Analysing Phenol-Formaldehyde Resin Reaction For Safe Process Scale Up

SYMPOSIUM SERIES NO 16 HAZARDS 25 215 IChemE Analysing Phenol-Formaldehyde Resin Reaction For Safe Process Scale Up David Dale, Process Safety Manager, SciMed/Fauske and Associates, Unit B4, The Embankment Business Park, Vale Road, Stockport, Cheshire. SK4 3GN. The production of phenolic resins is an old but still very active industry with uses today including heat shields, abrasives, coatings, composites, felt-bonding, laminates, moulding and many others. Over the years much safety work has been done characterising the process safety aspects of these challenging but industrially important phenol-formaldehyde reactions. Two types of phenol-formaldehyde resins, namely Novolac and Resole, can be formed depending on the formaldehyde to phenol ratio. In this paper we will briefly discuss various calorimetry techniques (differential scanning, reaction and adiabatic), what data each can provide, and then show how they can be used in conjunction with each other to fully characterise the energies involved in a generic Resole recipe. The obtaining of such thermal data is an essential part of the process if we are to make informed decisions about safe scale up and due to the magnitude of these heat releases we will show how important it is to control the intended heat of reaction and also the latent crosslinking energy that is liberated should the reaction runaway. Having obtained this data we will then demonstrate by the use of simple calculations what these energies mean for safe scale up strategies and discuss and present options for how this may be achieved based on typical plant set ups utilising options such as traditional jacket or reflux (condenser) control, addition control, temperature and pressure staging. A real life example will also be used to aid this discussion and hopefully this paper will give an appreciation for the challenges faced by phenol formaldehyde resin manufacturers and how various calorimetry techniques are essential in providing an insight into process performance, understanding how to prevent runway reactions and providing the data to necessary to design protective relief systems. Introduction Producing phenolic resin is a very old but still active industry. Von Bayer first reported the polycondensation of phenol with aldehydes in 1872. In 192 Blumer had the first industrial process producing shellac from a phenol formaldehyde reaction. Baekeland made the first thermosetting plastic in 199. Today phenolic resins see a wide variety of uses such as ablation (heat shields), abrasives, coatings (can lining), composites, felt-bonding, foams, foundry (casting), friction, laminating (PCB), molding, proppants (fracking), refractory, rubber, substrate saturation (paper) and wood bonding (plywood, particle board). There are two types of phenol formaldehyde resins. When the formaldehyde to phenol ratio is less than 1 a Novolac resin is formed. These resin reactions are acid catalysed, react to completion and the resulting product requires a separate crosslinking agent (typically hexamine) to produce the final resin. A Resole resin is formed when the formaldehyde to phenol ratio is greater than 1 (typically 1 to 3). These reactions are based catalysed and intentionally not reacted to completion. The Resole product carries pendent methylene hydroxy moieties (not stable in acid) that allow the resin to be self-crosslinking at higher temperatures. It is these reactive groups that make the Resole version of this chemistry particularly hazardous due to this latent but elevated temperature activated reactivity. Figure 1: Schematic of Resin Types In this paper we demonstrate the use of Reaction Calorimetry, Differential Scanning Calorimetry (DSC), and Adiabatic Calorimetry to characterise the energies involved in a generic Resole recipe. Our example generic recipe has a formaldehyde to phenol ratio of 2.2 and is 31% wt phenol, 21% water, 4% catalyst (5% aq. sodium hydroxide) with the aldehyde source being 5% aq. formaldehyde (44% wt). Figure 2 shows the normalized (W/kg phenol) heat flux profile for a batch reaction at 5 C of this recipe initiated by a 1 minute addition of the catalyst. Figure 3 shows an expanded plot where the addition of the catalyst is clearly visible. Phenol is a weak acid while 5% sodium hydroxide is a strong base so the formation of sodium phenolate (the active species in the polycondensation) is a rapid and complete reaction. The sloped appearance of the heat flow profile during the catalyst 1

(g), (W/kg phenol) ( C), (%) (g), (W/kg phenol) ( C), (%) SYMPOSIUM SERIES NO 16 HAZARDS 25 215 IChemE addition is due to the growing rate of the polycondensation reaction underlying the addition limited kinetics of the phenolcaustic reaction (square-wave response). The area under the heat flow profile curve represents the total heat of our batch reaction and integrating yields a normalized heat of reaction of -92 kj/kg phenol. Dividing the total heat by the thermal mass (mass x heat capacity) calculates the theoretical temperature rise under adiabatic conditions due to the intended heat of reaction. For this recipe the calculated adiabatic temperature rise was +91 C. Phenol Formaldehyde Batch at 5 C 5 1 45 4 9 8 35 7 3 6 25 5 2 4 15 3 1 5 2 1-1 1 2 3 4 5 6 7 8 9 Figure 2: Normalised Profile for Batch Reaction at 5 C. 5 45 Phenol Formaldehyde Batch at 5 C (expanded) 1 9 4 8 35 7 3 6 25 5 2 15 4 3 1 5 2 1.1.2.3.4.5 Figure 3: Normalised (Expanded) Profile for Batch Reaction at 5 C. So here just from the intended heat of reaction we see a large adiabatic potential, as from 5 C a loss of cooling scenario could result in a temperature rise to 141 C where the vapour pressure of water would be 54 psig. If we take the reaction product from the batch reaction at 5 C and look at it in the DSC we see the scan shown in Figure 4. 2

(W/g) (W/g) SYMPOSIUM SERIES NO 16 HAZARDS 25 215 IChemE.5 Size: 9.1 mg Method: Ramp @ 2 C/min.25 122.19 C. -.25 55.48 C 91.79 C 32.6 J/g -.5 5 1 15 2 25 3 35 4 Temperature ( C) Figure 4: DSC Of Reaction Product Mixture In the DSC scan we see the latent crosslinking energy from the pendent methylene hydroxy groups in the Resole reaction mass. Taking the integrated energy from the scan, 32.6 J/g, and dividing by the heat capacity of the reaction mass tested (3.1 J/g C) we calculate an additional +98 C adiabatic temperature rise potential. Adding this higher temperature activated energy to our previous intended reaction energy totals a temperature rise +189 C! To confirm that the intended heat of reaction under loss of cooling conditions can raise the temperature of the reaction to a temperature where the latent crosslinking energy can be realized, adiabatic calorimetry is needed. Figure 5 shows the same recipe run in the Fauske VSP2 adiabatic calorimeter. Thermosetting Catalyst add Intended Reaction Figure 5: Adiabatic VSP test Trace on Generic Reaction Mixture In the VSP2 the reaction is initiated by injection of the catalyst at 5 C and proceeds under adiabatic conditions. Indeed the VSP2 trace shows the intended reaction heat does carry the system into the temperature range where the crosslinking reaction occurs and further temperature rise is realized. Though not covered in this paper, the temperature rise rate and pressure rise rate data from such an experiment can be used to design a proper vent size for plant reactor for this process. The information gathered and illustrated by these three instruments definitively shows how important it is to control the intended heat of reaction in this Resole phenol formaldehyde resin process. Not only is there plenty of intended reaction energy to deal with but there is also latent crosslinking energy waiting to be liberated should the reaction runaway. While these reaction are easy to control in a laboratory reaction calorimeter via jacket cooling, to answer the question how easy is it to control with scale we can now look at some simple calculations to show how these thermal challenges offered by the phenol formaldehyde process can be handled with scale. Normally in batch equipment one would rely on jacket cooling to remove the reaction heat and control temperature. If we scale our phenol formaldehyde batch process to 2 kg of phenol and run it in a 12, litre (diameter 2.2 m) the stirred volume for our generic recipe would be 35 litres corresponding to a heat transfer area of 9 m 2 (A). If the tank was glass 3

(g), (W/kg phenol) ( C), (%) SYMPOSIUM SERIES NO 16 HAZARDS 25 215 IChemE lined steel, a typical heat transfer coefficient (U) would be ~3 W/m 2 K. Assuming a maximum temperature difference between the reactor and jacket (ΔT) of 5 C, our maximum cooling capacity would be UAΔT = 3 x 9 x 5 = 135, W. For the 2 kg phenol batch size the normalized cooling capacity is 67.5 W/kg phenol. However, practitioners of phenol formaldehyde processes universally cool their processes through the use of oversized condensers and reflux of water most of the time under various levels of vacuum. Data from an industrial plant showed the cooling capacity of their condenser to be 38.4 K BTU/min. This cooling capacity would be the equivalent of condensing 176 kg water/hr or converting to our current unit of choice, 675,236 W. Normalizing to our 2 kg batch size yields 338 W/kg phenol. Figure 6 shows the expanded plot of our 5 C batch reaction though this time with the normalized cooling capacities for the glass line steel jacketed reactor and that provided by the previously mentioned reflux condenser presented as horizontal dotted lines. Also shown are similar lines for a Hastelloy reactor (U = 5 W/m 2 K), and a stainless steel reactor (U = 1 W/m 2 K). Please note that all the values for the heat transfer coefficient are at the upper end of the performance range for each material of construction. Certainly jacket fouling and choice of heat transfer fluid can affect these assumed U values, as well. 5 Phenol Formaldehyde Batch at 5 C (expanded) 1 45 9 4 35 Q-Condenser 8 7 3 25 2 Q-SS 6 5 4 15 1 5 Q-Hast Q-GLS 1.1.2.3.4.5 Q - GLS Q - Hast. Q - SS Q - Condenser 3 2 Figure 6: Batch at 5 C showing heat flow versus time overlaid with various cooling capacity scenarios One can easily see that jacket cooling in either the glass lined steel reactor or Hastelloy reactor is inadequate as a significant portion of the heat flux profile exceeds the cooling capacity line. One can see that jacket cooling in a stainless steel vessel is close to having enough cooling capacity and possibly with a slightly longer catalyst addition it may well have been enough. Even with cooling by reflux the cooling capacity line is just shy of completely surpassing the peak heat rate of this batch process. So what other options are available for scale up of these types of processes? If one is fortunate enough to have aqueous formaldehyde as the aldehyde source then running the process semi-batch with a controlled addition of formaldehyde will help. Figure 7 shows an expanded plot of a semi-batch version of our generic phenol formaldehyde recipe. Here the initial rate corresponded to a 4 hour addition of 37% aq. formaldehyde to the precharged phenol, water and catalyst. 4

True (W/kg), Reaction Mass (kg) Tr ( C), (%) (W/kg phenol), Formaldehyde Add (g) ( C), (%) SYMPOSIUM SERIES NO 16 HAZARDS 25 215 IChemE Phenol Formaldehyde Semi-Batch - Formaldehyde Addition 7 C (expanded) 35 315 Q-Condenser 1 9 28 245 8 7 21 Q-SS 6 175 14 15 Q-Hast 5 4 3 7 2 Q-GLS 35 1 Dosing Mass -.25.25.75 1.25 1.75 2.25 Formaldehyde Dosing Mass Q - GLS Q - Hast Q - SS Q - Condenser Figure 7 :Semi- Batch plot for generic recipe After 3 minutes the addition rate was doubled then the addition rate increased again to three times the original rate then slowed to 1.5 times the original rate through the end of the addition. One can see addition limited kinetics (square-wave pattern) for the initial addition rate and double that rate but accumulation begins when the original rate was tripled. More important is to notice that the controlled addition of formaldehyde, at least at the initial 4 hour rate, brings the stainless steel, Hastelloy and even the glass lined steel reactor jacket cooling into play. Reflux cooling provides plenty of cooling capacity margin for the semi-batch process. But not all phenol formaldehyde recipes use aqueous formaldehyde as the aldehyde source thus allowing for semi-batch operation, and there is no guarantee that the semi-batch process makes the same material as its batch counterpart. Also water is a necessary burden in phenol formaldehyde processing. While it is convenient to have during the main reaction, before end use water will be stripped off to concentrate the resin before final thermosetting. This costs cycle time and generates waste needing disposal. One answer to using less water is to use paraformaldehyde, a polymer of formaldehyde which is a solid thus avoiding the water that comes along with using aq. formaldehyde. Being a solid though, paraformaldehyde is batch loaded. Figure 8 shows a different recipe where the paraformaldehyde is present in a ~2:1 ratio versus aqueous formaldehyde, more phenol and overall less water than our generic recipe. This run was carried out batch at 7 C with a ten minute catalyst addition. Batch at 7 C 1 12 8 1 6 8 4 2 Q-Condenser 6 4-2.5 1 1.5 2 2.5 3 3.5 4 True Q - Condenser (2K kg) 2 2 Figure 8: Reaction using paraformaldehyde at 7 C 5

(W/kg Phenol) Reaction Temperature ( C), (%) SYMPOSIUM SERIES NO 16 HAZARDS 25 215 IChemE For comparative purposes this recipe has a normalized heat of reaction of -943 kj/kg phenol and an adiabatic temperature rise of +143 C. Also shown on the plot is the 338 W/kg phenol reflux cooling capacity line and we can clearly see that this recipe could not be safely run, in this manner, at the 2 kg scale. So how might you scale a specific recipe that does not lend itself to controlled addition of one of the components? Staging of temperature, catalyst and even vacuum can be done to help keep the rate in check. Figure 9 shows an actual plant process whereby the temperature and catalyst addition is staged. Plant Mimic Run, 37 C to 67 C, Catalyst Add in Thirds 6. 8. 5. 7. 4. 3. Q-Condenser (1x) ZXReactor Temperature 6. 5. 2. 4. 1. 3.. Q-Condenser (3.2 x) 2. -1. 1. -2.. -3. -1. -4. -2. 1.5 2. 2.5 3. 3.5 4. Experiment Time (hr.) Q-Condenser (1X) Q - Condenser (3.2X) Tr % Conv. Figure 9: Laboratory Run of Actual Plant Process To mimic the plant process a background temperature ramp was performed adding the catalyst in thirds at appropriate times corresponding to the points where they establish and confirm temperature control. The three vertical spikes are the catalyst addition points and were done so in the lab much faster than the actual catalyst would be added in the plant. Notice the initial endotherm at the start where there is significant undissolved paraformaldehyde going into solution with reaction which helps offset exothermic reaction heat at the lower temperatures. By the time the last third of the catalyst is added 35% of the reaction has been completed and throughout the entire process reflux cooling is more than adequate, at least for the current batch-size (1X). Also shown is the normalized cooling capacity line if this recipe were to be scaled up by a factor of 3.2 (3.2X). Here you can see that by the second catalyst addition the reaction rate exceeds the 3.2X condenser cooling capacity. Based on this data it was decided not to scale this process up to the larger batch size. Possible changes to enable safe plant scale up, include controlled addition of the aqueous formaldehyde portion of the recipe and/or switching out some or all of the paraformaldehyde for aqueous formaldehyde, provided equivalent material can be made with the new recipe. Summary In summary we hope this paper has given an appreciation for the challenges faced by phenol formaldehyde resin manufacturers. In particular, that reaction calorimetry provides a vital window into process performance and helps gather the data necessary to understand how to prevent a runaway reaction together with simple heat rate scale up calculations. Thermal screening and adiabatic calorimetry are key to understanding the total upset scenario and provides the data necessary to design a relief system to protect the vessel in case of runaway. 6