Tunable High-Finesse Narrow Bandpass Fabry - Perot Filter: V.B. Markov, A.I. Khizhnyak, V. Goren, W.B. Cook
Tunable High-Finesse Narrow Bandpass Fabry - Perot Filter: V.B. Markov, A.I. Khizhnyak, V. Goren, W.B. Cook
Tunable High-Finesse Narrow Bandpass Fabry - Perot Filter: V.B. Markov, A.I. Khizhnyak, V. Goren, W.B. Cook
PACS: 42.79. Ci
Abstract. This paper discusses the results of the analysis and experimental
characterization of a narrow bandpass optical filter based on the Fabry – Perot
interferometer configuration with a variable spacing between the mirrors allowing for a
relatively wide spectral tunability. Such a filter, with a high-throughput bandpass and
sufficiently large aperture and acceptance angle, is of practical interest for a high-
resolution spectrometry and remote sensing in the visible and infrared spectral regions.
The Fabry – Perot filter (FPF) can be designed in a compact single-assembly
architecture that can be accommodated within existing instruments and should provide a
stable performance under variable thermal and mechanical conditions, including space
and airborne platforms. Possible applications of the filter include high-resolution multi-
spectral imaging, terrain mapping, atmosphere and surface parameters measurements,
and detection of chemical and biological agents.
465
Semiconductor Physics, Quantum Electronics & Optoelectronics. 2004. V. 7, N 4. P. 465-473.
0
3.1. FPF transmissivity in terms of the mirror
0 0.1 0.2 0.3 0.4 0.5 characteristics
/
A1 T1
The solution of the wave equation in the homogeneous
Fig. 3. The filter maximum transmissivity T0 exceeds 50% medium of the spacer permits to express the FPF
level in the shaded area. transmissivity in as [9]:
466
Semiconductor Physics, Quantum Electronics & Optoelectronics. 2004. V. 7, N 4. P. 465-473.
[ ] −1
X-Y-Z controller
T = T0 1 + F sin 2 (ϕT )
HLD1 HLD2
Beamsplitter D1
, (5)
D2
PD-1 Pin-hole
where maximum transmissivity T0 and the F-factor can
be expressed in terms of the transmissivity (T1, T2) and
D3
PD-2
reflectivity (R1, R2) of the mirrors M1 and M2, as follows:
(
T0 = T1T2 1 − R1 R2 ),2
(6)
(1 − ),
Diode laser
CCD-camera 2
Chopper F = 4 R1 R2 R1 R2 (7)
and ϕT is the total interference phase:
He-Ne laser
ϕT = k nS LS − ϕ M . (8)
Fig. 4. Schematic of the experimental set-up.
Here k nS LS is the optical length of the spacer (LS is
the mirror spacing thickness), and the phase shift ϕM is
the median phase of the reflection coefficients of the two
mirrors, determined by the structure of the multi-layer
mirrors (Fig. 2):
ϕ M = (ϕr1 + ϕr1 ) / 2 , ϕrj = arg(r j ) , ϕ M ≤ π , (9)
Eqs (6) - (9) can be derived in a consistent way [17]
using the transfer matrix approach [18]. The phase shift
ϕr upon reflection from a dielectric multi-layer mirror
has nearly linear dependence on the wavevector k near
the center kc of the reflection band [15-17],
ϕ r (k ) ≅ −bπ(k − k c ) / k c , for k – kc << kc (10)
a b with the coefficient b ~ 5 > 0 at the realistic values of the
parameters of the mirror. Therefore, the total
interference phase shift ϕT in Eq. (8) is also a linear
function of the wavevector,
ϕT (k ) = kΛ S − bπ ; b = (b1 + b2 ) / 2 , (11)
Fig. 5. PC screen showing the photodiode PD output signal
where
(upper trace) and the Michelson interferometer transmittance
(lower trace) vs. incremental (a) and decremented (b) voltage Λ S = nS LS + bλ c / 2
applied to the piezo actuator. is an effective optical length of the cavity spacer for
central wavelength λc (Fig. 1). The value of ΛS includes
the contribution of bλc / 2 that can be interpreted here as
a mirror penetration length [15].
Based on the grating-mirror model with the
coupled-wave approach [19], we have shown earlier [17]
that linear dependence of the phase change ϕr on the
wavevector (Eq. (10)) holds within the mid-band range
up to a half-width Δb k of the reflection band of the
multi-layer dielectric mirror,
k − kc ≤ 0.5Δ b k ≅ 0.2(Δn / n )kc (12)
where Δn = nH − nL , and n = (nH + nL ) / 2 are the
difference and median of the refractive indices of the
mirror layers (see Fig. 2), while the band-center
wavevector is kc = π/( n lM) and lM is the layer-pair
thickness. The phase-shift coefficient b is
b ≅ (nS / n ) ±1 (n / Δn) tanh( NΔn / n )
for nS ≤ n , (13)
Fig. 6. Experimental dependence of the light intensity on the where the sign (±) in the power depends on the value of
filter output with incremental voltage applied to piezo actuator. the front layer refractive index, and is (+) for nH and (–)
467
Semiconductor Physics, Quantum Electronics & Optoelectronics. 2004. V. 7, N 4. P. 465-473.
for nL, respectively. N is the number of the layer pairs, The same ratio (regardless of the bandpass number m)
and we assume that the refractive index of the substrate holds between the bandpass spectral width (Δpk) and free
is matched to the median refractive index n in the range (Eq. (15)) in terms of the wavevector,
multi-layer coating. (If nS > n , then nS and n have to Δ pk = Δ f k f , (20)
be switched in the first two factors in Eq. (13)).
and between their relative values,
3.2. Bandpass spectral position, free range, and tuning δ p = Δ p k / k c = δ f / f = 1 (mc + b) f , (21)
range
where mc is the number of bandpass as defined by
It follows from Eq. (5) and (11) that the spectral Eq. (17).
positions km of the bandpass maxima are: It follows from this analysis that if δ t = 0.01 and
k m = (m + b)π Λ S , m = 1, 2,K (14) δ p = 0.5 ⋅ 10 −4 are of interest, then the finesse should to
Therefore, regardless the bandpass number m, the be at least f = 200.
bandpass free range in terms of the wavevector is
3.4. Control of the bandpass spectral position
Δ f k = π/ ΛS . (15)
Because the free range defined by Eq.(15) has to be For applications in a high-resolution spectroscopic
larger than the required tuning range, Δf k ≥ Δt k, the imaging, the FPF bandpass peak position (km or
spacer length should be small enough and should satisfy λm = 2π / km) has to be controlled with high precision. In
the following condition this analysis we assume that the precision control should
be better than one-fifth of the bandpass width (Δp k or
Λ S ≤ δt−1λ c / 2 , (16)
Δp λ, see Fig. 1). Such a requirement restricts the
where δt = Δt k / kc is the relative value of the required tolerated relative error to (see Eq. (21)):
tuning range (δt ≅ Δt λm / λc for λm ≈ λc). If a relatively ~ ~
Δλ m / λ c ≅ Δkm / kc ≤ δ p / 5 , (22)
wide tuning range is of interest (δt ≥ 0.02), then the
phase-shift contribution (with b ~ 5) to the spacer that is ~ 10–5 for the narrow bandpass filter.
effective length ΛS is significant for determining the The spectral position of the bandpass peak is
relation of the free range (Eq. (15)) to the mirror spacing determined by the spacer optical length nSLS that can be
LS. The value of LS required to position the bandpass tuned by varying either the mirror spacing LS or the
peak at the center of the reflection band can be found spacer refractive index nS. It follows from Eq. (14) that
from Eq. (14): the peak wavelength is a linear function of the spacer
optical length,
nS LS = mc λ c / 2 (17)
λ m = (2nS LS + bλ c ) (m + b) . (23)
where mc is the number of this bandpass. According to
Therefore, both the mirror spacing LS and the refractive
the definition of ΛS, the corresponding relative value of index nS of the spacer have to be controlled with the
the free range (Eq. (15)) is related to this bandpass same relative precision (Eq. (22))
number as ~ ~
δ f = Δ f k / kc = 1 (mc + b) . (18) ΔLS / LS + ΔnS / nS ≤ δ p / 5 . (24)
Therefore (see Eqs (17) and (21)), the following absolute
To meet the requirement of Eq. (16) this bandpass
precision is required for the mirror spacing LS :
number should satisfy the condition mc ≤ δt–1 – b, so the
mirror spacing should be small according to Eq. (17). (25)
For a typical requirement of δt ~ 0.01, this yields the ~
that is ΔLS ≤ 0.3 nm for λ c ≅ 600 nm with f = 200.
following restrictions: mc ≤ 90 and nSLS ≤ 45λS. Such a high precision can be achieved using piezo-
electric actuators with closed-loop control [16].
3.3. Bandpass spectral width Due to varying the spacer optical length the tuning
coefficient for the bandpass peak wavelength Eq. (23) is
According to Eq. (5) for the FPF transmissivity, the
bandpass width (at half-maximum) in terms of the Ct = Δλ m / Δ (nS LS ) = 2 /( m + b) . (26)
interference phase shift ϕT is equal to 2 F at F >> 1, To shift the bandpass over its entire free range from the
band center to an adjacent bandpass, the spacer optical
while the phase spacing between the adjacent bandpass
length, according to Eq. (17), should be changed by
peaks is equal to π. Thus, the phase shift ratio of the
Δf (nsLs) ≅ λc / 2. One can see that the influence of the
bandpass spacing over the width (the interferometer
phase-shift on the bandpass spectral position is
finesse [9]) is
significant for the precise tuning of a narrow bandpass
f = (π / 2) F . (19) within a wide free spectral range. According to Eqs (21)
468
Semiconductor Physics, Quantum Electronics & Optoelectronics. 2004. V. 7, N 4. P. 465-473.
a) b)
1
0.8
1 2 3
0.6
0.4
0.2
0
0 5 10 15 20
Voltage, V
Fig. 7. Normalized intensity of transmitted light vs. voltage applied to piezo actuator. Solid line shows FPF transmission for
diaphragm D1 of 4 mm diameter, and broken line is for transmission through total aperture. Figure (b) shows the details of the
first peak in (a).
1 2 3 4
5 6 7 8
Fig. 8. Light distribution at the filter output when illuminated by the expanded laser beam (the diaphragm D1 in
Fig. 4 is open). The circled area is for the output beam selected through the diaphragm D2 (Fig. 4). The numbers
on the frames correspond to the numbers in Fig. 7 (b).
This is certainly the case for δ f ≥ 0.01 and δ p ≤ 10 −4 It follows from Section 3 that in order to attain a high
finesse (f ≥ 200) needed to build a narrow-bandpass
with b ≥ 1 . tunable FPF, it is essential that the reflectivity of the
469
Semiconductor Physics, Quantum Electronics & Optoelectronics. 2004. V. 7, N 4. P. 465-473.
mirrors should be close enough to 100 %. According to so that the filter background rejection level is determined
Eqs (7) and (19), entirely by its finesse f, and for f ≥ 200 , the background
(
f = π R 1− R , ) (27) rejection is T0 / Tb ≥ 1.6 ⋅ 10 −4 ≥ 40 dB .
where R = R1R2 is an average reflectivity of the 4.2. Tolerance to the environmental variations
mirror pair. Therefore, for achieving the required filter
finesse f >> 1, the reflectivity R has to satisfy the In order to control the peak wavelength position of the
following condition narrow bandpass (δp ≅ 0.5·10–4) with the precision
defined by the requirements from Eq. (22), according to
R ≅ 1− π/ f , (28)
Eq. (24) the uncontrolled variations of the spacer
resulting in R ≅ 0.984 for f = 200. Note, that the refractive index nS have to be limited as
~
relative error for the finesse (Eq. (27)) and, therefore, for ΔnS / nS ≤ δ p / 5 ≅ 10−5 . (34)
the bandpass width (Eq. (21)) is the same for (1 − R ) .
In its most common design, the FPF has an air
Following from Eq. (6) for the maximum
spacer with nS = na, where na is the refractive index of
transmissivity T0, an average transmissivity
air. The value of na is close to 1.0 and varies upon the
T = T1T2 of the mirror pair is determined by T0 and R temperature t, atmospheric pressure P, and relative
(or f ) as: humidity H [20]. It is convenient to analyze the
variations of na in terms of the air refractivity ηa,
T = (1 − R ) T0 ≅ π T0 f . (29)
ηa = na − 1 . (35)
Thus, it follows from Eq. (29) that in order to achieve
the maximum transmissivity T0 = 50 % with the finesse The requirement of Eq. (34) yields the limitation on the
absolute variations of the air refractivity (Eq. (35)),
f = 200, an average transmissivity T should be
T ≅ 0.011. ~
Δηa ≤ δ p / 5 ≅ 10 −5 . (36)
Maximum transmissivity T0 of the filter with a pair
of identical lossless mirrors would be as high as 100 %. At normal conditions, t0 = 15 °C, P0 = 1000 mb and
However, due to non-zero absorbance (A > 0) of the H0 = 50 %, the air refractivity is ηa 0 ≅ 27.2 ⋅ 10 −5 at
mirror structure, as follows from to the energy balance,
λ = 680 nm. For small deviations of the environmental
the magnitude of T0 (in Eq. (6)) always remains lower
variables t, P, and H from normal conditions the
than 100 %,
dependence of the deviation of ηa from ηa 0 on these
Ti = (1 − Ri ) − Ai , (i = 1, 2). (30)
variables is practically linear [20]:
In order to satisfy the requirement of T0 ≥ 50 %, the
absorbance of the mirror has to be low, ηa − ηa 0 ≅ 0.097(t − t0 ) ⋅ 10 −5 , (37)
−5
Ai / Ti ≤ 0.4 or Ai / (1–Ri) ≤ 0.3. (31) ηa − ηa 0 ≅ 0.027( P − P0 ) ⋅10 , (38)
−2 −5
In addition, the maximum transmissivity T0 (Eq. (6)) ηa − ηa 0 ≅ 0.075 ⋅ 10 ( H − H 0 ) ⋅ 10 . (39)
of the filter is reduced if the transmission characteristics
of the two mirrors differ. Indeed, even without Following from Eqs (37) - (39), the variations of the air
absorption, T < 1 − R if R2 ≠ R1. According to Eq. parameters (near normal conditions) allows to tolerate
(30) maximum transmissivity T0 of the FPF can the requirement of Eq. (36) within the following limits:
~ ~ ~
be approximated by the following function of the Δt ≤ 10 °C , ΔP ≤ 37 mb , ΔH ≤ 100 % . (40)
relative absorbance A1 /T1 and reflectivity difference
ΔR / T : 4.3. Angular acceptance of the FPF
T0 ≅ (1 − ΔR / T1 ) (1 + A1 / T1 − 0.5ΔR / T1 ) ,2
(32) For off-axis incidence (at an angle θ ≠ 0), Eq. (8) for the
where we denoted ΔR = R2 – R1 > 0 and made a interference phase can be modified by replacing the
practically reasonable assumption |A2 - A1| << ΔR << R1. absolute value of the wavevector by its z-component [9].
The graph of Eq. (32) with T0 = 0.5 in the plane of Therefore, the bandpass peak spectral position kmθ is
variables A1 / T1 and ΔR / T1 confines the area where the related to its value at normal incidence as:
maximum transitivity is higher than 50 % (see Fig. 3). k mθ cos θ = k m . (41)
According to Eqs (5) and (19), the background
transmissivity Tb can be expressed as: The acceptance angle θa for a spectroscopic imaging
FPF can be defined as an angle of incidence at which the
[
Tb = T0 1 + (2 f / π) 2 ]
−1
, (33) bandpass peak shifts from its normal-incidence position
by one-fifth of the bandpass width (in accordance with
470
Semiconductor Physics, Quantum Electronics & Optoelectronics. 2004. V. 7, N 4. P. 465-473.
471
Semiconductor Physics, Quantum Electronics & Optoelectronics. 2004. V. 7, N 4. P. 465-473.
transmitted beam. It shows a non-uniform intensity displacement Δ⊥ of the beam across the window of the
distribution with different maximum position for FPF is Δ⊥ = 0.38 mm. Thus, the value of Δ⊥ is one-tenth
different values of the voltage applied to the of the selected beam diameter at the detector. These
piezoactuator. It follows from this data that the distance measurements suggest that for the conditions of the
between the mirrors of the filter is a function due to the experiments, the non-flatness of the mirrors doesn’t have
position along the mirror surface, which would be a significant effect on the results obtained.
expected when the mirrors are non-planar.
In finding the shift of the mirror position for 5.5 Broadband light source (diode laser)
changing from the maximum transmission in the filter characterization
center to the maximum transmission at the boundary of
the filter, we conclude that the mirrors are concave and In this set of the experiments, a broadband light source
estimate the difference of the mirror spacing between was used for FPF characterization. The diode laser
peripheries and the center as ≈ 9 nm, so the non-flatness operating in a multi-frequency regime at λ = 685 nm
of the mirrors across the one-inch diameter section is served for this purpose. Fig. 10 shows that the FPF
approximately λ/70 for λ = 632.8 nm. resolution is higher than the bandwidth of the diode-laser
light. It is also evident from this figure that the distance
5.4 Acceptance angle of the filter between the corresponding peaks in the transmission
function is different due to the wavelength difference
The acceptance angle of the FPF was defined and between the He-Ne and diode lasers. If the displacement
measured as the FPF transmissivity at different incident of the mirror is a linear function of the voltage applied to
angles of the incoming collimated laser beam. The the piezoactuator, the simple relation between two
intensity of the transmitted beam at its central part was transmission wavelengths holds:
measured for several values of the incident angle. The V (λ1 ) λ1
= , (44)
V (λ 2 ) λ 2
size of the diaphragm in front of the photodetector was
selected to ensure minimal bandwidth at normal
incidence. Fig. 9 illustrates a typical dependence of the where V(λn) is the voltage variation for tuning between
detected signal upon angular deviation of the FPF, with two transmission maxima at the wavelengths λn. Our
every experimental point on the graph corresponding to estimations satisfy Eq. (44) with an error rate of less
an average value of multiple measurements, and with the than 5 %.
solid line calculated from the expression Fig. 10 illustrates the difference in the structure of
1 the spectrum for two orders of the FPF transmission.
T= (43)
1 + (2 f / π ) sin 2 (π mc cos θ)
2 This is mostly due to the instability of the oscillating
spectrum of the diode laser, what is typical when such a
that describes the transmissivity of the FPF as the laser operates without stabilization of its characteristics.
function of the incident angle for the bandpass at the
center λc of the mirror reflection band (mc is this
bandpass number, see Eq. (17)). 6. Conclusions
The following values were used for calculating the
transmissivity as the function of the incident angle:
φ = 70, mc = 2LS / λc = 500. The value of φ was In this paper, we present the results of the analysis and
estimated based upon the free range and spectral width experimental characterization of a narrow bandpass
of the bandpass at normal incidence of the laser beam, tunable Fabry – Perot filter. It follows from the
and the value of the parameter mc follows from the performed analysis that the contribution of the phase
shift of the light wave at its reflection from the FPF
spacer thickness equal to LS ≈ 160 μm. With these
mirrors to the effective optical length of the spacer plays
values, the free spectral range is Δf λ ≈ 1.18 nm, and the
an essential role in the precise spectral tuning of the
bandpass width is Δp λ ≈ 0.017 nm. filter bandpass. For the FPF with dielectric mirrors, we
Fig. 9 shows a good correlation between the have shown analytically that the dependence of this
measured angular dependence of the transmissivity and phase shift on the wavevector is linear up to a half-width
Eq. (43), illustrating a low sensitivity of the of the mirror reflection band, and derived a simple
acceptance angle to the mirror non-flatness. Indeed, for a expression for the phase-shift coefficient in terms of the
given spacer thickness LS and finesse φ, the total path l mirror parameters. It follows from the performed
of the beam inside the FPF can be estimated as analysis that the phase shift has to be accounted for the
l = φ⋅2LS, as the magnitude of finesse indicates the filter with a bandpass width narrower than ~ 10–4 and a
number of the interfering beams, i.e., the number of the tuning range wider than ~ 10–2 of the operational
passes through the interferometer. For the chosen values wavelength.
of the key parameters of the FPF this results in We have also analyzed the tolerances to the
l = 22.4 mm. By tilting the FPF on 1º the spatial absorbance of the mirror structure and reflectivity
472
Semiconductor Physics, Quantum Electronics & Optoelectronics. 2004. V. 7, N 4. P. 465-473.
difference between the two mirrors (for the required high Laboratory prototype double-etalon Fabry – Perot
throughput of the bandpass filter), to variations of the interferometer for remotely sensing atmospheric
environmental conditions (for the necessary precision of ozone // SPIE Proc. 3437, p. 158-166 (1998).
the bandpass spectral position control), and to the 8. W.L. Wolfe, Introduction to imaging spectrometers
deviation of the incoming light angle of incidence from // Ibid. TT25 (1997).
normal incidence (for the possibility of imaging 9. M. Born and E. Wolf, Principles of optics,
applications). Pergamon, Oxford (1980).
For experimental studies, we measured the FPF 10. R.P. Stanley, C. Weisbuch, et al., Impurity modes in
transmissivity function using a narrowband emission of one-dimensional periodic systems: The transition
He-Ne laser and found a significant broadening of the from photonic band gaps to microcavities // Phys.
bandpass with the increase of the filter aperture. By Rev. A48, p. 2246 (1993).
observing a non-uniform cross-sectional distribution of 11. Figotin and V. Gorentsveig, Localized
the light intensity in the transmitted non-collimated electromagnetic waves in a periodic layered
beam with different maximum-intensity spots at dielectric medium with a defect // Ibid. B58, 180-
different values of the mirror spacing it can be shown 188 (1998).
that this effect is due to the mirror non-flatness. We also 12. J.D. Joannopoulos, R.D. Meade, and J.N. Winn,
investigated the dependence of the filter transmissivity Photonic crystals, Princeton University Press,
on the angle of incidence and found a low sensitivity of Princeton (1995).
this function regarding the aperture width. In addition, 13. D.I. Babic and S.W. Corzine, Analytic expressions
we demonstrated an application of the filter width and for the reflection delay, penetration depth, and
found the transmissivity function to be the spectral absorptance of quarter-wave dielectric mirrors //
characterization of a broadband diode-laser emission. IEEE J. Quantum Electron. 28, 514–524 (1992).
Acknowledgments 14. L.R. Brovelli and U. Keller, Simple analytical
expressions for the reflectivity and penetration depth
The authors would like to thank NASA Langley of a Bragg mirror between arbitrary media // Opt.
Research Center for their support and funding of this Communs 116, p. 343–350 (1995).
project. 15. E. Garmire, Theory of quarter-wave-stack dielectric
mirrors used in a thin Fabry–Perot filter // Appl.
1. H. Buisson, Ch. Fabry, H. Bourget // Astrophys. J. Opt. 42, p. 5442-5449 (2003).
40, p. 241 (1914). 16. S. Weidong, L. Xiangdong, H. Biqin, et al.,
2. P. Jacquinot // J. Opt. Soc. Amer. 44, p. 76 (1954). Analysis on the tunable properties of MOEMS filter
3. V.B. Markov, V. Shishkov, Bragg diffraction with based on Fabry – Perot cavity // Opt. Communs 239,
multiple internal reflection // SPIE Proc. 1238, p. p. 153-160 (2004).
30-40 (1990). 17. V. Goren, A. Khizhnyak, and V.B. Markov,
4. V.B. Markov, V. Shishkov, Bragg diffraction with Coupled-wave analysis of the Fabry – Perot
multiple internal Reflection // Ibid. 1238, p. 30-40 interferometer with multi-layer dielectric mirrors
(1989). accounting losses, to be published.
5. W.B. Cook, H. Snell, P. Hays, Multiplex Fabry – 18. P. Yeh, Optical waves in layered media, Wiley,
Perot interferometer // Appl. Opt. 34, p. 5263-5277 New York (1988).
(1995). 19. H. Kogelnik, Coupled wave theory for thick
6. H.F. Dobele, and J.H. Massig, Application of a hologram gratings // BSTJ 48, N 9, p. 2909-2947
Fabry – Perot spectrometer to the measurements of (1969).
spectral line shifts much smaller than line width // 20. J.C. Owens, Optical refractive index of air:
Ibid. 15, 69-72 (1976). dependence on pressure, temperature and
7. A.M. Larar, W.B. Cook, and R.S. Lancaster, composition // Appl. Opt. 6, N 1, p. 51-59 (1967).
473