Intrinsic magnetic moments of gold nanoparticles - Physical Review ...

PHYSICAL REVIEW B 83, 174446 (2011)

Intrinsic magnetic moments of gold nanoparticles Chi-Yen Li, Chun-Ming Wu, Sunil K. Karna, Chin-Wei Wang, Daniel Hsu, Chih-Jen Wang, and Wen-Hsien Li* Department of Physics and Center for Neutron Beam Applications, National Central University, Jhongli 32001, Taiwan (Received 11 January 2011; revised manuscript received 24 February 2011; published 31 May 2011) In this study the intrinsic magnetic moments developed in 4 nm bare fcc Au nanoparticles were investigated. The magnetization curves reveal a Langevin magnetic field profile, reflecting the existence of spontaneous magnetic moments. Neutron diffraction measurements detect a ferromagnetic moment. The magnetic diffraction patterns are consistent with a moment of μZ = 0.14 μB points in the crystallographic [111] direction at 5 K. Magnetostatic interparticle interactions weaken the localized magnetic moments but enhance the conduction electron density. These behaviors can be understood by considering the merging of the 5d and 6s electron bands triggered by finite-size effects and the lifting of band mixture by interparticle interactions. DOI: 10.1103/PhysRevB.83.174446

PACS number(s): 75.20.−g, 75.45.+j, 81.07.Wx

In their bulk form noble metals are known to be weakly diamagnetic. However, it is now known that the magnetic properties of nanosized particles can differ dramatically from their bulk counterparts. Spontaneous spin-polarized moments have been found in polymer-capped1–7 and cappingfree8,9 noble-metal nanoparticles. Amazingly, a great variety of magnetic phenomena, such as giant paramagnetism,10 superparamagnetism,11 and even permanent magnetism12 have all been found in polymer-capped Au nanoparticles. These changes are understood to originate from the strong chemical affinity of Au atoms to the capping molecules.6,13 The appearance of magnetic moments in thiol-capped Au nanoparticles has been suggested3 to be associated with the 5d localized holes created through the covalent Au-S bonding between the surface Au atoms and the capping S atoms. The strong affinity between them can induce a noticeable amount of charge transfer. As a result, electrons in the surface atoms of Au nanoparticles will redistribute, which induces hybridization between the 5d and 6s electrons, such that the energy of the 5d electrons is closer to the Fermi level. Consequently, unoccupied 5d state densities are created in the surface atoms of Au nanoparticles that generate ferromagnetism.14,15 An electron transfer of ∼0.1 electrons per atom can be expected,14 which seems to be enough to induce the observed large magnetic anisotropies.3 This scenario, however, cannot be used to understand the magnetic moments observed8,9 in cappingfree bare Au nanoparticles, where electron redistribution cannot be initiated from the bonding between the surface atoms and the capping molecules. In this paper we address the question of the origin of the appearance of magnetic moments in capping-free Au nanoparticles. Magnetization and neutron diffraction measurements reveal the existence of intrinsic magnetic moments in 4 nm Au nanoparticles. Magnetic susceptibility measurements performed for various packing fractions of the nanoparticle assembly show that magnetostatic interparticle interactions can trigger electron redistribution in the localized 5d and extended 6s bands, which directly affect the magnetic moments developed in Au nanoparticles. The Au nanoparticles are fabricated employing the gascondensation method. High-purity (99.999%) gold spheres (2 mm in diameter) are evaporated in an Ar atmosphere at a pressure of 0.5 torr, using an evaporation rate of 1098-0121/2011/83(17)/174446(5)

˚ 0.05 A/s. The evaporated particles are collected on a nonmagnetic ceramic plate, placed 20 cm above the evaporation source and maintained at 77 K. After restoration to room temperature, the nanoparticles, which are only loosely attached to the collector, are stripped off by gently knocking the collector plate. The samples thus obtained are in powdered form and consist of a macroscopic amount of individual Au nanoparticles. There are no substrates or capping molecules on these nanoparticles and they are kept in a vacuum at all times. X-ray diffraction (XRD), atomic force microscopy (AFM), transmission electron microscopy (TEM), x-ray fluorescence (XRF), energy dispersive x-ray spectroscopy (EDXS), and atomic absorption spectroscopy (AAS) are used to characterize the sample. Size analyses based on the AFM and TEM images (inset to Fig. 1) reveal a log-normal size distribution centered at 4.0(2) nm and a half width at half maximum of 1.5(2) nm, as shown in Fig. 1. All the x-ray diffraction peaks can be associated with the fcc Au structure, with a lattice constant at room temperature that is ∼1.4% smaller than that of bulk Au. The mean particle diameter determined for the nanoparticle assembly is 4 nm, which is obtained by fitting the diffraction peaks to the diffraction profiles of finite-sized particles, assuming the same size distribution as that obtained from the AFM and TEM images. Chemical analysis by means of XRF and AAS is also performed to search for impurities. The XRF spectra are taken with a Rigaku ZSK Primus II spectrometer, employing the standard setup to scan through 15 mg of the Au nanoparticles. No fluorescence lines other than those from Au are detected. The AAS spectra are taken with a Perkin-Elmer AAnalyst 700 spectrometer, employing a graphite furnace to atomize 20 mg of the Au nanoparticles and a cathode discharge lamp for detecting Fe and Ni. There are no identifiable traces of oxidation phases or elements other than Au in the XRD/EDS/XRF/AAS spectra. We estimate the impurity phases in the sample to be less than 15 parts per million. The magnetization and ac magnetic susceptibility measurements are performed on a physical property measurement system, manufactured by Quantum Design, employing the standard setups. To avoid any aggregation among the Au nanoparticles, the powder is loosened by shaking at 50 Hz for 5 min in a Vortex-Genie mixer. The powdered sample (65 mg)

174446-1

©2011 American Physical Society


80 60 40 20 0

2

4

6

8

10

12

14

16

1

0

7.5 K

3

10 K 15 K

2 1

300 K 55 K

1.5 1.0 0.5

4 nm Au T = 2.1 K

2

5

1

0.0 3

-0.5 -1.0

4

-1.5

MP

0

-0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 Applied magnetic field H (kOe) a

0

Particle diameter (nm)

5K

-2

100

4 nm Au MI

4

M 10 emu / g

= 4.0(2) nm Total counts = 412 HWHM = 1.5(2) nm

120

-2

Number of particles

140

Magnetization ( 10 emu/g )

LI, WU, KARNA, WANG, HSU, WANG, AND LI

10

20

30

40

5

0

Applied magnetic field H a (kOe)

FIG. 1. (Color online) Size distribution obtained from the AFM and TEM images of the Au particle assembly, revealing a log-normal size distribution (solid curve) with a center at 4.0(2) nm and a half width at half maximum of 1.5(2) nm. The insets show three representative TEM images of the particle assembly, revealing a spherical sharp for the nanoparticles.

is then packed into a nonmagnetic cylindrical holder made of Macor (Corning Incorporated), with an inner diameter of 4.2 mm. The initial mass density of the nanoparticle assembly is 4.6% that of bulk Au, with a sample height of 5.3 mm which is much shorter than the height (20 mm) of the sensing coil. The mean interparticle edge-to-edge separation is 1.5 times the particle diameter. The background signal from the holder is ∼2% that of the sample. Figure 2 shows the variations in magnetization M with the applied magnetic field Ha of the 4 nm Au nanoparticle assembly, taken at several representative temperatures. Three components, each with its own thermal characteristics, can be identified from the M(Ha ) curves. In the low-Ha regime, M increases rapidly with increasing Ha , becoming saturated at Ha ∼ 2 kOe. This component, marked MP , can be described using a Langevin profile (solid curves) L(x) ≡ coth(x) − 1/x, where x ≡ μP Ha kB T , μp is the mean particle moment, and kB is the Boltzmann constant. The appearance of this Langevin profile can be understood as indicating the randomly oriented assembly of magnetic Au nanoparticles with a mean particle moment μp at a temperature T that are aligned by Ha . A value of μp = 166 μB is thus extracted from the fits for the assembly at 5 K.16 Inflection points in M(Ha ) are clearly evident at Ha ∼ 2 kOe, showing the existence of another component in M which disappears above 15 K. This component, marked MI , can be described using a Brillouin profile of order J = 1/2 (solid curves), which can be understood as the condensation of quantum-confined electrons into the Zeeman-split spin-polarized states.9 Lenz diamagnetic responses become dominant in the high-Ha regime, such that M gradually decreases with increasing Ha . Magnetic hystereses associated with MP are evident at all temperatures studied. A representative M(Ha ) hysteresis loop obtained at 2.1 K is shown in the inset to Fig. 2, revealing a coercive field of 40 Oe. Surprisingly, the coercive field increases (rather

FIG. 2. (Color online) Magnetization curves of the 4 nm Au particle assembly, taken at several representative temperatures. The solid curves indicate the fits to the profile discussed in the text. Two magnetic components, marked MP and MI , appear in the M(Ha ) curves taken below 15 K. MI disappears above 15 K, whereas MP remains visible even at 300 K. The inset shows the hysteresis loops associated with MP obtained at 2.1 K.

than decreases) noticeably with increasing temperature. It reaches 74 Oe at 300 K. The appearance of magnetic hysteresis indicates that the anisotropy energy of the 4 nm fcc Au particles is higher than their thermal energy at 300 K and the relaxation time is noticeably longer than the measuring time. Similar magnetization profiles have also been observed9 in 3.5 nm icosahedral Au particles. The spontaneous magnetic moment and induced Zeeman magnetization of the 3.5 nm icosahedral Au particles are factors of 2.8 and 14, respectively, larger than those of the present 4 nm fcc Au particles. This is understandable since the number of atoms packed in a 3.5 nm icosahedral Au particle is a factor of 1974 923 = 2.1 lower than that of a 4 nm fcc Au particle. It appears that the atomic geometry of the particles can significantly affect the magnetism developed in nanoparticles. The existence of magnetic moments in 4 nm Au particles is confirmed by the neutron diffraction measurements. These measurements were conducted at the Bragg Institute, ANSTO, using the high intensity powder diffractometer Wombat, employing Ge(115) monochromator crystals to select an incident ˚ and a cylindrical vanadium can wavelength of λ = 1.5419 A to hold the nanoparticles (∼0.9 g). Figure 3(a) displays the difference pattern between the diffraction patterns taken at 5 and 100 K, revealing noticeable increase in intensity upon cooling from 100 to 5 K. The slightly asymmetric profiles of the magnetic peaks are the direct result of slight differences in the peak positions of the 5 and 100 K patterns, but are not associated with low-dimensional magnetic ordering of any kind. These magnetic intensities appear on top of the nuclear ones, showing a ferromagnetic arrangement for the Au moments. The widths of the magnetic peaks are the same as those of the nuclear ones, indicating that the magnetic moments are distributed over the whole nanoparticle, rather than located solely on the surface. A ferrimagnetic-like moment arrangement, where the surface moments and the

174446-2

INTRINSIC MAGNETIC MOMENTS OF GOLD NANOPARTICLES

(a) 4 nm Au

Ha= 0 Oe )

(420)

p

0.9 0.6

0 0.3

30

45

60

75

90

105

120

0

135

(b)

1.45

4 nm Au

T=5K

1.40

0.01

(400)

1.35 4 nm Au (111) (400) (220)

H = 1 kOe

0.00

(200)

Integrated intensity (arb. units)

0.02


0.03

1.30 1.25 1.20

Temperature (K)

0

5

10

15

20

25

30


0.04

10

20

30

40

50

60

70

Packing fraction f (%)

Scattering angle 2 (deg)

Magnetization (emu/g)

T=5K

1.2

2

(400)

(220)

2000

(331)

(222)

(200)

4 nm Au

1.5 B

(311)

(111)

4000

1.8

I5 K - I100 K

( 10

Intensity (counts / 190 min)

6000


1.15

Applied magnetic field Ha (kOe) FIG. 3. (Color online) (a) Difference pattern between the neutron diffraction patterns taken at 5 and 100 K, revealing increases in the intensity upon cooling from 100 to 5 K. (b) Variations in magnetization with Ha together with that of the integrated intensity of the (400) reflection. The inset shows the temperature dependence of four reflections, revealing a ∼5% increase in the intensities upon cooling from 300 to 5 K.

core moments point in different directions, has been identified in 9 nm spherical magnetite particles17 and suggested for 3.5 nm Au particles.9 Unfortunately, the resolution of the present study is not sufficient to resolve the relative spin arrangement between the surface and the core moments if they point in different directions. The variations in the reflection intensity with applied magnetic field scale nicely with the MP component in the M(Ha ) curves. A representative plot is illustrated in Fig. 3(b). Note that the Ha -induced Zeeman component MI will not appear in the neutron diffraction intensity. The temperature dependence of the (111), (200), (220), and (400) reflections are shown in the inset to Fig. 3(b). The intensities increase by ∼5% upon cooling from 300 to 5 K. The diffraction pattern shown in Fig. 3(a) can be described (solid curve) reasonably well by assuming the development of a ferromagnetic moment of μZ = 0.08 μB points in the [111] crystallographic direction. This μZ = 0.08 μB indicates the moment developed upon cooling from 100 to 5 K. A magnetic moment of μZ = 0.14 μB is obtained by comparing the diffraction intensities obtained at 5 and 300 K. This magnetic moment for the 4 nm Au particles is underestimated, since the magnetization measurements show that the Curie temperature is above 300 K.

FIG. 4. (Color online) Effects of the packing fraction on the mean particle moment of the 4 nm Au particle assembly, revealing that μp decreases progressively with increasing packing fraction but reaching a nonzero finite value. The inset displays a schematic representation of the device used to adjust the packing fraction of the particle assembly.

Using the holder shown in the inset to Fig. 4, the packing fraction f of the nanoparticle assembly for the magnetization measurements can easily be adjusted by turning the tap cap. This setup allows us to fine-tune the packing fraction and perform measurements on the very same nanoparticles at different packing fractions. Similar M(Ha ) curves are obtained at higher packing fractions. Interestingly, the mean particle moment μp decreases progressively with increasing packing fraction but reaches a nonzero finite value when the nanoparticles are densely packed, as shown in Fig. 4. The value of μp decreases by a factor of 5.4 as f increases from 4.6% to 72%. In addition, the saturation MI also decreases progressively with increasing packing fraction. It reduces by a factor of 1.5 as f increases from 4.6% to 72% at 5 K. This reduction rate increases noticeably at higher temperatures, reflecting the thermal population characteristics of Zeeman magnetization. These observations reveal that increasing the packing fraction brings the assembly to approach its bulk form. The x-ray diffraction patterns taken on the f = 4.6% and 72% assemblies are indistinguishable, showing that the particle size is not significantly altered by cold pressing when adjusting the packing fraction. In a loosely packed assembly the nanoparticles are weakly linked together, so that magnetostatic interparticle interactions are insignificant. Increasing the packing fraction will progressively close up the spatial separations between the nanoparticles, which increase the significance of magnetostatic interparticle interactions. It is mainly the magnetostatic interparticle interactions that result in the reduction in μp , reflecting the fact that the density of localized magnetic electrons is reduced by the magnetostatic interparticle interactions. The effects of magnetostatic interparticle interactions on the density of the conduction electrons, on the other hand, can be examined through the magnetic susceptibility measured at various packing fractions. Figure 5(a) displays the temperature dependences of the in-phase component χ of the ac magnetic susceptibility, taken at representative packing fractions. The χ (T ) curves shift

174446-3

LI, WU, KARNA, WANG, HSU, WANG, AND LI


-5

' 10 emu/g Oe

6

4 nm Au

5 4

(a)

Solid lines: ' = 0 + C /T

f = 6.3 % Hac = 15 Oe

3

3

F req. = 10 H z f = 22.5 %

2 f = 72 % 1 0

25

50

75

100

125

150

Temperature (K)

(b)

-5

3.5

5.2

3.0

4.8

2.5

4 nm Au

4.4

2.0 4.0

1.5

3.6

1.0 0

10

20

30

40

50

60

-5

0

5.6

C (10 emu K/g Oe)

(10 emu/g Oe)

4.0

70

Packing fraction f (%) FIG. 5. (Color online) (a) Temperature dependence of the in-phase component of the ac magnetic susceptibility, taken at several representative packing fractions. (b) Plot of the variations in χ 0 and Curie constant C with packing fraction for a direct comparison.

progressively to lower values upon increasing the packing fraction. They can be described using the Curie C/T profile plus a constant value χ0 (solid curves). The variations in the Curie constant C and χ0 with packing fraction are illustrated in Fig. 5(b). It can be seen that C increases while χ0 decreases with increasing packing fraction. Figures 4 and 5(b) clearly show that magnetostatic interparticle interactions affect μp , χ0 , and C coherently. Although the band structure of bulk Au can be separated into five narrow 5d bands and one broad 6s band to accommodate the additional eleven electrons, at some values of wave vector all six electron bands are closer together.18 There are at least three effects that can alter the electron band structure and trigger electron redistribution upon reducing the particles to nanosizes. First, 37% of the atoms in a 4 nm fcc Au particle are located in the outmost surface layer, where disruption of lattice periodicity, known as the small-size effect, can result in an electron band structure that is significantly different from its bulk counterpart. Second, it is known that the Fermi energy of a particular dimension of a system is proportional to the electron density of that particular dimension. If the surface and core atoms of a nanoparticle gather the same number of electrons, the Fermi energy of the surface electrons is then slightly higher than that of the core electrons, because the

surface atoms are more closely packed due to the appearance of surface stress which deforms the surface of the nanoparticle into a curved one. Electron transfer from the surface region into the core can be anticipated for matching the Fermi energy of the two regions, which can result in noticeably different electron distributions for the core and surface regions. This transfer of charges from the surface atoms into the core has been found19–21 to be energetically favorable to stabilizing the core. Third, it should be noted that quantum confinement splits the electron bands near the Fermi level into discrete levels, which narrow the conduction band and reconstruct the electron distribution. When the nanoparticles are very loosely assembled so that interparticle interactions are not yet significant, the measurements mainly reflect the intrinsic properties of individual nanoparticles. The existence of localized moments is directly confirmed by the neutron diffraction measurements (Fig. 3). Although the moments are weak (μZ = 0.14 μB ), localized 5d holes do exist to reveal ferromagnetism. They remain visible even at 300 K. The mean particle moment μp is the net moment developed in a nanoparticle, which indicates the amount of uncompensated 5d holes for ferromagnetism. The Curie constant C is proportional to the density of the 6s conduction electrons. The observation of μp decreasing together with C increasing [Figs. 4 and 5(b)] with increasing packing fraction shows that the charge redistribution initiated by the magnetostatic interparticle interactions involves both conduction 6s electrons and localized 5d holes. This can happen only when the 6s and 5d bands are energetically close to each other. The general picture established so far for the appearance of magnetic moments in polymer-capped Au nanoparticles is that the strong chemical affinity between the capping molecules and the Au atom is a prerequisite for triggering electron redistribution. This creates localized 5d holes for ferromagnetism in the surface region. Ferromagnetism is thus associated only with the Au atoms on the surface while the core atoms remain diamagnetic. The present study shows further that magnetic moments do develop in bare Au nanoparticles, and the core atoms contribute to the ferromagnetism as well. It appears that 5d and 6s band mixture is visible in 4 nm Au particles. Bringing the nanoparticles close to each other will strengthen the links between them. These magnetostatic interparticle interactions will lead to broadening of the conduction band, lowering the Fermi energy of the surface electrons EFS and weakening the small-size effect. Broadening of the conduction band will push the electrons away from the confinement. The lowering of EFS will trigger a portion of the electrons to flow back to the surface region, which releases the uncompensated electrons back to the surface states for conduction. Weakening of the small-size effect will reduce the amount of modification in the electron band structure, which lifts the 5d-6s band mixture. All these effects push the 4 nm Au particle assembly toward the bulk Au state, but the finite-size character remains even when the particles are in mutual contact. This work is supported by the National Science Council of Taiwan, under Grant No. NSC 98-2112-M008-016-MY3.

174446-4

INTRINSIC MAGNETIC MOMENTS OF GOLD NANOPARTICLES *

[email protected] H. Hori, T. Teranishi, Y. Nakae, Y. Seino, M. Miyake, and S. Yamada, Phys. Lett. A 263, 406 (1999). 2 H. Hori, Y. Yamamoto, T. Iwamoto, T. Miura, T. Teranishi, and M. Miyake, Phys. Rev. B 69, 174411 (2004). 3 P. Crespo, R. Litrán, T. C. Rojas, M. Multigner, J. M. de la Fuente, J. C. Sánchez-López, M. A. Garc´ıa, A. Hernando, S. Penadés, and A. Fernández, Phys. Rev. Lett. 93, 087204 (2004). 4 Y. Yamamoto, T. Miura, M. Suzuki, N. Kawamura, H. Miyagawa, T. Nakamura, K. Kobayashi, T. Teranishi, and H. Hori, Phys. Rev. Lett. 93, 116801 (2004). 5 P. de la Presa, M. Multigner, J. de la Venta, and M. A. Garc´ıa, J. Appl. Phys. 100, 123915 (2006). 6 M. A. Garc´ıa, J. M. Merino, E. Fernández Pinel, A. Quesada, J. de la Venta, M. L. Ru´ız González, G. R. Castro, P. Crespo, J. Llopis, J. M. González-Calbet, and A. Hernando, Nano Lett. 7, 1489 (2007). 7 J. S. Garitaonandia, M. Insausti, E. Goikolea, M. Suzuki, J. D. Cashion, N. Kawamura, H. Ohsawa, I. G. de Muro, K. Suzuki, F. Plazaola, and T. Rojo, Nano Lett. 8, 661 (2008). 8 W.-H. Li, S. Y. Wu, C. C. Yang, S. K. Lai, K. C. Lee, H. L. Huang, and H. D. Yang, Phys. Rev. Lett. 89, 135504 (2002). 9 C.-M. Wu, C.-Y. Li, Y.-T. Kuo, C.-W. Wang, S.-Y. Wu, and W.-H. Li, J. Nanopart. Res. 12, 177 (2010). 10 I. Carmeli, G. Leitus, R. Naaman, S. Reich, and Z. Vager, J. Chem. Phys. 118, 10372 (2003). 11 P. Dutta, S. Pal, M. S. Seehra, A. Anand, and C. B. Roberts, Appl. Phys. Lett. 90, 213102 (2007). 1

PHYSICAL REVIEW B 83, 174446 (2011) 12

J. de la Venta, A. Pucci, E. Fernández Pinel, M. A. Garc´ıa, C. de Julián Fernández, P. Crespo, P. Mazzoldi, G. Ruggeri, and A. Hernando, Adv. Mater. 19, 875 (2007). 13 R. Litrán, B. Sampedro, T. C. Rojas, M. Multigner, J. C. SánchezLópez, P. Crespo, C. López-Cartes, M. A. Garc´ıa, A. Hernando, and A. Fernández, Phys. Rev. B 73, 054404 (2006). 14 Peng Zhang and T. K. Sham, Phys. Rev. Lett. 90, 245502 (2003). 15 C. López-Cartes, T. C. Rojas, R. Litrán, D. Mart´ınez-Mart´ınez, J. M. de la Fuente, S. Penadés, and A. Fernández, J. Phys. Chem. B 109, 8761 (2005). 16 The fits begin with the M(Ha ) curves taken at high temperatures, using a Langevin plus a linear diamagnetic term. The parameters obtained were then used as the starting values of the Langevin and diamagnetic parameters for the fits to the curves taken at low temperatures, where an additional Brillouin term was then employed. We encountered no difficulty in converging the fits without constraint. 17 K. L. Krycka, R. A. Booth, C. R. Hogg, Y. Ijiri, J. A. Borchers, W. C. Chen, S. M. Watson, M. Laver, T. R. Gentile, L. R. Dedon, S. Harris, J. J. Rhyne, and S. A. Majetich, Phys. Rev. Lett. 104, 207203 (2010). 18 N. W. Ashcroft and N. D. Mermin, Solid State Physics (Saunders College, Philadelphia, 1976), pp. 288–293. 19 C. M. Chang and M. Y. Chou, Phys. Rev. Lett. 93, 133401 (2004). 20 M. Pereiro and D. Baldomir, Phys. Rev. A 72, 045201 (2005). 21 M. Pereiro, D. Baldomir, and J. E. Arias, Phys. Rev. A 75, 063204 (2007).

174446-5