Kelly SwinneyDarryl J. Bornhop

Department of Chemistryand Biochemistry,Texas Tech University,Lubbock, TX, USA

Noninvasive picoliter volume thermometry basedon backscatter interferometry

Using the on-chip refractive index (RI) detector based on backscatter interferometry,sensitive, small volume, noninvasive thermometry can be performed. The current opti-cal configuration for the on-chip interferometric backscatter detector (OCIBD) is quitesimple and consists of an unfocused laser, an unaltered chip with a hemisphericalchannel and a photodetector. Alignment is straightforward with the only requirementbeing that the beam fully fills the channel. The interaction of an unfocused laser beamwith the uncoated etched channel with a curvature within the silica plate (chip)produces fringes whose positional changes scale with respect to the refractive index(RI), n, of the fluid in the channel. Due to the inherently high value of dn/dT for mostfluids and the high sensitivity of OCIBD to RI changes, the measurement of smalltemperature variations in sub-nanoliter volumes is possible. Performing OCIBD with a75 �m diameter laser beam on a silica chip that contains an etched channel with a40 �m radius facilitates noninvasive thermometry on a N-(2-hydroxyethyl)piperazine-(2-ethanesulfonic acid) (HEPES) solution in a 188�10–12 L probe volume with atemperature resolution of 9.9�10–4�C, at the 99% confidence level.

Keywords: Backscatter interferometry / Thermometry EL 4401

1 Introduction

Many chemical analyses problems exist that cannot cur-rently be studied effectively on-chip. For example, beingable to elucidate the thermal events induced by electro-kinetic flow in etched channels will allow a better under-standing of the heat production that dictates separationperformance in these unique systems. Using a noninva-sive thermometer, applications such as on-chip calorime-try could be performed, reaction dynamics could be mon-itored, the heat of mixing produced under low Reynoldsnumber conditions could be studied, flow sensing wouldbe possible and even time resolved enthalpies could beobtained.

However, to date no techniques exists that can facilitatesensitive, noninvasive temperature measurements direct-ly on-chip in picoliter volumes. While there are a few com-mercial solid state devices based on thermisitors [1, 2]available that can probe nano- or even picoliter volumes,these devices usually have poor performance tolerances(�0.2�C) preventing them from being compatible withanalyses that require millidegree sensitivity. Further sincethese type of sensors [1, 2] require direct contact of the

medium under investigation, e.g. noninvasive sensing isnot possible. Reports have been published suggestingthat sub-millidegree temperature changes can be mea-sured with thin film thermopiles [3–5] and thermal scan-ning microscopy [6, 7]. In order to obtain temperatureinformation using thermal scanning microscopy, the sen-sor must come into direct contact with the fluid and to theauthors’ knowledge no work has been published on-chip.In the case of thin film thermopiles, the detector wouldhave to be integrated directly into the chip in order tosense changes in temperature. A few reports have beenpublished [4, 5] for microbiosensors based on integratedthermopiles for the detection of analytes in whole bloodsamples using the heat produced from enzyme catalyzedreactions. However, these devices have not been shownto work in the kind of channels commonly used for on-chip HPLC and CE, large probe volumes were employed(1 �L) in order to make the temperature measurement,and thermal isolation of the cold and hot junctions of thesensor was required [4, 5].

Here, we show that a simple on-chip refractive index (RI)detector can be used to perform sensitive, small volume,noninvasive thermometry on-chip. This system, the on-chip interferometric backscattering detector (OCIBD),uses a simple optical configuration consisting of anetched channel containing a hemispherical feature, alaser and a photodetector [8–10]. Due to the inherentlyhigh value of dn/dT for most fluids, the measurementof small temperature variations, as a function of changesin refractive index, in sub-nanoliter volumes is possible.

Correspondence: Dr. Darryl Bornhop, Department of Chemistryand Biochemistry, Texas Tech University, Lubbock, TX, USAE-mail: djbornhop@ttu.eduFax: +806-742-1289

Abbreviations: OCIBD, on-chip interferometric backscatterdetector; RI, refractive index

2032 Electrophoresis 2001, 22, 2032–2036

WILEY-VCH Verlag GmbH, 69451 Weinheim, 2001 0173-0835/01/1006–2032 $17.50+.50/0

Figure 1. Block diagram of the OCIBD optical configura-tion.

Using OCIBD, it has been possible to make millidegreetemperature measurements in just 188 pL. Additionally,three different fluids, water and two buffer mixtures com-monly used in CE, are interrogated allowing the determi-nation of each fluid’s dn/dT response in a confinedvolume.

2 Materials and methods

2.1 OCIBD setup

The block diagram for the OCIBD is shown in Fig. 1.A more detailed discussion of the detector appears else-where [8]; yet, in general, a low power laser (either a HeNe(Melles Griot, Irvine, CA, USA) or a diode laser (MicroLa-sers Systems, Garden Grove, CA, USA)) was directedtowards a mirror positioned about 55 cm from the laser’saperture. The mirror was angled at 45� with respect to thelaser in order to steer the beam onto the chip and collect/redirect a portion of the interference pattern towardthe detector. The etched channel containing silica chip(Fig. 2) was located directly below the mirror and mountedonto a thermostated aluminum support block, which wasaffixed to an X-Y translation stage. The distance from thefront surface of the mirror to the chip was ca. 10 cm. Thelaser beam was directed onto the etched channel of thechip at a distance that is about 1 mm from the exit port.The resulting interference pattern was monitored using asolid state detector whose signal was digitized using ananalog to digital converter (PPIO-AIO8, CyberResearch,Branford, CT, USA) that was interfaced to a PC computer.The solid state detector used was a silicon avalanchephotodiode (APD) (Advanced Photonix, Inc., Camarillo,CA, USA) with an active surface area of 900 �m2 (0.3 mmdiameter) and a responsivity of 30 A/W at 632.8 nm. TheAPD was wired in the reversed bias configuration in serieswith a 15 M� resistor and powered by a 213-V powersupply, which was 2 V below breakdown. The APD wassecurely housed in an aluminum die cast box (Pomona

Figure 2. (A) Top view of the silica plate containing ahemisphere etched channel. (B) Cross-sectional end-onview of the etched channel in the silica plate.

Electronics, Pomona, CA, USA) and was mounted on ahigh-precision translation stage and configured to interro-gate the fringe shift directly. The photodetector waslocated approximately 28 cm from the front surface ofthe chip and was aligned on the edge of a fringe in orderto monitor the RI change induced fringe movement.Experimentally it was found that the best position for thedetector is on the edge of the sloping intensity gradient ofthe working fringe (usually 3rd fringe from the centroid) atabout I=1/e2 of the essentially Gaussian intensity distribu-tion. Placing the detector here allows a relatively large lin-ear dynamic range to be probed while maintaining highresolution. The chip was mounted on a custom designedaluminum black anodized assembly. This assembly wasthermally controlled using a Peltier thermoelectric coolingchip (Melcore, Trenton, NJ, USA) and a high-resolutionpower supply (ILX Lightwave, Bozeman, MT, USA) wiredin feedback with a calibrated thermister. In these experi-ments, a chip (Fig. 2A) with a 40 �m radius etched channel(Fig. 2B) was used. As shown in a top view, the silica chipscontained two crossing etched channels (Fig. 2A) withholes and reservoirs at each end to allow from introduc-tion and removal of sample from the system. An end-on

Electrophoresis 2001, 22, 2032–2036 Noninvasive picoliter volume thermometry 2033

Miniaturization

view of the etched channel is shown in Fig. 2B, illustratingthat the etched channel has the shape of a hemispherewith two 40 �m radii connected by a 10 �m flat section.

2.2 Evaluation of thermal responses

Thethermal responsesofdeionized/distilledwater,100mM

Tris/30 mM boric acid buffer solution, and a 20 mM HEPESbuffer solution were studied. All solutes used in theexperiment were reagent-grade or better. Experimentally,each fluid was hydrodynamicly injected into the etchedchannel, allowed to temperature and pressure stabilize,and then the temperature of the flow cell was manuallychanged in approximately 0.3�C increments. The samplewas allowed to achieve thermal equilibrium at eachtemperature resulting in stabilized fringe position andthe intensity change was quantified. Using the appropri-ate RI response versus voltage (fringe shift), calibrationcurves the absolute temperature change, the relativetemperature change, or the relative change in refractiveindex (dn) can be obtained. Additionally, the thermalresponse produced by an applied voltage, as under nor-mal CE conditions, was evaluated. In this case, a 20 mM

HEPES buffer solution was hydrodynamicly injected intothe etched channel and allowed to stabilize. Once ther-mal and pressure equilibrium was achieved, a voltagetypical of on-chip CE (2 kV), was applied across theseparation channel in the form of a square-wave. Theresulting RI signal was collected for the duration ofthe experiment.

3 Results and discussion

The ability to make millidegree, noninvasive temperaturemeasurements on-chip could facilitate many applicationsincluding reaction monitoring, flow sensing, and calori-metry. In general, temperature measurements are crucialand in many cases must be obtained noninvasively soas not to disturb the system. While this goal has beenpossible in capillaries using Raman [11], thus far this hasnot been possible on-chip in picoliter volumes. The preli-minary investigations shown here demonstrate that theOCIBD can be configured to perform real-time, noninva-sive thermometry on-chip. Specifically, the thermalresponse of OCIBD was studied and the dn/dT responsewas estimated for three different fluids / solutions (dis-tilled/deionized water, 100 mM Tris/30 mM boric acid, and20 mM HEPES). These values were obtained by first gen-erating calibration curves for RI signal versus temperaturefor each fluid. With the calibration curves, the relationshipbetween changes in the RI signal as a function of tem-perature can be established for each sample, the sensitiv-

Figure 3. Calibration curve with the RI signal versus tem-perature for a 40 �m radius channel filled with water.

ity of the detector determined and quantitative measure-ments of relative changes in temperature can be calcu-lated. Ultimately, these measurements can be used tostudy thermal events.

Figure 3 depicts the calibration curve of RI signal versustemperature for water. As noted in the experimental sec-tion, the RI signal was generated by monitoring the posi-tional change of a fringe as a result of a stepwise, experi-mentally induced temperature rise of the system. Asexpected, the relative change in RI of water to changesin temperature is linear [12]. Next, the temperature sensi-tivity or the minimal resolvable temperature change ofOCIBD was determined using the peak-to-peak noise(baseline dispersion) of the measurement as �. Usingthis approach, � was measured to be 0.9 mV yielding a2� detection limit of 7.5�10–3�C. So in the current config-uration and using a conservative approach, a temperaturechange of 7.5 millidegrees can be detected for a fluidcontained within a probe volume of 3.14 nL, “on chip”. Amore aggressive approximation of the sensitivity ofOCIBD can also be calculated. If the peak-to-peak noise(0.9 mV) is assumed to be random and Gaussian, then �is equal to 1/5 of the peak-to-peak noise or 0.18 mV.Using this value, the 2� detection limit is 1.4�10–3�C. Itwill be shown below that by using a diode laser in the opti-cal train, the temperature resolution can be improvedwhile further reducing the probe volume.

As mentioned previously, various parameters can beobtained from the calibration curves of RI signal versustemperature. For example, the dn/dT coefficient can beestimated for each fluid allowing a better understandingof each fluid’s refractive index sensitivity to changes intemperature. This dn/dTcoefficient was calculated by firstestablishing the rate of change of n with concentration,

2034 K. Swinney and D. J. Bornhop Electrophoresis 2001, 22, 2032–2036

Table 1. Temperature sensitivity and dn/dT responses of CE buffers

Fluid Temperaturecalibrationcurve slope(V/�C)

Detection limits (�C)a) Estimateddn/dT (�C–1)

Probevolume (L)

Water 0.265 7.7�10–3 b) 1.4�10–3 c) 1.3�10–4 3.14�10–9

100 mM Tris/30 mM boric acid

0.277 7.2�10–3 b) 1.3�10–3 c) 8.9�10–5 188�10–12

20 mM HEPES 0.362 5.5�10–3 b) 9.9�10–4 c) 1.2�10–4 188�10–12

a) Determined using 2� statisticsb) Based on peak-to-peak baseline dispersionc) Sigma determined as 1/5 of peak-to-peak noise

second finding the slope of the thermometry calibrationcurve, and then applying the following relationship,

�n = 0.01�concentration�(nwater – nglycerol) (1)

where concentration is expressed as weight percent(g/g%). This calculation will be demonstrated using theresults obtained for water and a HeNe laser. First, usingEq. (1), the concentration sensitivity of OCIBD (0.0265V/mM) was converted into a quantity with units of re-fractive index (�n) per volt (V), giving a value of 5.02�10–4 �n/V. Next, the dn/dT value of water was obtainedby multiplying 5.02�10–4 �n/V by the slope of the curve(0.265 V/�C) in Fig. 3 giving the experimentally determineddn/dT coefficient of 1.3�10–4 RIU/�C. This dn/dT re-sponse value for water, as determined by OCIBD, com-pares well with the previously reported value of 1.05�10–4 RIU/�C [13, 14] and was done noninvasively within3.14 nL.

Given the potential applicability of on-chip systems tobiological analyses and the need for universal detection,it is valuable to evaluate the temperature sensitivity anddn/dT responses for CE buffers. A 100 mM Tris/30 mM

boric acid buffer solution and the 20 mM HEPES bufferserved as the representative buffer systems. For theseexperiments, the HeNe laser was replaced with a diodelaser emitting a collimated beam with a diameter of75 �m. Reducing the size of the probe beam from0.8 mm (HeNe) to 75 �m (diode laser) facilitates a detec-tion volume reduction by a factor of 16.7 or from 3.1 nLto 188 pL. As shown in Table 1, reducing the probevolume from 3.14 nL to 188 pL does not compromisethe sensitivity of the OCIBD system allowing changesin temperature for fluids with similar dn/dT coefficientsto be measured at the millidegree level. In fact, whencomparing the calculated detection limits for water and20 mM HEPES (two fluids with similar dn/dT coefficients),utilizing the diode laser improved the limits of detectionby 28.6%.

Improved resolving power for on-chip temperature mea-surements were achievable when using the diode laserprimarily because the intensity stabilizing correction cir-cuit integrated in the laser package allows a reduction inthe predominant noise in the system, intensity instabilitiesof the probe source. No such intensity correction feed-back loop was employed for the HeNe laser, which areknown to have relatively large intensity fluctuations.Nonetheless, in both cases millidegree temperature mea-surements are obtainable on-chip and noninvasively.When performing temperature measurements withOCIBD, the dn/dT sensitivity of fused silica must be con-sidered. At 633 nm, the dn/dT coefficient of fused silicahas been measured to be 10� 10–6 RIU/�C [15]. Consid-ering that the dn/dT response for most fluids are at leastan order of magnitude higher [14], the RI signal producedas result of the silica plate heating up contributes lessthan 10% to the temperature signal measured by OCIBD.Therefore, the RI response sensed by OCBID is assumedto be primarily associated with the temperature change ofthe fluid in the channel, not the temperature induced RIchanges from the silica.

With the capability to make sensitive temperature mea-surements in hand a simple voltage cycling experimentwas performed. Figure 4 shows that for a 20 mM HEPESbuffer, application of 2 kV results in an appreciable ther-mal (RI signal) perturbation. It should be noted, that wehave found this potential noise source can, to a greatextent, be eliminated in RI-CE by using the appropriatetiming [9]. Yet, it would be quite problematic to performboth solute detection and thermometry by OCIBD. Whatthe data shown here do indicate, as does a more detailedreport [10] on this topic, that a greater insight into theinfluence of Joule heating in on-chip CE will be possible,thus facilitating further optimization of the desired separa-tion. Further, OCIBD may be valuable in monitoring tem-perature changes in PCR so that these systems can beoptimized in the on-chip format. Current investigations in

Electrophoresis 2001, 22, 2032–2036 Noninvasive picoliter volume thermometry 2035

Figure 4. RI signal – temperature rise resulting fromapplying a voltage across the 40 �m radius separationchannel in the chip. The Joule heating response was fora 20 mM HEPES buffer and an applied voltage (squarewave) of 2 kV.

our laboratories include the further application of OCIBDincluding the study of on-chip calorimetry in picolitervolumes, the measurement time resolved enthalpies andon-chip flow sensing.

4 Concluding remarks

In summary, it has been shown that the on-chip RI detec-tor based on backscattering interferometer can be usedto perform noninvasive thermometry with millidegreeresolution in picoliter volumes. All of the measurementswere obtained using a very simple optical train with noadditional focusing or collection optics and using a chipthat has no reflective coatings. Considering the trendtoward ‘system’ miniaturization, the availability of an ultra-sensitive, small-volume thermometer is desirable. OCIBDrepresents such a device.

The National Science Foundation and the Welch Founda-tion supported this research. Applied Biosystems Inc. andCarl Paul are acknowledged for supplying the chips andhelpful discussion on implementation. The kind donation,from ILX Lightwave of Bozeman MT, for the diode laserdriver, power supply and the thermoelectric cooler con-troller is recognized. Dimitry Markov is especially thankedfor his contribution and expertise with electrical circuits.

Received October 26, 2000

5 References

[1] Alpha Sensors, Micro-thermistor Series. www.alphasensors.com

[2] Weston Electronic Components Corp., NTC Surface MountThick Film Thermistors. www.wecc.com.

[3] Muehlbauer, M. J., Guilbeau, E. J., Towe, B. C., Anal. Chem.1989, 61, 77–85.

[4] Xie, B., Mecklenburg, M., Danilsson, B., Ohman, Ove, Win-quist, F., Anal. Chem. Acta 1994, 299, 165.

[5] Xie, B., Mecklenburg, M., Danilsson, B., Ohman, Ove, Win-quist, F., Analyst 1995, 120, 155.

[6] Mills, G., Zhou, H., Midha, A., Donaldson, L., Weaver, J. M. R.,Appl. Phys. Lett. 1998, 72, 2900.

[7] Hammiche, A., Boxec, L., Conroy, M., Pollock, H. M., Mills,G., Weaver, J. M. R., Price, D. M., Reading, M., Hourston,D. J., Song, M., J. Vac. Sci. Technol. B 2000, 18, 1322.

[8] Swinney, K., Markov, D., Bornhop, D. J., Anal. Chem. 2000,72, 2690.

[9] Swinney, K., Bornhop, D. J., Analyst 2000, 125, 1713.

[10] Swinney, K., Markov, D., Bornhop, D. J., Proc. Natl. Acad.Sci. USA 2001, in press.

[11] Liu, K. K. L., Davis, K. L., Morris, M. D., Anal. Chem. 1994,66, 3744.

[12] Tarigan, H., Kenmore, C. K., Neill P., Bornhop, D. J., Anal.Chem. 1996, 68, 1762.

[13] Weast, R. C. (Ed.), CRC Handbook of Chemistry and Phys-tics, CRC Press, West Palm Beach, FL 1978.

[14] Riddick, J. A., Burger, W. B. (Eds.), Organic Solvents, Wiley-Interscience, New York 1970.

[15] Crystran’s Quick Guide to Optical Silica, Crystran Ltd. http://crystran.co.uk.

2036 K. Swinney and D. J. Bornhop Electrophoresis 2001, 22, 2032–2036