LA=6039

C3Q COPY Reporting Date: July 1975Issued: October 1975

A&curate Determination of Impurity Concentrations

4

(10s(

in Plutonium Metals by Statistical Evaluation

of Analytical Data

by

C. J. MartenG. L. TietjenM. M. Horita

I

)

D alamosscientific laboratory

“ of the University of CaliforniaLOS ALAMOS, NEW MEXICO 87545

An Affirmative Action/Equal Opportunity Employer

uNITED STATES

ENERGY REsEARCH AND DCVELOPMCNT ADMINISTRATION

CONTRACT W.7405.ENG. S6

printed in tho Unibd Statoaof Autericu. Available fromNational Teohniad hifOIBMitiOli %rvfcx

U S Department of Commttrca52SSPorI Royal RoadSpringfield, VA 22151

Rico: printed Copy S4.00Microfiche $22S

l%im-W was preuud u m ●mumt d workspmwrdb’ lb United$WW tie-meat. NelUurlk UnkedSISleaw tk Unltd Elata E- ~h ●d Dwclapmcml M-mlnhtratton, nor qv atthelr anploycec. w *W &the& mm.Ira*. sukcontrach, or Iktr cmploye4s. auk ●ay

‘bwarnn v expcmu or implied, or asmimrs amy kd Ihkillty ar-wad hlIY for the K$ww. mmelrtrm. or ~Im:.Kv Iafaradom. .pgaratu, Hucl. . pmcossrwrnrnt, tkd [W w weuk! 8C4 lmr?iluc mriv.tdy 9WA

v

●

rkkts.

ACCURATE DETERMINATION OF IMPURITY CONCENTRATIONSIN PLUTONIUM METALS BY STATISTICAL EVALUATION

OF ANALYTICAL DATA

by

C. J. Martell, G. L. Tietjen, and M. M. Horita

ABSTRACI’

Analytical data from a plutonium-metal exchange program con-- 1. :. ducted by six ERDA laboratories are statistically evaluated. The-comm objective is an accurate determination of five metal impurities

~~-’ (aluminum, chromium, iron, nickel, silicon) in each of three~— r.=

3—-m; plutonium metals by using data from four analytical methods. The===ml __J~ statistical evaluation yields the weighted mean and its standard

z—~sg~ deviation for each method, plutonium metal, and impurity, using a

3-g( -procedure that minimizes the effect of outliers by assigning zero

!?~.-nr weights tQ the most extreme values and variable weights to the‘E’s;— __ remaining data. Where possible, weighted means from the various_~m~ analytical methods are Pooled.-

.I. INTRODUCTION

Well-characterized standard materials are in-valuable for making accurate analyticalmeasurements in any field, and in the nuclear fieldwell-characterized plutonium metals have not beenavailable. Data are presented here for three standardplutonium metals accurately characterized accor-ding to content of five impurities: aluminum,chromium, iron, nickel, and silicon.

Data from a plutonium-metal exchange programconducted by six ERDA laboratories have beenanalyzed, and three plutonium metals, identified asH, R, and P, have been carefully characterized.These well-characterized plutonium metals shouldbe useful in resolving measurement differencesamong laboratories, improving current analyticalmethods, developing new measurement capabilities,and providing data with which to resolve shipper-receiver differences.

By using these standard plutonium metals, alaboratory can evaluate its analytical performanceand take corrective action if its analytical results arein error. Also, work is under way using these three

plutonium metals in evaluating several types ofPuQ emission spectroscopy standards. --

Data from the plutonium-metal exchangeprogram, recorded quarterly from December 1971 toDecember 1974, will be considered here. Unfor-tunately, different reporting procedures were usedduring this period, and the same procedure was notused by all the laboratories at any given time. Theeffect of these differing procedures will be discussedlater.

Four methods of analysis were used: emissionspectroscopy; chemical method; spark source massspectroscopy; and atomic absorption. Of the four,only emission spectroscopy can be used to determineall five impurity elements in all three plutoniummetals. The chemical method determines only iron.Emission spectroscopy and the chemical methodhave been used in the exchange program longer thanthe other methods and so have contributed thelargest number of values to this report.

Spark source mass spectroscopy has been usedonly at the Rocky Flats laboratory. For calibrationpurposes, spark source spectroscopy uses the ironvalue determined by atomic absorption for a given

1

sample and gives values for aluminum, chromium,nickel, and silicon.

Atomic absorption has been used in the exchangeprogram for only a short time. The impurities deter-mined by atomic absorption include aluminum,chromium, nickel, and iron. Rocky Flats has alsoreported a small number of values for silicon.

A secondary purpose of this report is to suggest astatistical treatment for future plutonium-metal ex-change data that may lead to a consensus on impuri-ty concentrations while a supply of a particularbatch of plutonium metal still exists. These well-characterized metals could be considered “certified”plutonium metals. Additional determinations, suchas those for carbon, gallium, plutonium, anduranium, could also be so treated.

II. STATISTICAL METHODOLOGY

For each impurity, plutonium metal, andanalytical method, the data from the six par-ticipating laboratories are combined, and from 8 to127 data values are given in the initial evaluation. P-metal iron data taken before December 1972 byemission spectroscopy at two of the laboratoriesshowed a high bias and were deleted. The twolaboratories had been using PUOZ obtained from thedirect ignition of plutonium metal. However, astudy 1 made during 1972 showed that high ironvalues are obtained from P-metal when the particlesize of the PUOZ is not controlled. When thelaboratories began to control PUOZ particle size, thehigh bias for iron was no longer observed.

Because the data come from several laboratories,analytical methods, and reporting procedures, out-liers are likely to be present. It is difficult to deter-mine an underlying distribution for the data. Cer-tainly the assumption of normality is not ap-propriate. How, then, does one best estimate themean under these circumstances? An estimator thatis not very sensitive to the underlying distribution issaid to be “robust.” We believe that the methoddescribed in this report is the best method of robustestimation to date. It makes use of weighted meansand prescribes how the weights are to be assigned.Standard deviations of the weighted means are alsocalculated, but their properties are not yet wellknown.

If N,i= l,..., n,denotes the i-th observation andw, a weighting factor for the i-th observation, theweighted mean is defined as

I-w =

n

z ‘ixii=l .

.

,Let ri = Ixi – X~ denote the absolute value of the i-thresidual and define S to be the median residual. Theweights themselves are defi.ned iteratively as w I =I/ri sin (n/cS), where c is an arbitrary scaling factor(c = 1 in this report). Ten iterations are made star-ting with unit weights. This method assigns zeroweights to observations when ri > TS, large weightsto observations with small residuals, and smallweights to observations with large residuals.

The standard deviation of the weighted mean,SRW, is then calculated for each impurity in eachplutonium metal and for each analytical procedureby

ST =w

n’

1

U2

z

2r. w.

1i=l 1

2(nl-~) ~.

1. i =1

where n’ is the number of observations receivingnonzero weights.

The computer program, data, residuals, weights,and various means and standard deviations aregiven in Appendixes I and H.

111. RESULTS

The weighted mean and the standard deviation ofthe weighted mean for each impurity, plutoniummetal, and analytical procedure are summarized inTable I.

A weighted mean and its standard deviation, bothpooled from the analytical methods, are shown inTable H. These are computed as follows:

t?’ Y—

~Vr(pooled ) =1 W1+ ‘; XW2 + ‘i ‘w=

“nl+n’+nt

123

Impurity

Aluminum

Chromium

Nickel

Silicon

Iron

TABLE I

WEIGHTED MEAN AND ITS STANDARD DEVIATION FOR EACHIMPURITY FROM EACH ANALYTICAL METHOD

(micrograms per gram of plutonium metal)

PlutoniumMetal

HRP

HRP

HRP

HRP

HRP

EmissionSpectroscopy

174.6 + 2.922.5 + 0.555.5 + 1.3

196.7 + 2.845.0 * 0.948.6+ 0.9

441.0& 5.2101.3* 1.2137.2+ 1.5

180.8+ 3.934.6+ 1.024.4& 0.8

962.8& 10.5111.4* 2.2318.1+ 5.5

Spark SourceMass Spectroscopy

177.5 + 10.516.2 + 0.453.7 + 1.8

217.5 + 3.738.0 + 1.543.9 & 2.7

449.1 + 3.694.2 + 2.4

135.4 & 5.4

222.4 & 20.235.8 + 1.136.2 + 2.2

. .

.-

. .

AtomicAbsorption

183.5 + 2.534.6 k 2.4

--

171.5 * 1.148.9 + 1.0

--

437.5 ● 3.1108.1 + 1.6

--

129.7 + 3.632.6 + 0.9

-.

915.7 + 11.4112.9 * 2.1

--

TABLE II

WEIGHTED MEAN AND ITS STANDARD DEVIATION FOR EACHIMPURITY POOLED FROM SEVERAL ANALYTICAL METHODSa

(micrograms per gram of plutonium metal)

Plutonium Metals

Impurity H .R P

Aluminum 176.3 + 4.2 22.5 & 0.5b 55.3 * 1.3Chromium 196.7 + 2.8b 46.1 + 0.9C 48.2 + 1.1Nickel 440.9 + 4.7 102.5 + 1.4 137.0 * 2.1Silicon 185.8 + 7.7d 34.5 * 1.0 24.4 + 0.8bIron 949.2 & 8.8 112.7 + 2.2 322.0 + 4.2

Vata from all analytical methods are combined h compute the pooledweighted means and pooled standard deviations of the weighted meansunless indicated by superscripts.

%ased on emission spectroscopy.

%ased on emission spectroscopy and atomic absorption.

ChemicalMethod

---------- --

--------—----

------------

------------

947.0 & 4.3114.4 * 2.3326.4 + 2.1

%ased on emission spectroscopy and spark source mass spectroscopy.

and IV. SUMMARY AND RECOMMENDATIONS

[

: + cif2 s: + df3 s; !?dfl %

‘1 ‘2-w

~ (pooled) =3

ndfl + df2 + df3

r

where subscripts 1, 2, 3 refer to the variousanalytical methods, n’iis the number of observationsin the i-th mean with nonzero weight, and dfi = n; —1.

In three instances, values come only from emissionspectroscopy. This method has produced the largestamount of data, and the weighted mean of the emis-sion spectroscopy results is between the weightedmeans reported from the other two analyticalprocedures. In two other cases the weighted meansare pooled from two analytical procedures reportingvalues for an impurity. The weighted means notpooled in these cases are from analytical methodsthat have apparent biases for these impurityelements in these plutonium metals.

An approximate t-test, where

t=

is used to check for significant differences at the 0.05level among any two of the analytical methods. Withjudgment based on analytical experience, one canarrive at the same conclusion as the t-test in 80% ofthe cases. Because of the variety of reportingprocedures and the rounding and averaging ofresults, the estimates of precision calculated for thisreport, while correct for the data as used, are not en-tirely satisfactory.

In 20’% of the comparisons, the t-test gives a t-value marginally significant at the 0.05 level. Thedifferences, while statistically significant, are notpractically significant. In these few cases, analyticaljudgment indicates that the weighted means of thepertinent analytical methods can be pooled.

Differences between the weighted means for someimpurities (Table I) may indicate that biases exist.If there are such biases, they are not consistentamong the several impurities; i.e., one analyticalprocedure does not always yield higher results thananother. Because the methods are supposed tomeasure the same quantity and because there seemsto be no consistent bias between methods, theweighted means are pooled from the methods, ex-cept for those cases noted.

The values given in Tables I and 11should be usedwith judgment. Where there are apparentdifferences among weighted means, further ex-perimental work is recommended. To resolve thesedifferences, individual laboratories should use thevalues of Table II in a conscientious effort to ex-amine their analytical methods, especially wherelarge discrepancies occur between theirmeasurements and the ones reported here. In thisway, and with consistent reporting practices, the ac-cumulation of future data will be helpful in deter-mining impurity concentrations in plutoniummetals even more accurately than shown here.

ACKNOWLEDGMENTS

We gratefully acknowledge the efforts andcooperation of the many people at the participatinglaboratories who contributed the data discussed inthis report. The participating laboratoriesare: Atlantic Richfield Hanford Company, Hanford,WA; Savannah River Plant, Aiken, SC; Rocky Flats,Golden, CO; Mound Laboratories, Miamisburg,OH; Argonne National Laboratory, Chicago, IL; andLos Alamos Scientific Laboratory, Los Alamos, NM.

REFERENCES

1.C. J. Martell, “The Effect of Particle Size on theCarrier-Distillation Analysis of Pu02,” Los AlamosScientific Laboratory report LA-5454 (February1974).

.

2.D. F. Andrews, “A Robust Method for MultipleLinear Regression,” Technometrics 16, 523-531(November 1974).

i)

H

a

:.2.x-.u

z.

0?U

J

Ca.

0u

A-<0

11

Uau

m

.

5

.muu-.D

mm

mIr-1

.ml-lu

NN

N3

z+

0-

*.WZ

ZZ

*)-8

04

”zx

am

=&

&a

u

.,.,

6

..

-.a.“.

x.“xna-.x.

..

7

APPENDIX II

DATA, RESIDUALS, WEIGHTS, AND VARIOUS MEANS AND STANDARDDEVIATIONS FOR EACH IMPURITY, PLUTONIUM METAL, AND ANALYTICAL PROCEDURE

Analytical Method

Emission SpectroscopyMetal H

Metal P

Metal R

Chemical MethodMetal HMetal PMetal R

Spark Source MassSpectroscopyMetal H

Metal P

Metal R

Atomic AbsorptionMetal H

Metal R

Impurity

AlCrFeNiSiAlCrFeNiSiAlCrFeNiSi

FeFeFe

AlCrNiSiAlCrNiSiAlCrNiSi

AlCrFeNiSiAlCrFeNiSi

,

Page

*

91011121314151617181920212223

242526

272829303132333435363738

39404142 *434445 ●

464748

8

.“●

xu.a!..=,-.-,e“x..axM

r-*mN

*5*.-+

N*m

Jr-e

r-mO

mO

O***

*G*

OO

Om

Nm

-N=

--IN--IF

IWI

N1

9N

-0~tU

19

~N

FI-I

Ntim

,ut-lm

m000000-0

00

Oooeo

oeo

oo

Oooooeo

ooo

..

..

..

..

..

..

..

..

..

..

..0-....

..

9

zoW(IJ

:

10

e*w

#ti

Om

OO

-8*0

*e

nll-o

.-on

oe

we

oo

*tid*d

ti.9.

6J

CW

-F)F

I.-O

CU

W)N

ON

-N

NF

)WO

%(U

IUIU

f9w

0tU

fi~lU

WI~

fUo

oo

oo

co

oo

oC

oo

eo

oo

oo

e0

00

00

00

00

00

0s

..

....0

...

..

..

..

..

..

..

..

..

..

..

..

00

00

.m.

“xw.s

-l-”a-o

@J*e

mF

)*ti--8

*e’=

+e

I8a

ea

“-lQ

t-J*e

mO

ticfutv190Lw

.-*~N

0NF

J~

F)W

N15-1U

f9W~

tW7V

J0NtW

NN

F)190

Oo

oc

oo

oo

oo

oS

=?O

oe

oo

oo

oe

oC

Oo

oa

eo

oe

o.

..

..

..

..

..

..

..

..

..

..

..

..

..

...0

.0

0

.

.0

.x“.s.-a.x.xu

11

.aa..

x..3“

om

mo

oo

r-wr-m

ao

r-dm

mm

co

ina

tcw

tvwr-o

mm

mw

Om

ooeo-$o-tid00.-8.0--.”.

ati”e

aO

.de

HO

oo

oo

ee

oo

oO

oo

co

co

co

oo

00

00

00

00

0.

..

..

..

...*.....

..

..

..

..

..

0.

..

.0

00

00

aux.

.

u%u

.

12

..

zo0---1;

13

..

u

..”

....-.-

ulD

-““.

U----

xax

zz

.amuu

.

14

.

a“xu.s.-a.x“3*M.-,Ux“..-*.-.-U“n.

“

-.--w”

“m.”

00

0

15

n

us,-a.xs,9.4

.-.ax.-=.-““xw

NF

1--0

0N

’@*W

*0~*

W-*C

Ofu

eN

dc

wN

ti.O

dcuN

onlm

eti

00

00

00

00

00

00

0-0

00

..

..

..

..

..

..

..

..

.

..

16

..

a“x“.-“=,-.Q“

Flc

8J

E*N

=-w

cve

O?-”.-I

OF

tUO

dO

me

ae

Om

.c.c

u*m

4u

******.+m

ee

eo

eo

am

tna

c.e

00

00

00

00

--0

00

00

00

00

Oo

oo

oe

..

..

..

..

..

..

..

..

..

..

..

..

.0

00

tm---m

90

u““”

---

xa=

z%

17

zowu)

,“xUl

a-1izav.

-1“’(n

“a.x.xu“x,-.-Uuxu“a.

.-‘D.aux“

..

..

..

..

..

..

..

..

..

..

00

..

.;e

.

18

a2D-

%’

mu

r-un

:5+...

*mu

194Cuu

---““”---

xaz

zi

19

*.=,-axz=.a“x“..-%,-.a.xu

..

20

zomWtl

.“.

“.Im

olm-Id-.

uanxs.“s.lx“x

.

21

a

*ae

an

lr-m*w

i*N**em

wtir-om

oui-N

+a

al’o

ol~

ar-m

cm

ca

aa

ae

ea

l-.co

wm

tra0

00

GO

OO

OC

O0

00

0=0

00

00

00

00

0=

..

..

..

..

..

..

..

..

..

..

..

..

.0

““””----

xa

szz

...

22

.

z0in

msw

.!2

——

23

1

.-O

mO

Oa

*futw

drq

am

m*

Op

mm

um

r-*m

om

.tU

INW

luc-&

(l.&

.W

-0

-8-

NH

NcV

O@

tidW

OW

Oooooooooc

8=0

00

00

00

00

00

00

0z

..

..

..

..

..

..

..

..

..

..

..

..

.0

0

*r)tveo

-oN

*mI*lu

O(vm

ora

r)luo

-$c

uo

*nJr)O

dd

na

*oN

nJ

”tu-

NC

UN

C+J

NN

ON

O-m

Jo

do

aJ

(u0

00

00

00

00

0O

oe

eo

oo

co

oo

00

,00

0=

..

..

..

..

..

..

..

..

..

..

..

..

..

—0

00

..

24,

m.

z0.

a

zzas’

S-.D

-*--8OO

QF

-+H

w”+

m“

@*U

INF

IUll#l*~

-KIU

te

if)em

Oo

oo

oe

oo

oo

0=0

00

0=

..

..

..

..

..

..

..

..

25

.------

---——

——

—-——

--—

s

..a“xu

r-?-aw

ae

eu

em

?-vo*O

----N

tim4

ml

Nm

lN

Iwo

Jo

J.-

NO

IXO

(UC

UC

V(U

00

00

00

00

00

00

0-0

00

00

..

..

..

..

..

..

..

..

..

.0

0

.a.x

“.””--=x

asz%

26

x

us“E“xus.a.x““s

U)

win

““n

---““”---

27

“i“a.“K.x““s“lx

“

*mo

mN

oC

um

IWIU

N-*

..

..

0Ue

nlr)

.01,4

”1

A[

Ot-d

*u

OF

I**O

oe

oa

..

..

0

1.

.

..

,

x

0-00.

Onto

0000m

..

..

00

.2

a

“x“u“=.a“xM..

2Zma“x?.“

.s*Nm

00-0-0000.

..

.

.0.

30

hiva2v!

xa:Ina

00

MN

*.

.s

..

*Oua“xn.-“.-x.a“xu

31

Nm

.(U

*.-3

..

ruo

“.a9

r

a

--Ju--s

..

m-

.*m

x“m

f-

O-Q

“---3.

.0mo

m“

tnm

m

%

..

---uyn

-“””---xaszi

32

**“

du

l00

=.

.

.

xMs

mm

.-m.d

33

E$“-1m

.*a

m

OF

Iw

omO,w

-%

..

0

.

1

34

*n..-

::s

..

..

“*a

m

x“s

K#N

~-tu

-

-ome

35

a

x.

eu

”“

a1

.-

Un-Im

e

....

36

nd

“me

..

=.*

a

x

te

M.“mu

Oln

mm

Om

37

Imm

*m

mo

mm

e**O.

..

Ne

d

.-.UJ

a2>

u*C

“s.0.a.xN“s.94

.--a,.x.z,-S.-.a“x“

..

38

.

a.x“x,“.-.Qx..-3.-.“a“x“

rl.-w000@

lu@

InS

oo

om

au

lo

oc

cc

cc

e.

..

..

..

.0

00

ma

oo

oe

wd

U-lln

oo

om

**0

=0

00

00

0.

..

..

..

.e

oa

39

..

.-a

i5

,-

Zu

-zs

.xu

ooO

m*@

?N(u

O.e

00

*m*4

u1

0“0

06

++.00

..

..

...0.

00

0

x

,0.4.-a“x“.-..-S.-.“aux“

40

.

.s.a“x“,..

z

“s.*a“x““;,“.-U“xn.s“a“xu

cu*nlm

.-nJ-**d

-dO

dH

ddd

000000000.

..

..

..

..

d.*e

J*$

ow

*A

O.

od

”o

dd

000000000.

..

..

..

..

..-.m+xe

O*

AO

mw

om

NO

*eu

lm*lu

*ul

oo

co

oa

oe

o.

..

..

..

..

0

UI

---m”

“.0.9”---x

a=zz

..

42

...

na3,“.-a.x“..-=

.-.Hauxn

u

43

I

-z

a“%u..-3ua.

..

..

..

..

**mm

m*m

Om

ml-lm

mm

mo

0000O

GO

C.

..

..

..

.0

*

..

44

●✎✎

b

“x“s.-a“x““3.-1s.4.-*ax““94

.-x.-,-la“x“

0.m

-*cu

n*m

Lna*r?

l-mIcu

*.

..

..

..

..

..

...

..

.

c0~40n

ltiew4J0

0-1:

0-

45

“s“a.x.“%Ore

-.

ax““3

.-,-.-,a“xu.-x,-

Oorlln

*m@

.a*m

ot-u

.o.o

e.oa

-...

..

~8

-lm.c

Om

NO

cu@

*.Dw

.u.c.ao@

000000000

..

..

..

..

.0

WI

(’utu

alr-*c

om

tulJ

lN

m-?

NcP

oo*o

..

..

..

...

46

>

u.D

$*a“x““s,“

de

wfoeooulu!

-OU

lme

me

ee

+O

aoooaao

o.

..

....9

0

9-9

nb4u

.----x

a~zz

47

.,

u*m

.x““s.-iu.-,a.x.-“%“

.

..

rCJ

..

48