INSTRUMENT PROBLEMS

Heater Event Problem:

The response of this telescope was unusual in that:

1. The HE mass measurement, M1, showed a relatively large energy and time dependence,

2. A relatively large 'background' of events was present in the charge range near fluorine,

3. The L1 versus L2+L3 element tracks had an uncharacteristically blurred appearance.

The problem may be electronic instability, but was not investigated in detail; instead the data from this telescope were excluded from use in the measurements presented in Chapter 4, Tables 4.1 through 4.12.

V2 LET C Problem:

On April 1, 1978 the L1 detector of this telescope experienced a simultaneous shift in energy calibration and an increase in count rate, perhaps as the result of a light leak in the thin AL entrance window. Subsequent data from this telescope were not used in the measurements of Chapter 4, Tables 4.1 through 4.12.

A background effect specific to the iron group nuclei was identified in Sections 3.3 and 3.5 and necessitated the further study and special handling which is discussed here. The affected PHA events were clearly seen in Figure 3.7, a plot of the charge measurements Z1 versus Z2, as a clump of points near (Z1=26, Z2=28). Inspection of the L1, L2 and L3 pulse heights of the events in this clump revealed a very specific signature:

1. The events lie directly on the L1 versus L3 track for iron, but lie just above the L2 versus L3 iron track, indicating that the L1 and L3 pulse heights are normal, but that the L2 pulse height is abnormally large.
2. The events have relatively small L3 pulse heights (E{L3} less than about 170 MeV), indicating that the nuclei responsible for the events just barely penetrated detector L2.

The effect also occurs for a fraction of the iron nuclei which barely penetrate the L1 detectors. In this case, the affected events may be seen as the unusual clump of points near Z1=28 in Figure 3.6.

The above signature suggests a pulse height multiplication effect in the LET 35 micron surface barrier detectors that is very similar to one observed earlier in the response of such detectors to fission fragments. The effect (discussed by Walter 1969) occurs for nuclei which deposit a large ionization charge density in the depleted silicon near the gold electrode. The large ionization charge density is thought to induce tunneling of additional carriers (electrons) from the electrode through the thin oxide layer which separates the electrode frm the depleted bulk silicon. The orientation of the LET 35 micron detectors is consistent with this multiplication hypothesis: The gold electrodes face toward the L3 detector, such that the ionization charge density near a gold electrode is largest for nuclei which just barely penetrate the detector, as in (2) above.

To further study the problem, three detector PHA events with charge measurements Z1> 24 were re-analyzed. For each event the L1 and L3 energy measurements (E{L1} and E{L3}) were used to calculate a new charge measurement Z' and obtain an estimate E'{L2} of the energy deposited in detector L2 by solving:

T{L1} = R(E{L1} + E'{L2} + E{L3}, Z', M) - R(E'{L2} + E{L3}, Z', M)
T{L2} = R(E'{L2} + E{L3}, Z', M) = R(E{L3}, Z', M)

taking  M = 2.132 * Z'

R is the generalized range-energy function discussed in Section 3.4. T{L1} and T{L2} are the mean pathlengths for the L1 and L2 detectors respectively. The results are illustrated in Figure B.1, a plot of E{L2} (the measured L2 energy) versus E'{L2} (the L2 energy inferred from the L1 and L3 energy measurements). Normal events lie along the diagonal, while events with abnormally large L2 energy measurements lie above the diagonal. (Events below the diagonal may be due to L2 edge effects and other processes discussed in Section 3.5.2) The abnormal events occur predominantly at large values of E'{L2} which can only be obtained by nuclei which barely penetrate L2, consistent with observation (2) above. Figure B.2 shows two histograms of E{L2}/E'{L2}; one histogram for events which barely penetrate L2 (E{L3} < 170 MeV) and one for the remaining events (E{L3} > 170 MeV). These histograms are summed over all the LETs used on both Voyagers and indicate that (1) about 15 percent of the events with E{L3} less than 170 MeV have L2 energy measurements which are abnormally large by an average of about 12 percent and (2) the L2 energy measurements are essentially normal for events with E{L3} greater than 170 MeV.

The results for iron nuclei presented in Table 4.2 (8.7-15.0 MeV/nucleon) are based on PHA events selected as follows:

1. Normal iron events (i.e. those with normal L2 pulse height) were selected using the charge consistency requirement and charge boundaries discussed in Section 3.5, and were counted if the total energy measurement, E{L1}+E{L2}+E{L3}, was appropriate to the 8.7 - 15.0 MeV/nucleon interval.

2. Iron PHA events having abnormally large L2 pulse heights were identified as those events with charge measurements near (Z1=26, Z2=28). (Specifically, the requirement was (Z1+Z2)/2 > 25 and (Z1-Z2) < -0.85.) For each of the PHA events the L2 energy measurement was divided by 1.125 (the mean shift of the abnormal L2 energy measurements; see Figure B.2) to obtain a corrected value E*{L2}. Then the event was counted if E{L1}+E*{L2}+E{L3} was within the total energy interval corresponding to 8.7 - 15.0 MeV/nucleon.

PHA events with abnormal L2 pulse heights constitute about 15 percent of the iron nuclei counts presented in Table 4.2. Any additional systematic error (beyond that discussed in Section 4.4.1) in the iron nuclei counts due to the abnormal L2 pulse heights should be small. For example, if no corrections at all are applied to the abnormal L2 energy measurements (as was the case for the iron abundances reported earlier in Cook et. al. 1979 and Cook, Stone and Vogt 1980), the resulting iron abundance measurements differ from those of Table 4.2 by less than 5 percent.

In the two parameter analysis of iron (for the fluence measurements of Tables 4.4 through 4.11) the multiplication effect in the L1 detectors was taken into account as follows:

1. Iron PHA events with normal L1 pulse heights were identified as those events with charge measurement Z1 in the range 24.8 to 27.2 and were binned according to energy/nucleon as described in Section 3.4.

2. PHA events with charge Z1 in the range 27.2 to 32.0 and L2+L3 energy measurements of less than 170 MeV are primarily due to iron nuclei with abnormally large L1 pulse heights, rather than nickel or other nuclei. Therefore, the L1 energy measurements of these events were divided by 1.125 before energy/nucleon binning. The remaining events with Z1 in the range 27.2 to 32.0 (i.e. those with E{L2}+E{L3} > 170 MeV) are mainly nickel nuclei an were included with no L1 energy adjustment. All PHA events were sorted into energy/nucleon binns as if they were due to iron nuclei, exactly as in (1) above.

In the two parameter iron analysis, no special treatment was applied to account for multiplication effects in the L2 detectors. They do not significantly affect the Z1 charge measurement and at most shift a small fraction (about 5 percent) of the PHA events which should fall in th 7.8-8.7 MeV/nucleon interval into the 8.7 - 10.6 MeV/nucleon interval.

The inclusion of nickel and other nuclei (27 <= Z <= 32) increases the fluence measurements over measurements of pure iron by roughly the combined abundance of these elements relative to iron, or about 10 percent. However, the inclusion of these nuclei probably has only a negligible effect on the spectral index( gamma, see Section 4.5), since the energy spectra of the various elements of the iron group are likely to be similar. (Spectral index variations among the elements appear to be roughly ordered by nuclear charge Z, such that, in a given flare event, neighboring elements have similar spectral indices; see Section 5.2. Further, the nickel to iron ratio {8.7 -15 MeV/nucleon, Table 4.2} is nearly constant from flare to flare.)

The multiplication effect in the L1 detectors is potentially an additional source of systematic error (beyond those discussed in Section 4.5) in the iron fluence measurements of Tables 4.4 through 4.11 and the iron spectral index measurements of Tabl 4.12. Upper limits to these possible additional errors were obtained by re- computing the fluences and spectral indices without the L1 energy corrections discussed in (2) above. The changes in the iron fluence measurements were typically less than 10 percent for energy bins in the range 5.0-8.7 MeV/nucleon, and wer zero for the other higher energy bins. The iron spectral indices changed by less than 5 percent. Since the actual systematic error induced in the iron spectral indices is probably much less than 5 percent (and therefore is small compared to statistical error and the possible systematic error from other sources discussed in Section 4.5) this error is negligible.

From Appendix G of Dr. H.H. Breneman's thesis:

Appendix G Instrumental Anomalies and Other Problems

The analysis of data form the Voyager CRS LET and HET telescopes was complicated by several instrumental problems. These are described in detail elsewhere (Breneman 1984); they are summarized here. The Flare analysis was treated differently from interplanetary data.

This effect has been observed many times in numerous surfacebarrier detectors, both in flight and in the laboratory (Cook 1981, Breneman 1982). Particles passing completely through a detector sometimes yield pulse heights that are anomalously high By about 10 - 30%). It occurs most often for particles with high dI/dx in the detector in question, and therefore at a given initial energy, the errect occurs more often for elements higher on the charge scale; in the data it is most prominent for Fe. On a delta E versus E' plot, the effect appears as a more or less diffuse 'track' above and roughly parallel to the nominal track for the element, since delta E is anomalously high for the affected particles (Fig. G.1). A charge determination of such an event will of course be high, generally by ~2-3 charge units at Fe. Since the effect is strongly dependent on dE/dx, it is usually evident only in the delta E detector immediately before the E' detector. When z is calculated for 3-parameter events involving such anomalous pulse heights, Z2 is more strongly affected than Z1, since the anomalous pulse height has the role of delta E for Z2, while for Z1 the same detector PHA usually makes only a modest contribution to E' with delta E normal. On a Z1 versus Z2 plot (e.e., Fig. 3.1), the effect takes the form of a cluster fo events to the right of, and slightly above, the main cluster along the diagonal. All of the Voyager LETs show the effect for Fe; although its rate of occurrence varies somewhat between the different telescopes, it is generally in the range of ~5 - 10% of all the 3-parameter Fe SEP events in a given telescope. The fraction of 2-parameter events affected is larger, since the E' detector is thinner and therefore a larger fraction of the data has high dE/dx in the delta E detector. At least one LET (Voyager 1 LET A) shows evidence for the effect at charges as low as 20.

In the HETs the problem is worse in several respects. Its rate of occurrence at Fe, as a percentage of the total Fe event ample, is generally much larger than in the LETs (~40 % for Voayger 1 HET 2); it is clearly seen for elements as low on the charge scale as Mg in some telescopes (Voyager 1 HET 2 and Voyager 2 HET 2); and for 3-parameter events it can sometimes be seen occurring in either (or both) delta E detectors, rather than just the last one, resulting in several displaced clusters of events in those telescopes (Voyager 1 HET 1 and Voayger 2 HET 2).

Two actions were necessary to deal with this problem. For abundant elements affected by the problem, mainly Fe and Ni, the 3-parameter charge consistency requirement was made lenient enough to include the particles affected by pulse height multiplication. The rate of occurrence in LET for elements lower than Fe was negligible compared to other sources of uncertainty in the abundance determination, and no correction was made for these elements. Based on the observed rates in some HET telescopes for the abundant elements (e.e., Fig. G.2 for Voayger 1 HET 2), the rate of occurrence in these thelscopes for elements in the Z = 17-25 charge range was significant even though limited statistics make it less apparent and less quantifiable. However, HET data were not used for these elements for the reasons given in Section 3.3.

In addition, the energy loss in the delta E detector had to be corrected in an approximate way for affected events, so that the total incident energy, which is required for constructing energy spectra, would be accurate.

For 2-parameter events, there is no second independent determination fo Z to permit unambiguous separation of normal and abnormal events. This is not a serious problem for abundant elements, since the only abundant element significantly affected is Fe, which has no other elements of comparable abundance near it on the charge scale with which it could be confused. For rare elements the situation would be more serious, but as noted in Section 3.3, 2-parameter data were not used for rare element abundances due to excessive background contamination from other sources.

For each Block I or Block II LET event, there is a single tag bit which specifies from whih LET telescope in that Block the event originated. A near coincidence in the triggering of LETs A and B, as is possible during periods of very high count rates, can result in the bit being set for a LET A event, causing that event to be read out as a LET B event (similarly for C and D). Since the bit is ordinarily set only for LET B events and is otherwise not set, LET A events can be misidentified as LET B events but never the reverse.

The telescope identification bit is used in all subsequent data analysis to determine the appropriate detector thicknessess and gains to use in calculating energy losses in the detectors and, ultimately, the charge of the particle. If the telescope identification is erroneous, incorrect thicknesses and gains are used in the calculations, resulting in incorrect determinations of Z. The magnitude and sign of the discrepancy in Z depends only on the (coincidental) relationship between the thicknesses of the detectors in the paired telescopes, and, to a lesser degree, differences in the energy calibrations fo the respective detectors.

On a Z1 versus Z2 plot of 3-parameter Voyager data, this effect has the appearance of small clusters of events displaced slightly from the main clusters along the diagonal for all the more abundant elements. It appears only in plots of the B and D telescopes, sine it is events with these identifications which contain some misidentified particles. In the Voyager flight data the effect is noticeable only during flare period 7, for which the peak LET B singles rate is ~ 5 * 10³/sec, the highest seen during the Voayger mission through August 1984. Its rate of occurrence is about 3% at this peak rate, an average of about 1% for flare period 7 as a whole, and is the same for all elements for which statistics permit a measurement. On delta E versus E' plots of both 2- and 3-parameter data, these effects have the appearance of 'ghost' tracks falling between or partially overlapping the real tracks of nearby elements (Fig. G.3).

Later laboratory work using the backup GRS and pulse generators (Martin 1983) was able to reproduce the effect with greatly improved statistics, and verified the magnitude of the time constant (~6 micro-seconds) implied by the flight data while extending coverage to event rates more than an order of magnitude above the highest seen in the flight data.

The impact of this problm on the data analysis is relatively minor. For 3-parameter data, the previously chosen charge consistency requirement is restrictive enough to easily exclude the misidentified events; the amount of data lost is an insignificant 0.2% of the total, and the remaining data set is as 'clean' as that from the other telescopes. The problem is more serious for the 2-parameter data, since there is no second determination of Z to permit separation of the normal and abnormal events; it is an unremovable source of background in the data. For abundant elements this is unimportant, since the error introduced by this background is on the order of 1 % or less. But the problem would be serious for rare elements in cases where the 'ghost' track of an abundant element overlaps the true location fo a rare element, since even 1 % of an abundant element could seriously contaminate a much rarer element. However, as noted previously, the 2-parameter LET data are not useful in obtaining rare element abundances on account of other background contributions.

As a result of their exposure to intense charged particle fluxes in the inner Jovian magnetosphere during the 1979 Jupiter encounters, the L1 detectors of the LETs experienced radiation damage which can be modeled as a reduction in the 'effective thickness' of the detectors. It is thought to be due to the implantation of energetic oxygen and sulfur ions known to be present in the inner Jovain magnetosphere (Gehrels 1982). Although all LETs on both spacecraft were affected to some degree, the Voayger 1 LETs were affected much worse than those on Voyager 2, since the former spacecraft passed closer to Jupiter and experienced a more intense radiation environment. On each spacecraft, LET C was by far the most seriously affected; this telescope was sparially oriented so as to receive the most intense radiation exposure during the Jupiter encounters. (LET B on Voyager 2 experienced unrelated types of radiation damage during the Jovian encounter and has returned no data since the encounter.) The front detectors of the HETs, being much thicker than those of the LETs and pro- tected by a thicker window, showed no detectable effective thichness reduction.

The impact of the radiation damage on the post-Jupiter data is apparent as a shift in the location fo the element tracks on a delta E versus E' plot of data from flares 16 and 17, the first large post-Jupiter flares, relative to their location in plots of pre-Jupiter flares. Similarly, Z1 versus Z2 plots of flares 16 and 17 show Z values shifted from their proper values when Z is calculated using the detector thicknesses measured before launch, which served adequately for all pre- Jupiter flares. The change in effective thickness appears to be somewhat dependent on Z, with the magnitude of the reduction increasing with Z for any given detector. Furthermore, with the passage of time the radiation damage seems to gradually undergo a partial reversal. This 'annealing' effect is evident in data from the later large flares, 20 and 24, which show less shift in Z than do the flares immediately following Jupiter encounter.

In the analysis of post-Jupiter data, this problem was dealt with by adjusting the L1 detector thicknesses used in the calculation of Z so as to make the calculated charges fall in the proper places on the charge scale. Flares 16 and 17 were used to define the required shift for the more abundant elements; a linear or weakly quadratic function of Z was fit to these to define the Z dependence for all Z. Data from flare periods 20 and 24 were used in conjunction with the flare 16/17 data to mathematically characterize the time-dependence of the annealing effect, by fitting to a decaying-exponential function of time. This procedure, repeated for each LET, defined the L1 thickness to be used in analyzing any given post-Jupiter event. The adjustment of the L1 thickness was then incorporated into the iterative cycle for calculating Z. The radiation damage did not have a noiceable effect on the inherent charge resolution of the telescopes, so with the above modifications the post-Jupiter flare data could be treated the same way as the pre-Jupiter data.

Table G.1 lists the adjustment made to the thickness for each L1 detector cor carbon and iron at two different times during the post-Jupiter phase. The actual expression for the thickness L(Z,t) of each L1 detector was given by

L(Z,t) = L0 - deltaL0(Z) + K exp(delta t / 562.56)

where K is a constant, L0 is the pre-Jupiter thickness, delta t is the time since the Jupiter encounter in days, and deltaL0(z) is the linear or quadratic function of Z that closely fits the required thickness changes for the first post-Jupiter flares.

During the time period 1978 APR 3 - June 9, the L1 detector of LET C on Voayger 2 experienced, for unknown reasons, a temporary gain shift (~ 47 % decrease) and an associated excessively high L1 count rate (~ 9* 10³/sec). The gain shift and excessive count rate set in abruptly, remained nearly constant until about May 29, and then gradually reverted to their former levels; a very slight decline during the central phase was consistent with the decrease in the intensity of sunlight during the smae time period and suggests the possibility of a light leak in the telescope's aluminum window.

The effect on the data was to shift the locations of the element tracks on a delta E versus E' plot, and yield shifted charge estimates when nominal gain factors were used in the analysis. The only flares occurring duting this time period were the three large events of flare period 7. Since the gain was constant at the shifted value during these flares, the data could be analyzed by generating the appropriate gain factors by fitting the oxygen flight data from flare period 7 in the manner described in Appendix A. The resulting values are included in Table A.1.

This anomaly was previously noted (Cook 1981), and the telescope was rejected for analysis because of the problem. However, since the energy calibration used in the analysis is easily adjusted to compensate for the problem, since the charge resolution and background of the telescope do not seem to be affected by the problem and since the three 7 flares include a major fraction of all SEP data, it was decided to inlude Voayger 2 LET C in the analysis with the special treatment described above.

At certain times during the Voayger mission, the configuration of the CRS instruments was changed in ways that influence data analysis. At the beginning of each flight the LETs were configured to require triggering of the L3 detector for pulse height analysis; that is, only 3- parameter events were analyzed. About 12 days after launch the L3 coincidence requirement was removed, permitting both 2- and 3-parameter events to be analyzed. For Voyager 2 this occurred before the first flares were seen, but on Voyager 1 flares 1a and 1b occurred before the configuration was changed, so no 2-parameter events were obtained from these flares. A similar situation occurred on 17 Jun 1978 when Voyager 1 LET C was switched back to requiring L3 coincidence. Thus for all flares from 8 onward there are no 2-parameter events from this telescope. These situations required changes in the weightings of 2-parameter relative to 3-parameter events for the affected spacecraft and flares. There could still be a residual abundance bias in flares 1a dna 1b if the particle composition was energy-dependent and if the two spacecraft saw particles with different spectra, but the possibility of a bias comparable to the statistical uncertainty in the abundances is very unlikely.

Another important configuration change is the HET gain state. Normally the instrument cycles between high and low gain modes, but after Jupiter encounter HET 1 on Voyager 2 was switched to a high-gain-only mode, so no post-Jupiter SEP heavy ion data were obtained from this telescope. This required changes in the partile wwighting factors of HET relative to LET for Voyager 2 for all post-Jupiter flares.

The event weighting factors tabulated in Table 3.8 include the effects of all instrument configuration changes.

On 1982 Feb 8, the Voayger 1 CRS experienced a failure affecting the readout of PHA information from the Block I telescopes (LETs A and B and HET 1). The result of the failure is that in place of PHA2, the instrument reads out whichever of the three PHAs has the largest numerical value. If PHA2 happens to be numerically the largest pulse height, as is true over some energy ranges, the event is read out normally; otherwise some information si lost. The effect of this problem is that some of the 2-parameter events are lost completely, and that some of the 3-parameter events are degraded to 2-parameter events.