journa l h om epage: www.elsev ier .com/ locate /snb
arbon nanotube based multifunctional flame sensor
umit Mohanty, Abha Misra ∗
epartment of Instrumentation and Applied Physics, Indian Institute of Science, Bangalore, Karnataka 560012, India
r t i c l e i n f o
rticle history:eceived 25 August 2013eceived in revised form 19 October 2013ccepted 6 November 2013
a b s t r a c t
Carbon nanotubes (CNT) due to its multifunctional characteristics has been presented as a flame sensor bycombining both radiation and chemical sensitivity. Chemical functionalization enhances the sensitivityof CNT sensor toward any chemical modifications that are induced by the flame. Response of the sensoris revealed to be dependent on the measurement direction (longitudinal and transverse) as well as the
vailable online 15 November 2013
eywords:arbon nanotubeslameensorunctionalization
radiation intensity. A nonlinear relation between the sensitivity and its distance from the source is used tocalibrate the intensity of the flame. The present method allows a simpler approach for the flame detectionby utilizing a calibration scheme to operate at any particular bias current and tune its sensitivity withrespect to any working distance at a particular bias current.
Carbon nanotubes (CNT) have been extensively used in thehrust areas of material research due to its extraordinary properties,ecause of which CNT has also replaced various conventional mate-ials used in diverse applications. In addition, CNT has presented anbility to tune its conductance under different experimental con-itions, for example, upon being irradiated with electromagneticadiations of significant intensities, etc. [1,2]. It has been reportedhat CNT respond to electromagnetic radiations and hence, haveuccessfully presented as infrared (IR) sensitive material [3,4].wing to its excellent electronic and optoelectronic properties, it
s considered as an ideal material for IR photodetectors [4]. Thismplies that such a response of CNT can be exploited to detectarious sources of radiations, for example from flame or fire. Ineneral, flame-sensing materials in general should be very respon-ive to detect the presence of fires and at the same time shoulde capable of inhibiting false alarms. In such cases a novel prop-rty of an effective flame sensing material lies in its range ofetection and response time. Moreover, ignition of the flame isccompanied by emission of gases like hydrocarbons and predomi-antly carbon dioxide upon complete combustion of air, which canlock the conducting channels of the CNT through physisorption
hereby making it more electrically resistive in nature [5]. Whileroperties of CNT can be tuned by functionalizing it with variousunctional groups such as carboxyl, amine etc. it can also be used to
enhance its sensitivity toward the proposed application [6]. In thepresent work, we have devised a method to develop such a flamesensor, which is shown to be sensitive to the above-mentionedparameters.
Conventional flame or fire sensors distinguish between theflame radiation and background radiation by mainly utilizing fourprimary optical flame-sensing technologies such as detection ofultraviolet (UV), combined detection of UV/IR, multi-spectrum IR,and visual flame imaging. Optical characteristics of the flame suchas its shape and flickering provide a real time analysis through videoprocessing [7]. These methods work on optical monitoring throughcamera, which is realized as extremely complicated in handling aswell as real time data processing. In addition, an intelligent systemrequires a combination of smoke, fire and temperature detectorsin an assembled manner to combat false alarms by incorporatingextremely complex fault detection algorithms [8]. We have devel-oped a simple flame sensor based on multiwalled carbon nanotubes(MWCNT), which not only detects a flame but this technique canalso be used to estimate distance, hence the intensity along both thelateral and longitudinal directions of the flame. The sensor responseshowed a reversible nature after being exposed to a flame and theresulting response depends upon the intensity of emitted radia-tion. Through this work we explore a bilateral nature of sensor’sresponse along two different orientations of the flame, which can befruitfully utilized for various non-contact applications. For exam-ple, the short range of the response along one direction provides
the scope for developing a proximity sensor as a highly sensitivedevice. On the other hand the long-range detection capability inthe perpendicular direction can be materialized into fire safetyalarms.
MWCNT was synthesized using chemical vapor deposition using three-zone furnace. The end zones of the furnace were set at loweremperature while the middle zone of the furnace, which is theeaction zone, was set at 825 ◦C. Toluene and ferrocene were useds precursors acting as carbon source and catalyst, respectively.
chemical solution was obtained after mixing these chemicalsn a mass ration of 0.02, which was heated at 200 ◦C. Chemicalapors were carried into a reaction zone by argon gas flowing at
constant flow rate of 800 sccm (standard cubic centimeter perinute). MWCNT were grown on a silicon dioxide substrate and
crapped off from the substrate for chemical functonalization. Acidreatment procedure was used to achieve COOH functionalized
WCNT (COOH-MWCNT) as suggested by Khalili et al. by adding00 mg of pristine MWCNT to 150 ml mixture of H2SO4/HNO3 (3:1y volume ratio) [9]. The carboxylated MWCNT thus prepared wereispersed in a 20 ml solution of de-ionized water and isopropyllcohol using ultra-sonification [10]. Fig. 1a shows scanning elec-ron micrograph (SEM) of the as-grown, entangled microstructuref CNT. CNT solution was then drop casted on to the silicon diox-de substrate between two pre-deposited aluminum electrodess shown in Fig. 1b. The approximate size of the whole chip is
× 3 cm2. The distance between the electrodes was kept 100 �m.he amount of drop casted CNT on to the substrate determinedhe initial resistance of the MWCNT device that in turn dependspon the thickness of CNT layer (concentration of CNT betweenlectrodes).
Fixing an initial resistance of the device could control the con-entration of the CNT between electrodes in different devices. Theesponses from the CNT sensor devices are normalized with initialesistance to acquire a statistical variation in each experiment.
.2. Sensor-setup
The experimental setup of the sensor comprised of MWCNTevice (Fig. 1b) connected to a Keithley-2611 source meter ashown by schematic in Fig. 2a. A data acquisition module was usedo observe and collect a typical response of the device in �V, whenxposed to a source of flame such as a spirit lamp (which used ethyllcohol as a burning fuel).
Sensor response was measured both along the lateral and longi-udinal directions of the flame (Fig. 2b and c). A typical sensingesponse in the lateral direction of the CNT sensor is shown inig. 2d. A voltage response of the sensor upon applying a con-tant current of ∼12 mA (a randomly chosen currant value within
detectable range of the sensor) was measured with the peri-dic exposures of flame and is plotted with the exposure time ineconds. A stable and reversible response can be seen for the fiveycles of the flame exposure.
The effect of radiation only (without flame) was also evaluatedsing an IR source with variable power in the range of 18–23 dbm to
rradiate the device with an optical fiber. The sensor was subjectedo various bias currents ranging from 0 to 20 mA.
.3. Signal conditioning and data acquisition module
The response of the MWCNT device (∼�V) was amplified using signal-conditioning module as shown in Figs. 2a and 3a to a rangef mV using a Quantum X module. First phase of the signal con-
itioning circuit was a typical Wheatstone bridge, which yields aensitive response of the device. The initial and stable resistancewithout connecting to Wheatstone bridge) of a typical function-lized MWCNT device was measured around 180–200 �. Then,
uators B 192 (2014) 594– 600 595
this device was connected to a Wheatstone bridge with equal orslightly higher resistance arms (considering the fact that slightlydisparate arms would produce a very feeble deflection to counter-part the zero error which might be present), which provided anideal sensitivity for a bridge with equal arms and can be expressedas (�V/V)/(�R/R) = 1/4 [11]. The arms of a Wheatstone bridge wereemployed depending upon the resistance value of the MWCNTsensor and in case of a device with a different resistance value,the values of arms of the Wheatstone bridge could be employedaccordingly. The exposure to the flame caused the resistance of thedevice, (which acts as one of the arms of the Wheatstone bridge) todecrease, thereby inducing deflection in the bridge circuit.
An operational amplifier (op-amp) was employed to amplify thedeflection of the Wheatstone bridge circuit and an IC-OP07CP wasfabricated as an op-amp in differential configuration to add a gainto the circuit and hence the deflection in �V could be amplified(with the gain of Rf/Ri as shown in the circuit) to be approximately20 times of the original from the chosen value of resistances. Nowthe amplified response of the device was provided to the micro-controller for data acquisition. Arduino Uno 4.0 board was used asa data acquisition module that took an analog voltage of the signal-conditioning module and showed the digital response by couplingit to a basic 16×2 inches Liquid Crystal Display (LCD) like HitachHD 44780.
The algorithm of the Arduino Uno was developed on an Arduinotoolkit and can be calibrated to measure the voltaic response andproduce real time estimation of the distance. Fig. 3b shows a blockdiagram showing all the functional blocks of the sensor as describedin this section.
2.4. Experimental procedure
Initially, the functionalized MWCNT device was tested at vari-ous values of applied bias currents and the response to the flamewas monitored along both lateral as well as longitudinal directionsof the flame exposure. In other words, upon traversing along boththe directions, the device is exposed to the flame by sideways aswell as to the top, respectively as was shown earlier in Fig. 2b andc. In general the laws of heat transfer state that the heat producedsideways from a flame, which is primarily due to radiation is muchsmaller than compared to that from the top of the flame due to con-vection. Hamins and Bundy reported that the total flux producedfrom a candle flame at a radial distance (along the lateral direction)of 5 mm was around 40 kW/m2 whereas along the length of theflame at a height of 38 mm from its baseline was 90 kW/m2 [12].The distribution of total heat flux versus radial distance from theflame centerline at two longitudinal positions, which was recorded,clearly demarcates the convective heat transfer to be dominantover radiative heat flux as we move above the flame region [12].Hence observations were made along these two axes (as shown inFig. 2b) to distinguish the nature of electrical response in both thedirections as well as the effect of change in distance between flameand CNT device.
The experiments were conducted by subjecting the MWCNTdevice to a range of bias currents (3–15 mA in case of lateral direc-tion and 5–18 mA in vertical direction) refraining the amplifiedoutput of the microcontroller from going beyond its supply volt-age. A fixed bias current supply of 5 mA (in the lateral direction)and 10 mA (in the longitudinal direction), which gave more ami-cable and consistent response for all distances of exposure, wasused to observe any variation in the sensor response that occurreddue to variation in the distance. The electrical response of the sen-
sor increases the deflection in the Wheatstone bridge. Electricalresponses were recorded at different distances from the flame aswell as applied current for a time of 30 s of the flame exposure.Experiments were also conducted to reveal the response of the
596 S. Mohanty, A. Misra / Sensors and Actuators B 192 (2014) 594– 600
Fig. 1. (a) SEM image showing microstructure of the MWCNT film used in our study. (b) Schematic shows the functionalized MWCNT device for flame sensing.
Fig. 2. (a) Schematic shows the experimental set-up used for sensing flame using MWCNT sensor. (b) and (c) Show the lateral and longitudinal directions of the sensormeasurements. (d) Cyclical voltage response of the MWCNT sensor device while applying a fixed bias current and it is exposed to the flame that was kept at 4 cm distancefrom the sensor in the lateral direction.
Fig. 3. (a) Signal conditioning circuit used in fabrication of the flame sensor. (b) Block diagram of the circuit that comprises of the Wheatstone bridge and differential amplifier.
S. Mohanty, A. Misra / Sensors and Actuators B 192 (2014) 594– 600 597
F senso( . Insett e app
Mw
3
(rtirriccMgnrtposrwdvbtcfe
ig. 4. The parameters observed for the lateral displacement of the flame from ther). (b) The sensitivity �V/V of the sensor is plotted with the corresponding distancehe bias current passing through the sensor. (d) Sensor sensitivity is plotted with th
WCNT device to radiations by exposing to a source of IR radiationith variable laser power intensity.
. Results and discussion
The electrical response was first recorded from bare MWCNTwithout functionalization) device when exposed to a source ofadiation (spirit lamp as mentioned in sensor setup) i.e. a means toest and ultimately calibrate the sensor, showed a linear decreasen the measured resistance. The extent of bare CNT to recover itsesistance is feeble contributing to extremely slow and decayingesponse. This is because the gaseous emission of carbon diox-de gets adsorbed onto the surface of the CNT hence blocking itsonducting channels [9]. Therefore, the desorption of these gasesontribute to the slow recovery of the electrical response of theWCNT device [5]. However, functionalized CNT with carboxyl
roup aids to the recovery of the electrical response to the origi-al resistance upon periodic exposure to the flame [9]. Though theecovery of the electrical response is not as spontaneous as thato the periodic exposure of flame, the recovery time is still com-arable to the response time during exposure, thereby responsef the flame is reversible as apparent from the typical responsehown in Fig. 2d. Upon periodical exposure of the flame, theeversible/cyclical response of the voltage across it is monitoredhich shows a recoverable change in Ohmic resistance of theevice, which was seen clearly in Fig. 2d. The magnitude of thisoltaic change went up to 5 mV depending upon the variation ofias current across it and was also sensitive to the distance from
he flame. These two parameters primarily governed the extent ofhange in resistance in the device and are studied systematicallyurther along the lateral and longitudinal directions of the flamexposure.
r. (a) A plot of change in output voltage �V of the sensor with respect to distance shows a linear fit of the sensitivity with 1/r2. (c) Output voltage �V is plotted withlied bias current.
3.1. Lateral displacement of the sensor
Upon traversing along the lateral direction of flame, the elec-trical response of MWCNT device was revealed varying with thedistance. Fig. 4a shows continuous decrease in response, as sen-sor was moved farther along the axis. A maximum response wasrecorded between a distance of 1 and 2 cm, which is shown tosaturate after a farthest distance of 5 cm where the curve almostflattens. It appears to follow a non-linearly decreasing polynomialfunction. It is known that the intensity of radiation is a characteristicof the source (I0) and the distance r from it as Eq. (1).
I = I04�r2
(1)
Since the spirit lamp used in our experiment acts as a constantsource of intensity, the variation of distance is the driving forcebehind the observed characteristics and hence clearly the change inelectrical response seems to follow 1/r2 variation with the distance.A normalized change in voltage is shown in Fig. 4b and inset shows alinear relation with 1/r2. Henceforth the voltage change occurringin the device shows a linear dependence on the intensity of theradiation (I).
The sensitivity of the sensor can be expressed as (�R/R) whichcan be directly related to a ratio of the change in voltage acrossthe sensor (�V) to the applied bias voltage (V) across it as (�V/V),since V is same for the all the distances under observation, thebias current through the device being constant i.e. 5 mA. Hence,the sensitivity with respect to the distance so obtained i.e. �V/V isas shown in Fig. 4b. Also we notice that in case of extreme proximity(<1 cm) to the flame the response shoots up and going beyond this
proximity can permanently damage the device due to the heatingeffects, which causes larger error bars in both Fig. 4a and b. Its evi-dent that within this range the response is sufficiently high to bedetected and clearly distinguishable from the farther distances.
5 nd Actuators B 192 (2014) 594– 600
raiectdw
MbScv
eiedo
3
vdmsalt
Fdi
98 S. Mohanty, A. Misra / Sensors a
To estimate the sensitivity with respect to the varying bias cur-ent, the distance of the sensor from the flame was fixed at 3 cmnd different bias currents were applied from the source as shownn Fig. 4c. An increasing trend in change in voltage was noticed forach bias current but the initial bias voltage across the whole cir-uit was way higher than any voltaic change. So while calculatinghe sensitivity for each current it was noticed that 1/V term wasominant than the �V term resulting into a hyperbolic trajectoryith linearly decreasing slope as shown in Fig. 4d.
To separate the flame response with the IR exposure, theWCNT device was exposed to an IR source with variable laser
eam intensity and resulting electrical responses were recorded.uch higher power radiation has been reported to produce photourrents in carbon nanotubes prominent even at no external biasoltage across it [Eq. (1)].
It is shown in the present study that intensity is either depend-nt on the source of radiation or the distance from it (which followsnverse square law), and the change in voltage has a linear depend-ncy on intensity as was seen in Fig. 4a. Fig. 5 shows a linearependence of change in voltage of the MWCT sensor on intensityf radiation of the IR laser source i.e. I0 (from Eq. (1)).
.2. Longitudinal displacement of the sensor
The electrical response of the sensor was monitored in the trans-erse direction of the flame. Upon traversing in the longitudinalirection (which is straight top of the flame) the heating effect isuch more prominent and so is the influence of gaseous emission
uch as carbon dioxide which hinders its desorption from the CNTfter the flame exposure to a greater extent as compared to theateral direction due to the orientation of the sensor [5]. Therefore,he role of carboxyl functionalized MWCNT as mentioned earlier to
ig. 6. Sensor parameters observed for the longitudinal displacement of the flame fromistance (r). (b) The respective sensitivity �V/V of the sensor is shown with its correspon
n output voltage �V with respect to the bias current passed through the sensor. (d) Sens
Fig. 5. Change in the voltage is plotted with the varying intensity of the IR sourcewithout any external bias.
aid the desorption process, becomes even more significant as theconvective currents cause greater extent of heat transfer in the ver-tical direction also drive the emission of carbon dioxide along thesame direction [12]. Therefore, the device was able to recover itsoriginal resistance though the recovery is slower in comparisonto that in the lateral direction. The response to the flame is soprominent in this direction that the range of detection increasestremendously (more than 40 cm) as can be seen in Fig. 6.
The change in resistance (voltage) is measured to be almosttwice than that in lateral direction. The response of the sensortoward the flame exposure was recorded upon applying a constantbias current of 10 mA at various distances upon an exposure for 30 s
the sensor. (a) The change in output voltage �V of the sensor is plotted with theding distance from the flame and inset depicts its relation with 1/r2. (c) The changeor sensitivity is plotted with the applied bias current.
S. Mohanty, A. Misra / Sensors and Actuators B 192 (2014) 594– 600 599
cemen
arootwhb
actmttl
bbtbsdo
tFcgt
3
strtFv
i3flpdi
[
Fig. 7. Relative response time for (a) lateral displa
s shown in Fig. 6a. Upon bringing the device closer (<20 cm), theesponse in terms of change in resistance went beyond the limitf the permanent change in resistance and the effect of adsorptionf carbon dioxide is dominant. Similar to our previous observation,he sensor response was observed to be abiding inverse square lawith distance, as it is directly proportional to the intensity, andence the sensitivity with respect to the distance at this particularias current (10 mA) as shown in Fig. 6b and inset, respectively.
Upon varying the bias current for an exposure to the flame at fixed distance of 30 cm, it was noticed that the magnitude ofhange in voltage was much more than that in the lateral direc-ion as evident in Fig. 6c. Interestingly, now since the variation in
agnitude of voltaic change is higher and is comparable to that ofhe actual voltage output for each individual bias current, the �V/Verm cannot be approximated as hyperbolic but still follows theinearly decreasing trend as shown in Fig. 6d.
After conducting the studies relevant to characterize the sensorased on above discussed parameters, it was inferred that alongoth the axes of sensor displacement the sensitivity with respecto the distance is the parameter of paramount importance to cali-rate the sensor. For any fixed bias current through the sensor, theensitivity follows inverse square law of intensity thereby the slopeecreases as we go along the distance axes for both the orientationsf the sensor.
Upon reforming these results to be sensitivity versus per dis-ance square, we get the linear curves shown in the inset ofigs. 4b and 6b as predicted. Based on these curves, the sensoran be calibrated linearly and thus the microcontroller can be pro-rammed to predict the distance of flame from the sensor based onhe instantaneous value of sensitivity.
.3. Response time
In previous results the sensor response of the sensor was mea-ured at a fixed bias voltage with a variation in the distance fromhe flame. As seen above the rise and fall of the typical cyclicalesponse occurs linearly with time and hence the change in resis-ance of MWCNT follows a linear behavior with time as shown inig. 2d thus we can estimate the time it takes to reach a particularoltage value at different distances.
The relative response time in this direction for various distancess shown in Fig. 7b. So for a fixed bias current of 5 mA sensor takes0 s to reach a specific voltage value when the distance between
ame and sensor was kept at 5 cm along the lateral direction. Com-aring this maximum voltage value, a relative response time forifferent distances can be estimated using the distance sensitiv-
ty as shown in Fig. 7a. On the similar accounts considering the
[
[
t and (b) longitudinal displacement of the sensor.
longitudinal direction, the value of the voltage, which reaches a par-ticular voltage level at 45 cm in 30 s can be used to draw a similaranalogy.
4. Conclusions
We have fabricated and designed a flame sensor based on func-tionalized MWCNT and shown the nature of response and itssensitivity in both the lateral and longitudinal directions of theflame. The sensor thus fabricated can be calibrated to operate onany particular bias current and its sensitivity with respect to dis-tance at that particular bias current, which can be used to estimatethe working distance from the flame. Depending upon the range ofdetection and the nature of response, the sensor can be deployedin various fields depending upon the application. The far fieldresponse in the longitudinal direction has a good working range andcan be used as safety alarms whereas the near field response in thelateral direction can be capitalized to use as proximity sensor in lab-oratories. Materialistic research can be done in future to make thesensor more spontaneous thereby decreasing the response time.
References
[1] V. Kotipalli, Z. Gong, P. Pathak, T. Zhang, Y. He, S. Yadav, L. Quea, Light andthermal energy cell based on carbon nanotube films, Applied Physics Letters97 (2010) 124102.
[2] A. Suri, A. Misra, Coupling of photomechanical and electromechanical actua-tions in carbon nanotubes, Nanotechnology 24 (2013) 105501.
[3] Q. Zeng, S. Wang, L. Yang, Z. Wang, T. Pei, Z. Zhang, L. Peng, W. Zhou, J. Liu, W.Zhou, S. Xie, Carbon nanotube arrays based high-performance infrared pho-todetector, Optical Materials Express 2 (2012) 839.
[4] P.W. Barone, S. Baik, D.A. Heller, M.S. Strano, Near-infrared optical sensorsbased on single-walled carbon nanotubes, Nature Materials 4 (2005) 86.
[5] M. Oftadeh, B. Gholamalian, M. Hamadanian, Investigation of the interaction ofcarbon dioxide fluid with internal and external single-wall carbon nanotubesby DFT, Journal of Nanostructures 1 (2012) 213.
[6] K. Balasubramanian, M. Burghard, Chemically functionalized carbon nano-tubes, Small 1 (2005) 180.
[7] B.U. Toreyin, Y. Dedeoglu, U. Gudukbay, A.E. Cetin, Computer vision basedmethod for real-time fire and flame detection, Pattern Recognition Letters 27(2006) 49.
[8] R.C. Luo, K.L. Su, K.H. Tsai, Fire detection and isolation for intelligent buildingsystem using adaptive sensory fusion method, in: Robotics and Automation,2002. Proceedings. ICRA’02. IEEE International Conference on Robotics andAutomation, vol. 2, 2002.
[9] K. Soodabeh, A.A. Ghoreyshi, M. Jahanshahi, Carbon dioxide captured by multi-walled carbon nanotube and activated charcoal: a comparative study, ChemicalIndustry and Chemical Engineering Quarterly 19 (2013) 153.
10] A. Misra, J.R. Raney, A.C. Craig, C. Daraio, Effect of density variation and non-covalent functionalization on the compressive behavior of carbon nanotube
arrays, Nanotechnology 22 (2011) 425075.
11] G. Weiss, W. Bridge Sensitivity, Instrumentation and measurement, IEEE Trans-actions on Instrumentation and Measurement 18 (1969) 2.
12] A. Hamins, M. Bundy, S.E. Dillon, Characterization of candle flames, Journal ofFire Protection Engineering 15 (2005) 265.
umit Mohanty is a senior undergraduate student, working toward a degree innstrumentation Engineering at Department of Electrical Engineering at Indian Insti-ute of Technology (IIT), Kharagpur. He has been a research fellow at IISc, Bangalore.
tuators B 192 (2014) 594– 600
Abha Misra received her PhD degree in the field of thermal-mechanical behavior
of confined metal nanorods in the carbon nanotubes. She joined Indian Instituteof Science (IISc), Bangalore in 2010 as an assistant professor in Department ofInstrumentation and Applied Physics. She is investigating the sensing mechanismof carbon nanotubes and graphene. Her research interests also include fabricationand characterization of variety of sensors and actuators.
