Surface Science 499 (2002) L141–L147 www.elsevier.com/locate/susc
Surface Science Letters
5f Resonant photoemission from plutonium J. Terry a,1, R.K. Schulze a, J.D. Farr a, T. Zocco a, K. Heinzelman b, E. Rotenberg b, D.K. Shuh b, G. Van der Laan c, D.A. Arena d, J.G. Tobin
d,*
a
Los Alamos National Laboratory, Los Alamos NM 87545, USA Lawrence Berkeley National Laboratory, Berkeley CA 94720, USA c Daresbury Laboratory, Warrington WA4 4AD, UK Lawrence Livermore National Laboratory, Livermore CA 94550, USA
b
d
Received 19 September 2001; accepted for publication 29 October 2001
Abstract Experimental resonant photoemission (ResPes) results for a-Pu and d-Pu bulk samples are presented and compared to the results of an atomic model calculation. Both Pu samples exhibit limited agreement with the atomic model calculations. As expected, a-Pu appears to have more 5f valence band delocalization than d-Pu. Evidence of an enhanced sensitivity to surface corruption, by using synchrotron radiation as the excitation, is presented. Ó 2001 Elsevier Science B.V. All rights reserved. Keywords: Polycrystalline surfaces; Surface structure, morphology, roughness, and topography; Synchrotron radiation photoelectron spectroscopy
The valence electronic structures of the actinide metals and alloys in general and plutonium (Pu) in particular remain mired in controversy [1–5]. Interestingly, the various phases of Pu metal provide a microcosm of the metallic actinides as a whole [6–9]. Thus, unravelling the nuances of the interplay of electronic and geometric structures in Pu will illuminate the properties of all transuranic metals. In a sense, the behavior of the Pu 5f elec-
* Corresponding author. Tel.: +1-925-422-7247; fax: +1-925423-7040. E-mail address:
[email protected] (J.G. Tobin). 1 Present Address: Ill. Inst. of Tech. and the Adv. Ph. Source, ANL, Chicago, IL, USA.
trons is completely counter-intuitive. The dense phase, a, has some semblance of delocalization in the 5f valence bands and can be treated theoretically within single electron models such as the local density approximation (LDA). The a phase is monoclinic, which is a low symmetry ordering [10]. The less dense d-phase is fcc and exhibits evidence of localized and/or correlated electronic behavior. The fcc is a high symmetry phase which is normally associated with superior wave function overlap in d state metals. But herein is the key: the linear combinations of the 5f’s do not produce the nicely lobed wave functions with symmetry about the x, y, z and diagonal axes, as occurs for d states. Instead, the linear combinations of f-states have oddly lobed and badly directed wave functions
0039-6028/01/$ - see front matter Ó 2001 Elsevier Science B.V. All rights reserved. PII: S 0 0 3 9 - 6 0 2 8 ( 0 1 ) 0 1 7 5 7 - 5
CE RF A LE C E TT S E R C IE S N
SU
L142
J. Terry et al. / Surface Science 499 (2002) L141–L147
[11–13] which match very poorly with the high symmetry of the fcc structure. In the rare earths and lanthanides, this odd lobing is of no consequence: generally, the 4f valence states are inside the outer 5d, 6p and 6s electrons and can be treated as isolated, atomic-like orbitals that do not participate in chemical bonding. In the actinides, the 5f’s are less well shielded. Shielding is undoubtedly a key issue and relates to the volume changes. In fact, the 20% volume increase of Pu between the a and d phases is a reflection of the discontinuous 30% volume jump between Pu and Am, as one moves along the row of actinide elements [14–16]. Furthermore, the fcc d-Pu cannot be treated theoretically without ad hoc corrections, e.g., the addition of Coulombic repulsion terms to LDA formalisms [4,5]. Finally, we have not even begun to deal with the issue of dilute alloy formation and its impact upon the electronic structure of metallic actinides. The upshot of this is that the valence electronic structure of Pu in particular and the actinides in general is only poorly understood. To attempt to correct this situation, we have begun a series of synchrotron-radiationbased, photoelectron spectroscopy investigations at the Advanced Light Source in Berkeley, CA. The Pu 5f valence band measurements were performed using resonant photoemission. In a ResPes experiment, the valence band spectrum is collected as the photon energy is ramped through a core level absorption edge. In the case of Pu, the photon energy was scanned through the O4;5 (5d) absorption edge. Thus, the result is the 5d5f5f resonant photoemission decay from Pu metal. The final state can be reached by an X-ray absorption-Coulomb decay process, 5d10 5f n þ hm ! 5d9 5f nþ1 $ 5d10 5f n1 þ e, and by direct photoemission, 5d10 5f n þ hm ! 5d10 5f n1 þ e, where e is a photoelectron in a continuum state which has no interaction with the ionic state left behind. Interference between the different pathways results in asymmetric absorption features in the measured spectra. The experiments were performed at the spectromicroscopy facility (Beamline 7.0) at the Advanced Light Source in Berkeley, CA [17]. Based upon standard mean free path arguments, the sampling depth should be on the order of 20 to . The Pu samples were taken from a specially 50 A
purified batch of Pu metal. The Pu was zone refined and vacuum distilled while magnetically levitated [18]. The product of the purification process was a-Pu containing a total of 170 ppm impurities. A portion of the refined metal was alloyed with gallium to form the d-phase (fcc symmetry). The sample surfaces were prepared by repeated room-temperature, sputter-annealing cycles to minimize the amount of oxygen and other impurities dissolved in the sample or at grain boundaries, in a specially designed chamber attached to the sample introduction and analysis systems on Beamline 7.0. The transfer, preparation, and analysis chambers ensured that the Pu metal samples did not experience pressures greater than 108 Torr. This minimized any surface contaminants that could adversely effect the soft X-ray measurements. Photoelectron spectroscopy is a ‘‘photon in, electron out’’ process. Often, it can be simplified down to a single electron phenomenon, where the energy of the photon is absorbed and transferred over entirely to a single electron, while all other ‘‘spectator’’ electrons essentially remain frozen. An advantage of this is its simplicity of interpretation. But in many systems, it is possible to induce a process with heightened sensitivity and significantly increased cross sections, resonant photoemission (ResPes) [17,19–22]. Here, a second set of indirect channels open up, which contribute in concert with the original or direct channel of simple photoemission. Shown in Fig. 1 are ResPes results from polycrystalline a-Pu and single crystallite (large grain polycrystalline) d-Pu. The plots show intensity variations versus binding energy and photon (excitation) energy. Along the photon energy axis (90–160 eV), the classical signature of ResPes can be seen, an anti-resonance or intensity minimum (near 100 eV) followed by a maximum (near 125 eV). These energies correlate with the Pu 5d core level threshold at 102 eV. ResPes is a multielectron process which in this case involves 5d as well as 5f electrons. Along the binding energy axis, the ‘‘wall’’ is the threshold associated with the Fermi energy: no emission can occur above the Fermi energy. The overall spectral envelopes are approximately the same, as expected for two samples which are primarily Pu. Subtle yet signif-
SU N IE SC RS CE E A TT R F LE CE
J. Terry et al. / Surface Science 499 (2002) L141–L147
L143
Fig. 1. Two ResPes data sets are shown here, for a polycrystalline a (top) and single crystallite d (bottom). The pass energy of the analyzer was 1 eV. The plots show the intensity variations (z-axis) versus the binding energy of the states (the negative numbers in eV; zero is the Fermi energy) and photon energy (between 90 and 160 eV).
icant differences can be observed between the a and single crystallite d samples, particularly at the Fermi energy and near a photon energy of 130 eV. These results suggest that the valence electronic structure of Pu is dependent upon its phase and
chemical state. Nevertheless, overall the two sets of spectra strongly resemble each other and confirm the observation of Pu 5f ResPes. Additional measurements, such as core level spectroscopy of the Pu and tracking of oxygen and
CE RF A LE C E TT S E R C IE S N
SU
L144
J. Terry et al. / Surface Science 499 (2002) L141–L147
Fig. 2. (a) Core-level photoemission spectra from a large crystallite d-Pu sample and a polycrystalline a-Pu sample are shown. These spectra were collected with a photon energy of 850 eV and an analyzer pass energy of 6 eV. Note that two large components are visible in each spectrum, a sharp feature (S) at low binding energy and a broad feature (B) at higher binding energy. (b) The oxygen 1s and (c) carbon 1s regions. The carbon (binding energy of 280 eV) and oxygen (binding energy of 530 eV) are much more easily observed at 860 eV (––) than at 1250 eV (- - -).
carbon surface contamination, have been performed. Shown in Fig. 2 are the Pu 4f core level spectra for both samples. Here it is clear that there are obvious differences between the a and d spectra, which may be used to differentiate between the phases. It should be noted that these spectra confirm earlier work performed by Gouder et al. [23] and others [24–26] using laboratory sources. These spectra are sensitive to the bulk: it seems fairly likely that surface effects are not the dominant factor here since the kinetic energy is close to
400 eV and provides substantial bulk sensitivity. It has been argued in the past [23] that the relative sharpness of the peaks in a-Pu and the large satellite features in d-Pu are suggestive of delocalization in a-Pu and atomic or electron correlated effects in d-Pu. Interestingly, while the Pu 4f core level spectra (Fig. 2a) are consistent with previous results, the surface contamination work has lead to a surprising result, relative to previous laboratory studies (also Fig. 2b and c). Here the spectra at a photon energy of 1250 eV (i.e., approximately Mg Ka energy) suggests that the sample surface is ‘‘clean’’, but the spectrum at a photon energy of 860 eV indicates the presence of oxygen and carbon contamination. The increased sensitivity of the measurements made at photon energies available only at synchrotron radiation sources suggests that there may be more oxygen and carbon contamination than heretofore suspected, based upon measurements using laboratory sources. The point is that what was thought to be ‘‘clean’’ may in fact have had surface contamination. Based upon relative O and Pu spectral intensities for various peaks and photon energies and correcting for cross sectional effects, it is estimated that the O/ Pu concentration ratio is on the scale of 10% for the sampling depth. It is for this reason that we will treat the present ResPes results as representative of Pu in general and be fairly circumspect in terms of making phase specific conclusions. Now we will compare the experimental results to calculations based upon an atomic model. Theoretically, a-Pu should be more free-electronlike than d-Pu. Nevertheless, because of the possibility of the impact of partial oxidation, we will focus instead upon analysis within the atomic picture. At worst, one might expect that Pu ResPes from Pu-oxide would be reflective of a localized valence electronic structure. Here, we will be looking for trends instead of quantitative agreement. Finally, the theoretical understanding of ResPes has recently been considerably improved. Transition metal and lanthanide systems usually give a good agreement between experiment and theory [17,19–21]. The situation is muddled for the actinides [3,27]. In any case, it is highly useful for us to begin the analysis by comparing our Pu
SU N IE SC RS CE E A TT R F LE CE
J. Terry et al. / Surface Science 499 (2002) L141–L147
L145
experimental results to an atomic picture of 5f ResPes. Using the t-matrix approach, the transition probability for ResPes was calculated in first order in the dipole operator and infinite order in the Coulomb decay [21]. The decay half-width is then expressed by Ci ¼ pV 2
ð1Þ
where V is the super-Coster-Kronig matrix element, which is proportional to the radial integrals. The Fano asymmetry parameter (q) is given by qi ¼ ðDa Þ=ðpVDe Þ
ð2Þ
where Da ¼ h5djrj5fi is the matrix element for the 5d ! 5f absorption, and De ¼ h5fjrjei is the matrix element for 5f photoemission. If the transition probability for absorption (5d ! 5f) is much larger than for direct photoemission (5f ! e), the Fano parameter, q, is large resulting in lineshapes that are rather symmetric. As q approaches infinity, the Fano line shape goes over into a Lorentzian. It is important to note that C and q only have a simple meaning for a single line interacting with a continuum. In the presence of multiplet structure, intermediate states of the same symmetry interfere with each other. In this calculation, the interference was explicitly taken into account by diagonalization of the Hamiltonian for the connected final states, 5d9 5f nþ1 $ 5d10 5f n1 þ e. The details of the calculation are described elsewhere [28]. Constant initial state (CIS) spectra, with fixed binding energy and variable photon energy, are a measure of the absorption and resonant photoemission. (These are slices through the plots in Fig. 1, with fixed binding energies.) Both Pu experimental and calculated atomic results are shown in Fig. 3. The calculated 5d absorption spectrum (BE ¼ 12:0 eV) displays a broad structure with roughly three features. The separation into these distinct features is mainly due to the df and ff Coulomb interaction. The two features at lower energy have mainly J ¼ 7=2 and 5=2 character, while the feature at higher energy has mainly J ¼ 3=2 character. The line broadening is due to the decay where the Hartree–Fock (HF) value
Fig. 3. The CIS spectra are shown: (a) a-Pu, (b) d-Pu, and (c) atomic theory. (a) The calculated spectra are shown at the binding energies of 0.0, 0.5, 1.3, 2.1, 3.2, and 12.0 eV. The experimental spectra are shown at similar binding energies. Note that in both the calculated and measured d-Pu spectra, the intensity of the second and third peaks switch as the binding energy increased.
of the half-width C is about 3.5 eV. Other decay channels, such 5d5f6s, 5d5f6p, 5d6p6p, and
CE RF A LE C E TT S E R C IE S N
SU
L146
J. Terry et al. / Surface Science 499 (2002) L141–L147
5d6s6p, were taken into account in the calculation by including an additional Lorentzian of width C0 ¼ 1 eV. In the calculated spectra, the strong interference between the direct photoemission and the decay makes the absorption structures asymmetric with an intensity minimum around 97 eV due to the Fano anti-resonance. The experimental CIS spectra from d-Pu show a similar splitting into two peaks with a separation of about 8 eV (117 and 125 eV) and a possible shoulder at lower energy. (In a configuration interaction (CI) model, these satellites can be ascribed to configurations 5d9 5f n . With different n in X-ray absorption [but not in photoemission], the satellite intensities are usually small and a single-configuration approximation is often valid [21]. The reason for this is that after 5d ! 5f excitation, the 5d core hole is efficiently screened by the additional 5f electron.) The most successful application of the atomic model is its ability to reproduce the relative intensities of the second (theory: 110 eV; experiment: 117 eV) and third peaks (118 eV; 125 eV) in the data. The third peak is originally more intense than the second peak. As the binding energy increases (1.3 eV), the second peak becomes more intense than the third. As the binding energy is furthur increased the intensities reverses again. This suggests the presence of a remnant of atomic structure in the d-Pu. This similarity between theory and experiment was not as well observed in the CIS spectra from a-Pu, but still appears to be present. The diminishment of the effect in a-Pu is possibly due to the assumed greater delocalization of the f-electrons in a-Pu sample. High quality ResPes spectra of a-Pu and d-Pu have been collected and compared to the results of an atomic model calculation: Significant agreement has been found. Nevertheless, it is clear that an atomic model alone will not suffice to explain the observations of 5f ResPes in Pu. Supporting core level measurements have been made which confirm the a-Pu and d-Pu are different but which also suggest the need for care in dealing with surface corruption. In the future, refined sample preparation and enhanced inclusion of delocalization effects in the theory will be required for further phase specific analyses.
Acknowledgements This work was performed under the auspices of the US Department of Energy by LLNL (contract no. W-7405-Eng-48), LANL (contract no. W-7405- ENG-36) and LBNL (contract no. DEAC03-76SF00098). The Spectromicroscopy Facility (Beamline 7.0) and the Advanced Light Source were built and and are supported by the DOE Office of Basic Energy Research. The authors wish to thank Jason Lashley and Michael Blau for synthesis of the Pu samples.
References [1] [2] [3] [4] [5]
[6]
[7]
[8] [9] [10]
[11] [12] [13]
[14]
[15] [16] [17] [18] [19]
E.A. Kmetko, Cleaver-Hume Press Ltd., 1965. O. Eriksson et al., J. Alloy Compd. 287 (1999) 1. J.W. Allen, J. Electron Spectrosc. 78 (1996) 57. S.Y. Savrasov, G. Kotliar, Phys. Rev. Lett. 84 (2000) 3670. P. Soderlind, Phys. Rev. Lett., submitted for publication; P. Soderlind et al., Phys. Rev. B 55 (1997) 1997; P. Soderlind, Adv. Phys. 47 (1998) 956. F.W. Schonfeld et al., Plutonium constitutional diagrams, Progress in Nuclear Energy, Series V, Metallurgy and Fuels vol. 2 (1959) 579. R.O. Elliot et al. (Eds.), Thermal Expansion of some Delta Plutonium Solid Solution Alloys, Cleaver-Hume Press, London, 1960. F.W. Schonfeld, R.E. Tate, Los Alamos National Laboratory, Los Alamos, NM, 1996. G.T. Seaborg, W.D. Loveland, The Elements Beyond Uranium, Wiley, New York, 1990, p. 12. W. Wolfer et al., LLNL, 1999; W.H. Zachariasen, F.H. Ellinger, Acta Crystallogr. 16 (1963) 777; F.H. Ellinger, Trans. AIME 206 (1956) 1256. E.E Anderson, Modern Physics and Quantum Mechanics, W.B. Saunders Publishing, Philadelphia, PA, 1971. M. Karplus, R.N. Porter, Atoms and Molecules, The Benjamin/Cummings Publishing, Menlo Park, CA, 1970. H.G. Freidman et al., J. Chem. Educ. 41 (1964) 354; C. Becker, J. Chem. Educ. 41 (1964) 358; E.A. Ogryzlo, J. Chem. Educ. 42 (1965) 150. D.L. Olson, Electronic and Alloying Predictions of Pu Stability, Seminar at LANL, Los Alamos, NM, September 13, 1999. D.C. Wallace, Phys. Rev. B 58 (1998) 15433. S. Meot-Reymond, J.M. Fournier, J. Alloy Compd. 232 (1996) 119. S.R. Mishra et al., Phys. Rev. Lett. 81 (1998) 1306, and references therein. J.C. Lashley, J. Nucl. Mater. 274 (1999) 315. G. van der Laan et al., Phys. Rev. B 46 (1992) 9336.
SU N IE SC RS CE E A TT R F LE CE
J. Terry et al. / Surface Science 499 (2002) L141–L147 [20] K. Starke et al., Phys. Rev. B 55 (1997) 2672. [21] G. van der Laan et al., Phys. Rev. B 59 (1999) 8835; G. van der Laan et al., Phys. Rev. B 33 (1986) 4253. [22] A.J. Arko et al., J. Alloy Compd. 271–273 (1998) 826; A.J. Arko et al., J. Alloy Compd. 286 (1999) 14. [23] T. Gouder et al., J. Alloy Compd. 271–273 (1998) 841; T. Gouder et al., J. Electron Spectrosc. 101–105 (2000) 419;
[24] [25] [26] [27] [28]
L147
T. Gouder et al., Poster Presentation, VUV-12 Conference, San Francisco, CA, 1998. R. Baptist et al., J. Phys. F Met. Phys. 12 (1982) 2103. L.E. Cox, Phys. Rev. B 37 (1988) 8480. J.R. Naegele, J. Nucl. Mater. 166 (1989) 59. G. Kalkowski et al., Phys. Rev. B 35 (1987) 2667. G. van der Laan, in preparation.