1348 ANALYST, AUGUST 1992, VOL. 117 Similarly, any variation in distance between sample and detector in a replicate analysis can contribute towards random error. Possible sources of such error are the variation in the thickness of the bases of the sample holders above the detector. Beckerg calculated that the sample shape variation was 3% for a 1 mm change in the average height of a sample at 5 cm distance from a 13% efficient detector. Similarly, Potts and Husseyg have calculated such variation resulting from reposi- tioning of a sample. They worked on a range of sample-to- detector distances of 5-50 mm and found that a repositioning discrepancy of only 0.2 mm for a source-to-detector distance of 10 mm can cause an error of up to 4% in the measured activity of a sample. Fortunately, in this work, all these errors could be eliminated by counting at large distances from the detector.Total Analytical Variation Taking account of all the analytical uncertainties, the total error for a sample i would be the sum of all absolute and relative random errors consisting of irradiation variation (iv), counting variations (cv) and the counting statistics (cs) error: (4) According to Heydorn and Damsgard,4 analytical variation, a2, becomes insignificant if it is less than one-third of the total variance (9). This can be checked conveniently by calculating the index of determination (ID): For an ID 20.8, the analytical variation is less than one-third of the total variance and can be ignored; therefore, eqn. (2) reduces to: or Experimental Blank Rock Standards Standard samples of similar matrices were needed to check the irradiation variation in this work.Therefore, a synthetic standard was prepared. A blank rock (basalt) of 200 mesh particle size was used as a base matrix because the samples were also siliceous ores. Blank ore, weighing 19 g, was doped with 19 g of a 13 mg kg-1 gold solution (HAuCI4). The mixture was blended on an orbital shaker and then frozen at -30 "C before drying in a vacuum-dryer. After drying, the synthetic standard was re-ground with a Tema swing mill in an agate pot. Finally, it was blended in an end-to-end shaker for 4 h. Reference Materials The reference ores of the Canadian Certified Reference Material Project, MA series, are typical of high and low (waste rock) ore grades from the Macassa Mine at Kirkland Lake, Ontario, Canada.The ore contains quartz, wall rock inclu- sions, carbonates, and small amounts of sulfides, tellurides and native gold. The principal sulfide is finely disseminated pyrite. Most of the gold occurs as an electrum. The high-grade ore, MA-1, was replaced by MA-la and then MA-lb, as they became depleted. The low-grade ore, MA-2, has been replaced by MA-2a. Irradiation Facilities All the samples in this work were irradiated in the CONSORT I1 reactor at Imperial College, London. It is a swimming pool-type reactor of thermal power 100 kW. A number of irradiation facilities are available, two of which were used: a manual epithermal system called CT8 and a new pneumatic epithermal large-volume irradiation system (ELVIS).Both these systems are lined with 1 mm thick cadmium and have epithermal-neutron fluxes of 2.8 x 1014 n m-2 s-' and fast-neutron fluxes of 2 x 1015 n m-2 s-1. Gamma-ray Spectrometry All the samples and standards were counted on a Ge(Li) detector [full width at half maximum (FWHM): 1.8 keV, peak/Compton ratio: 36.3 and efficiency: 8.l%, at 1.33 MeV]. The gamma-ray spectra were measured with an ND6700 multichannel analyser. Results Evaluation of Analytical Errors Irradiation variations A set of twelve 1 g replicates of the synthetic rock standard was prepared in polyethylene containers (18 x 8 mm) and irradiated in the CT8 system to measure the combined effect of neutron-flux variation and the sample geometry. Another batch of synthetic rock standards was then prepared by the same procedure and sixteen 1 g replicates from that batch were used to measure the irradiation variation in the ELVIS.Counting geometry variations The spectrometry system at the reactor centre has a sample changer on which 12 samples can be counted automatically. Small plastic cups (51 mm diameter and 1 mm thick) are used for holding the sample containers over the detector. A slight variation (0.2 mm) in the thickness of these cups was found to contribute a 1% error to replicate variation, at 11 mm. The error was reduced by using those with minimum variation, and eliminated by counting samples at 60 mm from the detector. A second counting geometry error, which was reduced by counting samples at 60 mm, was due to the movement of the sample inside the container (i.e., the sample shape variation).This error was evaluated for those samples measured at 11 mm from the detector by counting a sample six times without shaking and six times with intermittent shaking. Evaluation of Sampling Constants When this work was started, only the core tube system, CT8, was available for epithermal-neutron activation, and CANMET had certified two reference ores: MA-1 and MA-2 (MA-lb and MA-2a were certified later, after their deple- tion). In order to avoid flux variation, two middle positions of CT8 were chosen and so a maximum of 12 capsules (18 X 8 mm) could be placed in them. Therefore, 12 replicates of four sample masses (1, 1.5,2 and 2.5 g) were prepared to carry out the replicate analyses. Replicates larger than 2.5 g could not be irradiated because the irradiation container (18 X 8 mm) only held a maximum of 2.5 g of powdered rock.All the samples were irradiated for 7.5 h. By the time MA-lb and MA-2a were certified, the ELVIS had also been installed. Therefore, two sample masses, 1 and 10 g, were chosen to carry out the analyses on MA-lb and MA-2a. The samples were prepared in containers measuring 18 x 8 mm and 22 X 35 mm, respectively. The capsules were sealed thermally to avoid leakage during pneumatic transfer. Sixteen 1 g replicates were irradiated for 1 h, and sixteen 10 g samples were irradiated for 15 min, sequentially.1348 ANALYST, AUGUST 1992, VOL. 117 Similarly, any variation in distance between sample and detector in a replicate analysis can contribute towards random error.Possible sources of such error are the variation in the thickness of the bases of the sample holders above the detector. Beckerg calculated that the sample shape variation was 3% for a 1 mm change in the average height of a sample at 5 cm distance from a 13% efficient detector. Similarly, Potts and Husseyg have calculated such variation resulting from reposi- tioning of a sample. They worked on a range of sample-to- detector distances of 5-50 mm and found that a repositioning discrepancy of only 0.2 mm for a source-to-detector distance of 10 mm can cause an error of up to 4% in the measured activity of a sample. Fortunately, in this work, all these errors could be eliminated by counting at large distances from the detector. Total Analytical Variation Taking account of all the analytical uncertainties, the total error for a sample i would be the sum of all absolute and relative random errors consisting of irradiation variation (iv), counting variations (cv) and the counting statistics (cs) error: (4) According to Heydorn and Damsgard,4 analytical variation, a2, becomes insignificant if it is less than one-third of the total variance (9).This can be checked conveniently by calculating the index of determination (ID): For an ID 20.8, the analytical variation is less than one-third of the total variance and can be ignored; therefore, eqn. (2) reduces to: or Experimental Blank Rock Standards Standard samples of similar matrices were needed to check the irradiation variation in this work. Therefore, a synthetic standard was prepared.A blank rock (basalt) of 200 mesh particle size was used as a base matrix because the samples were also siliceous ores. Blank ore, weighing 19 g, was doped with 19 g of a 13 mg kg-1 gold solution (HAuCI4). The mixture was blended on an orbital shaker and then frozen at -30 "C before drying in a vacuum-dryer. After drying, the synthetic standard was re-ground with a Tema swing mill in an agate pot. Finally, it was blended in an end-to-end shaker for 4 h. Reference Materials The reference ores of the Canadian Certified Reference Material Project, MA series, are typical of high and low (waste rock) ore grades from the Macassa Mine at Kirkland Lake, Ontario, Canada. The ore contains quartz, wall rock inclu- sions, carbonates, and small amounts of sulfides, tellurides and native gold.The principal sulfide is finely disseminated pyrite. Most of the gold occurs as an electrum. The high-grade ore, MA-1, was replaced by MA-la and then MA-lb, as they became depleted. The low-grade ore, MA-2, has been replaced by MA-2a. Irradiation Facilities All the samples in this work were irradiated in the CONSORT I1 reactor at Imperial College, London. It is a swimming pool-type reactor of thermal power 100 kW. A number of irradiation facilities are available, two of which were used: a manual epithermal system called CT8 and a new pneumatic epithermal large-volume irradiation system (ELVIS). Both these systems are lined with 1 mm thick cadmium and have epithermal-neutron fluxes of 2.8 x 1014 n m-2 s-' and fast-neutron fluxes of 2 x 1015 n m-2 s-1.Gamma-ray Spectrometry All the samples and standards were counted on a Ge(Li) detector [full width at half maximum (FWHM): 1.8 keV, peak/Compton ratio: 36.3 and efficiency: 8.l%, at 1.33 MeV]. The gamma-ray spectra were measured with an ND6700 multichannel analyser. Results Evaluation of Analytical Errors Irradiation variations A set of twelve 1 g replicates of the synthetic rock standard was prepared in polyethylene containers (18 x 8 mm) and irradiated in the CT8 system to measure the combined effect of neutron-flux variation and the sample geometry. Another batch of synthetic rock standards was then prepared by the same procedure and sixteen 1 g replicates from that batch were used to measure the irradiation variation in the ELVIS.Counting geometry variations The spectrometry system at the reactor centre has a sample changer on which 12 samples can be counted automatically. Small plastic cups (51 mm diameter and 1 mm thick) are used for holding the sample containers over the detector. A slight variation (0.2 mm) in the thickness of these cups was found to contribute a 1% error to replicate variation, at 11 mm. The error was reduced by using those with minimum variation, and eliminated by counting samples at 60 mm from the detector. A second counting geometry error, which was reduced by counting samples at 60 mm, was due to the movement of the sample inside the container (i.e., the sample shape variation). This error was evaluated for those samples measured at 11 mm from the detector by counting a sample six times without shaking and six times with intermittent shaking. Evaluation of Sampling Constants When this work was started, only the core tube system, CT8, was available for epithermal-neutron activation, and CANMET had certified two reference ores: MA-1 and MA-2 (MA-lb and MA-2a were certified later, after their deple- tion).In order to avoid flux variation, two middle positions of CT8 were chosen and so a maximum of 12 capsules (18 X 8 mm) could be placed in them. Therefore, 12 replicates of four sample masses (1, 1.5,2 and 2.5 g) were prepared to carry out the replicate analyses. Replicates larger than 2.5 g could not be irradiated because the irradiation container (18 X 8 mm) only held a maximum of 2.5 g of powdered rock. All the samples were irradiated for 7.5 h.By the time MA-lb and MA-2a were certified, the ELVIS had also been installed. Therefore, two sample masses, 1 and 10 g, were chosen to carry out the analyses on MA-lb and MA-2a. The samples were prepared in containers measuring 18 x 8 mm and 22 X 35 mm, respectively. The capsules were sealed thermally to avoid leakage during pneumatic transfer. Sixteen 1 g replicates were irradiated for 1 h, and sixteen 10 g samples were irradiated for 15 min, sequentially.1348 ANALYST, AUGUST 1992, VOL. 117 Similarly, any variation in distance between sample and detector in a replicate analysis can contribute towards random error. Possible sources of such error are the variation in the thickness of the bases of the sample holders above the detector.Beckerg calculated that the sample shape variation was 3% for a 1 mm change in the average height of a sample at 5 cm distance from a 13% efficient detector. Similarly, Potts and Husseyg have calculated such variation resulting from reposi- tioning of a sample. They worked on a range of sample-to- detector distances of 5-50 mm and found that a repositioning discrepancy of only 0.2 mm for a source-to-detector distance of 10 mm can cause an error of up to 4% in the measured activity of a sample. Fortunately, in this work, all these errors could be eliminated by counting at large distances from the detector. Total Analytical Variation Taking account of all the analytical uncertainties, the total error for a sample i would be the sum of all absolute and relative random errors consisting of irradiation variation (iv), counting variations (cv) and the counting statistics (cs) error: (4) According to Heydorn and Damsgard,4 analytical variation, a2, becomes insignificant if it is less than one-third of the total variance (9).This can be checked conveniently by calculating the index of determination (ID): For an ID 20.8, the analytical variation is less than one-third of the total variance and can be ignored; therefore, eqn. (2) reduces to: or Experimental Blank Rock Standards Standard samples of similar matrices were needed to check the irradiation variation in this work. Therefore, a synthetic standard was prepared. A blank rock (basalt) of 200 mesh particle size was used as a base matrix because the samples were also siliceous ores.Blank ore, weighing 19 g, was doped with 19 g of a 13 mg kg-1 gold solution (HAuCI4). The mixture was blended on an orbital shaker and then frozen at -30 "C before drying in a vacuum-dryer. After drying, the synthetic standard was re-ground with a Tema swing mill in an agate pot. Finally, it was blended in an end-to-end shaker for 4 h. Reference Materials The reference ores of the Canadian Certified Reference Material Project, MA series, are typical of high and low (waste rock) ore grades from the Macassa Mine at Kirkland Lake, Ontario, Canada. The ore contains quartz, wall rock inclu- sions, carbonates, and small amounts of sulfides, tellurides and native gold. The principal sulfide is finely disseminated pyrite. Most of the gold occurs as an electrum.The high-grade ore, MA-1, was replaced by MA-la and then MA-lb, as they became depleted. The low-grade ore, MA-2, has been replaced by MA-2a. Irradiation Facilities All the samples in this work were irradiated in the CONSORT I1 reactor at Imperial College, London. It is a swimming pool-type reactor of thermal power 100 kW. A number of irradiation facilities are available, two of which were used: a manual epithermal system called CT8 and a new pneumatic epithermal large-volume irradiation system (ELVIS). Both these systems are lined with 1 mm thick cadmium and have epithermal-neutron fluxes of 2.8 x 1014 n m-2 s-' and fast-neutron fluxes of 2 x 1015 n m-2 s-1. Gamma-ray Spectrometry All the samples and standards were counted on a Ge(Li) detector [full width at half maximum (FWHM): 1.8 keV, peak/Compton ratio: 36.3 and efficiency: 8.l%, at 1.33 MeV].The gamma-ray spectra were measured with an ND6700 multichannel analyser. Results Evaluation of Analytical Errors Irradiation variations A set of twelve 1 g replicates of the synthetic rock standard was prepared in polyethylene containers (18 x 8 mm) and irradiated in the CT8 system to measure the combined effect of neutron-flux variation and the sample geometry. Another batch of synthetic rock standards was then prepared by the same procedure and sixteen 1 g replicates from that batch were used to measure the irradiation variation in the ELVIS. Counting geometry variations The spectrometry system at the reactor centre has a sample changer on which 12 samples can be counted automatically.Small plastic cups (51 mm diameter and 1 mm thick) are used for holding the sample containers over the detector. A slight variation (0.2 mm) in the thickness of these cups was found to contribute a 1% error to replicate variation, at 11 mm. The error was reduced by using those with minimum variation, and eliminated by counting samples at 60 mm from the detector. A second counting geometry error, which was reduced by counting samples at 60 mm, was due to the movement of the sample inside the container (i.e., the sample shape variation). This error was evaluated for those samples measured at 11 mm from the detector by counting a sample six times without shaking and six times with intermittent shaking. Evaluation of Sampling Constants When this work was started, only the core tube system, CT8, was available for epithermal-neutron activation, and CANMET had certified two reference ores: MA-1 and MA-2 (MA-lb and MA-2a were certified later, after their deple- tion). In order to avoid flux variation, two middle positions of CT8 were chosen and so a maximum of 12 capsules (18 X 8 mm) could be placed in them.Therefore, 12 replicates of four sample masses (1, 1.5,2 and 2.5 g) were prepared to carry out the replicate analyses. Replicates larger than 2.5 g could not be irradiated because the irradiation container (18 X 8 mm) only held a maximum of 2.5 g of powdered rock. All the samples were irradiated for 7.5 h. By the time MA-lb and MA-2a were certified, the ELVIS had also been installed. Therefore, two sample masses, 1 and 10 g, were chosen to carry out the analyses on MA-lb and MA-2a.The samples were prepared in containers measuring 18 x 8 mm and 22 X 35 mm, respectively. The capsules were sealed thermally to avoid leakage during pneumatic transfer. Sixteen 1 g replicates were irradiated for 1 h, and sixteen 10 g samples were irradiated for 15 min, sequentially.1348 ANALYST, AUGUST 1992, VOL. 117 Similarly, any variation in distance between sample and detector in a replicate analysis can contribute towards random error. Possible sources of such error are the variation in the thickness of the bases of the sample holders above the detector. Beckerg calculated that the sample shape variation was 3% for a 1 mm change in the average height of a sample at 5 cm distance from a 13% efficient detector.Similarly, Potts and Husseyg have calculated such variation resulting from reposi- tioning of a sample. They worked on a range of sample-to- detector distances of 5-50 mm and found that a repositioning discrepancy of only 0.2 mm for a source-to-detector distance of 10 mm can cause an error of up to 4% in the measured activity of a sample. Fortunately, in this work, all these errors could be eliminated by counting at large distances from the detector. Total Analytical Variation Taking account of all the analytical uncertainties, the total error for a sample i would be the sum of all absolute and relative random errors consisting of irradiation variation (iv), counting variations (cv) and the counting statistics (cs) error: (4) According to Heydorn and Damsgard,4 analytical variation, a2, becomes insignificant if it is less than one-third of the total variance (9).This can be checked conveniently by calculating the index of determination (ID): For an ID 20.8, the analytical variation is less than one-third of the total variance and can be ignored; therefore, eqn. (2) reduces to: or Experimental Blank Rock Standards Standard samples of similar matrices were needed to check the irradiation variation in this work. Therefore, a synthetic standard was prepared. A blank rock (basalt) of 200 mesh particle size was used as a base matrix because the samples were also siliceous ores. Blank ore, weighing 19 g, was doped with 19 g of a 13 mg kg-1 gold solution (HAuCI4). The mixture was blended on an orbital shaker and then frozen at -30 "C before drying in a vacuum-dryer.After drying, the synthetic standard was re-ground with a Tema swing mill in an agate pot. Finally, it was blended in an end-to-end shaker for 4 h. Reference Materials The reference ores of the Canadian Certified Reference Material Project, MA series, are typical of high and low (waste rock) ore grades from the Macassa Mine at Kirkland Lake, Ontario, Canada. The ore contains quartz, wall rock inclu- sions, carbonates, and small amounts of sulfides, tellurides and native gold. The principal sulfide is finely disseminated pyrite. Most of the gold occurs as an electrum. The high-grade ore, MA-1, was replaced by MA-la and then MA-lb, as they became depleted. The low-grade ore, MA-2, has been replaced by MA-2a. Irradiation Facilities All the samples in this work were irradiated in the CONSORT I1 reactor at Imperial College, London.It is a swimming pool-type reactor of thermal power 100 kW. A number of irradiation facilities are available, two of which were used: a manual epithermal system called CT8 and a new pneumatic epithermal large-volume irradiation system (ELVIS). Both these systems are lined with 1 mm thick cadmium and have epithermal-neutron fluxes of 2.8 x 1014 n m-2 s-' and fast-neutron fluxes of 2 x 1015 n m-2 s-1. Gamma-ray Spectrometry All the samples and standards were counted on a Ge(Li) detector [full width at half maximum (FWHM): 1.8 keV, peak/Compton ratio: 36.3 and efficiency: 8.l%, at 1.33 MeV]. The gamma-ray spectra were measured with an ND6700 multichannel analyser.Results Evaluation of Analytical Errors Irradiation variations A set of twelve 1 g replicates of the synthetic rock standard was prepared in polyethylene containers (18 x 8 mm) and irradiated in the CT8 system to measure the combined effect of neutron-flux variation and the sample geometry. Another batch of synthetic rock standards was then prepared by the same procedure and sixteen 1 g replicates from that batch were used to measure the irradiation variation in the ELVIS. Counting geometry variations The spectrometry system at the reactor centre has a sample changer on which 12 samples can be counted automatically. Small plastic cups (51 mm diameter and 1 mm thick) are used for holding the sample containers over the detector.A slight variation (0.2 mm) in the thickness of these cups was found to contribute a 1% error to replicate variation, at 11 mm. The error was reduced by using those with minimum variation, and eliminated by counting samples at 60 mm from the detector. A second counting geometry error, which was reduced by counting samples at 60 mm, was due to the movement of the sample inside the container (i.e., the sample shape variation). This error was evaluated for those samples measured at 11 mm from the detector by counting a sample six times without shaking and six times with intermittent shaking. Evaluation of Sampling Constants When this work was started, only the core tube system, CT8, was available for epithermal-neutron activation, and CANMET had certified two reference ores: MA-1 and MA-2 (MA-lb and MA-2a were certified later, after their deple- tion).In order to avoid flux variation, two middle positions of CT8 were chosen and so a maximum of 12 capsules (18 X 8 mm) could be placed in them. Therefore, 12 replicates of four sample masses (1, 1.5,2 and 2.5 g) were prepared to carry out the replicate analyses. Replicates larger than 2.5 g could not be irradiated because the irradiation container (18 X 8 mm) only held a maximum of 2.5 g of powdered rock. All the samples were irradiated for 7.5 h. By the time MA-lb and MA-2a were certified, the ELVIS had also been installed. Therefore, two sample masses, 1 and 10 g, were chosen to carry out the analyses on MA-lb and MA-2a. The samples were prepared in containers measuring 18 x 8 mm and 22 X 35 mm, respectively.The capsules were sealed thermally to avoid leakage during pneumatic transfer. Sixteen 1 g replicates were irradiated for 1 h, and sixteen 10 g samples were irradiated for 15 min, sequentially.1348 ANALYST, AUGUST 1992, VOL. 117 Similarly, any variation in distance between sample and detector in a replicate analysis can contribute towards random error. Possible sources of such error are the variation in the thickness of the bases of the sample holders above the detector. Beckerg calculated that the sample shape variation was 3% for a 1 mm change in the average height of a sample at 5 cm distance from a 13% efficient detector. Similarly, Potts and Husseyg have calculated such variation resulting from reposi- tioning of a sample.They worked on a range of sample-to- detector distances of 5-50 mm and found that a repositioning discrepancy of only 0.2 mm for a source-to-detector distance of 10 mm can cause an error of up to 4% in the measured activity of a sample. Fortunately, in this work, all these errors could be eliminated by counting at large distances from the detector. Total Analytical Variation Taking account of all the analytical uncertainties, the total error for a sample i would be the sum of all absolute and relative random errors consisting of irradiation variation (iv), counting variations (cv) and the counting statistics (cs) error: (4) According to Heydorn and Damsgard,4 analytical variation, a2, becomes insignificant if it is less than one-third of the total variance (9).This can be checked conveniently by calculating the index of determination (ID): For an ID 20.8, the analytical variation is less than one-third of the total variance and can be ignored; therefore, eqn. (2) reduces to: or Experimental Blank Rock Standards Standard samples of similar matrices were needed to check the irradiation variation in this work. Therefore, a synthetic standard was prepared. A blank rock (basalt) of 200 mesh particle size was used as a base matrix because the samples were also siliceous ores. Blank ore, weighing 19 g, was doped with 19 g of a 13 mg kg-1 gold solution (HAuCI4). The mixture was blended on an orbital shaker and then frozen at -30 "C before drying in a vacuum-dryer. After drying, the synthetic standard was re-ground with a Tema swing mill in an agate pot.Finally, it was blended in an end-to-end shaker for 4 h. Reference Materials The reference ores of the Canadian Certified Reference Material Project, MA series, are typical of high and low (waste rock) ore grades from the Macassa Mine at Kirkland Lake, Ontario, Canada. The ore contains quartz, wall rock inclu- sions, carbonates, and small amounts of sulfides, tellurides and native gold. The principal sulfide is finely disseminated pyrite. Most of the gold occurs as an electrum. The high-grade ore, MA-1, was replaced by MA-la and then MA-lb, as they became depleted. The low-grade ore, MA-2, has been replaced by MA-2a. Irradiation Facilities All the samples in this work were irradiated in the CONSORT I1 reactor at Imperial College, London.It is a swimming pool-type reactor of thermal power 100 kW. A number of irradiation facilities are available, two of which were used: a manual epithermal system called CT8 and a new pneumatic epithermal large-volume irradiation system (ELVIS). Both these systems are lined with 1 mm thick cadmium and have epithermal-neutron fluxes of 2.8 x 1014 n m-2 s-' and fast-neutron fluxes of 2 x 1015 n m-2 s-1. Gamma-ray Spectrometry All the samples and standards were counted on a Ge(Li) detector [full width at half maximum (FWHM): 1.8 keV, peak/Compton ratio: 36.3 and efficiency: 8.l%, at 1.33 MeV]. The gamma-ray spectra were measured with an ND6700 multichannel analyser. Results Evaluation of Analytical Errors Irradiation variations A set of twelve 1 g replicates of the synthetic rock standard was prepared in polyethylene containers (18 x 8 mm) and irradiated in the CT8 system to measure the combined effect of neutron-flux variation and the sample geometry.Another batch of synthetic rock standards was then prepared by the same procedure and sixteen 1 g replicates from that batch were used to measure the irradiation variation in the ELVIS. Counting geometry variations The spectrometry system at the reactor centre has a sample changer on which 12 samples can be counted automatically. Small plastic cups (51 mm diameter and 1 mm thick) are used for holding the sample containers over the detector. A slight variation (0.2 mm) in the thickness of these cups was found to contribute a 1% error to replicate variation, at 11 mm.The error was reduced by using those with minimum variation, and eliminated by counting samples at 60 mm from the detector. A second counting geometry error, which was reduced by counting samples at 60 mm, was due to the movement of the sample inside the container (i.e., the sample shape variation). This error was evaluated for those samples measured at 11 mm from the detector by counting a sample six times without shaking and six times with intermittent shaking. Evaluation of Sampling Constants When this work was started, only the core tube system, CT8, was available for epithermal-neutron activation, and CANMET had certified two reference ores: MA-1 and MA-2 (MA-lb and MA-2a were certified later, after their deple- tion). In order to avoid flux variation, two middle positions of CT8 were chosen and so a maximum of 12 capsules (18 X 8 mm) could be placed in them.Therefore, 12 replicates of four sample masses (1, 1.5,2 and 2.5 g) were prepared to carry out the replicate analyses. Replicates larger than 2.5 g could not be irradiated because the irradiation container (18 X 8 mm) only held a maximum of 2.5 g of powdered rock. All the samples were irradiated for 7.5 h. By the time MA-lb and MA-2a were certified, the ELVIS had also been installed. Therefore, two sample masses, 1 and 10 g, were chosen to carry out the analyses on MA-lb and MA-2a. The samples were prepared in containers measuring 18 x 8 mm and 22 X 35 mm, respectively. The capsules were sealed thermally to avoid leakage during pneumatic transfer. Sixteen 1 g replicates were irradiated for 1 h, and sixteen 10 g samples were irradiated for 15 min, sequentially.1348 ANALYST, AUGUST 1992, VOL.117 Similarly, any variation in distance between sample and detector in a replicate analysis can contribute towards random error. Possible sources of such error are the variation in the thickness of the bases of the sample holders above the detector. Beckerg calculated that the sample shape variation was 3% for a 1 mm change in the average height of a sample at 5 cm distance from a 13% efficient detector. Similarly, Potts and Husseyg have calculated such variation resulting from reposi- tioning of a sample. They worked on a range of sample-to- detector distances of 5-50 mm and found that a repositioning discrepancy of only 0.2 mm for a source-to-detector distance of 10 mm can cause an error of up to 4% in the measured activity of a sample.Fortunately, in this work, all these errors could be eliminated by counting at large distances from the detector. Total Analytical Variation Taking account of all the analytical uncertainties, the total error for a sample i would be the sum of all absolute and relative random errors consisting of irradiation variation (iv), counting variations (cv) and the counting statistics (cs) error: (4) According to Heydorn and Damsgard,4 analytical variation, a2, becomes insignificant if it is less than one-third of the total variance (9). This can be checked conveniently by calculating the index of determination (ID): For an ID 20.8, the analytical variation is less than one-third of the total variance and can be ignored; therefore, eqn.(2) reduces to: or Experimental Blank Rock Standards Standard samples of similar matrices were needed to check the irradiation variation in this work. Therefore, a synthetic standard was prepared. A blank rock (basalt) of 200 mesh particle size was used as a base matrix because the samples were also siliceous ores. Blank ore, weighing 19 g, was doped with 19 g of a 13 mg kg-1 gold solution (HAuCI4). The mixture was blended on an orbital shaker and then frozen at -30 "C before drying in a vacuum-dryer. After drying, the synthetic standard was re-ground with a Tema swing mill in an agate pot. Finally, it was blended in an end-to-end shaker for 4 h. Reference Materials The reference ores of the Canadian Certified Reference Material Project, MA series, are typical of high and low (waste rock) ore grades from the Macassa Mine at Kirkland Lake, Ontario, Canada.The ore contains quartz, wall rock inclu- sions, carbonates, and small amounts of sulfides, tellurides and native gold. The principal sulfide is finely disseminated pyrite. Most of the gold occurs as an electrum. The high-grade ore, MA-1, was replaced by MA-la and then MA-lb, as they became depleted. The low-grade ore, MA-2, has been replaced by MA-2a. Irradiation Facilities All the samples in this work were irradiated in the CONSORT I1 reactor at Imperial College, London. It is a swimming pool-type reactor of thermal power 100 kW. A number of irradiation facilities are available, two of which were used: a manual epithermal system called CT8 and a new pneumatic epithermal large-volume irradiation system (ELVIS).Both these systems are lined with 1 mm thick cadmium and have epithermal-neutron fluxes of 2.8 x 1014 n m-2 s-' and fast-neutron fluxes of 2 x 1015 n m-2 s-1. Gamma-ray Spectrometry All the samples and standards were counted on a Ge(Li) detector [full width at half maximum (FWHM): 1.8 keV, peak/Compton ratio: 36.3 and efficiency: 8.l%, at 1.33 MeV]. The gamma-ray spectra were measured with an ND6700 multichannel analyser. Results Evaluation of Analytical Errors Irradiation variations A set of twelve 1 g replicates of the synthetic rock standard was prepared in polyethylene containers (18 x 8 mm) and irradiated in the CT8 system to measure the combined effect of neutron-flux variation and the sample geometry.Another batch of synthetic rock standards was then prepared by the same procedure and sixteen 1 g replicates from that batch were used to measure the irradiation variation in the ELVIS. Counting geometry variations The spectrometry system at the reactor centre has a sample changer on which 12 samples can be counted automatically. Small plastic cups (51 mm diameter and 1 mm thick) are used for holding the sample containers over the detector. A slight variation (0.2 mm) in the thickness of these cups was found to contribute a 1% error to replicate variation, at 11 mm. The error was reduced by using those with minimum variation, and eliminated by counting samples at 60 mm from the detector. A second counting geometry error, which was reduced by counting samples at 60 mm, was due to the movement of the sample inside the container (i.e., the sample shape variation).This error was evaluated for those samples measured at 11 mm from the detector by counting a sample six times without shaking and six times with intermittent shaking. Evaluation of Sampling Constants When this work was started, only the core tube system, CT8, was available for epithermal-neutron activation, and CANMET had certified two reference ores: MA-1 and MA-2 (MA-lb and MA-2a were certified later, after their deple- tion). In order to avoid flux variation, two middle positions of CT8 were chosen and so a maximum of 12 capsules (18 X 8 mm) could be placed in them. Therefore, 12 replicates of four sample masses (1, 1.5,2 and 2.5 g) were prepared to carry out the replicate analyses.Replicates larger than 2.5 g could not be irradiated because the irradiation container (18 X 8 mm) only held a maximum of 2.5 g of powdered rock. All the samples were irradiated for 7.5 h. By the time MA-lb and MA-2a were certified, the ELVIS had also been installed. Therefore, two sample masses, 1 and 10 g, were chosen to carry out the analyses on MA-lb and MA-2a. The samples were prepared in containers measuring 18 x 8 mm and 22 X 35 mm, respectively. The capsules were sealed thermally to avoid leakage during pneumatic transfer. Sixteen 1 g replicates were irradiated for 1 h, and sixteen 10 g samples were irradiated for 15 min, sequentially.1348 ANALYST, AUGUST 1992, VOL.117 Similarly, any variation in distance between sample and detector in a replicate analysis can contribute towards random error. Possible sources of such error are the variation in the thickness of the bases of the sample holders above the detector. Beckerg calculated that the sample shape variation was 3% for a 1 mm change in the average height of a sample at 5 cm distance from a 13% efficient detector. Similarly, Potts and Husseyg have calculated such variation resulting from reposi- tioning of a sample. They worked on a range of sample-to- detector distances of 5-50 mm and found that a repositioning discrepancy of only 0.2 mm for a source-to-detector distance of 10 mm can cause an error of up to 4% in the measured activity of a sample. Fortunately, in this work, all these errors could be eliminated by counting at large distances from the detector.Total Analytical Variation Taking account of all the analytical uncertainties, the total error for a sample i would be the sum of all absolute and relative random errors consisting of irradiation variation (iv), counting variations (cv) and the counting statistics (cs) error: (4) According to Heydorn and Damsgard,4 analytical variation, a2, becomes insignificant if it is less than one-third of the total variance (9). This can be checked conveniently by calculating the index of determination (ID): For an ID 20.8, the analytical variation is less than one-third of the total variance and can be ignored; therefore, eqn. (2) reduces to: or Experimental Blank Rock Standards Standard samples of similar matrices were needed to check the irradiation variation in this work.Therefore, a synthetic standard was prepared. A blank rock (basalt) of 200 mesh particle size was used as a base matrix because the samples were also siliceous ores. Blank ore, weighing 19 g, was doped with 19 g of a 13 mg kg-1 gold solution (HAuCI4). The mixture was blended on an orbital shaker and then frozen at -30 "C before drying in a vacuum-dryer. After drying, the synthetic standard was re-ground with a Tema swing mill in an agate pot. Finally, it was blended in an end-to-end shaker for 4 h. Reference Materials The reference ores of the Canadian Certified Reference Material Project, MA series, are typical of high and low (waste rock) ore grades from the Macassa Mine at Kirkland Lake, Ontario, Canada.The ore contains quartz, wall rock inclu- sions, carbonates, and small amounts of sulfides, tellurides and native gold. The principal sulfide is finely disseminated pyrite. Most of the gold occurs as an electrum. The high-grade ore, MA-1, was replaced by MA-la and then MA-lb, as they became depleted. The low-grade ore, MA-2, has been replaced by MA-2a. Irradiation Facilities All the samples in this work were irradiated in the CONSORT I1 reactor at Imperial College, London. It is a swimming pool-type reactor of thermal power 100 kW. A number of irradiation facilities are available, two of which were used: a manual epithermal system called CT8 and a new pneumatic epithermal large-volume irradiation system (ELVIS). Both these systems are lined with 1 mm thick cadmium and have epithermal-neutron fluxes of 2.8 x 1014 n m-2 s-' and fast-neutron fluxes of 2 x 1015 n m-2 s-1.Gamma-ray Spectrometry All the samples and standards were counted on a Ge(Li) detector [full width at half maximum (FWHM): 1.8 keV, peak/Compton ratio: 36.3 and efficiency: 8.l%, at 1.33 MeV]. The gamma-ray spectra were measured with an ND6700 multichannel analyser. Results Evaluation of Analytical Errors Irradiation variations A set of twelve 1 g replicates of the synthetic rock standard was prepared in polyethylene containers (18 x 8 mm) and irradiated in the CT8 system to measure the combined effect of neutron-flux variation and the sample geometry. Another batch of synthetic rock standards was then prepared by the same procedure and sixteen 1 g replicates from that batch were used to measure the irradiation variation in the ELVIS. Counting geometry variations The spectrometry system at the reactor centre has a sample changer on which 12 samples can be counted automatically. Small plastic cups (51 mm diameter and 1 mm thick) are used for holding the sample containers over the detector. A slight variation (0.2 mm) in the thickness of these cups was found to contribute a 1% error to replicate variation, at 11 mm. The error was reduced by using those with minimum variation, and eliminated by counting samples at 60 mm from the detector. A second counting geometry error, which was reduced by counting samples at 60 mm, was due to the movement of the sample inside the container (i.e., the sample shape variation). This error was evaluated for those samples measured at 11 mm from the detector by counting a sample six times without shaking and six times with intermittent shaking. Evaluation of Sampling Constants When this work was started, only the core tube system, CT8, was available for epithermal-neutron activation, and CANMET had certified two reference ores: MA-1 and MA-2 (MA-lb and MA-2a were certified later, after their deple- tion). In order to avoid flux variation, two middle positions of CT8 were chosen and so a maximum of 12 capsules (18 X 8 mm) could be placed in them. Therefore, 12 replicates of four sample masses (1, 1.5,2 and 2.5 g) were prepared to carry out the replicate analyses. Replicates larger than 2.5 g could not be irradiated because the irradiation container (18 X 8 mm) only held a maximum of 2.5 g of powdered rock. All the samples were irradiated for 7.5 h. By the time MA-lb and MA-2a were certified, the ELVIS had also been installed. Therefore, two sample masses, 1 and 10 g, were chosen to carry out the analyses on MA-lb and MA-2a. The samples were prepared in containers measuring 18 x 8 mm and 22 X 35 mm, respectively. The capsules were sealed thermally to avoid leakage during pneumatic transfer. Sixteen 1 g replicates were irradiated for 1 h, and sixteen 10 g samples were irradiated for 15 min, sequentially.