4. Technical Issues Which Have Been Resolved

A variety of measures have had to be taken to tackle particular reliability problems of the polarimeter. The difficulties which have had to be addressed are:

• Significant cross-talk was discovered between the signal output cables on the two channeltrons; this was largely eliminated by the installation of improved shielding within the polarimeter.

• The high voltage supply used to accelerate the electrons toward the thorium target was found to be unreliable; there was a considerable delay in the diagnosis of this fault due to the difficulty of finding a reasonably safe means of measuring the required potential difference of many $kV$. Once the supply had been discovered to be producing only $750V$, replacement of the offending control potentiometer was rapidly effected^4.1.

• The channeltrons are designed as single-electron counting devices, intended to function by using an electrostatic field to accelerate an incoming electron into a collision which liberates further electrons from the channeltron walls, which are in turn accelerated to produce further collisions. This process amplifies a single electron into a current pulse containing $\sim 10^8$ electrons, which is readily detectable using a digital counter. However, the count rates measured by this method were found to be anomalously low. This is believed to be because the rate at which electrons were arriving at the channeltrons was substantially above the saturation rate ($\sim 10^4e/s$) at which the channeltrons' detection efficiency begins to drop rapidly with increasing electron arrival rate^4.2. This was the case with specularly reflected beams of energies between $\sim 2keV$ and $\sim 3keV$ from magnetic and non-magnetic surfaces, even for the smallest incident beam currents available. Therefore, it was necessary to make adaptations to allow the use of the front ends of the channeltrons as Faraday cups, the current collected at which could be measured as a continuous flow, rather than as discrete electron arrival events. To this end, an ammeter of sufficient sensitivity to measure the currents involved (of order $1pA$) was procured, and a wire which connected the front ends of the two channeltrons, possibly allowing flow of current between them, was removed.

• With the high voltage supply to the thorium foil in operation, continuous current measurement revealed current spikes, occurring at a frequency of $\sim 3Hz$, peaking at what later transpired to be approximately the steady current which was obtained in successful measurements. The spikes were set against a background of substantially lower current, and were also detected as pulses in single electron counting mode, whether or not an accelerating potential difference was applied across the channeltrons. This is now believed to have been a consequence of the electrical connection to the thorium foil and surrounding assembly from outside the vacuum chamber having some freedom of movement inside its insulating casing, leading to a situation in which there was good electrical contact only for brief periods, occurring at $\sim 3Hz$, producing the spikes which represented the current in the intended operating condition. The effect has disappeared since the mechanical security of the connection was improved.

• Calculation of raw asymmetries (equation 2.6) from the measurements (section 3.3) on the reflected beam from $Cu/0.5ML$ $Mn/7ML$ $Co/Cu(001)$ shows larger uncertainties and slightly poorer consistency between the two runs of the experiment than the results (figure 3.6) from the use of the formula (equation 6.1) [10,14] for the elimination of multiplicative systematic errors presented here (section 3.4), suggesting that the multiplicative errors do exist and were subject to drift over the period between the two runs of the experiment, and that the use of the formula, with the polarization reversal achieved using opposite remanent magnetization directions of the film structure, was a necessary measure.

• The difficulty of adjusting the zero on the ammeter used for continuous current measurements, given its slow response both to the zeroing control and to changes in the current which it was measuring, led to a suspicion that there might be additive offsets which would require a fitting procedure of the kind introduced in section 6.2. The non-zero and variable values obtained for the channeltron currents at zero incident beam current using this procedure (chapter 3) indicate that such offsets do exist and are subject to drift; therefore, both the fitting procedure and the time ordering of the experiments were necessary precautions.