ARTICLE IN PRESS
Journal of Magnetism and Magnetic Materials 316 (2007) e879–e882 www.elsevier.com/locate/jmmm
Soft magnetic FeCoTaN film cores for new high-frequency CMOS compatible micro-inductors K. Seemann, H. Leiste, C. Ziebert Forschungszentrum Karlsruhe in der Helmholtz-Gemeinschaft, Institut fu¨r Materialforschung I, Hermann-von-Helmholtz-Platz 1, 76344 Eggenstein-Leopoldshafen, Germany Available online 19 March 2007
Abstract New high-frequency micro-inductors with thin magnetic film cores where developed by R.F.-magnetron sputtering and plasma beam as well as reactive ion etching. In order to realise soft magnetic films with magnetic resonance frequencies in the GHz range determined by frequencydependent permeability measurements, 6-inch Fe47Co36Ta17 and Fe37Co46Ta17 targets were used to deposit FeCoTaN-films by reactive R.F.magnetron sputtering in an Ar/N2 atmosphere. To obtain soft magnetic film properties with a marked uniaxial in-plane anisotropy needed for the high-frequency suitability, the films were annealed in a static magnetic field at CMOS temperatures of around 400 C. Due to the specific material composition the films possess a nanocrystalline microstructure with a low magnetocrystalline anisotropy. The film material was employed to realise different magnetic cores for new micro-inductor designs fabricated by the CMOS aluminium process. r 2007 Elsevier B.V. All rights reserved. PACS: 84.40.Az; 76.50.þg; 75.40.Gb; 75.60.Ch; 84.32.y Keywords: Nanocrystalline ferromagnetic thin film; Uniaxial anisotropy; High-frequency permeability; Magnetic domain; Micro-inductor
1. Introduction Many efforts have been undertaken to develop different highly permeable magnetic films to perform their frequency suitability and possible micro-electronic applications in passive devices [1–3]. Soft ferromagnetic films are barely used in micro-electronic high-frequency devices yet, but are becoming more and more important in the field of semiconductor components realised by CMOS processes. An essential requirement for the use in CMOS semiconductor components like, e.g., micro-inductors, is the thermal stability of the films during the production process between 400 and 500 C, i.e., the production process must not deteriorate the soft magnetic properties through temperature induced crystallisation. This can be met by refractory metal–nitride grain boundaries (e.g., TaN) acting as a grain refiner [4]. Not only the composition and constitution are responsible for the film performance, Corresponding author. Tel.: +49 7247 82 4255; fax: +49 7247 82 4567.
E-mail address:
[email protected] (K. Seemann). 0304-8853/$ - see front matter r 2007 Elsevier B.V. All rights reserved. doi:10.1016/j.jmmm.2007.03.126
however, these conditions for low frequency losses must be optimised in a first step. Besides high saturation polarisation, low coercitivity, appropriate in-plane uniaxial anisotropy and high resistivity [5], the film dimensions additionally impact the response to high-frequency fields they are exposed to. Micro-electronic devices demand magnetic films with small lateral dimensions which produces different magnetic properties in comparison to those laterally extended, caused by demagnetisation effects influencing the domain structure and hereby the frequency behaviour. In the present paper, we have realised magnetic films which were patterned to cores for micro-inductors. In this relation, an already introduced and entirely new micro-inductor design [6] as an application for those films will be discussed. 2. Experimental procedure Reactive R.F. magnetron sputtering was used to deposit the magnetic films on oxidised 5 5 mm 380 mm (1 0 0)silicon substrates. Two 6-in targets were employed with
ARTICLE IN PRESS K. Seemann et al. / Journal of Magnetism and Magnetic Materials 316 (2007) e879–e882
e880
different compositions, Fe47Co36Ta17 and Fe37Co46Ta17. To produce Fe–Co–Ta–N films the R.F. power of 250 W, the pressure of 0.5 Pa and the total gas flow of 50 sccm Ar/N2 were kept constant. The relative nitrogen flow was optimised to 2 sccm for both targets. This results in films with the following composition, Fe38Co30Ta8N24 and Fe33Co40Ta10N17, respectively. After the deposition process the films were annealed for 1 h at 400 C in a static magnetic field of about 50 mT to induce an in-plane uniaxial anisotropy. Magnitudes like the saturation polarisation J s , the coercitive field H c and the uniaxial anisotropy H u were measured with a vibrating sample magnetometer (VSM). Patterning of films to core structures and the micro-inductor components [6] were realised by ordinary ‘‘near ultra violet’’ lithography and physical or chemical dry etching on oxidised 4-in Si-wafers. The investigation of the domain structure was made by magnetic force microscopy (MFM). The frequency-dependent permeability of the films was measured with a broadband permeameter connected to an Agilent 8753ES vector network analyser up to 4.5 GHz where the micro-inductor was measured by means of a wafer prober set-up and an Agilent 8719D network analyser up to 4 GHz. 3. Results and discussion 3.1. Film materials development In order to optimise the frequency behaviour, i.e., to obtain a sharp resonance frequency permeability spectrum, the films must possess a uniaxial anisotropy as well as a material composition which guarantees magnetic homogeneity as far as possible. As an example, the uniaxial anisotropy field amounts to m0 H u ¼ 4 mT, determined by the easy and hard axis of polarisation of a representative Fe33Co40Ta10N17 film. The saturation polarisation J s is 1.3 T, where the coercitive field 1400 Re µr Fe38Co30Ta8N24 Im µr Re µr Fe33Co40Ta10N17 Im µr Modified Landau-Lifschitz-Theory
1200 1000 Permeability µr
800 600 400 200 0 -200 -400 -600 -800 0.1
1 Frequency f (GHz)
Fig. 1. Real and imaginary parts of the frequency-dependent permeability of about 380 nm thick Fe–Co–Ta–N films annealed in a magnetic field of 50 mT for 1 h at 400 C. The solid lines show the modified Landau–Lifschitz in combination with the Maxwell theory.
was around 0.1 mT. In Fig. 1, the frequency behaviours for films fabricated by means of the two targets mentioned above are shown. Both films exhibit a resonance frequency of around 2 GHz but have different permeability shapes. The film with the composition Fe38Co30Ta8N24 has a second relative but clear maximum at approximately 1 GHz due to magnetic or anisotropy in-homogeneities. This leads to an enhanced broadening of the permeability spectrum (increased full with half maximum 2pDf ¼ 2=t of the imaginary part) and therefore to higher losses in comparison to the frequency spectrum of a Co-rich Fe33Co40Ta10N17 film. This means, that the lifetime t of the resonance state decreases due to more damping through energy dissipation in the precessing system of magnetic moments. At this point, it is obvious which film system is predestined and used for the application as, e.g., cores in micro-inductors. 3.2. Micro patterned magnetic core components and their domain structure Patterning of the films to micro-core structures causes a crucial change in the permeability and frequency characteristics. Having finite dimensions, that is, the lateral dimensions are closer to the film thickness, demagnetising effects are decisively responsible for the resonance or cut-off frequency and domain structure. Due to the increasing demagnetisation field favourable and unfavourable magnetisation in domains with respect to the highfrequency field arise. This leads to a declining effective permeability number through loosing magnetic volume due to closure domains but to higher cut-off frequencies which are increasingly determined by the in-plane demagnetisation field if lateral dimensions become smaller and smaller (Fig. 2). Here, the uniaxial anisotropy, which also impacts the frequency, is more and more overshadowed by the demagnetising anisotropy. Therefore, an alternative concept of micro-cores was conceived to influence, i.e., to stabilise and increase a convenient domain structure with magnetic moments which are able to efficiently interact with the high-frequency field. To reduce the closure domain fraction, slits were incorporated preferably at the place where domain boundaries are probably located. This can be observed in, e.g., Fig. 3 where a 100 100 mm extract from a 100 mm wide and 500 mm long core structure shows less closure domains. In Fig. 3, a donut-like core structure with radial oriented slits for applications in, e.g., toroid-inductors tend to a similar domain pattern, i.e., an in-plane anisotropy preferably arranges in radial direction after annealing the ‘‘donut’’ even without a static magnetic field. Due to the slits higher demagnetisation effects drive the cut-off frequency to increased values. In Fig. 4, the frequency dependent permeability spectra of a 100 100 mm squared and a 100 mm wide slit core structure are demonstrated.
ARTICLE IN PRESS K. Seemann et al. / Journal of Magnetism and Magnetic Materials 316 (2007) e879–e882
e881
Fig. 2. MFM images of domain structures of 200 nm thick Fe33Co40Ta10N17 squares annealed in a magnetic field. The lateral dimensions are 40 40 mm and 100 100 mm, respectively.
Fig. 3. Extracted MFM images of domain structures of 200 nm thick Fe33Co40Ta10N17 core structures with domain stabilizing slits. The donut-like core was annealed without a magnetic field.
3.3. Micro-inductor design Imaginary part of permeability µr (arb. units)
core structure subdivided in 100 x 100 µm squares 100 µm wide core structure with slits
0.1
1 Frequency f (GHz)
Fig. 4. Imaginary parts of the frequency-dependent permeability of 200 nm thick Fe–Co–Ta–N core structures annealed in a magnetic field for 1 h at 400 C.
The inductor prototype already introduced in Ref. [6] has five subdivided squared cores with lateral dimensions of mainly 100 100 mm and is depicted in Fig. 5. The windings and cores of the micro-inductor are arranged in a way that they produce a magnetic flux nearly parallel to the substrate where eddy-currents in the substrate are lowered which can impact the inductance and quality factor in a quite negative way. By this inductor design, flux leakage is reduced in comparison to planar spiral shaped micro-coils without magnetic cores used in industrially fabricated R.F. circuits. After an ‘‘open deembedding’’ measurement procedure the micro-inductor exhibits an initial inductance of approximately 1.4 nH which is slightly lower for an inductor of the same type without a magnetic core. The initially higher frequency
ARTICLE IN PRESS K. Seemann et al. / Journal of Magnetism and Magnetic Materials 316 (2007) e879–e882
e882
The frequency dependent quality factor can be theoretically described by an equivalent circuit model, and in combination with the R.F. ferromagnetic resonance Landau–Lifschitz–Maxwell-model, if required.
4. Conclusion
Fig. 5. SEM image of a micro-inductor with 200 nm thick subdivided 100 100 mm and 40 40 mm Fe33Co40Ta10N17 cores.
5
Quality factor Q
4 3
Film material investigations should be carried out to obtain a low loss magnetic film frequency behaviour by optimising the composition. By this, the development of magnetic Fe–Co–Ta–N films has shown that they can be realised and incorporated as cores in passive highfrequency devices like micro-inductors. Due to the design of cores their domain structure and their operating frequency can be additionally influenced. It was observed that the quality factor of micro-inductors can be increased by means of cores up to the GHz range. By applying an appropriate equivalent circuit model the, e.g., quality factors can be excellently described. A good correspondence between the measured and calculated data could be attained.
2 micro-inductor with magnetic core equivalent circuit model with magnetic core micro-inductor without magnetic core equivalent circuit model without magnetic core
1
References
0 0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
Frequency f (GHz) Fig. 6. Measured and theoretical quality factor of the micro-inductor with and without magnetic cores.
dependent quality factor Q of the inductor with cores decreases at about 1.5 GHz which can be assigned to the cut-off frequency of the magnetic film core material (Fig. 6).
[1] C.-H. Lee, D.-H. Shin, D.-H. Ahn, S.-E. Nam, H.-J. Kim, J. Appl. Phys. 85 (1999) 4898. [2] M. Yamaguchi, K. Suezawa, Y. Takahashi, K.I. Arai, S. Kikuchi, Y. Shimada, S. Tanabe, K. Ito, J. Magn. Magn. Mater. 215–216 (2000) 807. [3] F. Liu, T. Ren, C. Yang, L. Liu, A.Z. Wang, J. Yu, Mater. Lett. 60 (2006) 1403. [4] B. Ma, F.L. Wei, X.X. Liu, C.T. Xiao, Z. Yang, Mater. Sci. Eng. B 57 (1999) 97. [5] V. Bekker, K. Seemann, H. Leiste, J. Magn. Magn. Mater. 296 (2006) 37. [6] K. Seemann, H. Leiste, V. Bekker, J. Magn. Magn. Mater. 302 (2006) 321.
