Calculation and evaluation of narcissus for

diffractive surfaces in infrared systems

Tao Liu, Qingfeng Cui,* Changxi Xue, and Liangliang Yang

Department of Optical Engineering, Changchun University of Science and Technology,
Changchun 130022, China
*Corresponding author: cuiqf@yahoo.com.cn

Received 6 December 2010; revised 24 February 2011; accepted 26 February 2011;

posted 2 March 2011 (Doc. ID 139080); published 27 May 2011

In infrared optical systems, the narcissus effect for diffractive surfaces should be calculated with specific
diffraction orders based on the diffraction efficiency. It is shown in this work that the diffraction order of
maximum diffraction efficiency varies with the change of the incident angle and wavelength of the back-
ward-traced narcissus flux. Meanwhile, yni, which is the paraxial evaluation criterion of narcissus in-
tensity for a refractive surface, is modified considering diffraction when a ray passes through diffractive
surfaces, and a practical example has been given. The analysis can be used to calculate and control the
narcissus intensity in infrared optical systems with diffractive surfaces. © 2011 Optical Society of
America
OCIS codes: 220.0220, 110.3080, 050.1965, 050.0050.

1. Introduction sign. Because of the simplicity of paraxial ray-tracing

Narcissus generated by the temperature difference
between a cold detector and a warm environment
has been a well-known problem for cooled infrared
(IR) imaging systems [1,2]. The optical system’s im-
age quality could be decreased by the reflected flux
that emits from the cold detector. Much effort has
gone toward analyzing and controlling the narcissus
phenomenon with approximate and accurate calcula-
tions and sequential and nonsequential ray tracing
[1,3–5].
In [1], the intensity and size of the narcissus ghost
are given in terms of paraxial ray data at the contri-
buting surfaces and the system f -number. The ana-
lysis showed that the radius of the narcissus cold
spot on the detector is −4ynif -number, where y is
the on-axis marginal ray height, i is the incident an-
gle of the normally traced marginal ray at the surface
under consideration, and n is the refractive index be-
fore the surface. In modern optical design software,
for example, ZEMAX [6], yni is calculated and con-
strained to control the narcissus effect in optical de-
0003-6935/11/162484-09$15.00/0
© 2011 Optical Society of America
methods, the effect of beam clipping by limiting the
apertures of lenses is not taken into account. A more
meaningful evaluation of narcissus involved calcu-
lating the equivalent temperature difference. In
[3,4], finite ray-tracing computation was done to cal-
culate the narcissus return signal. In those discus-
sions, a bunch of rays, which cover the aperture of
the cold stop, are traced backward from the detector
to the optical surface. Each of the emitted rays repre-
sents an equal amount of flux. The rays are traced
backward from the detector to the reflecting surface
and back to the detector after reflection. The ratio
of the returned rays is analyzed to assess the aper-
ture clipping and equivalent temperature difference
caused by narcissus.
Because of the distinct color dispersion character
and thermal property, a diffractive surface is often
used to decrease the total number of lenses used.
When there are diffractive surfaces, according to
[7], the narcissus effect is different from the trans-
mitted and reflected radiation. The authors pointed
out that the narcissus calculation must be considered
with specific orders of diffraction based on their effi-
ciencies. Higher diffraction efficiency leads to more

2484 APPLIED OPTICS / Vol. 50, No. 16 / 1 June 2011

narcissus contribution. That conclusion is used in [5]
to simulate the narcissus phenomenon in the nonse-
quential ray-tracing mode. In [7], however, the re-
flected narcissus radiation was evaluated only for
vertical incident rays on flat diffractive surfaces.
In real optical systems, the incident angles of a back-
ward-traced cold ray are usually nonzero. As the in-
cident angle varies, the order of maximum diffraction
efficiency would be changed. The diffraction effi-
ciency is also wavelength dependent, which means
the order of maximum diffraction efficiency differs
with shifts of wavelength. The relationships among
diffraction efficiency, incident angle, and wavelength
for backward-traced narcissus flux are discussed in
Section 2 for diffractive optical surfaces.
In IR optical system design, the influence of narcis-
sus is mainly related to narcissus intensity, which is
proportional to the reciprocal of the narcissus ghost
area. For simplification, the paraxial parameter yni,
which is derived from the calculation of the narcissus
ghost area, is often used to stand for the concentra-
tion of the reflected narcissus flux at the central de-
tector. The commonly used paraxial data, however,
are computed assuming the vertex radius of curva-
ture is an acceptable approximation to the surface
power over the entire aperture of the surface, and
the effect of diffraction is not taken into account.
Since the direction of propagation is deviated when
a ray passes through the diffractive surface, Snell’s Fig. 1. Reflection of backward-traced narcissus radiation for dif-
law is not applicable. In Section 3, the criterion of fractive surfaces: (a) back diffractive surface and (b) front diffrac-
yni is modified with consideration of the diffraction tive surface.
for a specified diffractive surface with maximum
diffraction efficiency. It is shown that diffractive sur- 1 0
faces can influence the narcissus effect in IR optical ϕs ¼ ðn l1 − nl2 Þ; ð1Þ
λ0
systems and need to be considered in the optical de-
sign process. where l1 and l2 are optical paths of the two neighbor-
ing parallel rays and n and n0 are the indices before
2. Diffraction Efficiency for Backward-Traced and after reflection. It should be pointed out that, in
Narcissus Flux Fig. 2(a), where the perpendicular foot is below the
optical surface, l1 is defined as a negative value fol-
A. Diffraction Order of Maximum Diffraction Efficiency for lowing our sign convention. In Fig. 2(b), because of
Design Wavelength λ0 the relatively small incident angle ichief , the perpen-
dicular foot is above the optical surface, and l1 is de-
Two IR lenses with back or front diffractive surfaces
fined as positive in such a case. l2 is always positive.
of positive power are shown in Fig. 1. In order to eval-
To calculate the value of l1 in Fig. 2(a), two right
uate the diffraction efficiency for backward-traced
triangles, ▵ABC and ▵AEF, have been set. From
narcissus rays, a bunch of rays emit from the detec- ▵ABC, we can find that the relationship between
tor, filling the cold stop, and are reflected back by dif- AB and AC as shown is
fractive surfaces.
First, for the back diffractive surface in Fig. 1(a), AC ¼ AB sin ∠ABC ¼ AB sin i0 chief ; ð2Þ
whose facets are sloped toward the incident wave-
front according to [8], each continuous phase grating where i0 chief is refraction angle.
structure could be modeled by N-level phase profiles, By substituting AB ¼ δy þ AE and AC ¼ AF − l1
as shown in Figs. 2(a) and 2(b). The two situations in into Eq. (2), where l1 is a negative value, we can
Fig. 2 are the same except that the perpendicular foot derive
point C is below the optical surface in Fig. 2(a) and
above the optical surface in Fig. 2(b). With the scalar AF − l1 ¼ ðδy þ AEÞ sin i0 chief ; ð3Þ
approximation for such a case, the phase difference
in waves between two neighboring subperiods for de- where δy is the step width between neighboring
sign wavelength λ0 can be expressed as subperiods.

1 June 2011 / Vol. 50, No. 16 / APPLIED OPTICS 2485

 l2 − BG ¼ ðδy − GHÞ sinð−ichief Þ: ð8Þ

From ▵BGH, the useful relationships are

δz δz δz
BG ¼ ¼ ¼ ; ð9Þ
cos ∠GBH cosð−ichief Þ cos ichief

GH ¼ δz tan ∠GBH ¼ δz tanð−ichief Þ: ð10Þ

By substituting Eq. (9) and (10) into Eq. (12), after

some rearrangement, we can obtain that

δz
l2 ¼ − sin ichief ðδy þ δz tan ichief Þ: ð11Þ
cos ichief

By inserting Eqs. (6) and (11) into Eq. (1) and con-
sidering i0 chief =ichief ¼ n=n0 and n0 ¼ −n, the expres-
sion for ϕs can be simplified to

2nδz cos ichief

ϕs ¼ − : ð12Þ
λ0

The maximum phase delay in waves of all sub-

periods in one period structure can be given as

2Nnδz cos ichief

ϕ0 ¼ Nϕs ¼ − ; ð13Þ
Fig. 2. Rays reflected by two neighboring subperiods of N-level λ0
phase profiles: (a) the perpendicular foot point C below the optical
surface and (b) the perpendicular foot point C above the optical where N is the number of phase levels within dimen-
surface (back diffractive surface). sion T ¼ Nδy.
Since the optimum depth for a diffractive optical
element (DOE) designed to operate at the mth order
It is easy to point out from the right triangle ▵AEF is d ¼ Nδz ¼ mλ0 =ðn0 − 1Þ, Eq. (13) can be simplified
that to
EF δz 2n cos ichief
AF ¼ ¼ ; ð4Þ ϕ0 ¼ − m; ð14Þ
cos ∠AFE cos i0 chief n0 − 1

where n0 is the refractive index of the diffractive ele-

AE ¼ EF tan ∠AFE ¼ δz tan i0 chief ;
where δz is the step height between neighboring
subperiods.
By relating Eqs. (4) and (5) with Eq. (3), after some
rearrangement, l1 can be solved:
δz
l1 ¼ − sin i0 chief ðδy þ δz tan i0 chief Þ:
cos i0 chief
With the above discussion, it is quite similar to set
two more right triangles ▵DFG and ▵BGH. As
can be seen from ▵DFG, the relationship between
DG and FG is
DG ¼ FG sin ∠DFG ¼ FG sinð−ichief Þ:
ment, and n ¼ 1:0 for the back diffractive surface in
ð5Þ Fig. 1(a), which means the incident medium is the
air.
Considering the front diffractive surface as shown
in Fig. 1(b) for the second case, the diffractive facets
are sloped away from the incident wavefront as
shown in Fig. 3. A very similar calculation can be
done. It should be noted, however, that the step
height δz is negative in the front diffractive surface
ð6Þ
compared with that in the back diffractive surface,
and the index n should be n0 , which means the inci-
dent medium is the lens material. Then it can be de-
duced that the maximum phase delay in one period
structure for the front diffractive surface is
2n0 cos ichief
ϕ0 ¼ m: ð15Þ
ð7Þ n0 − 1

By substituting DG ¼ l2 − BG and FG ¼ FH− As discussed above, the maximum phase delay for
GH ¼ δy − GH into Eq. (7), we can get oblique incident rays in one period structure of the

2486 APPLIED OPTICS / Vol. 50, No. 16 / 1 June 2011

 face. The reflected energy will be distributed in or-
ders in the vicinity of the nominal order m0 ϕ0. And
the maximum diffraction efficiency is reached in
an integer order m0 max next to m0 ϕ0 .
As a practice example, the narcissus diffraction ef-
ficiencies for a Ge DOE with back and front diffrac-
tive surfaces are calculated using Eqs. (16) and (18).
In Table 1, the diffraction efficiencies for different
orders at different incident angles are given. The cal-
culation is done under the assumption that m ¼ 1,
which is the most widely used design order.
Fig. 3. Rays reflected by two neighboring subperiods of N-level From Table 1, first it can be seen that it is neces-
phase profiles (front diffractive surface). sary for designers to choose those diffraction orders
that have significant reflected energy during the nar-
cissus simulation. Second, the maximum efficiency
diffractive surface is as follows: order will approach to 0 as the increase of incident
( 2 cos i angle, whether it is the back or front diffractive sur-
n0 −1 m
chief
for back diffractive surfaces face. That is because the nominal order m0 ϕ0 will
ϕ0 ¼ 2n0 cos ichief ; approach to 0 with the increase of ichief .
n0 −1 m for front diffractive surfaces
ð16Þ B. Polychromatic Considerations of Narcissus for a
Diffractive Surface
In [5,7] m0 max is used to stand for the order of max- For a diffractive surface, the maximum diffraction
imum efficiency: efficiency is usually achieved just at the design
wavelength, and the diffraction efficiency will devi-


− n02−1 m for back diffractive surfaces ate with the shift of the wavelength. This also hap-
m0 max ¼ 2n0 : pens to narcissus evaluation for a diffractive surface
n0 −1 m for front diffractive surfaces
over a wide spectral band.
ð17Þ As for wavelength λ, Eq. (13) should be changed
into
Equation (17) is, in fact, the simplification of
ichief ¼ 0°, which is the case of vertical incident. How- 2Nnλ δz cos ichief
ϕλ ¼ Nϕs ðλÞ ¼ − ; ð19Þ
ever, the definition of m0 max as the order of maximum λ
diffraction efficiency is not accurate or appropriate.
where ϕλ is the maximum phase delay in waves in
In [8], the diffraction efficiency of the m0 th
one period structure for wavelength λ.
diffraction order for an infinite number of phase
Combining d ¼ Nδz ¼ mλ0 =ðn0 − 1Þ with Eq. (19),
levels is when nλ is 1.0 for the back diffractive surface, we
obtain
η∞ 0
m0 ¼ sinc ½m − ϕ0
:
2 ð18Þ
2λ0 cos ichief
ϕλ ¼ − m: ð20Þ
In the scalar diffraction approximation, Eq. (18) is ðn0 − 1Þλ
valid at the condition that the diffraction order m0 is
an integer and the 100% diffraction efficiency can Considering the front diffractive surface, ϕλ can be
only be reached when ϕ0 is an integer and m0 − determined in a very similar manner:
ϕ0 ¼ 0. In the case of narcissus for a diffractive sur-
face, ϕ0 is not usually an integer, and, according to 2nλ λ0 cos ichief
ϕλ ¼ m; ð21Þ
Eq. (16), ϕ0 varies with the change of incident angle. ðn0 − 1Þλ
Thus, it is more appropriate to define a nominal
order m0 ϕ0, which equals to ϕ0 , the maximum phase where nλ is the refractive index of the lens material
delay in one period structure of the diffractive sur- at wavelength λ for the front diffractive surface.

Table 1. Efficiencies ηm 0 for Different Diffraction Orders at Various Incident Angles for a First-Order DOE on Ge

Diffraction Order for Back Diffractive Surface Diffraction Order for Front Diffractive Surface
Incident Angles −1 0 1 2 0 1 2 3
0° 0.685 0.171 0.027 0.011 0.011 0.027 0.171 0.685
20° 0.619 0.219 0.033 0.013 0.016 0.045 0.396 0.415
40° 0.424 0.388 0.044 0.016 0.000 0.002 0.994 0.002
60° 0.172 0.683 0.043 0.014 0.043 0.683 0.172 0.028
80° 0.017 0.957 0.010 0.003 0.467 0.347 0.042 0.016

1 June 2011 / Vol. 50, No. 16 / APPLIED OPTICS 2487

 Thus, the nominal order m0 ϕλ, which equals to ϕλ as
we discussed above, is as follows:

calculation of the radius of the “cold spot” for the cen-
− 2λðn
0 cos ichief
0 −1Þλ
m for back diffractive surfaces
m0 ϕλ ¼ 2nλ λ0 cos ichief
ðn0 −1Þλ m for front diffractive surfaces
ð22Þ
From Eq. (22) we can see that nominal order m0 ϕλ
for a back-tracing narcissus ray is a function of both
the incident angle of chief ray ichief and wavelength λ.
Thus, the maximum diffraction order m0 max ðλÞ, which
is an integer next to m0 ϕλ , may also be changed for
different wavelengths. The wavelength dependence
of these diffractive surfaces should be considered
for the calculation of the narcissus effect in a given
wavelength band.
3. On-Axis Paraxial Narcissus Intensity Evaluation for
a Diffractive Surface
To evaluate the intensity of narcissus for optical sur-
faces in an IR system, the radius of the circular cold
ghost area should be calculated for on-axis detector
pixels. A simplification for the narcissus evaluation
is the paraxial value yni, which is deduced with the
calculation of the on-axis cold ghost area. The yni val-
ue is often controlled in optical design software to be
greater than unity to defocus the return narcissus
rays. The paraxial ray tracing in optical design soft-
ware assumes that all desired surface power is
defined by the surface curvature and does not include
Fig. 4. Narcissus effect for diffractive surfaces.
2488 APPLIED OPTICS / Vol. 50, No. 16 / 1 June 2011
the effect of diffraction. With the introduction of a dif-
fractive surface, the calculation of yni should be
modified. In this paper, the modification is done by
: tral detector pixel using the Lagrange invariant. As
is shown in Fig. 4, one marginal ray is traced back-
ward from the center of the detector to a lens com-
posed of a diffractive surface. Here, the diffraction
effect has been taken into account to calculate the
propagation, and the ray with the maximum diffrac-
tion efficiency is traced back from the diffractive
surface to the focal plane. Then a pair of rays for
the Lagrange invariant is defined at the diffractive
surface.
As for the analysis of diffractive angles, we can
treat the diffractive surface here as an N-step blazed
grating, as shown in Fig. 5. The discussion focuses on
the design wavelength λ0 for simplicity and the chief
ray incident angle ichief ¼ 0°. Then the nominal order
m0 ϕ0 can be simplified to


− n02−1 m for back diffractive surfaces
m0 ϕ0 ¼ 2n0 : ð23Þ
n0 −1 m for front diffractive surfaces
The incident wavefront undergoes some phase
delay or advance of m0 ϕ0 λ (see Fig. 5), depending
on the sign of m0 ϕ0 . From what we discussed in
Subsections 2.A and 2.B, the diffraction order
m0 max , which we are most interested in, lies in the
vicinity of m0 ϕ0 . As N approaches infinity, θ, the dif-
 m0 max λ0
θ¼− − iB : ð27Þ
nB T

Substituting Eq. (27) into Eq. (26) gives

ym0 max λ0
I ¼ 2ynB iB þ : ð28Þ
T
At the central detector of the image plane (see
Fig. 4), the Lagrange invariant is as follows:

y0
I ¼ y0 n0 u0 ¼ − ; ð29Þ
2f #

where y0 is the height of the diffracted marginal ray

Fig. 5. Reflection of diffractive surfaces as N-step blazed
gratings.
fractive angle of the backward-traced marginal ray
for the m0 max th diffraction order, can be given by
using the grating equation:
for the m0 max th diffraction order at the detector plane
and is equal to the radius of the circular cold ghost
area, u0 is the angle of the forward-traced marginal
ray at the detector plane, n0 is the index of refraction
at the detector plane, and f # is the system’s f -
number.
By setting Eq. (28) equal to Eq. (29), the radius of
the circular cold spot y0 can be expressed as
2f #ym0 max λ0
y0 ¼ −4ynB iB f # − : ð30Þ
T

m0 max λ0 It should be mentioned that the yni value calcu-

sin θ ¼ − − sin iB ; ð24Þ lated by optical design software is always the product
nB T
of the paraxial marginal ray height times the index
times the angle of incidence for the forward-traced
where iB is the incident angle of the backward-traced
marginal incident ray. As for the refractive surface,
marginal ray.
It should be pointed out that nB ¼ 1:0 for the back
nF sin iF ¼ nB sin iB ; ð31Þ
diffractive surface and nB ¼ n0 for the front diffrac-
tive surface. where nF and iF are the index and the incident
At the diffractive surface, the Lagrange invariant
angle for the forward-traced marginal incident ray
is as follows:
as shown is Fig. 4. However, Snell’s law is not applic-
able for a diffractive surface. The diffraction property
I ¼ yd nB uB − ynB ud ; ð25Þ of the forward-traced marginal ray could be de-
scribed as
where I is the Lagrange invariant, yd is the height of
the diffracted marginal ray at the diffractive surface, mλ0
ud is the angle of the diffracted marginal ray for the nF sin iF − ¼ nB sin iB ; ð32Þ
T
m0 max th diffraction order, y is the height of the back-
ward-traced marginal ray at the diffractive surface, which means the marginal beam undergoes a phase
and uB is the angle of the backward-traced marginal delay of mλ0 in one period structure dimension T. For
ray from the central detector pixel. paraxial simplification, Eq. (32) can be changed to
Since yd ¼ y and uB − ud ¼ iB − θ, by substituting
into Eq. (25), we get mλ0
n F iF − ¼ n B iB : ð33Þ
T
I ¼ ynB ½iB − θ
; ð26Þ Relating Eq. (33) to Eq. (30) results in the following
expression:
where iB is the incident angle of the backward-traced
marginal ray, which is also the diffraction angle of 2f #yλ0
the forward-traced marginal ray. y0 ¼ −4ynF iF f # þ ð2m − m0 max Þ: ð34Þ
T
Using Eq. (24) and substituting the angle itself for
its sine value in this equation for paraxial simplifica- The phase function Φ for diffractive surfaces is
tion, we get usually expressed as

1 June 2011 / Vol. 50, No. 16 / APPLIED OPTICS 2489

 Fig. 6. (Color online) Layout of an 8–12 μm IR system with a cooled detector.

2π y0
Φ¼ ða y2 þ a2 y4 þ a3 y6 þ …Þ; ð35Þ − ¼ ynF iF þ N D ; ð39Þ
λ0 1 4f #

where a1 ; a2 ; a3 … are phase coefficients. where

From Eq. (35), the vertical spacing distance of one  
period of structure for diffractive surfaces designed m0 max
ND ¼ 2− ða1 y2 þ 2a2 y4 þ 3a3 y6 þ …Þ: ð40Þ
to operate at mth order can be deduced as m
 −1 Equations (39) and (40) could be used effectively as
1 dΦ
T¼ − an evaluation criterion for the intensity of the narcis-
2mπ dy sus influence caused by the diffractive surface. After
 −1
1 the maximum diffraction order m0 max is determined
¼ − ð2a1 y þ 4a2 y3 þ 6a3 y5 þ …Þ : ð36Þ with the discussion in Section 2, optical designers
mλ0
could incorporate Eqs. (39) and (40) into optical de-
sign software to control the narcissus influence of
Consequently, the size of the detector ghost areas
the diffractive surface.
can be obtained by using Eqs. (34) and (36) and is
As a practical example of the narcissus influence
given by caused by diffractive surfaces, an 8–12 μm IR system
with a cooled detector sealed by a window is de-
y0 ¼ −4ynF iF f # signed. As illustrated in Fig. 6, the optical system
  is designed with a front diffractive surface and a back
m0 max
− 4f # 2 − ða1 y2 þ 2a2 y4 þ 3a3 y6 þ …Þ: diffractive surface at the first lens and the fourth
m lens, separately. The surface types of the two diffrac-
ð37Þ tive surfaces are set as binary 2 in ZEMAX, and the
shape of the binary 2 surfaces are identical to the
As for the refractive surface, the ghost radius y0 is even asphere surface type, with the surface sag being
divided by −4f # to get the value of ynF iF , which is expressed by
often used as a criterion to indicate the strength of
narcissus for IR optics. By dividing both sides of cy2
z¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi þ α1 y2 þ α2 y4 þ α3 y6 þ …;
Eq. (37) by −4f #, the evaluation criterion of narcissus 1 þ 1 − ð1 þ kÞc2 y2
for diffractive surfaces, however, should be changed ð41Þ
to
where c is the curvature (the reciprocal of the radius),
y0 k is the conic constant, and α1 ; α2 ; α3 … are even as-
− ¼ ynF iF phere coefficients.
4f #
  The substrate material for lens 1 and lens 4 is
m0 max ZnSe and both of the elements are first-order DOEs.
þ 2− ða1 y2 þ 2a2 y4 þ 3a3 y6 þ …Þ:
m Tables 2–5 list the first-order parameters and detail
ð38Þ
Table 2. First-Order Parameters for the 8–12 μm
It can be inferred from Eq. (38) that, except from IR Optical System

paraxial parameters y, nF , and iF , the diffraction Entrance pupil diameter 47 mm

terms have influence on the narcissus effect. Cold stop diameter 7:5 mm
Setting the diffraction part of the formulas at the System focal length 127 mm
right-hand side of Eq. (38) to be N D , indicating the FPA dimensions 9:84 mm × 7:38 mm
influence of the narcissus effect caused by diffraction, FPA pixel pitch 30 μm × 30 μm
Cold stop f # 2.75
we can get

2490 APPLIED OPTICS / Vol. 50, No. 16 / 1 June 2011

 Table 3. Lens Data of the 8–12 μm IR Optical System: Surface Table 8. Paraxial Narcissus Values and Modified Narcissus Values of
Data Summary the IR Optical System

No. Type Radius Thickness Glass No. Type ynF iF y ynF iF þ N D ND

0 Standard Infinity Infinity Air 1 Refractive 4.435 23.561 — —
1 Standard 125.1762 8 ZnSe 2 Diffractive −4:572 22.681 −4:896 −0:324
2 Binary 2 864.4047 120 Air 3 Refractive −0:133 −5:892 — —
3 Standard −22:59946 6 Germanium 4 Refractive −1:381 −7:423 — —
4 Standard −24:61636 17 Air 5 Refractive 3.427 −9:387 — —
5 Standard 37.60503 6 Germanium 6 Refractive 0.319 −8:436 — —
6 Standard 50.25131 1 Air 7 Diffractive 2.470 −8:306 2.496 0.026
7 Binary 2 19.43353 7 ZnSe 8 Refractive 1.379 −6:082 — —
8 Standard 14.76969 9.319 Air
9 Standard Infinity 2.5 Germanium
10 Standard Infinity 3 — y
Stop Standard Infinity 20 — ρ¼ ; ð43Þ
Image Standard Infinity — —
RNorm

where RNorm is the normalized radius and here we

Table 4. Lens Data of the 8–12 μm IR Optical System: Even set RNorm ¼ 100. Thus, the relationship between
Asphere Coefficients the two types of phase coefficients is as follows:
No. Type α1 α2 8
>
> a1 ¼ 2πRλ20 A1
2 Binary 2 −3:549789 × 10−6 8:861466 × 10−8 >
< Norm
7 Binary 2 1:206406 × 10−3 −1:610080 × 10−7 a2 ¼ 2πRλ40 A2
: ð44Þ
>
>
Norm
λ0
> a ¼
: 3 2πR6Norm 4 A
Table 5. Lens Data of the 8–12 μm IR Optical System: Phase
…
Coefficients of Diffractive Surfaces in ZEMAX
Table 6 lists the converted phase coefficients a1, a2 ,
No. Type RNorm A1 A2 a3 of the two diffractive surfaces.
2 Binary 2 100 −1394:3655 1318.0636 Table 7 illustrates the diffraction efficiencies ηm0
7 Binary 2 100 −1898:1135 −46898:928 for different diffraction orders calculated using
Eqs. (16) and (18) with ichief ¼ 0° for the case of on-
axis analysis. From Table 7, we could tell that the
Table 6. Converted Phase Coefficients of Diffractive Surfaces m0 max th diffraction order for the back diffractive sur-
No. Type a1 a2 a3
face of the first lens is −1, and that m0 max ¼ 3 for the
front diffractive surface of the fourth lens. Using the
−4 −8
2 Diffractive −2:2192 × 10 2:0978 × 10 −1:1933 × 10−11 analysis functions of ZEMAX, the paraxial ynF iF and
surface
the value of y can be given. The parameters are listed
7 Diffractive −3:0209 × 10−4 −7:4642 × 10−7 2:3420 × 10−9
surface in Table 8. By combining Eqs. (39) and (40) and the
parameters in Tables 6–8, the modified evaluation
criterion of the narcissus intensity for the two diffrac-
the lens data of the optical system. It should be men- tive surfaces can be given. The values of ynF iF þ N D
tioned that the binary 2 surface in ZEMAX adds op- and N D for the diffractive surfaces, S2 and S7 in
Fig. 6, are listed in the last two columns of Table 8.
tical phase to the ray according to the following
Compared with the evaluation criterion of the narcis-
polynomial:
sus intensity, which is often controlled to be greater
than unity to defocus the return narcissus rays, the
Φ ¼ A1 ρ2 þ A2 ρ4 þ A3 ρ6 þ …; ð42Þ magnitude of the N D for S2 obviously cannot be ig-
nored; meanwhile, the N D for S7 is relatively small.
That is because the paraxial value of y for S2 is big-
where A1 , A2 , A3 … are the phase coefficients of ρ, ger than that of S7, and the exponential function ex-
which is the normalized radial aperture coordinate aggerates the difference. The phase plot and the line
and frequency versus aperture for the two diffractive

Table 7. Diffraction Efficiencies ηm 0 for Different Diffraction Orders at the Back and Front Diffractive Surfaces of the Two ZnSe Lenses

Diffraction Order for Back Diffractive Diffraction Order for Front Diffractive
Surface of The First Lens Surface of The Fourth Lens
Diffraction
Efficiency −3 −2 −1 0 2 3 4 5
ηm0 0.039 0.295 0.525 0.047 0.047 0.525 0.295 0.039

1 June 2011 / Vol. 50, No. 16 / APPLIED OPTICS 2491


Fig. 7. (Color online) Phase plot and the line frequency versus
aperture for (a) the back diffractive surface and (b) the front dif-
fractive surface.
surfaces are shown in Fig. 7. In Fig. 7, the downward
sloping curve is the phase plot in periods and the up-
ward sloping curve is the line frequency, with periods
per millimeter as the unit. From Fig. 7, we can see
that the line frequencies at the maximum aperture
of the two diffractive surfaces are more or less
1 period=mm. As the development of optical proces-
sing for diffractive surfaces, the line frequency could
reach 5–10 periods=mm, which may lead to even
more severe diffraction influence. Thus it is crucial
to take the diffractive effect into account in the pro-
cess of narcissus evaluation for IR optical systems.
4. Conclusion
The diffraction efficiency for backward-traced narcis-
sus flux and the paraxial evaluation for on-axis
narcissus intensity, considering the influence of dif-
fraction for diffractive surfaces, are discussed. The
2492 APPLIED OPTICS / Vol. 50, No. 16 / 1 June 2011
analysis in Section 2 showed that the order of max-
imum diffraction efficiency varies when the wave-
length and incident angle of the chief ray of the
narcissus flux are changed. The result could be used
for accurate calculation of narcissus intensity by
using backward ray tracing. In Section 3, extension
formulas, compared with traditional refractive nar-
cissus evaluation criterion yni, have also been de-
duced to evaluate the narcissus intensity caused
by diffractive surfaces. A practical IR optical system
with two diffractive surfaces has been designed to il-
lustrate the influence of diffraction for the intensity
of narcissus. It has been pointed out in this work that
the impact of diffractive surfaces should be consid-
ered in the optical design of IR systems with cooled
detectors. The discussion in Section 3 is based on
paraxial ray tracing and did not take into account
ray clipping by real apertures. The new narcissus
evaluation criterion, however, could be used as a
quick on-axis narcissus intensity estimate for diffrac-
tive surfaces, and the results could be considered for
the control of narcissus intensity in optical design
software.
This work is supported by the National Natural
Science Foundation of China (NSFC, Grant No.
60878030).
References
1. J. W. Howard and I. R. Abel, “Narcissus: reflections on retro-
reflections in thermal imaging systems,” Appl. Opt. 21,
3393–3397 (1982).
2. J. M. Lloyd, Thermal Imaging Systems (Plenum, 1975),
pp. 275–281.
3. K. Lu and S. J. Dobson, “Accurate calculation of narcissus
signatures by using finite ray tracing,” Appl. Opt. 36,
6393–6398 (1997).
4. J. L. Rayces and L. Lebich, “Exact ray tracing computation of
narcissus equivalent temperature difference in scanning ther-
mal imagers,” Proc. SPIE 1752, 325–332 (1992).
5. M. N. Akram, “Simulation and control of narcissus phenom-
enon using nonsequential ray tracing. I. Staring camera in
3–5 μm waveband,” Appl. Opt. 49, 964–975 (2010).
6. ZEMAX User Manual, Zemax Development Corporation, 3001
112th Avenue NE, Suite 202, Bellevue, Wash. 98004-8017,
USA, 2008.
7. J. B. Cohen, “Narcissus of diffractive optical surfaces,” Proc.
SPIE 2426, 380–385 (1995).
8. G. J. Swanson, “Binary optics technology: theoretical limits
on the diffraction efficiency of multilevel diffractive optical
elements,” Tech. Rep. 914, MIT Lincoln Laboratory, 1991.