PHYSICAL REVIEW A VOLUME 56, NUMBER 6 DECEMBER 1997

K a 1,2 and K b 1,3 x-ray emission lines of the 3d transition metals

G. Hölzer,1 M. Fritsch,1,2 M. Deutsch,2 J. Härtwig,3 and E. Förster1
1
X-ray Optics Group, Institute of Optics and Quantum Electronics, Friedrich-Schiller-University Jena,
Max-Wien-Platz 1, D-07743 Jena, Germany
2
Physics Department, Bar-Ilan University, Ramat-Gan 52900, Israel
3
European Synchrotron Radiation Facility, Boî te Postale 220, F-38043 Grenoble, France
~Received 14 May 1997!
The K a 1,2 and K b 1,3 emission spectra of the 3d transition metals Cr, Mn, Fe, Co, Ni, and Cu were
measured, employing a single-crystal diffractometer optimized for minimal instrumental broadening. The
high-accuracy diffractometer, and the interferometrically calibrated silicon crystal employed ensure absolute
wavelengths in the metric scale to a sub-part-per-million accuracy. An accurate analytic representation of each
line, obtained by a fit to a minimal set of Lorentzians, is presented. The absolute energies, linewidths, and
indices of asymmetry, derived from the data, agree well with previous measurements. So do also the intensity
ratios K a 2 /K a 1 and K b 1,3 /K a 1,2 , which are, however, slightly, but consistently, higher than previous values.
Possible origins for the observed Z-dependent trends are discussed. @S1050-2947~97!03012-6#

PACS number~s!: 32.30.Rj, 32.80.Hd, 31.30.Jv

I. INTRODUCTION of suitable spectra for 5 other elements, namely, Cr, Mn, Co,
Fe, and Ni as a first part of a study similar to that of Cu. The
Simultaneous multielectronic transitions within the atom results of these measurements are presented in this paper.
play an important role in determining the structure in and the Their strength lies in being the only set of K a and K b emis-
intensities of x-ray emission spectra. This is particularly true sion lines for ~almost! all of the transition metals, which
in the case of the 3d transition metals, whose asymmetric were measured on the same instrument under highly opti-
line shapes were attributed as early as 1928 to contributions mized conditions and on an absolute energy scale of sub-
from two-electron transitions @1#. Several other mechanisms parts-per-million ~ppm! accuracy. The optimization, based
such as conduction-band collective excitations @2#, exchange on a detailed study of the instrumental effects involved @8#,
@3#, and final-state interactions @4# were also suggested as ensures a minimization of the distortions introduced by them,
equally probable alternatives. In spite of an extended and and allows the removal of any remaining distortion by nu-
merical methods. These procedures are crucial for obtaining
extensive research effort over several decades @5# no final
reliable and accurate spectra, as found in the Cu study @5#.
agreement on the physics underlying these line shapes
Precise knowledge of the line shapes and their accurate rep-
emerged. Very recently, combining precision line-shape
resentation by analytical forms such as the sums of Lorent-
measurements and ab initio relativistic Dirac-Fock calcula- zians employed here and elsewhere @5,9# are also important
tions, we were able to show that the line shapes of the Cu for applications in high-precision x-ray diffractometry. These
K a and K b emission lines can be fully accounted for by are, for example, stress and grain size determination by line
contributions from 3d-spectator transitions only, in addition profile analysis of polycrystalline materials and high-
to the diagram ones @5# ~in the following, underlining de- precision crystal lattice parameter measurements @10#. Our
notes hole states!. This conclusion was strongly supported by results should prove, therefore, useful in these fields as well.
the considerably improved agreement with theory of the L This paper presents and discusses the measured spectra.
and M level widths and fluorescence yields, which now Accurate analytic representations of the line shapes are
could be derived from the diagram contributions only, after given, in terms of sums of Lorentzians fitted to the data, for
stripping off from the measured lines those due to the spec- the applications discussed above. Characteristic quantities
tator transitions @6#. High-resolution measurements of these such as positions of line maxima, linewidths, indices of
spectra photoexcited at energies near the K edge show asymmetry, etc. are determined and compared with the nu-
clearly the first appearance of the asymmetric features at an merous separate measurements of individual lines available
excitation energy coinciding with the calculated threshold for in the literature. Finally, the intensity ratios K a 2 /K a 1 and
creation of a 1s3d two-hole configuration, which is the ini- K b 1,3 /K a 1,2 are derived from the fits and critically com-
tial state of the 3d-spectator transitions @7#. While these re- pared with previous measurements.
sults support very convincingly a two-electron excitation ori-
gin for the line-shape asymmetry in Cu, it is clear that for
this interpretation to be conclusive, neighboring 3d elements II. EXPERIMENT
must be shown to have a similar behavior. To our knowl-
A. Measurement setup
edge, suitable high-resolution spectra for such a study are,
however, not available at present, nor has there been any The measurements were performed using a high-precision
in-depth study, such as that discussed above for Cu, for any single-crystal spectrometer @11# in the classical Bragg geom-
other element. We have, therefore, undertaken measurements etry. The instrument is of a high mechanical stability. Its

1050-2947/97/56~6!/4554~15!/$10.00 56 4554 © 1997 The American Physical Society

56 K a 1,2 AND K b 1,3 X-RAY EMISSION LINES OF THE . . . 4555

TABLE I. Measurement parameters for the line-shape determination. The angular widths D u s , which correspond to the width of the
spectral line, and D u d , which is the ‘‘spectrometer window,’’ as well as their ratio, are also listed. A positive asymmetry for a given
monochromator crystal reflection indicates grazing incidence relative to the crystal surface.

Element Line Reflection U @kV# / Bragg angle Asymmetry Slit width Dus Dud D u s /D u d
I @mA# ~deg! angle ~deg! hor. ~mm! ~s! ~s!

Cr Ka1 331 40 / 30 66.76 22.0 0.04 176.5 21.2 8.3

Kb 331 56.79 22.0 131.4 6.2
422 70.11 19.5 239.7 11.3

Mn Ka1 422 30 / 20 71.43 19.5 0.04 247.7 21.2 11.7

Kb 422 59.49 19.5 160.7 7.6

Fe Ka1 333 25 / 32 67.85 0 0.04 200.2 21.2 9.5

Kb 333 57.18 0 156.8 7.4
531 73.10 228.6 332.8 15.7

Co Ka1 531 30 / 35 77.00 228.6 0.04 291.4 21.2 13.7

Kb 531 61.99 228.6 223.0 10.5
620 70.69 43.1 338.6 16.0

Ni Ka1 620 30 / 35 74.48 43.1 0.03 214.5 15.9 13.5

Kb 620 60.87 43.1 245.0 15.4
444 73.11 0 449.6 28.3

Cu Ka1 444 40 / 32 79.30 0 0.03 328.3 15.9 20.7

333 47.48 0 65.1 4.1
Kb 444 62.63 0 265.8 16.7
553 40 / 30 79.91 12.3 0.10 770.6 48.5

mean total angle dividing accuracy in the absolute position coefficients from Henke et al. @14# for the wavelength at the
over the full angular interval of 2p is 0.12 arcsec and its maximum of each spectrum.
minimal angular step is 0.06 arcsec. A silicon crystal with
the surface orientation ~111! of WASO9 type ~obtained from B. Resolution and instrument function
Wacker Chemitronic, Burghausen! was used to measure the A major problem for any line-shape measurement, includ-
emission lines for most of the 3d transition metals. Its lattice ing those using single- and double-crystal spectrometry, is
parameter a was determined in the metrical system by means the evaluation of the ‘‘true’’ emission line from the mea-
of combined x-ray and optical interferometry @12# with an sured and instrumentally distorted profile. Preferably, the in-
accuracy of @13# Da/a59.6310 28 and allows the determi- fluence of the instrumental function should be minimized
nation of the emission line peak positions on an absolute already in the measurement process itself by an optimal se-
energy scale with high accuracy. lection of the experimental setup. This has been done in our
The incident beam collimator consists of a 390 mm long experiments on the basis of computer simulations of the
tube with crossed slits at both ends, directly attached to the spectrometer window function for our case @8#. The influence
x-ray tube. For all measurements the vertical slit heights of the collimator geometry, the crystal reflection properties
were fixed at 0.42 mm while the horizontal widths varied as ~as given by the dynamical theory of x-ray diffraction!, the
required. The total air-path length of the radiation from the tube arrangement and the absorption of the radiation in the
x-ray tube to the detector is 800 mm. The x-ray generator is anode, air and windows were taken into account. If the in-
a highly stabilized ID3000 from Seifert GmbH&Co KG. The strumental window is much narrower ~ideally a d function!
stability was tested in a 60-h run. No measurable drift of the than the spectral linewidth, the measured line profile corre-
x-ray intensity was observed. Stability against temperature sponds to the ‘‘true’’ emission line profile to a very good
variation was ensured by locating the spectrometer and the approximation. In this case, the measured data can be cor-
x-ray generator in separate basement rooms. The x-ray tube rected numerically for any small remaining distortion due to
excitation conditions are given in Table I and are always the instrumental function. It should be noted that in general
operated with the standards used in x-ray tube applications, the measured emission line is not a convolution of the
with excitation energies much higher than the corresponding ‘‘true’’ emission line and the instrumental window. Only un-
K-edge energy of the anode material. Line-shape variations der special conditions, selected according to our computer
with exciting energy are not expected in this region @7#. simulations and used in the measurements, can the measured
Temperature and air pressure were measured at each point data be approximated well by such a convolution, and the
and the intensity measured was corrected to correspond to evaluation of the ‘‘true’’ emission line profile can be done by
standard conditions (T5293.15 K, p5101.325 kPa!. The a deconvolution of the measured spectrum by the instrumen-
absorption in air and Be was calculated using absorption tal function @8#.
4556 HÖLZER, FRITSCH, DEUTSCH, HÄRTWIG, AND FÖRSTER 56

The ratio D u s /D u d of the spectral linewidth D u s to the K a , Co K b , Fe K a ) and about 100 000 counts ~Co K a , Ni,
beam divergence D u d ~spectrometer window! can be used as Cu! were collected at the maximum. For some lines a smaller
a measure of the achievable resolution. Note, however, that number of counts was used ~for Mn K b 25 000 counts and
in general, the spectrometer window function includes the for Fe K b 11 000 counts!. Rather than a single scan with a
effects of both the beam divergence ~collimator divergence long measuring time at each point a repeated scan technique
function! and the crystal’s angular acceptance function. Only was used, with a shorter time at each point and each scan
when the latter is of negligible width relative to the beam repeated from 20 to 40 times. With this strategy obvious
divergence, as is the case here, can the window function be runaway scans are easily excluded and the long-time stability
approximated well by the beam divergence D u d . The value of the generator is not stressed to its maximum extent. The
of D u s depends, among other parameters, strongly on the energy of the line is determined from the angular position of
Bragg reflection used. Since the calibrated silicon crystal has the intensity profile’s peak, through Bragg’s law. This profile
the surface orientation ~111!, the number of available reflec- is, under optimized conditions, a convolution of a wide in-
tions and especially of those providing a good resolution strumental function ~containing basically the wide collimator
~large Bragg angles! is very limited. D u s and D u d are given function and especially the wide spectral line itself! and the
in Table I. For a single-crystal spectrometer the ~horizontal! narrow dynamical reflection curve. The peak position deter-
slit widths of the collimator are the crucial factors for the mination is done best by calculating that intensity distribu-
width of the instrumental function. They should be as small tion @8# using the emission line shape ~with its exact absolute
as possible while still allowing a sufficiently high count rate energy scale still unknown! represented by the Lorentzian fit
for good statistics. The widths used for the line-shape mea- ~see Sec. III B!, and fitting it to the measured intensity dis-
surements were between 0.03 and 0.05 mm. The measured tribution, varying only its height and position @17#. Thus, to
emission lines were registered with about 1000 steps and determine the line position with high accuracy, its shape
measuring times between 100 and 200 s at each point. The must be known. Finally the determination of the peak posi-
total counts collected at a single point at the peak varied tion of the emission line in the absolute energy or wave-
between 20 000 and 50 000 counts for K a and 2000 and length scale, i.e., in eV or nm, respectively, is possible using
6000 for K b . In the measurements of the absolute peak po- the Bragg equation, the kinematical Bragg angle, and the
sition, discussed below, it was found advantageous to choose lattice parameter of the calibrated silicon crystal. The con-
these three parameters ~steps, time, and total counts! differ- version factor from the wavelength to the energy scale is @18#
ently. 8.065541~3!3105 ~m eV! 21 . A careful analysis of the accu-
Most of the K b spectra used for the determination of the racy of Bragg angle measurements and of the possible aber-
emission line shape were measured with two different reflec- rations was given in @19#. Since no deconvolution of the
tions. One was the same reflection used for measuring of the measured spectrum by the instrumental function is carried
K a doublet. This was not necessarily the one with the high- out in the case of the line position determination, the hori-
est resolution, but it guaranteed the highest possible compat- zontal slit width could be chosen larger than for the line-
ibility in the measuring conditions for the determination of shape determination in order to get more intensity. A width
the K b 1,3 /K a 1,2 intensity ratio. The second reflection was of ; 0.1 mm was selected.
optimized for high resolution ~see Table I!. The experimental
conditions were invariably set so that the effect of the instru-
III. RESULTS AND DISCUSSION
mental window function could be removed by deconvoluting
the measured data by the simulated instrumental function @8#. A. Introduction
The measured data were also corrected for the relative In this section we present the measured K a and K b emis-
change in absorption and in the integrated reflectivity over sion spectra of Cr, Mn, Fe, Co, Ni, and Cu. The ‘‘true’’
the wavelength range. No additional smoothing was applied emission line profiles were determined through deconvolu-
to the data. tion by the simulated instrumental function of the single-
The peak positions of the lines were determined on an crystal spectrometer @8#, as detailed above. Each line was
absolute energy-wavelength scale in separate measurements fitted by a sum of Lorentzians to get an accurate analytic,
according to the Bond method @15#. This method, developed though phenomenological, representation of the line shape
for the diffractometric measurement of lattice parameters of for use in x-ray line profile analysis and other applications.
nearly perfect crystals with high accuracy when the energy The line parameters, i.e., full width at half maximum
of the line is known accurately, may be used in reverse to ~FWHM!, index of asymmetry, and integrated intensity were
determine the absolute energy, when the lattice parameter of determined and are critically compared below with previous
the crystal is known accurately @16#. The optimum experi-
measurements, where available. Finally, the K a 2 /K a 1 and
mental conditions in that case are similar to those for the
K b 1,3 /K a 1,2 intensity ratios are calculated and discussed.
line-shape determination ~a large ratio D u s /D u d , and, in
particular, a large Bragg angle!, but the strategy for the mea-
B. Lorentzian fit
surement and the data analysis is different. The basic task is
to determine the kinematical Bragg angle of the selected An accurate analytical representation of the emission line
crystal reflection from the measured intensity distribution. shape is essential for a large number of practical applications
The basics of the measurement strategy are described in de- of x rays ~e.g., grain size and strain determination from
tail in @8,17#. 20–25 points were measured spanning the part single-crystal diffraction line profile analysis of polycrystal-
of the peak whose intensity I is >80% of the maximum. The line samples @10#, structure refinement from powder diffrac-
slit width is selected so that about 50 000 counts ~Cr, Mn tion patterns, etc.!. In both the classical and quantum theory
56 K a 1,2 AND K b 1,3 X-RAY EMISSION LINES OF THE . . . 4557

the natural shape of an emission line in the frequency or to a correct background determination. The precision of the
energy scale is a Lorentzian @20#. This is observed in practice fit is characterized by the weighted R factor R w , defined as
for a large number of K a emission lines. They are highly @23#

F G
symmetric and the K a doublet structure can be represented
n 1/2
accurately by a superposition of two Lorentzians, one each
for the K a 1 and the K a 2 lines @21#. (
i51
w i @ F i ~ obs! 2F i ~ calc!# 2
By contrast, it is well known that the experimentally ob- R w5 .
n
served K a and K b emission lines of the 3d transition metals
are pronouncedly asymmetric. Considering the basic Lorent- ( w i F i~ obs!
i51
2

zian shape of emission lines it was proposed to fit the shape

of those asymmetric lines empirically by a superposition of
Here the weighting variable w i 51/s 2i and s 2i is the vari-
several Lorentzians. The high precision achievable by this
ance due to the counting statistics. F i (obs) and F i (calc) are
approach was already demonstrated for the Cu K a and the
corresponding values on the measured and calculated line-
Cu K b spectra @5,8,9#. Each Lorentzian i is represented by
shapes, respectively. The complete set of parameters for a
three parameters: the position of the maximum E i , the rela-
multiple Lorentzian representation of the emission line
tive ~compared to the maximum of the emission line! peak
shapes of Cr, Mn, Fe, Co, Ni, and Cu are given in Tables II
intensity I i , and the full width at half maximum W i . For a fit
and III together with the R w values. The accuracy of the fits
of the measured lines a linear background ~parameters A and
is better for the K a doublet (R w <1.1%! than for the K b 1,3
B) has to be taken into account. Consequently the ‘‘true’’
line (R w <2.1%!. The worst fit is obtained for Fe K b with
~i.e., after the correction for the instrumental broadening!
R w 52.1%. Examples for the fits of the K a 1,2 and K b 1,3
emission spectrum S(E) is represented by
emission lines of Cr, Fe, and Ni are presented in Fig. 1. The
n high quality of the fit can be seen by the small and nonsys-
I
S~ E !5 ( i
i51 11 @~ E2E ! /W # 2
1BE1A. ~1! tematic residuals. The largest deviations between the mea-
sured and fitted profiles are obtained at the maximum, usu-
i i
ally with the fitted intensity being higher than the measured
In the first step of the fit the number n and the parameters one. The small regular oscillations, observed in the tails of
of the Lorentzians are selected manually to start the numeri- some of the emission lines, are obvious relics of the Fourier
cal refinement with optimal initial values of the parameters. deconvolution procedure.
This process starts always with the assumption of a doublet The values presented in Tables II and III allow a simple
for K a and for K b . The number of Lorentzians is increased and accurate analytic representation of the most frequently
in steps of 1 until either the integrated intensity of the last used x-ray emission lines from x-ray tubes. In this empirical
Lorentzian added is smaller than 1% of the integrated inten- decomposition no attempt was made to take into account the
sity of the entire emission line or the differences between physical origin of the asymmetry of these emission lines. A
measured and fitted profile over the entire energy range are successful physical interpretation of the observed asymmetry
smaller than the standard deviation 2s given by the measure- was already published @5# for Cu. A similar study for the
ment statistics. other lines presented here is in progress.
In the second step of the fit the approximate values of the
parameters for each Lorentzian and the background are nu-
C. Absolute wavelength determination
merically refined using the Levenberg-Marquard algorithm
@22#. The standard approach is slightly modified to improve The basic problem of high-precision x-ray spectroscopy is
the convergence of the fit. An optimization cycle is divided the extension of the metric ~or, now, the electromagnetic!
into n subcycles. In every subcycle the parameters of only a scale to the x-ray wavelengths @24#. Tabulations for x-ray
single Lorentzian are refined and fitted to an ‘‘artificial’’ pro- K a and K b emission line wavelengths have existed for sev-
file. It is constructed by removing the contributions of eral decades ~cf. Bearden @25#, whose data are identical with
n21 Lorentzians ~except the one to be optimized in this the data in Vol. IV of @23#! and the wavelengths or energies,
cycle! from the measured profile. The convergence limit respectively, are quoted with precisions extending down to
achieved is determined by the weighted sum of squares «: the ppm level. However, it must be noted that in almost all
cases this is only a relative precision, taking into account
m
1 only the accuracy of the angular position through the Bragg
«5 ( @ F i~ E i ,A,B,I 1 ,E 01 ,W 1 , . . . ,I n ,E 0n ,W n !
i51 M i
law. As x-ray wavelength determination invariably involves
crystal diffraction, an accurate determination of the absolute
2M i ~ E i !# 2 . ~2! wavelength on a metric scale requires that the lattice param-
eters of the crystal be known on an absolute metric scale to
M i is the measured intensity and F i is the intensity of the fit an accuracy higher than that of the wavelength to be deter-
profile, i.e., a function of the parameters of all Lorentzians mined. Almost all of the tabulated values @23,25# are based
and the background for each of the m points of the experi- on measurements that do not fulfill this requirement and are
mental emission line profile. This convergence criterion likely to be in error by more than 1025 on an absolute scale.
takes into account relative deviations between the measured The origin of this relatively low accuracy was for a long time
and the modeled profiles and is also sensitive to small ~ab- the lack of a proper absolute ‘‘ruler’’ for atomic-scale length
solute! deviations between the measured and the fitted pro- measurements. In the last 10–20 years extensive efforts were
files at the background level, i.e., it gives a higher sensitivity made at several national standards institutes, in particular at
4558 HÖLZER, FRITSCH, DEUTSCH, HÄRTWIG, AND FÖRSTER 56

TABLE II. Positions E i , widths W i , amplitudes I i , and integrated intensities I int obtained for the individual Lorentzians from a fit of
each measured line by the sum of the minimal number of Lorentzians that provides a reasonable fit to the data. Weighted R factors obtained
for the fits are also shown.

Ka1 Ka2
Peak E i ~eV! W i ~eV! Ii I int Peak E i ~eV! W i ~eV! Ii I int

Cr a 11 5414.874~2! 1.457~2! 0.822~2! 0.378 a 21 5405.551~3! 2.224~4! 0.386~1! 0.271

a 12 5414.099~6! 1.760~7! 0.237~2! 0.132 a 22 5403.986~3! 4.740~50! 0.036~1! 0.054
a 13 5412.745~16! 3.138~20! 0.085~1! 0.084
a 14 5410.583~30! 5.149~51! 0.045~1! 0.073
a 15 5418.304~38! 1.988~58! 0.015~1! 0.009
Rw 1.1%

Mn a 11 5898.853~2! 1.715~2! 0.790~2! 0.353 a 21 5887.743~3! 2.361~5! 0.372~2! 0.229

a 12 5897.867~6! 2.043~7! 0.264~2! 0.141 a 22 5886.495~13! 4.216~21! 0.010~1! 0.110
a 13 5894.829~22! 4.499~33! 0.068~1! 0.079
a 14 5896.532~15! 2.663~20! 0.096~1! 0.066
a 15 5899.417~17! 0.969~20! 0.007~3! 0.005
Rw 1.1%

Fe a 11 6404.148~2! 1.613~3! 0.697~2! 0.278 a 21 6391.190~4! 2.487~6! 0.339~1! 0.207

a 12 6403.295~4! 1.965~5! 0.376~2! 0.182 a 22 6389.106~22! 2.339~19! 0.060~1! 0.066
a 13 6400.653~19! 4.833~26! 0.088~1! 0.106 a 23 6390.275~33! 4.433~12! 0.102~1! 0.065
a 14 6402.077~12! 2.803~15! 0.136~1! 0.094
Rw 0.9%

Co a 11 6930.425~2! 1.795~2! 0.809~2! 0.378 a 21 6915.713~3! 2.406~4! 0.314~1! 0.197

a 12 6929.388~6! 2.695~8! 0.205~1! 0.144 a 22 6914.659~7! 2.773~10! 0.131~1! 0.095
a 13 6927.676~12! 4.555~15! 0.107~1! 0.127 a 23 6913.078~23! 4.463~35! 0.043~1! 0.050
a 14 6930.941~27! 0.808~27! 0.041~2! 0.088
Rw 0.5%

Ni a 11 7478.281~2! 2.013~2! 0.909~2! 0.487 a 21 7461.131~4! 2.674~5! 0.351~1! 0.250

a 12 7476.529~13! 4.711~15! 0.136~1! 0.171 a 22 7459.874~15! 3.039~21! 0.079~1! 0.064
a 23 7458.029~48! 4.476~78! 0.024~1! 0.028
Rw 1.0%

Cu a 11 8047.837~2! 2.285~3! 0.957~2! 0.579 a 21 8027.993~5! 2.666~7! 0.334~1! 0.236

a 12 8045.367~22! 3.358~27! 0.090~1! 0.080 a 22 8026.504~14! 3.571~23! 0.111~1! 0.105

Rw 0.7%

the National Institute of Standards and Technology, Gaith- measurements of the 3d transition metal emission lines with
ersburg ~MD! @26,27# and the Physikalisch-Technische a comparable precision.
Bundesanstalt, Braunschweig ~Germany! @12~a!# but also in The general strategy for the determination of the wave-
Japan @12~b!# and Italy @12~c!#, to bridge the gap in absolute lengths is presented in Sec. II B. The measured maximum
wavelength determination between optical and x-ray wave- positions of the K a 1,2 and the K b 1,3 lines are listed in Table
lengths. A combined x-ray and light interferometric method IV, along with those from the widely used tabulation of
capable of measuring lattice parameters on an absolute met- Bearden @25#. These data are presented in the original nota-
ric scale to a relative accuracy of 1027 or better was devel- tion, using the Å * unit and the probable error ~p.e.!. They
oped and the lattice parameter of highly pure silicon crystals are also converted to SI units in the next column using
were determined @12,26,27#. The silicon crystal used in our l(nm)5l(Å * )0.10001481(92) of the 1986 CODATA
measurement was calibrated based on such a measurement. least-squares adjustment @85# with the error listed as one
To date, only a few x-ray wavelengths were determined standard deviation ~1.48p.e.!. As can be seen, our values are
on an absolute scale using such calibrated crystals to an ac- systematically larger than the original values of Bearden
curacy Dl/l'1026 . The value with the highest accuracy is ~with the exception of Fe and Co K a 2 , which have negative
that of the Cu K a 1 , which was measured to an accuracy of deviations smaller than 431026 ) by an amount ranging up
@16# Dl/l'331027 . The data presented here are, to our to 2.4231025 ~Cu K a 2 ). A large part of the deviation is, of
knowledge, the first set of systematic absolute wavelength course, due to Bearden’s use of the x-unit, rather than the
56 K a 1,2 AND K b 1,3 X-RAY EMISSION LINES OF THE . . . 4559

TABLE III. Same as Table II, but for the K b 1,3 spectra. The weighted R factors here are 1.5–2 times worse than those obtained for the
K a spectra.

K b 1,3
Peak E i ~eV! W i ~eV! Ii I int

Cr ba 5947.00~1! 1.70~1! 0.670~5! 0.307

bb 5935.31~9! 15.98~21! 0.055~1! 0.236
bc 5946.24~1! 1.90~2! 0.337~5! 0.172
bd 5942.04~5! 6.69~8! 0.082~1! 0.148
be 5944.93~3! 3.37~4! 0.151~2! 0.137

Rw 1.7%

Mn ba 6490.89~1! 1.83~1! 0.608~5! 0.254

bb 6486.31~7! 9.40~9! 0.109~2! 0.234
bc 6477.73~8! 13.22~15! 0.077~1! 0.234
bd 6490.06~2! 1.81~2! 0.397~5! 0.164
be 6488.83~3! 2.81~4! 0.176~4! 0.114

Rw 1.6%

Fe ba 7046.90~9! 14.17~17! 0.107~2! 0.301

bb 7057.21~2! 3.12~2! 0.448~5! 0.279
bc 7058.36~1! 1.97~2! 0.615~7! 0.241
bd 7054.75~6! 6.38~8! 0.141~3! 0.179

Rw 2.1%

Co ba 7649.60~8! 3.05~1! 0.798~4! 0.449

bb 7647.83~3! 3.58~3! 0.286~4! 0.189
bc 7639.87~8! 9.78~12! 0.085~1! 0.153
bd 7645.49~6! 4.89~8! 0.114~3! 0.103
be 7636.21~20! 13.59~37! 0.033~1! 0.082
bf 7654.13~11! 3.79~16! 0.035~2! 0.025

Rw 1.6%

Ni ba 8265.01~1! 3.76~2! 0.722~4! 0.450

bd 8263.01~2! 4.34~3! 0.358~4! 0.258
bc 8256.67~10! 13.70~16! 0.089~1! 0.203
bd 8268.70~7! 5.18~9! 0.104~3! 0.089

Rw 1.9%

Cu ba 8905.532~2! 3.52~1! 0.757~3! 0.485

bb 8903.109~10! 3.52~1! 0.388~2! 0.248
bc 8908.462~20! 3.55~3! 0.171~2! 0.110
bd 8897.387~50! 8.08~8! 0.068~1! 0.100
be 8911.393~57! 5.31~8! 0.055~2! 0.055

Rw 1.0%

true SI standard, not available to him at that time. The SI D. Linewidth and index of asymmetry
converted Bearden data shows a much better agreement with 1. Introduction
our data, with almost random deviations. Nevertheless, upon
careful examination a mean deviation of 24.631026 is ob- A frequently used and convenient ~but not very accurate!
served between Bearden’s set and ours, possibly due to a quantification of the line shape is by its full width at half
small remnant systematic deviation in the conversion factor. maximum ~FWHM! and its index of asymmetry. Both values
Thus, the present results provide a large set of accurate, well allow a comparison with other experimental results and a
calibrated x-ray wavelengths, which should prove useful for general classification of our measurements in relation to
a variety of precision measurements involving x rays. other reference data. Note, however, that the line shape can-
4560 HÖLZER, FRITSCH, DEUTSCH, HÄRTWIG, AND FÖRSTER 56

FIG. 1. Measured K a and K b emission lines of Cr ~a!,~b!, Fe ~c!,~d!, and Ni ~e!,~f! ~points!. A fit by a sum of a minimal set of
Lorentzians ~solid line! is also shown for each line, along with the individual Lorentzians of the set ~dashed lines!. The fit parameters are
given in Tables II and III. The fit residuals ~DIFF! are shown under each spectrum, with thin lines denoting the 62 s values of the data,
where s is the standard deviation due to the counting statistics.
56 K a 1,2 AND K b 1,3 X-RAY EMISSION LINES OF THE . . . 4561

TABLE IV. The peak wavelengths of the measured lines, on an absolute metric scale. Measurement uncertainties are given in brackets,
and the widely used values of Bearden @23,25#, listed in Å * with their probable error ~pe! values, are also shown. The fourth column shows
these values in nm units, using the relation l ~nm! 5 l ~Å * )0.100 001 481 @85# and the standard deviation 1.483pe. Note the part-per-
million level of accuracy of the present results.

Line Reflection hkl l Bearden ~Å * ) Corr. l Bearden ~nm! l ~nm! ~this work! E ~eV! ~this work!

Cr K a 1 331 2.289 70~2! 0.228 973~3! 0.228 972 6~3! 5414.81~1!

Cr K a 2 2.293 606~3! 0.229 364 0~5! 0.229 365 1~3! 5405.54~1!
Cr K b 1,3 2.084 87~2! 0.208 490~3! 0.208 488 1~4! 5946.82~1!

Mn K a 1 422 2.101 820~9! 0.210 185~2! 0.210 185 4~3! 5898.80~1!

Mn K a 2 2.105 78~2! 0.210 581~3! 0.210 582 2~3! 5887.59~1!
Mn K b 1,3 333 1.910 21~2! 0.191 024~3! 0.191 021 6~4! 6490.18~1!

Fe K a 1 333 1.936 042~9! 0.193 607~2! 0.193 604 1~3! 6404.01~1!

Fe K a 2 1.939 980~9! 0.194 001~2! 0.193 997 3~3! 6391.03~1!
Fe K b 1,3 1.756 61~2! 0.175 664~3! 0.175 660 4~4! 7058.18~3!

Co K a 1 531 1.788 965~9! 0.178 899~2! 0.178 899 6~1! 6930.38~1!

Co K a 2 1.792 850~9! 0.179 288~2! 0.179 283 5~1! 6915.54~1!
Co K b 1,3 620 1.620 79~2! 0.162 081~3! 0.162 082 6~3! 7649.45~1!

Ni K a 1 620 1.657 910~8! 0.165 794~2! 0.165 793 0~1! 7478.26~1!

Ni K a 2 1.661 747~8! 0.166 177~2! 0.166 175 6~1! 7461.04~1!
Ni K b 1,3 444 1.500 135~8! 0.150 016~2! 0.150 015 2~3! 8264.78~1!

Cu K a 1 444 1.540 562~2! 0.154 058 5~3! 0.154 059 29~5! 8047.83~1!
Cu K a 2 1.544 390~2! 0.154 441 3~3! 0.154 442 74~5! 8027.85~1!
Cu K b 1,3 553 1.392 218~9! 0.139 224~2! 0.139 223 4~6! 8905.42~4!

not be described precisely by a simple asymmetric profile in account multiple measurements and multiple fits of each
all cases. Such an approximation is quite good for the K a 1,2 spectrum. They are always ,0.05 eV for the FWHM. The
lines but a very bad one for the K b 1,3 line, where a substruc- errors for the index of asymmetry are ,0.05 eV for K a 2 and
ture due to the two unresolved lines of the doublet, the K b 1 , 0.07 eV for K a 1 . The FWHM and the index of asymme-
and K b 3 lines, separated by less than their combined half try are compared in Figs. 2 and 3 with available previous
widths, is clearly visible ~see Fig. 1!. Thus, we have avoided measurements, not for all of which error estimates are avail-
assigning an index of asymmetry to the K b 1,3 lines. able.
In our measurements the FWHM and the index of asym-
metry were determined from the background-corrected 2. The K a linewidths
Lorentzian fits ~see above!. The index of asymmetry is de- As can be seen in Fig. 2, the previous measurements of
fined as the ratio of the half widths at half height on the low the FWHM values are rather broadly scattered, reaching a
and high energy sides of the peak @28#. In Table V the factor of two in extreme cases @Cr K a 1 in Fig. 2~a!#. Fur-
FWHM of K a 1 , K a 2 , K b 1,3 and the index of asymmetry for thermore, the deviations exceed the error bars, in the few
K a 1 , K a 2 are presented. The experimental uncertainties in cases where these are available, indicating probable system-
the values cited in the table were determined taking into atic errors. Also, a tendency of ‘‘clustering’’ is observed for

TABLE V. The full width at half maximum ~FWHM!, index of asymmetry, and corrected integrated
intensity ratios for the measured spectra.

FWHM ~eV! Index of asymmetry I(K a 2 )/I(K a 1 ) I(K b 1,3)/I(K a 1,2)

Ka1 Ka2 K b 1,3 Ka1 Ka2

Cr 1.88 2.52 2.53 1.38 1.01 0.139

Mn 2.47 2.92 2.97 1.49 1.17 0.51 0.139
Fe 2.55 3.14 3.53 1.66 1.24 0.51 0.144
Co 2.33 3.18 4.37 1.38 1.33 0.52 0.137
Ni 2.24 3.16 5.40 1.18 1.21 0.52 0.147
Cu 2.35 3.41 5.92 1.07 1.36 0.52 0.141
4562 HÖLZER, FRITSCH, DEUTSCH, HÄRTWIG, AND FÖRSTER 56

FIG. 3. Asymmetry indices for the K a 1 ~a! and K a 2 ~b! lines of

Cr, Mn, Fe, Co, Ni, and Cu, derived from the present measurements
and previous studies.

FWHM of the spectrometer window from the FWHM of the

measured emission line @29–31#. This approach is exact if
both the emission line profile and instrumental function have
a Lorentzian shape. Normally this is not the case. For emis-
sion lines with an obvious asymmetric shape it becomes less
exact but still gives a reasonable approximation even for the
K b line shapes @29–31#. The combined fit and deconvolu-
tion method of Lee and Salem @32# seems to disagree with
the general Z-dependent trend of all other measurements
shown in Figs. 2~a! and 2~b!.
Our experience indicates that the determination of the
FIG. 2. The width ~FWHM! of K a 1 ~a!, K a 2 ~b! and K b ~c! FWHM and the index of asymmetry is especially sensitive to
lines of Cr, Mn, Fe, Co, Ni, and Cu, as derived from previous a precise correction for the instrument broadening. In the
studies ~listed in the legend by first author name and reference case of an optimally adjusted spectrometer the instrument
number! and the present measurements. Note the weak maximum function is nearly symmetric and narrow ~see @8#!. A
near Fe for the K a lines and the monotonic increase with Z for the
‘‘broad’’ symmetric instrumental function tends to suppress
K b lines.
the line asymmetry and give systematically too small values
for the index of asymmetry. This effect cannot be corrected
most of the data points with a few obvious ‘‘runaway’’ after the measurement. Only experiments with a narrow
points. The fact that most of the previous measurements spectrometer window in an optimized spectrometer setup
yield higher values for the widths than ours indicates a prob- provide a precise determination of the index of asymmetry
able neglect of, or incomplete correction for, the instrumen- and the FWHM. Our data are determined with the same op-
tal broadening. Only few of the previous data were corrected timized spectrometer setup and with a well-defined correc-
for an instrumental broadening, although it was realized tion procedure for the narrow spectrometer window. They
early that the instrumental broadening has a significant effect have a higher accuracy than most of the other reference data
@29# and even simple corrections can lead to a significant ~see error bars in Figs. 2 and 3!. Additionally they fit into the
improvement in the results. The ‘‘true’’ FWHM of the emis- above-mentioned ‘‘clusters’’ and present the same general
sion line can be determined by a simple subtraction of the Z-dependent trends as the other more comprehensive refer-
56 K a 1,2 AND K b 1,3 X-RAY EMISSION LINES OF THE . . . 4563

ence data @28–31,33,34#. General trends for the Z depen- ments of Onoue and Suzuki @44#. The spread in the reported
dence of the FWHM and the index of asymmetry for the 3d values of the index of asymmetry is even larger for the K a 2
transition metals are therefore deduced in the following us- line than for the K a 1 as shown in Fig. 3~b!. Despite the large
ing our data. spread, a general trend for the Z dependence of the index can
The FWHM of the K a lines show a good agreement with be observed. For K a 1 a distinct maximum occurs for Fe,
most of the previous data. Some of these are systematically with the index decreasing monotonically towards both Cr
too large, as those of Edamoto @35# and Lehnert @36#. The and Cu. This maximum coincides with the maximum in the
good correspondence with the early data of Parratt @29#, FWHM, shown in Fig. 2~a!. The Z dependence of the asym-
which were corrected for resolution effects, show the effec- metry of K a 2 peak shows two characteristic deviations from
tivity of even a simple correction procedure. Good agree- K a 1 . The maximum asymmetry is shifted to Co and there
ment is also found with the measurements of Bearden @37# as exists a distinct minimum for Ni, which does not occur for
analyzed by Thomsen @21#. In addition, the Z-dependent K a 1 . The asymmetry increases once more for Cu. Both the
trend of the more comprehensive data sets @28–30,33,34# maximum for Co and the minimum of the asymmetry for Ni
agree with those in our data. Generally the FWHM of the are confirmed by most of the other measurements and should
K a 1 line increases from Cr to Fe, has a distinct local mini- be, therefore, considered to be real effects.
mum for Ni and increases once more for Cu. This trend is the
same for the FWHM of the K a 2 line although there a plateau 5. Z-dependent trends
is observed in the Fe-Ni region rather than a distinct mini- Several attempts have been made in the past to give a
mum. The FWHM of the K a 2 line is larger than that of the physically motivated, or even a purely empirical functional
K a 1 line by 0.5–1 eV. This can be accounted for by the representation for the experimentally observed Z dependence
reduced lifetimes due to the L 2 -L 3 M 4,5 Coster-Kronig tran- of the lines’ characteristic parameters, such as the width,
sition, which is forbidden in the free atom but allowed in the asymmetry index, etc., over a broad Z range, for use in prac-
solid @6,38,39#. tical applications. These attempts were successful for the
widths @33,45# and the K a doublet separation @46#. How-
3. The K b linewidths ever, distinct deviations between the smooth representation
and the measured values were observed in all these forms for
For the FWHM of K b 1,3 the number of available previous
20<Z<30, i.e., in the region of the 3d transition metals
data is rather limited @30,35,40–43#. They are presented in
~and, to a smaller extent, also for the 4d transition elements!.
Fig. 2~c!. Their agreement with our data is even better than
In this Z range the measured widths exceed significant the
for the K a widths. This is most probably due to the fact that
the considerably larger FWHM of the line reduces the rela- widths extrapolated from lower- and higher-Z elements.
tive contribution of the instrumental broadening. Neverthe- These deviations were ascribed to the unique electronic
less, the early data of Edamoto @35# and Tsutsumi @40# are structure of the atoms in this range, which have a partly or
systematically larger than recent measurements. This results fully populated 4s shell and an open 3d shell, giving rise to
most probably from instrumental effects. Unlike the K a complicated transition structures in the x-ray emission pro-
cess. Similar anomalous deviations were observed in the
lines the FWHM of the K b 1,3 peak increases monotonically
XPS linewidths of the same elements @39,47#. Finster @4# and
from Cr to Cu, indicating that the nonmonotonic behavior of
Pessa @28# attempted to correlate these deviations with the
the K a widths is probably due to the behavior of the L level
number of unpaired 3d electrons in the atom, as calculated
widths rather than the K level ones. The smaller overlap of
by them. While neither calculation result in an accurate cor-
the M and K wave functions, as compared to the K and L
respondence, they show a reasonable qualitative correlation
ones, may reduce the relative influence of possible similar-
with the Z dependence of the deviations. In particular, the
sized nonmonotonic contributions originating in the final
number of unpaired electrons is maximal for iron, which
state level widths.
corresponds well to the experimentally observed maxima of
the FWHM for K a 1 and K a 2 and for the maximum of the
4. Index of asymmetry of the K a lines
index of asymmetry for K a 1 .
Our results and previous data for the index of asymmetry Tsutsumi @3# proposed that the anomalous widths and
are presented in Figs. 3~a! and 3~b! for K a 1 and K a 2 , re- asymmetric line-shape results from an exchange interaction
spectively. The spread of the data is large, probably due to between electrons of the incomplete 3d shell and those of the
the high sensitivity of the index of asymmetry to the width of 2 p shell, left open by the x-ray emission process. This will
instrumental function as mentioned above. A broad instru- split the L levels into two sublevels each, and, thus, also each
mental function should result in a systematically too low of the K a lines into two close lines. Since the separation of
asymmetry. The data of Meisel @33# and Pessa @28# for the the two lines is smaller than their combined widths and their
narrow and highly asymmetric K a 1 line indeed show this intensities are unequal, the splitting will yield the skewed
tendency; they are systematically smaller than the majority line shapes observed. While this model leads to a qualitative
of the other measurements in Fig. 3~a!. It is clear from the improvement in the line-shape fits, and hence also of the
results that the asymmetry is generally larger for the K a 1 widths, the theoretically expected splitting is usually too
peak. Our indices of asymmetry for the K a 1 peak are larger small to explain the experimentally observed asymmetry and
than most of the other data. This is particularly true for Mn, width. Furthermore, it predicts symmetrical line shapes for
Fe, and Co, which have the highest asymmetry among the 3d atoms having complete 3d shells, like Cu, in obvious con-
transition elements. Our results agree well with those of tradiction with the measurements @48#. XPS measurements
Blochin @31#, Sorum @34#, and the high-resolution measure- @39,49,50# demonstrated that the excess width of the x-ray
4564 HÖLZER, FRITSCH, DEUTSCH, HÄRTWIG, AND FÖRSTER 56

lines results from a broadening of the L II and L III levels

beyond their intrinsic lifetime widths due to strong Coster-
Kronig ~CK! transitions. These are allowed in the solid
phase, though energetically forbidden in the isolated atom
~gas phase! case @39#. The inclusion of the CK transition
greatly improves the agreement of the total line shape and its
width with the experiment. Pease @51#, however, showed,
based on measured soft x-ray emission spectra of the
2 p→3s transition, that the CK transitions are insufficient to
account for the L level widths over the full 20<Z<30 range,
and that additional ~unspecified! nonlifetime effects need to
be included.
Clearly, the simplified approach of addressing only the
width and asymmetry of a line, is insufficient to account for
the shapes of complex spectra such as those of the 3d tran- FIG. 4. The K a 2 /K a 1 intensity ratio, as obtained in the present
sition metals. Thus, attempts have been made to fit the full measurements and previous studies. No reliable value could be de-
emission spectrum by a model based on various ab initio, or rived from the present measurements for Cr, for reasons detailed in
semiempirical, calculated transition arrays. Using relativistic the text.
Dirac-Fock ab initio calculations and high-resolution spec-
tral measurements, we have shown recently @5# that both the multiplet transition. Applying the Burger-Dorgelo-Ornstein
Cu K a and Cu K b emission spectra could be accurately rule @20# to this doublet an intensity ratio of 0.5 is expected.
reproduced assuming contributions from two transitions This value should be independent of the atomic number Z.
only: the diagram 1s→lp and the 3d spectator hole 1s3d The early measurements of Williams @54# for a selection of
→lp3d, where l52,3 for the K a and the K b lines, respec- twenty elements from Z524 to Z552 confirm these predic-
tively. The validity of this approach is further supported by tions. The results are statistically distributed and the average
observations of a strong energy dependence of the Fe K a K a 2 /K a 1 intensity ratio is determined to be 0.5003. Later
@52# and Ar K b @53# emission line shape on the exciting theoretical calculations @55# taking into account exchange
photon energy when photoexcited close to the K edge. For interactions presented slightly larger and also weakly
photons with insufficient energy to excite the initial 1s3d, Z-dependent values. These results are supported quantita-
but sufficient to do so for the lower-energy 1s hole state, a tively by the experimental observations of McCrary @56# and
much more symmetric and narrower line is observed. An Salem @57,58#, which also obtained values slightly larger
analysis of the emission lines of 3d transition metals other than 0.5 for the K a 2 /K a 1 intensity ratio. To the best of our
than Cu, based on Dirac-Fock ab initio calculations, as well knowledge, more recent experimental studies of the
as excitation energy-dependent measurements of emission K a 2 /K a 1 intensity ratio of the 3d transition metals are not
spectra near threshold are in progress @7#. available in the literature. For a verification of the theoreti-
cally predicted small deviations of K a 2 /K a 1 intensity ratio
E. The K a 2 /K a 1 and K b 1,3 /K a 1,2 integrated intensity ratios from 0.5 precise corrections of the measured intensities are
necessary. This is especially true for the energy-dependent
The accurate knowledge of the K a 2 /K a 1 and the K b 1,3/ integrated reflectivity of the crystal, which can be calculated
K a 1,2 integrated intensity ratios is required for a number of from the dynamical theory of x-ray diffraction @59#, the
practical applications of x rays, e.g., structure determination energy-dependent absorption of the radiation in the target
from powder diffraction patterns, like the Rietveld refine- itself, and the intensity contributions from satellite transi-
ment technique, etc. In precision single crystal structure re- tions included in the doublet lines ~see, for example,
finement methods a quantitative analysis of the intensity of a @5#!.These energy-dependent corrections are small for the
large number of reflections is required, over a large Bragg K a 2 /K a 1 intensity ratio, since the energies of both doublet
angle range. Intensity contributions from simultaneously ex- lines are only slightly different. They are considerably larger
cited characteristic emission lines, i.e., from the K a doublet for the K b 1,3 /K a 1,2 intensity ratio as we discuss below.
and/or the K a , b lines for a particular reflection cannot al- A more serious problem for the accurate determination of
ways be resolved experimentally. This is particularly true for the K a 2 /K a 1 intensity ratio is the partitioning of the inten-
small Bragg angles. The determination of the quantitative sity of the spectrum between the two lines. The intensity
contribution of each emission line to the measured reflection between the doublet lines is nonzero and thus its assignment
requires, in this case, an accurate foreknowledge of the in- is done using additional assumptions, such as an identical
tensity ratios K a 2 /K a 1 and K b 1,3/K a 1,2 . Finally, these ra- peak shape and/or a fixed ratio of the FWHM of the doublet
tios depend sensitively on the atomic structure. Thus they peaks @56#. These assumptions are not universally valid, par-
have been widely used also for critical evaluation of atomic ticularly in our case where the shapes of the K a 1 and the
structure model calculations. We now discuss the values of K a 2 peaks are significantly different ~see Table V!. The dif-
these ratios as obtained in our measurements. ferences are especially large for Mn, Fe, and Co.
In our experiments the energy-dependent corrections of
1. The K a 2 /K a 1 integrated intensity ratio
the integrated intensity and the absorption are performed ac-
Our results, along with previous measurements, are plot- cording to @8#. The intensity assignment to each of the dou-
ted in Fig. 4. The K a doublet transition is a special case of a blet peaks is done using the results of the decomposition of
56 K a 1,2 AND K b 1,3 X-RAY EMISSION LINES OF THE . . . 4565

the doublet into single Lorentzians. The intensity of each dynamical theory based computer code DIXI @63#.
doublet peak is the sum of all Lorentzians included in this The background is subtracted from the spectra by least-
peak. The results are shown in Fig. 4. They show a very squares fitting @22# of Lorentzians and a linear background.
good agreement with the theoretical predictions for Mn and This is crucial because small changes in the background
Fe, while the values determined for Co, Ni, and Cu exceed cause significant changes in the intensity ~in the percent
Scofield’s @55# theoretical values by 1–2 %, which are, nev- range!. The last correction concerns the self-absorption of
ertheless, still within the 2% experimental uncertainties. The the radiation produced in the x-ray tube anode. Because of
intensity partitioning method fails completely for chromium, the thick target, it is not possible to neglect this effect. We
since the Cr K a doublet in our measurement has a significant used the approach of Sewell et al. @64# who deduced an ana-
intensity in between the two peaks and a unique, objective lytical expression for the correction by using results of tracer
partitioning of this contribution between the two doublet experiments. The uncorrected and corrected values for the
peaks is not possible. We wish to point out that the K b /K a intensity ratios, along with the various correction
K a 2 /K a 1 intensity ratios in Table V hold for typical exci- factors, are given in Table VI. The formula for the corrected
tation conditions of x-ray tubes and therefore can be used in ratio is
analyzing diffraction and other data measured with such
tubes. Also note that both doublet lines include contributions I K b ~ cor! I K b ~ meas! S a T a R a
from satellite transitions and an exact determination of the 5 ,
I K a ~ cor! I K a ~ meas! S b T b R b
intensity ratio for the diagram transitions 1s→2p 1/2 and 1s
→2p 3/2 requires a separation of these satellite contributions where S, T, and R are the self-absorption, the transmission
using a suitable model for the whole transition array. This through air and Be, and the integrated reflectivity of the crys-
was demonstrated for the Cu K emission spectra, where con- tal, respectively. Strictly speaking, the measured intensities,
tributions due to spectator transitions of 20–30 % were de- I(meas), are given by an integral of the intrinsic, undistorted
termined @5#. In that case, both doublet lines were nearly intensity over the angular and energy spreads of the incom-
equally contaminated by these satellite intensity contribu- ing beam, with each intensity value multiplied by the value
tions and the pure I(1s→2p 1/2)/I(1s→2p 3/2) ratio is deter- of S, T, and R at that energy and incidence angle. Thus, to
mined to be 0.53 6 0.01, which is in good agreement with obtain the corrected intensity values, I(cor), inversion of this
the value presented in Table V. integral is required. However, as the angular width of the
crystal reflectivity curve is extremely narrow, we have ap-
2. The K b 1,3/K a 1,2 integrated intensity ratio
proximated the intensity variation over this range to be con-
In contrast to the K a 2 /K a 1 ratio, almost all published stant, thus allowing the integration to be carried out and ef-
measurements of the K b 1,3 /K a 1,2 ratio have been done us- fective correction factors to be expressed as individual
ing semiconductor detectors ~SD!, since the separation of multiplicative factors, yielding the equation above. While
these spectra, 500–900 eV for the transition elements, is con- this is admittedly an approximation, it should be a very good
siderably larger than the ;200 eV resolution of the SD. We one, considering the very narrow widths, a few arcsec only,
were able to find in the literature only a single experiment of the reflectivity curves of the perfect silicon crystal used in
using a crystal spectrometer ~CS! @58#. This state of affairs is the measurements. The necessity of these energy-dependent
due to the difficulties in correcting the CS-measured data for corrections becomes clear when comparing the uncorrected
the various effects discussed above, and, in particular, ac- and corrected K b /K a intensity ratios listed in the Table VI.
counting for the varying reflection properties of the crystal Most of the K b -emission lines were measured using two
@60#. In spite of this, the CS is preferable in principle to the different crystal reflections. The Cu K a spectrum was mea-
SD for such measurements due to its much higher resolution, sured with 3 different reflection orders, namely, 333, 444,
as compared to a SD, which reduces drastically the spurious and 620. These independent measurement allow a better as-
background due to bremsstrahlung and allows a clear sepa- sessment of the measurements the errors, in particular those
ration of x-ray satellites, such as the K a 3,4 or the K b 2,5 lines. originating in the crystal parameters. We estimated a preci-
As shown in the preceding section, the reflection properties sion of better than 4%. By using the same reflection in the
of a perfect crystal could be handled numerically. Further- K a and K b measurements the overall error in the ratio
more, for extracting exact ratios from a SD measured data should not exceed 5%, as errors of multiplicative nature can-
significant corrections are required as well @61,62#. The cel. The influence of the energy of the exciting electrons on
energy-dependent efficiency of the SD must be known and the K b /K a intensity ratio is negligible @65#. A comparison
the corrections for self-absorption and absorption in win- of this ratio for different excitation mechanisms is given by
dows and air, mentioned below, must also be carried out Perujo et al. @66#.
when using a SD. In Fig. 5 our values are compared with previous results.
The two CS-measured K a 1,2 and K b 1,3 spectra were re- Only theoretical results and experiments employing electron
corded during two different runs. To avoid complications the or photon excitation are included, since the prominent mul-
measuring conditions ~tube voltage and current, slit widths! tiple ionization in the case of heavy-ion excitation strongly
remained the same for both runs. The corrections applied to alters the K b /K a intensity ratio @32,67#. Also included in
the raw data after the measurements are as follows: the ab- the comparison are theoretical data from Scofield, Manson,
sorption in air and in the Be windows of the x-ray tube and and Jankowski. Of the theoretical results, Manson’s @68#
the detector for the different peak emission wavelengths was simple nonrelativistic Hartree-Slater calculations clearly un-
calculated using absorption coefficients from @14#. The re- derestimate almost all other experimental and theoretical re-
flection R int of the silicon crystal was calculated using the sults. The more sophisticated relativistic Dirac-Fock calcula-
4566 HÖLZER, FRITSCH, DEUTSCH, HÄRTWIG, AND FÖRSTER 56

TABLE VI. The raw and corrected K b 1,3 /K a 1,2 intensity ratio. The correction factors discussed in the
text are also listed.

Line and I(K b )/I(K a ) Transmission Self-absorption Reflectivity I(K b )/I(K a )

reflection raw ratio ratio ratio corrected
T a /T b S a /S b R a /R b

Cr K a ~331! 0.427 0.869

Cr K b ~331! 0.223 1.681 0.139
Cr K b ~422! 0.417 0.949 0.147

Mn K a ~422! 0.512 0.920

Mn K b ~422! 0.131 2.354 0.140

Fe K a ~333! 0.588 0.929

Fe K b ~333! 0.136 1.929 0.143
Fe K b ~531! 0.231 1.206 0.152

Co K a ~531! 0.655 0.931

Co K b ~531! 0.066 3.390 0.137

Ni K a ~620! 0.713 0.933

Ni K b ~620! 0.062 3.573 0.147
Ni K b ~444! 0.190 1.190 0.150

Cu K a ~444! 0.761 0.903

Cu K b ~444! 3.391 0.141

tions of Scofield @55# and of Jankowski and Polasik @69#, F. Fine structure of the K a doublet lines
which also include exchange effects, agree much better with Several studies reported observations of fine structure in
the experimental results. The recent results of Kücükönder et
the K a lines of the iron group elements. Schnopper and
al. @70# considerably underestimate all other measurements.
Kalata @71# reported such an observation for chromium, and
This may be due to an overcorrection for the particle size of
further reports include chromium ~Schnopper and Kalata
the powdered samples, used in the measurements. As can be
seen, our results agree well with most of the other values, @71#, Shah and Das Gupta @72,73#, and Priest @74#!, iron
except for Fe and Ni, which seem to be too high. In view of ~Priest @74#!, cobalt ~Shah and Das Gupta @73#!, and copper
the error bars, it is not clear if this is a real effect. We note, ~Sauder et al. @75#!. By contrast, a large number of early and
however, that Berényi et al. @65# also find an enhanced ratio recently published studies failed to observe such structure
for Ni. Our results are consistently at the upper part of the @28,29,31,34,76–78#. In particular, no evidence for a fine
experimental values’ spread, which could hint at a systematic structure was found in the high-resolution (Dl/l59.7
deviation in one of the corrections. 31026 ) double crystal measurements of Onoue and Suzuki
@44# for the chromium K a 1 line, or the studies of Deutsch
and co-workers @77,78# for Cu.
The resolution achievable in the single-crystal instrument
used here is limited by the narrow but finite width of the
spectrometer window. Using the values D u s /D u d from
Table I we roughly estimate that fine structure of width
smaller than ~0.05–0.12!3FWHM of the K a 1 line, i.e., sub-
structures with a FWHM smaller than 0.12–0.25 eV ~de-
pending on the different resolution settings for each mea-
surement!, cannot be resolved with our setup.
The accuracy of the present data is comparable with the
results from two- and three-crystal spectrometers. The
FWHM of the Cr K a 1 peak, which was measured with the
lowest-resolution setting of all of our measurements ~see
Table I! is 1.88 eV. It is found to be in excellent agreement
FIG. 5. The present, and previous, values of the K b 1,3/K a 1,2 with the 1.87-eV value of the high-resolution study of Onoue
intensity ratio. Note the almost constant value across the range, and and Suzuki @44#, measured using a double-crystal spectrom-
the strong underestimation by the theoretical values of Manson eter with an asymmetric setup, and the 1.87 eV obtained in
@68#, and measurements of Kücükönder @70#. the triple crystal spectrometer measurements of Shah and
56 K a 1,2 AND K b 1,3 X-RAY EMISSION LINES OF THE . . . 4567

Das Gupta @73#. This agreement clearly demonstrates the taking into account the significant Coster-Kronig cross sec-
good resolution and negligibly small instrumental broaden- tion. The curve of the K a widths vs Z shows a characteristic
ing in our measurements. In spite of the good resolution no peak near the iron, while that of the K b widths increases
fine structure of the type reported in the studies mentioned monotonically with Z. We also presented values for the
above was detected in any of the spectral lines. While fea- K b 1,3/K a 1,2 and K a 2 /K a 1 intensity ratios. Both are some-
tures of widths below our resolution, 0.1–0.25 eV, could what higher than previous measurements though still within
hypothetically remain undetected, the existence of such fea- the combined experimental error of the best previous sets.
tures is highly unlikely considering the lifetime widths of the Finally, an empirical fit by a minimal set of Lorentzians is
3d transition metals’ emission lines. presented for each spectrum, for use in accurate measure-
ments requiring a correction for nonmonochromaticity, like
IV. CONCLUSIONS structural refinements based on powder diffraction patterns.
The high-quality set of spectra obtained in this study pro-
We presented here results of a high-resolution and high- vides a broad and accurate data base for testing theoretical
accuracy determination of the K a and K b emission line calculations and predictions concerning the atomic structure
shapes of the 3d transition metals chromium, manganese, and excitation and deexcitation dynamics in the 3d transition
iron, cobalt, nickel, and copper. The measurements were elements. In particular, it is hoped that a physically moti-
done using the same optimized single-crystal spectrometer, vated study, now in progress, employing ab initio Dirac-
where special attention was given to minimizing the instru- Fock calculations and the present spectra will allow a defi-
mental broadening effects. The lattice parameter of the per- nite resolution of the long standing question of the origin of
fect silicon crystal employed was determined on an absolute the asymmetry of the x-ray emission lines of the 3d transi-
scale using combined x-ray and optical interferometry to a tion elements.
sub-ppm accuracy.
The most important results of this study are the determi-
nation of the absolute peak wavelengths of the lines on a ACKNOWLEDGMENTS
metric scale to an accuracy of Dl/l;1026 and the precise
measurement of the emission line shape. This is imperative Support by the Israel Science Foundation, Jerusalem, and
for the elucidation of the physical processes underlying the the Bundesministerium für Forschung und Technologie un-
highly asymmetric x-ray lineshapes of the 3d transition met- der Contract No. 05 5SJAAI7 is gratefully acknowledged.
als. The linewidths and indices of asymmetry, determined We thank P. Becker ~Physikalisch-Technische Bundesanstalt
from the measured spectra, compare well with the best pre- Braunschweig, Germany! for providing the calibrated silicon
vious measurements, and also with theoretical calculations crystal used in the absolute wavelength measurements.

@1# M. Druyvesteyn, Z. Phys. 43, 707 ~1928!. tone, and G. Zossi, Phys. Rev. Lett. 72, 3133 ~1994!.
@2# S. Doniach and M. Sunjic, J. Phys. C 3, 285 ~1970!. @13# D. Windisch and P. Becker, Phys. Status Solidi A 118, 379
@3# K. Tsutsumi and H. Nakamori, in X-Ray Spectra and Elec- ~1990!.
tronic Structure of Matter, edited by A. Faessler and G. Wiech @14# B. Henke, E. Gullikson, and J. Davies, At. Data Nucl. Data
~Fotodruck Frank OHG, München, 1973!. Tables 54, 181 ~1993!.
@4# J. Finster, G. Leonhardt, and A. Meisel, J. Phys. ~Paris!, Col- @15# W. Bond, Acta Crystallogr. 13, 814 ~1960!.
loq. 32, C-4 ~1971!. @16# J. Härtwig, S. Grosswig, P. Becker, and D. Windisch, Phys.
@5# M. Deutsch, G. Hölzer, J. Härtwig J. Wolf, M. Fritsch, and E. Status Solidi A 125, 79 ~1991!.
Förster, Phys. Rev. A 51, 283 ~1995!. @17# J. Wolf, Ph.D. thesis, Friedrich-Schiller-Universität Jena, 1994
@6# M. Deutsch, O. Gang, G. Hölzer, J. Härtwig, J. Wolf, M. ~unpublished!.
Fritsch, and E. Förster, Phys. Rev. A 52, 3661 ~1995!. @18# E. Cohen and B. Taylor, Rev. Mod. Phys. 59, 1121 ~1987!.
@7# M. Fritsch, C. C. Kao, O. Gang, and M. Deutsch ~unpub- @19# J. Härtwig and S. Grosswig, Phys. Status Solidi A 115, 369
lished!. ~1989!.
@8# J. Härtwig, G. Hölzer, J. Wolf, and E. Förster, J. Appl. Crys- @20# B. Agarwal, X-Ray Spectroscopy ~Springer-Verlag, Berlin,
tallogr. 26, 539 ~1993!. 1979!.
@9# H. Berger, X-Ray Spectrom. 15, 241 ~1986!. @21# J. Thomsen, J. Phys. B 16, 1171 ~1983!.
@10# J. Härtwig, G. Hölzer, E. Förster, K. Goetz, K. Wokulska, and @22# W. Press, B. Flannery, S. Teukolsky, and W. Vetterling, Nu-
J. Wolf, Phys. Status Solidi A 143, 23 ~1994!. merical Recipes in PASCAL ~Cambridge University Press,
@11# S. Grosswig, J. Härtwig, K.-H. Jäckel, R. Kittner, and W. Cambridge, 1989!.
Melle, Sci. Instrum. 1, 29 ~1986!. @23# International Tables for X-Ray Crystallography, edited by J.A.
@12# ~a! P. Becker, K. Dorenwendt, G. Ebeling, R. Laufer, W. Lu- Ibers and W.C. Hamilton ~Kynoch Press, Birmingham, 1974!,
cas, R. Probst, H.-J. Rademacher, G. Reim, P. Seyfried, and H. Vol. IV. Note that the values in Vol. III as well as those in the
Siegert, Phys. Rev. Lett. 46, 1540 ~1981!; ~b! K. Nakayama new International Tables of Crystallography, edited by A.J.C.
and H. Fujimoto, IEEE Trans. Instrum. Meas. 46, 580 ~1997!; Wilson ~Kluwer, Dordrecht, 1992!, Vol. C, are taken from a
~c! G. Basile, A. Bergamin, G. Cavagnero, G. Mana, E. Vit- compilation of Y. Cauchois and H. Hulubei, Tables de Con-
4568 HÖLZER, FRITSCH, DEUTSCH, HÄRTWIG, AND FÖRSTER 56

stantes et Donees Numeriques, I. Longueurs d’Ondes des @53# R. Deslattes, R. LaVilla, P. Cowan, and A. Henins, Phys. Rev.
Emissions X et des Discontinuites d’Absorption X ~Hermann, A, 27, 923 ~1983!.
Paris, 1947!, which differ from the values employed here. @54# J. Williams, Phys. Rev. 44, 146 ~1933!.
@24# R. D. Deslattes, Nucl. Instrum. Methods Phys. Res. B 31, 51 @55# J. Scofield, Phys. Rev. A 9, 1041 ~1974!.
~1988!. @56# J. H. McGrary, L. V. Singman, L. H. Ziegler, L. D. Looney, C.
@25# J. Bearden, Rev. Mod. Phys. 39, 78 ~1967!. M. Edmonds, and C. E. Harris, Phys. Rev. A 4, 1745 ~1971!.
@26# R. Deslattes and A. Hennins, Phys. Rev. Lett. 31, 435 ~1973!. @57# S. I. Salem and R. J. Wimmer, Phys. Rev. A 2, 1121 ~1970!.
@27# R. Deslattes and E. Kessler Jr., IEEE Trans Instrum. Meas. 40, @58# S. I. Salem, T. Falconer, and R. Winchell, Phys. Rev. A 6,
92 ~1991!. 2147 ~1972!.
@28# V. Pessa, E. Suoninen, and T. Valkonen, Phys. Fenn. 8, 71 @59# Z. G. Pinsker, Dynamical Scattering of X-Rays in Crystals
~Springer, Berlin, 1978!.
~1973!.
@60# K. Heinrich, C. Fiori, and R. Myklebust, J. Appl. Phys. 50,
@29# L. Parratt, Phys. Rev. 50, 1 ~1936!.
5589 ~1979!.
@30# G. Brogren, Ark. Fys. 23, 219 ~1962!.
@61# B. Dhal and H. Padhi, Phys. Rev. A 50, 1096 ~1994!.
@31# M. Blochin and I. Nikiforov, Bull. Acad. Sci. USSR 28, 780
@62# J. Campbell, A. Perujo, W. Teesdale, and B. Millman, Phys.
~1964!. Rev. A 33, 2410 ~1986!.
@32# P. Lee and S. Salem, Phys. Rev. A 10, 2027 ~1974!. @63# G. Hölzer, computer codes DIXI and FITSPK ~unpublished!.
@33# A. Meisel and W. Nefedov, Z. Chem. 1, 337 ~1961!. @64# D. Sewell, G. Love, and V. Scott, J. Phys. D 18, 1233 ~1985!.
@34# H. Sorum, J. Phys. F 17, 417 ~1987!. @65# D. Berényi, G. Hock, S. Ricz, B. Schlenk, and A. Valek, J.
@35# I. Edamoto, Rep. Res. Inst. Tohoku Univ. A 2, 561 ~1950!. Phys. B 11, 709 ~1978!.
@36# U. Lehnert, K. Merla, and G. Zschornack, Nucl. Instrum. @66# A. Perujo, J. Maxwell, W. Teesdale, and J. Campbell, J. Phys.
Methods Phys. Res. B 89, 238 ~1994! B 20, 4973 ~1987!.
@37# J. Bearden and C. Shaw, Phys. Rev. 48, 18 ~1935!. @67# G. Paic and V. Pecar, Phys. Rev. A 14, 2190 ~1976!.
@38# J. Matthew, J. Nuttall, and T. Gallon, J. Phys. C 9, 883 ~1976!. @68# S. Manson and D. Kennedy, At. Data Nucl. Data Tables 14,
@39# J. Fuggle and S. Alvarado, Phys. Rev. A 22, 1615 ~1980!. 111 ~1974!.
@40# K. Tsutsumi and H. Nakamori, J. Phys. Soc. Jpn. 25, 1418 @69# K. Jankowski and M. Polasik, J. Phys. B 22, 2369 ~1989!.
~1968!. @70# A. Kücükönder, Y. Sahin, E. Büyükkasaps, and A. Kopya, J.
@41# H. Sorum, O. Weng, and J. Bremer, Phys. Status Solidi B 109, Phys. B 26, 101 ~1993!.
335 ~1982!. @71# H. Schnopper and K. Kalata, Appl. Phys. Lett. 15, 134 ~1969!.
@42# H. Sorum, Phys. Status Solidi B 113, 197 ~1982!. @72# M. Shah and K. Das Gupta, Phys. Lett. 29A, 570 ~1969!.
@43# H. Sorum and J. Bremer, J. Phys. F 12, 2721 ~1982!. @73# M. Shah and K. Das Gupta, J. Phys. Soc. Jpn. 37, 1069 ~1974!.
@44# K. Onoue and T. Suzuki, Jpn. J. Appl. Phys. 17, 439 ~1978!. @74# J. Priest, J. Appl. Phys. 42, 4750 ~1971!.
@45# M. Krause and J. Oliver, J. Phys. Chem. Ref. Data 8, 329 @75# W. Sauder, J. Huddle, J. Wilson, and R. LaVilla, Phys. Lett.
~1979!. 63A, 313 ~1977!.
@46# W. Nefedov, Bull. Acad. Sci. USSR 28, 816 ~1964!. @76# G. Brogren, Phys. Rev. 96, 589 ~1954!.
@47# J. Fuggle, J. Electron Spectrosc. Relat. Phenom. 21, 275 @77# M. Deutsch and M. Hart, Phys. Rev. B 26, 5558 ~1982!.
~1980!. @78# N. Maskil and M. Deutsch, Phys. Rev. A 37, 2947 ~1988!.
@48# M. Hienonen, J. Leiro, and E. Suoninen, Philos. Mag. B 44, @79# A. Nigavekar and S. Bergwall, J. Phys. B 2, 507 ~1969!.
175 ~1981!. @80# P. Sarode, X-Ray Spectrom. 22, 138 ~1993!.
@49# L. Yin, I. Adler, M. Chen, and B. Crasemann, Phys. Rev. A 7, @81# C. Bhuinya and H. Padhi, J. Phys. B 25, 5283 ~1992!.
897 ~1973!. @82# V. Slivinsky and P. Ebert, Phys. Rev. A 5, 1581 ~1972!.
@50# R. Nyholm, N. Martenson, A. Lebulge, and U. Axelson, J. @83# N. Rao, S. Reddy, G. Satyanarayana, and D. Shastry, Physica
Phys. F 11, 1727 ~1981!. C 138, 215 ~1986!.
@51# D. M. Pease, Phys. Rev. B 44, 6708 ~1991!. @84# L. Coelho, M. Gaspar, and J. Eichler, Phys. Rev. A 40, 4093
@52# P. Chevallier, M. Tavernier, and J. Briand, J. Phys. B 11, L171 ~1989!.
~1978!. @85# E.R. Cohen, B.N. Taylor, CODATA Bull. No. 63 ~1986!.