CHAPTER 16 • Oceanic DOC Measurements
307
poor accuracy can arise from the sample injection system, from memory effects and
from sensitivity to instability of instrumental parameters (temperature, vibrations etc.).
Due to the rapid expansion after injection, the stability of the catalyst bed becomes
an important issue. After long use with marine water a salt crust is formed in the top
part of the catalyst and, surprisingly, improved peak sensitivity and shape is achieved.
This was attributed to the formation of a funnel shape of the precipitate that could
help guide the sample into the catalyst (Skoog et al. 1997). The use of a loop-type injector, allowing precise injections under closed atmosphere, improves the precision
(0.6% RSD for n = 12 on Antarctic seawater samples with 85 jlM C level, 4.0% RSD for
n = 4 on distilled water with 7 jlM C level) and eliminates the possibility of contamination during injection (Qian and Mopper 1996).
The memory effect arises from void volume at the top of the oxidation column (Qian
and Mopper 1996) and from adsorptive properties of the catalyst (Cauwet 1994). It is
evident when injection of sea water or high OC standards are followed by injections
of distilled water (DI): the first injection of DI gives signals always larger than the following replicate injections of Dr. This effect originates from incomplete combustion,
and especially from flaking off of uncombusted OC in the cold zone at the top of the
column. It can be reduced by eliminating the void volume by direct injection into the
high temperature zone through a platinum tube, or accurately direct the injected flow
on the hot zone. As far as the type of catalyst is involved, quartz or silica have lower
memory effects than alumina, which, being an amphoteric oxide, may adsorb CO2
depending on the partial pressure of CO2 and column temperature. In this case the
memory effect is proportional to the injected amount of C. However, the results from
injection of 14C-Iabelled organic material showed that the oxidation of organic carbon to gaseous compounds on Pt/alumina beads is complete, and that the catalyst is
not a source of carryover signal between injections. The blank signal from further
injection of water was attributed to carbonaceous compounds contained in the catalyst particles (Skoog et al. 1997).
As for column packing, the recovery (based on peak area) is independent on oxidation column temperature in the range 650-900 °C (Sharp et al. 1995; Qian and
Mopper 1996) and the flow rates of oxygen, air or even nitrogen (Skoog et al. 1997).
Low recoveries at lower temperatures may be due to memory effects, which may become worse with decreasing temperature. However these parameters influence the
sensitivity, that is the slope of the calibration curve, and the peak broadening. The peak
broadening changes significantly with temperature and the nature of the analyte
(see Fig. 16.3). Often the peak originated by natural samples contains two shoulders.
Suzuki et al. (1992) suggested that the two shoulders are an effect of compounds having different combustion rates. However, due to the flash evaporation of the sample,
the two shoulders could also be caused by pressure effects on the detector, or to the
cooling down of the catalyst enough to combust only part of the sample, leaving the
remaining carbon to combust after reheating (Skoog et al. 1997).
The decrease in sensitivity at low temperatures or high flow rates is related to the
decrease in oxidation efficiency due to the lower rates of oxidation and shorter residence time in the column. Conversely, the increase in precision at lower flow rates
is somewhat offset by peak broadening, which may adversely affect the integration
precision. For high precision a very stable gas flow, especially during injection, is
needed.
307
poor accuracy can arise from the sample injection system, from memory effects and
from sensitivity to instability of instrumental parameters (temperature, vibrations etc.).
Due to the rapid expansion after injection, the stability of the catalyst bed becomes
an important issue. After long use with marine water a salt crust is formed in the top
part of the catalyst and, surprisingly, improved peak sensitivity and shape is achieved.
This was attributed to the formation of a funnel shape of the precipitate that could
help guide the sample into the catalyst (Skoog et al. 1997). The use of a loop-type injector, allowing precise injections under closed atmosphere, improves the precision
(0.6% RSD for n = 12 on Antarctic seawater samples with 85 jlM C level, 4.0% RSD for
n = 4 on distilled water with 7 jlM C level) and eliminates the possibility of contamination during injection (Qian and Mopper 1996).
The memory effect arises from void volume at the top of the oxidation column (Qian
and Mopper 1996) and from adsorptive properties of the catalyst (Cauwet 1994). It is
evident when injection of sea water or high OC standards are followed by injections
of distilled water (DI): the first injection of DI gives signals always larger than the following replicate injections of Dr. This effect originates from incomplete combustion,
and especially from flaking off of uncombusted OC in the cold zone at the top of the
column. It can be reduced by eliminating the void volume by direct injection into the
high temperature zone through a platinum tube, or accurately direct the injected flow
on the hot zone. As far as the type of catalyst is involved, quartz or silica have lower
memory effects than alumina, which, being an amphoteric oxide, may adsorb CO2
depending on the partial pressure of CO2 and column temperature. In this case the
memory effect is proportional to the injected amount of C. However, the results from
injection of 14C-Iabelled organic material showed that the oxidation of organic carbon to gaseous compounds on Pt/alumina beads is complete, and that the catalyst is
not a source of carryover signal between injections. The blank signal from further
injection of water was attributed to carbonaceous compounds contained in the catalyst particles (Skoog et al. 1997).
As for column packing, the recovery (based on peak area) is independent on oxidation column temperature in the range 650-900 °C (Sharp et al. 1995; Qian and
Mopper 1996) and the flow rates of oxygen, air or even nitrogen (Skoog et al. 1997).
Low recoveries at lower temperatures may be due to memory effects, which may become worse with decreasing temperature. However these parameters influence the
sensitivity, that is the slope of the calibration curve, and the peak broadening. The peak
broadening changes significantly with temperature and the nature of the analyte
(see Fig. 16.3). Often the peak originated by natural samples contains two shoulders.
Suzuki et al. (1992) suggested that the two shoulders are an effect of compounds having different combustion rates. However, due to the flash evaporation of the sample,
the two shoulders could also be caused by pressure effects on the detector, or to the
cooling down of the catalyst enough to combust only part of the sample, leaving the
remaining carbon to combust after reheating (Skoog et al. 1997).
The decrease in sensitivity at low temperatures or high flow rates is related to the
decrease in oxidation efficiency due to the lower rates of oxidation and shorter residence time in the column. Conversely, the increase in precision at lower flow rates
is somewhat offset by peak broadening, which may adversely affect the integration
precision. For high precision a very stable gas flow, especially during injection, is
needed.