Research Article Vol. 56, No.

15 / May 20 2017 / Applied Optics 4561

Realization of uniform and collimated light

distribution in a single freeform-Fresnel
double surface LED lens
XIONG HUI,* JIE LIU, YUNJIA WAN, AND HAOBO LIN
School of Physics and Electronic Technology, Hubei University, Wuhan 430062, Hubei Province, China
*Corresponding author: xionghui@fudan.edu.cn

Received 24 February 2017; revised 28 April 2017; accepted 4 May 2017; posted 4 May 2017 (Doc. ID 287493); published 19 May 2017

In this work, a design method of a freeform-Fresnel composite surface lens is proposed, which can be applied on
LED lighting optics to realize uniform and collimated light output in a single lens element. The lower lens surface
is of freeform type, which transforms the light distribution from the LED sources to a uniform one, and the upper
lens surface is of the Fresnel type, which further transforms the lights to a collimated beam. Through the
optimization of lens parameters, the obtained lens has the advantages of compact volume, wide acceptance angle,
high light extraction efficiency, and high design accuracy. © 2017 Optical Society of America
OCIS codes: (080.4225) Nonspherical lens design; (220.2740) Geometric optical design; (220.3630) Lenses; (220.4298) Nonimaging
optics.

https://doi.org/10.1364/AO.56.004561

1. INTRODUCTION 2. MATHEMATICAL MODEL

The realization of uniform and collimated light output in LED A. Design Procedures
lighting design is a challenging topic [1]. Traditional lens struc- Figure 1 shows the light path passing through the lens and
ture, such as the freeform lens, can only realize either uniform or the geometric parameters. To realize a uniform light distribu-
collimated light output [2–4], while a uniform as well as colli- tion through the lower surface of the lens, the curvature of the
mated light distribution requires multiple lenses and reflectors surface can be calculated from the differential equation:
[5] or a highly curved lens surface [6]. The imperfect optics in-
clude a limited numerical aperture, low radius-height ratio, and d y1 n cos φ1 − n2 cos φ2

tan β
1
low light extraction efficiency. Recently, a single lens based LED d x1 n2 sin φ2 − n1 sin φ1
collimator has been proposed by delicate combination of free- n1 cos φ1 − cos φ2

;
form surfaces [2,7]. The collimation performance is promising, sin φ2 − n1 sin φ1
while the illumination uniformity can be further improved. x1
Addressed at these problems, this work proposes a freeform- cos φ1
pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi;
x 1  y 21
2
Fresnel double surface lens design method. The lower lens r − x1
surface is of freeform type [8,9], which transforms the light cos φ2
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi; (1)
distribution from the LED sources to a uniform one, and r − x 1 2  H − y1 2
the upper lens surface is of the Fresnel type [10], which further
where n1 and n2 are the refractive index of lens material and air,
transforms the lights to a collimated beam. Moreover, the
respectively, and (x 1 , y 1 ) are the coordinates of points on the
method is given in a general way that can be applied to
lower surface. Assuming the LED source is of the general
LED sources of different Lambert index. The obtained
Lambert type, of which the angular dependence of luminous
composite lens has a large acceptance angle and large radius-
intensity takes the form:
height ratio, thus holding the advantages of high light extrac-
tion efficiency and a compact volume. The results are useful for I θ
I 0 cosm θ; (2)
designing long distance and high intensity LED lighting mod-
ules. Based on a numerical solution, three-dimensional (3D) in which m is the Lambert index. From Eq. (2), the cumulative
modeling of the lens is carried out for light path simulation, distribution function of light intensity within the emission
and the result verifies the validity and high accuracy of the angle θ0 can be calculated by the integral of I θ in solid
method. angle:

1559-128X/17/154561-05 Journal © 2017 Optical Society of America

 4562 Vol. 56, No. 15 / May 20 2017 / Applied Optics Research Article

multiple-tailored optics [11]. The following are the design

considerations:
First, to enhance the light extraction efficiency of the lens,
the radius of the lower surface R 1 should be as large as possible
to increase the maximum acceptance angle (denoted by αc ) col-
lected from the light source. However, the maximum of R 1 is
limited by the critical condition, which dictates the longest
distance light can reach on the lower surface:
y1 dy

1: (5)
x1 d x1
Second, to enhance the luminous intensity of the light spot,
Fig. 1. Geometric parameters of light path passing through the a small R 2 is expected. However, a small R 2 demands a large
composite lens. curvature of the lower surface, hence resulting in a small R 1
inferred from Eq. (5), which is adverse to a high light extraction
efficiency. For energy saving purpose, priority should be given
Z Z to the light extraction efficiency, so the value of R 2 is set based
I c θ0 
I 0 cosm θ sin θdθdϕ on the values of R 1 and h1 , which are tuned to obtain a large αc .
It is calculated to be:
2π y

1 − cosm1 θ0 : αc
arctan 1 jx 1
R 1 :
m1 x1
For ϕ2 in Eq. (1), r is the radius of light spot corresponding Third, regarding the lens thickness, a small h is favorable for
to θ0 , which can be determined from the equation: practical usage (h ≈ h2 − h1 ). However, if h is too small, it also
sffiffiffiffiffiffiffiffiffiffiffiffi sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
increases the incident angle of light onto the upper surface [α in
I c θ0  1 − cosm1 θ0
r
R2
R : (3) Eq. (4)], such that the curvature of the Fresnel ring becomes
I c θc  1 − cosm1 θc 2 large. It instead increases the lens thickness, and the light dis-
Here, θc is the maximal emission angle receivable by the tribution on the upper surface may deviate from a uniform one.
lens, and R 2 is the maximal radius of the light beam incident Fourth, for the number of Fresnel rings (denoted by S), it is
on the upper surface. By solving Eq. (1) numerically, the known that a large S helps reduce the height of each ring, hence
coordinates of the lower surface (x 1 , y 1 ) are obtained. being beneficial to decreasing the lens thickness. However, a
Regarding the upper surface, however, it requires the surface large S is required for precise processing, and may increase the
coordinates should be close in height, to keep the uniform dis- cost of lens fabrication. So a compromise needs to be made on S.
tribution of light incident on the surface. Therefore, we choose In our design, intermediate values are chosen for R 1, R 2 , h1 ,
the Fresnel lens structure. By devising the curve of each Fresnel and h2 , and αc is determined self-consistently. For m
1, the
ring, light can be refracted to a collimated light beam. With the parameters used in the simulation and the output performance
increase of rings number, the height of the rings can be reduced, index are listed in Table 1 according to defined terminol-
so as to further reduce the deviation of light distribution from a ogy [12].
uniform one.
For a certain point (x 2 , y 2 ) on the upper surface, the 3. RESULTS AND DISCUSSION
corresponding (x 1 , y 1 ) on the lower surface can be found from A. Design Illustration and Lens Performance
solving the equations: In simulation, the Runge–Kutta method to the 4th order is
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi applied for a high-precision numerical solution [13]. The
1 − cosm1 θ0 x1
x 2
R2 ; where sin θ0
pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi:
1 − cosm1 θc x 1  y 21
2
Table 1. Parameters Used in Simulation and Output
The line connecting (x 1 , y 1 ) and (x 2 , y 2 ) gives the light path Performance Index of the Lens
in the lens. On the other hand, a collimated beam for light Parameter Physical Meaning Value
incident at angle α with respect to the horizontal line demands
that: h1 Distance of the lower lens surface 30 mm
h2 Distance of the upper lens surface 100 mm
d y2 n1 cos α n cos α R1 Height of the lower lens surface 68.5 mm

tan γ

1 : (4) R2 Height of the upper lens surface 90 mm
d x2 n2 − n1 sin α 1 − n1 sin α
n1 Refractive index of lens material 1.35
It gives the local curvature of the upper surface. Therefore, S Number of Fresnel rings 40
the coordinates of the upper surface (x 2 , y2 ) can be obtained. Performance Physical Meaning Value
B. Design Considerations αc Maximum acceptance angle of the lens 61.5°
In fact, the parameters of the lens are independent of each ξ Light extraction efficiency 77.3%
other. An optimal light output requires a delicate balance h2 − h1 ∕R 2 Thickness/height ratio of the lens 1.29
Ic Luminous intensity in light spot center 29.5 lux
among these parameters, as a general requirement in
 Research Article Vol. 56, No. 15 / May 20 2017 / Applied Optics 4563

Fig. 2. (a) Simulated curve of the upper (blue line) and lower (black
line) lens surface. (b) Illustration of three-dimensional model of the
lens.

calculation accuracy is Oδ5 , where δ is the differential step.

The obtained lens curve is shown in Fig. 2(a). The width of
each Fresnel ring is 2.25 mm, and the smallest and average
height are 0.47 and 4.39 mm, respectively. From the coordi-
nates of the double surfaces, a 3D lens model can be estab-
lished, and the result is shown in Fig. 2(b). The radius/
height ratio is 1.29, giving the lens a compact volume. In prac-
tice, polythene and PMMA materials can be used to fabricate Fig. 3. (a) Simulated light path passing the lens. (b) Light spot on
the lens, due to the advantages of high light transmission, easy the upper surface that shows the intermediate output from the lower
shaping in molds, and low cost. surface. (c) and (d) Light spots on the reception plane for H
300
The LED light source is assumed to be a Lambert emitter in and 2000 mm, respectively.
a circular shape, of which the radius is 1 mm. As Table 1 lists, a
αc exceeding 60° is obtained, which converts to a light extrac-
tion efficiency of 77.3% for the LED source of standard
Lambert distribution. h2
100 mm and the height of the
highest Fresnel ring is 12 mm, approximately 12% of the
height of the upper surface. Therefore, the light rays incident
on the Fresnel ring can be well approximated as uniformly
distributed.
Based on the 3D model of the lens, the light path transmit-
ting the lens can be traced in the Monte Carlo method. The
reception plane is placed at H
300 and 2000 mm, to evalu-
ate the light uniformity at short and long distance, respectively.
A sum of 50,000 light rays are used for simulation, emitted Fig. 4. Normalized angular light intensity distribution of the colli-
from the LED source of 1 lm luminous flux. The result is mated beam on a far-field receptor in (a) polar and (b) rectangular
shown in Fig. 3. As an intermediated result, Fig. 3(b) shows coordinates. The red line in (a) shows the intensity of stray beams,
the light spot created by the lower surface on the upper surface. which have been amplified by 103 times.
The uniform light spot agrees with the design expectation,
which is used to calculate the slope of each Fresnel ring on
the upper surface. intensity contrast indicates that, most light rays are refracted
It is clearly seen that, after light passes through the lens, a into the collimated beam. Furthermore, the fast decay of the
highly collimated and uniform emergent light beam is ob- lineshape beside the normal direction in Fig. 4(b) implies a uni-
served, and the light spots shown in Figs. 3(c) and 3(d) at form intensity distribution that gives the light spot a sharp
H
300 and 2000 mm are distributed within radius R 2 . edge. The result shows good agreement with mathematical
Around the perimeter of the circle, the light spots are slightly design expectations.
denser than in the central area. Although the dense belt be- As a further check of the lighting effect, the light spots are
comes wider in Fig. 3(d) as H increases, the overall uniformity converted into a luminous intensity distribution on the recep-
is kept at an acceptable level. tion panel. The results are mapped into color scales, as shown
To examine the light intensity information, Fig. 4(a) shows in Fig. 5. A compact luminous circle is observed. In the central
the normalized angular intensity distribution on a far-field re- area of Fig. 5(a), the intensity approaches a maximal constant of
ceptor in the polar coordinate, and Fig. 4(b) shows the rectan- 28 lux, while in Fig. 5(b) the maximum appears in the perim-
gular counterpart. The collimated beam in Fig. 4(a) is marked eter area of an intermediate radius. By dividing the intensity
in blue, and the stray beams are marked in red, of which the distribution into grids [14], the root mean square of deviation
intensity is amplified by 103 times. The field and beam angle is is evaluated to be 3.8 and 4.0 lux for Figs. 5(a) and 5(b), re-
estimated to be less than 1° (the sampling resolution). The high spectively, as a quantitative measure of the uniformity of the
 4564 Vol. 56, No. 15 / May 20 2017 / Applied Optics
Fig. 5. Distribution of luminous intensity on the reception plane
for (a) H
300 mm and (b) H
2000 mm.
lens output. Due to a finite size of the light source, this result
falls much short of the thermodynamic limit of collima-
tion [15].
The above results show the advantages of the lens design,
which casts and focuses most emission from the light source
with a high accuracy, and thus realizes a high and uniform
luminous intensity even at a long distance.
The collimation of the output beam can be quantitatively
examined by evaluating the luminous intensity at the center
of the light spot as a function of H, and the results are shown
in Fig. 6(a). When H increases from 300 to 2000 mm, the
intensity increases monotonously by 6.7% from 29.5 to
31.5 lux, indicating the beam is slightly concentrated by the
lens. The design errors primarily come from the Fresnel surface,
as discussed previously.
Figure 6(b) shows the root mean square of luminous inten-
sity deviation in the light spot as a function of S. With S in-
creasing from 20 to 40, the root mean square decreases from
Fig. 6. (a) Luminous intensity at the light spot center as a function
of H from 300 to 2000 mm. (b) Root mean square of luminous in-
tensity deviation as a function of S from 20 to 50.
Research Article
9.2 to 3.8 lux. When S further increases to 50, the decrease
tends to saturate, implying S
40 is a threshold above which
the refracted light approaches a uniform and collimated beam.
B. Influence of Fresnel Ring Number
To investigate the influence of the Fresnel ring number, the
light spot on the reception plane with S
20 and 30 is simu-
lated, and the results are shown in Fig. 7. For reference, the
lateral curve of the upper lens surface corresponding to each
case is also shown above the spot figure.
When S
20, as Figs. 7(a) and 7(c) show, the maximal
Fresnel ring height exceeds 24 mm, and the spot clearly shows
a ring-like feature, with light intensity at the edge being lower
than at the center. There are two reasons accounting for this
result. First, the larger the ring height is, the light intensity
on the upper surface is further deviating from a uniform dis-
tribution, such that the light intensity at each ring becomes
different. Second, for rings of a large height, part of refracted
light rays may hit neighboring rings, leading to an unpredict-
able light path. When S
30, as Fig. 7(d) shows, the beam
uniformity becomes better, but the edge still holds a lower spot
density. When S increases to 40, as Fig. 3(d) shows, a uniformly
distributed light spot is obtained. Accordingly, S
40 is
adopted from the previous discussion.
C. Light Source Size Effect
Real light sources are of finite sizes. To see the size effect, the
lens performance under the illumination of the surface light
source with different areas are simulated. Assuming the shape
of the light sources is spherical, with the radius of 1, 2, and
3 mm, respectively, the angular distribution of light intensity
by tracing each light ray on a far-field acceptor is shown in
Fig. 7. The sample plane contains the normal direction of
the Fresnel surface.
As Fig. 8 shows, the light spots are primarily distributed
along the 90° axis (the normal direction), which implies a
highly collimated beam output. Here, we use the ratio (denote
by N ) between the upper surface height R 2 and the radius of
the light source to measure the relative size of the lens over the
Fig. 7. (a) and (b) Lateral curves of the Fresnel surface with S
20
and 30, respectively. (c) and (d) Light spots on the reception plane at
H
300 mm with S
20 and 30, respectively.
 Research Article
Fig. 8. Angular distribution of light intensity by tracing each light
ray on a far-field acceptor. The radius of the light source is (a) 1 mm,
(b) 2 mm, and (c) 3 mm.
light source. N
90, 45, and 30 in Figs. 8(a), 8(b), and 8(c),
respectively. When N decreases, the angular width of the pri-
mary beam area slightly increases, indicating the light beam
tends to diffuse as long as the assumption of a Lambert point
light source fails. It leads to larger spot area, and less sharp spot
edge. However, the overall collimation is kept at a high level.
The result suggests a wide adaption scope of the lens design on
the light source size.
D. Dispersion Effect
To probe into the dispersion effect, the refractive index of the
lens is slightly changed to take the response to a different wave-
length into account. In ordinary dispersion, n1 becomes larger
for violet light and smaller for red light, respectively.
It is found that, for violet light, a larger n1 results in a
stronger concentrating effect, so the spot center shows a higher
intensity. For the red light, on the other side, the uniformity
and collimation of the light beam improves. As revealed in
Fig. 6, the lens tends to exhibit a slight concentration effect
at a long distance. A smaller n1 of the red light compensates
for the concentration effect, which is beneficial to realize an
ideally collimated output. Considering the different propaga-
tion path of red and violet light in the lens, a lens structure
of position dependent refractive index can be proposed for
highly demanding applications. The relevant study will be car-
ried out in the future.
E. Disadvantageous Factors
Compared with previous results, our design shows improved
illumination uniformity. However, the lens aperture in our
model is larger, and the cost in fabricating the Fresnel surface
is higher than a normal freeform surface. Taking all the factors
into consideration, the lens type we propose can serve as
an alternative to LED collimator solutions for demanding
applications.
4. SUMMARY
To summarize, we propose a freeform-Fresnel double
surface lens design that realizes uniform and collimated light
Vol. 56, No. 15 / May 20 2017 / Applied Optics 4565
distribution on LED sources. The lower surface of the lens
transforms the Lambert emission from the LED source to a
uniformly distributed pattern, while the upper surface further
transforms the light to a collimated beam. The results can be
referred to as high intensity and long-distance LED lighting
design, which reduce the layout of optical elements for
obtaining uniform and collimated light beams to a single lens.
Potential factors that may influence the lens performance were
analyzed. By taking advantage of the different surface struc-
tures, this method gives the lens the features of compact vol-
ume, wide acceptance angle, high light extraction efficiency,
and high design accuracy.
Funding. Science and Technology Foundation for Youth
Talents of the Educational Commission of Hubei Province
of China (Q2015002).
REFERENCES
1. B. V. Lumileds Holding, Secondary Optics Design Considerations for
SuperFlux LEDs, Application Brief (2015).
2. J.-J. Chen and C.-T. Lin, “Freeform surface design for a light-emitting
diode-based collimating lens,” Opt. Eng. 49, 093001 (2010).
3. H. Wu, X. Zhang, and P. Ge, “Double freeform surfaces lens design for
LED uniform illumination with high distance-height ratio,” Opt. Laser
Technol. 73, 166–172 (2015).
4. Z. Zhen, S. Cheng-yue, F. Qian, and Z. Chun-hua, “LED
collimating lens based on free-form surface,” J. Appl. Opt. 33,
1058–1062 (2012).
5. E. Aslanov, L. L. Doskolovich, and M. A. Moiseev, “Thin LED collima-
tor with free-form lens array for illumination applications,” Appl. Opt.
51, 7200–7205 (2012).
6. C. Chen and X. Zhang, “Design of optical system for collimating
the light of an LED uniformly,” J. Opt. Soc. Am. A 31, 1118–1125
(2014).
7. J. J. Chen, T. Y. Wang, K. L. Huang, T. S. Liu, M. D. Tsai, and C. T. Lin,
“Freeform lens design for LED collimating illumination,” Opt. Express
20, 10984–10995 (2012).
8. H. Ries and J. Muschaweck, “Tailored freeform optical surfaces,”
J. Opt. Soc. Am. A 19, 590–595 (2002).
9. L. Gui, X. H. Sun, S. Y. Yin, C. L. Du, and Y. Xiang, “Modeling error
analysis and compensation design for free-form uniform illumination
lens,” Acta Optica Sinica 35, 1022008 (2015).
10. A. Davis and F. Kuhnlenz, “Optical design using Frensel lens,” Optik
Photon. 2, 52–55 (2007).
11. A. Goldstein and J. M. Gordon, “Double-tailored nonimaging reflector
optics for maximum-performance solar concentration,” J. Opt. Soc.
Am. A 27, 1977–1984 (2010).
12. R. Winston, J. C. Minano, and P. G. Benitez, with contributions by N.
Shatz and J. C. Bortz, Nonimaging Optics (Elsevier Academic, 2005).
13. Y. Wan, J. Liu, H. Lin, and X. Hui, “Fresnel lens of freeform surface for
realizing uniform light distribution on general LED light sources,” Laser
Optoelectron. Prog. 53, 062201 (2016).
14. Y. Wan, H. Lin, J. Liu, and X. Hui, “Design of freeform LED lens real-
izing uniform light distribution of high convergence,” Laser
Optoelectron. Prog. 53, 122202 (2016).
15. J. M. Gordon and D. Feuermann, “Optical performance at the thermo-
dynamic limit with tailored imaging designs,” Appl. Opt. 44, 2327–
2331 (2005).