An official website of the United States government
Here's how you know
Official websites use .gov
A
.gov website belongs to an official
government organization in the United States.
Secure .gov websites use HTTPS
A lock (
) or https:// means you've safely
connected to the .gov website. Share sensitive
information only on official, secure websites.
As a library, NLM provides access to scientific literature. Inclusion in an NLM database does not imply endorsement of, or agreement with,
the contents by NLM or the National Institutes of Health.
Learn more:
PMC Disclaimer
|
PMC Copyright Notice
A computation is made of energy levels, wave functions and transition probabilities of the Fe3+ ion in Al2O3. The crystal field parameters used were those determined by Symmons and Bogle at 4 K. The magnetic field direction is described by the angles θ and φ indicating the directions with respect to and around the c axis of the crystal. The values of θ go from 0 to π/2 with π/12 intervals, the angles φ are 0 and 2π/3 corresponding to the two nonequivalent sites of the crystal. The transition probabilities are given for dipole radiation in three polarization directions and for ultrasonic work the six components of the quadrupole transitions were computed.
Keywords: Energy levels of Fe+ + + in Al2O3, iron doped Al2O3, magneto elastic tensor, paramagnetic resonance, quadrupole transitions, spin hamiltonian, transition probabilities, ultrasonic (paramagnetic) resonance, ultrasonic transition probabilities, wave functions of Fe+ + +
1. Introduction
As a result of recent work in ultrasonic paramagnetic resonance [1],1 it became desirable to obtain a complete set of transition probabilities for quadrupole radiation, besides the known values for the dipolar transitions. We have programmed the calculations of these and related quantities in order to select a priori angles and field strengths that would enable us to observe specific components of the transition probabilities. There are some uncertainties with regard to the selection rules for ultrasonic absorption [2–4]. The first question is whether dipolar, quadrupolar, or a combination of these determines the line strength of the absorption. The next question is whether components of other multipoles are playing a role in the process.
In the available literature [5–8] the wave functions are never quoted. Consequently, it was necessary to recalculate these quantities in order to evaluate the transition probabilities. A new program was constructed for that purpose; the details are described in appendix I.
The physical properties and the spin hamiltonian are described in Sections 2 and 3.
The angle θ is the angle between the magnetic field and the c axis of the lattice which is a three-fold axis oriented along a body diagonal of the crystal field. There are two nonequivalent sites in the lattice. The angle φ depends on the ion in the unit cell. The crystal field is rotated over π/3 if one goes from one ion site to the other. Hence the spectrum is actually a superposition of two spectra. The differences are in general small, but noticeable.
The values for θ in which the calculation was performed are 0, π/12 … π/2 rad. for φ = 0 and π/3. The field was varied from 0 to 0.5 Teslas (5000 Gauss) in steps of 0.025 T, but not all values are incorporated in the tables. At low field values and or near the crossing points, 10−3 T steps were used.
The crossing or non-crossing of the levels constitutes a certain practical difficulty in the labelling of the wave functions. We decided simply to list the levels in increasing order and label the wave functions accordingly.
2. Properties of the Crystal and Discussion of Previous Work
The free ion Fe3+ has the configuration 3d5. The 3d electron shell is half filled and the resultant orbital momentum L is equal to zero. The singlet ground state 6S has a six-fold spin degeneracy. The first four excited states of the free ion are 4G, 4P, 4D, and 4F [9]. The optical spectrum has been studied by Moorjani and McAvoy [10].
The Al2O3 lattice, as described by Geschwind and Remeika [11] is shown in figure 1. It exhibits a three-fold symmetry around an axis perpendicular to the plane of the O2−. The lattice presents the properties of the 3 m point group (D3d in the Schoenflies notation) but each site has only a 3 (C3) point of symmetry.
Portion of the Al2O3 lattice, the two inequivalent sites are located in two adjacent planes for instance a and b (ref. [10]).
The specific directions for the propagation of sound waves have been studied by Borgnis [12]. The c axis and also the x axis (two-fold rotational symmetry often called “a” axis) are directions for purely compressional waves. There are two nonequivalent sites, which are identical except for a rotation of π/6 around the c axis. Consequently, the spectrum consists of the superposition of two slightly different spectra. The differences are large enough to be seen, i.e., the lines are clearly separated in certain directions and for certain fields. Ultrasonic propagation and absorption in Al2O3 were studied by Brian and Meister [13]. The velocity of sound for longitudinal waves is 10.9 × 105 cm/s, for shear waves, 6.41 × 105 cm/s. The Fe3+ ion, when introduced in the Al2O3 lattice occupies the sites of an Al ion. It is surrounded by 6 ions that are held responsible for the electric field: the O2− ions. The positions of these ions determine the symmetry of the crystal field. Although deviations are known, the site symmetry can be considered bo be C3 as mentioned above.
The crystal field has a main component of trigonal symmetry and a smaller component of cubic symmetry. The total angular momemtum J = S = 21/2 is split into three Kramers doublets whose degeneracy can be removed only by a magnetic field. For fields large compared to the zero field splitting the ion will be virtually free. Although the zero field splitting is small for S-state ions, most of the ultrasonic work was done in the low field region; hence we are mainly interested in the numerical values of the transition probabilities when magnetic field and crystal field are comparable.
This level structure is confirmed by many paramagnetic experiments [14–16]. The energy levels are most conveniently described by a spin hamiltonian with an effective spin S = 5/2. Abraham and Pryce proposed [7] the following expression:
The D term corresponds to the trigonal or tetragonal crystalline field components and incorporates a part of the spin-spin interactions. A slight distortion from cylindrical symmetry around the z axis is described by the E term. The cubic field is described on a set of orthogonal axes such that the z axis (trigonal axis of the crystal) lies in the (111) direction. They considered the E term of no importance. Bleaney and Trenam [18] introduced another term for the fourth degree part of the trigonal field and wrote the spin hamiltonian
which became the generally accepted form. Basically, this is the most general form if one imposes the following conditions:
The triangular rule requires that the main quantum number of the spin hamiltonian for d-electrons should be 4 or less.
Inversion symmetry rules out even values for the main quantum number.
The C3 symmetry implies that the azimuthal quantum number can only take the values 0, ±3.
For sufficiently high fields (gβB ≫ a) the energy levels have been calculated. Bleaney and Ingram [19] obtained by perturbation calculation the matrix elements of the allowed transition in the absence of the fourth order terms. This calculation was completed by Bleaney and Trenam [18]. The results are:
In these expressions B is the magnetic field corresponding to a transition at the frequency v and B0 = hv/gβ.
where (l, m, n) are the directions cosines of H referred to the axes of the cubic crystalline field. θ is the angle between H and the trigonal axis and
3. Calculation of Eigenvalues and Eigenfunctions of the Spin Hamiltonian
The general expression for spin hamiltonians, using the conventions introduced by Stevens [21] as reviewed by Hutchings [22], is given by
where n is the “length” of the vector, m the azimuthal quantum number of its projection. The three restrictions mentioned in section 2 lead to a spin hamiltonian with three undetermined parameters
(1)
On the basis of the description of the crystal some coefficients are dominant, in particular, the term representing the trigonal field called D = 3 Next in importance is probably the field of cubic symmetry, oriented in such a way that a body diagonal of the cube coincides with the three-fold axis of the trigonal field.
This field can be described in a coordinate system with the axes along the four-fold axes of the cube or in a system with the axes along the three-fold axis (plus two orthogonal axes, to be defined) of the crystal.
On the first set of axes we have in operator polynomials
This expression can be transformed into the second set of axes [22]. In figure 2 the z axis is in the (111) direction. The x axis is in the zoζ plane (fig. 3) and y is in the ηξ plane:
Finally it is assumed that the cube is distorted (elongated or compressed) along the z axis. This is accomplished by adding a term F/180 = Adding all contributions we have for the coefficients in eq (1)
To insert the magnetic field in this matrix we use polar coordinates:
where φ is the angle between the field and the xz plane as defined in figures 2 and 3. In the case of θ = 0 the eigenvalues and eigenfunctions can be calculated directly:
where
Note that for θ = 0 (fig. 4) the levels | −5/2> and | 1/2> do not cross, contrary to figures published elsewhere [5]. For θ different from zero the diagonalization cannot be accomplished by simple algebraic operations. The computation of these orientations is conveniently performed by computer. Moreover, each calculation has to be repeated for at least two different values of φ, differing by 60°, in order to obtain the spectra of both sites.
Figure 2. Orientation of the x y z reference axes with respect to the ξ, η, ζ axes.
The solid line corresponds to φ = 0, the dotted line to φ = 60°. The Magnetic Field is in 10−4 · Tesla.
These coefficients can be easily translated in experimentally observable quantities with the exception of the sign of a.
The matrix describing the crystalline field expressed on the basis of the free spin eigenfunctions is:
4. Transition Probabilities
If the spin system previously described is placed in an rf magnetic field, Brf, the time dependent interaction hamiltonian
will induce transitions between the energy levels.
The probability of a transition between the level Ei characterized by a wave function ψi and the level Ej characterized by ψj is proportional to the square of the matrix element of Hint between ψi and ψj:
Since the ψi in most cases are not pure spin states, we have
and
(2)
This expression was calculated for three polarizations of the magnetic field, along x, y, and z.
If an ultrasonic wave is applied to the lattice the periodic displacement of the ions will induce transitions between the energy levels, through the modulation of the crystalline electric field. This can be described assuming an interaction hamiltonian including two terms, one describing the modulation of the g-factor, the second one the modulation of the crystalline field parameters.
In the case of the Fe3+, since the static spin hamiltonian includes, besides two other terms, and one would expect them to be also modulated by the ultrasonic wave. We have, however, neglected these two terms on the basis that the ratios and are very small (of the order of 2 and 5%) provided that considerations about the static behavior can be carried over to the dynamic one. With these hypotheses the interaction hamiltonian can be written:
The first term is referred in the literature as the dipolar term, the second as the quadrupolar one. The relative importance of these two terms was discussed by Mattuck and Strandberg [3] who showed that in most cases the second one is dominant and experimental data seemed to confirm that fact [23–27] in the case of non-S state ions. In this expression δg and d| are two second rank tensors related to the lattice strain ϵ as
where Fijkl and Gijkl are two fourth rank tensors (called the magnetoelastic tensors). Using the Voigt notation [28], they can be written as 6 by 6 matrices. Considering the fact that the point group symmetry at each Fe3+ site is C3 the number of independent components of F or G is only 10. For the first site the G tensor can be written:
The tensor for the second site is similar except for the sign of G52, G25, G45, and G16. The interaction hamiltonian describing the “quadrupolar” term can now be written:
Since the d’s (or the G’s) are not known for Fe3+ in Al2O3 and since we want to make it possible through a comparison with experimental data to evaluate the G’s, we have calculated the matrix elements between the different wave functions of each term:
(3)
the six (symmetrized) quadrupole moments.
5. Presentation of the Data
The energy level diagrams (figs. 4–10) were obtained by plotting the results of the calculation using steps of 0.02ST (250G) except near crossover or noncrossover points, where the steps were decreased. The dotted lines represent the data for ϕ = 60°.
Figure 10. Plot of the energy levels of Fe3+ in Al2O3.
The solid line corresponds to φ = 0, the dotted line to φ = 60°. The Magnetic Field is in 10−4 · Tesla.
The tables of numerical results use intervals for: θ, φ, and B (in 10−4 Tesla) larger than the ones computed in order to save space, but still maintain a representative sample. Each table consists of two parts:
The normalized wave functions, where every other column gives the real part, alternated by the imaginary part. The first column is the corresponding energy in cm−1. Note that the levels are arranged in ascending order. The lowest value is then called number one, the next lowest number two, etc…. This is to establish the correspondence with the next part of each table.
Tables that give the probabilities, i.e., absolute square of the matrix elements for dipole and quadrupole transitions between the energy level differences indicated. The column on the left gives the initial value i and the final value j which are of course interchangeable. The next column gives the positive energy difference in Gigahertz. The multiplication factor is 28.05 GHz/Tesla. The next three columns indicate the probabilities for a dipole transition in the x direction, the y direction, and the z direction. For θ = 0, the first two are equal. The last six columns are the quadrupolar transition probabilities. The symbol PSXSX stands for the probability corresponding to an SXSX transition. Similarly, PSXSY corresponds to 1/2 (Sx Sy + Sy Sx). The tables are for φ = 0, except for θ = 45°, where φ = 0, 30, 60, 90 is used to indicate the variation with φ.
Figure 5. Plot of the energy levelsof Fe3+ in Al2O3.
The matrix to be diagonalized is a 6 by 6 hermitian matrix. This matrix is equivalent to a 12 by 12 real symmetric matrix replacing each element:
The wave functions are treated similarly. The functions in the 12 column vector represent the real part of the wave functions at the odd positions, the imaginary part at the even positions. The diagonalization is performed using the Jacobi’s method [29]. This method consists of the following steps:
The first subroutine will select the largest off-diagonal element aij; it will diagonalize the 2 by 2 matrix consisting of aii, ajj, and aji. A second subroutine will rotate the wave functions |i> and |j> to accomplish this diagonalization. As a result of this transformation all elements will be modified. The procedure is then repeated by looking again for the largest matrix element (which may be larger than the first matrix element). Beyond a certain iteration, the largest element will decrease.
The convergence of the method is based on the relation:
where λi are the eigenvalues of the matrix. One can estimate the number of steps necessary to obtain elements of a certain magnitude; it is proportional to (n2 − n) where n is the dimensionality of the matrix [29]. The matrix is considered diagonalized if the off diagonal elements are less than a predetermined value. Actually, a weighted criterion was used based on the second order correction of the perturbation theory:
and ϵ was chosen to be 10−5. Smaller values led to “accidental” divergences. The machine was prevented from using this criterion if aii was equal to ajj. The number of iterations range from 4 (the minimum) to about 150.
Finally two more subroutines were used, one to order the levels, and one to calculate the dipolar and quadrupolar transition probabilities as indicated in eqs (2) and (3).
Footnotes
1
Figures in brackets indicate the literature references at the end of this paper.
Contributor Information
Jacques Lewiner, Physics Department, Catholic University of America, Washington, D.C. 20017.
Paul H. E. Meijer, Institute for Basic Standards, National Bureau of Standards, Washington, D.C. 20234 Physics Department, Catholic University of America, Washington, D.C. 20017.
7. References
[1].Kagan M. R., Lally J. F., Meister R., and Meijer P. H. E., Phys. Rev.
166, 234 (1968). [Google Scholar]
[2].Mattuck R. D., and Strandberg M. W. P., Phys. Rev. Letters
3, 369, (1959). [Google Scholar]
[3].Matturk R. D., and Strandberg M. W. P., Phys. Rev.
119, 1204 (1960). [Google Scholar]
[4].Dobrov W. I., Phys. Rev.
134, 734 (1964). [Google Scholar]
[5].Howarth D. J., Royal Radar Establishment (1960). [Google Scholar]
[6].Cross C. S., and Kikuchi R., Technical Report, University of Michigan Z;
1238 (24
June
1959). [Google Scholar]
[7].Unpublished figures obtained from the Bell Tel. Lab.
[8].Lessin I., U.S.A.F. Cambridge Report AFCRL
66–692 (1966). [Google Scholar]
[9].Moore C. E., NBS (US) Circ. No. 467. [Google Scholar]
[10].Moorjani K., and Mc Avoy N., Phys. Rev.
132, 504 (1963). [Google Scholar]
[11].Geschwind S., and Remeika J. P., Phys. Rev.
122, 751 (1961). [Google Scholar]
[12].Borgnis F. E., Phys. Rev.
98, 1000 (1955). [Google Scholar]
[13].Desilets E. H., and Meister R., J. Appl. Phys.
36, 261 (1965). [Google Scholar]
[14].Kornienko L. S., and Prokhorov A. M., JETP
33, 805, 1955. (Trans. JETP 6, 620, 1958). [Google Scholar]
[15].Bogle G. S., and Symmons J. F., Proc. Phys. Soc. (London)
73, 531, (1959). [Google Scholar]
[16].Symmons H. F., and Bogle G. S., Proc. Phys. Soc. (London)
79, 468, (1962). [Google Scholar]
[17].Abragam A., and Pryce M. H. L., Proc. Roy, Soc.
A205, 135, (1951). [Google Scholar]
[18].Bleaney B., and Trenam R. S., Proc. Roy. Soc. (London)
223, 1 (1954). [Google Scholar]
[19].Bleaney B., and Ingram D. J. E., Proc. Roy. Soc.
A205, 336, (1951). [Google Scholar]
[20].de L.
Kronig R., and Bouwkamp C. J., Physica
6, 290, (1939). [Google Scholar]
[21].Stevens K. W. H., Proc. Phys. Soc. (London)
A65, 209 (1952). [Google Scholar]
[22].Hutchings M. T., Solid State Physics
16, 227 (1964). [Google Scholar]
[23].Shiren N. S., and Tucker E. B., Phys. Rev. Letters
6, 105, (1961). [Google Scholar]
[26].Dobrov W. I., Phys. Rev.
134, 734 (1964). [Google Scholar]
[27].Guermeur R., Joffrin J., Levelut A., and Penne J., Compt. Rend. Acad. Sc. (Paris)
260, 108 (1965). [Google Scholar]
[28].Nye J. F., Physical Properties of Crystals (Oxford University Press, 1957) and [Google Scholar]; Birss R. R., Symmetry and Magnetism (North Holland Publishing Company, Amsterdam, 1964). [Google Scholar]
[29].Wilkinson J. H., Algebraic Eigenvalue Problems (Oxford, Oxford University Press;
1965). [Google Scholar]