zer is counted as follows: each antico- 
incidence channel output is counted by 
a scale-of-4 and a mechanical register, 
each coincidence channel by a scale-of- 
10 and a register. 
Recording equipment provides in- 
formation about the progress of the 
counting during overnight operation. 
When the scale-of-1,000 that counts 
the anticoincidence output of the sin- 
gle-channel analyzer reaches full-count 
it resets. This actuates a pulse-form- 
ing circuit that feeds a signal to a sin- 
gle-pen millivolt recorder and thus 
marks the recorder paper. The paper 
moves at a constant 3 in./hr, so the 
counting rate at any time can be calcu- 
lated from the spacing between marks. 
Similarly, the counting rates of the 
lowest and the highest anticoincidence 
channels are recorded using scales-of- 
100. The pulses in all four anticoinci- 
dence channels are added together and 
recorded also, using a scale-of-1,000. 
Until more recording equipment ar- 
rives only these most important count- 
ing rates can be recorded. 
Temporary faults that affect the 
counting rates will be apparent from 
the record. The appropriate period 
can be deleted so that an erroneous 
answer is not obtained for the counting 
Tate. 
It is worth noting the high standard 
of performance demanded of the elec- 
tronic equipment and the proportional 
counter if the correct counting rate is 
to be obtained. The statistical accu- 
racy possible from this equipment is 
approximately 0.1 cpm, ie., 6 eph. 
The voltage applied to the propor- 
tional counter is 10,000 volts for a 
3-atmosphere filling. A count will be 
registered for all pulses greater than 
2 mv at the anode of the counter. A 
small rapid change in counter anode 
voltage, due to corona discharge or cur- 
rent leakage across the insulators, can 
cause spurious pulses. To avoid this, 
counter voltage must not change more 
than 1 part in 5,000,000 per 100 usec. 
If such pulses do occur the counting 
rate in channel 1 will change, the other 
channels not being affected. The 
counting rates from the other channels 
can thus be:used for calculating sample 
activity for this particular run. 
Spurious anticoincidence pulses can 
also result from a change in efficiency 
of the ring counters and pulse analyzer. 
The coincidence counting rate is 600 
cpm, so that 0.02% change in ring- 
counter or pulse-analyzer efficiency will 
Filling: 78 cm CO, 
Pure filling. 
/ 
i 25p of 
air added 
1 bt 
\ 
M 
100 p of 
air added 
rai 
Coincidences (cpm) 
4400 4600 4800 5000 §200 
Counter Voltage (volts) 
FIG. 5. Curves used to check for electronegative impurities give coincidence counts 
in channel 1 (heavier lines) and channels 2 + 3 + 4 (lighter lines) as a function of 
counter voltage 
Filling: 78 cm CO, 
Coincidences 
on 
(2) 
Anticoincidences: 
“living” sample 
+ background 
(cpm) 
w D 
fe) fe) 
Coincidences (cpm) 
nm 
oO 
Anticoincidences: 
radon - background 
‘Anticoincidences: 
background 
vv 
o 
=] 
(= 
o 
12) 
S) 
& 
fe) 
2) 
& 
c 
a 
fo) 
3600 4000 4400 4800 4200 
Counter Voltage (volts) 
FIG. 6. Total coincidence and anticoincidence counting rates as a function of 
proportional-counter voltage for a 78-cm CO, filling. Flat plateaus suggest low 
electron attachment 
CO: as a Counter Gas 
The results in our laboratory show that previous difficulties with CO, as a 
counter gas were due to electronegative impurities and not to the COz itself. 
These electronegative impurities were separated from the COz but have not yet 
been identified. For reliable proportional counting, the loss of electrons due to 
electronegative impurities should be less than a few per cent. To keep the elec- 
tron loss less than 1%, the concentration of O2 must be less than 1 in 10°, 
the concentration of Cl, must be less than 1 in 10". 
When we first considered using CO: we were not very optimistic in view of the 
measurements of English and Hanna (11). These workers were unable to get 
saturation even with the purest CO. commercially available, indicating that 
electron attachment was taking place. Freedman and Anderson (12) had used 
a CO.-filled proportional counter for biglogical applications, but the plateau 
slope of 10% again indicated electron attachment. De Vries and Barendsen 
(6) have applied CO» counting to radiocarbon dating, using a small counter. 
193 
