Sintering Prevention

3B2v8:06a=w ðDec 5 2003Þ:51c XML:ver:5:0:1

Prod:Type:COM pp:126ðcol:fig::NILÞ

MAGMA : 23188

ED:V:Lakshmi PAGN:N:Manjunatha SCAN:Raj

ARTICLE IN PRESS

1 3

Journal of Magnetism and Magnetic Materials ] (]]]]) ]]]–]]] www.elsevier.com/locate/jmmm

5

9 11

Sintering prevention and phase transformation of FePt nanoparticles Y. Dinga, S.A. Majeticha,, J. Kimb, K. Barmakb, H. Rollinsc, P. Sidesd

13

F

7

a

O

Physics Department, Carnegie Mellon University, Pittsburgh, PA 15213, USA b MSE Department, Carnegie Mellon University, Pittsburgh, PA 15213, USA c Idaho National Engineering and Environmental Laboratory, Idaho Falls, ID 83415, USA d Chemical Engineering Department, Carnegie Mellon University, Pittsburgh, PA 15213, USA

O

15

PR

17 19

25 27 29

D

Two approaches attempted to overcome FePt nanoparticle sintering during the transformation to the high coercivity L10 phase, which currently limits the use of these nanoparticles in data storage media. High-pressure treatment of dilute nanoparticle solutions failed to prevent sintering due to surfactant decomposition above 360 1C. By pre-annealing nanoparticle monolayers to decompose the surfactant, and then coating with an immiscible SiO2 matrix, sintering was prevented with annealing temperatures up to 700 1C. r 2004 Published by Elsevier B.V.

TE

23

Abstract

EC

21

PACS: ’; ’; ’ Keywords: ’; ’; ’

R

31

R

33

41 43 45 47

C

39

N

37

Self-assembled FePt nanoparticle arrays have been the subject of much interest due to their potential use as ultrahigh density magnetic storage media [1]. A key advantage of the self-assembly approach is the high degree of uniformity in the grain size and position. In bulk phase FePt, the chemically ordered high magnetocrystalline aniso-

U

35

O

1. Introduction

Corresponding author. Tel.: +1-412-268-3105; fax: +1-

412-681-0648. E-mail address: [email protected] (S.A. Majetich). 0304-8853/$ - see front matter r 2004 Published by Elsevier B.V. doi:10.1016/j.jmmm.2004.07.011

tropy L10 phase is thermodynamically more stable. In direct synthesis of L10 phase, FePt nanoparticles has been reported [2], but so far this method leads to polydisperse particles that cannot selfassemble into arrays. Monodisperse FePt nanoparticles as made are FCC, and must be annealed to high temperatures to induce the transformation. Unfortunately this annealing process sinters the particles together, destroying the key advantage over existing magnetic recording media. Sintering occurs because the surfactant coating around the particles decomposes at
400 1C. Without the

49 51 53 55 57 59 61 63

MAGMA : 23188 ARTICLE IN PRESS Y. Ding et al. / Journal of Magnetism and Magnetic Materials ] (]]]]) ]]]–]]]

9 11 13 15 17 19 21

TE

23 2. Experimental

27

Three nanometer FePt nanoparticles were prepared using standard high temperature chemical methods [1]. The surfactant coating was a 1:1 mixture of oleic acid and oleylamine. The particles were washed with ethanol, centrifuged, and redispersed in hexane several times to improve the monodispersity. Three nanometer FePt nanoparticles dissolved in hexane were heated under high pressures using a specially designed stainless steel cell [6]. Two samples from the same batch were heated to 360 1C at
340 atm for 45 min and 450 1C at
300 atm for 5 min, respectively. The high-pressure treated nanoparticles were structurally and magnetically characterized to look for evidence of the FCC to L10 phase transformation and particle sintering. After high-pressure treatment, more surfactant was added, and drops of solution were evaporated on transmission electron microscopy (TEM) grids. The samples were observed using a 200 keV JEOL 2000 TEM. Glancing angle X-ray diffraction (XRD) measure-

39 41 43 45 47

R

R

O

37

C

35

N

33

U

31

EC

25

29

F

7

O

5

ments were made with a Philips X’Pert highresolution X-ray diffractometer operating at Cu Ka radiation (l ¼ 0:154 nm), to look for the development of the L10 phase and evidence of line narrowing associated with grain growth. Here the angle between the incident X-ray beam and the sample surface was fixed at 61, and detector angle 2y was scanned from 101 to 1001. The particle moment and coercivity were measured using a quantum design superconducting quantum interference device (SQUID) magnetometer at 10 K. To prepare over-coated monolayer arrays, particles were first dissolved in 1,1,1-trichloroethane (CH3CCl3), and then a droplet of the solution was evaporated on oxidized Si wafers that had a 100 nm SiO2 surface layer. The oxidized Si wafers were used to prevent silicide formation. The sample was then annealed in a glass tube under vacuum (400 1C, 30 min) to decompose most of the surfactant [3]. After this initial annealing stage at 400 1C, there was no measurable weight loss after heating to higher temperatures. To prevent the FePt nanoparticles from oxidizing after the surfactant decomposed, the wafer was temporarily coated with eicosane (C20H42) for transfer to the sputtering chamber. Eicosane is a solid at room temperature but boils at a relatively low temperature. Its chemical structure makes it unlikely to bind to the particle surface. The wafers were then heated under vacuum (200 1C, 30 min) to thoroughly remove the eicosane. The base pressure of the sputtering chamber was less than 5  107 Torr. A 2 min pre-sputtering process was used to stabilize the RF plasma. Next, 100 nm SiO2 layers were deposited by plasma sputtering of a silica target, with 5 mTorr of flowing Ar. The sputtering time was 1 h, at a rate of
0.03 nm/s to ensure a smooth film. While 100 nm is too thick to be usable for data storage media, excess silica was used in these preliminary experiments to ensure complete coverage. An excess of silica will not matter for fundamental studies of the transformation temperature as a function of particle size. Following the SiO2 over-coat, the wafers were then annealed in flowing H2 for 30 min at different temperatures, ranging from 400 to 700 1C. In order to image the particles after processing, and directly determine whether sintering had

O

3

steric barrier, the particles coalesce at temperatures sufficient for the L10 phase transformation [3]. Here, we report the findings of two new approaches to overcome this sintering problem. In the first, high pressures were used to heat a solution of the particles to high temperature, in order to initiate the phase transformation while the particles were well-separated. The high pressure was used to prevent the liquid from evaporating while the temperature was increased. In the second approach, particles were self-assembled into monolayer arrays, and then heated to moderate temperatures in order to decompose the surfactant [4] without sintering. The arrays were then over-coated with an immiscible silica matrix, and annealed at higher temperatures to transform the FePt into the L10 phase. Here sintering is prevented by replacing the surfactant coating with SiO2, which is chemically stable up to 700 1C, well above the L10 transformation temperature [5].

PR

1

D

2

49 51 53 55 57 59 61 63 65 67 69 71 73 75 77 79 81 83 85 87 89 91 93 95

MAGMA : 23188 ARTICLE IN PRESS Y. Ding et al. / Journal of Magnetism and Magnetic Materials ] (]]]]) ]]]–]]]

9 11

F

7

O

13 3. Results and discussion

17

After the high-pressure treatment at 360 1C and
340 atm for 45 min, the 3 nm FePt nanoparticles remained in solution, and did not respond to a magnet. After heating to 450 1C at
300 atm for 5 min, most of the particles precipitated and responded weakly to a magnet. Both samples were washed with ethanol and redispersed in hexane, in order to improve the surfactant coating and monodispersity. Fig. 1 shows that the 3 nm FePt particles treated at 360 1C were still able to selfassemble into arrays. However, TEM showed that the particles treated at 450 1C were found to have agglomerated and partly sintered together. The high pressure did not significantly change the surfactant decomposition temperature [4]. Once the surfactant decomposed, interparticle collisions caused agglomeration, even in dilute solution. The porous carbonaceous remnants of the decomposed surfactant did not prevent sintering.

27 29 31 33 35

EC

25

43

47

Fig. 1. TEM image of 3 nm FePt nanoparticle arrays formed from particles treated at 360 1C and
340 atm for 45 min.

53 55 57 59 61 63 65 67 69 71 73 75 77

83 85 87

N

45

51

81

89

U

41

49

79

C

37 39

Fig. 2. Magnetic moment at 10 K as a function of the applied field, for 3 nm FePt nonoparticles treated at high pressure for various temperatures.

R

23

R

21

O

19

TE

15

O

5

Fig. 2 shows the magnetization curves of the FePt nanoparticle solutions at 10 K. As made, the particle moment is extremely low due to structural disorder [4]. For the solution treated at 360 1C, the magnetic moment increased by a factor of four, but the coercivity was still
100 Oe. After treatment at 450 1C, the coercivity increased to 1800 Oe, and the magnetization curve showed clear saturation behavior expected for a ferromagnet. XRD showed that the structure of the FePt nanoparticles treated at 450 1C was mainly FCC, though there was evidence of a weak (1 1 0) peak from the L10 structure in the XRD glancing angle scan (Fig. 3). The chemical order parameter S is

PR

3

occurred, thinned Si3N4 TEM grids were also used as substrates. The same procedure was used to prepare particles on the Si3N4 grids, except that 20 nm SiO2 layer under-coats and over-coats were used to retain electron transparency. The room temperature coercivity of the overcoated samples was measured after annealing by an alternating gradient force magnetometer (AGM). The particle morphology after annealing was observed by TEM to see if the particles had sintered.

D

1

3

91 Fig. 3. Glancing angle XRD pattern of FePt treated at various temperatures and pressures. The broad peaks at around 201 are from the excess surfactant. The peak at around 561 in 450 1C pattern is from the off Bragg condition (3 1 1) diffraction of the Si substrate. The peak at around 321 in 450 1C pattern is from (1 1 0) diffraction of FePt L10 phase.

93 95

MAGMA : 23188 ARTICLE IN PRESS Y. Ding et al. / Journal of Magnetism and Magnetic Materials ] (]]]]) ]]]–]]]

ð1Þ

7

25 27 29 31 33 35 37 39 41 43 45 47

TE

23

EC

21

R

19

R

17

O

15

C

13

N

11

U

9

where I(hkl) is the intensity of the (hkl) diffraction peak. From Fig. 3, S was estimated to be 0.3070.05, indicating that the ordering is far from complete. Moreover, the narrowing of the FCC diffraction peak widths at this temperature confirms the TEM observation of particle sintering and grain growth. From these results alone, it is unclear whether the transformation has begun because the sample has reached a threshold temperature, or because some particles have sintered and reached a threshold size, above which the L10 phase is most stable. A small fraction of the transformed particles could dominate the coercivity. From Fig. 3, the calculated FCC lattice constant a was 0.390 nm for the FePt treated at 360 1C and 0.383 nm for the FePt treated at 450 1C. In comparison, the lattice constant of bulk Pt is 0.392 nm and the lattice constant of bulk FCC FePt is 0.382 nm [8]. This shift in the lattice constant following the heat treatment could be due to changes in the chemical homogeneity. When a FePt nanoparticle forms, a Pt seed forms first, and then Fe deposits around the Pt cluster. Pt nucleates more readily, and in the absence of Fe(CO)5, small Pt particles of
1 nm diameter are formed, even at much lower temperature. It is unclear how much interdiffusion occurs during this deposition process. With minimal diffusion, we would expect a Pt-rich core and an Fe-rich shell, with a gradient of composition. In the extreme case, a 3 nm Fe50Pt50 nanoparticle would have a pure Pt core of 2.4 nm diameter and would be surrounded by a 0.3 nm pure Fe shell. For the FCC structure with a lattice constant a
0.39 pffiffiffi nm, the distance between nearest neighbors, ð 2=2Þa; is only
0.3 nm, which means that the shell is just one layer of Fe atoms. With more Pt in the particle center, the average lattice constant determined by XRD would be dominated by Pt and very close to the 0.392 nm value of bulk Pt. Below temperatures

F

½I ð110Þ =I ð111Þ experimental ; ½I ð110Þ =I ð111Þ bulk

O

S2 ¼

O

5

where there is significant atomic diffusion (o400 1C), the atomic configuration won’t change significantly. Once atomic diffusion begins, the lattice constant would shift toward that of bulk FePt. We have observed similar lattice spacings in 5 nm FePt particles annealed without high pressure. The surfactant coating was the key to keeping particles from sintering, yet the organic surfactant decomposes at about 400 1C even when treated under high pressure. Even with the lowest feasible transformation temperatures, the surfactant breakdown limits the high-pressure approach. SiO2 was chosen as an alternative barrier because it is stable and does not react with Fe or Pt up to 700 1C. Fig. 4 shows TEM images of 3 nm FePt particles as initially self-assembled on a SiO2 surface, with the surfactant decomposed, and with a SiO2 over-coat after annealing at high temperature. This demonstrates that neither the annealing at 400 1C nor the coating process sintered the particles. Fig. 5 shows how the surfactant residue enables particle sintering at high temperatures, in cases where the particles were not thoroughly washed to remove excess surfactant. Here, the sputtered SiO2 overcoats the remnants of excess surfactant surrounding the particles. During the high temperature anneal, a few of the particles grew in size due to sintering. However, surrounding each of the sintered particles, the TEM images show a diffuse background associated with carbonaceous material. Presumably, the decomposed surfactant is porous enough that particle sintering occurs. Fig. 6 shows the room temperature coercivity of the unagglomerated 3 nm FePt nanoparticle arrays deposited on an oxidized Si wafer with a SiO2 overcoat and then annealed at different temperatures. The results show that the 3 nm FePt nanoparticle coercivity began to increase at 560 1C. The value of Hc for these particles is much smaller than the thousands of Oersteds reported for sintered FePt arrays. The small Hc of 3 nm FePt particles could occur because the phase transformation is incomplete, or because strains from the SiO2 coating prevent full transformation. However, it could also arise because the L10 phase is not the stable state for very small particles. The

PR

3

widely used to characterize the degree of phase transformation from FCC to L10 phase [7]. S is defined by the following equation:

1

D

4

49 51 53 55 57 59 61 63 65 67 69 71 73 75 77 79 81 83 85 87 89 91 93 95

MAGMA : 23188 ARTICLE IN PRESS Y. Ding et al. / Journal of Magnetism and Magnetic Materials ] (]]]]) ]]]–]]]

5

49

3

51

5

53

7

55

9

57

11

59

F

1

O

13 15

19

O

PR

17

Fig. 4. TEM images of FePt nanoparticles at different processing stages. All samples used thinned Si3N4 grids with a 20 nm base layer of sputtered SiO2. (a) A monolayer of surfactant-coated particles self-assembled by evaporating a drop of CH3CCL3 dispersion. The length scale of the ordering is lower than on a standard carbon or SiO-coated grid, but the nearest neighbor separation is the same; (b) Part of the monolayer after decomposing the surfactant by a pre-annealing process, in which the grid is sealed in an evacuated ampoule and heated in to 400 1C for 1 h. Note that the particle size and interparticle spacing are unchanged; (c) Part of the monolayer after sputtering a layer of 20 nm SiO2, and then annealing at 600 1C for 30 min. The particle size and spacing are preserved by this process.

D

21

TE

23 25

EC

27 29

41 43 45 47

71 73 75

81

4. Conclusions

89

last hypothesis has been suggested previously, based on annealing of FePt clusters deposited in ultrahigh vacuum [9]. The threshold average size for transformation to the L10 phase was 3.5 nm. Further experiments on monodisperse particles with larger sizes are underway.

Two new approaches were tried to overcome the sintering problem in the FePt nanoparticle phase transformation, high-pressure treatment and SiO2 over-coating. The results showed the important connection between the surfactant decomposition and sintering. The surfactant decomposition tem-

91

O

85

C

39

69

Fig. 6. Coercivity versus the maximum annealing temperature, for 3 nm FePt particles on an oxidized Si wafer, over-coated with SiO2 and annealed at different temperatures for 30 min.

N

37

67

Fig. 5. TEM image of annealed FePt particles with SiO2 overcoat showing some particle growth. Note that surrounding each of the larger particles is a diffuse halo due to the remnants of excess surfactant. If excess surfactant remains, the sputtered SiO2 is unable to form an impervious barrier between the particles. At the high temperatures needed for the phase transformation, atomic diffusion is significant and sintering will occur unless there is an immiscible barrier between the particles.

U

35

65

79

R

33

63

77

R

31

61

83

87

93 95

MAGMA : 23188 ARTICLE IN PRESS Y. Ding et al. / Journal of Magnetism and Magnetic Materials ] (]]]]) ]]]–]]]

9 11 13 15 17 19 21

Acknowledgments 23

35

[1] S. Sun, C.B. Murray, D. Weller, L. Folks, A. Moser, Science 287 (2000) 1989. [2] B. Jeyadevan, K. Urakawa, A. Hobo, N. Chinnasamy, K. Shinoda, K. Tohji, D. Djayaprawira, M. Tsunoda, M. Takahashi, Jpn. J. Appl. Phys. 42 (2003) L350. [3] Y. Ding, S. Yamamuro, D. Farrell, S.A. Majetich, J. Appl. Phys. 93 (2003) 7411. [4] Y. Sahoo, H. Pizem, T. Fried, D. Golodnitsky, L. Burstein, C.N. Sukenik, G. Markovich, Langmuir 17 (2001) 7907. [5] J.W. Harrell, S. Wang, D.E. Nikles, M. Chen, Appl. Phys. Lett. 79 (2001) 4393. [6] The experimental set-up for the high-pressure treatment studies consisted of a Shimadzu LC-10AT HPLC pump, a high-pressure reactor (model MS-11, High Pressure Equipment Company), a 1’’ tube furnace, pressure relief valve and high-pressure valves and tubing. The sample temperature was monitored using a k-type thermocouple in contact with the sample. [7] T. Shima, T. Moriguchi, S. Mitani, K. Takanashi, Appl. Phys. Lett. 80 (2002) 288. [8] PCPDFWIN v. 2.02, JCPDS-International Centre for Diffraction Data, 1999. [9] Y.K. Takahashi, T. Ohkubo, M. Ohnuma, K. Hono, J. Appl. Phys. 93 (2003).

37

TE

EC

R R O C N

27

U

25

We would like to thank Data Storage System Center (DSSC) at Carnegie Mellon University. S.A.M. acknowledges support from the National Science Foundation Grant ] CTS-0227645. H.W.R. acknowledges support from the US

31

References

F

7

29

33

O

5

Department of Energy through the INEEL Laboratory Directed Research and Development (LDRD) Program under DOE Idaho Operations Office Contract DE-AC07-99ID13727.

O

3

perature was not significantly changed under high pressure. Though the particles in solution were well separated, on average, once the surfactant decomposed, they sintered together. Measurements confirmed that sintering occurred at temperatures sufficient for the onset of the FCC to L10 FePt phase transformation. The SiO2 over-coating approach replaced the surfactant with an inorganic silica coating. Here the particles did not sinter even when annealed at 600 1C. The onset of the increase in coercivity started at 560 1C. Compared with earlier results with the same annealing temperature but without the SiO2 coating to prevent sintering [1], we found a lower coercivity. There may be a minimum size for the transformation to the L10 phase. This approach is more promising for the formation of self-assembled FePt arrays for magnetic recording media.

PR

1

D

6

39 41 43 45 47 49 51 53 55