Use of dielectric spectroscopy for real-time in-situ reaction monitoring.

Coating resin manufacturing requires knowledge of the extent of reaction during resin synthesis so the appropriate actions can be taken (addition of the next reactant, reaction termination, etc.). This article reports the results from experiments conducted to survey the utility of dielectric spectroscopy (DES) as a real-time, in-situ technique to monitor the extent of reaction during synthesis of three low molecular weight resins that are representative of components used in coatings formulations. The resins made were based on very different chemistries: (1) a 100% solids polyacrylate functional oligomer from the Michael reaction between a polyacrylate monomer and an acetoacetate ester; (2) an acrylate functional monomer from the reaction between an epoxy ester and acrylic acid; and (3) a solvent-based isocyanate-terminated polyurethane prepolymer from the reaction between a mixture of diols and excess diisocyanate. In all three cases, very good to excellent correlations were found between continuous real-time DES output and the values of characteristic QC parameters (viscosity, acid number, epoxy equivalent weight, % NCO, ATR-FTIR peak heights for reactants and products, and GPC data) determined by off-line analysis of samples taken periodically during the reactions.

Keywords: Dielectric spectroscopy, DES, FDEMS, coating resin synthesis, residual monomer, Michael reaction, acetoacetate, polyacrylate, acrylic acid, esterification, viscosity, acid number, epoxy equivalent weight, polyurethane prepolymer, residual isocyanate, weight percent NCO, infrared spectroscopy, ATR-FTIR, titration, in-process samples, off-line testing, real-time monitoring, in-situ monitoring, gel permeation chromatography, GPC, electrochemical impedance spectroscopy, QC/QA testing, manufacturing process control

**********

During coating resin production, it is critical to know the extent of the reaction (amount of time needed before going on to the next step). This issue has traditionally been addressed by taking in-process samples during the reaction and evaluating them off-line for the critical reaction parameters that are appropriate indicators of reaction completeness. Examples of such tests used in quality control include titrations (acid number, epoxy equivalent weight, and residual isocyanate), product solubility measurements (cloud point), product viscosity (Gardner, bubble seconds, cone and plate), chromatographic techniques (GC, LC, GPC), and the strength of a characteristic peak in the ATR-FTIR spectrum, to name a few. The most serious problems with this approach are the effort it takes to collect the samples and the challenges of performing the analyses in a timely manner in a production environment. Under these circumstances, if the key parameter is changing rapidly, it is possible that a practical (low) sampling frequency will miss the endpoint with the result that the resin will be out of specification by the time the data can be acted upon.

An inadequate solution to this enigma is to make resin specifications very broad, but this usually results in products with large performance differences at the extremes of the upper and lower control limits. The most desirable solution is to monitor at least one key parameter continuously during synthesis. While the power consumption of the reactor impeller may be useful to imply reaction viscosity, this approach has limited sensitivity and is best suited to situations where very high viscosity products are formed at the reaction temperature, something not often encountered with resins targeting coating applications.

Dielectric spectroscopy (DES), sometimes also referred to as frequency dependant electromagnetic sensing (FDEMS), is capable of monitoring changes in the response of materials to oscillating electrical fields in real time. Work performed in thermoset composites has clearly demonstrated that DES can distinguish between slight differences in resin cure state during processing and cure in the tool, (1) especially at high degrees of conversion, with very high sensitivity based on the effects of changing network structure on ionic conductivity (diffusion) and dipolar and molecular (rotational) relaxation rates. Some work has also demonstrated correlation with changes in liquid resin characteristics of the type measured by in-process sampling during coating resin synthesis. (2)

The objective of this work was to generate continuous DES data during the synthesis of several different low viscosity resins that are representative of those used in coatings. Correlations between this data and traditional QC type data from in-process samples would then be sought in order to establish the potential utility of this technique for use during manufacturing lower molecular weight, liquid resins that are representative of resin types used in coatings applications.

BASIC THEORY BEHIND DES

Dielectric spectroscopy is based on the tendency of ions and dipoles to orient in an electric field (Figure 1). Ions often originate as impurities in the raw materials or corrosion products from piping, but can also be generated during synthesis (chloride, for example, from synthesis of epoxy resins from epichlorohydrin), or be added on purpose (catalysts). Dipoles result when atoms with unequal electronegativities are attached to each other in a chemical bond. Typical examples include water and functional groups present in resins, such as alcohol (-OH), halogen (i.e., chloride: C-CI), ester (-CO-OR), acid (-CO-OH), amide (-CO-NH-R), urethane (-O-CO-NH-R), and so on.

If the electric field is reversed, the dipoles will reorient with the field and the ions migrate (diffuse) to the other electrode. As the field reversal rate (frequency) increases, it becomes increasingly difficult for the ions and dipoles to keep up, so a phase delay develops and the separation of the ions or orientation of the dipoles becomes less distinct (lower signal strength). As the groups on either side of the dipole become larger, as in a growing polymer, it becomes more difficult for them to reorient with the field, independent of the frequency of reversal, and signal strength (capacitance) decreases. Finally, as the system becomes more viscous, ion diffusion (conductivity) becomes slower, and signal strength decreases with increasing frequency. These behaviors are referred to as frequency dependant or complex permittivity.

These behaviors can be represented by a series of equations that bear resemblance to material mechanical and rheological parameters. As in mechanics (where complex modulus G* = G' - iG") and rheology (where complex compliance J* = J' - iJ"), complex permittivity (where [epsilon]* = [epsilon]' - i[epsilon]") has both real (storage) [epsilon]' and imaginary (loss) [epsilon]" components that relate to what happens to energy put into the system. Changes in these characteristics are based on structural changes that originate at a submolecular level. Furthermore, because DES interrogates the system with very small probes (ions and dipoles), subtle differences in chain architecture or network structure that affect ion diffusion or molecular relaxation processes can be detected, which in turn can be correlated to changes in macroscopic properties such as [T.sub.g] and viscosity.

[FIGURE 1 OMITTED]

Dielectric measurements determine the voltage and current between a pair of electrodes, allowing capacitance and conductance to be calculated. Capacitance is a measure of energy storage, while conductance is a measure of dissipated or lost energy. When a material is placed between a pair of electrodes, the applied potential polarizes the charge distribution within the material (the ions and dipoles orient in relation to the applied field, as in Figure 1).

Dielectric measurements are typically made at a range of frequencies and an impedance analyzer interprets the data. By making the impedance measurements over a range of frequencies, the frequency dependence of capacitance (C, energy storage) and conductance (G, energy loss) can be determined. Conductance G is important at higher temperatures, lower frequencies, and lower bulk viscosities, where the diffusion of the conducting ionic species can keep pace with the field reversals. Therefore, in an uncured coating resin, ionic conductance dominates the impedance.

The specific conductivity, [sigma], is defined as [sigma] = [[epsilon].sub.o][omega][epsilon]", where [[epsilon].sub.o] is the permittivity of a vacuum, [omega] is the frequency, and [epsilon]" is the dielectric loss factor. The inverse of conductivity is resistivity, [rho], defined as [rho] = 1/[sigma]; a commercial instrument manufacturer refers to this as "ion viscosity" to qualitatively call to mind the effects of network changes on ion and dipolar mobility, in analogy to rheological viscosity. Conductivity also strongly depends on the number of ions present, their type (which substantially influences ion mobility), and conduction mechanism. For example, changing the H-bonding functionality during polymer synthesis (when making epoxy, polyamide, polyurethane, and polyester networks, among others) will influence conductivity by changing the nature of proton-conducting species in the system, which changes charge mobility concurrently with molecular weight growth. (3) Therefore, since [sigma] = [SIGMA][N.sub.i]*[[mu].sub.i], where [[mu].sub.i] is the mean ionic mobility and [N.sub.i] is the ion concentration of ion type i, in cases where ion concentration and type do not change, conductivity changes are related to changes in polymer and network structure and temperature through their influence on ion mobility.

The ionic conductivity and rheological resin viscosity bear an inverse relationship through the Stokes-Einstein equation, [sigma] = [[eta].sup.-a], where, in reality, the exponent a is typically 0.7-0.9 instead of the theoretical 1.0. Consequently, it is possible to imply changes in resin rheological viscosity by measuring changes in [epsilon]". These changes can be correlated to other resin properties that are influenced by molecule-scale changes in the system. Therefore, conductivity (in units of siemens/cm) or resistivity (in units of ohm-cm) are convenient parameters to use for correlations between resin electrical properties and macroscopic thermomechanical and rheological resin pro-perties.

[GRAPHIC OMITTED]

COMMENTS ON DATA INTERPRETATION

Any empirical correlations depend on the quality of the data correlated. While this is obvious, it speaks to the need for both consistency when collecting the data and verification that the parameters recorded are valid. The high degrees of correlation between changes in the continuous data from the DES and off-line data from in-process samples collected during this work support the assumption that the parameters measured were meaningful with respect to the reactions run.

The commercial DES probe used contains both the concentric field electrodes and a thermocouple, so temperature versus time data at the electrode were recorded continuously along with impedance data during the entire course of the reactions. One of the most obvious correlations found in the lab work was the influence that slight changes in reaction temperature had on the resistivity during "isothermal" hold periods. Variations in reaction temperature due to heater cycling around the setpoint (less than about [+ or -] 2[degrees]C in these cases) caused corresponding small variations in resistivity to be superimposed on its "real" value. These variations were accommodated by plotting the "moving average" of resistivity (over a 5-10 min interval, exceeding the heater cycle period) versus the off-line parameters. In pilot plant or commercial scale reactors, these temperature variation cycles will have a much longer period and smaller amplitude due to the larger reaction mass and thermal "inertia," so their effects on resistivity will be correspondingly smaller.

In addition, because of the effects that temperature and reaction mixture composition have on resistivity, any changes made to the reaction components (or conditions) are clearly visible in the resistivity data plotted versus time. This means that most catalyst or reactant additions are clearly visible, and should be characteristic in magnitude for the materials and compositions. Therefore, this running record can be used as a first order check to verify that the process run in production corresponded to the procedure specified and to help identify sources of product lot-to-lot variation.

There were two pairs of reactions run to synthesize this oligomer. The intended differences were the catalysts used; inadvertently, their cycle times to reach the off-line QC endpoint also varied. Since the two ionic inorganic catalysts were added as 50 wt% solutions in water, the identity, charge, and concentration of the ionic species, and the water content (0.20 and 0.67 wt%) in the reactions, differed. As a result, the resistivity profiles for these reactions are similar but offset, and as expected, the resistivity for the pair of reactions run at both higher ion and water concentrations (-132 and -133) were offset towards lower values.

Figure 2 shows the reaction temperature profiles and Figure 3 shows the resistivity data for the four reactions. In Figure 2, the acrylate was added over about an hour to control the exotherm, and the reaction was continued for two and one-quarter hours. Temperature cycling during the isothermal hold between 70[degrees]C-75[degrees]C is clearly visible. In the small reaction mass of these examples, the period from peak to peak was about eight and one-half minutes. The sudden drop in temperature near two and one-half hours was due to the addition of polyacrylate, followed by reheating and a mild reaction exotherm. With the increase in reaction mass, the thermal cycling decreased to about a 20 min period.

The corresponding resistivity profile at 1 kHz is reproduced in Figure 3. During the one-hour addition of the acrylate, resistivity increased along with temperature. This is both an indication of reaction (increasing the rheological viscosity of the reaction mixture) and dilution of the conducting species. During the isothermal hold, the temperature cycling is responsible for slight variations in resistivity values. The nearly constant (average) value of resistivity during this time is indicative of very little additional chemical reaction taking place, consistent with the results of offline analysis [attenuated total reflection-Fourier transform infrared, gel permeation chromatography (ATR-FTIR, GPC)] of in-process samples. It is important to note that even though the cycle times for these four reactions differed, the final resistivity values were both comparable and characteristic of the two different sets of reaction mixtures when they were "finished" by offline viscosity measurement criteria.

Changes in resistivity at reaction temperature during consumption of polyacrylate were very well correlated with a low shear viscosity measured at room temperature, signal strength from ATR-FTIR for residual acrylate C=C-H wag at 812 [cm.sup.-1], and residual polyacrylate as determined by GPC.

Figure 4 shows the close relationship between resistivity at reaction temperature and low shear viscosity at room temperature for these reactions. The data also clearly illustrate the differences due to small "purposeful" variations in reaction composition. Importantly, however, the data for each pair conducted with the same catalyst are still coincident, so "normal" compositional variation based on weighing or transfer errors during manufacture can be accommodated by this technique.

[GRAPHIC OMITTED]

Figure 5 shows an excellent correlation between resistivity and the ATR-FTIR peak absorbance values for acrylate C=C-H wag (812 [cm.sup.-1]). The intensity of this signal is a measure of residual acrylate groups present in the reaction mixture (both bound to the oligomer and from unreacted polyacrylate). Again, the data fall onto two curves, indicative of the sensitivity of the technique to reaction composition. Figure 6 shows more directly that there is, as expected, a strong correlation between oligomer resistivity and conversion of polyacrylate reactant. The scatter in the data obtained for the reactions -128 and -130 also shows that there were apparently some difficulties in obtaining good (consistent) quantification of polyacrylate by GPC in the first pair of reactions (lower water content).

The related correlation between rheological viscosity and polyacrylate by GPC is shown in Figure 7. It illustrates the expected relationship between product viscosity and the consumption of polyacrylate measured offline from in-process samples. Based on the plot in Figure 7, since the data for reaction -130 falls consistently to the left of the other data points with similar viscosity (lower C=C conversion), it appears this data set is the source of the reduced correlation with the GPC data in Figure 6.

This assertion is reinforced by Figure 8, where again the data for reaction -130 consistently fall below the rest of the data. Finally, Figure 9 shows an excellent correlation between the low shear viscosity and ATR-FTIR acrylate absorbance taken from offline analysis of in-process samples.

Epoxy Ester-Acrylic Acid Monomer

Synthesis of the epoxy ester acrylate (EPAC) monomer is outlined in Scheme 2.

Results from duplicate reactions to make this monomer were compared. The reaction temperature profile (Figure 10) shows their differences in heat-up schedule on the way to the final hold until reaction "completion" as determined by standard offline measurements (acid number, AN, and epoxy equivalent weight, eew). Temperature cycling during the isothermal hold between 115[degrees]C-117[degrees]C is clearly visible, with a peak-to-peak period of about 10 min. The duplicate reactions exhibit different 1 kHz resistivity profiles in the early stages of the reaction where there are differences in temperature and reaction rate. They become very similar towards the end of the runs (Figure 11) because the reaction is going to completion, and therefore, the DES response is asymptotically approaching a final value representative of the product mixture. Resistivity falls sharply during the early phase of the reaction as the ionic catalyst is dissolving in the reaction mixture (increased [N.sub.i]).

[GRAPHIC OMITTED]

The best correlation was found to be between resistivity and AN, but very good correlations were also found between resistivity and eew or ATR-FTIR signal strength for several ester (C=O)-O-C[H.sub.2] bending modes (1065, 1191, and 1408[cm.sup.-1]).

Both AN and eew change over a wide range during the course of the reaction, so correlations made with these traditional metrics were made. Figure 12 shows the excellent relationship observed between resistivity and AN, while Figure 13 shows a somewhat reduced, but still very good, correlation between eew and resistivity. This difference in "goodness" of correlation is probably due to the more ambiguous titration endpoint for eew, resulting in greater error in calculated eew values.

Since epoxy is consumed in this reaction, monitoring disappearance of the epoxy peak near 910 [cm.sup.-1] by ATR-FTIR was considered. However, this is a weak peak, so while changes observed in the early portion of the reaction can be useful, it is inappropriate as a reaction endpoint indicator, because the precision is low at low epoxy concentrations. There were also significant changes in the strength of three -CO-O-C[H.sub.2]R ester bending mode peaks (1408, 1191, 1065 [cm.sup.-1]), so correlations between resistivity and these signals were evaluated. In Figure 14, the correlation coefficient clearly varies by peak, being best for the 1191 [cm.sup.-1] peak (the strongest absorbance, ~0.51) and being worst for the 1408 [cm.sup.-1] peak (the weakest absorbance, ~0.25). However, even in the worst case, correlation was still very good ([R.sup.2] = 0.83). Correlations between data collected offline from the in-process samples were also made. In this case, as expected, the correlation between AN and eew was excellent (Figure 15).

NCO-Terminated Polyurethane Prepolymer

This prepolymer was made according to Scheme 3.

Duplicate reactions were conducted that differed by the time to reaction "completion," as determined by standard offline mea-surements (% NCO and viscosity). These two reactions were run under nominally identical conditions, but since so little catalyst is required for this reaction (only "one drop"), and drop-size consistency is difficult at this scale (the catalyst solution is viscous), cycle times differed by about 30% (Figure 16). However, this is where the strength of this technique becomes obvious. By watching the resistivity output at 100 Hz (Figure 17), the reaction can be stopped just at the critical time when the NCO level drops within the specification.

An excellent correlation between weight percent (wt%) residual isocyanate and resistivity (Figure 18), was found. The correlation between resistivity and bubble second viscosity at 25[degrees]C (Figure 19) is also excellent. Finally, as expected by the high degree of correlation between DES and offline, in-process sample quality control data, the relationship between the traditional metrics collected is also excellent (Figure 20).

CONCLUSIONS

This work has shown that for a variety of coatings resin types made with very different chemistries and possessing different final molecular weights, the traditional in-process metrics used to follow the progress of the reactions and assign the reaction endpoints showed very good to excellent correlation with resistivity values determined in-situ and in real time by DES. This work has also shown that small "purposeful" differences in reaction composition can lead to distinctly different DES response curves when correlated with the offline data, offsetting the response and changing the value of the DES endpoint. However, small differences in composition caused by weighing and transfer variability are not expected to cause such shifts. Furthermore, this work has shown DES output to be sensitive to variations in reaction temperature, so comparison of DES data should be limited to that collected within narrow, but realistic, temperature ranges. DES appears to have good potential for use as a real-time, in-process monitoring system for making coatings type resins. The only caveat to using this technique effectively is that during resin process development, correlations between DES sensor output and appropriate traditional offline in-process sample responses must be generated.

EXPERIMENTAL DETAILS: RESIN SYNTHESES AND TESTING

General Procedures

Resin syntheses were all conducted in four-neck one-liter round bottom flasks with mechanical stirring, ther-mostatted electrical heating, and dry air or nitrogen sparge as necessary. The DES probe was one-half inch in diameter and was inserted through an airtight Teflon bushing. Electrode configuration was circumferential; the inner electrode was circular (~4 mm diameter) with about a 2 mm gap between it and the outer electrode. Netszch Instruments loaned a commercial DEA 230/2 analyzer, probe, and Eumetric[TM] software to Sun Chemical for these evaluations. The software was loaded onto a standard laptop computer and the data files were converted to Microsoft[TM] Excel file format for analysis.

Acrylated Acetoacetate Michael Oligomer

Samples were taken at decreasing time intervals for analysis as the reaction progressed. FTIR spectra were recorded using a Thermo-Electron AVATAR 370 Smart ARK with a 45[degrees] HATR Germanium plate. Peak height of the acrylate C=C-H wag peak at 812 [cm.sup.-1] was determined in absorbance on spectra normalized to full scale for the 986 [cm.sup.-1] ester CO-O-R bending peak for the sample taken immediately after polyacrylate addition. Viscosity was determined by a TA Instruments AR1000 at 25[degrees]C under 10 [sec.sup.-1] shear rate using a 4 cm/2[degrees] cone. GPC analyses were performed using a Waters system in THF on a Phenomenex Phenogel 10,000[Angstrom], 1,000[Angstrom], 100[Angstrom], 50[Angstrom] column set. To calculate residual monomer, GPC data files were converted to Microsoft Excel file format and the area for the major polyacrylate peak determined at constant chromatogram area = 1000, normalized to the initial weight percent of polyacrylate in the first sample.

Epoxy Ester-Acrylic Acid Monomer

Samples were taken at increasing time intervals for analysis as the reaction progressed. FTIR spectra were recorded using a Thermo-Electron AVATAR 370 Smart ARK with a 45[degrees] HATR Germanium plate. Peak height analysis of the epoxy band near 910 [cm.sup.-1] was inadequate since this is a weak band and sensitivity is poor for low levels of residual epoxy. Consequently, ester CO-O-C[H.sub.2]R bending mode peak heights at 1065 [cm.sup.-1], 1191 [cm.sup.-1], and 1408 [cm.sup.-1] were determined in absorbance on spectra normalized to full scale for the 1728 [cm.sup.-1] C=O peak for the sample taken after acrylic acid addition was complete. The best correlation was generated using the strongest peak (1191 [cm.sup.-1]). Acid number was determined by manual titration using 0.1000 N KOH solution in methanol and 1% phenolphthalein in ethanol as indicator. Epoxy equivalent weights were similarly determined by manual titration using 40% tetraethylammonium bromide in glacial acetic acid as a reagent, 0.1000 N HCl[O.sub.4] in glacial acetic acid as titrant, and 1% crystal violet in glacial acetic acid as indicator.

NCO-Terminated Polyurethane Prepolymer

Samples were taken at 30 min intervals for analysis as the reaction progressed. Weight percent NCO was determined by manual titration using 2.0 N di-n-butyl amine in toluene as a reagent, aqueous 1.0 N HC1 solution as titrant, and dilute bromcresol green in isopropanol as indicator. Bubble second viscosity was determined at 100[degrees]F according to ASTM D1545.

ACKNOWLEDGMENTS

I wish to thank Mr. Makoto Namura, a visiting scientist from our parent company, Dainippon Ink & Chemicals (Japan), for his experimental skill while conducting the polyurethane prepolymer syntheses and analyses.

I would also like to thank Mr. David Shepard and Netszch Instruments for their generous loan of the DEA 230/2 analyzer, probe, and Eumetric software, and training and advice.

Finally, I wish to thank Professor David Kranbuehl for his constructive and helpful critique of the original manuscript.

References

(1) (a) Kranbuehl, D.E., in Dielectric Spectroscopy of Polymeric Materials: Fundamentals and Applications, Runt, J.P. and Fitzgerald, J.J. (Eds.), American Chemical Society, Washington, D.C., Chapt. 11, pp. 303-328, 1997. (b) Kranbuehl, D.E., in Processing of Composites, Dave, R.S. and Loos, A.C. (Eds.), Hanser Gardner Publications, Chapt. 4, pp. 137-157, 2000.

(2) (a) Neag, C.M., Rohn, A., and Bode, D., Proc. North American Thermal Analysis Society, Toronto, 1994. (b) Shepard, D., Crowley, T., and Choi, K.Y., J. Appl. Poly. Sci., 55, 1361 (1995). (c) Kranbuehl, D. and Domanski, A., "In Situ Monitoring of Coating Polymerization, Cure, and Aging Using Frequency Dependent Dielectric Monitoring," JCT COATINGSTECH, 1, No. 6, 48 (2004).

(3) The author was reminded of this in a private communication with Kranhuehl, D.E., February 2006.

(4) Chalmers, J.M., "Infrared Spectroscopy in the Analysis, Characterization, and Testing of Coatings," JCT COATINGSTECH, 2, No. 18, 50 (2005).

(5) Lachowicz, A., Gaudl, K.U., Nahm, S.H., and Grahe, G.F., European Patent 1,431,320 Al, June 23, 2004.

APPENDIX

Michael Oligomer

[FIGURE 2 OMITTED]

[FIGURE 3 OMITTED]

[FIGURE 4 OMITTED]

[FIGURE 5 OMITTED]

[FIGURE 6 OMITTED]

[FIGURE 7 OMITTED]

[FIGURE 8 OMITTED]

[FIGURE 9 OMITTED]

Epoxy-Acrylic Acid Monomer

[FIGURE 10 OMITTED]

[FIGURE 11 OMITTED]

[FIGURE 12 OMITTED]

[FIGURE 13 OMITTED]

[FIGURE 14 OMITTED]

[FIGURE 15 OMITTED]

Polyurethane Prepolymer

[FIGURE 16 OMITTED]

[FIGURE 17 OMITTED]

[FIGURE 18 OMITTED]

[FIGURE 19 OMITTED]

[FIGURE 20 OMITTED]

Steven H. Nahm -- Sun Chemical Ink*

* Senior Scientist, Daniel J. Carlick Technical Center, 631 Central Ave., Carlstadt, NJ 07072. Email: [e-mail].