Sensitivity gauge
variations
in the Bayard-Alpert
ionization
Part two: Simulation of ion collection, for collimated beam input and chaotic gas input, using a digital computer received 2 March 1968; accepted 23 March 1968
D J Turner+ and C Priestland,
Central Research laboratory,
Edwards High Vacuum International
Limited,
Manor Royal, Crawley, Sussex
Editor’s Note
A Bayard-A/pert ionization gauge with end-caps on the grid has been found by many workers to have a higher sensitivity than one with an open-ended grid. Since the assumption of nearly 100 per cent collection efficiency is reasonable for a closed gauge, a simulation of the open gauge has been carried out to determine if this explains the effect satisfactorily. It was found that the loss of sensitivity by not having end-caps on the grid could be only partially explained in this way.
1. Introdllction In experiments described by Priestland and Holland (1968)‘, a Bayard-Alpert gauge with a closed-ended grid structure gave a reading 2.7 times as large as an identical gauge with an open ended grid when a collimated gas beam passed along its axis. Groszkowski (1965)’ has reported an “approximately threefold” sensitivity gain for the same gauge types in an experimental apparatus which would appear to give more chaotic gas conditions. Freytag and S&ram (1963)*, with a slightly smaller gauge, reported a 30 per cent gain for air, and Schulz (1957)p using helium with a gauge similar to Priestland and Holland /
/
efficient ion collection in the open ended gauge. There was the possibility that, due to the different configurations of the electric fields in the two cases (Figure l), there was a difference in the number of ions produced but it was felt that this would not explain the wide variations observed. The first explanation proposed (to account for the effect using a collimated beam) was that the time required for an ion to be collected was greater than that for the ion’s initial axial velocity to carry it out of the end of the gauge. (In the closed gauge the ion would be reflected and thus would not escape.) However, for the particular gauge and gas flow considered (see Figure 2) the maximum time to collection (for an ion produced at the grid) was approximately 1O-6 sect while a gas tthe equation of motion of an ion accelerating transverse plane is:
along a radius-vector
in the
a. Open-ended gauge d2r f -- 0 dt2 -
--
v,1
and this gives the time to collection
b. Closed-ended gauge Figure 1.
T_
Electric field in the gauge
gave a sensitivity “at most 20 per cent higher than the ordinary Bayard-Alpert gauge”. It appeared from these that the difference was due to less*Author’s new address: Industrial Measurement and Control Research Assn, South Hill, Chislehurst,
Vacuum/volume
18lnumber
6.
Group, Kent
Scientific
Instrument
Pergamon Press Ltd/Printed
for an ion produced
at the grid as
2r,Ine) m [
1
-Zn(!i)]
% -PC )I m
(2n + l)n!
n=o
where re -collector radius ra =grid radius l’a -potential difference between collector and grid e =ion charge m =ion mass The infinite series is found to converge quite rapidly
in Great Britain
319
D J Turner and C Priest/and:
Sensitivity
variations
in the Bayard-Alpert
ionization
gauge
I
c
LO.OO238 z
Dimensions in metres
Maximum-length
path
‘X
Figure 2. Layout of the system molecule travelling at the average velocity would take approximately IOW set to traverse the entire length of the gauge. Since this could only explain a small part of the effect it appeared that tangential velocity components due to incomplete collimation were causing ions to spiral round the collector in orbits which did not intersect it. Comsa (1966)5 has analysed ion collection in random gas due to this effect. Using his equations, a collection factor of 74 per cent was obtained for ions formed near the grid with higher factors for ions formed closer in, and an overall collection efficiency of 87 per cent. Similar orbits in a closed gauge would decay due to the field distortion at the gauge ends (Figure 1). Criteria have been given by Turner (1964)” for determining whether a given ion is collected, orbits, or arrives at the grid with sufficient energy to pass through. Since the directions of travel of molecules in the collimated beam could not be specified by any distribution law, the capture probability could not be calculated directly. It was decided therefore to simulate both gauge situations (collimated beam and random gas input) on a computer. These differed only in the way in which ions were generated. The programmes were written in Algol, using ICT input/output procedures. 2. Programme
for collimated gas beam
This commences with an analysis of gas flow in the collimators in order to achieve the correct directional distribution. In determining points of ionization, account had to be taken of the fact that a molecule entering the gauge at a very oblique angle would be in the ionizing field for a much shorter distance than one travelling more nearly parallel with the axis and thus would be less likely to be ionized. Allowance also had to be made for the ionization free region near the collector (Section 2.3). 2.1 Flow in the collimators. The collimators were two tubes each 0.00476 m diameter x 0.028 m long, spaced a distance of 0.017 m apart. Molecules were considered to have been injected with an isotropic distribution and re-emission from a surface was assumed to follow the cosine law. All lines and points were considered on a set of three-dimensional Cartesian co-ordinates with origin in the centre of the entry aperture and z-axis along the axes of the two collimators. Since the system had rotational 320
symmetry it was assumed, for simplicity of the equations, that x -0 at the injection point and at re-emission points. Molecules were considered to have been lost if passing back through the entrance orifice or falling between the two collimators. The direction of motion of the molecule was specified in terms of two angles. Firstly, the angle between the direction of motion and the normal (0). For injection this angle was rectangularly distributed between 0 and n/2 while for reemision it had a cosine distribution over the same range. The angle describes a cone with its apex at the point of emission (or injection) and with the molecule travelling from the apex along its slant surface. The second angle specifying the direction of motion was a rectangularly distributed angle between 0 and ZZ, referred to as 4, which was the angle of rotation of the line around the surface of this cone, from a datum point. For injection, where the normal was the line with direction cosines (O,O,l), the direction cosines of the direction of motion are, I 7 sin 0 . sin 4 171: sin 0 . cos rf,
n
cos 0
For reflection where the normal had direction cosines (O,-l,O), (due to the point of re-emission being at y equals plus 0.00238 m), the direction cosines of the direction of motion are, I L= sin 0 . sin a m = -cos 0 n == sin 0 . cos 4 Finally, when a molecule was found to have been transmitted through the collimators its point of intersection with the entry plane of the grid structure was found to determine whether it entered the gauge or passed outside. The grid was 0.02 m diameter and its entry plane was 0.013 m from the exit plane of the collimators. 2.2 Ionization of the gas entering the gauge. Strictly speaking, an individual molecule has a very low probability of being ionized. Therefore, in order that the programme should not be unreasonably long, it was necessary to consider only those molecules likely to be ionized, while still allowing for different probabilities of ionization due to varying path length in the gauge. The probability that a molecule will be ionized in a distance x,
D J Turner and C Priest/and: Sensitivity variations in the Bayard-Alpert ionization gauge assuming the ionizing efficiency of the field to be the same at all
points, is given by p(x)= 1 -exp
-3 (
L)
where L =mean distance to ionization. Since very few molecules are ionized we know that, for motion in the gauge, xt < < L. In this case expanding the above equation gives us
and the probability becomes $
of ionization
in the range x to (x + dr)
Thus, in the system being considered, the proba-
bility of ionization for a given molecule is proportional to the distance it travels inside the gauge, and ionization is equally likely to occur at all points along its path. The maximum distance a molecule could travel inside the gauge was 0.411 m. This distance is travelled by a molecule being emitted from one lip of the exit aperture of the collimator and passing through the exit aperture of the gauge at a point diametrically opposite to this (see Figure 2). The programme length was reduced by ignoring all molecules travelling farther than this before ionization and assuming that if a molecule travelled a distance 0.0411 m in the gauge the probability of ionization was 1. The probability of ionization for an individual molecule was then equal to the actual distance it travelled in the gauge, divided by 0.0411 m. This ensured that the relative probabilities of ionization for different trajectories was proportional to the path length in the gauge. In fact the dual aim of determining if a molecule was ionized and, if so, where, was achieved by multiplying 0.0411 m by a rectangularly-distributed random number in the range 0 to 1 to obtain the distance travelled by the molecule before ionization. From this, and the molecule’s point of entry into the gauge, and direction of motion, the point of ionization was found and this was examined
I
0
o*oos
Radial distance (m.) Figure 4. Ionizing and collection efficiencies in the experimental gauge
to see if the molecule had already passed out of the gauge. This
was complicated by the possibility of the molecule passing through the ionization free region before being ionized, in which case its trajectory was considered in two parts. 2.3 Ionizing efficiency as a function of radial distance. Due to the collector being negative with respect to the filament, electrons are prevented from approaching closer to it than a certain distance. Furthermore, as electrons approach this limiting radius their energies become too low for them to produce an appreciable number of ions. Figure 3 shows the ionizing efficiency of electrons in argon against their energy (Barnard, 1953)‘, and Figure 4 shows ionizing efficiency plotted against radial distance for the system used. This was obtained from Figure 3 and the known relationship between potential and radial distance. For simplicity it was assumed that ionizing efficiency is zero for radii less than 0.0013 m and constant outside this region. 3. Programme for random gas flow in the gauge
Figure 3. Ionizing efficiency of electrons in Argon
The probability that an ion will be formed at a given point is the product of two probabilities. (a) that a molecule passes through that point (b) that a molecule which does pass through the point is ionized there. The Iirst is, in a sense, a measure of the pressure at the point. Since, for random gas flow, this will be constant throughout the gauge, it follows that the first probability is independent of the position of the point. As has been shown previously, the probability that a molecule entering a uniform ionizing field will be ionized in the distance x to (x + dr) is equal to dx/L, where L is the mean distance to ionization. It should be noted that this is independent of the distance (x) travelled before the point is reached. Thus, the second probability given above is also independent of the position of the point. If a molecule is only in the field for a very short distance, this reduces the probability of it being ionized but does not affect the probability of it being ionized at a given point. It follows from this that the probability of ionization is constant throughout the gauge. 321
0 J Turner and C Priesfland:
Sensitivity variations in the Bayard-Alpert
ionization
gauge
The axial co-ordinate of the point of formation of an ion will be given by a rectangularly distributed random number in the range 0 to 0.04. The radial distance to the point of ionization was considered initially to be rectangularly distributed in the range 0.0013 to 0.01. However, since the probability of ionization or injection within the radii r to (r+dr) depends on the volume within the range this would not be true, and the chaotic gas programme was therefore run using both distributions. (Sections 5.3 and 5.4). The direction of motion of the ion is then specified in terms of the same two angles as described in section 2.1. 4 is generated as before and 0 is now rectangularly distributed in the range 0 to x. From this point on, both programmes proceed in the same way.
4. Ion trajectory in the gauge If the molecule under consideration was ionized it was necessary first to obtain a value for its velocity, which had a Maxwellian distribution. This states that the probability that a molecule has a velocity between I( and (U+du) is given by
Figure 6. Ion-orbit configurations in the transverse plane
4.1 Ion velocity.
P(u,u+du)=$(&~ u2enp( -g)du
(1)
where vz = mass of the molecule k -= Boltzmann’s constant T = temperature of the gas The velocity found was then considered as separate axial and transverse components and, in the case of the transverse component, the angle (A) between the direction of this component and the radius vector was also found (see Figure 5).
inward-moving ion will have a velocity of the same magnitude as an outward moving ion). (b) rmin > r, and r,,,,, > rg (u, i u, and 14~> uR). The ion must pass through the grid whether it is initially outward moving (case 2) or inward moving (case 6). (c) r,i, < r, and r,,