Multifunctional binary diffractive optical


elements for structured light projectors
OMRI BARLEV* AND MICHAEL A. GOLUB
Department of Electrical Engineering, Tel Aviv University, Tel Aviv 69978, Israel
*
omribarlev@gmail.com

Abstract: A set of diffractive optical elements for multiple-stripe structured illumination was
designed, fabricated and characterized. Each of these elements with a single layer of binary
surface relief combines functions of a diffractive lens, Gaussian-to-tophat beam shaper, and
Dammann beam splitter. The optical investigations of laser light patterns at 20° fanout angle
reveal up to 88% diffraction efficiency, high contrast, and nearly diffraction limited
resolution. The developed technology has the potential for reducing complexity, number of
optical components, power consumption and costs of structured light projectors in mobile and
stationary 3D sensors.
© 2018 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
OCIS codes: (050.5745) Resonance domain; (150.6910) Three-dimensional sensing; (050.1970) Diffractive optics;
(050.1950) Diffraction gratings; (150.2945) Illumination design; (090.2890) Holographic optical elements.

References and links


1. B. R. Jones, H. Benko, E. Ofek, and A. D. Wilson, “IllumiRoom: Peripheral Projected Illusions for Interactive
Experiences,” in Proc. SIGCHI Conf. Hum. Factors Comput. Syst. - CHI ’13 (2013), pp. 869.
2. T. Dutta, “Evaluation of the Kinec sensor for 3-D kinematic measurement in the workplace,” Appl. Ergon. 43(4),
645–649 (2012).
3. I. Oikonomidis, N. Kyriazis, and A. Argyros, “Efficient model-based 3D tracking of hand articulations using
Kinect,” in Procedings of the British Machine Vision Conference 2011 (2011), p. 101.1–101.11.
4. T. Leyvand, C. Meekhof, Y. C. Wei, J. Sun, and B. Guo, “Kinect identity: Technology and experience,”
Computer 44(4), 94–96 (2011).
5. A. F. Abate, M. Nappi, D. Riccio, and G. Sabatino, “2D and 3D face recognition: A survey,” Pattern Recognit.
Lett. 28(14), 1885–1906 (2007).
6. J. J. Aguilar, F. Torres, and M. A. Lope, “Stereo vision for 3D measurement: accuracy analysis, calibration and
industrial applications,” Measurement 18(4), 193–200 (1996).
7. J. Batlle, E. Mouaddib, and J. Salvi, “Recent progress in coded structured light as a technique to solve the
correspondence problem,” Pattern Recognit. 31(7), 963–982 (1998).
8. J. Salvi, J. Pagâ, and J. Batlle, “Pattern codification strategies in structured light systems,” Pattern Recognit.
37(4), 827–849 (2004).
9. R. L. Morrison, S. L. Walker, and T. J. Cloonan, “Beam array generation and holographic interconnections in a
free-space optical switching network,” Appl. Opt. 32(14), 2512–2518 (1993).
10. V. E. Levashov and A. V. Vinogradov, “Analytical theory of zone plate efficiency,” Phys. Rev. E Stat. Phys.
Plasmas Fluids Relat. Interdiscip. Topics 49(6), 5797–5803 (1994).
11. D. Prongué, H. P. Herzig, R. Dändliker, and M. T. Gale, “Optimized kinoform structures for highly efficient fan-
out elements,” Appl. Opt. 31(26), 5706–5711 (1992).
12. Jenoptics (MEMS Optical), Web site: https://www.jenoptik-
inc.com/download/77/entertainment/2241/diffractive-diffusers-5.
13. Holoeye, Web site: https://holoeye.com/diffractive-optics.
14. Holo/Or, Web site: http://www.holoor.co.il.
15. Sylvain Hallereau (System Plus Consulting), Apple IPhone X - IR Dot Projector (2017).
16. F. Lu, F. G. Sedgwick, V. Karagodsky, C. Chase, and C. J. Chang-Hasnain, “Planar high-numerical-aperture
low-loss focusing reflectors and lenses using subwavelength high contrast gratings,” Opt. Express 18(12),
12606–12614 (2010).
17. Y. Zhou, M. C. Y. Huang, C. Chase, V. Karagodsky, M. Moewe, B. Pesala, F. G. Sedgwick, and C. J. Chang-
Hasnain, “High-index-contrast grating (HCG) and Its applications in optoelectronic devices,” IEEE J. Sel. Top.
Quantum Electron. 15(5), 1485–1499 (2009).
18. A. B. Klemm, D. Stellinga, E. R. Martins, L. Lewis, G. Huyet, L. O’Faolain, and T. F. Krauss, “Experimental
high numerical aperture focusing with high contrast gratings,” Opt. Lett. 38(17), 3410–3413 (2013).
19. M. Khorasaninejad, W. T. Chen, R. C. Devlin, J. Oh, A. Y. Zhu, and F. Capasso, “Metalenses at visible
wavelengths: Diffraction-limited focusing and subwavelength resolution imaging,” Science 352(6290), 1190–
1194 (2016).

#331253 https://doi.org/10.1364/OE.26.021092
Journal © 2018 Received 9 May 2018; revised 18 Jul 2018; accepted 18 Jul 2018; published 1 Aug 2018
Vol. 26, No. 16 | 6 Aug 2018 | OPTICS EXPRESS 21093

20. A. Arbabi, R. M. Briggs, Y. Horie, M. Bagheri, and A. Faraon, “Efficient dielectric metasurface collimating
lenses for mid-infrared quantum cascade lasers,” Opt. Express 23(26), 33310–33317 (2015).
21. F. Aieta, P. Genevet, M. A. Kats, N. Yu, R. Blanchard, Z. Gaburro, and F. Capasso, “Aberration-free ultrathin
flat lenses and axicons at telecom wavelengths based on plasmonic metasurfaces,” Nano Lett. 12(9), 4932–4936
(2012).
22. D. Rosenblatt, A. Sharon, and A. A. Friesem, “Resonant grating waveguide structures,” IEEE J. Quantum
Electron. 33(11), 2038–2059 (1997).
23. O. Barlev, M. A. Golub, A. A. Friesem, and M. Nathan, “Design and experimental investigation of highly
efficient resonance-domain diffraction gratings in the visible spectral region,” Appl. Opt. 51(34), 8074–8080
(2012).
24. M. A. Golub, A. A. Friesem, and L. Eisen, “Bragg properties of efficient surface relief gratings in the resonance
domain,” Opt. Commun. 235(4–6), 261–267 (2004).
25. E. Noponen, A. Vasara, J. Turunen, J. M. Miller, and M. R. Taghizadeh, “Synthetic diffractive optics in the
resonance domain,” J. Opt. Soc. Am. A 9(7), 1206–1213 (1992).
26. M. G. Moharam, T. K. Gaylord, and R. Magnusson, “Criteria for Raman-Nath regime diffraction by phase
gratings,” Opt. Commun. 32(1), 19–23 (1980).
27. D. L. Dickensheets, “Imaging performance of off-axis planar diffractive lenses,” J. Opt. Soc. Am. A 13(9),
1849–1858 (1996).
28. O. Barlev and M. A. Golub, “Resonance domain surface relief diffractive lens for the visible spectral region,”
Appl. Opt. 52(7), 1531–1540 (2013).
29. M. Schmitz, R. Brauer, and O. Bryngdahl, “Gratings in the resonance domain as polarizing beam splitters,” Opt.
Lett. 20(17), 1830 (1995).
30. O. Barlev, M. A. Golub, A. A. Friesem, D. Mahalu, and M. Nathan, “Fabrication of high-aspect-ratio resonance
domain diffraction gratings in fused silica,” Opt. Eng. 51(11), 118002 (2012).
31. M. Quintanilla and A. M. de Frutos, “Holographic filter that transforms a Gaussian into a uniform beam,” Appl.
Opt. 20(5), 879–880 (1981).
32. M. A. Golub, I. N. Sisakyan, and V. A. Soifer, “Infra-red radiation focusators,” Opt. Lasers Eng. 15(5), 297–309
(1991).
33. M. Duparré, M. A. Golub, B. Lüdge, V. S. Pavelyev, V. A. Soifer, G. V. Uspleniev, and S. G. Volotovskii,
“Investigation of computer-generated diffractive beam shapers for flattening of single-modal CO(2) laser
beams,” Appl. Opt. 34(14), 2489–2497 (1995).
34. C. Zhou and L. Liu, “Numerical study of Dammann array illuminators,” Appl. Opt. 34(26), 5961–5969 (1995).
35. B. E. A. Saleh and M. C. Teich, Fundamentals of Photonics, 2nd ed. (Wiley, 2007).
36. O. Barlev and M. A. Golub, “Coherent imaging with a resonance domain diffractive lens in laser light,” Appl.
Opt. 55(18), 4820–4826 (2016).
37. M. A. Golub and I. Grossinger, “Diffractive optical elements for biomedical applications,” in Proc. SPIE 3199,
Biomedical Systems and Technologies II (1998), pp. 220–231.
38. H. Iwaoka and T. Shiozawa, “Aberration-free linear holographic scanner and its application to a diode-laser
printer,” Appl. Opt. 25(1), 123–129 (1986).
39. Y. Ono and N. Nishida, “Holographic disk scanners for bow-free scanning,” Appl. Opt. 22(14), 2132–2136
(1983).
40. O. Barlev and M. A. Golub, “Chromatic dispersion of a high-efficiency resonance domain diffractive lens,”
Appl. Opt. 54(19), 6098–6102 (2015).

1. Introduction
Three-dimensional (3D) sensing with reconstruction of depth data of a 3D scene is a rapidly
expanding field with a variety of methods and growing number of applications. The most
recent and popular applications are in entertainment [1], posture recognition [2], tracking of
hand articulations [3], and facial recognition system [4,5]. Stereovision [6], a common
method for 3D sensing, triangulates the 3D positions of the scene based on recording images
of the scene from two or more points of view and finding correspondences between them.
Today, in the emerging structured light (SL) technique, regular illuminators of a 3D scene are
substituted by specialized structured light projectors that sequentially project a set of tailored
SL light patterns onto the scene. Subsequent recording of respective set of images enables
computational reconstruction of the depth data of the scene. A basic and common sets of SL
patterns are “binary-encoded” patterns which consist of stripes [7,8], wherein the spatial
density of stripes increases by a factor of two at every consecutive pattern, as shown in Fig. 1.
Deviations of the stripes from their regular positions encodes depth data of the scene and
supports its computational reconstruction. This way, the accuracy of acquiring the depth data
by triangulation is improved, and becomes highly dependent on the sharpness, uniformity,
and contrast of the projected SL patterns.
Vol. 26, No. 16 | 6 Aug 2018 | OPTICS EXPRESS 21094

The creation of high contrast SL patterns with high power, frame rate, and large fanout
angle is quite challenging. Dynamic projectors based on spatial light modulators, digital
micro mirror devices, or scanning mirrors may have some disadvantages as SL projectors. For
example, in mobile application they overshoot for the creation of pre-determined patterns
which comes on the expense of cost, complexity and bulkiness. Moreover, they have limited
time frame rate and power limitation, as well as wavelength range restrictions, especially in
IR. The latter may be crucial for outdoor, and mid-range scanning (5-10 meters) applications.
In particular, the liquid-crystal have power threshold and absorb infrared light while the
scanning mirror projectors may have laser safety problems. Relatively small number of pixels
in SLMs and large few-micrometer pixel pitch restricts space-bandwidth product and fidelity
of the SL patterns. In contrary, modern static DOEs enjoy sub-nanometer spatial accuracy,
which enables diffraction limited fidelity of SL patterns. Further to advantages of static
DOEs, the costs of SLMs and electronics far exceed costs of the binary DOEs in mass
production. Static projectors based on computer generated “scalar“ diffractive optical element
(DOE) are relatively simple, compact, and have low cost [9]. However, they have tough limits
in fanout angle, spatial resolution of patterns, and light efficiency. For example, binary zone
plate efficiency is capped at 40% for a relatively small NA of ~0.035 [10]. The on-axis
Kinoform structure with continuous surface relief described at [11] experimentally
demonstrated beam splitting of 1 to 9 with 0.6° fanout with high diffraction efficiency, albeit
with even lower effective NA=0.005. Diffractive diffusers and beam splitters are [12–14]
capable of large fanout angles. However, the diffusers feature a kind of speckle noise and
random spots whereas, the beam splitters are essentially restricted to patterns composed of
points. Further to that, the scalar DOEs, diffractive diffusers and beam splitters all require
additional bulky optical components for laser light collimation and projection of the SL
patterns, as in the recently debuted iPhone X dot projector [15].
The challenge in modern diffractive optics, in our opinion, is in combination of large
fanout angles, high NA, separation of useful diffraction order from ghost ones, high
diffraction efficiency, capability for creation of complex and high-resolution light patterns.
Well-known scalar diffractive optics, based on Fourier optics and iterative Fourier transform
algorithms, fails in matching the challenge, because of physical, computational and
technological limitations. This paper meets the challenge by using off-axis design for non-
paraxial diffractive lens with high NA of ~0.17, Gaussian-to-top-hat shaper and binary beam
splitter. The fine structure of diffractive zones was designed with rigorous diffraction theory
that is a must for large fanout angles and off-axis configuration with high diffraction
efficiency. High diffraction efficiency of the DOEs compared to SLMs enables to use lower
power light sources for achieving same output powers at the SL patterns.
For a sole DOE to shape structured light patterns in laser light, it must combine at least
three functions: imaging of the point source (laser emitter or fiber edge) to the plane where
patterns are expected to be sharp, converting the point source to a certain structure intensity
distribution and finally, multiplexing that structure to compose SL pattern.

Fig. 1. Set of binary-encoded SL patterns for 3D sensing.


Vol. 26, No. 16 | 6 Aug 2018 | OPTICS EXPRESS 21095

Nano-technologies have rapidly advanced in the last decade and are enabling new DOEs
that overcome usual restriction of traditional ones. Such new DOEs include high contrast
dense gratings [16–18], metalenses [19–21], grating waveguide structures [22] and resonance
domain surface relief structures [23–26]. Specifically, resonance domain DOEs have a period
comparable to the wavelength, feature Bragg effects, have high diffraction efficiency [23],
fast focusing with high numerical aperture (NA) [27,28], tens of degrees separation angles
between the diffraction orders, tailored aberrations and polarization control [29]. Their
fabrication as binary structures on fused silica (FS) with well-established techniques of
electron beam lithography (EBL) and reactive ion etching (RIE) is rather straight forward
[30].
In this paper, we present DOEs that comprise all imaging, shaping and multiplexing
functions in a single binary surface relief layer. In our knowledge, it is the first time that a
DOE with complicated wide-angle fanout pattern and high diffraction efficiency is proposed,
theoretically analyzed, designed, fabricated and optically tested with success. The DOEs
project a sequence of binary-encoded light patterns at the visible wavelength. The patterns
reach up to 64 tophat stripes at a distance of one meter with 20° fanout out angle. Sections 2
and 3 are dedicated to derivation of the phase function of the DOE as a mathematical
superposition of an imaging diffractive lens, map-transformation Gaussian-to-top-hat beam
shaper [31–33], and Dammann grating beam splitter [34]. In section 4 we use numerical
calculations with rigorous conical diffraction to estimate the efficiency. The design was
implemented in section 5 as a single layer of transmissive binary surface relief profile.
Finally, section 6 presents the optical experiments. The results show fine resolution, high
contrast and high efficiency of up 88%.
2. Gaussian-to-tophat imaging shaper
In this section we apply geometrical optics for design of the SL DOE with imaging and
shaping functions, which we will call from now as an imaging shaper. The multiplexing by
adding a phase function of a Dammann beam splitter will be considered in the next section.
First, Eqs. (1)-(9) describe a model of a diverging polarized beam that is incident on the DOE.
Then Eqs. (10)-(13) provide a phase function for imaging and shaping. Finally, Eq. (14)
expresses the periods of a local diffraction grating.
The imaging shaper DOE converts an oblique diverging Gaussian light beam into a flat
tophat (stripe) intensity pattern, as shown in Fig. 2. Figure 3 shows typical diffraction grooves
of the imaging shaper, which can be considered as a continuous set of local diffraction
gratings. In order to define the local diffraction gratings, we used geometrical optics to
calculate the required smooth phase function.

Fig. 2. (a) Intensity distribution of the oblique incident Gaussian beam at the DOE plane. The
DOE boundary corresponds to the intensity level exp(−8) ; (b) Tophat intensity at the output
plane.
Vol. 26, No. 16 | 6 Aug 2018 | OPTICS EXPRESS 21096

Fig. 3. Imaging shaper as a set of adjacent local diffraction gratings.

Fig. 4. Off-axis layout for the imaging shaper DOE design. (a) View from top; (b) side view.

For the imaging shaper design in the resonance domain of diffraction, we used the off-axis
optical scheme shown in Fig. 4. Figure 4(a) shows view from a top of an optical table, (b)
shows a side view. We defined a Cartesian coordinate system x, y with origin O at the DOE
plane, axis z serving as normal to the DOE and respective unit vectors xˆ , yˆ , zˆ . For high
diffraction efficiency, the chief ray of the incident beam subtends the Bragg angle of
incidence
λ
sin θ B = , (1)
2Λ 0

to the DOE normal, where Λ 0 is the base period of the diffraction grooves at the center of the
imaging shaper and λ is the wavelength of the incident beam.
We assumed that the beam waist of the incident beam (e.g. cleaved end of a single mode
fiber coupled to a laser) has the mode field semi-diameter σ 0 (where the intensity level is
e −2 of that at the origin), and is placed in front of DOE at an oblique distance l0 along the
chief ray with angle θ B , as shown in Fig. 4. The unit polarization vector e0 at the beam waist
is defined with its orthogonal p and s polarization components e0 p , e0 s ( e02 p + e02s = 1 ) wherein
s polarization corresponds to the axis y in Fig. 4. The propagation direction vector of the
incident beam at the DOE point x, y can be obviously expressed as

( x + sin θ B l0 ) xˆ + y yˆ + cos θ B l0 zˆ
N inc ( x, y ) = , (2)
R ( x, y, l0 )

where we defined a function

R ( x, y, l ) = l 2 + x 2 + y 2 + 2sin θ B l x , (3)
Vol. 26, No. 16 | 6 Aug 2018 | OPTICS EXPRESS 21097

such that R ( x, y, l0 ) provides distance from the center of the incident beam waist to a point
x, y at the DOE plane. To propagate the oblique incident beam towards the DOE plane, we
used equations detailed in [35] for a vectorial Gaussian beam at the far-field zone, that should
be applied to Cartesian coordinates x′, y ′, z ′ where z ′ is aligned with the chief ray of the
oblique incident beam and defined with same origin O at the DOE as coordinates x, y . Then,
with the coordinate transformation
x′ = cos θ B x, y ′ = y, z ′ = sin θ B x (4)

we obtained expressions for intensity I inc ( x, y ) , semi-diameter σ ( x, y, l0 ) , phase


ϕinc ( x, y ) and unit polarization vector e ( x, y ) of the incident beam at the DOE plane as

 cos 2 θ B x 2 + y 2  λ
I inc ( x, y ) = I 0 exp  −2  , σ ( x, y, l0 ) = R ( x, y, l0 ) , (5)
 σ ( x, y, l0 ) 
2
πσ 0


ϕinc ( x, y ) = k  R ( x, y, l0 ) − l0  , k = , (6)
λ

e ( x, y ) =
( cos θ l e
B 0 0p
− sin θ B ye0 s ) xˆ + ( sin θ B x + l0 ) e0 s yˆ − ( x + sin θ B l0 ) e0 p + cos θ B ye0 s  zˆ
. (7)
( x, y, l ) − ( − ye + cos θ B xe0 s )
2 2
R 0 0p

Based on Eq. (5) we chose the boundary of the DOE as a contour line
cos 2 θ B x 2 + y 2 = 4σ 2 ( x, y, l0 ) where the Gaussian beam intensity level is exp(−8) relative to
its peak intensity. This contour line with an implicit equation
2 2
 x − x0   y 
 A  + B  =1 (8)
   
describes an ellipse with major and minor half-axes A and B and a lateral x-shift x0 that are
defined by

2λ l0 cos θ B σ 0 π − 4λ 2λ l0 cos θ B 4λ sin θ B l0


2 2 2 2

A= , B= , x0 = . (9)
cos θ Bσ π − 4λ
2 2
0
2 2
cos 2 θ Bσ 02π 2 B − 4λ 2 cos θ Bσ 02π 2 − 4λ 2
2

To implement functionality of imaging from the off-axis point source located at distance
l0 in front of the DOE to a distance l in our actual optical arrangement of Fig. 4, we have
added the off-axis lens [36] with off-axis angles θ B . Such lens images the center of the beam
waist from l0 in front of the DOE to l after the DOE, i.e. converts phase ϕinc ( x, y ) in Eq. (3)
of diverging incident beam to the phase −k  R ( x, y, l ) − l  of a converging output beam, and
accordingly has the phase function

ϕlens ( x, y ) = −k  R ( x, y, l ) − l + R ( x, y, l0 ) − l0  , (10)

where the function R ( x, y, l ) was defined in Eq. (3).


To proceed with shaping of the tophat, we used one-dimensional on-axis paraxial solution
for the far-field tophat shaper detailed in [37], with the phase function
Vol. 26, No. 16 | 6 Aug 2018 | OPTICS EXPRESS 21098

ka   2x  σ   2x2   
ϕ ( x; a, σ , C ) =  x erf   − 1 − exp  − 2   (11)
l erf ( 2C σ  )  σ  2π   σ   

that converts the Gaussian intensity with parameter σ and aperture size 2C to flat tophat
intensity distribution with size 2a at distance l . Respective two-dimensional (2D) paraxial
solution for the on-axis tophat shaper is separable in coordinates x, y . Further to that, the
angular tilt θ B of the beam shaper to the incident beam in our off-axis optical scheme leads to
modified equations wherein coordinates x, y exchange to x cos θ B , y per Eq. (4).
Accordingly, the phase function of the 2D far-field tophat shaper that converts the Gaussian
intensity with parameter σ DOE and aperture size 2 A cos θ B × 2 B to flat top-hat intensity
distribution with size 2a × 2b at distance l can be expressed as
ϕth , Far ( x, y ) = ϕ ( x cos θ B ; a, σ DOE , A cos θ B ) + ϕ ( y; b, σ DOE , B ) , σ DOE = σ ( 0, 0, l0 ) . (12)

Then, the phase function of the imaging shaper ϕth is a superposition of the off-axis lens and
the 2D far-field beam shaper
ϕth ( x, y ) = ϕth , Far ( x, y ) + ϕlens ( x, y ) . (13)

The spatial frequency vector ν ( x, y ) and period Λ ( x, y ) of the local diffraction grating of the
imaging shaper vary with coordinates x, y and can be calculated through the 2D gradient of
the smooth phase function of Eq. (13)
1 1  ∂ ∂ 
ν ( x, y ) = ∇ ⊥ϕth ( x, y ) , Λ ( x, y ) = , ∇⊥ =  ,  , (14)
2π ν ( x, y )  ∂x ∂y 

such that angular orientation of the normal to the fringes of the local diffraction gratings is
ν ( x, y ) Λ ( x, y ) , as shown in Fig. 3. The base period in these notations is
simply Λ 0 = Λ ( 0, 0 ) .

3. From single tophat to multi-stripe patterns


Off-axis imaging and shaping of the previous section provided single tophat stripe of the SL
pattern. This section deals with multiplexing to convert the single tophat stripe to entire SL
pattern with number of stripes of nearly equal power. In a set of subsequent mathematical
equations, Eqs. (15) and (16) describe a model of the SL DOE, which generates entire pattern
that comprises of N tophat stripes. For that, Eqs. (17), (18) and Table 1 define binary phase
of Dammann grating in a single grating period. Finally, Eq. (19) explains how to build the
diffraction grooves of the SL DOE.
The transition from the imaging shaper to the SL DOE that reconstructs N tophat stripes
(N )
is done by adding a phase function ϕ spl ( x ) of the N-th order binary Dammann grating [34] to
the phase ϕth ( x, y ) of the imaging shaper per Eq. (13). Accordingly, the phase function of
the SL DOE is

ϕ sl( N ) ( x, y ) = ϕth ( x, y ) + ϕ spl


(N )
( x ) = ϕth, Far ( x, y ) + ϕlens ( x, y ) + ϕ spl( N ) ( x ) . (15)

The phase function of the Dammann grating splits the light power into N equal intensity
diffraction orders and yields multiplexed system of the tophat stripes instead of originally
single one.
Vol. 26, No. 16 | 6 Aug 2018 | OPTICS EXPRESS 21099

The phase of the even-order Dammann grating beam splitter [34], in our off-axis
arrangement, is a periodic function of the coordinate x cos θ B with period Λ (splN ) defined as

λl
Λ (splN ) = (16)
0.5 Δ (
N)

for an even number of top-hats, where Δ ( N ) = b N is a distance between centers of adjacent


tophat stripes in the image plane. The phase of the Dammann beam splitter can be calculated
through a function ϕ1( N ) ( χ ) of a dimensionless variable 0 ≤ χ ≤ 1 reduced to a single period
with length 1.0

(N)
ϕ spl ( x ) = ϕ1 ( χ ) , χ =
1
Λ (splN )
( )
mod x cos θ B , Λ (splN ) , (17)

(
where mod ⋅ , Λ (spl )
N
) is a minimal non-negative residue of the first argument after subtraction
of multiples of Λ (splN ) . In case of even number of tophats, ϕ1 ( χ ) is calculated as

 ϕ0.5 ( χ ) , 0 ≤ χ ≤ 0.5
ϕ1 ( χ ) =  . (18)
 0.5 (
ϕ χ − 0.5 ) , 0.5 < χ ≤ 1
+ π

The piecewise constant function ϕ0.5 ( χ ) with alternating values 0 and π is defined on half of
the interval 0 ≤ χ ≤ 0.5 . The normalized values of N coordinates where the phase ϕ1 ( χ ) or
ϕ0.5 ( χ ) changes from 0 to π are summarized in Table 1 based on [34].
For design of diffraction grooves of the SL DOE, we first used Eq. (13) for the smooth
phase function of the tophat beam shaper and solved numerically the isoline equation
ϕth ( x, y ) = 2π ⋅ n, (19)

where n is number of the diffraction groove in the imaging shaper. The width of each
diffraction groove was chosen with duty cycle (DC) 0.5, i.e. as half of the local period per Eq.
(16). Important to mention, that an optimal design for maximal efficiency should use a
varying, rather than a fixed, DC and slant angle of diffractive grooves. However, these may
cause substantial technological complications in DOE fabrication and is outside the scope of
this paper. The addition of the Dammann beam splitter phase ϕ spl(N )
( x ) with alternating values
of 0, π was implemented by inverting the depth of binary diffraction grooves where the
Dammann splitter value was π which is equal to shifting the diffraction grooves by half of the
local period.
Table 2 details the set of parameters 2a, 2b, and Λ (splN ) for a set of the eight SL DOEs that
generates SL pattern set with up to 64 stripes for the fanout angle of 20°. DOE#1 with N = 0
corresponds to a fully uniformly illuminated SL pattern within entire fanout angle. The
DOE#2 with N = 1 corresponds to SL pattern with one stripe corresponding to half of the
fanout angle. Both DOE#1 and 2 provide single tophat stripes and accordingly do not need a
Dammann beam splitter phase.
Vol. 26, No. 16 | 6 Aug 2018 | OPTICS EXPRESS 21100

Table 1. Coordinates of alternating values 0 and π in Dammann gratings, following [34].

DOE# N Normalized coordinates for 0 to π phase changes


3 2 0.5.
4 4 0.22, 0.44, 0.5.
5 8 0.061, 0.176, 0.208, 0.317, 0.5.
6 16 0.140, 0.175, 0.221, 0.267, 0.356, 0.395, 0.439, 0.452, 0.5.
0.055, 0.089, 0.110, 0.133, 0.173, 0.195, 0.210, 0.230, 0.248, 0.330, 0.348,
7 32 0.401, 0.433, 0.441, 0.465, 0.484, 0.5.
0.010, 0.017, 0.029, 0.040, 0.047, 0.06, 0.07, 0.08, 0.09, 0.10, 0.12, 0.14,
0.212, 0.246, 0.262, 0.28, 0.288, 0.296, 0.317, 0.324, 0.337, 0.361, 0.382,
8 64
0.396, 0.412, 0.423, 0.433, 0.438, 0.448, 0.45944, 0.466, 0.474, 0.483, 0.489
0.5.

Table 2. DOEs parameters for binary-encoded stripe patterns

DOE# N
Λ (splN ) , Δ ( N ) , mm 2a, mm 2b, mm
µm
1 0 ∞ - 352.7
2 1 ∞ - 176.3
3 2 3.63 176.4 88.2
4 4 7.25 88.2 44.1
5 8 14.50 44.0 22.0 352.7
6 16 29.01 22.0 11.0
7 32 58.01 11.0 5.5
8 64 116.02 5.6 2.8

4. Efficiency in rigorous diffraction

In order to calculate the diffraction efficiency η sl( N ) of the SL DOE, we calculated separately
(N )
the efficiency of the imaging shaper ηth( ) and the Dammann beam splitter η spl
N
and
multiplied them

η sl( N ) = ηth( N ) η spl


(N )
. (20)

For the calculation of the diffraction efficiency ηth( N ) of the imaging shaper we approximated
the imaging shaper DOE as a set of adjacent local resonance domain diffraction gratings with
same groove depth h , and DC but slowly varying period, as shown in Fig. 3. Based on the
propagation vector N inc ( x, y ) per Eq. (2) and polarization vector e ( x, y ) per Eq. (7) of the
incident beam and the periods Λ ( x, y ) and angular orientation of the groove fringes
ν ( x, y ) ⋅ Λ ( x, y ) per Eq. (14), we calculated parameters of conical diffraction at distinct
points x, y at the DOE. The parameters are as follows: the angle of incidence 0 ≤ θinc ≤ π 2
referenced to the normal z to the DOE, the azimuthal angle 0 ≤ φincν ≤ 2π in the DOE plane
referenced to the spatial frequency ν ( x, y ) of the local diffraction grating and the
polarization angle β inc of the incident wave, which is defined in the plane that is
perpendicular to N inc and referenced to the plane of incidence. With such input, we used the
rigorous coupled-wave analysis (RCWA) method and Rsoft DiffractMOD software to
calculate the local diffraction efficiencies η1 ( x, y ) of the local gratings. Finally, the
diffraction efficiency of the imaging shaper was calculated as
Vol. 26, No. 16 | 6 Aug 2018 | OPTICS EXPRESS 21101

ηth( N ) =
η ( x, y ) I ( x, y) dxdy .
1 inc
(21)
 I ( x, y)dxdy
inc

An important parameter for optimal diffraction efficiency is the DOE base period
Λ 0 = Λ ( 0, 0 ) = λ ( 2sin θ B ) at the DOE origin, which support off-axis arrangement of Fig.
4(a) with Bragg angle of incidence per Eq. (1). The base period should be chosen to get
diffraction efficiency highest available at θ B and remaining such at slightly changed periods
Λ and angles θ inc that occur at the DOE periphery. We consider here transverse electric (TE)
polarized light at the beam waist of the incident beam, while generalization to the transverse
magnetic (TM) polarized light can be done in a straightforward way. Figure 5 shows results
for the first order diffraction efficiency of a resonance domain diffraction grating as a
function of θ inc , Λ for λ = 0.6328µm, groove depth h = 1.082µm, and DC 0.5, for TE
polarization in the model of the classical rigorous diffraction, as applicable for the local
gratings of our imaging shaper at y = 0 and arbitrary x. Also plotted in Fig. 5 are curves for
angle of incidence θ inc as a function of the local grating period Λ ( x, 0 ) of DOE #1 with
l0 = 7.6mm , l = 1m , for given distinct base periods Λ ( 0, 0 ) of 0.51µm, 0.61µm and
0.71µm. Results of Fig. 5 illustrates that the plot with optimal base period Λ ( 0, 0 ) = 0.61μm
is confined to the region with the highest diffraction efficiency.

Fig. 5. TE first order diffraction efficiency of resonance domain diffraction gratings as a


function of θ inc , Λ, for λ = 0.6328µm, h = 1.082 µm, DC = 0.5 ; Plotted curves show θ inc as a
function Λ ( x, 0 ) for DOE#1 with l0 = 7.6mm , l = 1m , σ 0 = 2.06μm for basic period Λ 0
of 0.51(dotted line), 0.61 (dashed line), and 0.71µm (dot dash).

Figure 6 shows first order local diffraction efficiencies η1 ( x, y ) of top-hat beam shapers
#1,2,3,8 per Table 2, with the parameters Λ 0 = 0.61μm , l0 = 7.6mm ,
l = 1m , h = 1.082μm , σ 0 = 2.06μm and λ = 0.6328μm , for the case of TE polarization
( e0 p = 0, e0 s = 1 ) at the beam waist of the incident beam.
Vol. 26, No. 16 | 6 Aug 2018 | OPTICS EXPRESS 21102

Fig. 6. (a)-(d) Local diffraction efficiency η1 of imaging shapers of SL DOEs #1,2,3,8 with
parameters; Λ 0 = 0.61μm , l0 = 7.6mm , l = 1m , DC = 0.5 , h = 1.083μm , λ = 0.6328μm ,
l = 1m , σ 0 = 2.06μm , and TE polarization.

(N )
The diffraction efficiency η spl of the Dammann beam splitter is defined as the ratio
between powers of the N useful diffraction orders to the total power of all diffraction orders.
(N )
Our numerical calculations of η spl are specified in Table 3 and match those published in [34].
Table 3 summarizes our theoretically estimated diffraction efficiencies for the set of the
SL DOEs per Eqs. (20) and (21).
Table 3. Parameters and theoretical performance data for SL DOEs.
(N )
DOE#. N η spl ηth( N ) η sl( N )
1 0 1.00 0.91 0.91
2 1 1.00 0.95 0.95
3 2 0.81 0.78
4 4 0.70 0.67
5 8 0.76 0.73
6 16 0.81 0.96 0.78
7 32 0.83 0.80
8 64 0.80 0.77

5. Fabrication
Eight DOEs #1 to #8 were fabricated with the following parameters; λ = 0.6328μm ,
θ B = 31o σ 0 = 2.06μm , l0 = 7.6mm , Λ 0 = 0.61μm , h = 1.082μm , DC = 0.5 , 2 A = 3.6mm ,
2 B = 3mm , x0 = 217μm , l = 1m , b = 176mm , a (0) = b , a( N ) = b / ( 2N ) ,
N = 1, 2, 4,8,16,32, 64 in accordance to parameters in Table 2. The ratios b / l and a (0) l
correspond to ±100 fanout angle. Based on the DOEs parameters and our in-house software
we created GDSII files that controlled direct e-beam writing of the spatial pattern in the EBL
Vol. 26, No. 16 | 6 Aug 2018 | OPTICS EXPRESS 21103

process. The fabrication process consisted of the following steps: (a) direct e-beam writing of
the spatial pattern in e-beam resist layer (ZEP 520A) with EBL machine Raith EBPG5200,
(b) transferring the recorded pattern to a fine Chrome mask by with ion beam milling with
ATC-2020-IM machine from AJA, and (c) transferring the spatial pattern from the Chrome
mask to the 0.5mm thick FS substrate by CHF3 plasma RIE with Nextral NE 860 machine
from Unaxis. The etch-depth in FS was controlled by duration of the RIE process applied to
calibration samples in advance. The etch-depth was measured with environmental scanning
electron microscopy (ESEM) at the cross section after mechanical cutting. Figure 7 shows
surface relief profile of the SL DOE#8 with N = 64 stripes, wherein (a) shows plot of
( 64 )
theoretical phase ϕ spl of the Dammann grating over half of a period Λ (spl64) / cos θ B , (b) shows
top-view ESEM image, wherein inversions of the depth of the diffraction grooves are in
match to the theoretical phase of the Dammann grating per (a), (c) provides zoom-in of a
region of image (b), and (d) shows mechanically cut cross section of a sample SL DOE.

Fig. 7. ESEM images of SL DOE#8 that projects N = 64 stripes. (a) Plot of theoretical phase
ϕ spl
( 64 )
of the Dammann grating over half of a period Λ (spl64 ) / cos θ B ; (b) Top-view image,
wherein depth inversion of the diffraction grooves is linked to theoretical phase per (a); (c)
zoom-in of a region of image (b); (d) mechanically cut cross section of a sample SL DOE.

6. Experiments
This section is dedicated to experimental investigation of the performance of SL DOEs. In
particular, we provide the main performance data: the patterns contrast, uniformity, and
sharpness. For the optical characterization of SL DOEs we used a pigtailed laser diode with
wavelength of 642nm (Thorlabs LP642-SF20) coupled to a polarization maintaining single
mode fiber (SMF) as light source, a power sensor (Thorlabs S120C), a focusing lens with
focal length 30mm and a square aperture with dimensions 25.4mm x 25.4mm.
Vol. 26, No. 16 | 6 Aug 2018 | OPTICS EXPRESS 21104

6.1 Diffraction efficiency


The optical scheme for measurement of diffraction efficiency of SL DOEs is shown in Fig. 8.
The DOE was mounted on a rotating stage, and the fiber polarization orientation was set in
the direction of y-axis that is perpendicular to the optical Table 1. TE polarization. The square
aperture was centered around the chief ray of the first diffraction order at a distance about
72mm from the DOE in order to block the ghost orders of the Dammann beam splitter. The
focusing lens was placed adjacent to the square aperture to collect the light power P1 onto the
power sensor. As a reference, we measured the light power emitted from the SMF PSMF , with
the same lens and the power sensor placed right opposite the unobscured SMF incident beam.
The diffraction efficiency of the SL DOE was calculated as
P1
η sl( N ) (θ B ) = , (22)
η Fresnel PSMF

where Fresnel transmission coefficient η Fresnel was measured according to the procedure
detailed in [28] to compensate for the TE Fresnel losses incurred at the smooth surface of the
FS DOE substrate.
Table 4 summarizes the results of the diffraction efficiency measurements. Some
difference from the theoretical efficiencies can be explained by fabrication defects, which
may be present in some parts of the DOE aperture.
Table 4. Theoretical and experimental diffraction efficiencies

DOE#. N
η sl( N )
Theoretical Experimental
1 1 0.91 0.88
2 1 0.95 0.81
3 2 0.78 0.68
4 4 0.67 0.67
5 8 0.73 0.72
6 16 0.78 0.76
7 32 0.80 0.72
8 64 0.77 0.73

6.2 Pattern quality


Crucial parameters for determining the accuracy of 3D sensing using structured light patterns,
are the contrast, uniformity, and sharpness of the projected patterns. In order to evaluate the
SL DOEs patterns we used the optical scheme of Fig. 4 wherein a white screen was placed
perpendicular to the chief ray of the diffracted beam at a distance l = 1m, in accordance to the
design of previous section. Actual optical configuration is shown in Fig. 9. The SL pattern
that was diffracted from the white screen was captured with a metal–oxide–semiconductor
(CMOS) camera (Cannon EOS 30D) which was mounted on the same optical axis as the
diffracted chief ray. Another CMOS sensor (Aptina MT9J003) was placed directly in the
image plane, for accurate measurements of local intensity distribution in the pattern.
Vol. 26, No. 16 | 6 Aug 2018 | OPTICS EXPRESS 21105

Fig. 8. The optical scheme for measurement of the diffraction efficiency of the SL DOEs.

Fig. 9. Photo of the optical setup for experimental evaluation of the SL patterns.

Figures 10 and 11 show the experimental structured light patterns that were reconstructed
from DOE#1-8 with different numbers of the tophat stripes and common dimensions 352.7
mm x 352.7mm for the fanout angle of 20° at distance l = 1m. Some geometrical distortions
that make the boundaries of the tophat stripes curved rather than straight lines can be
explained by the use of the paraxial tophat shaper phase function ϕth , Far in off-axis layout and
can be corrected as was demonstrated in [38,39]. Few local defects in the patterns, most
notable in Figs. 10(a) and 10(b) stem from some stitching errors and nonuniformity in EBL
fabrication process of the DOEs.
For the SL pattern of Fig. 11 reconstructed from DOEs #7 and 8, the tophat widths 2a
were small enough (5.5mm and 2.8 mm) to be directly captured locally with the CMOS
sensor whose active sensor area is 6.4mm X 3.6mm, as shown in fragment images of the
patterns in Figs. 11(c) and 11(g). Plots for intensity in the local cross sections of the pattern in
Fig. 11(e), 11(f), 11(i), 11(j) and 11(k) show that lowest intensity is virtually zero and the
noise level is quite low. Therefore, the resonance domain DOEs provide high 96% contrast in
structured light patterns, even for dense 64 stripes. The widths of the rise zones at the
boundaries at the tophat stripes, at levels 0.1 to 0.9 in normalized intensity, were calculated
from the plots and summarized in Table 5. The difference between the theoretical and
experimental widths of the rise zones can be attributed the DOEs chromatic dispersion, whose
impact is estimated as 1.8 mm lateral shift of the chief ray for 1nm spectral width of the laser
diode [40].
Vol. 26, No. 16 | 6 Aug 2018 | OPTICS EXPRESS 21106

7. Conclusion
We designed, fabricated, and characterized a set of eight transmissive resonance-domain
surface relief DOEs for binary-encoded structured light patterns. The patterns with 20° fanout
angles consist up to 64 stripes with angular pitch of 0.32°. Each of the DOEs demonstrated
combined optical functions of imaging, Gaussian-to-tophat beam shaping, and beam splitting,
all implemented in a single layer of surface relief with binary profile, with experimental
diffraction efficiency of up to 88%. The DOEs are implemented in fused silica material that is
suitable for high power lasers. Combination of large fanout angles, separation of useful
diffraction order from ghost ones, high diffraction efficiency, high resolution and contrast
reveals a breakthrough of the described DOEs compared to wide-spread scalar diffractive
optics.
The average efficiency presented in this paper of the eight DOEs is 74%. Several
measures could further improve the efficiency, such as advancing to a continuous phase for
beam splitting design, varying DC and slant angle of the grooves in binary design. However,
varying DC would require more computational resources as numerical results scan an optimal
DC for each local grating and, more importantly, substantial complicate the fabrication
process. The fabrication of varying slant would be possible after substantial modification of
RIE fabrication. Higher fanout angles are possible as well, but, on the expense of efficiency.
Some geometrical distortions from straight lines that are seen in the stripes of the SL patterns
could be optimized and tailored to the off-axis layout, in further research efforts.
Compactness of the SL projectors may be crucial for placing 3D sensor in mobile
platforms, including smartphones with 3D camera for face recognition and virtual reality
applications. High power could enable 3D sensors to perform outdoor and increase their
working distance. The developed technology could also find applications in such fields as
laser micromachining, beam shaping, optical trapping of atoms and particles, pattern
recognition, and optical computing.

Fig. 10. Experimental structured light patterns with dimensions 352 mm x 352 mm shot from a
white screen at distance l = 1m, from DOEs with following numbers N of tophat stripes: (a) #1,
N = 0; (b) #2, N = 1; (c) #3, N = 2; (d) #4, N = 4; (e) #5, N = 8; (f) #6, N = 16;
Vol. 26, No. 16 | 6 Aug 2018 | OPTICS EXPRESS 21107

Fig. 11. Experimental structured light patterns reconstructed from DOE#7 and 8 with 32 and
64 tophat stripes, at distance l = 1m with dimensions 352 mm x 352 mm. (a),(b) are photos that
were shot from a white screen; (c),(g) local images of the patterns direct sensed by the CMOS
camera, and (d),(h) their 1-D cross section plots; (e),(f),(i)-(k) magnified parts of (d),(h). The
dashed lines show theoretical results whereas the thick line is 200 points moving average.
Table 5. The contrast and rise length for the SL patterns reconstructed from DOEs #7, #8
DOE Experimental Rise length (10% to 90%), mm
N
No. Contrast Theoretical Experimental
7 32 0.527 1.85
0.98
8 64 0.475 0.936

Funding
Kamin program of the Office of the Chief Scientist, Ministry of Economy.

Anda mungkin juga menyukai