Particle size determination of superparamagnetic .alpha.-iron in

Jul 9, 1985 - of applied magnetic field.1 With iron catalysts, the magnetization can conveniently be measured byMossbauer spectroscopy. However ...
0 downloads 0 Views 371KB Size
4898

J . Phys. Chem. 1985, 89, 4898-4900

Particle Size Determination of Superparamagnetic a-Fe in Carbon-Supported Catalysts by in Situ Mossbauer Spectroscopy Per Helvig Christensen, Steen Mrarup, Laboratory of Applied Physics II, Technical University of Denmark, DK- 2800 Lyngby, Denmark

and Johannes W. Niemantsverdriet* Laboratory of Inorganic Chemistry, Eindhoven University of Technology, 5600 M B Eindhouen, The Netherlands (Received: July 9, 1985)

Mossbauer spectra of a carbon-supported iron catalyst, consisting of 3.5 wt % Fe on the carbon black Carbolac-l, show that all metallic iron is superparamagnetic both at 300 and at 80 K. Spectra measured in applied magnetic fields up to 1 T confirm that the signal is due to a-Fe. Analysis of the magnetic field dependence of the spectra shows that the average diameter of the particles is 2.5 & 0.2 nm. The sample is particularly suited for fundamental investigations of superparamagnetic relaxation and surface magnetism in metallic iron particles.

Introduction Superparamagnetic behavior of particles in supported catalysts is a useful phenomenon, as it allows determination of the particle size by analysis of the magnetization of the sample as a function of applied magnetic field.' With iron catalysts, the magnetization can conveniently be measured by Mossbauer spectroscopy. However, superparamagnetism has only rarely been observed in metallic iron,2 although it is very common in small particles of iron oxide. It has been suggested that a small particle size is not a sufficient condition for the Occurrence of superparamagnetism, and that an additional requirement is that the distance between the particles is such that magnetic interactions between the magnetic moments of the particles become negligible.3-4 Magnetic dipole4ipole and exchange coupling tend to align the moments of the particles and thus prevent superparamagnetism. Recently, Jung et al.596have reported that very small particles of metallic iron (a-Fe) with diameters in the 2-3-nm range can be obtained by impregnating a porous carbon black called Carbolac-l (C-l) with an aqueous solution of iron nitrate, followed by reduction in hydrogen a t 675-725 K. Mossbauer spectra of these samples measured in situ under H2 at 295, 80, and 4 K confirmed that mast of the iron is in the metallic state and revealed that about 60% of the a-Fe was superparamagnetic at 80 K.4 The purpose of this Letter is to describe the preparation of a carbon-supported iron catalyst in which all a-Fe exhibits superparamagnetism at 300 and 80 K and the characterization and determination of the size of the a-Fe particles by means of Mossbauer spectroscopy in external magnetic fields. Experimental Section Samples were prepared as follows. Fe203, 90% enriched in the isotope 57Fe(Oak Ridge), was reduced in flowing H2 at 700 K for 24 h and next dissolved in 2 N HN03. This solution was mixed with Fe(N0,)3.9H20 dissolved in H 2 0 and the mixture was added dropwise to the Carbolac-1 (950 m2/g, Cabot) support, until the incipient wetness point was reached. The Fe/C-1 samples contain 3.5 wt % of iron, 13% of which is 57Fe. Impregnated samples were ~~~

( I ) Selwood, P. W. "Chemisorption and Magnetization"; Academic Press: New York, 1975. (2) Marup, S . ; Dumesic, J. A.; Topsne, H.In "Applications of MGssbauer Spectroscopy", Vol. 11, Cohen, R. L., Ed.; Academic Press: New York, 1980; P 1. (3) Marup, S . J . Magn. Magn. Mater. 1983, 37, 39. (4) Niemantsverdriet, J. W.; van der Kraan, A. M.; Delgass, W. N.: Vannice, M. A. J. Phys. Chem. 1985, 89, 6 7 . f 5 ) June. H.-J.: Walker P. L.: Vannice. M. A. J . Catal. 1982. 75. 416. (6j Jung; H.-J.;'Vannice, M. A,; Mulay,'L. N.; Stanfield, R. M.: Delgass, W. N. J . Catal. 1982, 76, 208.

0022-3654/85/2089-4898$01.50/0

carefully dried in air, at 295 K for 12 days, at 325 K for 24 h, at 345 K for 48 h, at 365 K for 72 h, and finally at 400 K for 56 h. A dried sample was pressed into a self-supporting wafer and mounted in the Mossbauer in situ reactor described previously.' Reduction of the Fe/C-1 samples was carried out in the in situ reactor in flowing hydrogen which had been purified in a Pd diffusion cell at 390 K for 1 h, at 535 K for 0.5 h, at 615 K for 16 h, and finally at 675 K for 16 h. It is our experience that careful drying and reduction procedures are essential for obtaining small iron particles. Mossbauer spectra were measured with a constant acceleration spectrometer. Spectra were obtained in applied magnetic fields between 0.01 (the remanence field of the electromagnet) and 1.03 T. The magnetic fields were applied perpendicular to the y-ray direction. The measurements were carried out with the particles exposed to hydrogen. Spectra were folded after measurement in order to remove the geometrical background. Velocities are given with respect to the isomer shift of a-Fe at room temperature.

Results and Discussion Mossbauer spectra of the reduced Fe/C-1 sample, measured at 80 and 300 K in different applied magnetic fields, are shown in Figure 1. The spectra obtained without an applied magnetic field consist mainly of a broad single line with an isomer shift equal to that of a-Fe. Note that these spectra also contain a shoulder at a higher velocity. We attribute this component to iron in the ferrous state. As shown previously, this Fe2+component appears as an unresolved doublet at 300 K6 and as a well-resolved doublet at 80 K.4 The spectra measured in an applied magnetic field show magnetic splitting. The magnitude of the splitting increases with increasing applied magnetic field. The relative line areas in the magnetically split spectra are close to 3:4: 1 : I :4:3 as expected for a sample magnetized perpendicular to the y-ray direction. The lines in the spectra shown in Figure 1 are quite broad. This is presumably due to a particle size distribution which results in a distribution in magnetic hyperfine fields when an external field is applied. Part of the line broadening may also be due to differences in the hyperfine fields of surface atoms and atoms in the interior of the particles. The average hyperfine field, Bobsd,M ~ determined from the positions of the most well-defined lines in the spectra, Le., lines 2 and 5 at 80 K and lines 3 and 4 at 300 K. The magnetic field dependence of Mossbauer spectra of superparamagnetic particles has been discussed in detail in earlier ( 7 ) Clausen, B. S.; Marup, S.; Nielsen, P.;Thrane, N.: Topsae. H J . Phi..\. E . 1979, 12, 439.

0 1985 American Chemical Society

S

Letters

The Journal of Physical Chemistry, Vol. 89, No. 23, 1985 4899

..

. . ,

15‘

2

4

6

8

-8 -6 -4 -2 0

I

1.o

I

2.0

I

3.0

1

6.’ (T-‘) Figure 2. Magnetic hyperfine splittings as a function of the reciprocal applied magnetic field. The full lines are linear fits according to eq 3 . The dotted line is a plot of the exact expression as given by eq 2.

100 K

BO K

-8 -6 -4 -2 0

.

. .

.. ..

2

4

6

8

Velocity (mm B ’ )

Figure 1. Mossbauer spectra of the reduced Fe/C-l catalyst at 80 and 300 K obtained in different applied magnetic fields.

publication^.^*^-^ In short, the magnetic energy 0: a small ferromagnetic particle in an applied magnetic field, B, is given by E =E , - ~ B (1) where E, is the magnetic anisotropy energy and ji is the magnetic moment of the particle which is given by the product of the magnetization, M , and the particle volume, V. For applied magnetic fields of the order of 1 T, the anisotropy energy is normally much smaller than the Zeeman energy of ferromagnetic and therefore the first term in eq 1 may be neglected. At temperatures above the superparamagnetic blocking temperature the magnetic hyperfine splitting in the Mossbauer spectrum is then proportional to2*338

where Bo is the saturation hyperfine field, L { is the Langevin function, k is Boltzmann’s constant, and T i s the temperature. For p B / k T >> 1 one may use the approximation (3)

According to eq 3 a plot of IBobsd - dl as a function of B’ gives a straight line with slope B o ( k T / p )and intercept Bo, and if the magnetization is known the particle-volume can be determined.2,8 Figure 2 shows plots of ( B o b d - B( as a function of B’ for the spectra of the carbon-supported iron particles in Figure 1. The results at both 300 and 80 K are in accordance with the linear relationship predicted by eq 3. From the figure we find that Bo = 32.7 f 2.0 T at 300 K and Bo = 34.3 f 0.5 T at 80 K. These values are within experimental accuracy identical with those of bulk a-Fe after corrections for the influence of the demagnetizing field ( B , = 0.7 T) in small particle^.^^^** The isomer shift and the quadrupole coupling constant are also indistinguishable from the values of bulk a-Fe. (8) Marup, S.; Topsae, H.; Clausen, B. S. Phys. Scr. 1982, 25, 713.

It is therefore concluded that the main component in the spectra is due to a-Fe. The slopes of the straight lines in Figure 2 correspond to magnetic moments of (1.35 f 0.15) X J T-’ at 80 K and (1.6 f 0.3) X J T1 at 300 K, respectively. If we assume that the particles are spherically shaped and that their magnetization is equal to that of bulk metallic iron, the magnetic moments correspond to a particle diameter of 2.5 f 0.2 nm. The size determination of the iron particles in the Fe/C-1 catalyst is based upon eq 3. In the derivation of this expression two approximations have been made. First, the _magnetic anisotropy energy E, in eq 1 is much smaller than ,GB, and second, p B / k T >> 1. The diameter of the iron particles is only about 2.5 nm and, consequently, the magnetic moment of the particles is small. Moreover, the magnetic anisotropic energy constant may be very large since it increases with decreasing particle size.2 Therefore, the validity of these approximations needs to be verified. To start with the seco_nd assu-mption, we note that Figure 2 shows that the value of IBobsd - BI at 300 K for an applied field of 0.55 T is only about 50% of the saturation field Bo. This implies that p B / k T is about 2, Le., not much larger than 1. The dotted line in Figure 2 represents the more precise expression, eq 2 . It is seen that the error introduced by using the linear approximation, eq 3, is smaller than the experimental uncertainty. In microcrystals the magnetic anisotropy energy constant is normally much larger than the bulk value for the magnetocrystalline anisotropy.2.8 For example, the magnetic anisotropy energy constant in 6-nm a-Fe microcrystal^^^^ is of the order of lo5 J m-3 which is larger than the contribution from magnetocrystalline anisotropy by about one order of magnitude, illustrating that contributions such as shape, surface, and stress anisotropy dominate the magnetocrystalline anisotropy. In such cases the anisotropy is expected to be uniaxial rather than cubic2 The value of the magnetic anisotropy energy constant, K , for particles with uniaxial anisotropy can easily be calculated if the superparamagnetic relaxation time is At 80 K the line width of the zero field spectrum is about 1.4 mm s-’ and the spectrum contains a broad component. This indicates a relaxation time in the range 0.5-5.0 ns. From these values we estimate that K = (7.0 f 3.0) X lo5 J m-3. For 6-nm a-Fe particles in hydrogen,8 K is of the order of 1.2 X lo5 J m-3. The larger value (9) Morup, S.; Clausen, B. S.; Topme, H. J . Phys. 1980, 4 1 , C1-331.

J. Phys. Chem. 1985,89, 4900-4903

4900

found in the present particles is in accordance with the expected contribution from the surface due to the larger surface to volume ratio. With a magnetic anisotropy of this size the first term of eq 1 is approximately equal to the second term at B = 0.55 T, and therefore it cannot be neglected. In order to estimate the error in the particle size determination, introduced by neglecting the magnetic anisotropy energy, we have computer-simulated spectra of superparamagnetic a-Fe with various values of K and in different applied magnetic fields.I0 These investigations show that in the present case the error in the particle size determination by use of eq 3 is less than 5%. It is interesting to note that the isomer shift and quadrupole splitting of the 2.5-nm a-Fe particles are within experimental error equal to those of bulk a-Fe. Also Bo, the saturation hyperfine field, is, after correction for the influence of the demagnetizing field, equal to that of bulk a-Fe, within experimental error. Thus, (10) Clausen, B. S.; Christensen, P. H.; Mmup, S., to be published.

in spite of the extremely small particle size, the particles behave, apart from the superparamagnetic relaxation phenomena, essentially as bulk a-Fe. The present Fe/C-1 catalyst provides a very interesting sample for fundamental studies of superparamagnetic relaxation in metallic iron. As a-Fe particles with a diameter of 2.5 nm contain less than 700 atoms, of which about one third are at the surface, the Fe/C-1 catalyst offers also possibilities to investigate the influence of adsorbed gases on the magnetic behavior of the metal atoms at the surface. Measurements of the Fe/C-1 sample at temperatures below that of liquid nitrogen and in magnetic fields up to 7 T are now in progress.

Acknowledgment. The authors acknowledge support from the Netherlands Organization for the Advancement of Pure Research (Z.W.O.) and the Danish Council for Scientific and Industrial Research. Registry No. Iron, 7439-89-6; carbon, 7440-44-0.

Electronic Excitations in Bicyclohexylidene Michael Allan, Institute for Physical Chemistry, University of Fribourg, Switzerland

P. A. Snyder, Department of Chemistry, Florida Atlantic University, Boca Raton, Florida 33431

and M. B. Robin* AT& T Bell Laboratories, Murray Hill, New Jersey 07974 (Received: July 23, 1985)

A reinvestigation of the electronic excited states of the olefin bicyclohexylidene has been carried out using the techniques of photoelectron, electron impact, and multiphoton ionization spectroscopies. Our studies show that the multitude of sharp structures in the vapor spectrum (4000G65 000 cm-’) are due to Rydberg excitations originating at the a MO of the olefin chromophore. These excitations are near degenerate as a consequenceof the extreme alkylation of the olefin and are separable only by combining photoelectron and threshold electron impact techniques. The anomalously high intensities of the overlapping Rydberg bands are due to two underlying valence excitations, whereas only one (T a*)is expected. Two transitions also appear in the crystal spectrum in the same energy region and correlate with the underlying valence excitations of the vapor phase rather than with the more prominent Rydberg promotions; it is suggested that the second valence transition is due to a r(CH2) K* intramolecular charge transfer involving the cyclohexyl rings and the double bond. The two-photon selection rules evident in the multiphoton ionization spectrum of bicyclohexylidene point to C2, symmetry in the vapor phase.

-

-

Introduction In general, the parade of electronic excitations in the alkyl olefins below 8 eV is quite regular, with the pattern of a 3s, a 3p, and K 3d Rydberg excitations interrupted only by the a* valence excitation in the vicinity of the intrusion of the K transition to 3s.’ Valence transitions between a and CT valence MO’s occur at yet higher energies and rarely have been identified with any confidence. The various Rydberg excitations in alkyl olefins have been identified as such on the basis of (a) the term values of the excitations (ionization potential minus the excitation energy), (b) the spectral response to external perturbation on going from the vapor phase to a condensed phase, and (c) their oscillator strengths, f. Elaborating on the above points, the term values of Rydberg transitions terminating at 3s, 3p, and 3d Rydberg orbitals fall in the ranges 27 000-21 000,21000-18000, and 14000-1 1000 cm-I, respectively, with the variation within each range related directly

to the number and bulk of the pendant alkyl groups. Note that in the limit of very large alkyl groups (10 or more carbon atoms per molecule), the term values of transitions terminating at 3s and 3p in olefins are nearly equal. When studied in absorption, the Rydberg bands of a molecule’s vapor-phase spectrum are so strongly perturbed on going into solution or into a solid phaseZthat they often are cited as “missing” in the condensed-phase spectra. Studies of this effect using the circular dichroism of chiral derivatives show that the Rydberg bands upon external perturbation are not “missing” but are broadened and shifted to higher frequencie~.’.~,~ Lowering the temperature of such solutions further shifts the CD band centers to higher frequencies. In contrast, a valence excitation such as .rr K* will be shifted to lower frequency upon entering the condensed phase, and lowering the temperature has no further effect on the transition frequency. As for the oscillator strength distinction between Rydberg and valence transitions in the alkyl

(1) M. B. Robin, “Higher Excited States of Polyatornic Molecules”, Vol. 111, Academic Press, New York, 1985.

(2) A. Gedanken, B. Raz, and J. Jortner, J . Chem. Phys., 58, 1178 (1973). (3) A. F. Drake and S. F. Mason, Tetrahedron, 33, 937 (1977). (4) K. P. Gross and 0. Schnepp, Chem. Phys. Lett., 36, 531 (1975).

-

--

-

0022-3654/85/2089-4900$01.50/0

-

0 1985 American Chemical Societv