乳剂聚合反应（VC2，ACM）中实时监控检测方案(激光拉曼光谱)

Real-time Monitoring of Polymerisations inEmulsions by Raman Spectroscopy-Modelling andChemometrics Introduction Real-time monitoring of emulsionpolymerization by Raman spectroscopy hasbeen performed. Ramannspectroscopy hasadvantages； over otherr techniquesbeingimplemented for process monitoring. It is well-suited for emulsion polymeri-zation reactionsbecause tthe aqueouss solvent does notinterfere with the analyte signal by eithercontributing to the Raman signal or by maskingthe analyte signal by self-absorption processes.In addition， as it is a scattering phenomenon，the turbidity of an emulsion process is not asevere impediment. On the other hand there are experimentaldifficulties arising from the possibility that thereis fluorescence emission interfering with theRaman signal， and from the difficulties inextracting signals through windows in reactorwalls or in slip streams. In order to avoid theseproblems a near infrared excitation laser (785nm) and an immersion probe dipped directly inthe reaction medium have been used. Two parallel ways for data analysis have thenbeen undertaken. The first is based on amathematical deconvolution of the Ramanbands whereas the second relies on moresophisticated multivariate analysis(Chemometrics). Experimental conditions Thee reactionntto be followedl wasaaCCO-polymerization of vinyl dichloride (VC2) andmethyl acrylate (ACM) (90：10). A glass reactor was especiallydesigned for thisfeasibility study，enabling the insertionof the Raman probevia the cover of thereactor (fig.1). Thepressure wasmaintained atatmosphere level， and the temperature of the medium was increased to 32°C. The quantityof initiator had to be adjusted in order to inducethe polymerization process and to keep thereaction time acceptable(4-5 hours). TheInduRam，，a JobinYvonn on-lineRamananalyser，I，waslused for these tests， andincludesiafiber-coupled immersion probe.Collection optics on this probe can be adaptedto match the optical properties and otherphysical constraints of the medium. Principle of the measurements Polymerization normally proceeds by breakingthe carbon double bond of one monomer； thesecond monomer is then attached via the freeelectrons. Because these bonds generate veryintense Raman bands， one can follow theirdisappearance asSthe polymerisation takesplace. In1summary， the evolution of 1thereaction can be monitored by following theintensity of the bands associated with the C=Cvibrations. Figure 2： Raman immersion probe For this example， the estimated reaction timewas 5 hours. The automated acquisition andstorage of the Raman data was programmedfrom the LabSpec software. A spectrum wasrecorded and saved every five minutes. Thisset of accumulatedl spectra enabled ananalysis， a posteriori， of the reaction. Preliminarv measurements offtthe puremonomers were necessary to identify theircharacteristic Raman features (fig.2). Moreover，thesespectra arealsouseful for ithesubsequent data analysis. Figuree3： Raman Spectra offthee rpuremonomers (ACM： blue； VC2： red) The bands associated with the carbon doublebonds of VC2 and ACM located at 1612 and1632? cm respectively are ratherwellseparated. This enables a univariate analysisof the data， in which the intensity of selectedRaman bands is monitored. Figure 4 illustrates the changes in the Ramanspectrum during the chemical transformation.The reaction starts slowly but then significantlyspeeds up； the monomers tend to disappearmore and more quickly as the synthesisedpolymer (broad band at 1410 cm") appears. Inaddition， because the double bonds of the twomonomers are distinguishable in the spectra，these results show that the conversion rate ofthe VCis higher than that of the ACM. Carefulinspection of the spectra show the intensityratio of the bands at 1612 and 1632 cm-1is notconstant along the reaction. Raman data analysis In order to extract information concerning theprogress of the reaction， modelling underLabSpec was done. Figure 4： Raman spectra recorded at differentstages of the reaction (about every 20 mins) For this， .model. spectra were selected. Theseare， on one hand， spectra of the two puremonomers and on the other hand， a spectrumcollected at the very end of the reaction， in which theebandiassociated with thesynthesized polymer is clearly visible (fig.5).Thedeconvolution operation of LabSpecprovides a calculation of the contributions ofeach of these model spectra to the currentspectrum. The concentration profiles of theconstituents during the reaction are drawn fromthe quantitative values derivedfrom thedeconvolution (fig.6). The mechanisms of themonomer conversions and of the co-polymersynthesis can then be observed， leading to abetter understanding of the properties of thefinal polymer. Figure 6 shows that thebeginning of the polymerization itself seems tooccur approximately one hour after the additionof the initiator. These profiles also enable oneto estimate the time of the transformationprocess to 3hr 50min and confirm the higherconversion rate ofVC. Figure 5. Example of modelling of the current spectrum from spectra of pure monomers andof the medium at the end of the reaction whenmainly the polymer is present. Figure 6. Concentration profiles of the differentreaction constituents. In order to monitor more complex reactions，data analysis based on a multivariate approach(Chemometrics) is expected to be needed.Indeed， when the number of species increases，it is more likely that Raman bands overlap，making the deconvolution method either impossible or rather inaccurate. The advantageof Chemometrics is that it takes into account amore extended part of the spectrum or eventhe entire spectrum. It allows one to extractinformation that may not be accessible at first，when one deals with complex spectra. ThePLS algorithm was used for building up themodel， that was tested by cross validation(fig.7). The training set consisted of 15 spectrataken at equal intervals during the reaction. Figure 7. Results from the cross-validationprocedure The model constructed from the training setproduced concentration predictions that weremade on spectra of the same reaction but thatwere not part of the training set (tab.1). Theresults obtained show/ agood agreementbetween the predictions and the theoreticalvalues given by the reference method. HORIBAJOBIN YVON France ： The following step was to apply this model to asimilar reaction， for whichthe spectrumacquisition time is reduced by a factor 2 (tab.2). The agreement between the predicted andexpected values is not as good but the rangeof the predictions is still acceptable. In fact，part of the discrepancies is due to changes inthee reaction conditions ((initiator quantity，temperature) as well as collection parameters(integration time). In reality when the modelhas been trained to account for such changes，itwill bemuchmore robustto processvariations. Time interval betw.thereaction startand thespectrum collected %ACM %VC2 Polymer(arbitratyunits) Theor. Pred. Theor. Pred. Theor. Pred. 40 mns 8，20 8，16 83，5 80，2 21，3 22，7 1h40 8，64 8，68 74，3 64，8 28，6 30，2 3h20 0，9 1，1 3，8 2，3 91，9 88，6 Table 1 Time intervalbetw. thereaction startand thespectrum collection %ACM %VC2 Polymer(arbitraty units) Theor. Pred. Theor. Pred. Theor. Pred. 50 mns 10，8 6，6 80，1 88，7 31，2 35，9 1h00 4，83 7，82 76，5 78，8 39，2 25，6 1h45 8，3 7，92 50，9 65，9 40，3 27，8 Table 2 Conclusion Raman spectra， in conjunction with Multivariate(Chemometric) Analysis， have beendemonstrated to provide real-time informationon the progress of a polymerisation reaction.As shown by this example， these results canprovide unexpected information on the detailsof the reaction. in this case， the inequivalentreaction rates of the two monomers. Suchinformation ultimately enablesthe processengineer to optimise his process. ( F ax ： +33 ( 0)3 20 59 18 08. Email ： raman@jobinyvon.fr www.jobinyvon.fr ) ( USA： HORIBA Jobin Yvon Inc.， 3880 P a rk Avenue， Edison， NJ 08820-3012. Tel：+1-732-494-8660， F ax： + 1-732-549-2571. E m ail ： raman@jobinyvon.com www.jobinyvon.com Japan ： H ORIBA Ltd.， JY Optical Sales Dept.， 1 -7-8 Higashi-kanda， Chiyoda-ku， Tokyo 101-0031. T el： +81 (0)3 3861 8231 ， Fax ： +81 (0)3 3861 8259 . Email： raman@horiba.com Germany： +49 (0)62518475-0 I taly： +39 0 2 57603050 UK： + 44 (0)20 8204 8 142 3/3 ) ORIBAExplore the future ORIBAExplore the future Raman spectra, in conjunction with Multivariate (Chemometric) Analysis, have been demonstrated to provide real-time information on the progress of a polymerisation reaction. As shown by this example, these results can provide unexpected information on the details of the reaction. in this case, the inequivalent reaction rates of the two monomers. Such information ultimately enables the process engineer to optimise his process.