Journal of Chromatographic Science, Vol.

41, November/December 2003

Analysis of Low Levels of Oxygen, Carbon

Monoxide, and Carbon Dioxide in Polyolefin
Feed Streams Using a Pulsed Discharge
Detector and Two PLOT Columns

Downloaded from https://academic.oup.com/chromsci/article/41/10/545/372141 by guest on 27 February 2024

David B. Wurm, Kefu Sun, and William L. Winniford
The Dow Chemical Company, 2301 Brazosport Blvd., Freeport, TX 77541

Abstract rately detecting various poisons at this low level.

In this paper, a method for detecting sub-ppm levels of the
A gas chromatography (GC) method is developed for rapid analysis common catalyst poisons CO, CO2, and O2 in polypropylene feed
of polyolefin feed streams for the catalyst poisons CO, CO2, and streams is presented. The method uses gas chromatography
O2. The method uses an HP MoleSieve column in parallel with a (GC) with pulsed discharge detection (PDD) and is capable of
CP-PoraPLOT Q column and a pulsed discharge detector (PDD). detecting each of these components in the sub-ppm range. This
Detection limits for each of the potential poisons are between 50 method is general in nature and should be readily applicable to
and 250 ppb. For a 10-ppm standard, the precision of the method many types of polyolefin work.
was ± 4.2% for oxygen, ± 7.8% for carbon dioxide, and ± 2.0% for In modern “high yield” polypropylene polymerization sys-
carbon monoxide. In addition to the polyolefin feed stream, tems, extremely high purity propylene is necessary for efficient
nitrogen and hydrogen feed streams are also analyzed. In each polymerization. Compounds that bind to the active centers of
case, sampling is observed to be a critical issue, with air
the catalyst such as carbon monoxide, mercaptans, and arsine
contamination of the sample cylinder often the limiting step in
determining the true level of oxygen. It is also noted that large
can only be tolerated in ppb levels (1). Other poisons such as
amounts of argon are present in the standards when nitrogen is carbon dioxide, oxygen, and water can be tolerated at the slightly
used as a balance gas. Because the trace oxygen peak partly higher levels of 2–5 ppm. Molecular sieves are effective at
coelutes with the larger argon peak, it is suggested that helium be removing water and carbon dioxide from systems, and catalytic
used as the balance gas for all standards. This general experimental processes are available to remove other poisons. However, these
arrangement should be effective when applied to feed streams for precautions do not guarantee that all catalyst poisons will be
other polymers as well. removed from a system. Purification beds can fail because of
being overloaded or not properly regenerated. There are many
opportunities for feed stream integrity to be compromised.
When catalyst poisoning occurs, the negative financial impact is
Introduction tremendous. For these reasons, a fast and reliable method for
determining the presence of catalyst poisons in the low- to sub-
At the heart of all polyolefin polymerizations is catalyst chem- ppm range is critical.
istry. The type of catalyst dictates polymer properties such as Traditional GC detection methods are limited in their abilities
molecular weight, polydispersity, stereospecifity, and even to detect catalyst poisons in this range. The thermal conductivity
microstructure. The efficiency of the polymerization is also detector (TCD), though universal, suffers from relatively poor
largely governed by catalyst activity. For these reasons, it is crit- sensitivity. The flame ionization detector (FID) is more sensitive,
ical to ensure that losses in catalyst activity are kept to a min- but responds primarily to compounds with a hydrocarbon back-
imum. However, polymer catalysts are extremely sensitive to bone. The helium ionization detector (HID) is an extremely sen-
contaminants in the reaction system such as moisture, air, and sitive detector and responds to the permanent gases, but the
trace levels of other contaminants such as mercaptans. It suffices detector has several disadvantages: (a) the response is highly
to say that almost any compound with π or nonbonding elec- dependent on the applied voltage, (b) the source can become
trons is a potential catalyst poison. Because less than 1 ppm of easily contaminated, and (c) the radioactive source material
catalyst poison in a reaction system may be enough to cause a must be licensed and closely monitored. The PDD offers the sen-
significant decrease in catalyst activity, there is a need for accu- sitivity and universal response of the traditional HID with none

Reproduction (photocopying) of editorial content of this journal is prohibited without publisher’s permission. 545
 Journal of Chromatographic Science, Vol. 41, November/December 2003

of the drawbacks (2). For these reasons the PDD is the detector Experimental
of choice for many applications in which traditional HIDs were
previously used. There are several types of the PPD including The experimental configuration is shown in Figure 1. An HP
pulsed discharge photoionization detector (PDPID), pulsed dis- 5890 A gas chromatograph (Agilent Technologies, Wilmington,
charge electron capture detector (PDECD), and pulsed discharge DE) was used to carry out the separations. An HP MolSieve
emission detector (PDED). The abbreviation “PDD” often refers column (Agilent Technologies, 15-m × 0.53-mm i.d., 50-µm
to the pulsed discharge detector operated in photoionization coating) and Varian/Chrompack PPQ (30-m × 0.53-mm i.d., 20-
mode. Cai et al. have shown that the PDECD compares favorably µm coating) (Varian, Middelburg, the Netherlands) were used in
with traditional radioactive electron capture detectors (ECD) for parallel as shown in Figure 1. The temperature was held at 30°C
pesticide analysis (3,4). The PDECD has also been shown to be for 1.5 min, then ramped at 20°C/min for 1 min, and held at
effective for environmental applications such as the analysis of 50°C for 7.5 min. The carrier gas flow rate was 12 mL/min, and
polychlorinated biphenyl (PCBs) and chlorofluorocarbons the sample loop was 100 µL. The Model D-1 PDD (Valco
(CFCs) (5,6). Wentworth et al. have shown that qualitative iden-

Downloaded from https://academic.oup.com/chromsci/article/41/10/545/372141 by guest on 27 February 2024

Instruments, Houston, TX) was used. The PDD temperature was
tification of organohalogen- and sulfur-containing compounds 140°C, and a high PDD discharge gas flow rate of 45 mL/min was
can be achieved by coupling a PDED with a PDPID using pure used for the megabore capillary columns. Columns were baked
helium as the discharge gas (7). Sun et al. also have developed a overnight at high oven temperature to ensure that the PDD
chlorine-selective PDED that is promising for a broad range of background was low. All standards were custom mixed by Scott
environmental analysis applications (8–10). Judging from the Specialty gases (Plumsteadville, PA). Typically, a large amount of
versatility of the PDD, the application of the detector to petro- sample was used to purge the fittings of all residual air that
chemical analysis should be something that is straightforward in entered the system when switching sample containers until the
principle. A PDD operated in photoionization mode was used in constant O2 and N2 concentrations were obtained. Polyolefin
this study. PDD refers to pulsed discharge photoionization feed samples were collected in a piston cylinder under high
detector in this paper. enough pressure to maintain the samples in a liquid state.
Even though the PDD can detect the permanent gases and Simply bleeding a small amount of sample into the sample loop
carbon dioxide in low levels, chromatographic separation that was an effective means of delivering the sample.
allows for quantitation of CO, CO2, and O2 is not a trivial matter.
Molecular sieve or zeolite columns separate molecules based
(predominantly) on size. These columns are popular because of
their ability to separate the permanent gases. However, water, Results and Discussion
carbon dioxide, and other polar compounds adsorb on these
columns (11–13). Samples containing water and other polar As alluded to previously, it was necessary to use two columns
compounds can often be analyzed using porous-layer open in parallel to adequately resolve and quantitate the components
tubular (PLOT) columns coated with porous synthetic polymers of interest. In order to demonstrate the feasibility of this
of styrene and divinylbenzene. In order to analyze for all of the approach and to determine retention times and elution order of
compounds of interest here, an HP MoleSieve column and a the analytes, standards were run on each of the columns sepa-
PoraPLOT Q (PPQ) were used in parallel as shown in Figure 1. rately. Figure 2 shows a chromatogram for a 1000-ppm mixture
This experimental arrangement worked nicely for the perma- containing all of the analytes for the PPQ column. As expected,
nent gases including H2, O2, N2, CH4, CO, as well as CO2. there was no resolution of the fixed gases, but the carbon dioxide
peak clearly eluted at approximately 2.2 min. The same separa-
Intensity (mV)

Time (min)

Figure 1. Schematic diagram of experimental set up. Figure 2. Chromatogram of 1000-ppm standard using PPQ column at 60°C.

546
Journal of Chromatographic Science, Vol. 41, November/December 2003

tion using the MoleSieve column is shown in Figure 3. Here, the column. There should be two peaks for methane, as this compo-
fixed gases were readily resolved. Because polar compounds and nent elutes from both columns, and several coeluting peaks
carbon dioxide adsorb to molecular sieve columns, no peak was caused by poor resolution of the permanent gases on the PPQ
evident for carbon dioxide. By connecting the columns in par- column. This is what was seen, as demonstrated in Figure 4. A
allel as shown in Figure 1, one should expect to see discreet custom standard containing 10 ppm by mole CO, CO2, and O2
peaks for all of the permanent gases because of the MoleSieve with N2 as the balance gas was then analyzed, resulting in an
unsatisfactory chromatogram, as shown in Figure 5. As seen in
this figure, the carbon dioxide coeluted with the major nitrogen
peak. Also, the oxygen peak appeared to be much larger than one
would expect for a 10-ppm standard. Closer inspection revealed
that there were actually two peaks, as shown in Figure 6. From
prior experience with the elution order on a MoleSieve column,
Intensity (mV)

it was reasoned that the larger of the two peaks was possibly

Downloaded from https://academic.oup.com/chromsci/article/41/10/545/372141 by guest on 27 February 2024

argon. A call to the vendor of the mixed-gas blend confirmed this
hypothesis. According to the vendor, high-purity nitrogen typi-
cally contains up to 500 ppm argon. For better results, a 10-ppm
custom standard with helium as the balance gas was obtained. A
chromatogram for this standard is shown in Figure 7. The
expected results were seen with excellent resolution of the per-
Time (min)
manent gases, and carbon dioxide clearly resolved. Detection
limits for each of the components at 3 times signal to noise was
Figure 3. Chromatograpm of 1000-ppm standard using MoleSieve column 50–250 ppb, depending on how noisy the baseline was. The rela-
at 60°C.
tive standard deviation was typically 2–8%.
Intensity (mV)
Intensity (mV)

Time (min) Time (min)

Figure 4. Chromatogram of 1000-ppm standards using the PPQ and Figure 6. Chromatogram showing the large amount of argon typically found
MoleSieve columns in parallel. in samples that have nitrogen as the balance gas.
Intensity (mV)

Intensity (mV)

Time (min) Time (min)

Figure 5. Chromatogram of 10-ppm standards with nitrogen as the balance gas. Figure 7. Chromatogram for 10-ppm standards with helium as the balance gas.

547
 Journal of Chromatographic Science, Vol. 41, November/December 2003

A typical chromatogram for propylene feed is shown in attributed to air contamination. Spiking of a sample with
Figure 8. There was at least 30 ppm oxygen in each of the ini- ethane and comparison of retention times revealed that the
tial samples that were analyzed, as well as a large amount of peak at 4 min was ethane, an innocuous constituent of the
moisture. Because this amount of oxygen and moisture would sample. Although extreme care was taken during subsequent
result in an almost immediate loss of catalyst activity, it was samplings of the propylene feed, there was always several ppm
suspected that contamination occurred during the sampling of oxygen present. It could not be unambiguously determined
process. The nitrogen peaks were also sufficiently large as to be if this was caused by sampling issues or if there was in fact this
level of oxygen in the sample. This method is capable of deter-
mining trace ppm O2 in process streams, but sampling is no
doubt the limiting factor in determining low levels of oxygen in
conventional cylinder samples.
Production nitrogen and hydrogen feed streams were also
analyzed in the same manner as was the propylene feed. Figure 9

Downloaded from https://academic.oup.com/chromsci/article/41/10/545/372141 by guest on 27 February 2024

Intensity (mV)

shows a chromatogram for the first set of nitrogen samples that

were received. Because the nitrogen is supposed to be ultrahigh
purity, the huge peaks for oxygen (1.1 min) and water (7 min)
immediately indicated that the sample was possibly contami-
nated with outside air. A second round of samples for which
much more care was taken during the sampling process revealed
that there was less than 500 ppb of oxygen in the nitrogen
stream, as shown in Figure 10. No other contaminants were
Time (min)
detected. This ruled out the nitrogen stream as a source of con-
Figure 8. Typical chromatogram of polyolefin feed sample. tamination. Figure 11 shows a chromatogram for the hydrogen
feed stream. Again, oxygen and water were shown to be present
in the sample in large quantities. Poor sampling was probably
the reason for this, as well.

Conclusion
Intensity (mV)

A GC method employing PPD for the analysis of polyolefin

feed streams was presented. The detection limits for the cata-
lyst poisons CO, CO2, and O2 were between 50–250 ppb, with
relative standard deviations of 2–8%. Standards using helium
as a fill gas were clearly superior to standards using nitrogen
when the PPD was used. Interpretation of oxygen present
Time (min) in the feed streams was ultimately limited by sampling
Figure 9. Chromatogram for nitrogen feed stream when proper sampling care techniques, with large amounts of air contamination probably
was not taken. contributing to most of the oxygen and water present in
the samples.
Intensity (mV)

Intensity (mV)

Time (min)
Time (min)
Figure 10. Chromatogram of nitrogen feed stream when care was take during
the sampling process. Figure 11. Chromatogram of hydrogen feed stream.

548
Journal of Chromatographic Science, Vol. 41, November/December 2003

References selective detector for gas chromatography. J. Chromatogr. A 872:

119–40 (2000).
8. K. Sun, W.E. Wentworth, and S.D. Stearns. Development and char-
1. E.P. Moore, Ed. Polypropylene Handbook. Hanser Publishers, New acterization of chlorine-selective pulsed discharge emission
York, NY, 1996, p. 299. detector for GC. J. Chromatogr. A 872: 141–65 (2000).
2. W.E. Wentworth, S.V. Vasnin, S.D. Stearns, and C.J. Meyer. Pulsed 9. K. Sun, W.E. Wentworth, and S.D. Stearns. Applications using
discharge helium ionization detector. Chromatographia 34: 219–25 the chlorine-selective pulsed discharge emission detector.
(1992). J. Chromatogr. A 872: 167–78 (2000).
3. H. Cai, W.E. Wentworth, and S.D. Stearns. Characterization of the 10.K. Sun, W.E. Wentworth, and S.D. Stearns. Characterization of chlo-
pulsed discharge electron capture detector. Anal. Chem. 68: rinated compounds using a dual chlorine-selective pulsed dis-
1233–44 (1996). charge emission detector-helium pulsed discharge photoionization
4. H. Cai, S.D. Stearns, and W.E. Wentworth. Pulsed discharge elec- detector system. J. Chromatogr. A 872: 179–89 (2000).
tron capture detector operating in the constant-current mode by 11. J. de Zeeuw, R.C.M. de Nijs, and L.T. Henrich. Adsorption chro-
means of feedback DC bias voltage. Anal. Chem. 70: 3770–76 matography on PLOT (porous-layer open-tubular) columns: a new
(1998). look at the future of capillary GC. J. Chromatogr. Sci. 25: 71 (1987).
5. W.E. Wentworth, J. Huang, K. Sun, Y. Zhang, L. Rao, H. Cai, and

Downloaded from https://academic.oup.com/chromsci/article/41/10/545/372141 by guest on 27 February 2024

12.L. Henrich. Recent advances in adsorption chromatography for
S.D. Stearns. Non-radioactive electron capture detector. analysis of light hydrocarbons in petrochemical-related materials.
J. Chromatogr. A 842: 229–66 (1999). J. Chromatogr. Sci. 26: 198 (1988).
6. G. Gremaud, W.E. Wentworth, A. Zlatkis, R. Swatloski, E.C.M. 13.T.G. Andronikashvili, V.G. Berekzin, N.A. Nadiradze, and L.Ya.
Chen, and S.D. Stearns. Windowless pulsed discharge photoioniza- Laperashvilli. Increased effectiveness of chromatographic columns
tion detector application to qualitative analysis of volatile organic packed with zeolites in the separation of mixtures of some organic
compounds. J. Chromatogr. A 724: 235–50 (1996). compounds. J. Chromatogr. 365: 269–77 (1986).
7. W.E. Wentworth, K. Sun, D. Zhang, J. Madabushi, and S.D.
Stearns. Pulsed discharge atomic emission detector: an element- Manuscript accepted August 29, 2003.

549