ML19331C585
| ML19331C585 | |
| Person / Time | |
|---|---|
| Site: | Crane  |
| Issue date: | 09/05/1979 |
| From: | Pollack G MICHIGAN STATE UNIV., EAST LANSING, MI |
| To: | Gilinsky V NRC COMMISSION (OCM) |
| References | |
| NUDOCS 8008190073 | |
| Download: ML19331C585 (7) | |
Text
. \\ W. T h 3 -,-
b hilCHIG AN STATE UNIVERSITY COLLt.GE OF NNit'RAL $CILNCE
- DEPART &lLNT OF PHYSICS FAST LANSING * >!!CHIGAN
- 46124 EC;i.:syc,g. 3 September 5,1979 C011gh The Honorable Victor Gilinsky Cornissioner U.S. Nuclear Regulatory Commission 1717 H Street, N.W.
Washington, D.C.
20555
Dear Dr. Gilinsky:
Here is my report in response to the questions you asked me to look into l33 concerning Xe emissions from the accident at the Three Mile Island Nuclear
~
Station.
In order to get this information I had discussions with many scientists and engineers mostly of NRC, but also with others to whom I was referred.
I also read relevant NRC reports and other related material.
Ilhat I've tried to do is te synthesize this information into a coherent picture of the problem of 133 Xe emission.
The report is ordered under three headings each of which is a question or a few questions.
l33 I. How much Xe was emitted as a result of the TMI accident? How are the 133 estimates of Xe emission made? How reliable are the estimates?
The direct answer to these questions is that at least three independent 133 estimates of total Xe emissions have been made by people. The estimates are 6
6 6
2 x 10, 2.4 x 10, and 13 x 10 Ci.
The estimates are rough ones at best, i.e.
they make use of rough approximations, and are probably only reliable to within factors of two or three.
However, they represent responsible sensible work and are probably the best estimates one can make under the circumstances.
Background:
Xenon-133 is produced in the fuel of a reactor by fission of 235 U
during normal operation.
The TMI reactor had run for 140 equivalent full 133 power days at the time of the accident and the total amount of Xe in the 6 0 0 8190 07J ;
-'0 o.
6 inventory at t' hat time was 140 x 10 Ci. This number comes from the origin code, a computer calculation of the radioactive isotopes present as a result of pro-duction minus decay, and is generally believed to be reliable.
If all of this l33 Xe were emitted at the beginning of the acetient we'd get the maximum amount l33 6
of Xe emitted, i.e. an upper bound is 140 x 10 Ci.
We have good evidence from DOE envirormental samples (helicopter and ground l33 surveys) that the main radionuclide emittr.d during the accident was Xe That l33 means that some Xe got free from the fuel elements and into the air, probably by the pathway:
fuel elements to coolant water, then out of the reactor building into the auxiliary building, then through the filters and into the air via the stack.
During normal operation the emissioris from the reactor are measured at the stack by a combination of batch spectral analysis, which measures the relative amounts of the various radionuclides, and a stack monitor, which measures the gross number of curies being emitted.
Unfortunately the stack monitor went off scale about two or three hours after the accident started and it stayed off scale.
Had that not happened, i.e. had there been a suitably high intensity monitor in the stack, we could have used the stack monitor to estimate the total emissions.
Since this can't be done we have to use more indirect, less reliable means.
I found out about three of these and they are described briefly with their results below.
A.
Estimate from measured doses and meteorological data.
This estimate is described in NUREG-0558 and I discussed it in some detail with Dr. F.J. Congel of the NRC Radiological Impact Section.
Essentially what was done here was to take the doses as measured by thermo-luminescent dosimeters in place before and during the accident and to combine these doses with meteorological dispersion factors to get estimates of the strength 6
133 of the source of the radiation.
This gives an estimate of 13 x 10 Ci of Xe
. emitted.
The raw dosimeter data for this estimate come from the Teledyne Isotopes dosimeters in place at 20 onsite and offsite locations. The raw meteorological data coine from a tower, on the island, which reads wind magnitude and speed and the air temperature at various elevations. These kinds of data are gathered routinely. The meteorological data is analyzed by a meteorology group and they calculate tables of x/Q, the meteorological dispersion factor. This factor is the ratio of the steady state radioactive concentration at a point in the field 3
(x in Ci/m ) to the steady state rate of emission of radioactivity from the stack source (Q in C1/sec).
Thus the ratio x/Q is kind of a response function, similar to Green's functions used in formal physics. These factors depend in general on direction and distance from the source.
133 Probably the main source of uncertainty in this estimate of Xe emission is due to meteorological factors.
In general values of WQ arc thought to be good to a factor of two or three out to a few miles; beyond that the accuracy is worse.
At TMI the uncertainty was fairly large due to light winds.
Another potential source of uncertainty is in determining the dose: to which the TLD's have been exposed.
However the National Bureau of Standards has made a preliminary study of the calibrations and that appears to be O.K. (i.e. tie uncertainty is about +25% and -30%).
6 133 My conclusion is that this value (13 x 10 Ci of total Xe emission)may be off by a factor of two or three mainly due to meteorological uncertainties.
B.
Estimate from delayed grab samples.
This estimate was made by Dr. Andrew P. Hull of Brookhaven National Labora-tory and the work was reported at a recent Health Physics Society Meeting.
133 He obtained data on the actual Xe concentrations in about five grab samples of the effluent taken at the stack.
The samples were taken between April 6-10 several days after the accident, which was on March 28.
He plotted these data as
'r
.- a function of time and extrapolated back to t = 0 using a straight line fit.
He l33 0
than sumed up the total Xe emission and got the value 2 x 10 Ci.
The main strength of this method is that it relies on direct measurements l33 of Xe coming from the stack.
The main weaknesses are: first, as mentioned'above, the data were taken well after the accident and second, there is probably no real reason to believe that a straight line extrapolation back to the time origin is an accurate representation of what actually happened.
C.
Estimate from a remote monitor.
This estimate was described to me by Dr. Carol D. Berger, a health physicist at Oak Ridge flational Laboratory.
It is independent of the previous estimates 6
l33
~
and gave a value of 2.4 x 10 Ci of Xe emitted.
Dr. Berger's technique was to use readings from a monitor, in the auxiliary building, that saw from a distance of 40 feet a duct that feeds the stack. This monitor saw part of what went up the stack all the time and it never went off scale, as did the main stack monitor.
She calibrated the remote monitor against the main monitor during the short time that the main monitor was still on scale.
Associated with this calibration is an uncertainty of a factor of two, but there are other indications that the calibration may have been much better than that.
There also is some uncertainty introduced into this method by the necessity of assuming the distribution of radioactive gases in the effluent.
This distribu-fion was taken from the inventrey at the moment of shutdown, a reasonable choice.
In some ways this is, in principle, the most reliable of the available estimates.
D.
I learned in the course of this work that neith Woodard, a consultant 133 6
of the utility, estimated the total Xe emission to be 15 - ?.0 x 10 Ci.
I unfortunately do not know what information he used as a basis for his estimate."
l33 II.
Could we determine the total Xe emission from TMI in other ways?
I know of two other ways which in principle might be used to determine
~
,. l33 the Xe emission.
One way depends on conservation of Xe and the other depends on Xe which is adsorbed on the filters that precede the stack.
l33 l33 A.
Determining total Xe emitted by conservation of Xe We know from the core inventory at the time of the accident that there were 6
133 l33 initially 140 x 10 Ci of Xe present.
If no more fission or production of Xe I33 occurred after this time, then the total amount of Xe present at a time at (days) after the accident is 140 x 106,-Aat Ci, where A = 0.13153 day ~I (Xel33 33 has a half-life of 5.27 days).
The idea is to measure the total Xe still in the reactor (containment volume, core, coolant water) and compare this with the above. Whatever is missing has escaped out of the reactor into the environment.
The difficulties with making this work, are:
(a)
Xel33.has a short half-life so that measurements must be made shortly after the accident before it has decayed away to sut" reshold amounts; even tht:n it is a difficult measurement, and l33 (b)
It is necessary to measure radiation from Xe within the reactor where there is so much other radiation present.
6 l33 As a practical example: Suppose for illustration that 20 x 10 Ci of Xe had 6
133 escaped at the time of the accident and we were thus left with 120 x 10 Ci of Xe contained on March 28.
By August 20, when the present study was undertaken, that would have decayed to 120 x 106,-A145 = 0.63 Ci still contained within the reactor if no more had escaped.
This means, in order to detect.whether any Xe escaped in-133 itially it would be necessary,hp August 20,to measure the total Xe present to better than about + 0.1 Ci out of a total of 0.6 Ci all against the large back-ground of other radionuclides in the reactor.
That's probably much too hard to do.
There is an interesting application of this principle to the TMI accident, and l33 an associated puzzle.
On June 29, NRC engineers measured a Xe concentration of 1.5 x 10-2 Ci/ml in the gas space of the containment volume. The volume of this 6
gas space is 2 x 10 cubic feet, the volume of water in the containment vessel is 525,000 gallons, the solubility of Xe in water is about 0.1,
and at between 3/28 and 6/29 is 93 days.
From these data I calculate that on l33 6/29 there were present about 850 Ci of Xe in the gas containment volume, and l33 a negligible 3 Ci of Xe in the water volume.
This means that, if we assume l33 that no Xe escaped or was added to the containment building after the accident, there were about 850 e"93 = 170 x 106 l33 Ci of Xe present initially. That l33 implies that essentially all of the Xe escaped from the fuel into the contain-6 ment building.
In fact 170 x 10 Ci is even somewhat larger than the origin code value. The puzzle is that working from this same data, NRC engineers have con-cluded that half the noble gas was released from the fuel rods.
I do not under-stand the discrepancy between our conclusions.
l33 B.
Detm1 mining total Xe emitted by analysis of filteFs.
l33 The idea here is that all of the Xe that comes out of the stack first pbses through HEPA filters and charcoal filters. The charcoal filters are 2" j
thice, beds of carbon and some of the Xe remains on the charcoal filters. The mechanism for this is probably surface adsorption of Xe and subsequent diffusion into the bulk carbon.
l33 In principle it might be possible to tell something about how much Xe l33 passed through the filters from measurements of how much Xe they retained.
However I believe that this cannot with oresent knowledge tell us anything quanti-133 tatively useful about the Xe emission. The difficulty is that we would have to understand the interaction between Xe and carbon very well, then we'd have to know the entire history of the filters since they were exposed to Xe: Did 1
they ever get wet? Did fresh air blow over them? Did they ever get warm? etc.
l33 These are the kinds of things that determine the connection between how much Xe l33 the filters contain now and how much Xe passed through them.
I understand from Mr. John T. Collins that some of his group of engineers are working with the charcoal fili.ers from the TMI - 2 reactor.
p, 3
III.
What can be done so that Xe emissions and doses will be measured reliably in any future accidents?
In retrospect the best single way to insure that these emissions are reliably monitored is to install a stack monitor that will not go off scale during high radioactivity emissions. This may require a series of monitors which work from low concentrations to high ones with overlapping ranges.
There is a closely related question: How can the population doses due to l33 Xe emissions be measured more accurately? This can probably most conveniently be done by increasing the number of TLD's that are routinely in place, by more dense distribution of their locations, and by using TLD's that are well calibrated l33.
for the gamma (81 kev) and beta (350 kev) emissions of Xe 133 IV.
How can Xe emission be prevented in any fut"re accident?
This is difficult to answer optimally since the pathway for the emission isn't yet clear.
However, the following may work: Xe has boiling point and triple' point temperatures of,respectively,165 K and 161 K.
Probably if one provided "".rfaces which were cooled to liquid nitrogen temperatures and over l33 which the Xe had to pass in close proximity, then a large fraction of the Xe could be condensed and held until it decayed. The same kind of condensing plates 85 would probably work with Kr whose boiling point and triple point temperatures 0
are,respectively,120 K and 116 K.
There are a few practical problems in implementing this but the method should be straightforward.
For example it would l33 probably suffice just to cool charcoal filters when the Xe emissions were high and let the Xe be condensed and retained on tham.
Report submitted by Y
Gerald L. Pollack
~
' Professor, Physics Department
,