Institut ffir Quantenoptik, Universitat Hannover, Welfengarten 1, D-3000 Hannover 1, Federal Republic of Germany
Received March 24,1989; accepted May 23,1989
A novel fiber-optic humidity sensor is described. It is based on reversible sorption of water from the ambientatmosphere in a porous thin-film interferometer that sits on the tip of a fiber. The sorbed water changes therefractive index of the thin films and thus the reflectivity of the interferometer; the resulting modulation of thereflected intensity is detected. The sensor is insensitive to electromagnetic interference and aggressive chemicalsand is extremely small (micrometers).
Fiber-optic sensing is a rapidly expanding field, yetthere is a lack of sensor elements for humidity. This isregrettable because there is a need to control humid-ity, along with temperature, for air conditioning inboth residential and office buildings and for a varietyof industrial processes. We present a novel fiber-optic sensor for humidity that is extremely small (mi-crometers), not sensitive to electromagnetic fields,and withstands aggressive chemicals.
The operation of the sensor relies on the fact thatvapor-deposited thin films, since they are widely usedfor all kinds of mirrors and antireflection coatings,have a certain degree of porosity. Such films are actu-ally more a fractal sponge than a compact layer.1 Thevoid fraction depends on the details of the depositionprocess and can in some cases easily reach 10-20%.
It has been known for awhile that moisture from theambient atmosphere can diffuse into the pores. Aswater fills the voids the effective refractive index ofthe film rises. The index variation becomes manifest,e.g., as drift and lack of reproducibility in dielectricfilters with steep cutoff slopes.2 It is precisely thisindex change, which is normally a nuisance, that isexploited here for a measurement of the humidity.The proposal hinges on the observation that there arefilm materials in which the adsorption and desorptionof water molecules inside the pores is perfectly revers-ible.
Many conventional methods for the measurementof humidity make use of moisture adsorption in porousmaterials. The humidity-dependent length change ofa hair3 is the best-known example. Resistive sensorsutilize the change in conductivity of hygroscopic saltsembedded in a porous substrate.3 Capacitive sensors4
are now widely used; they use a flat substrate of porousplastic or ceramic, with electrodes on both surfaces.As water enters the pores it affects the dielectric con-stant, and the resulting change in capacity can beevaluated. A recently proposed fiber-optic sensor5utilizes a porous fiber into which humidity can mi-grate. A dye that changes color in the presence ofwater is deposited in the pores. Humidity can then beassessed from absorption spectroscopy.
There are difficulties that can arise with these prin-ciples. Hair is damaged in dry atmosphere. Electri-cal sensors can give wrong readings when condensa-
tion causes leak currents. A dye deposited in thepores might be washed out. None of these problemsarises with the sensor proposed here.
Let us remark that there are also methods that donot utilize porous materials.3' 6 Direct weighing of thewater content of a known air volume, e.g., throughfreezing out, is a time-consuming procedure. Thepsychrometric principle makes use of the temperaturedifference of wet and dry thermometers when both arefanned with the air to be measured. The dew-pointmethod involves measurements of the temperaturedrop required until condensation on a cooled polishedsurface occurs. These methods cannot be miniatur-ized and are thus unsuitable for use as sensor ele-ments.
We comment below on two recent proposals for inte-grated-optic humidity sensors7' 8 and first describe theproposed sensor. Consider a thin-film Fabry-Perotinterferometer as shown in Fig. 1. A spacer layer ofhigh-index material is sandwiched between two par-tially reflecting mirrors, each of which consists of astack of layers of alternatingly high- and low-indexmaterial. It is well known that a Fabry-Perot resona-tor has multiple resonances that are separated in fre-quency by the free spectral range, FSR = c/2nd, wherec is the vacuum speed of light, n is the index, and d isthe geometrical thickness of the spacer (for simplicitywe ignore a correction to d that arises because thesurface of reflection is not identical to the spacer-
==SiO 24 4 ,=TiO 2
fron lasermoisture
to detector
fiber Ha sensor -*Fig. 1. Cross section of a thin-film Fabry-Perot resonator(not to scale). The overall thickness of the stack of films isapproximately 2 Am.
Fig. 2. Schematic of the setup for the humidity measure-ment. L's, Lenses; PBS, polarizing beam splitter; A/4, quar-ter-wave plate; ref, reference voltage for zero-point calibra-tion.
mirror interface but rather a little inside the mirrorstacks). If one irradiates light tuned to a wing of aresonance, variations in the spacer index are sensitive-ly translated into variations of transmitted or reflect-ed intensity. With the spacer index being humiditydependent, we can assess humidity through a simpleintensity measurement.
For the materials we chose TiO2 (high index) andSiO2 (low index), which are among the most durablematerials both mechanically and chemically. Theyhave low absorption over a wide spectral range so thatsensors can be designed for any wavelength in thevisible or near infrared. Our first experiments on thematerials' characterization were performed with de-vices that were grown on 1-in.-diameter BK 7 sub-strates, rather than on fibers, and were designed tohave a resonance peak at the design wavelength XDnear the 514-nm line of an argon-ion laser.
The films were grown in an in-house facility at 600K, with typical growth rates of 0.7 nm/sec for TiO2 and2 nm/sec for SiO2 . Additional oxygen pressures of0.15 ,bar for SiO2 and 0.3 ,gbar for TiO2 were used sothat of the various titanium oxides mostly TiO2 wasformed. The structure of the devices is (substrate)(HL)kHI(LH)m, where each H (L) denotes a XD/4layer of the high- (low-) index material. A suitablechoice is k = m = 2, which results in mirror reflectivi-ties of approximately 0.8, and 1 = 14 for a spacerthickness of (14/4)XD-
Originally the devices were made for a differentpurpose. A detailed account of the characterizationof their properties is given elsewhere, 9 thus only a briefreview is given here.
From the wavelength dependence of the small-sig-nal resonance angle the index of the spacer layer isfound (nfilm = 2.29). Comparison with the TiO2 bulkindex (nbulk = 2.7) yields a void fraction of approxi-mately 15%. When the film temperature (and thusthe moisture content) is cycled repeatedly, no indica-tion of degradation is found; it thus seems that theindex variation is completely reversible. The mea-sured values of the index as a function of temperatureor of ambient humidity make good fits to Langmuirisotherms, so one can speculate that interior surfacesin the pores of the film are covered with monolayers ofwater. From the fits a desorption energy of 0.3 eV/molecule is derived. In addition, the number of sitesfor adsorption is obtained; from that number one canconclude that the interior surface is approximately 75times larger than the outside surface. The total water
load is obtained from the weight difference of the wetand dry devices; the result is approximately 1 ptg/cm2
of exterior surface in humid air. This number isroughly half of the upper limit given by the availablepore space (the void fraction). The response time ofthe index is determined by the diffusion of water mole-cules through the film stack.
This characterizes the materials for the humiditysensor fairly completely. We now turn toward filmstructures deposited on the tip of standard multimodecommunication fibers. In this case the film thicknessis designed such that there is a wing of a resonancepeak at 790 nm, the emission wavelength of a laserdiode to be used (see below). To obtain a reasonableresponse curve it is important to get the correct filmthickness and thus the correct resonance wavelength.If several fibers were coated at once, closely packedtogether, it would be possible to achieve such closetolerances that a calibration would be required onlyonce for the whole batch.
Such a fiber is part of a hygrometer setup shown inFig. 2. Light from a laser diode is launched into thefiber. At the far end the sensor reflects some of thelight back; this light is separated from the incominglight by means of polarization optics and is sent to aphotodiode. The detector signal is processed by astandard operational amplifier for gain and zero-pointsetting suitable for the readout.
For convenience we used an original-equipment-manufacturer module from a consumer compact discplayer in the setup. It contains a laser diode (withintensity stabilization), the optics, and the detector.All that is additionally required are the detection am-plifier and meter, a power supply, and the fiber itself.The compact disc player modules are made in largequantities and are therefore inexpensive.
The sensing end of the fiber sits in a small chamberthrough which dry or humidified nitrogen can beblown. In this way the humidity can be set between0% and approximately 80%. A commercial capacitivehygrometer1 0 located next to the fiber tip is used as areference instrument. A response curve can thus bemeasured (Fig. 3).
We also attempted to measure the temporal re-
5
I: 4
._
'3L-
2
I
0 20 40 60percent relative hunidity
80 NO0
Fig. 3. Steady-state response of the fiber-optic sensor mea-sured with the setup of Fig. 2 and using a commercial hy-grometer as the reference.
Fig. 4. Transient response of the sensor. The relative hu-midity of the gas blown into the test chamber is switchedbetween 85% and 11%.
sponse of the fiber sensor to sudden changes in humid-ity. Figure 4 shows a response time of clearly less thana minute. The difficulty is to make the humidity inthe chamber change rapidly enough so that a stepfunction is approximated. Under our experimentalconditions the fiber sensor and the reference instru-ment reacted with comparable temporal responses.We therefore suspect that at least part of the responsetime, and probably the obvious difference in the re-sponse to humidity increase and decrease, is caused bythe slow rates at which moisture is blown into or out ofthe chamber. Still, the observed response is fairly fastfor humidity sensors.
We emphasize that our monolithic sensor is all di-electric so that it is suitable for measurements in elec-tromagnetic fields. It also carries little energy so thatit is suitable for hazardous and explosive environ-ments. Since the sensor contains only SiO2 and TiO2,it holds the promise of being chemically stable. Wetested this assumption by immersing sensors for sever-al minutes in concentrated sulfuric acid or mixtures ofconcentrated sulfuric, nitric, and hydrochloric acid.The devices were then briefly rinsed in tap water andblown dry. We also put them into an aqueous solutionof sulfurous acid. This harsh treatment did no harmto the sensors. This remarkable stability seemsunique and may make measurements in noxious ex-haust fumes feasible.
The device is also extremely small (dimensions inthe micrometer range), and thus can be placed in hard-to-reach places. We are not aware of any other hu-midity sensor that uses so little space, so little of thegas to be measured, and traps so little water (-0.1 ng)as the sensor described here. We therefore believethat applications will be found in which this sensor canbe applied without competition, for example, for mea-surements in very small hollow spaces, such as insideof living organisms, or inside the insulating materialsused in buildings. The immunity to electric and mag-netic fields also suggests applications in high-voltageinstallations, where it can be important to monitorhumidity.
Our sensor has an advantage over the integrated-optic sensor proposed in Ref. 7, which is based on thevariation in the critical incoupling-outcoupling angleof a waveguide whose surface can adsorb water. Thatapproach gives a comparatively large, nonmonolithic
device that requires precise angle adjustment. Anapproach to measure the heat transport off an inte-grated-optic Mach-Zehnder interferometer 8 requireswire connections and high-temperature (2000C) oper-ation, thus the device is not suitable for use in electro-magnetic fields nor in reactive or explosive environ-ments. It is also extremely sensitive to convection, sothe air to be measured must not flow.
Our sensor has a temperature dependence that isdescribed by the Langmuir isotherm mentioned aboveand detailed in Ref. 9. Near the calibration tempera-ture it is more nearly the relative, rather than theabsolute, humidity that is being measured. For preci-sion measurements over an extended temperaturerange one may add a temperature sensor and apply therequired corrections to obtain either value. As withany intensity-detection scheme, precision is also im-paired by variations in cable losses. This could beavoided by using two wavelengths, one on each slope ofthe resonance, because then a ratio of two intensitiescould be evaluated that would be independent of extralosses. Source wavelength drift as an error source isnot critical since the humidity-dependent resonanceshift is large (several nanometers).
With all their unique features the devices proposedhere are reasonably easy to manufacture, so they couldbe inexpensive when made in sufficient numbers.Also, fiber-optic systems have a potential for multi-plexing. For example, fiber networks could easily beplaced in cable or air-conditioning conduits in largebuildings, ships, or airplanes for remote humiditymeasurement.
The author thanks J. Schulte, J. Ebert, and D. Ris-tau for their support and for stimulating discussionsand D. Ruiter and K. Tetzlaff for their skillful techni-cal assistance.
References
1. For recent studies of thin-film growth see, e.g., J. R.Banavar, M. Kohmoto, and J. Roberts, Phys. Rev. A 33,2065 (1986); A. Mazor, D. J. Srolowitz, P. S. Hagan, andB. G. Bukiet, Phys. Rev. Lett. 60, 24 (1988); R. P. U.Karunasiri, R. Bruinsma, and J. Rudnick, Phys. Rev.Lett. 62, 788 (1989).
2. J. Schild, A. Steudel, and H. Walther, J. Phys. (Paris) 28,C2-276 (1967); S. F. Pellicori and H. L. Hettich, Appl.Opt. 27, 3061 (1988).
3. A. Wexler, ed., Humidity and Moisture (Reinhold, NewYork, 1965), Vol. 1.
4. P. E. Thoma, J. 0. Colla, and R. Stewart, IEEE Trans.Components Hybrids Manuf. Technol. CHMT-2, 321(1979).
5. M. R. Shahriari, G. H. Sigel, Jr., and Q. Zhou, in Digestof Conference on Optical Fiber Sensors (Optical Societyof America, Washington, D.C., 1988), p. 373.
6. F. Kohlrausch, Praktische Physik, 23rd ed. (Teubner,Stuttgart, 1985), Vol. 1, pp. 397-406.
7. K. Tiefenthaler and W. Lukosz, Opt. Lett. 9,137 (1984).8. A. Enokihara, M. Izutsu, and T. Sueta, Appl. Opt. 27,
109 (1988).9. F. Mitschke, G. Ankerhold, and W. Lange, Appl. Phys. B
48, 467 (1989).10. Testotherm Model Hygrotest 6300, 7825 Lenzkirch,
