X-ray secondary-emission spectrometry or XRF spectrometry is a nondestructive method of qualitative and quantitative analysis for elemental composition of samples. XRF analysis is based on the measurement of wavelengths and intensities of X-ray spectral lines emitted by secondary excitation.
An ARL 8420 XRF spectrometer was installed in the X-ray Laboratory on the JOIDES Resolution prior to the Leg 102 transit to Portugal. During ODP, scientists on board studied ~4000 samples from 56 legs before the spectrometer was removed from the ship after Leg 189.
The ARL 8420 was a fully automated, wavelength-dispersive spectrometer using a 3-kWh rhodium X-ray tube as the excitation for both major oxides and trace elements. Scientists and technicians on Legs 109 and 111 developed the first set of ODP XRF standard operating procedures. Few changes were made to the XRF hardware or the standard operation procedures during the period that the XRF system was on board. Table T36 is a summary of XRF data operations during ODP.
In Table T37, the major oxides and trace elements routinely collected with the XRF are listed with an example of the operating parameters of the system (Leg 125). Parameters such as Line, Crystal, and Detector did not change except when the krypton detector was replaced by a flow-proportional counter (FPC) on Leg 136. The Collimator setup should not have changed much; however, some legs designated the settings as "fine" and "coarse," whereas others designated the settings as "fine" and "medium." Peak Angle and Background Offset changed slightly with each calibration. Background angles were adjusted to avoid overlap from other elements.
Most samples analyzed by the ODP XRF were first powdered because of the inhomogeneity of rock and sediment samples. The powder was processed into fused glass disks to be run for major oxide analyses, and pressed pellets to be run for trace element analyses. The descriptions of sample preparation varied some between legs. The descriptions below are representative of the general procedures.
Samples, ~10 cm3 of rock, had saw marks and unwanted material removed by wet-grinding on a silicon carbide disk mill. The samples were ultrasonically washed in distilled water and then methanol for 10 min and dried at 110°C for at least 2 hr. Larger pieces were reduced to <1 cm diameter by crushing between two plastic disks in a hydraulic press. Powders were produced by grinding pieces <1 cm in diameter in a Spex Shatterbox using a tungsten carbide grinding vessel for 60–120 s, depending on the size of the vessel. The powder was transferred to clean paper and then to a sample vial and labeled. The vessel was thoroughly cleaned and prepared for another sample.
Samples were treated like hard rock samples except that the oven drying was replaced by freeze-drying for at least 12 hr. After the sample was dried, it was ground and treated in the same manner as the hard rock sample.
Unlithified sediment samples were problematic. Sometimes the mud was washed to remove chloride and then freeze-dried. Other times the mud was only freeze-dried because it was believed that washing would remove other elements, not just the chloride contamination. After the sample was dried, it was ground and treated in the same manner as the hard rock sample.
Approximately 1.5 g of the rock powder was carefully weighed and ignited in an ash furnace for at least 5 hr at 1000°C for hard rock and 900°C for sediments. If the sample likely contained muscovite, biotite, amphibole, or carbonates, the sample was ignited for at least 6 hr. Because the powders were dried before ignition, the loss values resulting from the amount of adsorbed water (H2O–) were assumed to be negligible.
Fused glass disks were created for major oxide analysis in order to reduce matrix effects and variations in background (Claisse, 1956; Rose et al., 1962; Norrish and Hutton, 1969). These disks were made by mixing 7.20 g (20% La2O3) lithium tetraborate flux with 0.600 g ignited rock powder. This sample/flux mixture was melted at 1030°C in Pt-Au crucibles for 6–10 min and poured into Pt-Au molds using a modified Claisse Fluxer apparatus. The 12:1 flux to sample ratio had been found to sufficiently reduce matrix effects to the point where matrix corrections were unnecessary for normal basaltic to granitic composition ranges.
Pressed pellets were used for the trace element analyses. Pressed-powder pellets were made by mixing 7 g fresh rock powder with 30 drops polyvinyl alcohol binder and pressing the mixture into an aluminum cap with 7 tons of pressure. A minimum of 5 g of sample usually guaranteed the pellet would be "infinitely thick" for rhodium K-series radiation.
To compute trace element concentrations from measured X-ray intensities, an offline calculation program was written by J.W. Sparks based on routines modified from Norrish and Chappell (1977) and Reynolds (1967). After the computer and software upgrade on Leg 149, trace element calculations were done by the data acquisition software.
Most XRF analytical methods are based on comparison of the unknown sample's line intensities to one or more well-characterized standards. The standards must be similar to the samples in physical form, elemental concentrations, and matrix composition. The calibration for a leg was performed using standards chosen to best provide the range of values that were expected from the cored rock. Table T38 lists the standards most often used on the JOIDES Resolution. These standards were chosen because of their similarity to oceanic basalts, granites, and ultramafic rocks.
Oceanic sediments are harder to characterize because of the variety of materials that make up the sedimentary column (e.g., organic material, calcium carbonate, clay, or silica). Developing standards that would span the expected range of values was very difficult.
Major oxide and trace element data from XRF analyses were logged on log sheets. The log sheets were brought back to ODP/TAMU at the end of each leg, entered into the S1032 database, and microfilmed for archival purposes. The S1032 database was used to store data until Leg 134. Starting with Leg 134, the data were saved in files which were brought back at the end of each leg and archived on the ODP/TAMU servers.
The data model for XRF data can be found in "Janus XRF Data Model" in "Appendix R." Included are the relational diagram and the list of the tables that contain data pertinent to XRF, the column names, and the definition of each column attribute. ODP Information Services Database Group was responsible for the migration of pre-Leg 171 data to Janus. An additional column of information was added to the XRF_SAMPLE table after the migration of data. This column, Sample_type, was added to allow scientists to describe the type of rock that was analyzed. In turn, scientists would be able to extract data based on specific rock types. The sediment or rock designations were not usually stored in the original data files, and it was necessary to extract that information from the Initial Reports volumes. Due to constraints of time, it was not possible to complete this part of the migration.
XRF major oxide and trace element data can be retrieved from Janus Web using a predefined query. The XRF query Web page allows the user to extract data using the following variables to restrict the amount of data retrieved: leg, site, hole, core, section, depth range or latitude and longitude range. Often, replicate samples were analyzed for the major oxides. Those data, when available, were uploaded separately. The Web query reports the replicate data on separate lines. In addition, the trace element data will also be reported on a separate line.
Table T39 lists the data fields retrieved from the Janus database for the XRF predefined query. The first column contains the data item, the second column indicates the Janus table or tables in which the data were stored, and the third column is the Janus column name or the calculation used to produce the value. "Description of Data Items from XRF Query" in "Appendix R" contains additional information about the fields retrieved using the Janus Web XRF query and the data format for the archived ASCII files.
The XRF data were very high quality, even though the ship environment made getting accurate measurements more difficult. XRF analytical methods are based on comparison of the unknown sample's line intensities to one or more well-characterized standards. For this reason, XRF analyses done on volcanic, mafic and ultramafic samples are the highest quality because there are standards that characterize the range of elemental concentrations most often found in these types of rocks. It is more difficult to create standards that would encompass the range of elemental concentrations found in sedimentary environments. Migration of the sample type data was completed for 20 of the 56 legs.