JOURNAL OF OPT OELECTRONICS AND ADVANCED MATERIALS Vol. 19, No. 7 - 8, July – August 2017, p.

486 - 491

Numerical simulation and transmission power analysis of

tapered optical fiber sensors
I. NAVRUZ* , M. BILSEL, F. ARI
Department of Electrical and Electronics Engineering, Faculty of Enginee ring, Ankara University, Ankara, Turkey

In this paper, the geometrical design parameters of tapered optical fibers and their effects on sensing of refractive index are
investigated. The transmission power of the sensor is simulated by using finite difference beam propagation method. The
simulations are run by changing geometrical parameters which are length of transition region, length of waist region and
waist diameter. The simulation results show that design parameters strongly affect the sensor performance including
sensitivity and sensing range. Moreover, it is shown that the sensing range can be adjusted and also the sensor sensitivity
can be enhanced using the optimized design parameters.

Received June 30, 2016; accepted August 9, 2017)

Keywords: Beam propagation method, Optical RI sensor, Tapered fiber

1. Introduction the waist region are shown by L2 and Rw, respectively, in

the figure.
Optical fibers are used in data transmission and
various types of applications such as illumination in
inaccessible places, fiber laser, fiber optical splitter or
coupler, optical amplifier, fiber optical filter, and also
optical sensor. In recent years an increasing interest has
been observed in optical fiber sensors, since they offer
some superior features when compared to traditional
electrical sensors. Immunity from electromagnetic
interference, remote sensing ability, light weight, usage in
hazardous environments and high data carrying capacity
are some of the advantages of these sensors. Moreover
easy fabrication, low-cost and fast responsivity make them Fig. 1. A typical biconical tapered fiber sensor
powerful sensors especially for biomedical and chemical
sensing applications. These sens ors are capable to sense When the wave propagates through the taper region
strain/stress [1-3], temperature [4-5], pressure [6] and indicated by L1 length in Fig. 1, the fundamental mode will
biological or chemical analytes [7-9]. be degenerate, because the field distribution varies
A standard single-mode fiber can be transformed into depending on the core and cladding diameters along the
a refractive index (RI) sensor by tapering a small part of it, way. According to local tapering slope, the optical power
thus the resultant sensor is known as biconical tapered in the fundamental mode is transferred to a few closest
fiber sensor. Biconical tapered fibers are produced by higher order modes. Fiber diameter decreases to waist
heating and stretching the fiber simultaneously, so that the diameter at the end of first tapered region so core is almost
diameter of cladding and core are reduced at same rate or wholly disappeared. As a result, the cladding acts as a
[10-13]. At the tapered region, core and cladding become new core and the medium surrounding the fiber is now the
indistinguishable. As a result, almost all of the light is new cladding. When a change occurs in the refractive
guided by the cladding which results an increase in the index of the medium, propagation constants of the modes
evanescent field. A typical biconical tapered fiber is will change due to the large difference of the RIs of
shown in Fig. 1. It includes three regions: regions with cladding and medium. This will lead to optical power
declining and ascending diameter at both ends and a oscillations in the response of tapered fiber and it may also
region with constant diameter commonly known as waist cause transmission loss at the end of the fiber output.
in the middle. The uniform fiber region before descending Tapered fibers can be classified as adiabatic and non-
region will be called as input fiber and the uniform fiber adiabatic according to tapering slope of both end regions
region following the ascending region will be called as with lengths L1 as seen in Fig. 1. In the adiabatic structure,
output fiber. In this paper, heating and stretching processes small relative change should be allowed in decreasing
are assumed to be done perfectly, so the lengths of both fiber radius at taper region. If the tapering slope from
declining and ascending regions are equal and they are unaltered fiber to tapered fiber is very gradual, single-
specified by L1 in the figure. The length and diameter of mode is only coupled into lowest order mode (HE11 )
 Numerical simulation and transmission power analysis of tapered optical fiber sensors 487

without coupling to higher order modes so most of optical refractive index of the medium is varied between 1.30 and
power remains in this fundamental mode and loss of the 1.40. The length of uniform fiber called as output fiber
light is negligible. following the ascending region must be sufficiently long
When the slope of ascending and descending regions to ensure that light is coupled to single mode at the output
at both ends of the tapered fiber is very steep, the structure of uniform fiber [19]. In our designs, the length of the
is called nonadiabatic. In such a structure, fundamental output fiber is about 2 cm. The wavelength of light is 1.55
mode is coupled into higher order modes, mostly into first µm and mode calculation is evaluated for standard single
two lower order modes HE11 and HE12 [14]. These modes mode fiber before launching the light into the input fiber.
propagate along the taper with different phas es depending
on taper length, RI of fiber and surrounding medium. All
modes couple back to the fundamental mode after leaving 2. Simulation results
waist region. As a result of coupling fundamental mode
into higher order modes, the relative phase difference and In order to investigate the effect of design parameters
output intensity are given as [15] on optical BTF sensor characteristics, firstly, the
simulation was run for sensors having different waist

   n ( z )dz
l diameters. The normalized power transmissions obtained
 mn   mnl  m ( z)  (1)
0 for different waist diameter values are shown in Fig. 2. In
these simulations, other geometrical parameters L1 and L2
N N are kept constant at 1200 and 1500 µm, respectively.
I I
m1
m 2 
mn n1
I m I n cos mn (2) When the waist diameter is increased, the transition
geometry of tapered region becomes smooth in such a
form that it can be called more adiabatic. On the other
where l is length of the taper, N is the total number of hand, reducing the waist diameter causes a steeper
modes supported by tapered fiber, Im is the intensity of transition region so-called more nonadiabatic.
Nonadiabaticity resulting from reduction of the waist
HE1m mode and βmn is the difference of propagation
diameter leads to high coupling ratio between unaltered
constants of the modes. Changing either β or l will cause
fiber and tapered fiber that increases power oscillation at
a shift on power transmission versus wavelength or RI
the output of the fiber as shown in Fig. 2. This result can
axes. Phase difference between propagation constants of
be expected if it is considered that a smaller waist
any two modes is related to effective RI,
diameter stimulates the interaction of the guided field with
the surrounding medium. In addition, smaller waist
2
  neff (3) diameter decreases sensing range of the sensor. It is
 clearly seen from Fig. 2 that sensor with diameter of 20
µm has the longest RI sensing range. However, sensitivity
where  is wavelength, neff is effective RI which decreases because of little power fluctuations which
cannot be measured due to sensor’s detection limit. By
depends on refractive index of fiber and the surrounding changing the waist diameter, a specific RI measurement
medium. range with acceptable power resolution can be adjusted.
The finite difference beam propagation method
(BPM) is a powerful technique to simulate lightwave
propagation in tapered optical fibers like slowly varying
optical waveguides. This method is based on slowly
varying envelope approximation that solves Maxwell's
equations with paraxial approximation by using finite
differences in place of partial derivatives. More detailed
information and numerical implementation about this
method are given in [16-18], and a few commercially
available software packages, for example by Optiwave and
RP Photonics companies, can be found.
In this paper, the normalized transmission power of
BTF sensor is simulated numerically using BPM. The
transferred light powers related to RI are analyzed for
different sensor designs by changing its geometrical Fig. 2. Normalized power transmission of BTF sensors
design parameters, L1 , L2 and Rw. Finally, the effect of with different waist diameters, 8, 11, 14 and 20 µm
these parameters on the RI sensitivity and sensing range of
the sensor are investigated.
The uniform fiber parameters are compatible with The sensitivity of the sensor can be defined as
standard single mode fiber parameters of 9 µm core radius, P / n for approximately linear part of optical power
125 µm cladding radius, 1.45 and 1.445 core and cladding fluctuation curve. R-squared value which is commonly
refractive indices, respectively. In all simulations used as an indicator of how well the linear curve fits the
488 I. Navruz, M . Bilsel, F. Ari

data can test the linearity of sensor. The R-squared value In the second step of simulation studies, the influence
reaches to 1 at maximum when the fitted line is exactly of length L1 on sensor characteristic has been investigated
linear. In this paper, R- squared value for sensitivity by keeping waist length L2 constant at 1500 µm. The
analyses is calculated as 0.996 at minimum so that it can length of transition region L1 is varied from 800 µm to
be considered as excellent. In this way, the sensitivity in 1300 µm with steps of 100 µm while waist diameter was
normalized transmitted power per RI unit is calculated as 14 µm. The results are presented in Fig. 4(a-f).
15.786 RIU-1 and 2.795 RIU-1 for two sensors with The variation of L1 parameter caused significant
different waist diameters of 11 and 20 µm, respectively. amount of variation in power oscillation curves. When
Also, these sensors can measure refractive index between Fig. 4 is examined in detail, it is seen that L1 affects the
1.371-1.391 and 1.340-1.394 intervals and the results are position of the oscillation. However oscillation frequency
plotted in Fig. 3. is not affected. In other words, the power transmission
graphics have one peak (negative or positive) in the RI
range of 1.30-1.40. Simulation results obtained using
smaller step size for L1 are given in Fig. 5(a) and (b). In
these simulations Rw and L2 are kept constant at 14 µm and
1500 µm, respectively, and L1 is changed between 930-
970 µm for Fig. 5(a) and 1080-1120 µm for Fig. 5(b) with
5 µm steps.
As seen from both figures, when the parameter L1
increases with 5 µm step size, power oscillation shifts
towards right and a little change in its shape is observed as
well.

Fig. 3. Normalized power transmission and sensing

range represented by fitted line for waist diameters
Rw=11 and 20 µm

Fig. 4. Normalized power transmission for variation of L 1 between 800-1300 µm, Rw=14 µm, L 2=1500 µm
 Numerical simulation and transmission power analysis of tapered optical fiber sensors 489

This shift caused by variation of L1 was observed in sensitivity does not exhibit a significant amount of
all the simulations for different sensor designs. Both the difference.
shifting effect and the little change in oscillation shape
will change the sensitivity of the sensor and sensing range
of the sensor will also be affected. For example, if it is
desired to measure RI between 1.36 and 1.40, the sensor
design with L1 =970 µm in Fig. 5(a) can be chosen as the
most suitable. On the other hand, the power variation
versus RI for the sensor design with L1 =1080 µm in Fig
5(b) is almost linear between 1.30 and 1.36. Thereby this
is the most suitable design to make measurement in that
interval.

Fig. 6. Normalized power transmissions for longer L 1 values

Another geometrical parameter that can affect

characteristics of biconical tapered fiber sensors is the
length of waist region that is symbolized as L2 . The
simulation results obtained for different values of L2 are
shown in Fig. 7. The parameter L2 is increased with 50 µm
step size between 1000-1400 µm and 5 µm step size
between 1430-1470 µm in Fig. 7 (a) and (b), respectively.
Here Rw and L1 parameters are kept constant at 14 and
1000 µm, respectively.

Fig. 5. The simulation results for variation of L 1 by step

size of 5 µm between (a) 930-970 µm, (b) 1080-1120 µm,
when Rw=14 µm, L 2=1500 µm.

When L1 is increased to a few thousand micrometers,

transition from untapered fiber to tapered fiber becomes
very gradual. In this study, L1 is increased to nearly 5000
µm to analyze its effect on power transmission when Rw
and L2 were kept constant at 14 and 1500 µm,
respectively. Fig. 6 demonstrates power transmission of
sensors with different values of L1 between 1600-5350 µm
and sensing ranges with fitted R-squared value. The
minimum calculated R-squared value is 0.996. Some
specific values for L1 were chosen among lots of
simulations to compare the power transmission of sensors
whose characteristics are similar. Using the curves in Fig Fig. 7. The simulation results obtained for different
length of waist region, L 2 that increased by step size of
5, RI sensing ranges of the sensors are calculated as 0.057,
(a) 50 µm between 1000-1400 µm and (b) 5 µm between
0.044, 0.040 and 0.035 for L1 equal to 1600, 3450, 4400 1400 - 1470 µm, other parameters are Rw = 14 µm,
and 5350 µm, respectively. It is clearly noted that as L1 L1=1000 µm
increases RI sensing range becomes narrow but sensor
490 I. Navruz, M . Bilsel, F. Ari

As seen from Fig. 7(a), the normalized power and increased with 800 µm steps without changing other
transmission curves exhibit different oscillations those design parameters, the obtained power oscillation curv es
have single positive or negative peak point in the RI range are again similar to each other as seen in Fig. 9(b). Here
of 1.30-1.40 for different values of L2 . Increasing L2 the curves have positive peaks around RI of 1.38. The
causes the power oscillations to shift like the effect of L1 . power oscillation curves in Fig. 9(a) show negative trends
Therefore, the sensing range and the sensor sensitivity can within RI range of about 1.34-1.38, while the curves in
also be controlled with L2 . The normalized power Fig. 9(b) show positive trends approximately the same
transmission curves for L2 between 1430 and 1470 µm are range of RI. As it is seen from Fig. 8 and Fig. 9, many
given in Fig. 7(b). It is seen that the sensor designed for sensor responses including desired RI sensing range and
L2 =1430 µm can operate in range of 1.300-1.374 with the sensitivity can be obtained by controlling the parameter L2 .
calculated sensitivity of 3.3128 RIU-1 . On the other hand,
the sensor with L2 = 1470 µm can operate in range of
1.339-1.397 with sensitivity of 5.9038 RIU-1 .
In order to analyze the sensor characteristics of
biconical tapered fiber having longer waist region, the
value of L2 has been increased from 1800 to 1950 µm in
Fig. 8(a) and from 5100 to 5250 µm in Fig. 8(b) with 50
µm step size without changing Rw and L1 parameters. It is
clearly seen that as L2 increases the frequency of
oscillation also increases. The reason for the increase in
fluctuation frequency is that the longer waist region allows
the optical power to be transferred to higher order modes,
thus, higher slope changes occur in the power oscillations.
This response of the sensor can be used to perform more
sensitive measurement in a special range of RI.

Fig. 9. Normalized power transmission curves whose

characteristics are similar to each other when L 2 is
changed with 800 µm step size. The initial value of L2
is chosen as (a) 2050, and (b) 1950 µm

When L2 is increased by 800 µm step size, it is

obtained that the sensor sensitivity is improved. However,
the value of 800 µm step size is specific to this sensor
design and it may be altered for different sensor designs in
order to compare analogous power oscillation curves. The
fitted lines are shown as black dotted lines in Fig. 9 and
the sensitivities are calculated as 10.36 and 30.18 RIU-1
for the sensors with L2 values of 2050 and 5250 µm,
respectively.
Fig. 8. Normalized power transmissions for longer The sensitivity and sensing range variation versus
L 2 increased by step size of 50 µm between (a) length L2 calculated from Fig. 9(a) and (b) are shown in
1800-1950 µm (b) 5100-5250 µm the first and last three columns of Tab. 1, respectively. As
seen in the table, the sensitivity variation with L2 is not
The response of the sensor by changing L2 with 800 linear. However, it is clear that the sensitivity increases as
µm steps is illustrated in Fig. 9. It is noted that oscillation L2 increases. On the other hand, the sensing range
curves in Fig. 9(a) for different L2 values are similar to decreases while sensitivity increases.
each other. Here the curves have negative peaks around
1.38 RI. When the initial value of L2 is chosen as 1950 µm
 Numerical simulation and transmission power analysis of tapered optical fiber sensors 491

Table 1. Sensitivity and sensing range variation versus References

waist length
[1] S. W. Choi, J. Lee, B. H. Oh, H. S. Park, International
Journal of Distributed Sensor Networks , Article ID:
894780, 9 pages (2013).
[2] E. Li, IEEE Photonics Technology Letters
19, 1266 (2007).
[3] C. R. Dennison, P. M. Wild, D. R. Wilson,
M. K. Gilbart, Measurement Science and
Technology 21, 115803 (2010).
[4] S. Zhu, J. W. Xing, F. F. Pang, T. Y. Wang, Journal
of Shanghai University 15, 101 (2011).
[5] M. K. Saxena, S. D. V. J. Raju, R. Arya,
R. B. Pachori, S. V. G. Ravindranathc, S. Kher,
S. M. Oak, Optics & Laser Technology
65, 14 (2015).
[6] N. Wu, Y. Tian, X. Zou, Y. Zhai, K. Barringhaus,
X. Wang, Sensors and Actuators B: Chemical
181, 172 (2013).
3. Conclusion [7] Q. Meng, X. Dong, K. Li, Y. Ni, B. Xu, Z. Chen,
IEEE Sensors Journal 14, 1813 (2014).
In this study, the optical power response of biconical [8] S-H. Ohk, A. K. Bhunia, Food Microbiology
tapered fiber as a refractive index sensor is numerically 33, 164 (2013).
simulated for various waist radius, length of transition [9] M. I. Zibaii, H. Latifi, Z. Saeedian, Z. Chenari,
region and length of waist region values symbolized as Rw, Journal of Photochemistry and Photobiology B:
L1 and L2 , respectively. These geometrical design Biology 134, 55 (2014).
parameters of BTF sensors strongly affect their RI sensing [10] S. W. Harun, K. S. Lim, C. K. Tio, K. Dimyati,
characteristics especially sensor sensitivity and sensing H. Ahmad, Optik - International Journal for Light and
range. When Rw is increased gradually from 8 to 20 µm, Electron Optics 124, 538 (2013).
the sensing range increases, but sensitivity decreases. [11] A. Leung, K. Rijal, P. M. Shankar, R. Mutharasan,
Variation of L1 leads to significant changes in the power Biosensors and Bioelectronics 21, 2202 (2006).
oscillations; however, the oscillation frequency is not [12] S. Pricking, H. Giessen, Optics Express
affected. If L1 is increased a few µm, the transmission 18, 3426 (2010).
power oscillation versus RI slightly shifts towards to right. [13] H. Latifi, M I. Zibaii, S. M. Hosseini, P. Jorge,
When L1 is increased to a few thousand µm, the sensing Photonic Sensors 2, 344 (2012).
range decreases but sensitivity doesn’t increase. On the [14] G. Cohoon, C. Boyter, M. Errico, K. Vandervoort,
other hand, if L2 is increased a few µm, the transmission E. Salik, Optical Engineering 49, 034401 (2010).
power oscillation versus RI slightly shifts towards to right [15] Y. Tan, J. Lou, H. Xu, J. Huang, W. Shen,
similar to effect of L1 . When L2 is increased to a few International Symposium on Photoelectronic
thousand µm, the sensitivity increases in a slightly Detection and Imaging, Proc. of
narrower sensing range as a result of higher change in the SPIE 8914, 891407 (2013).
power oscillation due to excitation of higher order modes. [16] Y. Chung, N. Dagli, IEEE Journal of Quantum
This response of the sensor can be used to perform more Electronics 26, 1335 (1990).
sensitive measurement in a particular range of RI. [17] S. A. Shakir, R. A. Motes, R. W. Berdine,
IEEE Journal of Quantum Electronics
47, 486 (2011).
Acknowledgements [19] M. S. Wartak, Computational Photonics, an
Introduction with MATLAB. Cambridge
This work was supported by the Scientific and University, 2013.
Technological Research Council of Turkey (TUBITAK) [20] W. Klaus, W. R. Leeb, Optics Express
under Project: 114E040. 15, 11808 (2007).

__________________________
*
Corresponding author: [email protected]