Optical and Quantum Electronics 28 (1996) 1629-1639

Fibre-optic evanescent field absorption

sensor based on a U-shaped probe

B. D. G U P T A , H. D O D E J A , A. K. T O M A R
Physics Department, Indian Institute of Technology, New Delhi - 110016, India

Received 4 December 1995; revised 9 April; accepted 12 April 1996

A fibre-optic evanescent field absorption sensor based on a U-shaped sensing probe is

described. The influences of fibre core diameter, bending radius of the probe and the
refractive index of the fluid on the sensitivity of the sensor are evaluated experimentally.
The results are compared with the theoretical results obtained using geometrical optics
and based on two-dimensional treatment. A good qualitative agreement is found
between them.

1. Introduction
In the past few years, evanescent field fibre-optic absorption sensors have received considerable
attention due to their widespread applications in continuous monitoring of concentrations of
reactants in chemical processes and in the study of the absorption spectra of liquids and pastes
[1-14]. Further, these sensors can be used in the chemical industry for distributed sensing, as
well as for sensing at isolated locations. Much of the research to date on evanescent field fibre-
optic absorption sensors has used plastic-clad silica (PCS) fibres. These sensors are based on
attenuated total reflection (ATR) spectroscopy. In ATR spectroscopy, plastic cladding is
removed from the middle portion of the fibre and the exposed core is surrounded by an absorb-
ing fluid. The light is launched from one end of the fibre and is detected at the other end. The
transmitted power depends on the absorption of the evanescent field penetrating into the clad-
ding region formed by the absorbing fluid. Thus, the presence of the fluid and its concentration
are determined by observing the transmitted power through the fibre. The evanescent field
absorption, for a given length of the unclad fibre, depends on the number of ray reflections
per unit length of the unclad fibre and the penetration depth of the evanescent field in the
sensing region. The penetration depth is given by:
A
dp = 27rnl (sin2 0 - sin 2 0c) 1/2 (1)

where A is the free space wavelength of the light launched into the fibre, nl is the refractive
index of the core, Oc = [sin-l(nE/nl)] is the critical angle of the sensing region with respect
to the normal on core-cladding interface, n 2 is the refractive index of the absorbing cladding
(fluid), and 0 is the angle of the ray with the normal to the core-cladding interface. As 0
approaches Oc, the penetration depth increases, which results in the increase in evanescent
absorption and hence the sensitivity of the sensor. In practice, n 2 < nel, ncl is the refractive
index of the cladding (plastic) and, therefore, the numerical aperture of the sensing region is

0306-8919 9 1996 Chapman & Hall 1629

 B. D. Gupta et al.

higher than that of the fibre. In other words, the critical angle of the fibre is higher than that of
the sensing region. Thus, 0 of a guided ray in the fibre cannot approach 0c of the sensing region
if the sensing region is straight and uniform in diameter. Hence, one cannot achieve significant
evanescent field penetration in the sensing region, and as such the sensitivity of the sensor is
limited. To increase the sensitivity, the use of selective ray launching in the fibre [7, 12] and
tapering of the sensing region have been suggested [8, 14]. In the selective launching, the
sensitivity is increased by launching only those rays which make angles close to the critical
angle of the sensing region; while in the case of taper the angle of a ray approaches the critical
angle as it propagates. The angle of the ray can also be brought close to the critical angle of the
sensing region if the sensing part of the fibre is bent. Based on this, fibre-optic refractive index
sensors with a U-shaped probe have been reported in the literature [15-17].
In the present paper we have carried out a detailed experimental study of the fibre-optic
evanescent wave absorption sensor based on the U-shaped sensing region. The performance
of the sensor has been studied with different fibre core radii, bending radii of the fibre, refract-
ive indices and concentrations of the dye in the solvents. It has been shown that the sensitivity
of the sensor increases with decreasing bending radius of the probe and also with the increase in
refractive index of the fluid. To compare experimental results we have evaluated the character-
istics of the sensor with a geometrical optics approach based on a two-dimensional treatment.
The experimental and theoretical results have been found to match qualitatively.

2. Theory
Consider a step index multimode optical fibre whose cladding has been replaced locally by an
absorbing fluid. If P0 is the power transmitted by the fibre in the absence of an absorbing fluid,
then the power transmitted by the fibre in the presence of the fluid is given by:
P = Po exp(-TL) (2)
where L is the length of the unclad portion of the fibre and 7 is the evanescent absorption coef-
ficient of the fluid. In the case of a ray making an angle, O, with the normal to the core-cladding
interface in the sensing region, 7 is given by [18]:
c~)~n2 cos/9 cot 0
7(0) ----27rpn21cos2 0c (cos 2 0c _ cos2 0 sin 2 0r (3)

where 0~ is the skewness angle, c~ is the bulk absorption coefficient of the fluid and p is the
radius of the fibre core. It may be noted from Equation 3 that for a fixed value of 0, the
evanescent absorption coefficient is maximum when the ray is meridional (i.e. 0~ -- 7r/2). It
decreases with the increase in skewness. Recently, we have shown [I0] that if the refractive
index of the absorbing fluid does not vary significantly with the concentration, c, of the fluid
within the desired concentration range, then the sensitivity [ - 1 / P (dP/dc)] of the sensor is
proportional to L,~/a. Thus, for a given L, 3'/a directly defines the sensitivity. It means that
the sensitivity will be maximum if the meridional rays are launched in the fibre. To propagate
meridional rays in the fibre, light is launched into the fibre from a collimated source (such as a
laser) using a microscope objective. The objective is used to focus the beam onto the fibre end
face at the axial point. The numerical aperture of the objective is assumed to be greater than
that of the fibre, so that all the bound rays can be excited in the fibre. For such an illumination,
the power, dP, arriving at the axial point on the fibre end face between angles 00 and 00 + d00 is
proportional to (tan 00/cos 2 00)d00, where 00 is the angle of the ray with the axis outside the
fibre [18]. Using Snell's law and 0 = 90 ~ - 01, where 0, is the angle of the same ray with

1630
Fibre optic evanescent field absorption sensor based on a U-shaped probe

the axis of the fibre inside the core, we can write:

dP oc n2 sin 0 cos 0 dO
(1 - cos2 0) 2
Hence, the effective evanescent absorption coefficient for the meridional rays launched into the
sensing region for the angles (with respect to the normal to the core-cladding interlace) in the
range (01, 02) is given by:

"Teff(0,,02) = , , ( 1 ~ 1 2 c-~s20)2 ,/(0)d0 (1 - n ~ cos20) 2 dO (4)

For all bound rays to propagate in the sensing region, 01 = sin -1 (ncl/nz) and 02 = 90 ~
Let us consider a U-shaped sensing region as shown in Fig. 1. To simplify the theoretical
treatment of this probe we consider a two-dimensional approach. Under this approach, the
rays in the sensing region will be mainly meridional and will confine in the plane of bending.
As mentioned above, the sensitivity increases with decreasing skewness and, therefore, in the
case of the two-dimensional approach, the calculated sensitivity will be more than that in
the actual case. For a ray to propagate through the sensing region its angle with the normal
to the core-cladding interface in the sensing region should be greater than the critical angle
of the sensing region. In a U-shaped fibre there can be two possibilities; either a ray can
undergo total internal reflection (TIR) both at the outer and inner surfaces (Fig. 1, continuous
ray), or it undergoes TIR only at the outer surface and does not reach the inner surface at all
(dotted ray). If 0 represents the angle that a guided ray makes with the normal to the core-
cladding interface in the straight fibre then the corresponding angle ~b at the outer surface of
the bent core is given by the expression:

sin ~ = kR + 2p] sin 0 (5)

Cladding, nc~
,/
2~p~~ C
, \ o\ \r e \, \ \ \ \n \ \ \ ~ \ \ \ \ \ \.-,,.--
't

Figure 1 Geometry of the U-shapedsensing region and the representation of a meridional ray in it.

1631
 B. D. Gupta et aL

where h is the distance at which the ray is incident on the entrance of sensor from the core-
cladding boundary and R is the bending radius. Similarly angle 6 of the ray at the inner surface
of the bent core is given by:

sin6 = ( - ~ ) sin0 (6)

Thus the angle 0 is transformed to angles ~ and 6 according to Equations 5 and 6, respectively.
Hence, the effective evanescent absorption coefficient for the outer surface of the sensing
region will be given by:

COS3 0 dO dh
[%fe(q~l, @2)]outer= 2127rp(n~ - n22)] x [jo jr (1 - n 2 cos 2 0)2(n~2 sin 2 0 - 1) 1/2

_. I?L~ (7)
- 07J
where

= sin-1 r ,. + h)ncl ]
and
( R +h'~
~2 = sin-1 \R + 2p)

Similarly, the effective evanescent absorption coefficient for the inner surface of the sensing
region can be written as:

aAni X [ f 2 P f '2 cos3OdOdh

[%ff(61,62)]i~n~, ----2127rp(n2 _ n 2 ) ] LJO J6 (1 -n2cos2O)2(n~2sinZO- 1) 1/2

2p 62 sin 0 COS0 dO dh]

where

,51=sin-1 [(R
L ~+ h)ncl]-j
and 62 = 90 ~
Thus the total effective evanescent absorption coefficient in the case of a U-shaped sensing
region is given by:
")'eft = ['7eff(~l, ~2)]outer + ["Yeff(~l, (~2)]mner (9)
If the total internal reflection occurs only from the outer surface then the second term in
Equation 9 vanishes. As mentioned above, for a given length of the sensing region, the sensi-
tivity is proportional to "//a; we calculate this ratio to evaluate the characteristics of the sensor.
To calculate 7 / a we choose PCS fibre of numerical aperture (NA) = 0.2, with nl = 1.457
and p = 100#m. These values give n d = 1.443 and 82 ~ as the critical angle of the
fibre. Further, we choose A = 663 nm, the wavelength corresponding to the peak evanescent

1632
Fibre optic evanescent fieM absorption sensor based on a U-shaped probe

absorbance of methylene blue dye, which has been used mostly. The values of % f f / a for dif-
ferent values of n2 and R are calculated using Equations 7-9, and are shown in Fig. 2 (contin-
uous curves). The dotted curve corresponds to R = oo (i.e. for the straight sensing region). It
may be noted that the value of %ff/oL increases with the increase in the refractive index of the
liquid, n2. Further, for a fixed value of n2 the value of % f f / a or the sensitivity increases as the
bending radius, R, decreases. These results also show that as the sensitivity increases the avail-
able range for refractive indices decreases. The sensitivity of the sensor for n2 = 1.335 and the
range of usable refractive indices for different values of bending radius for a fibre of NA = 0.2
a r e summarized in Table I.
To see the effect of a fibre's numerical aperture on the sensitivity of the sensor we have car-
ried out similar calculations for NA = 0.4. In this case nd = 1.401 and the critical angle of the
fibre is 74 ~ In Fig. 3 the values of % f f / a have been plotted as a function of n2 for different
values of R for a fibre of NA = 0.4. Comparing these results with those plotted in Fig. 2,
we find that the sensitivity has increased significantly with the increase in the numerical aper-
ture of the fibre for a fixed bending radius of the sensing region. In Table II we have summar-
ized the results obtained for NA = 0.4.

3. Experimental
The experimental set-up to study the characteristics of the U-shaped probe is shown in Fig. 4.
To fabricate the probe, plastic-clad silica (PCS) fibres of 200 and 600 #m core diameters and
NA = 0.17 were used. The cladding was removed from a small portion of the central region of
the fibre. To bend the unclad portion it was first exposed to a flame and then slowly bent until it
became U-shaped. For 200 #m fibre, the temperature of the flame was in the range 400-500~

32

28

10J/
i ,

24

2O

x 16

12

8
j j s j~"

I t ,,I I ....... I i I ,,,1 ........ I

0
1.335 1.355 1.375 1.395 1.415 1.425
n2

Figure 2 Variation of "T~f/~ as a function of n2 for three different values of R for a fibre with NA = 0.2. ( - - - )
straight fibre.

1633
 B. D. Gupta et al,

TABLE I Values of "~en/c~and the range of functional refractive indices of the fluid for NA = 0.2 of the fibre

Design % f f / a for R a n g e o f n2 Ratio o f sensitivity with respect to

n z = 1.335 the straight case at nz = 1.335

Straight 1.06 • 10 -4 <1.443 1.0

Bent 2.54 x 10 -4 <1.420 2.4
(R = 1,5 cm)
Bent 3.42 • 10 -4 <1.410 3.2
(R = 1 . 0 c m )
Bent 6.84 x t 0 -4 <1.385 6.5
(R = 0.5 c m )

24

20
R;o /
I~= 1.5 cm
16
%
x 12

0I I I .... I 1 ................. I
1.335 1.345 1.355 1.365 1.375 1.385
n2

Figure 3 Variation of "~e~/ct as a function of n2 for three different values of R for a fibre with NA = 0 4. ( - - - )
straight fibre.

while for 600 #m fibre the temperature was 700~ The temperature of the flame was controlled
by proper mixing of liquid petroleum gas and oxygen. The uniformities of the core diameter
and the bending radius were checked using a travelling microscope. The experiments were
performed only on those probes that were uniform in core diameter and the bend was close

T A BLE I1 Values of 3'eff/~ and the range of functional refractive indices of the fluid for NA = 0.4 of the fibre

Design %ff/a for R a n g e o f n2 Ratio of sensitivity with respect to

n2 = 1.335 the straight case at n2 = 1.335

Straight 5.25 • 10 -4 <1 401 1.0

Bent 7 04 x 10 -4 <1.380 1.3
(R = 1.5 c m )
Bent 8.35 / 10 `-4 <1.370 1.6
(R = 1.0 era)
Bent 14.68 • 10 -4 <1.345 2.8
(R = 0 . 5 c m )

1634
Fibre optic evanescent field absorption sensor based on a U-shaped probe

Optical
I He-Ne 1 0 fibre Power
laser r meter
Microscope
objective

~ Glass cell
Unclad sensing
region
Absorbing
liquid

Figure 4 Experimentalset-up used to evaluatethe performanceof the sensor.

to U-shape. The light from an H e - N e laser operating at 632.8 nm was focused using a micro-
scope objective (NA = 0.4, x20) on the input face of the fibre. The total length of the fibre in
all the experiments was about 50 cm. The bent portion of the fibre was fixed in the cover of a
glass cell of 5 ml capacity using an adhesive paste. The fluid used for the experiments was

0.3

0.2
E
Q

0.1
o/
I 1 i I i I i t
0 10 20 30 40
Concentration(llmOl)

Figure 5 Variationof the evanescentabsorptioncoefficientof methyleneblue dye as a functionof its concen-

tration in deionized water. The fibre was of 600/~m core diameter, the bending radius of the probe was
0.213cm and n2 was 1.33. (9 experimentalpoints.

1635
 B. D. Gupta et al.

1.5

1.0

E
p..

0.5

0 i I = I I I , I
0 10 20 30 40

Concentration (lamol)

Figure 6 Variation of the evanescent absorption coefficient of methylene blue dye as a function of its concen-
tration in deionized water. The fiber was of 200/~m core diameter, the bending radius of the probe was 0.1 cm
and n2 was 1.33. ((3) experimental points.

methylene blue dye dissolved in deionized water. To have different refractive indices of the
solvent, glycerol was mixed with the water in different ratios. The refractive index was
measured using an Abb6 refractometer. The output end of the fibre was connected to a
power meter. After each measurement with dye, the cell and the unclad portion of the fibre
were cleaned to remove the adsorbed dye molecules from their surfaces.To clean, 5% KOH
in ethanol was used as reported in [19]. In the experiment, the transmitted powers through
the fibre were measured separately for the dye solution, P, and the solvent, P0, in the cell.
These values, along with the length of the unclad portion of the fibre, were used to calculate
the evanescent absorption coefficient, 7, from Equation 2.

4. Results and discussion

To study the sensitivity of the fibre-optic evanescent field absorption sensor in detail, experi-
mental investigations have been carried out with different fibre core diameter, bending radius
of the probe and the refractive index of the fluid surrounding the probe. In the present study, a
solution of methylene blue dye has been used as the absorbing fluid. Figure 5 shows the varia-
tion of the evanescent absorption coefficient with the concentration of the dye dissolved in
deionized water for 600#m core diameter fibre and 0.213cm as the bending radius of the
probe. The variation is non-linear in nature, as reported in [7], [12] and [19]. It may be
noted that at 10#mol concentration, the experimental value of 7 is 0.19. If we compare this
value with those obtained for a straight uniform core (R = oc) and tapered probe (taper
ratio---1.67) [12] we find that the U-shaped probe is more sensitive than the uniform and

1636
Fibre optic evanescent field absorption sensor based on a U-shaped probe

0.35

0.30

'T
oE 0.25

0.20

I I I I I,
.33 i .34 1.35 1.36 1.37 1.38

Refractive index
Figure 7 Variation of evanescent absorption coefficient of the dye as a function of the refractive index of the
solvent. The fibre used was of 600pm core diameter and the bending radius of the probe was 0.213cm. The
concentration of the dye was 10/zmol.

tapered probes. This is because of the high value of the evanescent absorption coefficient in the
case of the U-shaped probe. In the case of uniform and tapered probes the experimental values
for 10#mol concentration are 0.031 and 0.074, respectively. In Fig. 6 we have plotted similar
results, but for 200 #m core diameter fibre and 0.1 cm bending radius of the probe. Again, the var-
iation of evanescent absorption coefficient with concentration is non-linear. Comparing these
results with those of Fig. 5 we find that the values of 7 have increased many-fold. This is because
of small values of both the fibre core diameter and the bending radius in Fig. 6.
Figure 7 shows the variation of the evanescent absorption coefficient with the refractive
index of the solvent for 10#mol dye concentration. The probe used was the same as in
Fig. 5. The refractive index of the solvent was changed by mixing glycerol with deionized
water and its value was measured using an Abb6 refractometer. It can be seen from the figure
that as the value of refractive index increases the evanescent absorption coefficient and hence
the sensitivity of the sensor increases. These results are qualitatively in agreement with the
theoretical results plotted in Fig. 2.
To see the effect of bending radius on the sensitivity of the sensor, experiments have been
carried out with two different bending radii, 0.10 and 0.15 cm. The fibre used for the experi-
ment was of 200 #m core diameter. The experimental results obtained for 10 #mol dye concen-
tration are shown in Fig. 8. Two points may be noted from the figure. First, the value of "7
increases as the refractive index of the solvent increases. These results are consistent with

1637
 B. D. Gupta et aL

R = 0.10 ~ / / ~ 9

0.72

= ,

0.68

E
t~

0.64

0.60

.I I I I I
1.33 1.34 1.35 1.36 1.37 1.38
Refractive index
Figure 8 Variation of evanescent absorption coefficient of the dye as a function of the refractive index of the
solvent. The fibre core diameter was 200#m. Filled circles correspond to 0.10cm bending radius, while the
open circles correspond to 0.15 cm bending radius of the probe. The concentration of the dye was 10/~mol.

those obtained with 600 #m core diameter fibre and shown in Fig. 7. Second, for a given
refractive index, the decrease in bending radius increases the value of "7 and hence the sensi-
tivity of the sensor. Again, these results are in agreement with the theoretical results reported
in Figs 2 and 3.
It may be noted that in the theoretical treatment presented in Section 2, the evanescent
absorption coefficient, "7, is directly proportional to the bulk absorption coefficient, c~, of the
fluid, which is normally proportional to the concentration of the absorbing material in solution
(Beer's law). This implies that the experimental results in Figs 5 and 6 should be straight lines,
which they are not. In the literature, various reasons have been assigned to this discrepancy
between theory and experiment. A detailed theoretical explanation has been given by Payne
and Hale [20]. According to them, the non-linearity occurs because each mode in the fibre
has a different absorption coefficient and the output from the sensor is the sum of the individual
mode powers. In contrast, Ruddy et al. [7] and Safaai-Jazi and Petersen [19] mentioned that the
non-linearity in experimental results occur due to the adsorption of dye on the silica core
surface of the fibre. In fact, Safaai-Jazi and Petersen [19] have experimentally shown that
the non-linearity of the response can be eliminated if the reactivity of the dye with the silica
surface of the core is prevented. In the theory given above we have not taken into account
these points. However, as shown above, our theoretical results on the dependence of evanescent
absorption coefficient on the refractive index of the fluid and the bending radius of the probe
are qualitatively in agreement with the experimental results.

1638
Fibre optic evanescent field absorption sensor based on a U-shaped probe

5. Conclusions
In summary, we have carried out a detailed experimental study on the fibre-optic evanescent
field absorption sensor with U-shaped probe. The effects of fibre core diameter, bending radius
and refractive index of the solvent on the sensitivity of the sensor have been studied. The
experimental results have been found to match, qualitatively, the theoretical results based on
a two-dimensional treatment and meridional rays analysis.

Acknowledgements
The authors are grateful to Professor A. K. Ghatak for constant encouragement. The present
work is partially supported by the Council of Scientific and Industrial Research (India).

References
1. P. H. PAUL and G. KYCHAKOFF,Appl. Phys. Left. 51 (1987) 12.
2. S. SIMHONY, A. KATZIR and E. M. KOSOWER, AnaL Chem. 60 (1988) 1908.
3. C. A. VILLARRUEL, D. D. DOMINGUEZand A. DANDRIDGE,Proc. SPIE 798 (1987) 225.
4. S. SIMHONY, I SCHNITZER, AmKATZIR and E. M. KOSOWER, J. AppL Phys. 64 (1988) 3732.
5. L SCHNITZER, A. KATZIR, U. SCHIESSL,W, J. RIEDEL and M TACKE, J. Appl. Phys, 66 (1989) 5667.
6, V RUDDY, Fiber Integrated Opt. 9 (1990) 142.
7, V. RUDDY, B. D. MACCRAITH and J, A, MURPHY, J. Appl. Phys. 67 (1990) 6070.
8. L. C. SHRIVER-LAKE,G. P. ANDERSON, L P, GOLDEN and F. S. LIGLER, Anal. Lett. 25 (1992) 1183.
9. S, R LOWRY, T. E. MAY and A. BORNSTEIN,Appl. Spectrosc. 47 (1993) 30.
10. B. D. GUPTA, A, SHARMA and C. D, SINGH, Int. J. Optoelectron. 8 (1993) 409.
tl, B. D. GUPTA and C. D. SINGH, Appl, Opt, 33 (1994) 2737.
12, B. D GUPTA, C D SINGH and A. SHARMA, Opt, Engng. 33 (1994) 1864.
13, B. D. GUPTA and C. D. SINGH, Fiber Integrated Opt. 13 (1994) 433.
14. B D. GUPTA, A. K. TOMAR and A. SHARMA, Opt. Quantum Electron. 27 (1995) 747.
15. T TAKEO and H. HATTORI, Jpn. J. AppL Phys. 21 (1982) 1509.
16. T. TAKEO and H. HATrORI, Proc. SPIE 1544 (1991) 282.
17. T. TAKEO and H. HATTORI,Appt. Opt. 31 (1992) 44.
18. A. W. SNYDER and J. D. LOVE, Optical Waveguide Theory (Chapman & Hall, London, 1983) pp. 63-88.
19, A. SAFAAI-JAZIand J V. PETERSEN, Opt, Laser Technol. 26 (1994) 399.
20, F. P. PAYNE and Z. M. HALE, Int, J. Optoelectron. 8 (1993) 743.

1639