In permanent magnet applications, response often scales with volume or dimension in power-conversion and magnetostrictive applications, even in film form. In microelectromechanical devices it is necessary to explore versatile methods of dense film deposition with film thicknesses approaching one micron. In this study, we present a wet chemical route to hard magnetic cobalt ferrite (CoFe2O4) films to produce films with large coercivity, controllable thickness, saturation approaching that of the bulk, and smoother morphology than state-of-the art sputtered or pulsed-laser-deposited films. The development of etching and releasing processes demonstrates how these films are suitable for precise engineering in a variety of form factors and applications.

Cobalt ferrite (CoFe2O4) is a ubiquitous hard magnet that possesses outstanding magnetoelastic, magnetooptical, and photomagnetic, properties.1–3 Most notably, its large saturated moment and coercivity drive its use in biomedical devices, high-density magnetic recording media, and noncontact force and torque sensors.4–10 The coercivity (as high as 4 kOe for nanocrystals, compared to ∼250 Oe for Fe3O4 at room temperature)11–16 of cobalt ferrite is often a crucial property in materials selection, regardless of film or particulate form of the material.12–16 The large magnetocrystalline anisotropy of CoFe2O4 makes it a candidate for fabricating devices require strong directional magnetostrictive behavior. To that end, coupling it with piezoelectric materials has been a staple of microelectromechanical devices.17,18 When such a device must achieve a high-quality factor through resonance, the acoustic resonance frequency depends on the precise geometry of the film and the thickness of the film must is typically hundreds of nanometers or more, and should be rigorously controlled.19 

The most widely applied technologies of growing cobalt ferrite thin films include ion-beam sputtering,9 pulsed-laser deposition (PLD),20 and chemical solution deposition (CSD) including sol-gel method.6,21–24 Studies applying sputtering and PLD can usually produce fully-dense films with roughness on the order of 3 nm,20,25 but low thermal conductivity of the target limits the typical sputtering rates to 0.7Å/s,9 while the cost and maximum area of uniform thickness from PLD can be prohibitive. Existing CSD methods have been developed to avoid these problems, but no films have been shown to exceed thickness of 200 nm,21 have sufficient density to survive etching steps without pinholes, or have rms roughness less than 3 nm.21 Existing reports of sol-gel-derived CoFe2O4 suffer from ubiquitous pinholes, and correspondingly show roughness consistent with the grain size.22–24 These limitations, combined with the brittleness of ferrites, have hindered efforts to pattern microscale devices on high-quality CoFe2O4 thin films.

To satisfy the needs of various applications of cobalt ferrite in nanoscale and microelectromechanical devices, we present a wet chemical route of growing high-quality ferrite films, with smooth films that surpass the thicknesses achievable by sputtering and PLD, a dense microstructure, magnetic saturation near the bulk value, and coercivity over 4 kOe.

Cobalt ferrite films were fabricated by a chemical solution deposition and spin-coating process. Fe(NO3)3·9H2O (99%, Acros Organics) and Co(NO3)2·6H2O (99%, Aldrich) were added in a 2:1 ratio in 2-methoxyethanol (99+%, Acros Organics) to obtain a 0.05M solution. Over the progress of development in this study, we found that 2-methoxyethanol outperformed other solvents (deionized water, methanol, and ethanol) in wettability (and therefore film thickness uniformity). After mixing, an orange-colored solution was obtained and aged at room temperature for 24 hours. The aged solution retained its stable state for at least one month at room temperature without degrading film quality.

Silicon (100) substrates were cleaned by rinsing with acetone and isopropyl alcohol, then coated with nitrate solution at a spinning speed of 3000 rpm for 30s, and then the coated wafer was directly transferred onto a hot plate preheated 200 °C for one minute. Subsequent edge-bead removal with a fiber-free wipe at this step was necessary to prevent stress concentration at the edge of the substrate. The solidified films were then annealed in air in a preheated furnace at 600 °C for 10 minutes to remove organics and form the desired cubic spinel phase CoFe2O4. Layer stacking was achieved by repeating the coating and annealing process stated above to obtain films with increased thickness.

Etching and patterning steps were performed on a 300 nm ferrite film with SPR220 photoresist, and are summarized in Figure S1 in the supplementary material. The ferrite was wet-etched 85 wt% phosphoric acid (H3PO4, J. K. Baker) at 80°C (approximate rate 40 nm/min), and partial release of the patterned film was achieved by etching the underlying Si with XeF2.

The film morphology was observed with atomic force microscopy (AFM, Asylum Research MFP-3D AFM) in tapping mode and scanning electron microscopy (SEM, Hitatchi S-4700). The crystal structure of the film was characterized with a Bruker D8 Mo-Kα powder X-ray diffractometer (XRD) and analyzed with Rietveld refinement via the TOPAS program. The magnetic behavior of the film was characterized with a Quantum Design MPMS3 vibrating sample magnetometer.

The XRD pattern (with a fixed incident glancing angle of 2°) of a cobalt ferrite film with thickness of 480 nm (8 layers) deposited onto (100) Si is shown in Figure 1(a) in the supplementary material, with all peaks fit to the known cubic spinel structure. This pattern indicates the same structure of the XRD pattern in Figure 1(b) in the supplementary material of a powder sample prepared from the same solution precursor and the same annealing route using Rietveld refinement. Along with confirming phase purity, the film XRD pattern indicates that no significant texturing is evident in the film.

FIG. 1.

XRD patterns obtained with Mo-source diffractometer on powder sample and 480 nm thick film sample that originated from the same solution precursor and underwent the same annealing process. Both data showed good match with theoretical peak position calculated via Rietveld refinement.

FIG. 1.

XRD patterns obtained with Mo-source diffractometer on powder sample and 480 nm thick film sample that originated from the same solution precursor and underwent the same annealing process. Both data showed good match with theoretical peak position calculated via Rietveld refinement.

Close modal

A top-view SEM of the sample with a single deposition layer, as shown in Fig. 2(a and b), has isolated grains with size of 30 nm randomly dispersed on a smooth film of cobalt ferrite with grain size of ∼20 nm, all of which are strongly adhered to the substrate. The anomalous grains may arise from heterogeneity in the oxidation of the underlying Si wafers, or due to a slight shrinking of the soft-baked film upon thermal contraction. In any case, the following depositions of layers 2-7 show that the cobalt ferrite covers the Si substrate with a monodisperse grain size, so uniform nucleation and growth is aided by the first coating acting as a seed layer. The top-view SEM images of a sample with 4 layers is shown in Fig. 2(c and d).

FIG. 2.

SEM top view of a single-layer CoFe2O4 deposition (a, b), with scattered larger particles evident. These particles disappear in subsequent layers, as shown in a sample with four layers (c, d) with a uniform surface and densely packed grains.

FIG. 2.

SEM top view of a single-layer CoFe2O4 deposition (a, b), with scattered larger particles evident. These particles disappear in subsequent layers, as shown in a sample with four layers (c, d) with a uniform surface and densely packed grains.

Close modal

Micrographs of fractured cross sections of films with 3, 5, and 8 layers are shown in Fig. 3, and again the uniform thicknesses of the films are evident. The SEM thickness measurements of the fractured cross-sections of the eight samples with increasing layer depositions are plotted in Fig. 4. Error bars in Fig. 4 refer to the imaging resolution of the SEM measurement. In fact, the thickness is significantly more uniform than these error bars, which is confirmed subsequently by AFM. The data reveal a linear relationship between the number of layers deposited and the film thickness. The cross-sectional view also shows that the repeated deposition does not result in an undesirable layered structure, which complicates subsequent fabrication steps and has been observed in prior sol-gel-derived films.22 The grains size shown in cross-sectional SEM proves a uniformity from the bottom to the top layer, and supports the homogeneity of physical properties across the samples. The thermal expansion coefficients of silicon substrate and CoFe2O4 are known to be 3.5 × 10-6 K-1 and 1 × 10-5 K-1, respectively.20 Our experience is that thermal mismatch led to film delamination in films of more than eight layers (480 nm). Adding a buffer layer or using an expansion-matched substrate would lead to thicker available films.

FIG. 3.

SEM cross-section images of samples with three layers (a-b, about 200 nm), five layers, (c, about 350 nm) and eight layers (d, about 500 nm).

FIG. 3.

SEM cross-section images of samples with three layers (a-b, about 200 nm), five layers, (c, about 350 nm) and eight layers (d, about 500 nm).

Close modal
FIG. 4.

Film thickness measured from cross-sectional SEM shows a linear increase with the number of successive depositions. Error bars are derived from the SEM resolution while determining the edge position, and do not represent the roughness of the film.

FIG. 4.

Film thickness measured from cross-sectional SEM shows a linear increase with the number of successive depositions. Error bars are derived from the SEM resolution while determining the edge position, and do not represent the roughness of the film.

Close modal

AFM images shown in Fig. 5(a) and (b) show the surface topology of samples with one and five layers (with thicknesses of 50 nm and 250 nm, respectively), and the latter is representative of all samples from 2-7 (Figure S2 in the supplementary material). For sample 1, the AFM-determined RMS roughness including the larger grains sticking out the film was 2.1 nm, while the roughness excluding those grains is 1.5 nm. For samples 2-7, the AFM-determined RMS roughnesses are between 1.1-1.4 nm. The calculated roughness for all flat deposition is consistent, and along with the dense grain packing shown via SEM, the film quality is very comparable with and even better than published roughness of cobalt ferrite thin films from previous studies with sputtering and PLD techniques.9,16,20 To date, the RMS surface roughness achieved via sol-gel method and PLD are both reported to be about 3 nm,20,21 smoother films have been reported with PLD at pressures below 2 Pa, but at the cost of crystallinity and deposition rate. The orientation preference of PLD-deposited films is also strongly affected by the substrate structure and orientation.25 

FIG. 5.

AFM images showing the morphology and topology of a sample with a single deposition layer (a), where the scattered outlier particles are masked to avoid skewing the z-range, due to their height of about 22 nm. A sample with five layers (b) displays the typical roughness profile of all samples 2 – 7 layers of coating, showing a uniform and flat surface with no outlier particles. Additional AFM is shown in Figure S3 in the supplementary material.

FIG. 5.

AFM images showing the morphology and topology of a sample with a single deposition layer (a), where the scattered outlier particles are masked to avoid skewing the z-range, due to their height of about 22 nm. A sample with five layers (b) displays the typical roughness profile of all samples 2 – 7 layers of coating, showing a uniform and flat surface with no outlier particles. Additional AFM is shown in Figure S3 in the supplementary material.

Close modal

The magnetic hysteresis loops of eight samples in both in-plane and out-of-plane orientation are shown in Fig. 6(a and b). Bulk cobalt ferrite sample has saturation magnetization of 80.8 emu/g,6,26 and as shown in Fig. 6(b), the out-of-plane Ms for all samples are between 70 – 80 emu/g, quite close to the bulk value even in thin film form. Some uncertainty in this value arises because we normalize the film moments by their measured areas and thicknesses, and assume 100% dense films (5.2 g/cm3). The fact that the calculated Ms are comparable to the bulk value indicates that the actual density of the deposited film is very close to the theoretical density, which further confirms its quality. The out-of-plane Ms is always greater than that of the in-plane measurement, which is most likely due to interfacial strain at CoFe2O4/Si interface, and the internal stress accumulated through thermal cycling during the deposition process. This magnetization mismatch is also observed in other similar magnetostrictive spinel ferrimagnetic systems, including manganese ferrite and zinc ferrite, since the larger out-of-plane compressive strain tends to align spins normal to the substrate, which makes the z-direction the easy axis.27–29 

FIG. 6.

Hysteresis loops extracted from M vs. H measurements of all eight samples in the in-plane (a) and out-of-plane (b) directions. The Ms of bulk sample of cobalt ferrite of 80.8 emu/g is marked as a red dashed line.

FIG. 6.

Hysteresis loops extracted from M vs. H measurements of all eight samples in the in-plane (a) and out-of-plane (b) directions. The Ms of bulk sample of cobalt ferrite of 80.8 emu/g is marked as a red dashed line.

Close modal

The coercivity, Hc and the saturation magnetization, Ms, of samples measured in both in-plane and out-of-plane directions are shown as functions of thickness in Fig. 7(a and b). Since Ms depends on the mass of measured material measurement, we generate error bars from the uncertainty in our thickness measurement. Meanwhile, Hc is independent from the mass and is exact. In our work, the 500 nm-thick sample (eight layers) exhibits out-of-plane coercivity of 7.16 kOe, which is the largest coercivity reported in CoFe2O4 thin films.13,20–25 Our values of Hc and Ms are inversely correlated: samples with smaller Hc have a greater Ms, and vice versa. This behavior is mediated by the grain structure. Hence, the grain size of each sample was measured and calculated by counting the number of grains appearing in 1 um2 AFM scans, and the average grain size with standard deviation, which was determined via grain counting on obtained high-resolution AFM images, of each sample is plotted and shown in Fig. 7(c). According to the comparison of the plots of Hc, Ms, and the grain size, Dg, it was summarized that the Ms is positively correlated to the grain size, as Hc is negatively correlated to it. The saturation magnetization of CoFe2O4 decreases with smaller grain sizes due to disordered spins at grain boundaries.30 The opposite correlation between grain size and coercivity is due to the multi-domain nature of the grains, which slightly exceed the typical critical domain size, Dc, of 35 nm for CoFe2O4.31 Grain boundaries serve as domain-wall-pinning defects, and increasing their concentration increases Hc. Despite a small variation in grain size between the samples, the trend is strongly mirror in the magnetization behavior. Variability in our process likely comes from subtle differences in the heating ramp when each layer is placed in a preheated furnace. Standardizing this heat treatment for all samples would likely provide additional uniformity and control.

FIG. 7.

Saturation magnetization (a) and coercivity (b) extracted from hysteresis loops obtained from M vs. H measurements are plotted along with the grain size measured (c) of samples with various number of depositions. The critical single domain size, Dc, of 35 nm is marked as dashed line in (c). The error bar of grain size is from size measurement via AFM imaging. The error bar of Ms is from the thickness determination.

FIG. 7.

Saturation magnetization (a) and coercivity (b) extracted from hysteresis loops obtained from M vs. H measurements are plotted along with the grain size measured (c) of samples with various number of depositions. The critical single domain size, Dc, of 35 nm is marked as dashed line in (c). The error bar of grain size is from size measurement via AFM imaging. The error bar of Ms is from the thickness determination.

Close modal

Micrographs of etched ferrite films are shown Figure S3(a) in the supplementary material, with lines that are well-patterned on the micron scale and rougher on the nanometer scale of the polycrystalline CoFe2O4 grains. This roughness is typical of the wet etching process but could be improved by optimizing the etchant concentration and temperature. Most importantly, we verify the conformality and density of our ferrite film to demonstrate its usefulness in a variety of form factors by a releasing test, where the exposed Si is etched by XeF2, thereby undercutting the ferrite. Fig. 8(a) shows micrographs of the overall pattern of the etched and partially suspended film (10 μm toward inside from the sidewalls were released). The fiber-like feature is Si residual from releasing with Si etchant. The suspended region appears lighter in secondary electron imaging due to its more insulating character, and it is clear that no pinholes are formed in the wide view in Fig. 8(a and b), and the interior of the ferrite in Fig. 8(d). Figures 8(b and c) show the edge of the suspended film. The sponge-like region below is the etched Si wafer, and cracking is only evident at high-angle curves in the suspended wedge-shaped film.

FIG. 8.

SEM images of a CoFe2O4 film with thickness of about 300 nm that underwent wet-etching and XeF2 releasing process showing (a) top view of the overall pattern, (b) close view of the suspended CoFe2O4 (light) over they holey Si below. A closer view of the wedge-shaped edge of the suspended ferrite is shown in (c), and the morphology of the unsuspended area retaining dense grain packing and clear grain boundaries is shown in (d).

FIG. 8.

SEM images of a CoFe2O4 film with thickness of about 300 nm that underwent wet-etching and XeF2 releasing process showing (a) top view of the overall pattern, (b) close view of the suspended CoFe2O4 (light) over they holey Si below. A closer view of the wedge-shaped edge of the suspended ferrite is shown in (c), and the morphology of the unsuspended area retaining dense grain packing and clear grain boundaries is shown in (d).

Close modal

In conclusion, the need for a versatile method to deposit insulating hard magnetic films with significant thickness (hundreds of nm) is demonstrated for CoFe2O4, with coercivity values typical of nanocrystals, but saturation moments representative of the bulk material. A dense thin film with low porosity and high uniformity was formed by a wet chemical method and exhibits the largest reported out-of-plane coercivity of 7.16 kOe and saturation magnetization very close to the bulk value. Only small variations in these parameters occur as a result of the changing grain structure during processing. Regardless of grain size, and despite the multi-step deposition process, the films are extremely smooth, with rms roughness about 1.5 nm. Since the continuity and conformality of the ferrite film are clear, optimizing wet etching conditions for the desired resist/substrate combinations is the only necessary step to utilize these films in a variety of micron-scale devices.

The detailed microfabrication process and supporting experimental results are included in the supplementary material (Figure S1 – S3).

This work was supported by the 2016 SPAR program of the Defense Advanced Research Projects Agency (DARPA). Cleanroom work was performed in the Micro and Nanotechnology Laboratory Facilities, University of Illinois at Urbana Champaign. Characterization was performed in the Materials Research Laboratory Central Research Facilities, University of Illinois at Urbana Champaign.

1.
V.
Pallai
and
D. O.
Shah
,
J. Magn. Magn. Mater
163
,
243
(
1996
).
2.
R.
Skomski
,
J. Phys.: Condens. Matter
15
(
2003
).
3.
M.
Sugimoto
,
Journal of the American Ceramic Society
82
,
269
(
1999
).
4.
S.
Amiri
and
H.
Shokrollahi
,
Materials Science and Engineering: C
33
,
1
(
2013
).
5.
M.
Colombo
,
S.
Carregal-Romero
,
M. F.
Casula
,
L.
Gutiérrez
,
M. P.
Morales
,
I. B.
Böhm
,
J. T.
Heverhagen
,
D.
Prosperi
, and
W. J.
Parak
,
Chemical Society Reviews
41
,
4306
(
2012
).
6.
K.
Maaz
,
A.
Mumtaz
,
S. K.
Hasanain
, and
A.
Ceylan
,
Journal of Magnetism and Magnetic Materials
308
,
289
(
2007
).
7.
J. A.
Paulsen
,
A. P.
Ring
,
C. C. H.
Lo
,
J. E.
Snyder
, and
D. C.
Jiles
,
Journal of Applied Physics
97
,
044502
(
2005
).
8.
A. V.
Ramos
,
M. J.
Guittet
,
J. B.
Moussy
,
R.
Mattana
,
C.
Deranlot
,
F.
Petroff
, and
C.
Gatel
,
Applied Physics Letters
91
,
122107
(
2007
).
9.
S.
Schnittger
,
C.
Jooss
, and
S.
Sievers
,
Journal of Physics: Conference Series
200
,
072086
(
2010
).
10.
Q.
Song
and
Z. J.
Zhang
,
Journal of the American Chemical Society
126
,
6164
(
2004
).
11.
S.
Ammar
,
A.
Helfen
,
N.
Jouini
,
F.
Fiévet
,
I.
Rosenman
,
F.
Villain
,
P.
Molinié
, and
M.
Danot
,
Journal of Materials Chemistry
11
,
186
(
2001
).
12.
Y.
Cedeño-Mattei
and
O.
Perales-Pérez
,
Microelectronics Journal
40
,
673
(
2009
).
13.
M. V.
Limaye
,
S. B.
Singh
,
S. K.
Date
,
D.
Kothari
,
V. R.
Reddy
,
A.
Gupta
,
V.
Sathe
,
R. J.
Choudhary
, and
S. K.
Kulkarni
,
The Journal of Physical Chemistry B
113
,
9070
(
2009
).
14.
T.
Meron
,
Y.
Rosenberg
,
Y.
Lereah
, and
G.
Markovich
,
Journal of Magnetism and Magnetic Materials
292
,
11
(
2005
).
15.
A. S.
Ponce
,
E. F.
Chagas
,
R. J.
Prado
,
C. H. M.
Fernandes
,
A. J.
Terezo
, and
E.
Baggio-Saitovitch
,
Journal of Magnetism and Magnetic Materials
344
,
182
(
2013
).
16.
B. G.
Toksha
,
S. E.
Shirsath
,
S. M.
Patange
, and
K. M.
Jadhav
,
Solid State Communications
147
,
479
(
2008
).
17.
M. I.
Bichurin
,
D. A.
Filippov
,
V. M.
Petrov
,
V. M.
Laletsin
,
N.
Paddubnaya
, and
G.
Srinivasan
,
Physical Review B
68
,
132408
(
2003
).
18.
M. I.
Bichurin
,
V. M.
Petrov
, and
G.
Srinivasan
,
Physical Review B
68
,
054402
(
2003
).
19.
C.
Wu
,
Y.
Huang
,
Y.
Cui
,
Z.
Shen
,
Y.
Ma
,
C.
Chen
,
S.
Xie
,
H.
Du
,
Y.
Li
, and
S.
Li
, in “
Film thickness induced magnetic anisotropy and ferromagnetic resonance in Fe56Co24B20 amorphous films prepared by pulse laser deposition
” (
2015
), p.
1
.
20.
A.
Raghunathan
,
I. C.
Nlebedim
,
D. C.
Jiles
, and
J. E.
Snyder
,
Journal of Applied Physics
107
,
09A516
(
2010
).
21.
J. Y. P.
Jae-Gwang Lee
,
Y.-J.
Oh
, and
C. S.
Kim
,
Journal of Applied Physics
84
(
1998
).
22.
M.
Millaty
,
M. Y.
Lee
,
D. H.
Kim
,
B. W.
Lee
, and
C.
Liu
,
Journal of Magnetics
19
,
227
(
2014
).
23.
T. E.
Quickel
,
V. H.
Le
,
T.
Brezesinski
, and
S. H.
Tolbert
,
Nano Letters
10
,
2982
(
2010
).
24.
X.
Tang
,
L.
Jin
,
F.
Chen
,
R.
Wei
,
J.
Yang
,
J.
Dai
,
W.
Song
,
X.
Zhu
, and
Y.
Sun
,
Applied Physics Letters
109
,
152406
(
2016
).
25.
R. S.
Hassan
,
N.
Viart
,
C.
Ulhaq-Bouillet
,
J. L.
Loison
,
G.
Versini
,
J. P.
Vola
,
O.
Crégut
,
G.
Pourroy
,
D.
Muller
, and
D.
Chateigner
,
Thin Solid Films
515
,
2943
(
2007
).
26.
I.
Sharifi
,
H.
Shokrollahi
, and
S.
Amiri
,
Journal of Magnetism and Magnetic Materials
324
,
903
(
2012
).
27.
R.
Zhang
,
M.
Liu
,
L.
Lu
,
S.-B.
Mi
, and
H.
Wang
,
Journal of Materials Chemistry C
3
,
5598
(
2015
).
28.
J. M.
Iwata
,
R. V.
Chopdekar
,
F. J.
Wong
,
B. B.
Nelson-Cheeseman
,
E.
Arenholz
, and
Y.
Suzuki
,
Journal of Applied Physics
105
,
07A905
(
2009
).
29.
Y. J.
Yang
,
M. M.
Yang
,
Z. L.
Luo
,
C. S.
Hu
,
J.
Bao
,
H. L.
Huang
,
S.
Zhang
,
J. W.
Wang
,
P. S.
Li
,
Y.
Liu
,
Y. G.
Zhao
,
X. C.
Chen
,
G. Q.
Pan
,
T.
Jiang
,
Y. K.
Liu
,
X. G.
Li
, and
C.
Gao
,
Journal of Applied Physics
115
,
173505
(
2014
).
30.
B.
Issa
,
I. M.
Obaidat
,
B. A.
Albiss
, and
Y.
Haik
,
International Journal of Molecular Sciences
14
,
21266
(
2013
).
31.
A.
López-Ortega
,
E.
Lottini
,
C. d. J.
Fernández
, and
C.
Sangregorio
,
Chemistry of Materials
27
,
4048
(
2015
).

Supplementary Material