Method of glass selection for color correction in

optical system design

Bráulio Fonseca Carneiro de Albuquerque,1,2,* Jose Sasian,2 Fabiano Luis de Sousa,1
and Amauri Silva Montes1
1
National Institute for Space Research –Engineering and Space Technology (INPE/ETE), Av dos Astronautas, 1.758,
12227-010 São José dos Campos – SP-Brazil
2
College of Optical Science, University of Arizona, 1630 East University Boulevard, Tucson, Arizona 85721, USA
*
[email protected]

Abstract: A method of glass selection for the design of optical systems with
reduced chromatic aberration is presented. This method is based on the
unification of two previously published methods adding new contributions
and using a multi-objective approach. This new method makes it possible to
select sets of compatible glasses suitable for the design of super-
apochromatic optical systems. As an example, we present the selection of
compatible glasses and the effective designs for all-refractive optical
systems corrected in five spectral bands, with central wavelengths going
from 485 nm to 1600 nm.
©2012 Optical Society of America
OCIS codes: (220.0220) Optical design and fabrication; (220.3620) (080.3620) Lens system
design; (220.1000) Aberration compensation; (080.0080) Geometric optics; (080.2720)
Mathematical methods (general).

References and links

1. P. Mouroulis, “Broadband achromatic telecentric lens,” Nasa Tech Briefs, NPO-44059, (2007).
2. J. L. Rayces and M. Rosete-Aguilar, “Selection of glasses for achromatic doublets with reduced secondary
spectrum. I. Tolerance conditions for secondary spectrum, spherochromatism, and fifth-order spherical
aberration,” Appl. Opt. 40(31), 5663–5676 (2001).
3. R. D. Sigler, “Glass selection for airspaced apochromats using the Buchdahl dispersion equation,” Appl. Opt.
25(23), 4311–4320 (1986).
4. C. Gruescu, I. Nicoara, D. Popov, R. Bodea, and H. Hora, “Optical glass compatibility for the design of
apochromatic systems,” Sci. Sin. 40(2), 131–140 (2008).
5. P. Hariharan, “Superachromatic lens combination,” Opt. Laser Technol. 31(2), 115–118 (1999).
6. P. Hariharan, “Apochromatic lens combinations, a novel design approach,” Opt. Laser Technol. 29(4), 217–219
(1997).
7. R. I. Mercado and P. N. Robb, “Color corrected optical systems and method of selecting optical materials
therefor,” U.S Patent, 5,210,646, (1993).
8. P. N. Robb, “Selection of optical glasses. 1: two materials,” Appl. Opt. 24(12), 1864–1877 (1985).
9. N. V. D. W. Lessing, “Selection of optical glasses in superachromats,” Appl. Opt. 9(7), 1665–1668 (1970).
10. T. R. Sloan, “Analysis and correction of secondary color in optical systems,” Appl. Opt. 9(4), 853–858 (1970).
11. M. Herzberger and N. R. McClure, “The design of superachromatic lenses,” Appl. Opt. 2(6), 553–560 (1963).
12. R. R. Willey, Jr., “Machine-aided selection of optical glasses for two-elements, three-color achromats,” Appl.
Opt. 1(3), 368–369 (1962).
13. R. E. Stephens, “Four-color achromats and superchromats,” J. Opt. Soc. Am. 50(10), 1016–1019 (1960).
14. W. S. Sun, C. H. Chu, and C. L. Tien, “Well-chosen method for an optimal design of doublet lens design,” Opt.
Express 17(3), 1414–1428 (2009).
15. I. Ono, Y. Tatsuzawa, S. Kobayashi, and K. Yoshida, “Designing lens systems taking account of glass selection
by real-coded genetic algorithms,” in Proceedings of IEEE International Conference on Systems, Man and
Cybernetics (Institute of Electrical and Electronics Engineers, New York, 1999), 7803–5731.
16. Y. C. Fang, C. M. Tsai, J. Macdonald, and Y. C. Pai, “Eliminating chromatic aberration in Gauss-type lens
design using a novel genetic algorithm,” Appl. Opt. 46(13), 2401–2410 (2007).
17. L. Li, Q. H. Wang, X. Q. Xu, and D. H. Li, “Two-step method for lens system design,” Opt. Express 18(12),
13285–13300 (2010).
18. R. E. Fischer, A. J. Grant, U. Fotheringham, P. Hartmann, and S. Reichel, “Removing the mystique of glass
selection,” Proc. SPIE 5524, 134–146 (2004).
19. W. J. Smith, Modern Optical Engineering (McGraw-Hill, Inc., 1990).

#162763 - $15.00 USD Received 8 Feb 2012; revised 27 Mar 2012; accepted 27 Mar 2012; published 4 Jun 2012
(C) 2012 OSA 18 June 2012 / Vol. 20, No. 13 / OPTICS EXPRESS 13592
20. P. N. Robb and R. I. Mercado, “Calculation of refractive indices using Buchdahl’s chromatic coordinate,” Appl.
Opt. 22(8), 1198–1215 (1983).
21. J. Branke, K. Deb, K. Miettinen, and R. Slowinski, Multiobjective Optimization: Interactive and Evolutionary
Approaches (Springer-Verlag, Berlin, 2008).
22. N. Srinivas and K. Deb, “Multi-objective function optimization using non-dominated sorting genetic algorithm,”
Evol. Comput. 2(3), 221–248 (1994).
23. J. Rayces and M. R. Aguilar, “Selection of glasses for achromatic doublets with reduced secondary color,” Proc.
SPIE 4093, 36–46 (2000).
24. N. Lopez, O. Aguirre, J. F. Espiritu, and H. A. Taboada, “Using game theory as a post-Pareto analysis for
renewable energy integration problems considering multiple objectives,” in Proceedings of the 41st International
Conference on Computers & Industrial Engineering, 678–683 Los Angeles, (2011).
25. O. Aguirre, H. Taboada, D. Coit, and N. Wattanapongsakorn, “Multiple objective system reliability post-Pareto
optimality using self organizing trees,” in Proceedings of IEEE International Conference on Quality and
Reliability (Institute of Electrical and Electronics Engineers, New York, 2011), 225–229.
26. E. Zio and R. Bazzo, “Clustering procedure for reducing the number of representative solutions in the Pareto
front of multiobjective optimization problems,” Eur. J. Oper. Res. 210(3), 624–634 (2011).
27. X. Blasco, J. M. Herrero, J. Sanchis, and M. Martínez, “A new graphical visualization of n-dimensional Pareto
front for decision-making in multiobjective optimization,” Inf. Sci. 178(20), 3908–3924 (2008).
28. J. C. Ferreira, C. M. Fonseca, and A. Gaspar-Cunha, “Methodology to select solutions from the Pareto-optimal
set: A comparative study,” in Proceedings of the 9th annual conference on Genetic and evolutionary
computation, (ACM, New York, NY, 2007), 789–796.
29. V. Venkat, S. H. Jacobson, and J. A. Stori, “A Post-optimality analysis algorithm for multi-objective
optimization,” Comput. Optim. Appl. 28(3), 357–372 (2004).
30. C. A. Coello Coello, “Handling preferences in evolutionary multiobjective optimization: a survey,” in
Proceedings of the 2000 Congress on Evolutionary Computation (Institute of Electrical and Electronics
Engineers, New York, 2000), 30–37.
31. SCHOTT N. America, Inc., “Optical glass catalogue- ZEMAX format, status as of 13th September 2011,
http://www.us.schott.com/advanced_optics/english/tools_downloads/download/index.html?PHPSESSID=utt2cbk
96nlk3gf7gjpb7ggt54#Optical%20Glass

1. Introduction
Multi-spectral imaging instruments are widely used in scientific instrumentation. The different
spectral bands, which frequently go beyond the visible region of spectrum, can allow
extraction of many important features and chemical-physical information from the objects
been imaged.
Common examples of this kind of instrument are multispectral satellite remote sensing
cameras, multispectral microscopes, and astronomical telescope multispectral cameras. These
instruments sometimes cover spectral bands from the UV to the thermal IR. Depending on the
instrument specification, an all-reflective solution for the optical system, which would be
chromatic aberration-free, is not always feasible, due to some disadvantages of this kind of
solution [1]. The design of a refractive optical system that can cover a wide spectral band
providing good image quality is not an easy task. According to Rayces and Aguilar [2], two
barriers impose limitations on an optical system performance, light diffraction and chromatic
aberration.
The chromatic aberration in imaging forming optical systems is a well-known issue
studied since the XVII century. As pointed out by Sigler [3], this topic has been one of most
investigated in optical design.
Several graphical and mathematical methods for the selection of optimum glass
combinations for the correction of chromatic aberration have been proposed [2–13]. However,
the problem of glass selection is wide in scope and in our opinion is not yet completely
solved. Even a recent publication about optimum glass selection [14] brings no relevant
contributions for the subject in our opinion.
Some contemporary methods propose the use of evolutionary [15,16] and hybrid [17]
optimization algorithms for the optimal glass selection. Despite of reporting excellent results,
we believe these techniques can be computationally demanding due to the extremely huge
number of different possibilities available even for a reasonable simple optical system.
Moreover, these methods do not guaranty that the best set of glasses has been found.

#162763 - $15.00 USD Received 8 Feb 2012; revised 27 Mar 2012; accepted 27 Mar 2012; published 4 Jun 2012
(C) 2012 OSA 18 June 2012 / Vol. 20, No. 13 / OPTICS EXPRESS 13593
 Fischer et al [18] mention that the glass selection in optical design has a mystique and
tends to be both a science and an art. Our goal in this paper is to present a synthesis method
that systematizes the task of glass selection for the design of color corrected optical systems,
making this an objective task. The proposed method is based on the unification of two other
methods proposed in the literature [2,7] with some important contributions added, and a multi-
objective approach for the problem.
In the next Section we present the motivation that conducted us to the development of this
new method of glass selection. In the Section 3 we present the background for the proposed
method, in Section 4 we explain the proposed method itself. In Section 5 we present one
example of the application of the method. Concluding this work in Section 6.
2. Motivation
The studies presented herein were motivated by the request for the feasibility study of a single
refractive optical system capable of covering and providing excellent image quality in five
spectral bands, going from the blue (0.45-0.485-0.52 µm), passing through green (0.52-55-
0.59 µm), red (0.63-0.66-0.69 µm), NIR (0.77-0.83-0.89 µm) and reaching the SWIR (1.5-1.6-
1.7 µm) spectral region.
Optical system covering wide band with good image quality is a challenge due to the
chromatic aberration. In this case a detailed study of how to design a broadband system like
this had to be conducted. The success of such system lies on the selection of the right set of
optical glasses used in the design [2,7,9,11,16]. For this purpose we started to study the
available methods of glass selection in the literature. During this survey, we identified points
that could be better developed in the available methods and we ended up with an improved
technique that we present in this paper.
Using the developed method it was possible to design a preliminary system for the five
spectral bands camera, which complies with the main requirements imposed to the system, as
we show in Section 5.
3. Background of the proposed method
After studding many methods of glass selection available in the literature [2–13], we realized
that the one proposed by Mercado and Robb [7] is the most theoretically rigorous and general.
The Mercado-Robb method considers in the formulation different number of glasses used in
the set, as well as different number of wavelengths for which the minimization of chromatic
aberration is desired. It is possible to affirm that other methods presented in the literature [e.g
4–6,9,11,13.] can be seen as special cases of this method.
Despite the general formulation of the Mercado-Robb method and the excellent discussion
provided in reference [7], the authors solve the problem for practical purposes just in some
specific cases. The cases for two glasses corrected from 2 to n wavelengths are very well
presented with all the necessary details. Nevertheless, for more than two glasses, only some
particular cases are discussed. One reason is that for more than two glasses, the metric
adopted to define how good a specific set of glasses is at color correction for n specific
wavelengths, becomes difficult to define and interpret geometrically. Furthermore, the
adopted method to calculate the optical power of each glass type does not have a general
equation, also becoming complicated in these cases. The method presented in this paper
improves the Mercado-Robb method with contributions that address some of the practical
implementations issues.
In spite of the method of color correction proposed by Rayces and Aguilar [2] being
limited to two glasses and three wavelengths, it establishes and makes use of some metrics
that appear not to have been reported before in glass selection theory. These metrics are not
related to color correction but are important for verifying if a set (in their case a pair) of
glasses can provide a successful design. In contrast to other glass selection methods, the
Rayces-Aguilar method uses as input not only the wavelengths, but also the focal length and

#162763 - $15.00 USD Received 8 Feb 2012; revised 27 Mar 2012; accepted 27 Mar 2012; published 4 Jun 2012
(C) 2012 OSA 18 June 2012 / Vol. 20, No. 13 / OPTICS EXPRESS 13594
the numerical aperture of the designed system. In the method presented in this paper, we
incorporate this metrics proposed by Rayces and Aguilar [2].
With different metrics of dissimilar physical natures for each possible glass arrangement,
the use of a multi-objective approach was very convenient and helpful to filter out the non-
dominated solutions and organize them in different Pareto rankings, helping the selection of
the most appropriate glass combination solution for the problem.
3.1-The Mercado and Robb method with some new contributions
The index of refraction of optical materials is a function of the wavelength. Several
mathematical models have been proposed to describe this dependence. Some are based in
physical models other are simply empirical functions [19]. One of these models, proposed by
Buchdahl [20] is given by Eq. (1).
N (λ ) = N 0 + ν 1ω (λ ) + ν 2ω (λ ) 2 + ⋯ + ν nω (λ ) n (1)
This model, as many others, is based on a Taylor series. N represents the refraction index
for wavelength λ. N0 is the refraction index in a reference wavelength λ0, and ω is a function
of the wavelength λ that is called chromatic coordinate:
δλ
ω= (2)
1 + αδλ
where δλ = λ-λ0, and α is a universal constant taken as 2.5 [7]. The dispersion coefficients νn,
are particular to a given glass. This dispersion equation proposed by Buchdahl converges
rapidly and can model optical glasses to a very good accuracy using only a few terms in the
series [20].
The develop method of glass selection for color correction takes advantage of this
dispersion equation. If a set of glasses is needed to minimize the chromatic aberration for n
wavelengths, Eq. (1) is expanded to include up to the n-1th algebraic power term. Then a
system of linear equations is obtained to compute the dispersion coefficients νn of each glass
were the number of unknowns is equal to the number of equations.
By passing N0 to the left side of Eq. (1) and by dividing both sides by the constant N0 –1,
we obtain:
n −1
D(λ ) = ∑ηiω (λ )i (3)
i =1

where: D ( λ ) = δ N ( λ ) / ( N 0 − 1) ; δ N (λ ) = N ( λ ) − N 0 and ηi = ν i / ( N 0 − 1) . The term D(λ)

is called dispersive power.
This equation is very important since the method presented in [7] is mainly based on it.
The optical power φ of a lens is defined as the inverse of the it’s focal length f:
1
φ= (4)
f
The optical power of a single thin lens for a wave λ is given by the relationship:

φ ( λ ) =  N ( λ ) − 1 ( C1 − C2 ) (5)

where C1 and C2 are the lens curvature.

For a specific optical material and a defined optical power, the quantity (C1 – C2) must be
a constant, conveniently called K. Thus we can write:

φ ( λ ) =  N ( λ ) − 1 K (6)

#162763 - $15.00 USD Received 8 Feb 2012; revised 27 Mar 2012; accepted 27 Mar 2012; published 4 Jun 2012
(C) 2012 OSA 18 June 2012 / Vol. 20, No. 13 / OPTICS EXPRESS 13595
 As a consequence, the power of a thin lens at λ0 can be expressed by:

φ ( λ0 ) =  N ( λ0 ) − 1 K (7)
By use of D ( λ ) = δ N ( λ ) / ( N 0 − 1) , and δ N (λ ) = N ( λ ) − N 0 , together with Eq. (7), we
can write for the optical power:

φ ( λ ) = φ ( λ0 ) 1 + D ( λ ) (8)

For a system of k thin lenses in contact, the resulting optical power for the reference
wavelength λ0 is computed by:
k
Φ ( λ0 ) = ∑ φ j ( λ0 ) (9)
j=1

Using Eq. (8) and Eq. (9), the total optical power for any wavelength λ can be written as:
k
Φ ( λ ) = Φ ( λ0 ) + ∑ φ j ( λ0 ) D j ( λ ) (10)
j =1

Assuming that each one of the k lenses is made out of a different glass, where k≥2, the
mathematical conditions for having an achromatized optical system in n wavelengths, where
n≥2, can be given by:
Φ ( λ1 ) = Φ ( λ 2 )
Φ ( λ2 ) = Φ ( λ 3 )
(11)
⋮
Φ ( λn −1 ) = Φ ( λ n )
Using Eq. (10), Φ(λ1), Φ(λ2), Φ(λ3), ..., Φ(λn) can be transcribed in the following form:
Φ ( λ1 ) = Φ ( λ0 ) + φ1 ( λ0 ) D1 ( λ1 ) + ⋯φ k ( λ0 ) Dk ( λ1 )
Φ ( λ2 ) = Φ ( λ0 ) + φ1 ( λ0 ) D1 ( λ2 ) + ⋯φ k ( λ0 ) Dk ( λ2 )
(12)
⋮
Φ ( λn ) = Φ ( λ0 ) + φ1 ( λ0 ) D1 ( λn ) + ⋯φ k
( λ0 ) Dk ( λn )
Using the set of Eq. (12), the conditions for achromatized optical systems are:
φ1 (λ0 ) ⋅ ( D1 (λ1 ) − D1 (λ2 ) ) + ...φk (λ0 ) ⋅ ( Dk (λ1 ) − Dk (λ2 ) ) = 0
φ1 (λ0 ) ⋅ ( D1 (λ2 ) − D1 (λ3 ) ) + ...φk (λ0 ) ⋅ ( Dk (λ2 ) − Dk (λ3 ) ) = 0
(13)
⋮
φ1 (λ0 ) ⋅ ( D1 (λn −1 ) − D1 (λn ) ) + ...φk (λ0 ) ⋅ ( Dk (λn −1 ) − Dk (λn ) ) = 0
The difference in the dispersive power of a particular glass j over the wavelength range
λ1< λ< λ2, can be written in simplified form as:
D j (λ1 , λ2 ) = D j (λ1 ) − D j ( λ2 ) (14)
By using the dispersive power definition in Eq. (3), we can rewrite Eq. (14) in terms of the
chromatic coordinate ω as:
n −1
D j (λ1 , λ2 ) = ∑ηij ω i (λ1 ) − ω i (λ2 )  (15)
i =1
Thus we can write the conditions for achromatized optical systems, expressed in Eq. (13),
in matrix form as:

#162763 - $15.00 USD Received 8 Feb 2012; revised 27 Mar 2012; accepted 27 Mar 2012; published 4 Jun 2012
(C) 2012 OSA 18 June 2012 / Vol. 20, No. 13 / OPTICS EXPRESS 13596
 ∆Ω ⋅η ⋅ Φ = 0 (16)

where, ∆Ω is a square matrix of order n-1 x n-1:

 ( ω1 − ω2 ) ( ω12 − ω2 2 ) ⋯ ( ω1n −1 − ω2 n −1 ) 
 
 ( ω2 − ω3 ) (ω2 2 − ω32 ) ⋯ ( ω2 n −1 − ω3n −1 ) 
∆Ω =  ⋯  (17)
 
 ⋮ ⋮ ⋮ ⋱ ⋮ 
(ωn −1 − ωn ) ( ωn −12 − ωn 2 ) ⋯ ( ωn −1n −1 − ωn n −1 ) 
η is a matrix of order n-1 x k:
 η11 η12 ⋯ η1k 
 
 η21 η22 ⋯ η2k 
  (18)
η = η31 η32 ⋯ η3k 
 ⋮ ⋮ ⋱ ⋮ 
 
 η(n−1)1 η(n−1)2 ⋯ η(n−1)k 

Φ is a matrix of order k x1:

 φ (λ ) 
 1 0 
 φ (λ )  (19)
Φ = 2 0 
 ⋮ 
 K λ0 )
φ ( 

and 0 is a matrix of the n-1 x 1 order:

 0 
 
 0  (20)
0 = 0 
 ⋮ 
 
 0 
The matrix ∆Ω is a square and doubtless nonsingular. As a consequence its inverse
( ∆Ω −1 ) exists. Multiplying both sides of Eq. (16) by ∆Ω −1 results in a condition to obtain a
solution free from chromatic aberration for all the wavelengths defined:
η ⋅Φ = 0 (21)
Equation (21) has a nontrivial solution (i.e. Φ ≠ 0 ) if and only if the matrix η rank is
lower than k (i.e. not a full rank matrix). This happens when there is a perfectly linear
dependence among the columns of matrix η . Nevertheless, for any practical and meaningful
situation, where k≤n-1, the linear dependence will virtually never be mathematically exact. As
a consequence, the matrix η rank will always be equal to k. This result makes the rank of
matrix η an inefficient metric either to identify sets of glasses that are free from chromatic

#162763 - $15.00 USD Received 8 Feb 2012; revised 27 Mar 2012; accepted 27 Mar 2012; published 4 Jun 2012
(C) 2012 OSA 18 June 2012 / Vol. 20, No. 13 / OPTICS EXPRESS 13597
aberration in the defined wavelengths, or to compare the residual chromatic aberration among
the different possible combination of glasses.
To solve this problem, Mercado and Robb provide a geometrical interpretation of the Eq.
(21). In this way, they suggest a geometric metric to verify how good a set of glass is at color
correction for a given set of wavelengths. The metric is easy to understand and visualize for
the case of two glasses. Nevertheless, for more than two glasses the interpretation changes and
becomes complicated. Another drawback is that the metric has no physical meaning.
In this paper we propose a different metric to verify how good a specific set of glasses is at
minimizing the chromatic aberration for a given set of wavelengths. The proposed metric has
a general form, not depending on the number of glasses used in the combination, and has a
direct physical meaning. This new metric is presented and explained in some paragraphs
ahead.
To minimize or correct the chromatic aberration, not only a specific set of compatible
glasses must be selected, but also the right optical power for the lenses made with each one of
these materials must be used. To calculate the optimum power of each glass that minimizes
the chromatic aberration, both Eq. (16) and Eq. (9) are used. To simplify the computation, we
normalize the focal length of the optical lens system for λ0. As a consequence Eq. (9)
becomes:
k

∑φ (λ ) = 1
j =1
j 0 (22)

This equation can be written in a matrix form as:

S ⋅Φ = 1 (23)

where S is a row vector of order 1xk, with all elements equal to one.
Putting together Eq. (16) and Eq. (22) we obtain:
 S 
  ⋅ Φ = ê (24)
 ∆Ω⋅ η 

where ê is a column vector of order nx1 with the first element equal to one and the others
zero as shown below:
 1 
 
 0  (25)
ê =  0 
 ⋮ 
 
 0 
 S 
Defining G =   , and assuming that n ≥ k and k≥2, we estimate an optimum Φ
 ∆Ω ⋅η 
applying in Eq. (24) the least square method, what results the following equation:

( )
ˆ = G t ⋅ G −1 ⋅ G t ⋅ eˆ
Φ (26)

Equation (26) computes the optimum power of the lenses made with each one of the glasses
considered in the set that minimizes the square sum of the chromatic change of power for the
n defined wavelengths. We point out that the equations provided by Mercado and Robb, to
compute the optical powers, are only related to some specific situations and do not use all the
glass information available. In contrast, the equation presented herein is general, uses all the

#162763 - $15.00 USD Received 8 Feb 2012; revised 27 Mar 2012; accepted 27 Mar 2012; published 4 Jun 2012
(C) 2012 OSA 18 June 2012 / Vol. 20, No. 13 / OPTICS EXPRESS 13598
glass dispersion coefficients available, and provides the minimum chromatic aberrations for
the glass set considered, in the n given wavelengths.
Now, it is possible to use the vector Φ̂ in Eq. (16), to obtain the minimum chromatic
change of power CCP , as expressed in Eq. (27).
ˆ
CCP = ∆Ω ⋅η ⋅ Φ (27)
Our metric to verify how suitable a specific set of glasses is for minimizing chromatic
aberration, for a given set of n wavelengths, is now established as the modulus of the vector
CCP .
As we have normalized the optical power (see Eq. (22)), we obtain an excellent
approximation for the chromatic focal shift by multiplying the vector CCP by the desired
effective focal length F for the optical system.
 f (λ ) − f (λ ) 
 2 1

 f (λ3 ) − f (λ2 )  (28)
  ≈ CCP ⋅ F
 ⋮ 
 f ( λ n ) − f (λn−1 ) 

As the chromatic focal shift is proportional to CCP it clearly gives physical meaning to our
metric.
3.2-The Rayces-Aguilar method
Rayces and Aguilar [2] proposed a method of glass selection where not only the chromatic
correction is considered, but also aberrations, that according to the authors, cannot be
corrected, namely spherochromatism and fifth order spherical aberration.
The Rayces-Aguilar method is based on an exhaustive search of combination of pairs of
glasses. The possible arrangements of glasses, deriving from a glass catalog, are tested. For
each glass set possibility the power of the glasses are computed to produce a thin achromatic
doublet solution for the two extreme wavelengths considered. The chromatic aberration for
the middle wavelength, also called the secondary spectrum, is computed. Based on the power
of the elements of the doublet and on the desirable aperture of the system, a first weeding out
of potentially useless solutions is carried out. This eliminates solutions with steep curves,
what is an indication of high-order monochromatic aberrations, which are difficult to correct
or balance. In the next step, the radius of each surface is computed to produce an aplanatic
solution to third-order approximation using structural aberration coefficients. Paraxial rays are
then traced to compute third–order sphero-chromatism and fifth-order spherical aberration.
Based on the magnitude of these aberrations, a second glass arrangement elimination is
carried out.
The output of the Rayces-Aguilar method is a table with solutions that comply with the
limits imposed for each aberration, ranked according to the secondary spectrum value. The
method provides a certain level of confidence for glass combinations solutions that may
provide a successful design.
3.3-The multi-objective approach
Despite being frequently considered as mono-objective, practical optimization problems have
in general more than one objective or criteria, which usually are conflicting. In the problem of
finding the best glass combinations for color correction we pursue more than one objective
and therefore the use of a multi-objective approach [21] is appropriate.

#162763 - $15.00 USD Received 8 Feb 2012; revised 27 Mar 2012; accepted 27 Mar 2012; published 4 Jun 2012
(C) 2012 OSA 18 June 2012 / Vol. 20, No. 13 / OPTICS EXPRESS 13599
 When a problem is treated as multi-objective, usually there is a set of solutions, and not
only one solution. One solution in this set cannot be considered, in principle, better than
another solution in the same set, because at least it will be worse than another solution in one
aspect or objective. This set of solutions is known as non-dominated solutions. When these
solutions are plotted in the objective-functions space they form the thus called Pareto front.
To illustrate the ideas of dominance, non-dominance and Pareto front consider Fig. 1
where blue and red dots represent solutions of a multi-objective problem as plotted in the
objective-function space F1 and F2.

Fig. 1. The graph shows solutions for a generic min -min multi-objective problem, plotted in the
objective-functions space F1 and F2. Dominated solutions are represented in blue, while red
dots represent non-dominated solutions.

Objective functions are metrics used to evaluate a specific characteristic of a solution. The
example considered in Fig. 1 represents a two objective minimum-minimum problem. This
means that the smaller the values of F1 and F2 the better the solution is.
The red dots represent the non-dominated solutions while the blue dots represent the
dominated solutions. A given solution “A” is considered dominated when there is a solution
“B” with at least one of objectives better than the objectives at solution “A”. The dominance
relationship among the dominated solutions can also be considered; for example, the solution
1 dominates solutions 2 and 3 [21].
When the non-dominated solutions are plotted on the object
objective
ive function space, they form
the so-called Pareto front, represented in the Fig. 1 by the dashed line connecting the red dots.
Another useful concept used in this paper is the Pareto rank [22]. For a set of possible
solutions of a specific problem the dom
dominance
inance definition can be applied several times. Each
time the previous non-dominated solutions are removed, giving place to the formation of a
new Pareto front. The different Pareto fronts that result are classified by ranking. For example,
in Fig. 1 the red dots forming the Pareto ranking equal to 1. Solutions 5, 6, 1 and 4 are the
ones forming the Pareto ranking 2 and so on.
The dominance, non-dominance, Pareto front and Pareto rank concepts can be used for
multi-objective problems containing any number of objectives.
4. The synthesis method of glass selection
With the background presented in the last section, the explanation of our method of glass
selection becomes straightforward. Its implementation involves several steps.
Step 1. As input data for the method, the designer must provide the effective focal length
F, the f number F/#, the n wavelengths that covers the desired spectral range, and the number
of the primary wavelength λ0. A glass catalog and the number of glasses used in the
combination (i.e. 2, 3, 4, etc) must also be specified.
Step 2. At the outset, the first n-1 dispersion coefficients ηi, are calculated for each glass in
the catalog. For that, the n specified wavelengths and their respective refractive index in the
corresponding glass are used in Eq. (3). This results in a system of linear equations with n-1

#162763 - $15.00 USD Received 8 Feb 2012; revised 27 Mar 2012; accepted 27 Mar 2012; published 4 Jun 2012
(C) 2012 OSA 18 June 2012 / Vol. 20, No. 13 / OPTICS EXPRESS 13600
equations and n-1 unknowns, that when solved provides the ηi dispersion coefficients. With
the specified wavelengths, the matrix ∆Ω is then calculated using Eq. (17).
Step 3. Next, all possible arrangements for the glasses from the specified catalog are
performed. For each possibility the optimum normalized power of each glass is computed
using Eq. (26). The sum of the absolute power of each arrangement, given by Eq. (29) below,
is used as a metric for the first weeding out. As pointed out by Rayces and Aguilar [2], high
power elements have steep surfaces that result in large monochromatic aberrations, involving
higher orders of aberration. This first cut eliminates potentially useless solutions. The metric
used here is different from the one presented in [2]. It is more general in terms of the number
of glasses used in the combination. This metric has been suggested in [11].
k
F1 = ∑ φ j (λ0 ) (29)
j =1

The user must set the maximum value for F1. The glass arrangements that have F1 values
larger than the specified value are discarded. This metric is not just used to eliminate potential
useless solutions but can also be used as one of the metrics in the multi-objective approach
proposed in this work. The next steps and calculations are only performed for the
arrangements that comply with the F1 limit imposed.
The vector CCP is than calculated by Eq. (27). The modulus of this vector, called F2 (
F2 = CCP ) can also be used in the multi-objective analysis. The smaller the value of F2 the
better the color correction the set of glasses provides as explained in Section 3.1.
Step 4. Following, a thin lens aplanatic solution for wavelength λ0 is found for each
candidate glass arrangement. To find the aplanatic solution, the system structural coefficient
for spherical aberration Ξ and coma Χ are set equal to zero, with the power of each glass
element calculated using Eq. (26). We ended up with the following set of equations.
k
Ξ = ∑ ξ j = 0; (30)
j=1

k
Χ = ∑ χ j = 0; (31)
j =1

 1 1   φ1 (λ0 ) 
 N1 ( λ0 ) − 1  r − r  =  F 
 1 2  
⋮ (32)
 1 1   φk (λ0 ) 
 N k ( λ0 ) − 1  r − =
  
 (2 k ) −1 r( 2 k )   F 
To find the aplanatic solution it is necessary to solve the above set of equations for r1 to
r2k.
For the case of a doublet, k = 2, there are four equations and four unknowns resulting in a
straightforward solution. As Eq. (30) has a quadratic dependence as a function of the radius
(see appendix A in [23]), two different aplanatic solutions can be obtained for each glass
arrangement. The best solution is retained where the definition for a better solution is based in
the metric F3 as explained ahead.
For k≥3 there are more unknowns than equations. For solving the set of equations in an
analytic and fast way, some constraint equations are added to make the number of unknowns
equal to the number of equations. For example, the case where k = 3 (triplet), two options for
the constraint equations are possible r3 = r2, or r5 = r4. The system can then be solved for both
cases; in each case two solutions exist, which means four total possible solutions. Once more

#162763 - $15.00 USD Received 8 Feb 2012; revised 27 Mar 2012; accepted 27 Mar 2012; published 4 Jun 2012
(C) 2012 OSA 18 June 2012 / Vol. 20, No. 13 / OPTICS EXPRESS 13601
only the better solution is retained. This same idea can be expanded for k>3. The solution for
the set of equations where k≥3 in not so trivial and is made with the help of a computer.
For each one of the possible retained solutions, the fifth-order spherical W060 ( λ0 ) and the
sphero-chromatism W040CL ( λ1 ⋯ λn ) wave aberration coefficients are calculated according to
the algorithm presented in [23]. The fifth-order spherical is calculated for the reference
wavelength λ0. The sphero-chromatism is calculated for all possible combinations of the input
wavelengths, and the worse case is assigned for the set.
Step 5. The third and last metric used in the multi-objective analysis is then computed by
the sum of the normalized fifth-order spherical W060 and normalized sphero-chromatism
W040CL wave aberration coefficients according to Eq. (33).

F3 = (W040CL + W060 ) (33)

where [2]:
14 ⋅W060 (λ0 )
W060 = (34)
20 7
and
14 ⋅ W040CL (λ1 ⋯ λn )
W040CL = (35)
6 5
The metric F3 is also used to define which of the possible aplanatic solutions for a specific
glass set is the best one, as mentioned above.
Step 6. For all the possible set of glass arrangements complying with the maximum
allowed metric F1, the best aplanatic solution is stored in a table with its respective F1, F2 and
F3 metric values. The data stored in the table are organized as shown in the Fig. 2. The r’s are
the radius of curvature of each surface and φ’s are the normalized optical power of each thin
lens.

Fig. 2. Format of the table where the data for each glass arrangement best aplanatic solution is
stored.

Step 7. The solutions are then organized into different Pareto ranks using the metrics F1,
F2 and F3.
Step 8. At last, a post-Pareto analysis is applied in the first or in the firsts Pareto ranks,
organizing the solutions in the out-put table from the best to the worse trade-off solutions.
In summary, the glass selection for the design of optical systems with reduced chromatic
aberration can be seen as a multi-objective optimization problem where the goal is to
minimize at the same time the objective functions F1, F2 and F3, subjected to: F1 ≤ Constant;
to Eq. (30), (31) and (32), and to some additional constrains when k≥3 (e.g. r3 = r2, or r5 = r4,
for the case when k = 3). The method we used here to solve the problem was an exhaustive
search.
The method is also represented in a flowchart form in Fig. 3.

#162763 - $15.00 USD Received 8 Feb 2012; revised 27 Mar 2012; accepted 27 Mar 2012; published 4 Jun 2012
(C) 2012 OSA 18 June 2012 / Vol. 20, No. 13 / OPTICS EXPRESS 13602
 Start

Calculate the ηi
Input data for the method: coefficients for
EFL, F/#, wavelengths,λ0,Glass all the glasses i=1
Catalog, N ͦ of Glasses used in the catalog.
in the arrangement,etc Define matrix ∆Ω

Calculate the
optical
Test Glass
N Were all Y
Compute F1 power of each possible Glass type
arrangement i
lens arrangements
in the set tested

Is F1 greater
than the maximum
Y i=i+1
allowed by
the user

N
Compute F2

Calculate the
Calculate F3 For the
possible aplanatic
For all aplanatic arrangement i
solutions for the
solutions store the solution
arrangement i
with the lowest F3

Apply the
Table of glass
dominance concept
arrangements
End Post Pareto using F1, F2 and F3
organized according
Analysis and organize the
to the post-Pareto
solution in
methodology used
Pareto ranks

Fig. 3. Flowchart of the proposed method of glass combination selection.

4.1-Post Pareto analysis

The Pareto front, or the Pareto rank 1, specifies the global non-dominated trade-off solutions
for the problem. In practice, the designer has to pick one solution from this set for designing
the optical system. Despite one solution in the Pareto front not being in principle considered
better than other solution in the same front, it is evident and intuitive that a discrimination
among the less satisfactory trade-offs and the most promising solution can be done. This
process of selecting a solution is called decision-making. Many methods for supporting this
process, also known as Post-Pareto analysis, can be found in the literature [24–30].
The task of post-Pareto analysis is not so easy; especially when the number of candidate
solutions is large and the number of objectives is greater than two, which is the case.
Depending on the number of glasses in the catalog and the number of glasses used in the
combination, hundreds of solutions are usually obtained in the Pareto front.
For this work, we used two methods of post-Pareto analysis described in sub-sections
4.1.1 and 4.1.2.
4.1.1-Minimum F2
Organizing the solutions in the Pareto front by the most important metric is intuitive to
perform the post Pareto analysis. In this paper the color correction is the most important
metric, given by F2. Rayces and Aguilar [2,23] also propose the organization of the output
table of their method by the color correction index, in their case given by the secondary color.
We recommend the use of this method if the number of glasses used in the set is much lower

#162763 - $15.00 USD Received 8 Feb 2012; revised 27 Mar 2012; accepted 27 Mar 2012; published 4 Jun 2012
(C) 2012 OSA 18 June 2012 / Vol. 20, No. 13 / OPTICS EXPRESS 13603
than the number of waveleng
wavelengths
ths defined and the spectral band is broad, covering different
regions of the spectrum. The best glass combination is not necessarily in the first line of this
table but probably among the first ones. The final choice will be made by the designer, in this
case, it is important to look for a solution with a low F3 but at the same time keeping F2 as low
as possible. Other glass parameters can also be considered in this final choice.
4.1.2-Minimum distance to the origin
Suppose a generic multi-objective problem with two objective functions O1 and O2, where the
goal is to minimize both functions. Suppose also that the Pareto front for this problem in the
objective function space can be represented as the line plotted in Fig. 4. This is in fact a very
usual shape for a Pareto front in a min-min problem. In this Figure we highlight the “knee”, a
region where the best trade-off solutions lays.

Fig. 4. Typical Pareto front for 2 objective min-min problem, showing the attributes used in the
post-Pareto analysis.

Looking at Fig. 4, it is possible to say that the bigger the length of vector gi the less
satisfactory trade-offs solution i provides. This vector gi connects the origin of the system to a
solution i on the Pareto front, having objectives values O1i and O2i. Due to possible different
physical meanings of the objective functions, completely different numerical values ranges
may be represented in each axis. This difference in range can be a problem for the use of the
vector numerical length as a metric. However, we can work out this issue through the
normalization of each one of the objectives. This can be done dividing O1i by Ō1, O2i by Ō2,
and so on, for each solution i. The solutions can than be organized according to the value gi ,
given in its general form by Eq. (36).
2
O 
m
gi = ∑  ob ,i  (36)
ob =1  Oob 

These Ōob values are not necessarily the highest values in the range of the solutions for
each objective as we show in Fig. 4. For instance, in this work we defined this normalization
factor for each variable as the value that accumulates 90% of the solutions used in the
analysis. Organizing the solutions in the Pareto front in a new table using the metric given by
Eq. (36), from the lowest to the highest, supports in a very nice way the decision-making.
keeping the final choice for the designer that should be limited among the firsts lines in the
table.
We recommend the use of this method if the number of glasses k is within n ≥ k > n/2. For
the case when n is equal to k, F2 is zero, so only functions F1 and F3 are used to calculate gi .
When k is lower than n, the use of only F2 and F3 to compute gi is recommended. Again, the

#162763 - $15.00 USD Received 8 Feb 2012; revised 27 Mar 2012; accepted 27 Mar 2012; published 4 Jun 2012
(C) 2012 OSA 18 June 2012 / Vol. 20, No. 13 / OPTICS EXPRESS 13604
best glass combination does not necessarily lie in the very first line of this table but probably
among the first ones, and the designer must make the final choice.
5. Example
In this section it is presented an example for the application of the glass selection method
proposed in this work. The specification of the lens system that motivated this development,
described in Section 2, is used as the example. Our intention is not to present a final design for
the problem but to show how the method can be used to effectively design a multi-spectral
lens system. In Table 1 it is shown the most important features specified for the optical
system.
Table 1. Basic Requirements for the optical system used as example.
ITEM REQUIREMENT
Effective Focal Length 250mm
F/ Number (f#) 5
Field of view ± 9 degrees
Spectral Bands 0.450-0.520; 0.520-0.590; 0.630-0.690;
0.770-0.890 e 1.5-1.7µm
Maximal Distortion 3%
Field relative illumination Constant ± 3%
Back focal length (BFL) Large enough to fit the spectral bands
beamsplitter (more than 50mm).
Design Resolution (MTF) Close to diffraction limit for all bands in
sagittal and tangential directions.
The inputs for the method can be extracted from the spec in Table 1. The focal length and
the F/# are taken directly from the table. For the wavelengths, the central values for each
spectral band were used: 0.485, 0.55, 0.66, 0.83 and 1.6 microns. The primary wavelength λ0
was set to 0.83 microns the due to its proximity to the central wave of the whole spectrum
covered by the instrument.
The newest available Schott glass catalog was selected [31] to run the method. However,
some specific glasses from this catalog were discarded: Lithotec-CAF2, N-PK51, N-PK52A,
N-FK51A, P-PK53, N-PSK53A and N-PSK53. Despite these glasses being very good options
for color correction, they were rejected due to their undesirable thermal behavior. Optical
systems designed with these glasses are potentially sensitive to temperature changes.
Normally, for small changes of temperature, the effect can be compensated with refocusing,
however, the application of the instrument object of this example, cannot afford either a
manual or an automatic refocus mechanism.
At first we ran the method for arrangements of two glasses. The limit F1 defined for this
case was 9. The post-Pareto analysis was applied only for the solutions in the Pareto ranking
1, using the method presented in Section 4.1.1.
In Table 2 we can see the first 10 rows of the output table, sorted from the smallest to the
biggest F2.
Table 2. Output table from the glass selection method for 2 glasses sorted by F2.
N° Glass 1 Glass 2 r1 r2 r3 r4 φ1 φ2 F1 F2 F3
771 N-BALF4 N-KZFS11 170.91 -37.58 -38.36 -351.12 4.63 -3.63 8.27 5.46E-04 16.28
606 N-BAK1 N-KZFS11 163.36 -47.77 -48.80 -403.20 3.82 -2.82 6.64 5.75E-04 8.69
1383 N-KZFS11 N-BAK1 111.63 37.08 36.53 3124.93 -2.82 3.82 6.64 5.75E-04 8.59
1417 N-KZFS11 N-SK2 129.72 38.16 37.98 3141.56 -2.90 3.90 6.79 7.36E-04 6.94
4394 N-SSK5 KZFS12 168.54 -58.44 -58.88 -1078.81 3.74 -2.74 6.47 2.50E-03 6.12
216 KZFS12 N-SSK5 134.49 42.57 42.37 1796.69 -2.74 3.74 6.47 2.50E-03 5.91
2092 N-LAK12 KZFS12 168.51 -74.49 -74.20 -2741.16 3.23 -2.23 5.47 3.72E-03 4.35
191 KZFS12 N-LAK12 145.49 50.04 50.19 1765.35 -2.23 3.23 5.47 3.72E-03 4.29
212 KZFS12 N-SK2 118.05 48.24 47.64 2732.27 -2.09 3.09 5.18 3.80E-03 4.32
596 N-BAK1 KZFS12 160.66 -65.53 -67.81 -355.23 3.03 -2.03 5.07 4.20E-03 4.92

We believe that the fourth line of Table 2 brings the best trade-off option for the
combination of 2 glasses for the problem. Solutions above the fourth line have F2 values
slightly smaller, however, the F3 values are significantly higher. Bellow the forth line the F2
values increase very fast.

#162763 - $15.00 USD Received 8 Feb 2012; revised 27 Mar 2012; accepted 27 Mar 2012; published 4 Jun 2012
(C) 2012 OSA 18 June 2012 / Vol. 20, No. 13 / OPTICS EXPRESS 13605
 Before we go for the design of the optical system with the selected pair of glass, we can
perform a roughly check to see if it is promising in terms of the color correction. It is known
that the tolerable depth of focus of a system can be given by [19]:
ε = ±2λ ( f #)2 (37)
Calculating Eq. (37), using the f# provided in the Table 1 and the selected primary
wavelength λ0 (0.83µm), results in ε = ± 0.0415mm. We can compare this value to the result
from the multiplication between F2 and the focal length F, which for the selected pair gives
0.184mm. This number is much higher than the calculated ε, telling us that the design with the
selected pair is not promising. Even with the lowest F2 value shown in the first line of Table
2, we cannot even get close to the calculated ε. The conclusion is that more glasses to the set
are necessary to design the desirable system with the glass catalog used.
In this case, we ran the method again with the same parameters but now for three glasses
in the set. With more glasses, the F1 limit was changed to 11. The most suitable post-Pareto
method in this situation is the one presented in Section 4.1.2, where the metric gi is
calculated using only F2 and F3.
In Table 3 we can see the first 15 rows of the output table resulted from the application of
the method using 3 glasses in the set. The solutions are sorted from the smallest to the biggest
gi .
For the case of three glasses, we selected two solutions among the first lines that we
believe to be good trade-offs. The chosen solutions are located in the fifth and thirteenth lines
of Table 3. The first one has a smaller F3 than the second one and also a better power
distribution among the lenses. On the other hand the second has a smaller F2. Calculating the
multiplication between F2 and the focal length for both solutions we get 0.047mm and
0.014mm respectively. These numbers reveals that these glass combinations are promising in
terms of color correction.
Figure 5 shows the chromatic focal shift for the two aplanatic thin triplet obtained from the
glass combination chosen from Table 3. The one in the left side of the Fig. 5 corresponds to
the glasses on the fifth line of Table 3 (N-BAF52, N-KZFS11 and N-BAK2) while the one in
the right side corresponds to the ones on the thirteenth line of Table 3 (N-KZFS8, P-SF68 and
N-SK2). This graphs reveals that despite not crossing the axis five times in the center of all
spectral bands, the shift is not greater than 33 microns for the combination N-BAF52, N-
KZFS11 and N-BAK2 and less than 13 microns for the combination N-KZFS8, P-SF68 and
N-SK2 for the central wavelength of each spectral band. This gives us confidence that the
design of the objective can be done. The residual chromatic focal shift can be compensated
with a slightly change on the position of each spectral band image plane, as each band will
focus in a different detector. In this case we can go for the design.
The output glass combination chosen after the application of the method proposed can
then be used to design an optical system either applying classical or evolutionary methods. In
this last one the advantage is the significantly reduction of the design space, decreasing the
number of glasses from thousands to just a few.

#162763 - $15.00 USD Received 8 Feb 2012; revised 27 Mar 2012; accepted 27 Mar 2012; published 4 Jun 2012
(C) 2012 OSA 18 June 2012 / Vol. 20, No. 13 / OPTICS EXPRESS 13606
 Table 3. Output table from the glass selection method for 3 glasses sorted by gi .

N° Glass 1 Glass 2 Glass 3 r1


r6 φ1 φ2 φ2 F1 F2 F3

5317 N-BAF52 N-KZFS11 N-SK4 145.26


-1056.62 3.12 -4.68 2.56 10.37 2.15E-04 2.98
35786 N-SK4 N-KZFS11 N-BAF52 135.68


-2128.74 2.56 -4.68 3.12 10.37 2.15E-04 2.99
11387 N-KZFS11 N-BAF52 N-SK4 -42.02


-55.13 -4.68 3.12 2.56 10.37 2.15E-04 3.03
25496 N-LAK8 N-KZFS11 N-BAK2 -20.89


-23.05 -3.21 -1.58 5.79 10.59 4.61E-04 1.49
5310 N-BAF52 N-KZFS11 N-BAK2 110.35


-1266.01 3.34 -4.71 2.38 10.43 1.88E-04 3.14
26091 N-LASF31A N-KZFS2 N-BAK1 -16.58


-17.83 -2.09 -2.03 5.13 9.26 5.11E-04 0.67
13718 N-KZFS11 N-SSK8 N-SK4 -45.56


-62.64 -3.71 2.41 2.29 8.41 4.28E-04 2.04
13672 N-KZFS11 N-SSK8 N-BAK2 -47.45


-65.80 -3.67 2.56 2.11 8.35 4.17E-04 2.14
9960 N-FK5 N-KZFS11 N-BAF51 165.17


-264.64 2.02 -4.37 3.35 9.74 3.18E-04 2.84
11541 N-KZFS11 N-BAK2 N-SSK8 -43.54


-58.25 -3.67 2.11 2.56 8.35 4.17E-04 2.23
6824 N-BAK2 N-KZFS11 N-SSK8 142.16


-444.22 2.11 -3.67 2.56 8.35 4.17E-04 2.28
37966 N-SSK8 N-KZFS11 N-BAK2 115.23


-1534.96 2.56 -3.67 2.11 8.35 4.17E-04 2.31
20485 N-KZFS8 P-SF68 N-SK2 98.06


812.98 -4.73 1.17 4.56 10.46 5.66E-05 3.56
2283 KZFS12 N-SF4 N-SK4 112.05


1464.73 -4.48 1.54 3.94 9.96 3.67E-04 2.68
11237 N-KZFS11 LLF1 N-SK14 -42.00


-54.57 -4.20 2.49 2.70 9.39 3.75E-04 2.63

Fig. 5. Chromatic focal Shift for the aplanatic triplets designed with glass combination (a) N-
BAF52, N-KZFS11 and N-BAK2, and combination (b) N-KZFS8, P-SF68 and N-SK2.

The lens design lay out for the glass combination N-BAF52, N-KZFS11 and N-BAK2 can
be seen in the top left side of Fig. 6, where each lens glass is identified. The system complies
with all the basic requirements presented in Table 3. The image quality is also shown in Fig. 6
through the MTF curves. The Quality is fair for the blue band and great for the other bands.

#162763 - $15.00 USD Received 8 Feb 2012; revised 27 Mar 2012; accepted 27 Mar 2012; published 4 Jun 2012
(C) 2012 OSA 18 June 2012 / Vol. 20, No. 13 / OPTICS EXPRESS 13607
 1.0
µm
Spectral band: 0.450-0.520µ
0.9
o o
N-BAF52 N-KZFS11
0.8 Diff. Limit T 0.00 S 0.00
o o
T 7.00 S 7.00
0.7 o o
T 9.00 S 9.00
0.6

MTF
0.5
0.4
N-BAK2 0.3
329.31mm
0.2
(a) 0.1 (b)
0.0
0 45 90 135 180 225 270 315 360 405 450
Spatial Frequency in cycles per mm
1.0 1.0
µm
Spectral band: 0.520-0.590µ µm
Spectral band: 0.630-0.690µ
0.9 0.9
o o o o
0.8 Diff. Limit T 0.00 S 0.00 0.8 Diff. Limit T 0.00 S 0.00
o o o o
T 7.00 S 7.00 T 7.00 S 7.00
0.7 o o 0.7 o o
T 9.00 S 9.00 T 9.00 S 9.00
0.6 0.6

MTF
MTF

0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 (c) 0.1 (d)
0.0 0.0
0 39 78 117 156 195 234 273 312 351 390 0 32 64 96 128 160 192 224 256 288 320
Spatial Frequency in cycles per mm Spatial Frequency in cycles per mm
1.0 1.0
µm
Spectral band: 0.770-0.830µ µm
Spectral band: 1.500-1.700µ
0.9 0.9
o o o o
0.8 Diff. Limit T 0.00 S 0.00 0.8 Diff. Limit T 0.00 S 0.00
o o o o
T 7.00 S 7.00 T 7.00 S 7.00
0.7 o o 0.7 o o
T 9.00 S 9.00 T 9.00 S 9.00
0.6 0.6
MTF

MTF

0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 (e) 0.1 (f)
0.0 0.0
0 26 52 78 104 130 156 182 208 234 260 0 14 28 42 56 70 84 98 112 126 140
Spatial Frequency in cycles per mm Spatial Frequency in cycles per mm

Fig. 6. Layout (a) and MTFs (b)(c)(d)(e)(f) for each spectral band and field position for the
design made with glass combination N-BAF52, N-KZFS11 and N-BAK2.

In Fig. 7 the layout for the lens system designed with the glass combination N-KZFS8, P-
SF68 and N-SK2 is shown in the top left side. The glass of each one of the lenses is identified
in the layout. Notice that the glass P-SF68 is present only in one lens. This reflects the big
difference in the power distribution between the 2 positive lenses as shown in Table 3.
The MTF curves for each one of the spectral bands for this system is also presented in Fig.
7. Again the image quality is fair for the blue band and excellent for the other bands for the
whole field of view.
For both systems presented, the MTF curves reported were obtained with each spectral
band focusing in its best focus. The prescription data for the lens shown in Fig. 6 and Fig. 7
are presented in Table 4 and Table 5 in the end of this section. The systems were designed
using only spherical lenses.

#162763 - $15.00 USD Received 8 Feb 2012; revised 27 Mar 2012; accepted 27 Mar 2012; published 4 Jun 2012
(C) 2012 OSA 18 June 2012 / Vol. 20, No. 13 / OPTICS EXPRESS 13608
 1.0
µm
Spectral band: 0.450-0.520µ
0.9
N-SK2 o o
0.8 Diff. Limit T 0.00 S 0.00
o o
T 7.00 S 7.00
0.7 o o
T 9.00 S 9.00
0.6

MTF
0.5
0.4
0.3
P-SF68 N-KZFS8
0.2
282.08mm
0.1
(a) (b)
0.0
0 45 90 135 180 225 270 315 360 405 450
Spatial Frequency in cycles per mm
1.0 1.0
µm
Spectral band: 0.520-0.590µ µm
Spectral band: 0.630-0.690µ
0.9 0.9
o o o o
0.8 Diff. Limit T 0.00 S 0.00 0.8 Diff. Limit T 0.00 S 0.00
o o o o
T 7.00 S 7.00 T 7.00 S 7.00
0.7 o o 0.7 o o
T 9.00 S 9.00 T 9.00 S 9.00
0.6 0.6

MTF
MTF

0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
(c) (d)
0.0 0.0
0 39 78 117 156 195 234 273 312 351 390 0 32 64 96 128 160 192 224 256 288 320
Spatial Frequency in cycles per mm Spatial Frequency in cycles per mm
1.0 1.0
µm
Spectral band: 0.770-0.830µ µm
Spectral band: 1.500-1.700µ
0.9 0.9
o o o o
0.8 Diff. Limit T 0.00 S 0.00 0.8 Diff. Limit T 0.00 S 0.00
o o o o
T 7.00 S 7.00 T 7.00 S 7.00
0.7 o o 0.7 o o
T 9.00 S 9.00 T 9.00 S 9.00
0.6 0.6
MTF

MTF

0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0.0
(e) 0.0
(f)
0 26 52 78 104 130 156 182 208 234 260 0 14 28 42 56 70 84 98 112 126 140
Spatial Frequency in cycles per mm Spatial Frequency in cycles per mm

Fig. 7. Layout (a) and MTFs (b)(c)(d)(e)(f) for each spectral band and field position for the
design made with glass combination N-KZFS8 P-SF68 and N-SK2.

Despite being very good designs, the systems presented in this example might not
represent final designs for the system that motivated the development of the glass selection
method proposed herein. Probably more elements in the system will be needed in order to
comply with all the detailed optical requirements necessary for the system, as well as to
accommodate the beam splitters necessary for the spectral bands separation in the different
detectors. Although, care was taken to design systems that are very representative in order to
show the feasibility of the project. For example we avoided the use of some glasses with
potential thermal problems, and also cemented lens that would facilitate the design,
controlling easier the lateral color, but would not be desired for the final system due to some
thermo mechanical constraints.

#162763 - $15.00 USD Received 8 Feb 2012; revised 27 Mar 2012; accepted 27 Mar 2012; published 4 Jun 2012
(C) 2012 OSA 18 June 2012 / Vol. 20, No. 13 / OPTICS EXPRESS 13609
 Table 4. Prescription data for the system shown in Fig. 6.
System 1. EFL = 250mm; F/# = 5.
Surf Radius (mm) Thickness (mm) Glass
OBJ Infinity Infinity
1 114.506 7.737 N-BAF52
2 1551.850 2.887
3 −787.580 8.220 N-KZFS11
4 58.764 2.534
5 60.880 8.455 N-BAK2
6 173.824 59.804
STO Infinity 20.929
8 125.540 10.496 N-BAF52
9 −80.372 2.299
10 −75.012 7.000 N-KZFS11
11 84.061 2.084
12 84.729 12.000 N-BAK2
13 −170.627 126.151
14 −69.911 7.833 N-KZFS11
15 −207.232 50.887
Table 5. Prescription data for the system shown in Fig. 7.
System 2. EFL = 250mm; F/# = 5.
Surf Radius (mm) Thickness (mm) Glass
OBJ Infinity Infinity
1 120.751 6.425 N-SK2
2 −1753.068 4.809
3 −112.487 6.002 N-KZFS8
4 179.109 1.011
5 115.631 8.594 N-SK2
6 −119.451 0.130
STO Infinity 11.169
8 12071.420 6.000 P-SF68
9 −141.327 2.586
10 −107.188 8.000 N-KZFS8
11 62.025 3.207
12 68.587 12.000 N-SK2
13 159.482 40.000
14 163.086 12.000 N-SK2
15 −257.749 98.105
16 −82.596 12.000 N-KZFS8
17 −253.127 50.000

6. Conclusion
In the research of available methods and techniques of glass selection for color correction in
lens design, we realized that all presented approaches in the literature had some drawbacks
and/or missing points. During the literature survey two of the reviewed methods called our
attention: the Mercado and Robb method [7] and the Rayces and Aguilar method [2]. The first
one presents a technique in a very general form in terms of the number of wavelengths where
the achromatization is desired, spectral region, and number of glass material used in the set.
However, it has some practical implementation issues that limit its use to some specific cases.
The second is limited to the combination of only two glasses and three wavelengths.
Nevertheless, it proposes the use of some metrics very important in the identification of
promising glass combinations that can potentially provide good final designs.
Unifying these two mention methods, providing some original contributions to Mercado
and Robb technique that repair its practical issues, and using a multi-objective approach, a
new method of glass selection for color correction was developed. This new method offers

#162763 - $15.00 USD Received 8 Feb 2012; revised 27 Mar 2012; accepted 27 Mar 2012; published 4 Jun 2012
(C) 2012 OSA 18 June 2012 / Vol. 20, No. 13 / OPTICS EXPRESS 13610
significant advantages in the task of glass selection, leading to optimum choice of optical
glasses for specific problems, as pointed out and demonstrated here.
Along this paper we went from the background theory, passing through a detailed
description of the new proposed method and finally wrapped it up with a practical example.
The design examples demonstrated the power of the proposed method in fiddling compatible
glasses that are able to conduct to excellent final designs. The results of this paper offer
significant improvements to the problem of glass selection in lens design when color
correction is an important matter, converting this task into a systematic and objective work.
The efficiency of this new method will be investigated in future papers for other lens design
problems, involving different spectral bands and types of lens systems.
We intend to incorporate the presented method of glass selection together with
evolutionary optimization methods in lens design that we have been working on. The glass
selection method has the ability to identify the most promising glasses to be used in a certain
design, reducing the glass options from hundreds to just a few. As a consequence the design
space is reduced and simplified, what is a significant advantage for global search heuristic
methods.
In a near future, free stand-alone version of the software with the method presented here
will be provided. The software will be found in the site:
http://www.optics.arizona.edu/glassselectiontool. Meanwhile, we can provide free of charge
tables of glass combinations upon request for specific input data provided. The request can be
done by e-mail: [email protected].
Acknowledgments
B. F. C. Albuquerque gratefully acknowledges the Brazilian Research Council (CNPq) for
supporting these studies and the Kidger Optics Associates for also supporting these studies
through the 2010 Michael Kidger Memorial Scholarship in Optical Design.

#162763 - $15.00 USD Received 8 Feb 2012; revised 27 Mar 2012; accepted 27 Mar 2012; published 4 Jun 2012
(C) 2012 OSA 18 June 2012 / Vol. 20, No. 13 / OPTICS EXPRESS 13611