Table of Contents ECNDT '98 Session: Materials characterization

Monitoring of Curing Reaction of Polycondensating Thermosets at Press and Injection Moulding W. Stark1, J. Döring Federal Institute for Materials Research and Testing, 12205 Berlin, Unter den Eichen 87, Germany V. Bovtun Kiev Polytechnic Institute, KPI-2240, Peremogy Ave. 37, 252056 Kiev-56, Ukraine Ch. Kürten Engineering College Iserlohn, Frauenstuhlweg 31, 58644 Iserlohn, Germany Corresponding Author Contact: Wolfgang Stark Email: Wolfgang.Stark@bam.de

Introduction

Plastics may be roughly divided into two classes: thermoplastics moulded in the molten state, thermosets being chemically reactive and cure in the moulding process by crosslinking. Typical thermosets are resins of phenolic - PF, urea - UF, melamine - MF, melamine-phenolic - MP, epoxy - EP, unsaturated polyester - UP (with glass filler - bulk or sheet moulding compounds BMC, SMC). Thermosets are widely used for their low price and sometimes outstanding mechanical and electrical properties, especially at high temperatures. Unfortunately they run the risk of being under- or overcured which plays a substantial role in product quality. In the final product it is difficult or often impossible to determine the extent of curing.
That is the reason behind years of research work to develop methods to monitor the curing process on-line. In principle all methods showing the propagation of chemical reaction are suitable. They can be divided into methods which are directly sensitive to the chemical reaction and those which detect changes in microscopic material parameters. The chemical reaction is sometimes exothermic and can be detected by thermoanalysis (e.g. differential scanning calorimetry DSC). Infrared spectroscopy, nuclear magnetic resonance spectroscopy and chromatography are also typical methods in research and development. However, for practical reasons these methods are difficult to incorporate into the production process. The macroscopic attributes which change during cure are mechanical and electrical material parameters. Typical mechanical parameters are complex elastic or shear modulus and viscosity. These quantities can be measured by dynamic mechanical analysis DMA, ultrasound propagation, viscosimetry. Typical electrical parameters are dc and ac conductivity, dielectric permittivity and loss. All methods have been successfully used in basic investigations but only few methods seem to be suitable for on-line cure monitoring in the mould at high temperatures and pressures. Most basic developments were done in epoxy resin cure monitoring [1]. Due to its simplicity the dielectric method often is preferred. Also acoustic methods using ultrasound were tested. Based on experience with epoxy monitoring (published results) we started to adapt these methods to the specific demands of thermoset press and injection moulding.

Physical background of electrical and mechanical methods

The physical background is to monitor the change of mobility of movable parts of the molecular chain. Molecular group mobility is influenced by molecular structure. Drastic changes appear at critical temperatures, e.g. the glass-rubber transition and also when crosslinking occurs. In this way the mobility of molecular groups can be used as a microscopic probe for crosslinking.
If the molecular segments have polarisable groups they interact with an electric field. The vibration of these groups is characterised by characteristic time constants. Fortunately at electrical activation the frequency can be changed in an extremely wide range (up to 15 decades), that means dielectric spectroscopy can be used to find these characteristic time constants. The material parameters dielectric permittivity and loss can be calculated from dielectric measurements. Using a characteristic dipole relaxation time _D we can define permittivity as a function of frequency *(^w):


When there is an additional ohmic conduction " must be modified:

'- dielectric permittivity, ''- dielectric loss, _h - ' at high frequency limit, _l - ' at low frequency limit, - ohmic conductivity.
Therefore softening and crosslinking should influence the characteristic dielectric values via a change of _D and . The first term in _c'' indicates that the contribution of ohmic conductivity at low frequencies is high.
Also mechanical behaviour is determined by the mobility of molecular segments which are characterised by a relaxation time. The complex modulus (e.g. shear modulus G*(^w)) is therefore also dependent on the measuring frequency:
G_h - modulus at high frequency limit, G_l - modulus at low frequency limit. The relaxation time is influenced by temperature and crosslinking. Therefore the mechanical parameters such as modulus and therefore sound propagation give also valuable information on the moulding process.

Sample preparation

For the investigations all thermoset materials listed in the introduction were available as gratuitous test granulates from the industry. To prepare samples for mechanical and electrical preliminary investigations, the granulate was plastified (without precuring) in the range of 70 to 90 °C in an extruder and moulded as 100 mm long cylinder of 40 mm diameter. For the torsion pendulum rectangular plates of 60x10x1 mm³ were cut and polished. For dielectric measurements samples of 40 mm diameter, 0.5 mm thickness were prepared with sputtered electrodes.

Measuring equipment

Dielectric measurements were carried out in two frequency ranges: 100 Hz to 1 MHz and 5 to 20 GHz. In the first range commercial equipment with a temperature controlled sample holder (-150...300 °C) and a dielectric analyser were used (see Fig. 1a). For practical tests in the mould a dielectric sensor [2] and a plug in dielectric card for a computer were developed [3] (shown in Fig. 1b). The microwave equipment for basic and for realistic press moulding investigations is shown in Fig. 2 and Fig. 3.

Fig. 1: Dielectric equipment for frequencies up to 1 MHz: a) for basic investigations, b for on-line monitoring: 1 - sample holder with temperature regulation, 2 - sample with electrodes, 3 - analyser, 4 - PC, 5 - temperature controller, 6 - press mould form, 7 - movable stamp, 8 - dielectric sensor, 9 - thermocouple, 10 - PC plug in card

Fig. 2: Block-diagram of the rectangular waveguide impedance method: 1 - plate-shape sample; 2 - shorted waveguide; 3 - heating or cooling system; 4 - microwave generator; 5 - probe measuring line; 6 - indicator	Fig. 3: Block- diagram of the open-ended coaxial line resonator method: 1 - sample; granulated material, powder, bulk, etc.; 2 - pressform; 3 - coaxial sensor; 4 - heating system; 5 - microwave sweeping generator; 6 - microwave reflectometer; 7 - display.

Preliminary investigations of the curing process

For preliminary investigations of uncured material DSC, torsion pendulum and dielectric
spectroscopy were used. Starting at low temperatures the material was heated linearly. The heating was stopped after the crosslinking reaction. In a second measurement the same now cured sample was measured once more. Typical results for all three methods are demonstrated for novolac, the resin base of phenolic thermosets and original type PF51. In the DSC results PF51 (Fig. 4, curve 1) and novolac (curve 3) both show a distinct change of heat capacity at about 60 °C. This behaviour including the small peak is typical for the glass-rubber transition i.e. softening (segmental movement in the main chain). Starting at about 100 °C the exothermic crosslinking reaction leads to a typical reaction peak. At 175 °C a second reaction appears. This behaviour is known for phenolic resins [4]. For the cured material (curve 2) softening and crosslinking are absent.

Fig. 4: DSC measurements of 1-original PF51, 2-cured PF51, 3-novolac

Fig. 5: Torsion pendulum measurement of original and cured PF51

Pendulum measurements show a decrease in modulus which is related to softening begin at about 30 °C. At 120 °C an increase in modulus similar to that at the beginning sets in, which is caused by crosslinking. The loss modulus shows two maxima, typical for high energy loss by passing two times the resonance between measuring frequency and relaxation time.
The same behaviour was observed again in dielectric parameters. Softening of novolac and PF51 causes an increase, crosslinking causes a decrease. The very high ´ and ´´ values at low frequencies hint on high ohmic conduction on softening. High ´ is caused by interface polarisation (Maxwell-Wagner-effect). ´´ is directly influenced by ohmic conductivity.

Fig. 6: Dieletric spectroscopy of novolac and PF51 as function of temperature

Ohmic conduction depends on the number of charge carriers and their mobility. Assuming that at softening the number of charge carriers does not increase it is plausible that the existing carriers now become mobile.
Vice versa their mobility decreases by crosslinking. At temperatures over 180 °C there may be an increase in the number of charge carriers, as the modulus does not indicate distinct additional softening.
In the microwave region the fundamental properties of the "pure" substance, mobility of the molecules play the main role and are not masked by the conductivity. To prove the efficiency of the microwave methods we investigated 4 sorts of thermosets, belonging to 2 different types: PF, UF. Increasing of '(T) and "(T) at T > 50 °C corresponds to softening, decreasing at T > 110 °C - to curing. If the microwave and low frequency values of '(T) and "(T) are compared, we can see a big difference in the absolute values: at low frequencies ' 10², " 10² (tan 1); at microwaves ' 3 to 8, " 0.1 to 0.8 (tan 10^-2).
Low values of the microwave dielectric permittivity and losses prove that microwave dielectric parameters are defined by the fundamental polarization mechanism, caused by the dynamics and kinetics of the polymer structure parts [5,6]. Two maximaof ´(T) and ´´(T), corresponding to the curing process, are observed in UF and PF. The second maximum of microwave dielectric parameters is caused by a second kind of the crosslinking reaction as seen in Fig. 4 by DSC.

Investigation of the curing process at manufacturing conditions

To test the methods experiments using compression and injection moulding machines were carried out. The thermoset granulate was shuted into the preheated press mould. Then the mould was quickly closed, the resin pressed and time dependence of ´ and ´´ measured .The results for ´ using PF and UF at three frequencies are given in Fig. 8.

Fig. 8: UF and PF ' variation during press moulding at 160 °C for frequencies, 1 - 1 kHz, 2 - 10 kHz, 3 - 100 kHz

There is an important difference between both kinds of thermosets. UF shows the expected picture: after closing the mould and pressing softening is seen by increase of ´. After a short time the increase of temperature leads to crosslinking, and a decrease of ´ follows, as known from the basic experiments. This is the opposite for PF. Here, softening is also observed, but where a decrease caused by crosslinking is expected, an increase occurs which remains constant over a longer time period. UF and PF crosslinking reaction is a polycondensation reaction where by-products, such as ammonia (PF) or water (UF) are produced. In the mould these by-products have no chance to disappear. In the mould these by-products have no chance to disappear. Obviously in PF the conductivity is considerably increased by this effect so that hardening is masked. Therefore the dielectric method in the low frequency range does not work with all thermosets. Many tests gave the results, that crosslinking monitoring by the medium frequency dielectric method works well with EP, UF (known from literature) and also with UF (unknown before).

Fig. 7: Temperature dependence of the microwave dielectric parameters of two UF sorts - a, b and two PF sorts - c, d at 10 GHz during curing process (measurements in a rectangular waveguide, samples were prepared from the preformed pressed but not cured polymers).

Therefore monitoring of UF manufacturing in injection moulding process could be tested. UF is widely used for electrical installation articles. The influence of different production parameters was successfully simulated. Because of the limited space only one characteristic result - the influence of mould temperature - is given in Fig. 9. The sensor is situated at a distance of about 100 mm from the nozzle. That means when the resin comes in contact with the sensor the softening process is over and the material has already begun to cure. The signal reduces towards the end of reaction. There are valuable results for producers because the cycle time may be optimized and reduced.

Fig. 9: UF131.5 " variation during injection moulding, influence of mould temperature

One option is to increase the mould temperature, so that the heat transfer rate to the resin is increased in order to reach the starting temperature earlier and also to increase reaction speed. But without insight into the process this can not be done effectively because of the danger of under- or overcuring. Now we have the possibility to watch directly when curing is finished. At 170 °C an important reduction of cycle time seems to be possible.
To find also possibilities for cure monitoring of the thermosets which are unsuitable for low frequency dielectric method we selected methods which should not be influenced by conductive reaction products. These methods should use ultrasound or microwaves. With ultrasound and microwaves we were successful. The ultrasonic results are presented in the same proceedings [7].
For microwave measurements under realistic conditions a sensor was developed for use in the mould. The measuring arrangement for the open-ended coaxial resonator is given in Fig. 3. Results for PF in rectangular waveguide and at press moulding using the open-ended coaxial resonator are shown in Fig. 10.

Fig. 10: Temperature dependence of the microwave dielectric parameters of PF at 10 GHz during curing process. Curves 1 and 2 correspond to: 1 - measurements in a rectangular waveguide under no pressure, 2 - measurements by the coaxial sensor under conditions, close to the industrial press and injection moulding process, with granualated polymers inserted into the pressform

Both results correlate with each other. Products of the curing reactions do not influence the microwave dielectric parameters.

Summary

Process monitoring of thermoset curing is an important tool for quality control and increase of productivity. Dielectric methods in two frequency ranges were tested. Without pressure all classes of thermosets could be investigated successfully at low frequencies. Some thermosets under pressure result in high conductivity, caused by reaction by-products which mask the curing process. In this case microwave and ultrasound methods can be used. Both work well under production conditions.

References

1. S.D. Senturia, N. F. Sheppard, Advanced Polymer Science 80 (86) 1
2. Patent Nr. DE 297 11 861.7
3. Patent Nr. DE 196 01 947 C1
4. G. L. Brode, Phenolic Resins, in Kirk-Othmer (Ed.) "Encyclopedia of Chemical Technology", J. Wiley New York 1982, p. 384
5. W. Stark - Proceedings 8th Intern. Duroplasttagung, Würzburg 23-24.04.96, (1996)
6. V. Bovtoun, W. Stark, V. Borisov, R. Kunze, A. Grishin, Yu. Yakimenko - Proc. Int. Conf. on Dielectrics and Insulation, Budapest (1997) 59-62
7. J. Döring, W. Stark, G. Splitt, Proceedings 7th ECNDT Conference, Kopenhagen 1998

Acknowledgements

An important part of such a wide field of research work can be only successfully done by the assistance of a lot of co-workers. The authors wish to thank the following colleagues for close co-operation and friendly support:
BAM: Dr. W. Mielke, Dr. J. Kelm, Dr. Ch. Holland, Dr. G. Kalinka, Dr. R. Kunze, P. Fengler, B. Hantschke, V. Lehmann, J. McHugh, D. Neubert, C. Vogt, W. Roltz, M. Schneider, Engineering College Iserlohn: Prof. P. Thienel, B. Hoster
Kiew Polytechnic Institute: Prof. Yu. Yakimenko, Dr. V. Pashkov, V. Borisov, L. Kuznezova, E. Ljubak, Technical Association of Producers and Users of Thermosets (TV): K. Auracher, Bakelite Letmathe: M. Stahl.