You are on page 1of 26

Fiber and Integrated Optics

ISSN: 0146-8030 (Print) 1096-4681 (Online) Journal homepage: https://www.tandfonline.com/loi/ufio20

Luminescence-Based Optical Fiber Chemical


Sensors

P. A. S. JORGE , P. CALDAS , J. C. G. ESTEVES DA SILVA , C. C. ROSA , A. G.


OLIVA , J. L. SANTOS & F. FARAHI

To cite this article: P. A. S. JORGE , P. CALDAS , J. C. G. ESTEVES DA SILVA , C. C. ROSA ,


A. G. OLIVA , J. L. SANTOS & F. FARAHI (2005) Luminescence-Based Optical Fiber Chemical
Sensors, Fiber and Integrated Optics, 24:3-4, 201-225, DOI: 10.1080/01468030590922731

To link to this article: https://doi.org/10.1080/01468030590922731

Published online: 23 Feb 2007.

Submit your article to this journal

Article views: 195

View related articles

Citing articles: 3 View citing articles

Full Terms & Conditions of access and use can be found at


https://www.tandfonline.com/action/journalInformation?journalCode=ufio20
FIO 24(3-4) #52245

Fiber and Integrated Optics, 24:201–225, 2005


Copyright © Taylor & Francis Inc.
ISSN: 0146-8030 print/1096-4681 online
DOI: 10.1080/01468030590922731

Luminescence-Based Optical Fiber


Chemical Sensors

P. A. S. JORGE C. C. ROSA
Unidade de Optoelectrónica e Unidade de Optoelectrónica e
Sistemas Electrónicos Sistemas Electrónicos
Porto, Portugal, and Porto, Portugal, and
Faculdade de Ciências da Universidade Faculdade de Ciências da Universidade
do Porto do Porto
Porto, Portugal Porto, Portugal

P. CALDAS A. G. OLIVA
Unidade de Optoelectrónica e Universidade Nova de Lisboa
Sistemas Electrónicos Oeiras, Portugal
Porto, Portugal, and
Instituto Politécnico de Viana J. L. SANTOS
do Castelo
Viana do Castelo, Portugal Unidade de Optoelectrónica e
Sistemas Electrónicos
Porto, Portugal, and
J. C. G. ESTEVES DA SILVA Faculdade de Ciências da Universidade
Faculdade de Ciências da Universidade do Porto
do Porto Porto, Portugal
Porto, Portugal
F. FARAHI
University of North Carolina at Charlotte
Charlotte, North Carolina, USA

A scheme for the simultaneous determination of temperature and analyte concentra-


tion for application in luminescence-based chemical sensors is proposed. This scheme
is applied to an optical oxygen sensor, which is based on the quenching of the flu-
orescence of a ruthenium complex. Temperature measurement is performed using
the excitation radiation and an absorption long-pass filter. Preliminary results are
presented that show the viability of an oxygen measurement that is independent of
temperature and optical power level. The possibility of self-referenced temperature
measurements with semiconductor nanoparticles is also investigated. In order to op-
timize the sensor design, several different optical fiber probe geometries for oxygen
sensing are tested and compared, including different methods of coupling radiation

Received 30 October 2004.


Address correspondence to P. A. S. Jorge, Unidade de Optoelectrónica e Sistemas Electrónicos,
INESC Porto, Rua do Campo Alegre, 687, Porto 4169-007, Portugal. E-mail: pjorge@inescporto.pt

201
202 P. A. S. Jorge et al.

into the optical fiber system. Polyvinyl alcohol (PVA) and polyacrylamide membranes
are tested as supports for sensor immobilization in fiber-optical pH sensing devices
in aqueous solution. Some results are presented that show the feasibility of using
fiber-optical pH indicators for remote monitoring.

Keywords luminescence sensors, optical fiber, oxygen, temperature, pH, heavy


metals

Introduction
The measurement of chemical and biological parameters using optical sensors is an
expanding area of research with growing importance, especially in biomedical and en-
vironmental applications. Intrinsic immunity to electromagnetic interferences, increased
sensitivity, fast response times, and suitability to remote monitoring, when used with
optical fibers, are some of the characteristics that make optical sensors an appealing al-
ternative to conventional technologies. The combination of optical fiber technologies with
fluorescence spectroscopy techniques, together with the development of new materials for
sensor immobilization, has greatly contributed to the progress of optical chemical sensors
[1–4]. At present, with the environment and, consequently, the human health constantly
at risk, this is a very important prior area of technological development and, therefore, a
strategic area of research.
In this context, luminescence-based optical sensors are especially attractive due to
their high sensitivity and selectivity [5]. An important class of luminescent sensors is
based on the dynamic quenching of fluorescence. In this type of sensor the analyte to be
measured quenches the luminescent emission of a sensing dye. This sensing mechanism
is widely used for the determination of several analytes like oxygen, halides, and various
metals [6–8].
The intensity, I , and the excited state lifetime, τ , of the luminescent emission of
the sensing dye are both quenched in the presence of the analyte. These luminescence
parameters are related to the analyte concentration by the Stern-Volmer (SV) equation [9]:

I0 τ0
= = 1 + KSV [A] (1)
I τ

and

KSV = kq τ0 (2)

where I0 and τ0 are the luminescent intensity and the excited state lifetime in the ab-
sence of the quencher, respectively; KSV is the SV constant; and [A] is the analyte
concentration. The SV constant is related to the dynamic quenching rate constant, kq ,
a characteristic of the molecular species involved in the process, which describes the
dynamic collisional deactivation of electronic energy. The constant kq is directly pro-
portional to the quenching efficiency and to the diffusion coefficient. This way, dynamic
quenching mechanisms are diffusion-dependent processes that are clearly affected by tem-
perature. Therefore, knowledge of the sample temperature is necessary for the correct
determination of the analyte concentration by luminescence quenching methods.
Equation (1) shows that measurements of either the luminescence intensity or the
excited state lifetime can be used to determine the analyte concentration. However, the
direct measurement of either of these parameters has its associated problems. When
Luminescence-Based Sensors 203

intensity measurements are performed, a reference scheme is needed in order to com-


pensate for fluctuations of the excitation optical power. Otherwise, changes in the sensor
output caused by optical source instability, coupling efficiency fluctuations, leaching, and
photo bleaching of the indicator, can be mistaken with changes in analyte concentra-
tion. Conversely, direct lifetime measurements demand for high-speed electronics and
fast pulsed optical sources. These problems can be minimized using frequency domain
techniques [10].
If sinusoidal modulation is applied to the optical source, the resulting luminescent
emission will also be modulated with the same frequency. However, a phase delay, φ,
will be present between the excitation and the luminescent emission signal that is related
to the lifetime by:

tan(φ) = 2πf τ (3)

where f is the modulation frequency. Using Equations (1) and (3) the phase delay
information can easily be related to the analyte concentration. With this technique, low-
cost, high-brightness optical sources, like blue LEDs, which are suitable for excitation
of many sensing dyes, can be used in association with standard photodetection. This
very simple but powerful method, together with the development of new materials for
immobilization of the sensing chemistry in the optical fiber (sol-gels, polymers), has
greatly contributed to the development of a new class of analytical instrumentation.
Several fiber optic–based chemical sensing devices are already commercially available
[11, 12]. However, several problems still persist and there is much room for improvement.
The fact that the phase is directly proportional to the excited state lifetime implies that
it is also dependent on temperature. Although some optical temperature reference schemes
have been proposed, commercially available luminescent sensors still rely on conventional
technology for the measurement of temperature [13, 14]. This is incompatible with certain
applications where an all-optical fiber probe must be used.
Phase detection would be very sensitive and independent of the signal intensity if
the signal-to-noise ratio (SNR) was high. However, low efficiency of the luminescence
process, added to the use of broadband optical sources with standard photodetection and
optical fibers, makes it very difficult to fulfill this condition. In this perspective, the design
of the sensing system (sensing head geometry, level of optical power, etc.) is critical in
order to optimize sensor performance.
Another important class of luminescent sensors is based on fluorescent pH indi-
cators. The optical measurement of pH offers many advantages over traditional glass
electrodes. The use of sensitive luminescent methods for pH determination, when used
in combination with optical fibers, introduces the possibility of miniaturization and re-
mote monitoring. Additionally, pH indicators provide a base for other detection systems,
like biosensors and heavy metal ions. These are key features for environmental and
biomedical applications [15, 16].
An important step for implementation of optical fiber–based pH sensing is the immo-
bilization of the fluorescent indicators in a solid matrix on the fiber probe, as it determines
the sensor performance. An ideal immobilization membrane would effectively entrap the
indicator and preserve its optical properties, avoiding leaching and photobleaching, and
would establish a fast and reversible equilibrium with the aqueous environment. Some
progress has been made in this regard with the development of new sol-gel and polymer
materials; however, there is no universal solution available, since each application has its
own specific chemistry, and this is still an active area of research.
204 P. A. S. Jorge et al.

The determination of pH with measurements based on the luminescence intensity


need reference due to leaching and photobleaching and other sources of optical power
drift. Ratiometric detection is the most popular method for this purpose and it involves
the ratio of two fluorescent intensities at different wavelengths [17]. In this context, the
application of frequency domain techniques, an established technology in optical oxygen
sensing, is highly desirable. A scheme based on non-radiative energy transfer from a
luminescent long-lived ruthenium(II) complex (the donor) to a short-lived colorimetric
pH indicator (the acceptor) was proposed that allows conversion of the pH information
into a luminescence decay time signal [18].
In practical applications, the operation of pH sensors in the near infrared region of
the spectrum (NIR) is a major step toward low-cost instrumentation. Due to the develop-
ment of telecommunications, NIR optical fiber and semiconductor technology are widely
available at low cost. Additionally, excitation and detection in this wavelength range avoid
background noise from environmental and biological samples. Several NIR fluorescence
chemosensors for pH determination have already been developed [19]. However, fluores-
cence chemosensors for toxic heavy metal detection and the corresponding optical fiber
instrumentation are still in the first steps of development.
In this article several problems of luminescence-based optical fiber sensors are ad-
dressed. We propose an all-optical scheme for simultaneous determination of oxygen
concentration and of temperature that can be applied in any luminescence based sensor.
The possibility of using nanoparticle quantum dots in a luminescent probe for tempera-
ture determination is also assessed. Different methods of coupling the excitation radiation
into the fiber system are evaluated. Additionally, four different oxygen fiber probe ge-
ometries are tested and compared. Finally, the properties of some membranes for the
immobilization of pH sensors in optical fibers are addressed. Some results are presented
that show the viability of NIR optical fiber pH probes. Preliminary tests are made to
assess the possibility of adapting the pH fiber probes to the detection of metal ions.

Experimental

Temperature Measurement with a Long Pass Filter


Oxygen is a well-known quencher of many luminescent dyes. Ruthenium complexes
are specially suited for oxygen sensing applications. They are efficiently excited by
blue LEDs (470 nm), their emission (620 nm) is strongly quenched by O2 , and their
lifetimes are relatively long (0.5 to 6 µs), which make them suitable for frequency
domain applications. Additionally, the immobilization of these sensing dyes in sol-gel
and polymer matrices is a well studied and developed subject [20–22]. Some optical
solutions for the measurement of oxygen are already commercially available. For these
reasons, the temperature measurement scheme and the sensing probes geometry were
evaluated using a phase detection scheme applied to an oxygen sensor. The experimental
setup can be observed in Figure 1.
In order to obtain a temperature measurement that is independent of the oxygen
concentration, part of the excitation radiation (which is back-reflected from the sens-
ing probe) can be used. The spectral characteristics of this radiation are independent
of the oxygen concentration but can be made temperature dependent when propagating
through a colored glass long-pass filter placed in the distal end of the sensing probe.
The position of the cut-off wavelength of these filters has well-known linear temperature
dependence [23]. If the edge of the filter is made to coincide with the source spectral peak,
205
Figure 1. Experimental setup for simultaneous measurement of oxygen and temperature. Insert: the transmittance of low-pass filter for two
different temperatures and the LED emission spectrum.
206 P. A. S. Jorge et al.

the back-reflected blue radiation becomes temperature dependent and can be used to
determine temperature.
For the implementation of the proposed configuration, a high-brightness blue LED
(470 nm—Nichia) was used as excitation source and was sinusoidally modulated at
90 kHz. Light was injected into a multimode silica fiber coupler with core/cladding
diameters of 550/600 µm and a nominal 50/50 coupling ratio. The oxygen-sensing fiber
probes and the temperature-sensing fiber probe were connected to different outputs of the
fiber coupler and placed close together in an oxygen/temperature controlled environment.
For the temperature probe a mirror and a long-pass absorption filter with cut-off
wavelength of 455 nm (GG455—Schott glass) were placed in the tip of the fiber. The cut-
off wavelength of this filter is shifted to higher wavelengths with increasing temperature
by 0.08 nm/◦ C.
The luminescent signal (620 nm) from the ruthenium complex carrying the oxygen
information and the blue signal (470 nm) with the temperature information were guided
back to one of the input arms of the fiber coupler. These signals were separated at the
fiber output by a dichroic beam splitter. The luminescent signal was fed into a lock-
in amplifier and was compared with the reference signal that was used to modulate
the optical source. In these conditions the lock-in output was an amplitude/phase signal
proportional to the oxygen concentration. Conversely, the spectrum of the blue signal was
spatially distributed by a diffraction grating. Two photodiodes were then used to obtain
two signals, centered at 470 nm (P1 ) and at 500 nm (P2 ), with a bandwidth of ≈3 nm.
While signal P1 overlapped with the edge of the filter and became strongly temperature
dependent, signal P2 overlapped with a transmission maximum of the filter and remained
independent of temperature. This way, a signal that was independent of optical power
drifts and contained the information of the sensing head temperature could be obtained
with the ratio P1 /P2 .

Temperature Measurement with Quantum Dots


Semiconductor nanoparticles promise to be a powerful tool with potential to solve
many problems in luminescence-based applications [24–26]. Quantum dots (QDs) are
nanometer-sized particles of semiconductor material (e.g., CdSe). Due to the quantum
confinement effect, they have unique properties. QDs have relatively high quantum yields,
narrow fluorescence spectrum (FWHM of 30 nm), very broad absorption, outstanding
photostability, and the ability to tune optical properties by simply changing the size of
the particles. At present, they are readily available in a variety of emission wavelengths
(from 350 to 2000 nm) and can be provided in the form of colloidal suspensions suitable
to be immobilized in polymers or sol-gel–based materials.
QDs have been reported to have favorable characteristics to be used as temperature
probes in optical sensing applications [27]. It was observed that the wavelength max-
imum, width, and intensity of the QDs photoluminescence emission were all strongly
temperature dependent.
The possibility to perform temperature measurements independent of the optical
power level in the system was tested. Due to the presence of a wavelength shift, a sim-
ple detection scheme can be implemented in order to obtain self-referenced temperature
measurements with QDs. If two signals, S1 and S2 , corresponding to two narrow spec-
tral windows on opposite sides of the spectrum are normalized according to (S1 − S2 )/
(S1 + S2 ), the resulting normalized output will be proportional to temperature and in-
dependent of the optical power level in the system. In order to test the applicability of
Luminescence-Based Sensors 207

QDs as temperature probes in the context of luminescent chemical sensors, core-shell


CdSe-ZnS quantum dots with emission wavelengths ranging from 517 nm to 610 nm
were immobilized in non-hydrolytic sol-gel.
Several samples were excited by a blue LED (λmax at 470 nm, from Nichia). Optical
power was carried from the LED to the sample through a 4-mm-diameter fiber bundle.
An Ocean Optics S2000 miniature spectrometer (with a 600-µm fiber cable), connected
to a PC, was used for detection of the photoluminescence emission. The temperature
of each sample was controlled using a peltier cooling device. The emission spectrum of
each sample was recorded after thermal equilibrium was reached. To assess independence
of optical power drift, temperature measurements were performed using the proposed
processing scheme for different levels of the LED optical power.

Oxygen Fiber Probes


For this particular application, Tris(2,2 -bipyridine) ruthenium(II) chloride hexahydrate
[Ru(bpy)3 ] was used as the oxygen sensor with a peak emission at 620 nm and a lifetime
of approximately 1 µs. The sensing dye was used to dope a tetraethoxysilane (TEOS)-
based sol-gel solution. Several fiber probes were prepared by dip-coating silica fibers with
core/cladding ratio of 550/600 µm into the resulting solution. After adequate thermal
treatment, probes coated with a thin film of porous glass were obtained. The detailed
preparation procedures are described elsewhere [28]. Four different oxygen sensing probe
geometries were tested. To choose the best performing sensing head, a simplified setup
was used: the temperature probe was removed and the corresponding output of the fiber
coupler was immersed in index-matching gel. Also, the detection of the luminescent
signal was made using a photodetector with two cascaded colored glass long-pass filters
(OGG 550: cutoff at 550 nm, extinction coefficient 10−3 ).
The four geometries tested can be observed in Figure 2. In all the configurations
the excitation radiation and the luminescent signal were carried by the same fiber. The
first configuration, Figure 2a, is an extrinsic sensor where the sol-gel sensing film was
deposited on a glass slide that was placed inside the O2 chamber. The fiber tip, in direct
contact with the ruthenium-doped film, was used to excite and collect the luminescence
signal. All the other configurations were intrinsic: the sensing films were deposited di-
rectly onto the fiber tips. In configuration (b) the sensing film was deposited on a polished

Figure 2. Fiber probe geometries: (a) glass slide; (b) fiber tip; (c) uncladded fiber tip; (d) fiber
taper.
208 P. A. S. Jorge et al.

fiber tip. In configuration (c), the cladding of a 2-cm section of the fiber tip was removed
by chemical etching with hydrofluoric acid (HF) and then coated. Finally, in configura-
tion (d), 2 cm of the fiber tip were tapered (550 µm at the base, 300 µm at the tip) by
slow dipping in HF, and then coated.
In order to compare the sensitivity of the different geometries, some parameters were
evaluated. One of them was the quenching response defined by:
IN2 − IO2
Q= (4)
IN2
where IN2 and IO2 are the luminescence intensities in N2 - and O2 -saturated atmospheres,
respectively. To assess phase sensitivity independently of the modulation frequency, the
lifetime difference was measured:

τ = τN2 − τO2 (5)

where τN2 and τO2 are the excited state lifetimes in saturated atmospheres of N2 and O2 ,
respectively. Lifetime was estimated by plotting tan[φ] (Equation (3)) as a function of
modulation frequency. Additionally, the amount of detected luminescence signal, Pdet ,
and the back-reflected excitation radiation, Pblue , were measured and compared.

Coupling Systems
Using the same simplified setup, the efficiency of three different techniques of coupling
LED radiation into the fiber system were tested and compared: a standard microscope
objective (10×) system; a 10-mm-diameter ball lens; and direct butt-coupling of the fiber
to an LED whose encapsulation was partially removed and polished.
The configuration with best performance was used to test the simultaneous measure-
ment of oxygen and temperature.

pH Fiber Probes
The properties of two different membranes for immobilization of pH sensors were tested
and compared: polyvinyl alcohol (PVA) (cross-linked with glutaraldehyde) and polyacry-
lamide gel. The detailed preparation of these membranes and its doping with pH sensors
is described elsewhere [17, 29].
Both membranes were doped with 5(6)-carboxynaphtofluorescein (CNF). This pH
indicator was chosen for its versatility. It has a fluorescent response on the physiological
pH range (6.0 to 8.0), which is of particular interest to biomedical applications, it can
be indirectly used as sensor for metal ion recognition when it is coupled to a chelating
agent, it has absorption bands at 460 nm and 630 nm that are suitable to LED and diode
laser excitation, and it is a dual emission dye suitable for ratiometric detection (it emits
at 550 nm in acidic form, and at 700 nm in basic form) [17].
Immobilization was tested both in extrinsic (configuration (a)) and intrinsic (config-
uration (c)) fiber probes. For the extrinsic membranes the polymerization process was
executed in small molds. Conversely, the PVA membrane was applied to the fibers by dip
coating. The polyacrylamide membrane was covalently bounded to the silica core of the
fibers. The fibers were immersed into a glass capillary, with internal diameter of 650 µm,
filled with polymerization solution, and illuminated with a blue light (430 nm) (Hilux Cur-
ing Light 200). After photo-induced polymerization the membrane was bounded to the
silanized core but not to the glass capillary, which was easily removed.
Luminescence-Based Sensors 209

Figure 3. Experimental setup used to test the pH sensing membranes.

The membranes were tested in the experimental setup shown in Figure 3. The fiber
probes were connected to a multimode fiber coupler (50/50 coupling ratio). An He-Ne
laser (633 nm) was used as excitation source. For the detection of the luminescent signal
the output fiber was connected to a miniature CCD spectrometer (USB2000 Ocean Optics).
A magnesium ion [Mg(II)] sensor was tested by co-immobilization on PVA of the
CNF sensor with Eriochrome Black T (EBT). The absorption spectrum of EBT overlaps
with the absorption spectra of CNF. When Mg(II) is present the absorption spectrum of
Mg(II)-EBT shifts to shorter wavelengths, provoking the increase of the CNF fluores-
cence. Preliminary tests were performed with the setup in Figure 3 by dipping the sensing
fiber into the modified solution in order to access the response of the luminescent signal
to the presence of Mg(II).

Results and Discussion

Coupling Systems
The first step in assembling the experimental setup was to optimize the coupling system.
The results obtained with the different configurations tested can be observed in Table 1.
The lowest efficiency (2.6%) was obtained with a traditional coupling method (col-
limating and focusing with 10× microscope objectives). The main reason for this poor

Table 1
Power coupling efficiencies of the different techniques

Coupling system Coupling efficiency (%) Alignment

Microscope 2.6 Very sensitive


Ball lens 8.0 Sensitive
Butt coupling 550 µm core 8.0 Straightforward
Butt coupling 1 mm core 20.0 Straightforward
210 P. A. S. Jorge et al.

result was the high divergence of the optical source (≈15◦ ) and the high insertion losses of
the optical system. The efficiency of this configuration was very sensitive to the position
of the optical fiber relative to the focus of the optical system.
A higher efficiency was obtained with a 10-mm ball lens (8%). This improvement
was mainly due to the reduced insertion loss of a single glass sphere when compared
with the two microscope objectives. This system was less sensitive to misalignment when
compared with the previous one.
After polishing the LED encapsulation, the semiconductor emitting surface was only
100 µm away from the fiber tip and an efficiency of ≈8% was achieved by direct
butt-coupling. In this setup, fiber alignment was straightforward. After polishing, the
LED divergence angle increased to 45◦ , however, the proximity to the emitting area was
also increased. As a result, the power density reaching the fiber top was much higher.
Considering that the LED emitting area (≈900 µm diameter) was bigger than the fiber
core (≈550 µm diameter), better results are expectable for the case where larger fiber
core diameters are used. Indeed, for a fiber with a 1-mm core diameter, light coupling
efficiencies of ≈20% were achieved.
In the experimental setup of Figure 1 the butt-coupling configuration was used with
a 550-µm core fiber.

Oxygen Fiber Probes


The glass slide (configuration (a)) was used to tune the sol-gel properties. These properties
have a significant influence on the final characteristics of the sensing films. The resulting
glass porosity has a known dependence on the solution aging time and determines the
oxygen accessibility to the sensing dye and thus the sensitivity of the resulting sensor [28].
After determining the more suitable sol-gel parameters, several probes of geometries (a),
(b), (c), and (d) were coated under the same conditions: sol-gel solution aging time of
more than 48 h; dip-coating speed of 3 mm/s; and curing at ambient temperature for
24 h. The merit parameters of the probes with best characteristics are shown in Table 2.
Observation of the films obtained with configuration (a) under an optical microscope
revealed uniform films with no cracks. A perfilometer was used to measure the films’
thicknesses; thicknesses of 600 nm to 800 nm were found. The response times of the
sensing films when they underwent O2 /N2 saturation cycles were the same and were
approximately 9 s (framed to the 10% and 90% reference levels). In configuration (a)
the detected luminescence emission was very weak (Pdet ≈ 0.36 nW), and due to a low
signal-to-noise ratio the phase signal was unstable.

Table 2
Comparative results for the tested fiber probe configurations

Response
Geometry Q (%) τ (ns) time (s) Pdet /Pdet(a) Pblue /Pblue(b)

Slide 40 176 9 1.00 —


Tip 34–44 246 20–60 1.81 1.00
Uncladded 55 250 20 2.00 0.16
Taper 65 350 11–13 3.58 0.15
Luminescence-Based Sensors 211

With configuration (b) no uniform thin films could be obtained because they were
thicker in the tip than in the lateral surface, a consequence of the fabrication process.
However, the film deposited smoothly in the fiber lateral surface and in the tip the
film structure was also relatively flat. The response time of several fiber tips with this
geometry, with the same sol-gel parameters, varied from 20 s to 60 s. This shows that,
with dip-coating, it is not possible to repeatedly obtain films of uniform thickness at the
fiber tip. In spite of that, due to the increased thickness and a more efficient luminescence
coupling, a significant increase in SNR was observed. In comparison to configuration (a)
the value of detected luminescence, Pdet , increased by a factor of 1.8. Although the
coating parameters were similar in all geometries, smaller values for Q and τ were
obtained with this configuration. This can be related to film thickness and lower oxygen
accessibility.
In configuration (c), uniform thin films were obtained in the side of the fiber where
the cladding was removed. However, in the tip of the fiber the resulting films were
non-uniform and too thick. Response time of approximately 20 s was obtained. This
probably results from the averaging of a faster response from the side of the fiber with a
much slower response from the top. Also, some improvement in Q and τ parameters
was observed. However, no significant enhancement in luminescence coupling efficiency
was observed in comparison with configuration (b). This may indicate that, in spite of
the much larger emission area, the contribution from the film in the fiber side to the
luminescence signal is small due to poor guiding from this region.
With configuration (d), uniform thin film was obtained in the side as well as at
the fiber tip due to its geometry. Shorter response times clearly indicate that the overall
film thickness is similar to the one obtained with a glass slide. This is also shown by the
increase in Q and τ which also denote better quenching efficiency and sensitivity. With
this configuration, several sensing probes with very similar parameters were obtained,
showing good reproducibility. A great improvement in the SNR was registered in the
tests performed with this configuration, and Pdet was increased by a factor of 3.5 relative
to the glass slides. In Figure 4 a photograph of a fluorescent tapered fiber in a 100% N2
environment is shown.
In the last column of Table 2, the degree of detected back-scattered blue radiation
in each configuration (Pblue ) is presented. Configuration (b) shows a 6-fold increase in
the level of back-scattered radiation when compared with configurations (c) and (d).
Considering the respective levels of detected radiation (Pdet ) this implies that even after
attenuation by a factor of 1000 (obtained with a single filter), a significant amount of
blue light is present with the luminescent signal. The attenuated power at the excita-
tion wavelength (Pblue × 10−3 ) is still 18% of Pdet in configuration (b), 2.6% of Pdet in
configuration (c), and 1.4% of Pdet in configuration (d). Straightforward numerical calcu-
lations verify that this will introduce phase errors of 5.5, 0.7, and 0.4%, respectively [28].
In the experimental setup used to test the fiber probes, this problem was avoided by using
two cascaded long-pass filters corresponding to a 10−6 attenuation factor. In this case
the error dropped below a negligible 0.005% in all configurations. In general, double
filtering is not a desirable solution from a practical point of view, but it can be bypassed
by using filters with higher performance. Anyway, the desirable solution to this problem
is to use sensing probe geometries, which minimize the level of backscattered radiation.
We should note that when the back-scattered blue radiation is reduced, a more efficient
and uniform excitation in the sensing device is automatically accomplished.
The results obtained clearly show that tapering the fiber is a right step to producing
an optimal sensing probe and that both the intensity and phase measurement schemes will
212 P. A. S. Jorge et al.

Figure 4. Tapered fiber in a 100% N2 environment.

benefit from its properties. Therefore, a tapered fiber was used to test the simultaneous
measurement of oxygen and temperature (setup of Figure 1).
The intensity and the phase of the sensor output signal were recorded while the sens-
ing head was submitted to successive N2 /O2 saturation cycles. The respective results can
be observed in Figures 5a and 5b. From these results, a response time of approximately
10 s can be estimated. Also, it can be observed that the phase signal is more unstable,
indicating the need to improve the SNR in order to take full advantage of the frequency
domain detection scheme.
The sensor calibration curve, i.e., the Stern Volmer plot, was obtained by performing
measurements of the system response to several values of oxygen concentration. The
corresponding plot can be observed in Figure 6. The nonlinearity of the response curve
indicates that the sensing dye is not uniformly distributed within the sensing film and
that sites with different oxygen accessibilities coexist in the sol-gel matrix. This behav-
ior can be more accurately described with a more complex bi-exponential model [30].
Additionally, an adequate thermal treatment can render the sensor response more linear
at the expense of loosing some quenching efficiency.
Luminescence-Based Sensors 213

(a)

(b)

Figure 5. Response of tapered fiber to O2 /N2 saturation cycles: (a) luminescence intensity
response; (b) phase response.
214 P. A. S. Jorge et al.

Figure 6. Stern-Volmer plot obtained from the phase response of the tapered fiber.

In order to assess the sensitivity of the oxygen sensing probe to temperature fluctu-
ation, the output phase and the luminescent intensity were measured for several values
of temperature with a constant oxygen concentration of 21%. The results can be seen in
Figures 7a and 7b.
It is clearly shown that both the lifetime and the luminescence intensity of the sensing
dye depend not only on the oxygen concentration, but also on the temperature. In order
to extract the information about the oxygen concentration from the luminescent signal,
the temperature must be simultaneously determined.

Temperature Measurement with a Long Pass Filter


The configuration of Figure 1 was used to measure the temperature using the spectral
information of the blue radiation reflected by the temperature probe. The measurements
were performed under a constant oxygen concentration of 21%. As it can be seen from
Figure 8, the spectral characteristic of the reflected excitation radiation is rendered tem-
perature dependent by the presence of the long-pass filter. The temperature dependence
is maximum around 465 nm, nearby the LED spectral peak and coinciding with the band
edge of the filter. This dependence vanishes for longer wavelengths (>480 nm), where
the filter transmittance is maximum (insert in Figure 1). As the temperature increases, the
band edge of the filter shifts toward the red (0.08 nm/◦ C), and less radiation is transmitted
at 465 nm (insert in Figure 8).
In order to evaluate the influence of optical power level in the temperature mea-
surement, the outputs P1 and P2 were measured for two different radiation coupling
conditions, identified by HIGH and LOW. In the LOW condition the optical power
reaching the detector was 20% less than in HIGH condition. In Figure 9 the temperature
dependency of signal P1 at two different conditions, HIGH and LOW, are shown.
Luminescence-Based Sensors 215

(a)

(b)

Figure 7. Response of oxygen probe to temperature: (a) luminescence intensity response; (b) phase
response.
216 P. A. S. Jorge et al.

Figure 8. Spectral response of temperature probe (temperature range 19◦ C, 54◦ C). Insert: temper-
ature dependence of the return optical power at 465 nm.

Figure 9. Temperature response of P1 detected at 470 nm for 100% optical power (HIGH) and
80% optical power (LOW) (temperature range 15◦ C, 45◦ C).
Luminescence-Based Sensors 217

It can be seen that the optical power (P1 ) of the LED at the central wavelength is
a linear function of temperature. However, it is also clear that this measurement is not
independent of the optical power level (the 20% decrease in optical power can clearly
be observed in the resulting plots).
To achieve immunity to optical power fluctuations, the curves in Figure 9 were
divided by the optical power, P2 , which has negligible temperature dependence. To op-
timize the processing scheme, the ratio P1 /P2 was performed at two different reference
wavelengths. P2 was detected at 496 nm (Ratio A) and at 500 nm (Ratio B). The corre-
sponding results are shown in Figures 10a and 10b.
Comparing the processed signal with the original data, it can be observed that an
optical power drift of approximately 20% in P1 is reduced to a 4% variation in Ratio B
(Figure 10b), and to a 1% variation in Ratio A (Figure 10a). Both the 496 nm and
the 500 nm optical power signals have negligible temperature dependence. However, for
higher wavelengths the optical power available decreases very fast and noise is introduced
in the ratio operation. If the operation (P1 − P2 )/(P1 + P2 ) is performed instead, noise
is slightly decreased and the 20% change in P1 induces a 0.7% change in Ratio A and
1.1% in Ratio B. This shows that data dispersion comes from a low SNR and an increase
in the LED output power will benefit the sensor performance. Nevertheless, these results
indicate that the reference wavelength can be tuned in order to reduce further the residual
dependence of the system performance to the optical power fluctuations.

Temperature Measurement with Quantum Dots


Several samples of CdSe-ZnS QDs immobilized in sol-gel were submitted to temperature
changes while their spectral characteristics were monitored. In Figure 11, the spectral
behavior of the fluorescent emission of a thin sol-gel film (∼5 µm) doped with CdSe-Zn
QD (emission peak at 610 nm) can be seen. The sample temperature was increased from
11 to 48◦ C.
Data analysis showed that both the wavelength and the fluorescence intensity changed
in a linear and reversible way with temperature.
The behaviors of different samples (different emission wavelength, different sol-gel
parameters) were very similar. The rate of change in the intensity with temperature varied
from −0.7%/◦ C to −1.6%/◦ C. On the other hand, the rate of increase in the wavelength
with increase in temperature in all samples is approximately 0.2 nm/◦ C. This indicates
that the differences in the intensity response could be due to different immobilization
environments, distinct film thickness, etc. Because the wavelength dependencies of the
fluorescence signals to temperature are identical, it is reasonable to assume that the
intrinsic responses of the various QD samples to temperature are similar. Also, deviations
in linearity were observed in the intensity response (Figure 12a) for the lower temperature
range, which were due to some water condensation on the samples surface, which changed
the coupling conditions. This was not observed in the wavelength response, which was
linear through all temperature intervals. Therefore, this mechanism is particularly well
suited to perform self-referenced temperature measurement.
To test the self-reference scheme, two signals, S1 and S2 , corresponding to the
spectral windows 595–600 nm and 620–625 nm were detected using a CCD spec-
trometer and adequate software. The two signals were then normalized by calculating
(S1 −S2 )/(S1 +S2 ). Temperature measurements were performed using three different lev-
els of the LED optical power (PLED ): 100% PLED , 90% PLED , and 80% PLED . This was
achieved by changing the LED drive current. Figure 12a shows the QD luminescence
218 P. A. S. Jorge et al.

(a)

(b)

Figure 10. Temperature dependence of P1 /P2 , for 100% optical power (HIGH) and 80% optical
power (LOW). (a) P1 /P2 with P2 detected at 496 nm; (b) P1 /P2 with P2 detected at 500 nm.
Luminescence-Based Sensors 219

Figure 11. Temperature behavior of the emission spectrum of CdSe-ZnS QDs immobilized in
sol-gel (temperature range 11◦ C, 48◦ C).

intensity response to temperature changes for three different power levels. The corre-
sponding normalized outputs can be seen in Figure 12b.
Comparison of the plots in Figures 12a clearly shows that the luminescent emission
response strongly depend on the change in the excitation optical power. This possible
source of error can be eliminated by applying the proposed normalization scheme, Fig-
ure 12b. The dispersion of the normalized signal was within the noise limit of the system.
Although the emission spectrum of the QDs used to obtain the results in Figures 11 and 12
overlaps with the emission of most common oxygen sensors (620 nm), very similar be-
havior was observed with other QDs with lower emission wavelengths. Therefore, these
results confirm that QDs could be used to configure self-referenced temperature probes
for many chemical sensors.

pH Fiber Probes
Both the PVA and the polyacrylamide membranes showed good compatibility with the
CNF sensor.
When immersed into different buffer solutions the polyacrylamide sensing membrane
showed a fast and reversible response to pH changes in the surrounding medium. Both
the color and the fluorescent emission of CNF changed with change in pH. Very little
leaching was observed and it occurred mainly in the basic form. Due to covalent bonds
the adhesion to the fiber core was relatively strong. However, in its hydrated form the
membrane was soft and could be easily damaged. Additionally, due to the fabrication
procedure the end of the fiber tip was not efficiently coated. Hence, a very low signal-to-
noise ratio was obtained with the intrinsic configuration. Tapered fiber design is expected
to improve both the SNR and adhesion to the fiber tip.
220 P. A. S. Jorge et al.

(a)

(b)

Figure 12. Temperature response of QDs for three different levels of excitation optical power
(PLED = 100, 90, and 80%), for a temperature range from 11◦ C to 48◦ C. (a) Luminescence
intensity response; (b) normalized signal response.
Luminescence-Based Sensors 221

With the PVA membrane, a very fast and reversible response to pH was observed.
There was no apparent leaching of the sensor into the buffer solutions. The membranes
obtained were very robust and elastic and not easily damaged. However, adhesion to
the fiber core was very weak. Some modifications are in progress to improve the ad-
hesion of membrane to the fiber. In this work, only extrinsic membranes were tested.
Figure 13 shows the spectral response of CNF immobilized in PVA when the membrane
was immersed into solutions of pH 6.5, 7.5, and 9.9, respectively.
The resulting plot indicates an increase in the luminescence signal as the pH of the
buffer solution increases. The emission peak of the CNF fluorescence is around 700 nm
and can easily be discriminated from the laser emission. However, it can be seen that the
back-reflection of the laser is stronger than the fluorescence signal from CNF. The use of
optical high-pass filters may be needed in order to increase noise rejection and improve
the sensor signal-to-noise ratio. Also the use of tapered fiber probes will improve SNR.
Figure 14 shows a plot of intensity of the fluorescence peak as a function of the
solution pH over a pH range of 3.9–9.9.
These results indicate a good compatibility between the CNF sensor and the PVA
as immobilization membrane. Currently, work is in progress to improve adhesion to the
fiber to fabricate an intrinsic fiber probe.
The CNF sensor was combined with a chelatometric indicator (Eriochrome Black T,
EBT) in order to become responsive to metal ions. The sensitivity of the modified sensor
to the presence of magnesium was tested in pH = 9 buffer solutions. The spectral response
of the modified sensor to increased concentration of Mg(II) can be observed in Figure 15.
The results show an enhancement of the luminescent response that is proportional
to the concentration of Mg(II) in the solution (insert in Figure 15). The enhancement is
due to reduction of the interference of the absorption band of free EBT. This shows that
modified pH indicators could be used to detect heavy metal ions in aqueous environments.
Work is in progress to improve the sensitivity to Mg(II) and to other metal ions.

Figure 13. Emission spectra of CNF immobilized in PVA as function of pH (excitation at 633 nm).
222 P. A. S. Jorge et al.

Figure 14. Response of the PVA fiber probe as a function of solution pH (from 3.9 to 9.9).

Figure 15. Response of the modified CNF to the presence of increasing concentrations of
magnesium in sample solutions. Insert: Integrated luminescence intensity as function of Mg2+
concentration.
Luminescence-Based Sensors 223

Conclusion
Several technological challenges to optimize the luminescence-based optical fiber sensors
were addressed.
Four different fiber probes were tested using a luminescence phase detection scheme.
The tapered fiber tip structure showed the best performance and, in particular, the results
obtained from this design were very reproducible. This sensor design provides larger
excitation efficiency and reduced level of back-scattered excitation radiation. Light was
coupled into the fiber with higher efficiency and alignment constraints were more relaxed
with a butt-coupling geometry. The performance of the sensing system was assessed using
both phase and intensity measurements. An all-optical configuration for simultaneous
measurement of oxygen concentration and temperature was demonstrated. Encouraging
preliminary results were obtained. Further integration of the sensing head can be achieved
if the mirror and the absorption filter are placed on the top of the oxygen fiber probe. A
single fiber will then be enough and part of the fluorescence radiation that was lost in the
previous configuration can now be detected. Controlling the LED operating temperature,
its central wavelength can be tuned within a few nanometers. This can be used to optimize
the overlap of the filter transfer function and of the LED spectrum increasing the efficiency
of the optical power reference scheme. The principle of operation of this configuration
can easily be extended to other optical fiber sensors that are based on the quenching
of fluorescence. Preliminary results showed that QDs can be used to design a self-
referenced temperature probe in fluorescence-based sensors. Their availability in a wide
range of wavelengths opens the possibility of using them with different sensors and in
multiplexing applications.
The PVA and acrylamide as immobilization membranes together with CNF as sensing
dye present promising characteristics for environmental and biomedical applications. With
excitation at 633 nm, only one emission peak (700 nm) of CNF is excited, and there is no
possibility of implementing a ratiometric reference scheme. However, work is in progress
to implement frequency domain interrogation techniques and eliminate the dependency
to optical power level.
The results obtained during this course of research clearly show that luminescence
technique offers a practical path with a great potential to develop many novel fiber optic
sensing systems for environmental and biomedical applications.

Acknowledgments
Pedro Jorge acknowledges the financial support of FCT (Fundação para a Ciência e a
Tecnologia) and FLAD (Fundação Luso Americana para o Desenvolvimento). Financial
support from Fundação para a Ciência e Tecnologia (Lisbon) (FSE-FEDER) (project
POCTI/QUI/44614/2002) is also acknowledged.

References
1. Wolfbeis, O. S. 2004. Fiber optic chemical sensors and biosensors. Anal. Chem. 76:3269.
2. McDonagh, C. M., P. Bowe, K. Mongey, and B. D. MacCraith. 2002. Characterisation of
porosity and sensor response times of sol-gel–derived thin films for oxygen sensor applications.
Journal of Non-Crystalline Solids 306:138.
3. Kulmala, S., and J. Suomi. 2003. Current status of modern analytical luminescence methods.
Analytica Chimica Acta 500:21.
224 P. A. S. Jorge et al.

4. Holst, G., and B. Mizaikoff. 2002. Fiber optic sensors for environmental applications. In
Handbook of Optical Fibre Sensing Technology. J. M. López-Higuera (ed.). Great Britain:
John Wiley & Sons, 729–755.
5. Orellana, G. 2004. Luminescent optical sensors. Anal. Bioanal. Chem. 379:344.
6. Geddes, C. D. 2001. Optical halide sensing using fluorescence quenching: Theory, simulations
and applications—A review. Measurement Science and Technology 12:R33.
7. McEvoy, A. K., C. M. McDonagh, and B. D. MacCraith. 1996. Dissolved oxygen sensor
based on fluorescence quenching of oxygen-sensitive ruthenium complexes immobilized in
sol-gel–derived porous silica coatings. Analyst 121:785.
8. Zhoua, L.-L., H. Sun, X.-H. Zhang, and S.-K. Wu. 2004. An effective fluorescent chemosensor
for the detection of copper(II). Spectrochimica Acta Part A 61:61–65.
9. Lakowicz, J. R. 1999. Principles of Fluorescence Spectroscopy, 2d ed. New York: Kluwer-
Plenum.
10. Lakowicz, J. R., and I. Gryczynski. 1991. Frequency-domain fluorescence spectroscopy. In
Topics in Fluorescence Spectroscopy. New York: Plenum, 293–335.
11. Ocean Optics, Inc. 2004. TITLE. Available at http://www.oceanoptics.com.
12. PreSens (Precision Sensing GmbH). 2004. TITLE. Available at http://www.presens.de.
13. Liao, S.-C., Z. Xu, J. A. Izatt, and J. R. Alcala. 1997. Real-time frequency domain temperature
and oxygen sensor with a single optical fiber. IEEE Transactions on Biomedical Engineering
44(11):1114.
14. Klimant, I., M. Kuhl, R. N. Glud, and G. Holst. 1997. Optical measurement of oxygen and
temperature in microscale: Strategies and biological applications. Sensors and Actuators B
38–39:29.
15. Lin, J. 2000. Recent development and applications of optical and fiber-optic pH sensors. Trends
in Analytical Chemistry 19(9):541.
16. Wolfbeis, O. S., I. Klimant, T. Werner, C. Huber, U. Kosch, C. Krause, G. Neurauter, and
A. Durkop. 1998. Set of luminescence decay time based chemical sensors for clinical appli-
cations. Sensors and Actuators B 51:17.
17. Song, A., S. Parus, and R. Kopelman. 1997. High-performance fiber-optic pH microsensors
for practical physiological measurements using a dual-emission sensitive dye. Anal. Chem.
69(5):863.
18. Koronczia, I., J. Reichert, H.-J. Ache, C. Krause, T. Werner, and O. S. Wolfbeis. 2001. Sub-
micron sensors for ion detection based on measurement of luminescence decay time. Sensors
and Actuators B 74:47.
19. Fabian, J., H. Nakazumi, and M. Matsuoka. 1992. Near-infrared absorbing dyes. Chem. Rev.
92:1197.
20. Murtagh, M. T., D. E. Ackley, and M. R. Shahriari. 1996. Development of a highly sensitive
fibre optic O2 /DO sensor based on a phase modulation technique. Electronics Letters 32:477.
21. MacCraith, B. D., G. O’Keeffe, C. M. McDonagh, and A. K. McEvoy. 1994. LED-based fibre
optic oxygen sensor using sol-gel coating. Electronics Letters 30:888.
22. Xu, W., R. C. McDonough, III, B. Langsdorf, J. N. Demas, and B. A. DeGraff. 1994. Oxygen
sensors based on luminescence quenching: Interactions of metal complexes with the polymer
supports. Analytical Chemistry 66:4133.
23. Bass, M. 1995. Handbook of Optics, Vol. 1. LOCATION: MacGraw-Hill.
24. Murphy, C. J. 2002. Optical sensing with quantum dots. Analytical Chemistry 74:520A.
25. Sutherland, A. J. 2002. Quantum dots as luminescent probes in biological systems. Current
Opinion in Solid State and Materials Science 6:365.
26. Jaiswal, J. K., and S. M. Simon. 2004. Potentials and pitfalls of fluorescent quantum dots for
biological imaging. Trends in Cell Biology: Article in press.
27. Walker, G. W., V. C. Sundar, C. M. Rudzinski, A. W. Wun, M. G. Bawendi, and D. G. Nocera.
2003. Quantum-dot optical temperature probes. Appl. Phys. Letters 83(17):3555.
28. Jorge, P. A. S., P. Caldas, C. C. Rosa, A. G. Oliva, and J. L. Santos. 2004. Optical fiber probes
for fluorescence based oxygen sensing. Sensors and Actuators B 103:290.
Luminescence-Based Sensors 225

29. Wangbai, M., Z. Zhujun, and W. R. Seitz. 1989. Poly(vinyl alcohol)-based indicators for optical
pH and Mg(II) sensing. In Chemical Sensors and Microinstrumentation, no. 403. LOCATION:
American Chemical Society.
30. Boaz, H., and G. K. Rollefson. 1950. The quenching of fluorescence: Deviations from the
Stern-Volmer law. Journal of American Chemical Society 72:3435.

Biographies
Faramarz Farahi has been working in the area of optical devices for 19 years. He
has more than 130 papers and several patents in this area. He is currently a professor of
physics and optical science in the University of North Carolina at Charlotte.
José L. Santos graduated in applied physics (optics and electronics) from the Univer-
sity of Porto, Portugal, in 1983. In 1993 he was awarded a Ph.D from the same university
for research in fiber optic sensing. He is an associate professor of the physics department
of the University of Porto and is in charge of the Optoelectronics and Electronic Systems
Unit of INESC, Porto. His main research interests are in the optical fiber sensing field.
He is a member of OSA and SPIE.
Joaquim C. G. Esteves da Silva graduated in chemistry (inorganic and environ-
mental) from the University of Porto (1985), with a Ph.D. from the same University
in 1994. Presently he holds the position of associate professor of the chemistry depart-
ment of the University of Porto. Currently, his research interests focus in luminescence
spectroscopy (fluorescence, bioluminescence, and chemiluminescence) and in the devel-
opment of new luminescent methodologies for the analysis of chemical, environmental,
and biochemical systems.
Abel Oliva received his agricultural engineering degree from the University of
Buenos Aires in 1984 and his Ph.D. at the Hohenheim Universitaet, Stuttgart, Germany,
in 1989. He started a biosensors laboratory in the Instituto de Tecnologia Química e
Biológica, of the Universidade Nova de Lisboa/Portugal in 1992. His research inter-
ests and scientific activity include optical immunosensors and immunoassays for clinical
diagnostics and bioprocess monitoring.
Carla Carmelo Rosa received her technological physics engineering degree (1996)
and electrical engineering and computers sciences M.Sc. (1999) from Instituto Superior
Técnico, Lisbon, Portugal. Presently in the physics department of the sciences faculty
of Universidade do Porto, she is working in her Ph.D. in the Optoelectronics Group
of INESC-Porto. Her research interest are on fiber-based sensors with applications in
bio-medical fields, in particular low-coherence interferometry techniques and biosensor
development.
Pedro Jorge graduated in applied physics (optics and lasers) from the University
of Minho in 1996. He obtained his M.Sc. degree in lasers and optoelectronics from
the University of Porto in 2001. Currently he is a researcher in the Optoelectronics
and Electronic Systems Unit at INESC Porto working toward his Ph.D. on optical fiber
sensors based on fluorescence spectroscopy in the physics department of the University
of Porto.
Paulo Caldas graduated in applied physics (optics and lasers) from the University
of Minho in 1999. He received the M.Sc. degree in lasers and optoelectronics in the
physics department of the University of Porto in 2003. Currently he is a researcher in the
Optoelectronics and Electronic Systems Unit at INESC Porto and is teaching physics at
the Polythenic Institute of Viana do Castelo. He has been working in the area of optical
fiber sensors and biosensors.

You might also like