THE USE OF DSC AS A SCREENING TOOL IN CHEMICAL REACTION HAZARD ANALYSIS. P.G.Lambert* and G.Araery*

Transcription

1 THE USE OF DSC AS A SCREENING TOOL IN CHEMICAL REACTION HAZARD ANALYSIS P.G.Lambert* and G.Araery* This is a basic introduction into the use of a differential scanning calorimeter (DSC) as a screening tool in the evaluation of the potential thermal hazards of materials and mixtures. It is, therefore, aimed at those with little or no previous experience of the technique in chemical reaction hazard analysis. A broad range of topics is discussed including basic theory and instrumentation, advantages and limitations of DSC, the importance and variation of experimental technique, the use of the data obtained and the role of DSC as an integral part of a hazard evaluation system. KEY WORDS: DSC, THERMAL STABILITY, MAXIMUM SAFE OPERATING

2 handling system. The instrument currently in use in our laboratory is a Perkin-Elmer DSC-4. Other instruments are available from manufacturers such as Du-Pont, Setaram, Mettler Scientific Equipment, Netzsch and Stanton Redcroft. 2.THEORY AND PRINCIPLES When a sample is heated, there may be a change in its enthalpy. The purpose of differential thermal systems is to record the difference between the enthalpy change associated with this sample and an inert reference material. There are three types of system: (a) classical DTA; (b) 'Boersma' DTA; and (c) DSC. The most important attributes of these systems are illustrated in Figure 1. In all these dynamic systems the instrument is usually programmed to alter the sample temperature linearly over a given range at a predetermined rate. This is called the scan rate. For both classical and Boersma DTA, a single heat source is used to heat both sample and reference, temperatures being measured by sensors embedded in the sample and reference materials (classical), or attached to the sample holders (Boersma). The temperature difference between the sample and reference, AT = T s - T r, can be plotted against time. The magnitude of AT at any given time is proportional to: a) the heat capacities of sample and reference; b) any enthalpy change in the sample; c) the total thermal resistance to heat flow, R. The value of R depends on sample properties, the way the sample is packed into the sample pan, the quality of thermal contact between the sample pan and holder and, unfortunately, temperature. Because of these variables, it is not easy to convert the peak areas - gained from the AT vs. time plot - into energy units, and consequently DTA would not appear to be suitable for quantitative calorimetric measurements. Conversely, in DSC both sample and reference are provided with separate heaters. The temperature of the sample holder is kept the same as that of the reference holder by adjustment of the power to the individual heaters and if any temperature difference develops between the sample and reference (because of an exothermic or endothermic transition in the sample) the power input to the relevant heater is adjusted to compensate. This is the 'power compensation' principle. A signal, proportional to the difference between the power input to the sample and that to the reference, dtf/dt, is fed to a recording device and this is usually plotted against the average temperature of the sample/reference system. As the temperature increases a DSC thermogram is generated. With this set-up, R is unaffected by any change in the sample and for a small weight of sample in close contact with its sample pan and a low scanning rate, the thermal resistance of sample and pan, R s, is very small compared to the resistance between pan and holder, R 0 Variations in R 0 affect the DSC thermogram peak shape but not the area under the peak so true quantitative data may therefore be obtained. In practice, there is little difference between modern DTA and DSC for thermal hazards work. Only where very accurate results are required from small enthalpy changes do the above theoretical arguments have any practical significance. Figure 2 illustrates the main features of a typical scanning DSC trace. It can be seen that heat flow (measured in cal/sec or mw) is plotted on the ordinate and temperature (increasing from left to right) on the abscissa. By convention, endothermic transitions are plotted upwards and exothermic downwards. The temperature at which the thermogram first deviates from the baseline during an exothermic decomposition is called the onset temperature. The temperature corresponding to the maximum deflection of the curve (maximum power output) is called the peak temperature and the integrated area under the curve gives the total enthalpy change for the decomposition. 3.ADVANTAGES AND LIMITATIONS OF DSC One of the great advantages of DSC as a technique is that it is not dedicated to thermal hazard analysis work. In addition to examination of exothermic changes during reactions the apparatus can also be used to examine the purity of solid samples (many publications are available), study polymorphic crystalline forms, measure heat capacity^ and thermal conductivity and define glass transition temperatures in polymeric 86

3 materials.'* It is also possible to generate phase diagrams from DSC data to define the melting and solubility relationships of multi-component systems as a function of temperature.5 The instrument can therefore play a major part in a modern analytical laboratory and the initial cost of the instrument can more easily be justified on these grounds for the smaller chemical company working to a limited budget. Other advantages include the speed of the technique (a screening run can be performed in less than one hour compared to one day for the simplest ARC run), low running costs (sample pans are relatively cheap and data from many runs can be stored on one floppy disc if the instrument is interfaced with a computer), minimal operator time and the relative ease of generation of kinetic data.6 DSC however, has certain limitations in studying potentially hazardous reactions: (a) If standard non-sealed aluminium pans are used the magnitude of any enthalpy change detected may not reflect the relative hazard of a substance, since gas evolution and evaporative losses are not taken into account. This can be easily overcome and will be covered in more detail later. (b) Small, possibly non-homogeneous, samples can be unrepresentative of material from the chemical plant. (c) Measured onset temperature is a function of the sample heating rate. Again this will be covered more fully later. (d) Assessment of onset temperature from a DSC thermogram (by defining the point at which the trace first deviates from a baseline) can be very subjective. (e) The DSC test condition is essentially isothermal whereas plant decompositions usually occur in an adiabatic environment. (f) The sensitivity of the technique when used in the normal scanning mode is relatively low (1-5 Wkg-1). These and other limitations and advantages of DSC/DTA are discussed in detail by Hentze,' Duch et al. and Schulz et al." 4.EXPERIMENTAL METHODS The main factors that affect the appearance of a DSC trace and govern its suitability as a screening tool for the evaluation of thermal stability hazards are as follows: a) Calibration. The instrument must be calibrated both for temperature linearity in the scanning range and for accuracy of the heat flow scale. b) Encapsulation. The sample must be encapsulated in a suitable pan under the correct pressure/atmosphere. c) Sample selection. The sample must be representative and be of a weight appropriate to the sensitivity of the instrument and the pan type/volume. d) Sample heating rate. The optimum scan rate for a temperature scanning run or temperature for an isothermal run must be selected. These points are covered in greater detail below: a) Calibration. Before any quantitative measurements can be made the calorimeter must be calibrated to set the temperature scale accurately and to fix a calibration constant for the energy scale. With modern instruments this can be done semi-automatically with internally generated calibration signals but for higher precision, and with older generation machines, use is often made of high-purity metals with accurately known fusion enthalpies as calibration standards. The most commonly used standard is indium: AW(fusion) = 28.5 Jg"-*-, m.p C. An accurately weighed indium sample is placed in a sample pan (not Gold) and a thermogram of the melting peak is run at the desired heating rate. The fusion enthalpy can be obtained from the area under the melting peak and the true melting point obtained from the intersection between the continuation of the slope of the leading edge of the melting endotherm and the scanning baseline. The instrument's settings can then be 'trimmed' until these readings coincide with the true values. It may be found useful to use organic calibrants for calibrating the temperature scale when high precision is required, because of the much greater thermal conductivity of metal calibrants, but this is not normally necessary when using DSC as a 87

4 screening tool. With older DSC instruments exhibiting less linear baselines, more than one calibration standard may be required. Once satisfactorily calibrated, the instrument is ready to analyse a sample. b) Encapsulation. To encapsulate the test sample there is a choice of the following: (i) Aluminium pans. These are simple and cheap (approx. lop each) and usually have a circular lid which is crimped or cold-welded to form a seal often capable of withstanding 1-2 bar pressure. For materials that interact with aluminium, gold pans are also available. These types of pan are of limited application in thermal stability evaluation owing to the potential loss of energy from the sample by evaporation or gas evolution. (ii) Large volume volatile sample pans. These are usually made in stainless steel and sealed with a viton rubber seal capable of withstanding pressures up to 24 bar. They can be hazardous since excessive pressure build-up due to the vapour pressure of liquids or gaseous decomposition products can deform the capsule, relieve the force-fit seal and cause the capsule lid to be expelled at considerable velocity from the sample holder. This could cause considerable damage to the DSC instrument. If these capsules are used, it is advisable to minimise the final scan temperature and keep the sample weight as low as practicable. (iii) High-pressure capsules. These are commonly fabricated from stainless steel but can be gold-plated. They are usually made in threaded halves which screw together and form a seal on a copper, aluminium or gold-plated disc. They should be capable of withstanding pressures of up to 150 bar and are therefore ideal for suppressing the endothermic signal which would result from the volatilisation of sample material or from gas evolution during sample decomposition. This allows enthalpy changes occurring above normal atmospheric boiling points to be detected. The sample may also be confined in the capsule under a nitrogen atmosphere to prevent spurious oxidation reactions from being interpreted as genuine exothermic decomposition reactions. These types of capsule are used routinely in our laboratory. They are relatively expensive ( each) but can be re-used up to times with care, especially if pan inserts are used. It should be noted that sealed high-pressure capsules can show exothermic transitions due to hydrolysis with water at temperatures (cf. 200 C) where water would not be present in practice and care should therefore be taken with the interpretation of the DSC data given by aqueous samples. (iv) Glass ampoules. Taylor et al.*- have reported the use of sealed, flint glass ampoules which enabled DSC data to be obtained at up to 35 bar and with highly corrosive compounds. Their main use is to contain samples which, during their decomposition, show a marked catalysis by the metal from which the normal pans are made. DSC peaks are generally broader with glass ampoules owing to the lower thermal conductivity of flint glass relative to aluminium, but reasonable quantitative results can still be obtained. Extreme care should be taken in using glass ampoules with highly energetic materials or with samples expected to liberate gaseous decomposition products. (v) Standard aluminium pans used in a dedicated high-pressure DSC head. Certain DSC instrument manufacturers also market special DSC assemblies capable of being pressurised with an external inert gas such as nitrogen or a reactive gas such as oxygen to a set pressure. These give the advantages that rates of change of enthalpy may be enhanced by the effect of pressure and, in general, cheap aluminium pans may be used for sample encapsulation. Aluminium pans also have thinner walls than high pressure pans, giving higher sensitivity. The main disadvantage lies with the relatively nigh price of these specialist attachments. The application of pressure DTA (DSC) to thermal hazard evaluation has been discussed by Seyler.H Figure 3 illustrates the effect of changing the pan type on a DSC run. c) Sample selection. To ensure the generation of meaningful results, it is vital to select a sample which is representative of the actual material which is to be processed on the plant. This can be very difficult with the small sample sizes used in DSC, especially with non-homogeneous reaction masses and thick, tarry distillation residues. This is one of the major disadvantages of the technique in relation to thermal hazard evaluation. Provided a suitable sample can be selected, it can be accurately weighed into the chosen type of pan. Sample size is usually between 2 and 10 mg but this depends on the 88

5 nature of the material and the energetics of any likely decomposition. Some experience is required to gauge likely heat output during decomposition (and therefore weight of sample to use) from the chemical structure of the material. d) Sample heating rate. The pan can then be sealed and placed (in good contact with the sample holder) in the calorimeter along with an empty reference pan of the same construction. Both can then be simultaneously heated at a controlled scan rate or held isothermally at a set temperature whilst a measurement of the change of enthalpy of the sample is recorded as a function of temperature (or time in the case of an isothermal run). If a temperature-scanning run is performed, scan rate has a marked effect on the test results. Hentze^ has found that the observed onset temperature for the decomposition of 4-nitroaniline increased from 210 C to 300 C when the heating rate was changed from 0.1 Kmin"! to 5 Kmin'l. Figures 4 and 5 show this effect for 2-nitrophenol. It can be seen from Figure 5 that a graph of the logarithm of scan rate versus reciprocal peak temperature, gives a straight line. The slope of this line is proportional to the approximate activation energy for the decomposition. The ASTM method of determining Arrhenius kinetic constants of potentially hazardous materials" makes good use of this principle. It should be noted that temperatures used to generate this plot should be corrected for scanning rate dependence caused by the thermal lag of the sample holder and pan assembly. This can be done mathematically using an equation of the form: rcrue = T obs. - CdT/dt + D = where T^rue = actual temperature, Tobs. observed temperature, dt/dt = scanning rate in degrees per minute and C and D are constants. In general, low heating rates give the lowest observed exothermic initiation temperatures but detection sensitivity is also lowered. Exothermic self-heating of a sample will not be reliably detected unless it is significantly large when compared with the scan rate. Kinetic theory tells us that a substance should be undergoing decomposition, albeit at extremely slow rates, at temperatures much lower than the initiation temperature detected by a scanning method such as DSC. Therefore the use of initiation temperature on its own can mask the fact that time is also an important factor in an assessment of the stability of a substance.13 For the study of autocatalytic decompositions (which require an induction period), isothermal DSC can be usefully employed. The sample is maintained at a constant temperature in the region of interest and any enthalpy change recorded against time. In an autocatalytic material, ageing of the sample at a temperature below the 'normal' onset temperature can bring about a significant lowering of this onset temperature. If the temperature at which the sample is aged is sufficiently close to the 'normal' onset temperature, then decomposition may well occur actually during the ageing process. Figure 6 illustrates some typical DSC runs performed isothermally. As expected, it can be seen that the time to maximum decomposition rate is lowered from 8 hours to 4 hours when the ageing temperature is raised from 230'C to 250'C. This shows that time, as well as temperature, is of paramount importance in the setting of safe operating conditions for the distillation of a thermally unstable material, for example. 5.INFORMATION AVAILABLE FROM A DSC RUN Figure 2 shows what can be expected from a typical DSC scan. The questions which should be asked about a scan such as this are as follows: a) Are any exothermic transitions present? b) If so, how close to the intended plant operating conditions (or those resulting from any credible process deviation) is the onset of exotherm measured by the DSC? c) Is the onset temperature above or below the melting point of the material under test? d) What is the magnitude of any exotherm, i.e. how much energy is liberated during the decomposition? e) What is the rate at which this energy is evolved, i.e. how sharp is the 89

6 decomposition peak? f) If the run is isothermal, how long is it before an exotherm is observed at this temperature? Only when these questions have been answered can the data be interpreted. 6.INTERPRETATION AND USE OF DSC DATA FOR SCREENING PURPOSES In our laboratory, all reactants, products and intermediates from a particular process are screened by DSC using the techniques outlined in Section 4 above. A typical screen would be run on a 5 mg sample encapsulated in a high pressure pan at a heating rate of 5 Cmin"l. The end point for the run would normally be set to a temperature much higher than the proposed operating temperature of the process. If no exothermic transitions are detected in this initial screening run (i.e. the answer to question (a) above is no) the proposed maximum process temperature is considered to be safe. If exothermic activity is shown by the DSC trace then we would work to the so-called 100 C rule,-'-* i.e. the measured onset temperature is compared to the intended process temperature or the maximum possible temperature and if there is less than 100 C difference between the two, ARC studies would be considered necessary. If the onset temperature is greater than 100 C above the maximum process temperature or maximum possible temperature then we would perform more DSC runs (either scanning or isothermal depending on the particular case.) which may, on occasion, obviate the need for further examinations to be performed on the ARC. This type of approach is typical of the method normally employed when DSC is used as a tool for thermal stability screening. In the majority of cases this approach works well and the number of ARC runs required to assess the safety of a process can be cut dramatically by screening runs on DSC. We feel that it is unwise, however, to stick rigidly to systems such as the 100 C rule and all thermograms should be examined on their own merits. On occasions we have encountered materials that show a dramatic difference between onset temperatures measured by DSC and ARC. Figures 7 and 8 show the DSC and ARC traces for such a material, in this case an aliphatic nitro compound (Test compound X). It can be seen that the onset temperature as measured by DSC is over 125 C higher than that recorded by the ARC (raw data). We believe that this is due, at least in part, to sublimation of the material from the bottom to the top of the sample pan, thus dramatically increasing the thermal lag of the system. This has been confirmed by other tests, such as thermogravimetric analysis (TGA), which show the sample has a weight loss from just above ambient temperature. Examination of the DSC trace (with the benefit of experience) shows that the extremely high energy liberated during this decomposition, coupled with the high rate of energy release signals the need for further ARC testing. This example illustrates the possible dangers of an inexperienced person taking raw DSC data, perhaps from only a single run, and using it to define a maximum safe operating temperature for a plant process. If we are in any doubt about the onset and severity of any exotherm from a DSC trace then we always perform further tests. When unclear about the meaning of DSC data, other small companies who may not have access to other instruments such as ARC may be well advised to refer to specialist consultancies which offer an evaluation service. 7.CONCLUSIONS 1. DSC can be a useful technique for the screening of possible thermal instability in reactants, intermediates and products from a particular reaction. 2. It has the advantages of high throughput, low operator input and relatively low cost - especially if the initial purchase price of the equipment can be justified by its additional analytical uses. 3. Its main disadvantages lie with the small sample size used and the scanning rate dependence of the measured onset temperatures. With careful operation these can be largely overcome in the majority of cases. 4. Correct sample encapsulation is vital for the generation of meaningful data. In the majority of cases, sealed high pressure capsules (or a separate DSC pressure head) are required to suppress endothermic transitions which may mask a decomposition exotherm. 90

7 5. Where onset temperatures observed on DSC lie close to the proposed process operating temperatures, or if the observed exotherm is large then the use of adiabatic techniques with greater sensitivity or detailed kinetic analysis (classical or by DSC) is recommended. REFERENCES 1. R.Gygax, M.W.Meyer, F.Brogli. (1980) Thermal analysis C.F.Coates, W.Riddell. (1981) Chem. & Ind. 7 Feb p P.Lambert, G.Amery. To be published 4. J.H.Flynn. (1974) Thermochimica Acta B.Wunderlich. (1973) Thermochimica Acta A.A.Duswalt. (1974) Thermochimica Acta G.Hentze. (1984) Thermochimica Acta M.W.Duch, K.Marcali, M.D.Gordon, C.Heusler, G.J.O'Brien. Res. & Dev. Divn. Publn. No.582. E.I.Dupont De Nemours 9. N.Schulz, V.Pilz, H.Schacke. (1983) I.Chem.E. Symp.Ser. 82 B1-B6 10. G.R.Taylor, G.E.Dunn, W.B.Easterbrook. (1971) Anal.Chim.Acta R.J.Seyler. (1980) Thermochimica Acta ASTM method of testing for determining the Arrhenius kinetic constants for the screening of potentially hazardous materials. Perkin-Elmer Thermal Analysis Application Study No W.J.Fenlon. (1982) U.S. National Safety Council Chemical Newsletter 14. J.L.Cronin, P.F.Nolan, J.A.Barton. I.Chem.E. Symp.Ser (and references therein) (a) Classical DTA s R ^^ Single heat source (b) 'Boersma' DTA / S R 3 :-:::j "A -, \ ^-^ Single heat source (c) DSC LJBL Pt sensors Individual heaters Figure 1. DTA versus DSC 9!

8 Onset of MOtherm 1S7'C ENOO melting of sample Decomposition of sample (heat output) POWER (MWAIIS) xn 1 1 X Cp displacement EXP Of IS TEMPERATURE (OEG.O Figure 2. A typical scanning DSC trace. 12D.DQ 140.DD B0.00? D S C TEMPERATURE CO Figure 3. Tho influence of pressure on the decomposition of AZOD1CARBONAMIOE 92

9 2-NITROPHENDL WT: mq?q. OD deg/min (CLOSED CELL) 1U. UO d99/mln deg/min deg/min / deg/min ~~""~ deg/min > TEMPERATURE Figure 4'. The effect of SCAN RATE on o OSC trace of 2-NITROPHENOL (C) Dsc Peak temperature (1000/T (K)) Flgut-o S. Scan i-oto»t. tompoi-otuto plot for Z-NI rrdphenol 93

10 ,i\ iro ENOO Normal decomposition at 230 'C. POWER (mwaits) EXO w D- * 6 TIME [HOURS) :ic- ENOO POWER UWATTS) M' Autocatalytic decomposition at 250 'C. EXO w D" TIME (HOURS) jar ENOO POWER (-.WAITS) x Autocatalytic decomposition at 230 'C. z/.\! TIME (HOURS) Figure 6. Typical Isothermal DSC runs.

11 IChemE SYMPOSIUM SERIES No.115 A I **" WT mg ' "\ SCAN RATE: 5. DO deg/min 1 r PEAK FROM: TO ONSET: J/GRAM: M1N: HO TEMPERATURE CO Figure 7. DSC thermogram of TEST COMPOUND X DSC -co -Seif Heat Rate. i. ; t \ /,i. f! i! 100 SOO TEMPERATURE G (VY) 300 foo 5CC Figure 8. ARC Sslf hoot rata vs. tomp. plot for TEST COMPOUND X 95