APPLIED PHYSICS LETTERS 88, 182504 共2006兲
Quasiferromagnetism in semiconductors
T. Dubroca, J. Hack, and R. E. Hummela兲
Department of Materials Science and Engineering, University of Florida, Rhines Hall room 106,
Gainesville, Florida 32611
A. Angerhofer
Department of Chemistry, University of Florida, P.O. Box 117200, Gainesville, Florida 32611
共Received 20 January 2006; accepted 26 March 2006; published online 4 May 2006兲
Ferromagnetic hysteresis has been observed at room temperature in materials not consisting of
elements commonly associated with ferromagnetism, such as Co, Ni, Fe, or Mn-containing alloys.
In particular, we report on magnetic hysteresis seen in silicon prepared by two different techniques:
ion implantation 共Si and Ar兲 and neutron irradiation. Because the material investigated contains no
ferromagnetic elements, we tentatively call it “quasiferromagnetic.” The paramagnetic defects
present in these materials were investigated using electron paramagnetic resonance. We suggest that
these defects are one of the factors responsible for the observed macroscopic magnetic hysteresis
loop. © 2006 American Institute of Physics. 关DOI: 10.1063/1.2198483兴
Spin-transport electronics 共“spintronics”兲 is an evolving
device technology that functions using electron spins, either
alone or in conjunction with electron charge. Our research
focuses on understanding how certain semiconductors 共i.e.,
silicon, implanted with either silicon or argon ions or neutron
irradiated兲 causes a ferromagneticlike hysteresis loop.
Other investigators have speculated about the possibility
of ferromagnetism in argon and neon implanted silicon or
vapor-deposited amorphous Si but did not present a magnetic
hysteresis loop.1,2 Their conclusions were based on electron
paramagnetic resonance 共EPR兲 experiments yielding a g factor of 2.0055 which was interpreted as stemming from paramagnetic centers having a concentration between 1020 / cm3
and 1021 / cm3. Other investigators3–5 implanted manganese
up to 23% into Si and observed a ferromagnetic hysteresis
loop, which was interpreted to originate from the Mn d-shell
electrons. Finally, our group and Laiho et al. observed ferromagnetic hysteresis in spark-processed Si 共Ref. 6兲 or anodically etched porous Si,7 respectively.
In our current work, p-type silicon wafers having a resistivity of 3 – 5 ⍀ cm were implanted by silicon or argon
ions using an energy of 50 keV, while the current density
was kept at less than 10 ␮A / cm2. The sample stage was
water cooled throughout the implantation to diminish recrystallization. Another Si sample was neutron irradiated in a
thermal neutron reactor with a dose of 4 ⫻ 1016 cm−2.
Measurements of magnetization as a function of the applied magnetic field were conducted at various temperatures
using a commercial superconducting quantum interference
device 共SQUID兲 magnetometer.
Figure 1 shows the magnetization response of Si implanted into silicon at doses of 1 ⫻ 1016 cm−2 共1A兲 and 5
⫻ 1016 cm−2 共1B兲, respectively, measured at room temperature, while an insert shows the magnetization response of
sample 1B at 10 K. An unprocessed silicon wafer was also
measured 共not shown兲, and its known diamagnetic response
was confirmed. These samples display magnetic hysteresis
loops having a coercive field of 100 Oe, saturation fields of
600 Oe 共1A兲 and 700 Oe 共1B兲, and remanences of
a兲
Electronic mail: [email protected]fl.edu
0.10 emu/ cm−3 共1A兲 and 0.11 emu/ cm−3 共1B兲, respectively.
The comparison suggests that the magnetic response 共defined
as the area within the hysteresis loop兲 may be increased by
increasing the implantation dose. The insert of Fig. 1 shows
that essentially the same magnetic behavior is observed
when sample 1B is measured at 10 K.
Figure 2 depicts the magnetization response of samples
2A and 2B, which were implanted with argon at doses of 2
⫻ 1016 and 2 ⫻ 1017 cm−2, respectively, measured at room
temperature. Sample 2A displays a magnetic hysteresis loop
with a coercive field of 100 Oe, saturation field of 600 Oe,
and a remanence of 0.3 emu/ cm3. Sample 2B displays a
magnetic hysteresis loop with a coercive field of 100 Oe, a
saturation field of 600 Oe, and a remanence of 0.4 emu/ cm3.
Similar to Si into Si, a higher implantation concentration of
Ar yields a larger magnetic response. Moreover, it is noted
that implantation of argon into silicon causes a larger magnetic response than for Si into Si at a similar dose.
Figure 3 depicts the magnetization response at 10 K of
silicon irradiated by neutrons at a dose of 4 ⫻ 1016 cm−2. The
magnetization response of the neutron-irradiated sample was
FIG. 1. Magnetization vs magnetic field for sample 1A 共Si implanted into Si
at 1016 cm−2 dose, solid line兲 and sample 1B 共Si implanted into Si at 5
⫻ 1016 cm−2 dose with an energy of 50 keV, dashed line兲 measured and
implanted at room temperature. Insert: magnetization vs magnetic field for
sample 1B at 10 K.
0003-6951/2006/88共18兲/182504/3/$23.00
88, 182504-1
© 2006 American Institute of Physics
Downloaded 25 Feb 2008 to 146.139.50.23. Redistribution subject to AIP license or copyright; see http://apl.aip.org/apl/copyright.jsp
182504-2
Dubroca et al.
FIG. 2. Magnetization vs magnetic field for sample 2A 共Ar implanted into
Si at 2 ⫻ 1016 cm−2 dose, solid line兲 and sample 2B 共Ar implanted into Si at
2 ⫻ 1017 cm−2 dose with an energy of 50 keV, dashed line兲 measured and
implanted at room temperature.
significantly weaker than that for either Si into Si or Ar into
Si.
Secondary ion mass spectroscopy 共SIMS兲 reveals the
presence of silicon, oxygen, boron, and hydrocarbons 共results not shown for brevity兲. No other elements were observed within the detection limit. In particular, no iron 共limit
at 0.02 ppm兲 was detected. This excludes the possibility that
Fe is a contributing factor for the observed hysteresis loop. It
is suggested instead that the large quantities of paramagnetic
impurities should interact and produce hysteresis if not coherent ferromagnetism.
Figure 4 shows an EPR spectrum of sample 1A measured at room temperature. Also included is a curve showing
the sum of two model-generated Lorentzian lines, which displays the spectrum of an unprocessed silicon wafer, which
shows no evidence of unpaired spins. The modeled line
closely matches our data, with a correlation coefficient of
0.97. It is known that each Lorentzian line corresponds to the
resonance of a single type of dangling bond. In contrast, a
non-Lorentzian, such as a Gaussian line 共which is the sum of
Lorentzian lines兲, corresponds to a distribution of dissimilar
dangling bonds with slightly different g factors. Therefore,
we conclude that the EPR signal of sample 1A results from
two types of dangling bonds.
The first Lorenzian line has a peak that yields a g factor
of 2.0054, which has been attributed to a silicon dangling
Appl. Phys. Lett. 88, 182504 共2006兲
FIG. 4. EPR spectrum of sample 1A 共Si implanted into Si at 1016 cm−2,
square dots兲, its model using the sum of two Lorentzian lines’ first derivative
共solid line兲 and spectrum of an unprocessed silicon 共round dots scattered
around zero兲.
bond back-boned by three silicon atoms,8 consistent with
Refs. 1, 2, and 7. The second Lorenzian line has a peak that
yields a g factor of 2.0023. According to Tomozeiu et al.,9
this peak can be attributed to E⬘ centers typically present in
SiO2. In our case, a small but measurable amount of oxygen
is present, as determined by the SIMS analysis.
A common feature of our quasiferromagnetic materials
is the presence of a large number of paramagnetic centers
共i.e., unpaired spins兲. In addition, the EPR spectrum for ion
implanted Si is strikingly similar to the EPR spectrum for
spark-processed Si 共sp-Si兲, for which it was postulated that
the ferromagnetic hysteresis loop originates from dangling
bonds.6 It was shown that annealing sp-Si removes, in an
irreversible manner, the ferromagnetic behavior while, at the
same time, dramatically reducing the number of dangling
bonds.6 It is therefore suggested that dangling bonds substantially contribute to the ferromagnetic behavior in the presently investigated quasiferromagnetic materials.
It is further suggested that the unpaired spins of the dangling bonds interact to yield the quasiferromagnetic behavior
in the currently investigated materials. The interaction that
makes the exchange integral between the dangling bond
wave functions positive can be described in at least two different ways. First, the distance between the dangling bonds
may be large enough to avoid the formation of a covalent
bond. At the same time, the distance may be small enough to
permit the direct overlap of the wave functions of the two
electrons. As such, the triplet state may be the most stable
state, therefore inducing a positive exchange interaction at
the electronic level.
The second possibility is an indirect coupling of spins.
For example, Talapatra et al.10 suggested that the ferromagnetic hysteresis loop observed for carbon nanostructures implanted with nitrogen requires not only unpaired spins but
also double bonds. This would permit a delocalization of the
electrons and would create a mediated pathway for an indirect exchange interaction between unpaired electrons. Similarities between Si and C allow us to propose that an equivalent mechanism may be possible for Si. At present, neither a
direct nor an indirect exchange coupling between dangling
bonds is ruled out. An analysis that connects our observations with the density of paramagnetic defects utilizing
pulsed EPR and ENDOR is currently in progress.
1
A. F. Khokhlov and P. V. Pavlov, JETP Lett. 24, 211 共1976兲; Pis’ma Zh.
Eksp. Teor. Fiz. 24, 238 共1976兲.
FIG. 3. Magnetization vs magnetic field for sample 3 共silicon irradiated at a
2
T. R. Askew, H. J. Stapleton, and K. L. Brower, Phys. Rev. B 33, 7
dose of 4 ⫻ 1016 cm−2 with thermal neutrons兲 measured at 10 K.
Downloaded 25 Feb 2008 to 146.139.50.23. Redistribution subject to AIP license or copyright; see http://apl.aip.org/apl/copyright.jsp
182504-3
共1986兲.
F. M. Zhang, Y. Zeng, J. Gao, X. C. Liu, X. S. Wu, and Y. W. Du, J. Magn.
Magn. Mater. 282, 216 共2004兲.
4
L. Liu, N. Chen, S. Song, Z. Yin, F. Yang, C. Chai, S. Yang, and Z. Liu, J.
Cryst. Growth 273, 458 共2005兲.
5
M. Bolduc, C. Awo-Affouda, A. Stollenwerk, M. B. Huang, F. G. Ramos,
G. Agnello, and V. P. Labella, Phys. Rev. B 71, 033302 共2005兲.
6
J. Hack, M. H. Ludwig, W. Geerts, and R. E. Hummel, Mater. Res. Soc.
Symp. Proc. 452, 147 共1997兲.
3
Appl. Phys. Lett. 88, 182504 共2006兲
Dubroca et al.
7
R. Laiho, E. Lahderanta, L. Vlasenko, M. Vlasenko, and M. Afanasiev, J.
Lumin. 57, 197 共1993兲.
8
W. L. Warren, E. H. Pointdexter, M. Offenberg, and W. Muller-Warmuth,
J. Electrochem. Soc. 139, 872 共1992兲.
9
N. Tomozeiu, E. E. van Faassen, W. M. Arnoldbik, A. M. Vredenberg, and
F. H. P. M. Habraken, Thin Solid Films 420–421, 382 共2002兲.
10
S. Talapatra, P. G. Ganesan, T. Kim, R. Vajtai, M. Huang, M. Shima, G.
Ramanath, D. Srivastava, S. C. Deevi, and P. M. Ajayan, Phys. Rev. Lett.
95, 097201 共2005兲.
Downloaded 25 Feb 2008 to 146.139.50.23. Redistribution subject to AIP license or copyright; see http://apl.aip.org/apl/copyright.jsp