POLY-OLEFIN PROCESS GAS ANALYSIS BY MASS SPECTROMETRY

Transcription

1 JPACSM 44 POLY-OLEFIN PROCESS GAS ANALYSIS BY MASS SPECTROMETRY Peter Traynor Thermo ONIX Angleton, TX Robert G. Wright Thermo ONIX Winsford, Cheshire, UK KEYWORDS: PROCESS MASS SPECTROMETRY, POLY-OLEFIN PROCESS GAS ANALYSIS, POLYETHYLENE, POLYPROPYLENE, FAST ANALYSIS, MULTI-STREAM ANALYSIS, PROCESS CONTROL ABSTRACT World-wide poly-olefin production is currently undergoing rapid growth. For process control using rapid on-line measurement, it is expected that mass spectrometry (MS) will be increasingly favored. This paper describes the analysis of polyethylene and polypropylene process gas by MS. Sample data from high density polyethylene (HDPE) process analysis are presented and discussed. The advantages of using process MS for fast precise multi-stream analysis of many reactors in a large, modern poly-olefin plant are outlined. Latest developments in process MS technology are also detailed. INTRODUCTION Poly-olefin production has traditionally utilized on-line gas chromatography (GC) for process control. However, with the advent of faster, more precise analysis and the development of lower cost, and more compact and robust process instrumentation, MS will be increasingly favored. Analytically, the component concentrations required for control in poly-olefin production are easily measured by MS, on the basis that the overlapping peaks can be deconvoluted by mathematical treatment, e.g. by matrix inversion 1. Typical process stream compositions for Polyethylene (PE) and Polypropylene (PP) are shown in Table I, along with typical analytical precision obtained by process MS. Table I Typical Process Stream Compositions for PE and PP and Typical Analytical Precision Obtained by Process MS PE PP Conc. Std. Dev. Conc. Std. Dev. Hydrogen Ethylene Ethane Nitrogen Propylene Propane Butene Isobutane n-butane Hexene Hexene n-hexane In both cases the typical analysis time is 10 seconds for all components using process MS, compared with 10 minutes by process GC. Key control parameters are the following ratios of concentrations: PE Hydrogen / Ethylene 1-Butene / Ethylene Hexene-1 / Ethylene PP Hydrogen / Propylene Ethylene / Propylene 1-Butene / Propylene

2 JPACSM 45 The actual compositions (types of components and their concentrations) can be varied considerably depending on the type of poly-olefin being produced, e.g. low density polyethylene (LDPE), linear low density polyethylene (LLDPE) and high density polyethylene (HDPE) are three classifications of polyethylene according to density, but within each classification there are various process types, catalysts, and reaction conditions to achieve different grades of material. Because the uses for poly-olefins 2,3 and the range of production processes are so diverse, a key requirement for an on-line instrument is that it be sufficiently flexible to allow measurement over a wide range of component types and concentrations. Modern process MS, with its flexible and almost unlimited mass selection, is particularly well suited in this respect. The control of the parameters mentioned above is important for producing poly-olefin to a consistent specification for particular applications, e.g. in water pipes or parts for the automotive industry. EXAMPLE PROCESS MS DATA FROM HDPE PROCESS The data presented in this paper were obtained with a Thermo ONIX VG Prima process MS. The HDPE process that was being analyzed is the Phillips particle form slurry process using a loop reactor. In the Phillips process, HDPE is made from polymerizing ethylene with small amounts of 1-hexene and 1-butene as co-monomers to introduce side branches to the main (CH 2 ) n chains to the control the polymer density, while hydrogen is used to control the molecular weight. This process is currently the most widely used for HDPE. In this process an isobutane liquid carrier is used in combination with a catalyst slurry in a circulating loop reactor. The polymer is produced as particles suspended in the liquid carrier which settle at the bottom of the reactor and are removed as a slurry every 30 seconds and then dried in a flasher. It is this gas evolved in the flasher which is of interest, since its composition is identical to the composition in the polymerisation reactor, except for the addition of nitrogen introduced as a purge gas. The main interest is in the levels of hydrogen, ethylene, 1-butene, and 1-hexene. The MS is calibrated using individual gas components to determine their respective cracking patterns, along with a multi-component mixture to determine the relative sensitivities. Typical cracking pattern and relative sensitivity data are shown in Table II. One of the most demanding measurements is that of hydrogen, which needs to be measured with high accuracy at low levels, e.g mole %. However, the extent of interference from hydrocarbons being fragmented in the ion source to give H + 2, which interferes with the signal from H 2 gas, can be considerable (these interferences typically amount to 0.1 mole %). Any uncertainty in the extent of these hydrocarbon interferences is reflected in uncertainty in the measurement of hydrogen in the PE process. System precision was checked by periodically analyzing the certified mixture that was also used to calibrate for the different component MS sensitivities. Those values are shown in Table III in units of : Table II Typical Cracking Pattern and Relative Sensitivity Data for HDPE Flash Gas Analysis Mass Relative Sensitivity Hydrogen Nitrogen Ethylene Ethane 1-Butene Isobutane n-butane Hexene- 1 Hexene- 2 Hexane

3 JPACSM 46 Table III System Precision Demonstrated by Periodically Analyzing a Certified Mixture Mix A A B B Cert. Mean Std. Dev. Mean Std. Dev. Hydrogen Ethylene Nitrogen Isobutane n-butane Hexene Hexene n Hexane A = data immediately after re-calibrating B = data 3 days after calibration For the consideration of any new on-line analytical technique, the actual accuracy for real process samples (which might contain interfering unknowns) is also a very important criterion in assessing its applicability. The accuracy of MS was compared with GC by conducting a trial over several weeks based on taking samples from various HDPE reactors and comparing the MS analysis data and process GC data with a laboratory GC. The laboratory GC was regarded as having much higher precision compared with the process GC and the laboratory GC data were regarded as the benchmark. The comparison is shown in Figures 1 and 2, which show that the agreement between the MS and the lab GC is much closer than that between the process GC and lab GC. This demonstrates that MS can achieve a high level of accuracy for on-line control of HDPE production. Actual on-line process MS data from an HDPE plant during plant start-up are shown in Figures 3 and 4. These graphs show the advantage of fast on-line measurement, which gives an almost continuous trend of the critical component concentrations. For example, if the ethylene concentration were to suddenly become too great there is a great risk of excessive polymerization within the reactor and a blockage occurring. This can result in a considerable loss of production while the plant is taken out of service to clear the blockage with solvents. Fast analysis alerts the plant operator of abnormal parameters earlier and enables appropriate preventative action to be taken more quickly to avoid plant problems. Additionally, faster measurement allows the plant control to be operated closer to the ideal conditions, so that there are also economic benefits in terms of greater plant efficiency as well as reduced plant maintenance costs. An illustration of improved process monitoring using fast on-line measurement is shown in Figure 5 which graphs hypothetical data based on a sinusoidal variation in ethylene concentration in the process and the effect of the analysis time lag on accuracy. In a typical HDPE plant there are a number of reactors and these can all be monitored by a single process MS. The cycle time for each reactor will be given by the following equation: Cycle time = Number of reactors x 20 seconds The 20 seconds derives from the sum of the purge time required for a multi-stream sampling valve to switch from one stream to the next stream (10 seconds) and the analysis time (10 seconds). If the required cycle time (for optimum control) is actually two minutes, then it is possible to monitor six reactors with a single process MS. The traditional approach would be to monitor each of these reactors with a different process GC, but the reactor composition analysis would still only be updated every 10 minutes. The replacement of traditional mode of measurement with modern process MS provides lower cost, faster analysis, and improved process efficiency. Figure 5 illustrates the effect of variable process ethylene composition on accuracy of process analysis by fast measurement (MS) compared with slow measurement (GC), where a single MS is sequentially measuring six streams with a cycle time of two minutes while GC is only measuring a single stream with an analysis time of 10 minutes.

4 JPACSM L-GC H2 P-GC H2 MS H sample ref Figure 1. HDPE flash gas H2 analysis, comparing MS data and process GC data with laboratory GC data L-GC C2= P-G C2= MS C2= sample ref Figure 2. HDPE flash gas Ethylene analysis, comparing MS data and process GC data with laboratory GC data.

5 JPACSM 48 PROCESS MS INSTRUMENT CONSIDERATIONS FOR POLY-OLEFIN APPLICATION This section discusses some of the considerations for optimum performance of a process MS for poly-olefin analysis. The benefits of a system exhibiting the following performance characteristics have been discussed in a previous paper 1 : flat-topped peaks, two-dimensional focusing, high energy ion beam, fast scanning, and nonfractionating sample introduction. The importance of general considerations of area classification, operating environment, data communications, and presentation have also been reviewed in a previous paper 4. Analyzer Pressure Analyzer pressures for process MS are typically in the range 1x10-6 to 1x10-5 mbar, with the actual pressure being adopted for a given application dependent on the relative importance of various considerations (Table IV). Table IV Benefits of Operating at Lower or Higher Pressures Benefits of Lower Analyzer Pressure Longer filament life Less frequent ion source cleaning Better linearity Lower abundance of interfering ion-molecule species Less drift between calibrations Benefits of Higher Analyzer Pressure Improved signal to noise ratio Improved sample signal to background signal ratio Less conditioning Less memory It has been found that process MS applications involving high levels of heavier hydrocarbons are more successful at lower pressures, with the main consideration being reduced frequency of ion source cleaning required and reduced drift 5. Typically, the optimum analyzer pressure for poly-olefins is in the range of 1-2 x10-6 mbar 5. Vacuum Gauge A cold cathode vacuum (Penning) gauge is used as standard to interlock a process MS against vacuum failure (which can cause rapid degradation of ion optical components). This gauge is a rugged, sensitive, and low cost device. However, there is one drawback in its use for poly-olefin applications: the tendency for the cold cathode gauge to crack hydrocarbons to hydrogen. The interference with hydrogen analysis can amount to twice as much as that generated by hydrocarbon cracking in the mass spectrometer ionization chamber; moreover the extent of hydrogen interference due to the cold cathode gauge is considerably less stable than that of the ionization chamber. The problem can be eliminated by powering the gauge only during maintenance for vacuum monitoring, set-up, or troubleshooting and using an alternative gauge, e.g. a thermal conductivity (Pirani) gauge, for actual vacuum interlocking while in analysis or calibration mode. Filament Type Accelerated life tests have been performed with a Prima to determine the relative robustness of different electron emitting filament types for applications involving high levels of heavier hydrocarbons. The tests used a stream of 95% propane and 5% toluene, with pressures 10 3 higher than normal, and an emission (total electron production) current of 1 ma. A diverse range of filament materials, coatings, and geometries were considered in this study. Predictions of life under normal conditions were made using the simple assumption that filament lifetime is inversely proportional to analyzer pressure. The results are shown in Table V, where the different filament types are identified as Prfil-1 - Prfil-7. Table V Accelerated Filament Lifetime Tests Filament type Life (hours) at 1 ma and 10-3 mbar Predicted life (months) at 2x10-6 mbar Prfil Prfil Prfil Prfil Prfil Prfil Prfil The data are not intended to have absolute accuracy, but are useful to show general effects and the relative robustness of different filament types. These results have been borne out in real process analysis experience where a change from Prfil-1 to Prfil-3 filaments has produced an order of magnitude improvement in filament life 6. Actual life-times depend on the composition of the hydrocarbon process gas; there will be variations due to inhomogeneity of filament material, tolerances in construction, and configuration as well as other variances such as off/on cycling. The Prima uses Prfil-3 type filaments as standard for poly-olefin analysis.

6 JPACSM isobutane % mol ethylene :30 17:00 17:30 18:00 18:30 19:00 time Figure 3. Process MS data from an HDPE plant during plant start-up: isobutane and ethylene hydrogen % mol hexene hexene :30 17:00 17:30 18:00 18:30 19:00 time Figure 4. Process MS data from an HDPE plant during plant start-up: hydrogen and hexene-1 and hexene-2.

7 JPACSM % mol % ethylene error % ethylene MS error % ethylene GC time (min) Figure 5. Effect of variable process ethylene composition on accuracy of process analysis by fast measurement (MS) compared with slow measurement (GC), where single MS is sequentially measuring many streams (6) with a cycle time of two minutes while GC is only measuring a single stream with an analysis time of 10 minutes. Figure 6. Peak shapes obtained using (a) higher resolving power (b) lower resolving power, showing that (b) is a more fault tolerant configuration for hydrogen analysis.

8 JPACSM 51 Other Considerations There are a wide range of other factors in instrument design which influence the quality of the data produced by a process mass spectrometer. These factors include the following: Power supplies with higher specification, including less noise and drift and immunity from power glitches, surges etc. Higher resolution of mass position, higher mass range, faster mass setting and more stable mass scale. Higher resolution of low level peak intensities. Fewer electronic components and use of new components which have higher performance than older components. Better reliability, better diagnosis, and easier repair. Fault tolerant design. An example of a fault tolerant design, i.e. one which allows accurate measurement even under instrument fault conditions, is illustrated in Figure 6, which shows that with a wider peak width, the need for accurate mass scale alignment is much less critical. The Prima uses different collector apertures, which are selected by electrostatically deflecting the ion beam, to provide different resolving powers for low and high mass peaks. This permits the use of the widest possible peaks throughout the mass range, while avoiding overlap between adjacent mass peaks. Fault tolerant design is especially important in the case where one process MS replaces several process GCs, which could back each other up in the event of one failing. It is essential that a process MS in this situation exhibits a very high degree of reliability. CONCLUSIONS While world-wide poly-olefin production is currently undergoing rapid growth, it has already been demonstrated that process MS can provide very significant advantages: more precise and much faster analysis than traditional techniques. It is anticipated that MS will be increasingly used for more efficient control in poly-olefin production. Continuing developments in process MS hardware design are providing even higher performance, better reliability, and robustness. REFERENCES 1. R.G.Wright, Process Anal. Chem., Vol. IV, (1998). 2. Kirk Othmer, Encyclopedia of Chemical Technology, 4 th Edition, Wiley-Interscience, Vol. 17 (1996). 3. Chemistry in Britain, Sept 1998, P.Traynor, Process Anal. Chem., Vol. IV, (1998). 5. G. Josland, private communication. 6. R.G. Wright, unpublished data.