Development of a compact CO2 sensor based on near-infrared laser technology for enological applications

Appl Phys B (2009) 94: 725–733DOI 10.1007/s00340-009-3389-z

Development of a compact CO2 sensor based on near-infraredlaser technology for enological applications

M. Mulier · V. Zeninari · L. Joly · T. Decarpenterie ·B. Parvitte · P. Jeandet · G. Liger-Belair

Received: 9 September 2008 / Revised version: 9 January 2009 / Published online: 13 February 2009© Springer-Verlag 2009

Abstract This paper reports the development of an infraredlaser spectrometer using commercial diode laser emitting at2.68 µm. The instrument is designed to measure CO2 con-centrations above a glass poured with a sparkling liquid,such as beer or champagne in the present case. This spec-trometer was developed in order to realize the cartography ofCO2 outgassing in the headspace above various glasses. Weprovide details of the instrument design and data processing.Absorption lines were carefully selected to minimize inter-ferences from neighboring water vapor transitions. The in-strument performance allows to measure ambient CO2 con-centrations so that one can be very confident in the CO2

concentrations measurements above the glass. Some prelim-inary results on sparkling liquids such as beer and cham-pagne are presented and compared to a model describing theflux of CO2 discharging from glasses due to the contributionof bubbles.

PACS 07.07.Df · 07.57.Ty

M. Mulier · V. Zeninari (�) · L. Joly · T. Decarpenterie ·B. ParvitteGroupe de Spectrométrie Moléculaire et Atmosphérique, UMRCNRS 6089, UFR Sciences Exactes et Naturelles, Moulin de laHousse, BP 1039, 51687 Reims Cedex 2, Francee-mail: [email protected]: +33-03-26913147

M. Mulier · P. Jeandet · G. Liger-BelairLaboratoire d’Oenologie et Chimie Appliquée, URVVC EA 2069,UFR Sciences Exactes et Naturelles, Moulin de la Housse,BP 1039, 51687 Reims Cedex 2, France

1 Introduction

From a strictly chemical point of view, champagne andsparkling wines are multicomponent hydro-alcoholic sys-tems supersaturated with CO2-dissolved gas moleculesformed together with ethanol during the fermentationprocess. Champagne and sparkling wine tasting mainlydiffers from still non-effervescent wine tasting due to thepresence of carbon dioxide bubbles continuously risingthrough the liquid medium (the so-called effervescenceprocess) [1–3]. It is worth noting that approximately 5 litersof gaseous dissolved CO2 must escape from a typical 0.75liter champagne bottle. For a recent review about effer-vescence in glasses poured with champagne and sparklingwines, see, for example, [3] and references therein.

From the consumer point of view, the role of bubbling isindeed essential in champagne, in sparkling wines, and evenin any other carbonated beverage. Without bubbles cham-pagne would be unrecognizable, beers and sodas would bedefinitely flat. However, the role of effervescence is sus-pected to go far beyond the solely aesthetical point of view.Actually, in enology, effervescence is believed to play amajor role concerning flavor release and CO2 discharge inglasses poured with champagne and sparkling wines [4].The myriad of bubbles nucleating on the flute’s wall andtraveling through the wine’s bulk considerably enhances theperception of volatile organic compounds by considerablyenhancing exchange surfaces between the wine and the at-mosphere. However, each bubble collapsing at the wine’ssurface inevitably frees its tiny CO2 volume. Consequently,the inevitable counter party of the “exhausting” aromas ef-fect attributed to bubbles’ exchange surfaces, is to progres-sively bring some gaseous CO2 in the headspace above thewine’s surface. Actually, from the consumer point of view,the release of a sudden and abundant quantity of CO2 above

726 M. Mulier et al.

the champagne surface is known to strongly irritate the noseduring the evaluation of aromas. Moreover, it was demon-strated recently that the continuous flow of ascending CO2

bubbles through a sparkling wine strongly modifies the mix-ing and convection conditions of the liquid medium [4, 5].In turn, the release of the numerous volatile and potentiallyaromatic organic compounds from the wine surface (whichstrongly depends on the mixing flow conditions of the liq-uid medium) may be considerably enhanced compared tothe case of flavor release from a still non-effervescent whitewine [6].

Suffice to say that a strong coupling therefore exists inenology between rising bubbles, CO2 discharge and fla-vor release. This is the reason why glassmakers progres-sively became highly interested in proposing to consumersa new generation of champagne tasting glasses, especiallydesigned, with a well controlled CO2 release all along tast-ing [5]. The development of a tool providing an accurate andcontinuous measurement of gaseous CO2 in the headspaceabove glasses poured with carbonated beverages (in tastingconditions) is therefore considered as a first step helping usin better understanding the role of bubbling and glass-shapeon the kinetics of CO2 release from glasses showing variousglass-shape and effervescence conditions.

Laser diode spectrometry is an effective tool to pro-vide accurate carbon dioxide concentration measurements[7]. In particular, a new generation of distributed-feedbackGaInAsSb laser diodes emitting at room-temperature near2.7 µm are promising for the development of highly com-pact laser sensors [8]. The spectral emission propertiesof these lasers (monochromaticity, absence of mode-hops,wavelength tunability over a few reciprocal centimeters, alaser line width of ∼10 MHz, optical output power of afew milliwatts) are very suitable for the determination of gasconcentration by infrared absorption spectroscopy. Further-more, near 2.7 µm, carbon dioxide has strong rotational–vibrational transitions which are well suited for CO2 mea-surement with a small pathlength. By propagating the laserbeam near 2.7 µm over 60 cm in the spectrometer open to theatmosphere, roughly 15% of the laser energy is absorbed byambient CO2 in the troposphere (natural abundance of CO2

in air is around 380 ppm). This value is sufficient to obtainan error lower than a few percent in the CO2 concentrationretrieval for a measurement time of one second. By propa-gating the laser beam near 2.7 µm over 6 cm (typical diame-ter of the glass) in the spectrometer, roughly 50% of the laserenergy is absorbed by 2% of outgassing CO2. This kind ofabsorption corresponds to the best condition for the retrievalof the concentration.

The objective of the present work is to develop a com-pact sensor to measure CO2 concentrations above a glasspoured with a carbonated beverage (such as beer or cham-pagne in the present case). Such a sensor is described in

the next section of this paper. Details of the data process-ing are described in the subsequent section. Finally, we dis-cuss preliminary measurements achieved with the laser sen-sor. A set of concentration data obtained with beer or cham-pagne is recorded to demonstrate the capability of the sensorto measure CO2 concentrations above a glass poured with asparkling beverage.

2 Instrument design

The new-generation diode lasers with an emission wave-length around 3700 cm−1 (2.7 µm) are suitable for moni-toring carbon dioxide. The magnitude of the line strengthsmakes it possible to measure CO2 with a small absorptionpath length. However, considerable care must be taken inthe choice of the CO2 absorption lines because of potentialinterference with the ν1 and ν3 fundamental bands of watervapor. To avoid overlapping with neighboring H2O absorp-tion lines, the R18 transition of the (1001)I → (0000) bandof CO2 was selected [8]. The appropriate spectroscopic pa-rameters for this transition at 3728.41 cm−1 are an inten-sity of 5.712 × 10−20 (cm−1/(molecule cm−2)) and an air-broadening coefficient of 0.722 cm−1/atm [9].

The CO2 measurements are performed at high resolu-tion using a direct absorption spectrometer with a new-generation InGaAsSb tunable diode laser. The direct detec-tion technique is straightforward. A three-dimensional viewof the optical platform is shown in Fig. 1. This instrument isderived from the instrument described in [8] but significantimprovements have been done both in the instrument designand in the data processing. The improvements are detailedin the text.

The instrument is mounted on a 67 × 50 cm2 board.Thus, the instrument is more compact than that presented in[8]. The laser beam is collected by an aspheric ZnSe lens.The laser beam passes through the atmosphere above theglass. A ZnSe beam splitter (R = T = 50%) is used to sep-arate the laser beam into two parts. The reflected beam iscoupled with a Germanium Fabry–Pérot (FP) etalon L5940from Laser Components (free spectral range (FSR) = 45 ×10−3 cm−1) in order to obtain relative frequency calibra-tion. The FSR of the Ge Fabry–Pérot etalon is 4.5 times theetalon in [8]. This type of value is more interesting whenatmospheric pressure spectra are recorded. This etalon canbe used from 2 to 20 µm wavelength range. Its 2.54-cmlength and its finesse (typically 3) make it a quick-to-setup-and-ready-to-use and a very convenient component for theinstrument compactness and portability. Thus, this kind ofetalon permits obtaining a more compact instrument. Bothbeams are focused by a ZnSe lens on two InAs photodiodesmounted with thermistors on two-stage thermoelectric cool-ers from Judson. Ambient temperature and pressure condi-tions are measured by a weather transmitter WXT510 from

Development of a compact CO2 sensor based on near-infrared laser technology for enological applications 727

Fig. 1 A three-dimensionalview of the CO2 spectrometerdesigned to measure CO2concentration above glassespoured with carbonatedbeverages

Vaisala: barometric pressure is given with an accuracy ±0.5hPa at 0–30◦C and temperature is given with an accuracy±0.3◦C at +20◦C.

The laser diode is a commercially available distributedfeedback laser diode emitting at 2.68 µm from Nanoplus.This laser diode covers the frequency range from 3726 cm−1

to 3733 cm−1. The average output power is a few mW. Evenif this laser emission is similar to that of [8] this laser is anew one and some details permit to differentiate them. Forexample, the laser diode is operated at room temperature,typically at 30◦C (15◦C in [8]). This different set-point oper-ation is more interesting because it is further from the usuallaboratory temperature around 20◦C. Thus the Peltier ele-ment gives better temperature stability even if the lab tem-perature slightly varies. The threshold current is approxi-mately of 70 mA. A Thorlabs laser driver (model LDC 500)is used to control the laser current and a Thorlabs thermoelectrical controller (model TEC 2000) is used for the tem-perature control. The laser emission wavelength is scannedover the molecular transition by applying a triangular cur-rent ramp (10 ms period) to the laser diode. A Digital Sam-pling Oscilloscope Lecroy LT264 was used to perform dataacquisition in the original setup [8]. In order to enhance thevertical resolution, data acquisition is now performed by aData Translation DT9832 16 bits analog to digital sampler.2500 points are acquired over the 10 ms triangular laserscan. Typically, 20 successive scans are averaged to opti-mize signal to noise ratio. Two channels are acquired simul-taneously: the transmission over the open path (OP) and thetransmission from the Fabry–Pérot interferometer (FP). Anexample of a typical recorded spectrum is shown in Fig. 2.

The CO2 sensor is controlled by a portable computer. Theoperating program was developed with Matlab 7.2 software.

Fig. 2 An example of recorded experimental spectra about 10 minutesafter pouring: experimental absorption spectrum and Fabry–Pérotfringes

The program was converted to a standalone application us-ing the Matlab compiler. It performs the following tasks atperiodic time intervals: connection to the weather transmit-ter to acquire the weather parameters (RS232 connection)and connection to the DT9832 analog to digital sampler witha USB protocol to record the two signals (OP) and (FP).Weather data and spectra are then collected by the computer.Due to the communication protocols, the minimal time in-terval between consecutive measurements is currently about1 s.

3 Data processing

The main issue in processing spectra recorded at atmospher-ic pressure is baseline correction. For laboratory measure-

728 M. Mulier et al.

ments we usually determine the baseline by applying athird-degree polynomial interpolation over full transmissionregion on both sides of the absorption line. However, atground level, under standard conditions, there are no full-transmission regions in the experimental CO2 spectra: pres-sure broadening of the CO2 lines prevents to reach full trans-mission regions and furthermore several water vapor ab-sorption lines are close to the selected line. In order to im-prove the retrieved CO2 concentration, the data processinghas been modified in comparison with [8], and the followingoperations are now performed during data processing:

1. Fabry–Pérot (FP) signal is used to perform the wavenum-ber calibration. The frequency calibration is determinedby applying a fifth-degree polynomial interpolation to theinterference fringes in the FP signal. The absolute valueof the frequency is known from the position of the mainabsorption line in the transmission.

2. A ‘first guess’ for the baseline is obtained by subtractinga calculated absorptivity with Lorentzian line shape andfor mean ambient CO2 concentration to the natural log-arithm of the transmission. This baseline is modeled bya fourth degree polynomial interpolation. Polynomial de-gree is limited to 4th degree in order to smooth the imper-fect initial value of the absorptivity. In order to determinethe optimal degree for the polynomial, the same datasethas been processed using different values of the degreeand the value that gave minimum statistical dispersionhas been chosen. This procedure allows improvement ofthe statistical dispersion by a factor of 2 when comparedto the procedure described in [8]. This statistical disper-sion is typically about 10 ppm.

3. A least-squares fitting procedure is applied to minimizethe difference between the natural logarithm of the trans-mission and the sum of calculated absorptivity and base-line. During the fitting procedure, the [concentration(c) × length (L)] parameter, the main CO2 line centerand the coefficients of the polynomial baseline are ad-justed.

4. Before pouring the carbonated beverage, ambient CO2

concentrations are directly obtained by

c0 = [c × L]/L0 with L0 = 60 cm. (1)

5. After pouring, approximate CO2 concentrations over theglass are obtained by:

cg = [c × L] − c0(L0 − Lg)

Lg, (2)

where Lg is the glass diameter.

Another procedure would be to divide “after” measurementsby “before” measurements and then to directly fit the CO2

concentration. However, this technique needs high stability

Fig. 3 An example of recorded spectrum before pouring. The lineshape was fitted with a Lorentz profile. The absorption path length is60 cm, the temperature is 296 K and the pressure is 1000 mbar. Theretrieved CO2 concentration is around 370 ppmv. The estimated errorfrom the least square fitting is around 4 ppmv

of the laser; hence, good stability of the temperature. Thiscondition was not realized over our measurement time > 15minutes. This will be realized with the next version of theinstrument. One can note that both procedures neglect thehorizontal diffusion of CO2 in the ambient air. The deter-mination of an exact value for cg would require a betterknowledge of horizontal diffusion and of convection abovethe glass. This interesting problem will be studied soon.Anyway, (2) can be applied directly at least for the first 10minutes of measurement and it also gives a very good roughguess of the measured concentrations for the next part of themeasurements.

An example of a recorded spectrum of ambient CO2 isshown in Fig. 3, and a recorded spectrum of CO2 concentra-tions reached above a glass filled with champagne is shownin Fig. 4. The line shape was modeled with a Lorentz profileand the resulting residual is also displayed in the lower panelof Figs. 3 and 4. The residual (Exp-Calc) with the Lorentzprofile is less than 1% demonstrating the high quality of therecords. The estimated error from the least square fitting isaround 4 ppmv.

The atmospheric CO2 concentration measurement is il-lustrated in Fig. 5 and was obtained after a purge of thelaboratory room air. This result is in very good agreementwith the natural CO2 abundance: (370 ± 12) ppmv. Highervalues such as 450–500 ppm may be obtained if there is hu-man activity in the lab with closed windows. Although thisdispersion of 12 ppm is not yet sufficient for ambient CO2

monitoring, it is sufficient for our application where the con-

Development of a compact CO2 sensor based on near-infrared laser technology for enological applications 729

centration to be measured is typically between 30 000 and3000 ppm.

4 Measurements of gaseous CO2 concentration above aglass poured with a carbonated beverage

The designed instrument was then used to access measure-ments of CO2 concentrations above glasses poured withsparkling beverages. A lot of preliminary experiments havebeen necessary to develop a complete protocol for measure-ments, such as isolation of the set-up and cleaning procedureof the glass before it was filled.

Fig. 4 An example of recorded spectrum after the pouring of cham-pagne. The line shape was fitted with a Lorentz profile. The absorp-tion path length is 6 cm, the temperature is 296 K and the pressure is1000 mbar. The retrieved CO2 concentration is around 20 000 ppmv,i.e., 2%. The estimated error from the least square fitting is around 4ppmv

Standard commercial Champagne wine and beer holdingabout 10 g/L and 5 g/L of CO2-dissolved molecules, re-spectively, were used for this set of experiments. In orderto avoid the randomly located “bubbling environment” in-evitably provided in glasses showing natural effervescence,we finally decided to use, for this set of experiments, a singlestandard flute etched at its bottom (thus providing a “stan-dardized” and artificial effervescence). Champagne and beerwere thus poured into a standard commercial flute etchedat its bottom (as the one with the ring-shaped engravementdisplayed in Fig. 6). Glasses etched at their bottom are thusindeed easily recognizable with a characteristic bubble col-umn rising on their axis of symmetry.

Between the successive pouring and data recordings, theflute was systematically thoroughly washed in a dilute aque-ous formic acid solution, rinsed using distilled water, andthen compressed air dried. This drastic treatment forbids theformation of tartrate crystals on the flute wall as well as theadsorption of any dust particle acting as “natural” bubblenucleation sites [1–3]. Therefore, in this case, the CO2 re-lease out of the carbonated beverage is mainly related tothe bubble nucleation sites of the ring-shaped etching, sothat likely differences in CO2 concentrations above the glasspoured with various sparkling beverage are attributed only tophysicochemical differences between the beverages them-selves. Experiments were performed at room temperature(20 ± 2◦C).

The results of the experiment on isolation of the set-upare presented in the graph displayed in Fig. 7. A part of thefirst ten minutes of this experiment corresponds to Fig. 5.Then, after pouring champagne into the glass, the CO2 con-centration increases up to ∼2.5% and then decreases withtime. The black points of Fig. 7 correspond to CO2 con-centrations measured above the flute without box. There areprobably air movements in the room, thus leading to smallvariations of CO2 concentration in time. Even if this sit-uation is closer to the “real” conditions experienced dur-ing champagne or sparkling wine tasting, the results are not

Fig. 5 Atmospheric CO2concentrations measured beforepouring the sparkling beverageinto the glass. Different seriescorrespond to variousexperiments. All the series havebeen superimposed

730 M. Mulier et al.

Fig. 6 a At the bottom of thisflute, on its axis of symmetry,the glassmaker has engraved asmall ring (done with adjoininglaser beam impacts); b Singlelaser beam impact as viewedthrough a scanning electronmicroscope (bar = 100 µm);c Effervescence in this flute ispromoted from these “artificial”microscratches into the form ofa characteristic and easyrecognizable vertical bubblescolumn rising on its axis ofsymmetry (bar = 1 mm)(Photographs by G. Polidori andF. Beaumont)

Fig. 7 CO2 concentrationsabove a glass after champagnewas poured. Different seriescorrespond to variousexperimental conditions ofisolation of the set-up (see textfor more details)

quite satisfactory from a strictly scientific point of view. Thegray points of Fig. 7 correspond to the same measurementconditions with a big cardboard box to try to isolate the set-up from air movements. In this case, the retrieved concen-trations are less noisy. The dimensions of the “big” box are105 cm length × 70 cm width × 85 cm height. The use ofthis big box provides more satisfactory results and demon-strates that the laser spectrometer gives reliable measure-ments. Finally, the light-gray points are related to the same

measurements when adding a smaller box around the glass,thus reducing once again the inevitable air mass motionaround the set-up. The dimensions of the “small” box are30 cm length × 15 cm width × 40 cm height.The obtainedconcentrations are more and more close to each other withtime. However, it is clear that the use of this additional smallbox leads to an undesirable accumulation of carbon diox-ide above the glass. All these remarks are easier to verifyin Fig. 8. This figure corresponds to a zoom on the last part

Development of a compact CO2 sensor based on near-infrared laser technology for enological applications 731

Fig. 8 Zoom on the last part ofFig. 7

Fig. 9 CO2 concentrationsabove the flute all along the first25 minutes following thepouring of beer or champagne,respectively

of Fig. 7. One can also remark that the CO2 concentrationsmeasured with big and small boxes are significantly higherthan those measured without box, thus betraying an undesir-able and quite “artificial” accumulation of CO2. After theseexperiments we have finally decided to make all other mea-surements with the big box around the set-up. One can notethat the observed dispersion in Fig. 8 is greater than 12 ppmdue, for example, to fluctuations above the glass (convectionprocess and random burst of bubbles at the surface).

The next experiment has been realized in order to com-pare the kinetics of CO2 concentrations measurementsabove the flute poured successively with champagne andbeer, all along the first 25 minutes following pouring. Theresults of this experiment are presented in Fig. 9. A partof the first seven minutes of the experiment corresponds toFig. 5. Then, champagne or beer is poured into the glass,and CO2 concentrations are measured above the glass allalong the first 25 minutes following pouring. This experi-ment seems to demonstrate that the diffusion of CO2 abovethe two tested sparkling beverages is quite different. By useof this set-up and operating conditions, significant differ-

ences in CO2 concentrations above the glass were evidenceddepending on whether it was poured with champagne orbeer.

Immediately after pouring, CO2 concentrations mea-sured above the glass poured with champagne are approxi-mately twice those reached above the glass poured with beer(around 2.5% for champagne and only 1.3% for beer). Then,the decrease of the gaseous CO2 concentration above theflute which ensues is faster for champagne than for beer. Af-ter approximately 15 minutes the ‘Beer’ and ‘Champagne’curves are superimposed. We are logically tempted to won-der why such differences appear between champagne andbeer. The aim of the following paragraph is to discuss anddepict the contribution of CO2 bubbles collapsing at the liq-uid surface to the global kinetics of CO2 release from a glasspoured with a sparkling beverage.

5 Modeling the flux of CO2 discharging from glasses:the contribution of bubbles

For the reasons detailed above, since we used a single fluteengraved at its bottom for each data recording, differences

732 M. Mulier et al.

evidenced between the kinetics of CO2 release from cham-pagne and beer glasses are attributed only to physicochemi-cal differences between the sparkling liquid themselves. Ac-tually, the contribution of rising CO2 bubbles to the globalkinetics of CO2 release from a sparkling beverage is quiteeasily accessible by taking into account the number of nu-cleation sites found in the glass, the average frequency ofbubble production from a nucleation site, and the averagesize of a bubble as it reaches the liquid surface (to finallycollapse and release its CO2 content in the headspace abovethe liquid surface). Recently, models based on both classicaldiffusion and bubble rising velocity were developed in orderto propose scale laws likely to link the frequency of bub-ble nucleation (i.e., the number of bubbles released per sec-ond from a given nucleation site) as well as the size of CO2

bubbles rising in a carbonated beverage with some physico-chemical parameters of the liquid medium [10, 11]. The fre-quency of bubble formation from a single bubble nucleationsite, denoted f , was found to obey the following scaling law[10]:

f ∝ θ2(cL − kHP)

ηP(3)

where θ is the liquid temperature, cL is the bulk concen-tration of CO2 in the liquid medium, kH is the so-calledHenry’s law constant (i.e., the solubility of the CO2 mole-cules with regard to the liquid medium), P is the ambientpressure, and η is the liquid viscosity.

The diameter of a bubble reaching the liquid surface wasalso found to depend on various parameters. The diameter ofa bubble, denoted d , was found to obey the following scalinglaw [11]:

d ∝ θ5/9(

1

ρg

)2/9(cL − kHPB

PB

)1/3

h1/3, (4)

where ρ is the liquid density, g is the acceleration due togravity, and h is the distance traveled by the bubble fromits nucleation. PB is the pressure inside the rising bubble.Strictly speaking, the pressure PB inside the rising bubbleis the sum of two terms: (i) the atmospheric pressure P , and(ii) the Laplace pressure 4γ /d originated in the bubble’s cur-vature (γ being the surface tension of the liquid medium).However, the surface tension of champagne and beer beingof order of 50 mN m−1 [1], and bubbles’ diameters vary-ing from several tens to several hundreds of micrometers, itis worth noting that the contribution of Laplace pressure isclearly negligible in front of the atmospheric pressure P .

Finally, the contribution of the bubble nucleation processto the global flux of CO2 from a sparkling beverage pouredin the engraved flute may be accessed by multiplying thenumber N of nucleation sites found in the flute by the av-erage frequency f of bubble nucleation and by the average

volume v of a bubble collapsing at the liquid surface. There-fore, by combining the two above-mentioned scaling laws,the flux of gaseous CO2 released by bubbles rising and col-lapsing in a carbonated beverage poured into a glass, de-noted dV/dt , may be ruled by the following scaling law:

dV

dt= Nf v ∝ f d3 ∝ θ11/3

η

(1

ρg

)2/3(cL − kHP

P

)2

h. (5)

It is clear from (5) that the flux of CO2 released by bub-bles collapsing above the flute strongly depends on variousparameters. Nevertheless, under the same operating condi-tions, the only parameter which strongly differs betweenchampagne and beer is cL, i.e., the bulk concentration ofdissolved-CO2 (actually, the solubility of CO2 in cham-pagne is quite close to that in beer, and both the viscos-ity and density of champagne and beer are of the same or-der of magnitude [11]). Following the latter equation, thehigher the initial concentration of CO2 in the liquid bulk,the higher the flux of gaseous CO2 released from the liq-uid medium. Actually, champagne contains approximatelytwice more dissolved-CO2 than beer (around 10 g/L inchampagne against around 5 g/L in beer). As a result, dif-ferences in gaseous CO2 concentrations found above theflute poured with champagne and beer (shown in Fig. 9) aremainly attributed to differences in the initial bulk concentra-tion of dissolved CO2 between champagne and beer.

6 Conclusions and prospects

We have reported the development of a compact laser sensorto measure CO2 concentrations above a glass poured withsparkling liquids. The design of the sensor and the details ofthe data processing technique are reported. Atmospheric car-bon dioxide may be measured with this instrument. An up-grading of our sensor will include improved stabilization ofthe diode laser emission. The set-up was designed in a waythat allows many CO2 measurements by changing the posi-tion of the glass with regard to the laser beam (horizontallyas well as vertically). In a near future, we would like to ob-tain a complete cartography of gaseous CO2 concentrationsfound in the headspace restricted above champagne glasses,in tasting conditions. We plan to test the influence of theglass design on the kinetics of CO2 release from a sparklingbeverage, by use of glasses showing various glass shapeand engravement conditions, for example. We also plan totest the influence of other various parameters of the liquidmedium on the kinetics of CO2 release, such as its tempera-ture, viscosity, and liquid level in the glass, for example.

Acknowledgements The authors are grateful to champagne Moët &Chandon and Pommery for regularly supplying us with wine samples,

Development of a compact CO2 sensor based on near-infrared laser technology for enological applications 733

and to ARC-International for supplying us with glasses and for sup-porting our research. Authors are indebted to the Région Champagne-Ardenne, The Ville de Reims, the Conseil Général de la Marne and theMinistry of Research for financially supporting of the PhD of MaximeMulier.

References

1. G. Liger-Belair, M. Vignes-Adler, C. Voisin, B. Robillard, P. Je-andet, Kinetics of gas discharging in a glass of champagne: Therole of nucleation sites. Langmuir 18, 1294 (2002)

2. G. Liger-Belair, F. Beaumont, P. Jeandet, G. Polidori, Flow pat-terns of bubble nucleation sites (called fliers) freely floating inchampagne glasses. Langmuir 23, 10976 (2007)

3. G. Liger-Belair, G. Polidori, P. Jeandet, Recent advances in thescience of Champagne bubbles. Chem. Soc. Rev. 37, 2490 (2008)

4. G. Liger-Belair, J.-B. Religieux, S. Fohanno, M.-A. Vialatte, P.Jeandet, G. Polidori, Visualization of mixing flow phenomena inchampagne glasses under various glass-shape and engravementconditions. J. Agric. Food Chem. 55, 882 (2007)

5. G. Liger-Belair, F. Beaumont, M.-A. Vialatte, S. Jégou, P. Jean-det, G. Polidori, Kinetics and stability of the mixing flow patterns

found in champagne glasses as determined by laser tomographytechniques: likely impact on champagne tasting. Anal. Chim. Acta621, 30 (2008)

6. M. Tsachaki, A.-L. Gady, M. Kalopesas, R. Linforth, V. Athès,M. Marin, A. Taylor, Effect of ethanol, temperature and gas flowrate on volatile release from aqueous solutions under dynamicheadspace dilution conditions. J. Agric. Food Chem. 56, 5308(2008)

7. V. Zéninari, B. Parvitte, L. Joly, T. Le Barbu, N. Amarouche, G.Durry, Laboratory spectroscopic calibration of infrared tunablelaser spectrometers for the in situ sensing of the Earth and Mar-tian atmospheres. Appl. Phys. B, Lasers Opt. 85, 265 (2006)

8. L. Joly, B. Parvitte, V. Zéninari, G. Durry, Development of acompact CO2 sensor open to the atmosphere and based on near-infrared laser technology at 2.68 µm. Appl. Phys. B, Lasers Opt.86, 743 (2007)

9. L.S. Rothman, D. Jacquemart, A. Barbe, D. Chris Benner, M. Birk,L.R. Brown et al., The HITRAN 2004 molecular spectroscopicdatabase. J. Quant. Spectrosc. Radiat. Transf. 96, 139 (2005)

10. G. Liger-Belair, M. Parmentier, P. Jeandet, Modeling the kinet-ics of bubble nucleation in champagne and carbonated beverages.J. Phys. Chem. B 110, 21145 (2006)

11. G. Liger-Belair, Nucleation, ascension et éclatement d’une bullede champagne. Ann. Phys. Fr. 31, 1 (2006)