Designing experiments

In the previous section, a sample was dated using an experiment. The purpose of experiments in ACE is to test the effect of theory on calculated ages.  For example the sensitivity of calculated ages to muon contributions can be determined by making an experiment with no muon contributions, another with muon contributions, and comparing calculated ages calculating using both.  To make a new experiment we begin with the experiment browser, which can be invoked by clicking on the ‘Make Experiment’ button at the bottom of the sample browser.

Click on image to expand

Click on image to expand

The experiment browser lists all experiments that have been created, nuclide by nuclide and sorted into calibrated and uncalibrated experiments.  To see the details of a particular experiment just select it.  Here are the details for the 10Be Demonstration Experiment:

Click on image to expand

Click on image to expand

From this view we can see that the 10Be Demonstration Experiment uses Guyodo and Valet (1999) geomagnetic intensity changes, and has a calibration HLSL spallation production rate of 5.07 ± 0.33 atoms per grams per year.

So how would ages of 10Be sample differ if we had used a different record of geomagnetic intensity?  To answer this question we need to create a new experiment with a different record.  To do this we click in the ‘Create Experiment‘ button at the bottom left hand corner of the experiment browser.  A new window opens:

In this window we:

  • Choose which nuclide we want to work with (10Be, from the pull down menu)
  • Give the experiment a name (’10Be New Intensity’)
  • Choose a timestep (10 yrs, same as for the demonstration experiment).

This is step 1 of 6.  To continue we use the ‘Next’ Button, or select the next panel at the top named ‘Workflows’.

Click on image to expand

Click on image to expand

The workflow is the process through which the sample is passed.  Calibration is a workflow.  Dating is a workflow.  Ultimately ACE is expandable so that any relevant process (the calculation of steady state erosion ages, the comparison of scaling differences) can be thought of as a workflow.  In this panel there is only one option, which is ‘Non 36Cl Time Stepping Calibration’.  This workflow will calculate HLSL production rates.  When this workflow has been selected we continue by selecting ‘Next’, to the panel named ‘Calibration Set':

The calibration set selects the group of independently dated samples from which to estimate HLSL production rates.  New calibration sets can be imported into ACE just like normal samples, but they need the additional attributes ‘independent age’ and ‘independent age uncertainty’. New and existing calibration sets can viewed using the calibration set browser, and sets can be regrouped using the group editor. The original data sets are given as csv files on the ACE website (see here for 10Be), allowing for easy access.  We choose the Balco et al (2008) combination calibration set here, as the same set was used in the demonstration experiment, and we only want to change geomagnetic intensity. When this calibration set is selected, we continue by selecting ‘Next’, to the panel named ‘Factors’.

Factors refer to changeable components in the calibration/dating workflow and come from the scientific literature.  By default the factors are choice of scaling model, sea level, goemagnetic latitude and geomagnetic intensity.  The short labels here are more fully documented in the Data Collections section of the website, and users should consult the literature if they are unsure which choices to use. We want the same set of factors as for the demonstration experiment, with the exception of Geomagnetic Intensity, where we choose the combined Gyodo and Valet (1999)/Yamazaki and Oda (2005) paleointensity record.  When all choices have been made, we continue by selecting ‘Next’, to the panel named ‘Parameters':

Click on image to expand

Click on image to expand

The Parameters section is for miscellaneous variables in the code to be changed, and contains controls for the component of production resulting from fast and slow muons.  For nuclides other than 36Cl these are given in percentages, for 36Cl these are given in muons cm-2 yr-1.  By clicking ‘Next’, we are presented with a summary of our experiment and are ready to perform a calibration:

If all parameters correct, the ‘Create Experiment‘ dialog button will save this experiment as an uncalibrated experiment (10Be New Intensity) in the experiment browser:

This is useful if you need to design multiple experiments, as calibrating experiments can take some time. The experiment can then be calibrated to calculate HLSL production rates by selecting it and then clicking the ‘Calibrate Experiment‘ button.  When this button is pressed a window opens  showing the progress of the calibration:

The progress bar ends at the age of the oldest sample in the calibration database, if the progress is too slow we recommend you increase the time step (in the ‘Nuclide’ panel). The calibration computes the component of cosmogenic production which when multiplied by HLSL production rates would give the sample inventories, ie the sample dependent component of production.  In this manner the HLSL production rates can be estimated via a linear regression between sample inventory and non-HLSL production component.  The results of this regression are shown at the end of the calibration:

Click on image to expand

Click on image to expand

The x axis of this plot shows the observed inventory of each sample from the calibration dataset (minus any contribution from muons, which are hard to calibrate for), and the y axis shows the predicted inventory for this calibration.  The green line shows y = x, and if the samples are not reasonably close to this line something has gone wrong and you need to look more closely at your experiment choices. Errorbars show 1-sigma uncertainties in the sample inventory (x direction) and uncertainty in the predicted inventory (y direction), usually due to uncertainties in the independent age of the sample.  The regression is weighted by both of these uncertainties, so samples with large errorbars will likely by further away from y=x.  If you are happy with these result and would like to use this experiment to date other samples click ‘Yes’ at the base of this panel.  Then the experiment ’10Be New Intensity’ will appear besides the other calibrated experiments in the experiment browser:

Click on image to expand

Click on image to expand

By comparing this to our 10Be Demonstration experiment (by switching between them in the browser) two things are apparent.  First when there is no production due to muons HLSL spallation production rates are higher (5.25 ± 0.33 a g-1 yr-1 Default, 5.07 ± 0.33 a g-1 yr-1).  Second the match of observed inventory to predicted inventory when the Yamazaki and Oda (2005) record is used is slightly worse (Reduced chi-square 3.19 Default, 3.04 Yamazaki and Oda 2005). This experiment is now ready to be used in date samples in sample browser, here is a comparison of ages calculated for the samples supplied with ACE:

Click on image to expand

Click on image to expand

So from the above we see that the 10Be samples supplied with ACE are sensitive to the record of geomagnetic paleointensity changes to a few hundred years.

To watch a demonstration of this procedure click here (requires Quicktime)

Leave a Reply

You must be logged in to post a comment.