1 (1987) 145-161
Comparators
A C C U R A C Y A N D APPLICABILITY O F T H E k o - S T A N D A R D I Z A T I O N METHOD
The present paper deals with the accuracy and applicability of the ko-standardization technique
in NAA. Topics included are: user-oriented outline of the method, relevant nuclear data (ko, Q0,
~tc.), the non-1/E epithermal flux distribution, small detector separations and/or extended source
geometries, (n, 3') reactions with a Westcott.g :~ 1, primary interferences, subsequent (interrupted)
irradiations, the non-constancy of the neutron flux during irradiation, and a final account of the
accuracy. Although the paper is written in terms of the ko-methodology, a good deal of the
considerations can be transferred to most types of single-comparator standardization.
+
Research associate of the National Fund for Scientific Research.
146
F. DE CORTE et al.: ACCURACY AND APPLICABILITY
~,Au-factors
An experimental k0,Au factor is determined as /2,3/ :
with Fc~ = Cd-transmission factor for epithermal neutrons /10/. At the present
MTAA-77T986 Conference, the results for the extremely difficult case 96Zr(n,y)
9?Zr (Q0 = 248) are dealt with in detail by Simonits et al. /7/, As a new out-
put of ~he above mentioned efforts, the present paper is accompanied by a INW/
KFKI Interim Report /ii/ with a tabulation of k0=Au-factors for 94 analytical-
ly interesting isotopes. The Report also shows that, when calculating k O [Eq.
(2)3 from the data given in Refs /13,14/ [0,o~ and /15/ [y], the discrepancy
with the recommended ko-factorsis:
10" 147
F. DE CORTE et al.: ACCURACY AND APPLICABILITY
Q_0-values
The accuracy requirements on the Q0-values to be introduced in Eq. (I) are
less stringent than on the k0-factors , since the term [f + Q~(e)]/ [f +
Q0 (~)]
causes an error reduction. Nevertheless, Qo-values often being inaccurately
known in the literature, a careful choice or an experimental redetermination
is required, especially for cases with high QO and for application to ENAA
/lO/.
Following a similar philosophy as described above for the Gent-Budapest
cooperative determination of k0-factors , tabulations with recommended Q~-va-
lues were released in 1979 /16/ and 1984 /17/ for resp. 13 and 57 analytical-
ly interesting (n,y) reactions. Recently, we have updated and extended this
work - including determinations in the R4V4 channel of the DR-3 reactor -,
reaching a total of 72 (n,y) reactions. The very difficult cases of Q0-deter-
mination for 94Zr(n,y)95Zr and 96Zr(n,y)97Zr are discussed by 8imonits et el.
/7/ at the present MTAA-7/1986 Conference (see also below). Recommended or
carefully evaluated Q0-values are also listed in the INW/KFKI Interim Report
/11/. As a continuatlon of the above cited 1979- and 1984-papers, a third Q0-
tabulation is in preparation.
Although it is difficult in general to predict the residual uncertainty of
the analysis result caused by the uncertainty in Qo - this being dependent on
the magnitude of QO versus f -, it has been shown 727/ that on the average a
value of I% can be assumed.
9 2
i 9
wlth w i = (gF~,Fn/F)i/Er i' g belng noW the statistical weight factor. Based
on these appr6xlmatlons'one arrives at the conversion formula :
148
F. DE CORTE et al.: ACCURACYAND APPLICABILITY
and this is the corrected Qo(~)-factor which figures in Eqs (I) and (3).
Postponing for a moment the discussion on_the accuracy of the Qo/Qo(~)
converffion, i.e. the accuracy of the I/E I+~ - E concept, it is clea9 that
two additional parameters should he known : thernuelear constant Er' and the
neutro__n flux parameter ~.
E -values can be calculated from Eq. (6), with the introduction of liter-
ature ~ata foX the relevant resonance parameters. Ryves' papers /18,19/ con-
taining only Er's for a few isotopes, Moens et al. /20/ calculated values for
96 (n,y) reactlons, based on the 1973-edition of BNL-325 /21/. At the present
MTAA-7/1986 Conference, Jovanovic et al. /22/ are presenting an updating and
extension for 127 (n,y) reactions, based on the recent compilations of Mughabg-
hab et al. /13,14/ and including a discussion of accuracy.
The parameter ~ should be experimentally determined, and simple tech-
niques have beendeveloped /5/, namely the "Cd-covered multi-monitor" method,
the "Cd-ratio for multi-monitor" method and the "bare multi-monitor" method.
In the latter, suited for instantaneous ~-determination, the minimum number
of monitors is three, and the use of ~Zr-96Zr-197Au has been proposed /5/,
with the Zr-pair serving at the same time for f-monitoring. At the present
MTAA-7/1986 Conference, Simonits et el. /7/ deal extensively with the use of
Zr as an e- and f-monitor, emphasizing recent accurate remeasurements of the
input nuclear data.
.+Let us consider now the crucial question of the accuracy of the adopted
I/E l ~-Er approximations. Although much can be said theoretically/23-26/,
Table 1
,,Standard" monitors and nuclear data for use in the "Cd.ratio for multi-monitor" method for a-determination. Er
and Q0 values are based on MUGHABGHABet al.13,~4 Fcd=Cd-transmissiongfactor for epithermal neutrons, to
be multiplied with RCd
Monitor Er, eV Q0 T
149
F, DE CORTE et al.: ACCURACY AND APPLICABILITY
9it looks more conclusive to consider it from a practical point of~view. Re-
cently we have performed e-determinations, based on the "Cd-ratioifor multi-
monitor" method /5/, in 5 irradiation positions of three differens reactors
with a largely diverging neutron thermalization : channel "MILA" in the WWR-
M reactor, channels 14, 7 and 16 in the THETIS reactor and channel R4V4 in
the DR-3 reactor. We have introduced the use of 8 suited monitors, for which
the adopted nuclear data (E , Q , T ) - to be considered as "standard" data
0 12
for this application - can ~e assume~ to have an acceptable accuracy (see
Table I).
Fig. I shows for each channel one example of the ~-results obtained with
197Au-23eU-9~Mo-l~176 The adequacy of the linear fits is expressed as
(og E'r" *c
I R C # - I ) Oo(~<1
WWR -M / CH. MILA
, .,=o.o,,;o.o,o _ x:=~
n-APR-85 ~ f = 36.01 ti3,112 "
._z 0 4 - J U L - 8 5 ~ ~~n
3~35'~ " ~ I ~ ~ ~ X2 =O.~,TO
~ 0.00.
"1 ,I I . I .I, . I I
1.0 2,0 3.0 I~ Er
Fig. 1. Experimental a,determination in Various channels of the WWR-M reactor (Budapest), the THETIS reactor
(Gent) and the DR-3 reactor (Risb). Use is made of the ,,Cd-rafic for multi-monitor" method (5) with some
,,standard" monitors and nuclear data of Table 1
150
F. DE CORTE et al.:ACCURACY AND APPLICABILITY
Er- ~
log
(Rcd- I) Qo(~)
1 _z o 1
._
-- 04- JUL-1985 r *
X~ 1852
i F I I I I t I.
1.0 2.0 3~ log Er
Fig. 2. Check of the accuracy of the ~-determination in channel 7 of the THETIS reactor (Gent) by using two
different "standard" monitor sets
151
F. DE CORTE et al.: ACCURACY AND APPLICABILITY
Table 2
Qo-values determined in 5 channels/3 reactors, based on ~-determinatinn according to the
"Cd-ratio for multi-monitor" method. "Qo" is obtained without any correction for
"Qo" Q0 "Q0" Q0
-values are 338 eV for 96Zr(n,%) and 6260 eV for 9~Zr(n,7). Thus, these de-
r .
termlnatlons . are especlally
. . challenglng
. . the latter
for . reactlon, slnce It h as
to be assumed that the validity of the I/E I+~ approximation, shown to hold
up to 2560 eV [E_ (64Zn)], can be extrapolated to ~ 2.5 times higher neutron
energy. Without ~oing into detail, the final results of repeated determina-
tions are listed in Table 2, and the consistency of the results, after cor-
rection for ~, is striking.
Two practical remarks: ~/ ECd should be = 0.55 eV, especially for low-
QO monitors (Mn, etc.)/5/; thus, one should stick as close as possible to the
prescription: a quasi-point source centered in a large cylindric Cd-box with
I mm wall thickness and height/diam.= 2. B/ one should avoid different neu-
tron thermalization by polyethene spacers etc. in the containers for bare
and Cd-covered irradiation.
The propagation formula of the uncertainty on ~ towards the analytical ~
results has been given in a previous paper /3/. The residual uncertainty is
of course dependent on the flux parameters and on the isotope under conside-
ration, but on the average it amounts to ~ 2%. The residual uncertainty on
the analysis result, caused by the f-determination, was shown /27/ to be of
the order of 1.5%. A negligible residual uncertainty is induced by E , which
is, for isotopes with medium and high Q0-factors, usually known to w~thin 10%
accuracy.
152
F. DE CORTE et al,: ACCURACY AND APPLICABILITY
153
F, DE CORTE et al.:ACCURACY AND APPLICABILITY
the Westcott g-factor from unity. More precisely, the estimation of the error
introduced should be based on the variation of g with neutron temperature,
rather than on the g-value itself, and it is interesting to demonstrate this
statement in a simplified way with the example of 197Au(n,y) ISSAu, the ulti-
mate comparator. In general, a recommended k 0 ~_U -factor is the average of de-
terminations in several channels of the WWR-M and THETIS reactors, with neu-
tron temperatures from e.g. 20 to I00~ Thus, according to Ref./35/, gAu is
ranging from 1.0038 to i~0068. This variation of 0.3% (well within the expe-
rimental uncertainty) is - roughly speaking - transferred to the k O Au(eXp)
values. Hence, the average recommended k 0 A -factor is approximatel~ corres-
ponding to an average neutron temperature'o~ 60~ with gAu = 1.0053. If this
k^ 9 -factor is now used in an actual analysis at a neutron temperature he-
t~ 20 and 100~ the committed error on the resul~ing concentration turns
out to be maximum 0.15%, which is negligible. Similar cases are for instance
109 A
g(n,y) 11om~g~ 133 Cs(n,y) 134 Cs, I~8 Ba(n,y) I 3 9 ha, 181 Ta(n,y) 182 Ta, I8S W(n, )
187
W, 1 8 5 Re(n, T) ~ 8 6 Re, 2 3 2 Th(n,y) 2 3 3 m.h and 2 3 8 U(n,y) 2 3 8 U.
A somewhat more serious situation occurs for reactions like ~~
(see Table 3) with g-factors deviating a few percent from unity. Here, errors
of the order of I-2% are to be expected, which is still acceptable.
From analytical standpoint there are two cases where the g-factor is
drastically deviating from unity~ and thus significant errors will be commi-
ted : 1~IEu(n,y) and especially 176Lu(n,y) (see Table 3). In principle, de-
termination of Eu and Lu can also be performed via the iS~Eu(n,y) IS~Eu and
175 176m 9 .
Lu(n, T) Lu reactlons (for whlch recommended k0-factors are avallable).
However, for instance in non-destructive NAA of geological material, measure-
ment of 154Eu and 176mLu gives rise to poorer precision or even to a complete
loss of quantitative information (inherent poorer sensitivity for 154Eu; short
half-life and low gamma-energy for 176mLu), and this is - especially for the
rare earths - a regrettable situation.
Although up to now we have not performed any experimental work related to
the above problem, the solution is straightforward: one should change over to
the Westcott-convention, including a neutron temperature monitoring. The lat-
ter can be accomplished by the well-known technique of Lu-irradiation in two
irradiation positions: the one under investigation and a reference position
with known temperature. Note that this relative measurement was formerly re-
commended because of the uncertainty of the knowledge regarding the cross-
section for 176Lu(n,y)177Lu and the 177Lu decay scheme. In view of the recent-
ly re-evaluated data of interest [oQ = (2090+70) barn /14/; I0 = (1087+40) barn
/14/; y (208 keV) = (11.0+O.4)% /44f3, it might be interesting to expl~re the
possibility of an absolute T-monitoring with acceptable accuracy.
Finally, it is worthwile to dwell on the situation arising when a neu-
tron temperature monitoring would be performed. Since Lu would he coirradia-
ted, it can serve as a relative standard for Lu-determination, and the compli-
cation of T-monitoring and changeover to the Westcctt-convention is then only
necessary for the 151Eu(n,y) reactions. However, if recommended k~-factors
would be available for 152m Eu and 152 Eu (again, as above for Au, ~or T = 60 o C
and g = 0.8532), it appears from the g-factors of Table 3 - and follo~ing the
same reasoning as outlined above for Au - that the maximum committed error on
the Eu-concentration would be ~ 4.5%, which is in most cases acceptable for
chondrite normalization. Thus, it seems justified to determine the ~s2mEu and
~52Eu k0-factors, which - although physically not very meaningful - will lead
to reasonably accurate analytical results.
154
F. DE CORTE et al.: ACCURACY AND APPLICABILITY
Table 3
Isotopes for which application of the ko-method gives rise to small or significant errors (see text);
(a) = WESTCOTT/36/, (b) = GRYNTAKIS et al./35/, (c) = ENDF/B-V/37/,
(d) = MOENS/38/, (e) assumed to show a limited deviation from g (20 ~
Isotope Isotope
measured measured 100~
20*C 100*C 20~
Table 4
Some important primary interferences in comparator-type and absolute NAA;
crosS-section data from MUGHABGHAB et al./13/(o0, Io) and CALAMAND/39/(~-; + estimated);
* - from THIS WORK; ** - from Ref./40/; *** - based on Q0 from Ref./41/
formed by
Produced
(n,y) (n,2n)
isotope
8, % ~0' barn I0, barn Er' eV 8, % ~, barn 8, % ~, barn
Primary interferences
A problem inherent to many types of single-comparator NAA-standardization is
the ~ccurrence of primary interfering reactions of the type (n,n') and (n,2n),
causing a positive error if not corrected for. Some cases of practical impor-
tance are listed in Table 4, together with the relevant data for the isotopic
abundances and cross-sections.
The (n,n') interferences for 117msn and 135mBa were for the first time re-
cognized and elaborated during the course of our work. The (n,n') and (n,y)
cross-section values for the production of 135mBa were experimentally deter-
mined in exactly the same way as described previously for ~ T m s n /40/.
Table 5 shows the contribution from (n,y), (n,n') and (n,2n) to the ob-
served activity, for irradiation in three channels of reactor THETIS with
155
F. DE CORTE et al.: ACCURACY AND APPLICABILITY
Table 5
Contribution of (n, 7), (n, n') and (n, 2n) to the formation of TTmse, aTmsr, lllmCd, 11~mSn, laSmBa and
1a 7tuBa in three channels of reactor THETIS
...
THETIS Channel
Produced Formed 17 f 5
isotope by r162 = 15 Os/r = 42 r162 = 158
@s/r f = 1.6 ~s/r = 21 r162 = I18
= -0.028 = 0.052 = 0.11
low, medium and high thermalization. It is clear that the formation by (n,2n)
is negligible, except for 137mBa and for 117msn in a very hard neutron spec-
trum. On the other hand, the (n,n') interference is indeed a serious obstacle,
except for 77mse and 87msr. Nevertheless, corrections can of course be ap-
plied on condition that the fast flux component is known or is exPerimental-
ly determined; this cannot be considered as a serious restriction, sinc~ ~ -
monitoring is often necessary in N A A - whatsoever the standardization Fro--
cedure - so as to enable corrections for interfering (n,p) or (n,a) reac-
tions. In some favorable cases (high flux, low thermalization) it is in prin-
ciple possible to use Zr as a fast-flux monitor.Recently, the 238U spectrum
averaged cross-section for the reaction 9~ SgZr has been accurately
determined by two research groups, namely Sekine and Baba /42/ [(0, I04 +
0.005) mbarn] and Mannhart and Fabry /43/ [(0.103 + 0.004) mbarn]; 0the~ re-
lated nuclear data are : MZ~ = 91.22 ; @ (S~ = ~51.45 + 0.02)% /12/;
T (SSZr) = (78.43 ~ 0.08)h 744/; y (909.1 keV) (99.04_+--0.03)% /44/.
156
F. DE CORTEet al.: ACCURACYANDAPPLICABILITY
Note that it is even not necessary to measure the single comparator eve-
r~ time in the same counting position; if it is done, then of course e~ =
Ep. p,i
157
F. DE CORTE et al.: ACCURACY AND APPLICABILITY
Np/tm .......
cy0tirr ~(t-t. ~)
wD F(t)~e ir~ dt
158
F. DE CORTE et al.: ACCURACY AND APPLICABILITY
Table 6
Final account of the accuracy of k0-standardization in NAA; the overall uncertainty is obtained by quadratic
summation of the individual contributions
induced by contribution
k0 i%
Q0
2%
f 1.5%
eoine, corr. 1.5%
e -measur. and 2%
P conversion
Overall ~ 3.8%
- the general problem of how to prepare a standard with accurately known com-
position. According to our experience, systematic errors due to non-stoi-
chiometry of compounds or caused by pipetting of small volumes are no ex-
ception; they were detected and eliminated by our cooperative but indepen-
dent k0-determinations ;
- the problem of how to find or to prepare a standard suited for the analysis
of bulky samples which are measured close to the detector. Different Sp and
gamma-attenuation imposes in fact the same type of conversions as outllned
in the present paper.
Finally, it can be mentioned that the above estimated overall uncertain-
ty of < 4% was confirmed by actual analyses of various (standard) reference
materials /8, 47, 48/, showing at the same time the practical applicability
of the k0-standardization method.
The authors are highly indebted to L. MOENS (Gent), A. ELEK, E. SZABO (Budapest), K. HEYDORN, L. H.
CHRISTENSEN, J. JAKOBSEN, E. DAMSGAARD and U. JACOBSEN (Ris6) for their help in various ways. ~lanks
are due to M-L. MUNGHEN for the typographical work. The financial support of the Belgian National Fund for
Scientific Research is gratefully acknowledged.
References
159
F, DE CORTE et al.: ACCURACY AND APPLICABILITY
160
F. DE CORTE et al.: ACCURACY AND APPLICABILITY
11 161