COMMENTS ON astro-ph/0211534
This paper by Paolo Walter Cattaneo claims to reanalyze the results of the Homestake chlorine solar neutrino detector that were published in ApJ 496, 505 (1998) by using the table of individual run results that were attached to that paper. In this paper we clearly stated that the quoted results of the experiment were not based on the values provided in this table but rather on the combined data set of all 108 runs. The main reason for providing the table was to maintain consistency with previous conference and other reports that we produced.
There are a number of problems with using the single run results to obtain the final experimental result. Each run has only a small number of 37Ar atoms and so a small number of counts which are then fitted to the assumption of a two component time distribution, one that decays with the 37Ar lifetime, the signal, and another that is constant in time, the background. In order to obtain the total number of 37Ar decays, it is necessary to subtract the background rate, determined from the late part of the counting date from the early part of the counting data. Statistical fluctuations in the background can occasionally result in a subtraction in the early time region where the 37Ar decays lie that is greater than the total signal in this region. Such a subtraction results in a negative 37Ar signal. Since there is no physical reality in a negative signal, we arbitrarily set such runs to zero. This procedure produced an unusually large number of zero signal runs. Since Dr. Cattaneo does not mention anything specific about these runs, we presume that he used them as stated rather than assigning negative values to these runs. Had he treated these 11 or so runs correctly and assigned “negative” fluxes to them, his overall solar neutrino signal would have been lower. Of course, he could not do so since he did not know what negative value to assign.
Next, Dr. Cattaneo rejects runs with high backgrounds. We are well aware of these runs. On page 517, Section 7.1, of our ApJ paper we discuss 15 runs that require special treatment. A substantial fraction of these runs involved counting of the 37Ar decays on the surface of the earth. Seven of these were taken at an early stage of the experiment, with an earlier version of our counter readout electronics that had a smaller dynamic range. In this section we also point out that there are eight runs with backgrounds that are several times larger than those in the rest of the data set. As we indicated in that section of our paper, we first analyzed the remaining 93 runs (108 total runs – 15 “special treatment” runs) that had similar low backgrounds and were counted underground and then appropriately combined these additional 15 runs with the 93 runs. There was no statistical difference between the 93 run subset and the full data set.
The counting system was moved to the deep underground laboratory in 1977. During the next several years, the period between 1977 and 1982, a number of new counters were taken underground. As we described in our paper, the iron cathodes of these counters were slightly activated by cosmic rays while they were on the surface of the earth. The relevant activity, 55Fe, decayed away with a 2.7 year lifetime. To the best of our knowledge, this is the dominant source of the background in the 93 runs that were counted underground. The amplitude of this background for a given run depended on the length of time that the counter that was used had been underground. We provided neither the counter identification for each of the individual runs nor the history of that counter in the table since we did not anticipate that these data would be reanalyzed from the table without some consultation.
In addition to the above, there is also a very small, time independent background in the runs that were counted underground. We presume that this background is due to remnant background in the local counting apparatus that eludes our shield and our veto counter. The background of the post 1985 data, mainly from this source, is generally close to zero per run. This is the data region preferred by Dr. Cattaneo. We are not aware of any change of procedure for extraction, gas handling or counting in this time period. All of the final analysis of data, the results of which were presented in our ApJ paper, was done in the mid-1990’s. We treated all data time intervals in the same way.
There are only two ways in which we could get an artificially low 37Ar rate, either the 37Ar was never transferred from the detector into the counter or we incorrectly applied the energy and rise time selection cuts. In order to insure that the 37Ar was efficiently transferred we added some argon carrier gas, either 36Ar or 38Ar, to the detector before extraction. At the end of the one year counting period, the gas in the counter was analysed in a mass spectrograph and the overall gas extraction and transfer efficiency was determined. These data, which are presented on a run by run basis in our paper, indicate a transfer efficiency of about 95%, see figs 4 and 5 in our ApJ paper. The final results for each run were corrected for this efficiency. There is no indication of low efficiency for high background runs ruling out the possibility that 37Ar atoms were lost in these runs.
The second possibility is that drifts in the proportional counters of readout electronics resulted in valid 37Ar decays being rejected. A set of test pulses were sent through the readout electronics immediately after each counter pulse to check the gain and response of the electronic system. In addition, we calibrated each counter with an external 55Fe source every one or two months to check the counter response. There was a small change in counter gain during the first several days after filling due to the leakage of helium out of the counter. This effect was modeled and corrections were applied to the data.
For a single run with a small number of 37Ar atoms, the shift of a single event either into or out of the acceptance region can have a significant effect on the results of that run. In order to avoid this possibility, we combined all the runs together into a single data set and then studied the resultant rate as a function of the widths of the energy and rise time windows. We found no variation of 37Ar production rate as a function of window width ruling out the possibility that we lost events due to drifts in the electronics. These results were also presented in figs. 11 and 12 in our ApJ paper.
As Dr. Cattaneo correctly points out, most of the high background rate runs were in the first half of the data set. We have explained why above. We looked for long term variation of rate by dividing the data set into three time intervals, each about 7 years in duration and subjecting each subset to the same data analysis. The 37Ar production rate in each of these intervals was the same within the statistical uncertainty.
Dr. Cattaneo finally presents the results of a subset of 70 out of 108 runs. He assigns the same statistical and systematic errors to this subset as we have for the larger set. I find this a bit surprising since a smaller data set usually has a larger statistical error.
Dr. Cattaneo complains that we did not provide sufficient details of our data. There is no reasonable way to print all the data and calibration signals. As is the case with most present day experiments, the complete data set is rather massive. We feel we have provided a rather complete description of the data analysis procedure and given plots of the results at various intervals of the analysis. Page limit considerations limited the amount of material that we could present.
In section 2, Dr. Cattaneo explains why a series of previous papers, three of which were coauthored by John Bahcall, were all incorrect. He then proceeds to follow a “hint” in one of these Bahcall papers to “understand” the counting background. While John Bahcall is the leading expert on the solar model, is very well acquainted with our experiment and is a good friend and colleague, he is not a member of our experimental group and thus has participated in neither the experiment nor the data analysis. We would prefer that the author rely on our explanation of the background and not use a secondary source. We are not responsible for what others write. The use of secondary sources when the primary source is available should be discouraged. It merely leads to the propagation of misunderstanding of experimental details and data analysis.
In summary, I must recommend that this paper be rejected. It claims to critically review the results of the chlorine experiment, but does not. Rather, it misanalyses the wrong data set and arbitrarily rejects subsets of the data without any physical explanation for the rejection criteria and even misstates the experimental results of the low background subset of the data.
Sincerely,
Kenneth Lande