Identification and Characterization of Nonflammable Azeotropic Mixtures (Chapter 4)

CHAPTER IV Discussion This chapter contains a discussion of the methods provided in Chapter II, and the results provided in Chapter III. This discussion includes reasoning for the decisions that were made in this project, as well as data interpretation, and method evaluation. Solvent Selection The selection of the solvents used to make the azeotropes reported in Table 3 was far from random. All of the HFEs (hydrofluoroethers), Suprion, AE 3000, and Solkane 365 were chosen, because of their known flame-retardant abilities, which will be discussed in more detail later in this chapter. AE-3000s higher polarity and hydrogen bonding terms, and Solkane 365s higher dispersion term were promising. However, these substances tended to have solvency issues with hydrocarbon grease. CBTF was chosen, because of its VOC-exempt status, and tremendous dispersion term placing it close to TCE in Hansen space. This substance is flammable though, and had to be paired with a flame-retardant substance for safe use in a vapor degreaser. EL and tBAc seemed to be good possibilities as azeotrope components, because of their higher polarity and hydrogen bonding terms. However, all attempts to quench their flammability failed, and EL also had problems with miscibility. PFBI was a great solvent. However, it could not be used in solvent blends due to its rather rapid decomposition to form molecular iodine, and the fact that there was only a small supply of this solvent to test. Methanol was chosen, not only for its good solvency and high wetting index, but also because it tends to mix and dissolve in environments that would not be predicted by Hansen solubility parameter theory.10 While methanol is not VOC-exempt, it is a promising molecule for low-VOC blends, because of how VOC content is measured. Methanol is a very light molecule and VOC content is measured by mass of VOC per liter of solvent. So, there will be more molecules per gram of VOC when using methanol versus most other substances. However, all attempts to quench the flammability of methanol in an azeotropic mixture have failed. Fractional Distillation Fractional distillation was used in this study to find seven low-boiling azeotropic solvent blends. However, no actions were taken to investigate other types of azeotropes between any of these substances. Also, there was a goal of forming non-flammable azeotropes with good solvency against hydrocarbon grease, which led to pairing substances that have high solvency with substances that are non-flammable. So, there is a possibility that more azeotropes exist between these substances, or that some of these substances actually form double azeotropes. Since the boiling points of these azeotropes were obtained by reading a mercury or alcohol thermometer on the distillation apparatus, the readings are subject to an uncertainty of a tenth of a degree Celsius. However, it is rarely the case that the temperature reading does not fluctuate some during a distillation. In cases where there were small fluctuations, such as the temperature rising and dropping slightly, the boiling temperature was chosen as the temperature that the thermometer read most often within that fraction of distillate. Some of the boiling points were confirmed with a digital reading from a thermocouple later. All of the azeotropes that were made with this method had boiling points well below the boiling point of either component, with the exception of AZ6. AZ6 was an azeotrope between Suprion and CBTF, and its boiling point was the same as the boiling point of Suprion. However, this azeotrope consisted of 20 % CBTF by volume. There are examples of low-boiling azeotropes whose boiling points are very close to the boiling point of the lowest-boiling component. The water-ethanol azeotrope is an example of this with a boiling point that is only a few tenths of a degree lower than the boiling point of ethanol. Chemometric Analysis Since the components of each azeotrope were known, a chemometric analysis was the easiest way to measure the ratios of the components. In this case, a Raman chemometric analysis was performed, but other methods were tested. The other methods tested include gas chromatography with a flame ionization detector (GC-FID), nuclear magnetic resonance (NMR) analysis, and infrared (IR) spectroscopy. GC-FID was unfruitful, because no separation was obtained between the components of the mixture. This method also took much more time than the Raman analysis. Another student from this research group successfully performed NMR analysis to obtain the composition of azeotropes, however, the method was much more difficult and time consuming than the Raman analysis. An IR chemometric analysis was attempted, but the high-resolution instrument was too good at seeing the peak shift between the pure substances and the mixture. This led to issues with the process of fitting the component spectra to the azeotrope spectrum. The Raman analysis was performed with a lower resolution instrument, but some of the peak shift phenomenon was apparent. The RMS difference between the spectrum of the azeotrope and the spectrum composed of the weighted average of the components sometimes had a first-derivative nature, indicating a vibrational frequency shift in the spectrum of the mixture. However, the RMS difference between the location of a peak in the azeotrope and the location of the same peak in the pure substance was always very small. Since the Solver tool in excel was being used to find the ratios of the components, the analysis was performed with different initial composition ratios to ensure that Solver was reaching the global minimum for the RMS difference. For the same sample, Solver would always reach the same composition within 0.2 %. The fact that the Raman analysis was yielding volume percentages was also verified. Two mixtures of different volume compositions were analyzed using this chemometric method. The composition from the Raman analysis was within 1 % of the volume composition for both mixtures, and was never equal to the mass, or mole percent composition. All of the Raman measurements were performed at ambient temperature. No study was performed to see if the shifts in room temperature had an effect on the volume percentages obtained from this analysis. It is expected that slight changes in temperature would change the output of the analysis, because of the dependence of density on temperature. Azeotropes are usually defined in terms of mole ratio, and while the mole ratio will always be the same the volume ratio will change slightly when the temperature changes. As for the composition data seen in Table 3, it is only coincidence that AZ1 and AZ2 have the same composition values. This only occurred because the volume percentages were rounded to the nearest one percent. The one to one ratio of components in AZ4 was surprising, but it indicates that the intermolecular attractions that these substances share with each other are very similar to the intermolecular attractions that they share with themselves. AZ5 was very interesting, because it was a heterogenous azeotrope at room temperature. This azeotrope consisted of two immiscible layers that were each a mixture of the two components. Flash Point The flash point test was the first test performed on the azeotropes after determining their composition. This test was crucial to this project, because any substance that exhibits a flash point cannot be used in a vapor degreaser. The tag closedcup flash point tester was chosen to perform the flash point test, because of its electric heater, and imitation of the environment inside of a vapor degreaser. There is a level of controversy over the flash point test, because some methods report a flash point for a substance that is reported to not have a flash point when a different method is used. In ASTM D56, a flash is defined in the following way: “3.1.1.1 The specimen is deemed to have flashed when a flame appears and instantaneously propagates itself over the entire surface of the fluid. 3.1.1.2 Discussion—When the ignition source is a test flame, the application of the test flame may cause a blue halo or an enlarged flame prior to the actual flash point. This is not a flash and should be ignored.”.20 When a flash occurs during a flash point test, it is very apparent. However, when certain halogenated substances are heated in a flash point tester they participate in burning with the test flame by making the flame bigger, or changing the color of the flame. This is addressed in the standard as well by the following excerpt: “X1.1 A condition during flash point testing can occur with certain mixtures whereby the nonflammable component of the sample tends to inert the vapor space above the liquid, thus preventing a flash. Under this condition, the flash point of the material is masked resulting in the reporting of incorrect high flash point or no flash point. X1.2 This flash point masking phenomenon most frequently occurs with ignitable liquids that contain certain halogenated hydrocarbons such as dichloromethane (methylene chloride) and trichloroethylene. X1.3 Under this condition, no distinct flash as defined in 3.1.1 is observed. Instead a significant enlargement of the test flame and a change in the color of the test flame from blue to yellow-orange laminar flame is observed.”.20 The above information reveals that the halogenated hydrocarbons do not exhibit a flash, but will participate in burning. The flash point tests in this study were in agreement with this. The goal in this study was to use these halogenated hydrocarbons to make the vapor zone inert, thus making a mixture, with a flammable substance, that does not have a flash point. This would make the substance useable in a vapor degreaser. The substances in Table 3 that report a flash point as less than or equal to a certain temperature, are the substances that exhibited a flash at the starting temperature of the test. The case of AZ5 was particularly interesting, because the flash point of the azeotrope was a lower temperature than the flash point of ethyl lactate. Ethyl lactate was the only substance in the blend that had a flash point. It was deduced that this was the result of the increased vapor pressure of EL in the azeotrope, combined with the inability of HFE-7500 to quench the flammability of ethyl lactate. Solvency Screening and Solvent Wetting Index AZ 4 did not pass the solvency screening, because it had a Ptransferred of 0.37 %. The hope with this blend, was that the higher D term of HFC3 would combine with the higher P and H terms of AE3 to make a relatively good solvent. However, it can be seen in Figure 22 that the two components of this blend were both better than the azeotrope at removing hydrocarbon grease. From the data it appears that AZ6 and AZ7 were better than the components that the blends were made from, and that they were competitive with nPB and TCE. However, this was not attributed to the blends’ ability to dissolve the grease. In these cases the grease fell out of the vials in clumps, indicating that it was loosened from the glass, but not dissolved. This was attributed to the azeotropes’ increased wetting index compared to CBTF. The wetting index was increased because of an increased density and a decreased surface tension. Density The density measurements in Table 5 were all very good. The literature values were all within the uncertainty of the experimental measurements after interpolation to 25 °C. However, when the density versus temperature data was recorded, Suprion began to get cloudy at the low end of the temperature range. It was deduced that this led to a bad linear least squares fit when calculating the thermal expansion coefficient. This is the reason for the rather large uncertainty value on the thermal expansion coefficient and the density at 25 °C for Suprion. This uncertainty value for density doesn’t look very large, but when you compare it to the others, it is three times as large as the next highest uncertainty value. The density values for the azeotropes seem to be the expected values when compared to the literature values of the components. This is because the density is between the values of the two components, and closer to the value of the component that makes up the majority of the blend. Surface Tension The surface tension measurements for the pure substances were not included, because only two ambient measurements with slightly differing temperatures were measured for these substances. The two data points were too close in temperature to establish a reliable temperature dependence, and two data points is not enough to establish a statistically acceptable relationship. This method and instrument have been used to obtain good surface tension data of well-known substances, such as water, in the past. This removes the need for validation of this method. For the azeotropic mixtures, a high temperature and a low temperature were added when measuring the surface tension of the azeotropes. This gave a good temperature span for an interpolation and enough data points to create a statically significant relationship between temperature and surface tension for the azeotropes. The surface tension values for the azeotropes seem to be what is expected when compared to the literature values of the components. This is because the values are between the values of the components, but closer to the component with the smallest surface tension. Surface tension usually behaves this way with mixtures, because the lower surface tension molecules are surface active and break the interactions of the higher surface tension component. Viscosity The viscosity measurements were performed using the method with the most modifications. However, the modifications that were made only had the possibility of making the measurement better for substances that were stable in the temperature range of the density measurements. “Stable”, in this case, means that the substance does not undergo phase changes or other changes that could have a large effect on the density data. This excludes Suprion, because of the cloudy phenomenon that occurred at the low end of the temperature range. If this is taken into consideration, the use of interpolated densities in the calculation of viscosity should make the measurement more reliable, because it is difficult to ensure that the density and viscosity measurements take place at the same temperature without temperature-controlled environments. The use of wellknown standards to calibrate an instrument reduces error by removing the need to measure the dimensions of the viscometer, and allowing the temperature dependent viscometer constant to account for any changes in the viscometer’s dimensions when the temperature is varied. Also, a 30-frames per second video camera was used to time the ball-drop. This led to an uncertainty of ±1 frame or ±0.03 seconds in the timing, which is much better than any human could consistently do with a stopwatch. The literature value for HFE-7500 is far from being within the uncertainty of the measurement for unknown reasons. The measurement for Suprion is surprisingly close to the literature value, considering the suspect density data. However, the uncertainty value is quite large, which may be related to the cloudiness in the density measurement. The CBTF literature value is within the uncertainty of the measurement. Also, this measurement has a reasonably low uncertainty value. The azeotropic viscosity values seem to be reasonable, because they are between the literature viscosity values of the components. However, AZ6 seems to close to the center of the values of its components, while AZ7 is more than twice as close to the value for the component with the lower viscosity than it is to the other component. Azeotrope Prediction The HSPiP software has an azeotrope prediction model. Using this model, it is apparent that the boiling point prediction is very good, but the volume percent composition prediction does not work very well for the two examples shown in Figure 23 and Figure 24. In these two figures the volume composition and boiling point prediction using the HSPiP azeotrope prediction model is blown up at the bottom, and are compared to the experimental values located near the top. Figure 23. AZ 6 HSPiP volume composition and boiling point prediction compared to experiment. Figure 24. AZ 7 HSPiP volume composition and boiling point prediction compared to experiment. HSPiP acknowledges in the “help” documents that the results from the azeotrope prediction for fluorinated compounds may be unreliable due to a lack of data for fluorinated compounds. The two azeotropes that were tested in the HSPiP prediction model contained fluorinated compounds. The azeotrope prediction method that was produced in this study was tested on the seven azeotropes that were discovered in this project, which all contained fluorinated compounds. In these examples, the data needed to calculate the azeotropic point was taken from the HSPiP database. This data consists of values that were predicted based on the structures of the molecules. This could have had an effect on the prediction model’s ability to accurately predict the azeotropic composition, and this effect could be removed by using data from a database of experimental data. Also, the assumption that the volumes of the two components are additive when mixed, could cause much of the error with this prediction method. This error could be removed if a model could be found to accurately predict the real volume upon mixing, and how much of that volume each component would make up. CHAPTER V Conclusions The goal of this thesis was to provide methods that can be used to search for new azeotropes with specific desired properties, and methods to characterize these new azeotropes. All azeotrope possibilities that were examined in this study were low-boiling azeotropes composed of non-aqueous solvents, and the desired properties were for the azeotropes to be non-flammable and have good solvency against hydrocarbon grease. This is because the azeotropes were meant for use in a vapor degreaser as drop-in replacements for TCE and nPB. A secondary goal of this thesis was to discover a method for predicting if an azeotrope will form between two substances using the Hansen Solubility Parameters. This research uncovered seven new azeotropic solvent blends. In four of these azeotropes, the flammability was not quenched by the amount of flame retardant solvent in the blend. This means that these blends are failed attempts at producing azeotropes with the first desired property using the method outlined in this thesis. However, three of the azeotropes that were produced, were non-flammable. Out of these three, one did not pass the solvency screening against hydrocarbon grease. So, this one failed to meet the requirements of the second desired property. This still leaves two azeotropes that were non-flammable, and had some solvency against hydrocarbon grease. These solvents were not able to compete with TCE and nPB in terms of solvency though. The azeotropes in this study show that while the search method outlined in this thesis is not perfect, it is capable of producing azeotropic blends with the desired properties. The modified characterization methods that were produced in this study, were definitely improved from their original state, and the wholly-new Raman chemometric analysis was perfect for its purpose. The original characterization methods assumed that the density measurement was taken at the same temperature as the surface tension and viscosity measurements. However, in this study it was quickly realized that this is usually false unless all three measurements are performed on the same sample at the same time, which is impossible. The Raman analysis was quick, easy, and provided readily useable volume fractions for mixing the identified azeotropes in larger quantities. The implementation of property versus temperature relationships made the measurements of density, surface tension, and viscosity much more accurate due to the dependence of all of these properties on temperature. Surface tension and viscosity are dependent on density and these relationships helped to avoid the difficulty of implementing precise temperature control on the tests in order to achieve the same temperature between these tests and the density test. These relationships also helped to extrapolate or interpolate experimental values to 25 °C for reporting purposes, and comparisons to literature. Video-assisted data collection for flash point and viscosity, had a huge impact on these two methods. In the flashpoint test, the video served as visual evidence, and as reassurance in situations where the flash was not immediately obvious. In the viscosity measurements, the video was used to obtain more accurate timing for the ball-drop. Since this was traditionally done using a stopwatch, the timing was made much more reproducible and less reliant on the individual who performs the timing, or human reflexes. All the methods used in this thesis are inexpensive, and can be easily implemented with little budgetary impact. An azeotrope prediction model that relates the HSPs to azeotrope formation was discovered in this thesis. This model works fairly well for the azeotropes discovered in this project despite using predicted data and an assumption that is definitely not true in all cases. This azeotrope prediction method could save companies and researchers massive amounts of time and resources, because it provides a way to predict if two substances will form an azeotrope before going to the lab and attempting to make an azeotrope. A graduate student from the Department of Computer Science at SHSU, Nathan Thompson, has been contracted to write a program using this prediction model. With this program, it will be possible to take a giant list of compounds and find every combination of two substances that is predicted to form an azeotrope. If this program was available at the beginning of the SERDP project, there is a good chance that a better solution to the problem could have been found in the amount of time that was available for that project.