Dicumyl Peroxide Case Study – The Proper Application of Open Cell Test Methodology

Dicumyl peroxide is a commonly employed dialkyl peroxide used as a high temperature crosslinker in rubber and plastics applications [1]. A common use of dicumyl peroxide includes initiation of the polymerization of ethylene to produce low-density polyethylene (LDPE). Dicumyl peroxide is known to thermally decompose to acetophenone (high boiling liquid) and methyl radicals (gas), and the system has been extensively studied [2-5]. In terms of vent sizing, the thermal decomposition of dicumyl peroxide is classified as a gassy system otherwise known as a non-tempered gas generating reaction. The reaction mechanism is illustrated in Figure 1, with first-order Arrhenius kinetic parameters listed in Table 1.

Gassy System Fundamentals

Dicumyl peroxide is a commonly employed dialkyl peroxide used as a high temperature crosslinker in rubber and plastics applications [1]. A common use of dicumyl peroxide includes initiation of the polymerization of ethylene to produce low-density polyethylene (LDPE). Dicumyl peroxide is known to thermally decompose to acetophenone (high boiling liquid) and methyl radicals (gas), and the system has been extensively studied [2-5]. In terms of vent sizing, the thermal decomposition of dicumyl peroxide is classified as a gassy system otherwise known as a non-tempered gas generating reaction. The reaction mechanism is illustrated in Figure 1, with first-order Arrhenius kinetic parameters listed in Table 1. A gassy system is defined as one where noncondensable gas is generated in the venting region and there is an absence of tempering to control the reaction temperature. Because the reaction temperature cannot be controlled, the reaction is expected to run to completion and the highest gas generation rate must be accommodated in the relief vent area. The key parameters for relief sizing for gassy systems are the mass remaining in the vessel at the peak reaction rates and the maximum pressure rise rate.

Dr. Hans Fauske [6-8] originally suggested sizing relief devices for gassy systems considering homogeneous two-phase flow but pointed out that the assumption of no mass loss at peak reaction rates is overly conservative. The importance of early mass loss is backed by large scale testing on systems like hydrogen peroxide, dicumyl peroxide, t-butyl peroxy benzoate, and 3, 5, 5-trimethyl hexanoyl peroxide. Therefore, Fauske’s recommended strategy is to maximize the amount of early mass loss by requiring the use of rupture disks for gassy systems and setting the relief device as low as practically possible. This approach ensures consistent assumptions with the findings from large scale data: complete disengagement early in the reaction and at the peak reaction rates.

Adiabatic Calorimetry

Adiabatic calorimetry is the technique employed for measuring runaway chemical reaction rates. This approach measures temperature and pressure rise rates, and when a low-phi factor instrument such as the ARSST and/or VSP2 is used, the rates are directly scalable. There are several techniques for adiabatic calorimetry experiments: open cell tests, closed cell tests, closed-to-open tests, and scaled venting tests. This article focuses on open and closed system testing.

The primary difference between open and closed system testing is the relative void volume compared to the sample volume. In open system testing, the test cell is open to a larger closed containment vessel. In these systems, an inert backpressure is superimposed on the sample to suppress the vaporization of volatiles. In closed system testing, the test cell is isolated from the containment vessel, so the void volume is relatively small. Typically, the test cell headspace is evacuated at the start of a closed test. In these cases, the test cell pressure measurement can be representative of the vapor pressure of the system. Both open and closed system data are valuable and can be used for a variety of test goals. If the goal is vent sizing, often both open and closed tests, or two open system experiments with different backpressures, are required. Figures 2 and 3 illustrate the differences between an open and closed system VSP2 setup with the key difference being the vent line on the test cell for open system testing.

Open and Closed System Testing

Open and closed test configurations can lead to different test results such as different peak pressures and pressure rise rates, possibly different peak temperatures and temperature rise rates, and mass loss.

Potential differences in the test results between open and closed system experiments are the peak temperatures and pressures measured during the experiment; this can be due to tempering. Tempering is an endothermic process represented by the latent heat of vaporization. Experimentally, this can be observed as a plateau in the temperature as a function of time profile and a decrease in the temperature rise rate with a corresponding increase in the pressure rise rate. In a closed system experiment, the measured pressure will follow the vapor-liquid equilibrium with deviations caused by the generation of noncondensable gas. The maximum closed test temperature may be limited if the material has a high vapor pressure or if significant noncondensable gas is generated. Material cannot vaporize and leave the test cell in closed system experiments, so tempering of the material will not be observed and stalling of the reaction temperature and rates will not occur.

However, in closed system experiments, the available void volume is relatively small (often 40 ml headspace void compared to an 80 ml liquid sample in the VSP2). If the experimental procedure is not properly designed, this could potentially lead to liquid-full conditions at elevated temperatures.

Further, if gas generation occurs, the measured pressure in a closed system experiment will rapidly increase, potentially rupturing the test cell or prematurely stopping the experiment. It is also important to note that there is uncertainty in the void volume within a closed system experiment as the density of the material changes with temperature leading to liquid swell and a decrease in void. Bloating of the test cell itself (if there is excess internal pressure) can increase the void volume. Also, there is uncertainty in the amount of gas that goes into solution at high pressure. This can make quantification of gas generation difficult.

In an open system experiment, there is an excess gas void volume (typically about 3.8 liter of void in the VSP2 containment vessel), and therefore the measured peak pressures are much smaller allowing vigorous gas generation to be tracked throughout the reaction. The following equation converts the experimentally measured pressure rise rate (Pa/s) to the process scale gas generation rate (m³/s):

Open system experiments can exhibit tempering (something very beneficial to understand when sizing emergency relief systems). The temperature at which tempering occurs is dictated by the containment backpressure. Higher backpressures will suppress vaporization until a higher temperature (raising the boiling point), allowing the reaction mass to remain in the test cell and continue reacting adiabatically without boiling. If the material is nonvolatile (or at least is nonvolatile up to the maximum tested temperature), little difference is expected between the maximum reaction temperature and the temperature rise rates in an open versus closed system experiment. However, it is possible that some mass loss from the test cell during open system testing may occur due to other factors such as noncondensable gas discharged to the containment vessel, liquid entrainment inside gas bubbles during gas generation, and/or two-phase flow through the test cell vent. Decreasing the reactant mass can change the peak pressure rise rate achieved during testing.

Data Acquisition for Gassy Systems

However, in closed system experiments, the available void volume is relatively small (often 40 ml headspace void compared to an 80 ml liquid sample in the VSP2). If the experimental procedure is not properly designed, this could potentially lead to liquid-full conditions at elevated temperatures. The two key parameters for vent sizing of gassy systems are the mass remaining in the vessel at peak reaction rates and the maximum pressure rise rate. Fauske advocates using open cell tests performed with a superimposed backpressure equal to the maximum allowable accumulation pressure (MAAP) of the intended process vessel. This approach allows the full decomposition reaction to be captured without rupturing the test cell. A potential disadvantage of this approach is that some mass will leave the test cell. This effect is minimal if only gas is leaving the test cell. A potential solution to overcome significant mass loss during open cell testing is to increase the imposed backpressure, which in turn increases the gas density reducing the density difference between the gas and liquid thereby minimizing the superficial velocity through the test cell. Additionally, reducing the starting sample mass increases the test cell headspace and in turn decreases the potential for liquid entrainment from the test cell. This variation in mass remaining in the vessel can alter the peak pressure rise rate measured during testing impacting the relief requirement. Therefore, it is essential to perform the experiment in a way to best simulate the vessel in question.

Dicumyl Peroxide Case Study

Carson (2011) [5] presented calorimetric data for the decomposition of 40% dicumyl peroxide in 2,2,4-trimethyl-1,3-pentanediol diisobutyrate varying the starting backpressure and presented the resulting relief requirements. Carson noted differences in the calculated required vent area using open cell data with different superimposed backpressures (0 barg and 6 barg). The results show that increasing the back pressure from 0 barg to 6 barg causes the calculated required vent area to increase by a factor of 2-3, and the result from the open system experiment under 6 barg was very similar to a closed system experiment.

To explain the observations, it is appropriate to recall the expected effects of increasing the backpressure of the system:

• The gas density increases, reducing the density difference between the gas and liquid and reducing the superficial velocity through the test cell.
• Vaporization is reduced.

This is best illustrated using the drift-flux model. The dimensionless gas velocity (ψ) for 40% dicumyl peroxide with an initial pressure of 0 barg is 2.38 but is only 0.34 for an initial pressure of 6 barg. These values are plotted against the alpha-psi curve considering churn-turbulent (C0=1.5), bubbly (C0=1.2), and bubbly (C0=1.01) flow regimes in Figure 4.

Figure 4 indicates that complete vapor-liquid disengagement will occur after achieving a smaller void fraction in the high backpressure experiment given the three simulated flow regimes. Because of the increased quantity of two-phase flow prior to disengagement in the low backpressure experiment, a reduced reactant inventory can be expected along with a reduced peak pressure rise rate and smaller calculated vent size.

The Fauske team conducted additional experiments now utilizing 45% dicumyl peroxide in TBIX. One experiment was conducted with a backpressure of 0 psig and a second experiment was performed with a backpressure of 400 psig (28 barg). The predictions of the superficial velocity within the test cell for the two experiments is illustrated in Figure 5.

These experiments were performed to simulate loss-of-cooling scenarios with starting temperatures of 110°C. The test cell vent was fitted with an immersion heater and attached thermocouple (TC) which functioned as a “foaminess” detection device, shown conceptually in Figure 6. Power to the immersion heater was adjusted to maintain a temperature difference of 100°C between the immersion heater and the sample, with the expectation that foamy venting from the test cell would quench the immersion heater thus sensing the two-phase discharge.

In these experiments the immersion heater was placed inside the vent line and positioned so the tip of the thermocouple was just inside the test cell headspace. Figure 7 shows the immersion heater (left) installed in the VSP2 test cell vent line (right).

Figures 8 and 9 present the experimental results. In these figures, the sample temperature (blue solid line), guard heater temperature (blue dashed line), and immersion heater temperature (black dashed line) are plotted as a function of the experiment time. The initial immersion heater oscillations are due to testing the heater’s response time. The adiabatic reaction is evident immediately once the sample reaches 110°C with a sharp drop in the immersion heater temperature observed at a sample temperature of 165°C when the liquid sample first contacts the immersion heater.

A computer simulation was then performed considering the kinetic parameters outlined in Table 1. The simulation assumes complete reaction conversion and does not account for any mass loss due to venting. Figure 10 shows the experimental temperature rise rate as a function of temperature plotted against the simulation results. These results are consistent between the experiment and simulation up to approximately 210-215°C. While vaporization of the TBIX solvent is not expected at these temperatures, as it has a normal boiling point of 280°C, the dimensionless superficial velocity of this experiment would indicate that two-phase flow is predicted until a test-cell (vessel) void fraction of approximately 0.55 or higher is reached at which point complete disengagement would be expected. Further, the reaction product acetophenone has a normal boiling point of 202°C and may have contributed to tempering, further reducing the sample mass. The experiment also showed upward drift above approximately 235°C due to a nearly empty test cell.

The corresponding pressure rise rate data are shown in Figure 11. Here the measured peak pressure rise rate is approximately 150 psi/min and is a factor of ~4 less than the peak simulated pressure rise rate. This is expected because of early mass loss from the test cell (a combination of the noncondensable gas leaving the test cell, two-phase flow out of the test cell, and/or production of a volatile species vaporizing) and further confirmed by quenching of the vent line sensor above 165°C.

Figures 12 and 13 show results for a second experiment conducted with a backpressure of 400 psig (28 barg). In these figures, the sample temperature (blue solid line), guard heater temperature (blue dashed line), and immersion heater temperature (black dashed line) are plotted as a function of the experiment time. Again, there is some “tuning” of the immersion heater early in the test. Here, the adiabatic reaction is observed immediately once the sample reaches 110°C with a sharp drop in the immersion heater temperature observed at a sample temperature of 245°C due to the liquid sample contacting the immersion heater.

Figure 14 shows excellent agreement between the measured temperature rise rates and the simulation results.

The corresponding pressure rise rate data are shown in Figure 15. Here the measured peak pressure rise rates are also in excellent agreement with the simulation results. This agreement is attributed to the higher backpressure, which greatly reduces vaporization of the volatile components and reduces the superficial velocity, resulting in vapor-liquid disengagement at a lower void fraction.

These data can be easily used to size an emergency relief system using a tool such as FERST Powered by CHEMCAD [9].

Conclusion

Gassy systems present challenging reactive venting situations where the reaction temperature cannot be controlled by tempering, and therefore the relief vent size must accommodate the expected peak gas generation rate that the vessel might experience. Transient mass loss of reactants at scale is expected, largely due to two-phase flow, with potentially minor contributions from the release of noncondensable gas and possibly vaporization of volatile components without significant tempering. The loss of reactant inventory is important prior to reaching peak reactive conditions, as it decreases the peak generation rates (since the contributing reaction mass is smaller) and therefore decreases the relief requirements. Vent sizing experiments must be designed to ensure peak rates are measured that are applicable to full-scale conditions. Performing open cell adiabatic calorimetry testing with a backpressure equal to the vessel MAWP is the suggested approach for collecting the appropriate data such that Fauske’s simplified approach for vent sizing [7-8] can be employed.

References

1. https://www.arkema.com/usa/en/product/organicperoxide/luperox/dicupr/
2. DIERS User Group Round Robin (1996)
3. European DIERSUser Group Round Robin (2009)
4. L. Véchot, J. Kay, and J. Wilday, “Round robin vent sizing exercise on a gassy system: 40% dicumyl peroxide in butyrate solvent,” in proceeding of the Hazards XXII Conference, IChemE Symposium Series 156, 2011, no. 156, pp. 278–286.
5. Carson, D. et al., 2011, “Emergency Relief Vent Sizing for Runaway Chemical Reactions Using Different Methods,” 7th Global congress on Process Safety, Chicago, March 2011.
6. Fauske & Associates, Inc., 1983, “Emergency Relief Systems for Runaway Chemical Reactions and Storage Vessels: A Summary of Multiphase Flow Methods,” FAI/83-27, October 1983.
7. Fauske, Hans K., 2006, “Revisiting DIERS’ Two-Phase Methodology for Reactive Systems Twenty Years Later,” Process Safety Progress, Vol. 25, No. 3, 2006.
8. Fauske, H.K., 2000, “Properly Size Venting for Nonreactive and Reactive Chemicals,” Chem. Eng. Prog., February 2000.
9. FERST powered by CHEMCAD Version 1.0.0.16848, Fauske & Associates, LLC, 2024.