Upload
others
View
6
Download
0
Embed Size (px)
Citation preview
25
CHAPTER 2
TRANSISTION METAL DOPED GALLIM NITRIDE -
A DILUTE MAGNETIC SEMICONDUCTOR
2.1 INTRODUCTION
Diluted magnetic semiconductor (DMS) materials, which utilize
both the spin and the charge properties of carriers, have become attractive
because of the interest in investigating the fundamental physical properties of
such materials and in potential applications for many promising spintronic
devices (Ohno 1998, Fiederling et al 1999, Dietl et al 2000). Spintronics
refers commonly to a phenomenon in which the spin of electrons in a solid
state environment plays the determining role. These devices have wide
applications in spin-valve transistor, spin-light emitting diodes, non-volatile
memory, optical isolator, ultrafast optical switches. As GaN doped with
transition metal will retain ferromagnetism at practical operating temperature
as predicted by theoretical calculations and the other advantage is that there
are already an existing process technology for this material in other
applications such as photodedectors and light emitting diodes (LEDs).
There is a wide class of semiconducting materials which is
characterized by the random substitution of a fraction of the original atoms by
magnetic atoms. The materials are commonly known as semimagnetic
semiconductors (SMSC) or diluted magnetic semiconductors (DMS). The
most common DMS are II-VI compounds (like CdTe, ZnSe, CdSe, CdS, etc.),
with transition metal ions (e.g. Mn, Fe, Ni or Co) substituting the original
26
cation site. There are also materials based on IV-VI (e.g. PbTe, SnTe) and
recently III-V compound semiconductors (e.g. GaAs, InSb, GaN) crystals
(Russell et al 2004, Ham et al 2006). These mixed crystals (semiconductor
alloys) may be considered as containing two interacting subsystems. The first
of these is the system of delocalized conduction and valence band electrons.
The second is the random, diluted system of localized magnetic moments
associated with the magnetic atoms. The fact that both the structure and the
electronic properties of the host crystals are well known means that they are
perfect for studying the basic mechanisms of the magnetic interactions
coupling the spins of the band carriers and the localized spins of magnetic
ions. The coupling between the localized moments results in the existence of
different magnetic phases (such as paramagnets, spin glasses and
antiferromagnets).
(A) (B) (C)
Figure 2.1 Schematic diagram of (A) Pure or nonmagnetic semiconductor
(B) Dilute magnetic semiconductor (C) dilute magnetic
semiconductor with ferromagnetic interactions
The exchange interaction between spin of s, p electrons of host with
the d-electron of the doped transition–metal ion is responsible for magnetic,
optical and conductive properties of DMS materials (Furdyna 1998).
Possibility of ferromagnetism in GaMnN above room temperature based on
first principle calculations has been proposed (Dietl et al 2000). Calculation of
27
energetic and magnetic properties of manganese doped GaN with higher
concentration shows that Mn related cluster is energically favorable and is
responsible for the observed magnetism in the system (van Schilfgaarde and
Mryasov 2001). Investigations have shown that ferromagnetic behavior is
dominated by a short range double exchange mechanism in III-V compound
semiconductors when transition metal forms a random alloy (Sato et al 2001).
Ferromagnetic phase in (GaFe)N below 100K has been reported (Akinaga et
al 2000). Prediction of weak ferromagnetism in copper doped GaN and no
effect on magnetization with the creation of gallium vacancy is observed by
Rosa et al (Rosa et al 2007). DFT theory has shown that chromium doping in
GaN results in the formation of chromium cluster which exhibit anti
ferromagnetism (Cui et al 2005). Eu doping in GaN results in a high magnetic
moment of 6 µB (Goumari et al 2008). DFT calculations based on full potential
linearized augmented plane wave method have explored the possibility of
defect-induced magnetism in wurtize GaN (Hong 2008). The nitrogen
vacancy defect structure has no sign of magnetic state. Nonetheless, very
interestingly it has been found that GaN with a gallium vacancy defect can
show induced local magnetic moment in nitrogen atoms. The four nitrogen
atoms in the tetrahedron sites with the gallium vacancy as neighbor have
magnetic moments of 0.23 and 0.29 µB. The contribution of larger magnetic
moment is from the nitrogen which is at a larger distance from the gallium
vacancy.
In the recent past, much work has been carried out on manganese
doped GaN. The theoretical observation shows that Mn-Mn interaction is
responsible for ferromagnetism when Mn is buried inside the bulk but is
antiferromagnetic when the coupling takes place at the surface of GaN matrix
(Wang et al 2004). P-type GaMnN quantum well is ferromagnetic at room
temperature with weak spin-exchange interaction is reported (Kim et al 2005).
Prediction of room temperature ferromagnetic behavior in MOCVD grown
28
iron doped GaN is observed (Kuo et al 2008). We have studied nickel doped
GaN using TB-LMTO method and observed that the system having no defects
is ferromagnetic for lower percentage of dopant (Munawar et al 2010).
Theoretical and experimental result shows that copper doped GaN is
ferromagnetic with high spin polarization at Tc above room temperature
(Ganz et al 2010). Implantation of Co into p-type GaN grown by MOCVD
and reported ferromagnetism at 320 K, in most cases the damage induced due
to ion implantation is removed on samples annealing (Lee et al 2003). Weak
magnetic polarization has been predicted in defects or residual oxygen of
GaN samples (Rosa et al 2007). Investigations have been carried out on the
theoretical aspects and on the synthesis of cobalt doped GaN. Extensive
characterizations on the electrical, optical and magnetic properties have been
carried out and cobalt doped GaN is found to be a suitable candidate in the
field of spintronics.
2.2 TB-LMTO CALCULATION ON COBALT DOPED GaN
The calculations have been performed using TBLMTO method
based on the density functional theory (Anderson 1975). The kinetic energy
outside the muffin-tin spheres (where the potential is constant) is chosen to be
zero. The Wigner- Seitz cells located at the atomic or interstitial sites are
replaced by the atomic sphere and the spherically symmetric potential within
the sphere is extended to their boundaries. In energy calculations, we include
only s, p and d orbital’s and neglected all higher partial waves and integrate
the radial Schrödinger equation out to the boundary of the sphere. The local
density approximation (LDA) parameterized by Barth and Hedin (Barth and
Hedin 1972) was used for the exchange and correlation potential. Coulomb
potential energy caused by electron-ion interaction is described in which
orbital of Ga (3d4s), N (2p3s), Co(3d4s) and Ni(3d4s) were treated as valence
electrons. The accuracy of the total energy calculation within the density
functional theory, using LDA, is sufficient to predict which structure is
29
thermodynamically stable i.e., have lowest free energy. Empty sphere
were introduced in all cases in order to keep the overlap of atomic sphere
within 16 %.
GaN crystallize in the wurtzite structure with lattice constants
a = 3.188 Å and c = 5.189 Å (Lagerstedt. and Monemar 1979). Periodic
2x2x2, 4x4x2, 3x3x2 and 2x2x3 supercells of GaN matrix was used as the
calculation units. Each supercell has equal number of real and empty atoms.
Defects such as Ga and N vacancies are considered in the supercell. By
replacing Ga atom by transition metal namely, nickel and cobalt in these
supercells, we get the corresponding dopant concentration, Necessary care has
been taken not to exceed the doping level as beyond 12.5% for Ni. The
Wigner-Seitz sphere radii are chosen in such a way that the sphere boundary
potential was minimum and the charge flow in accordance with the electro
negativity criteria. Spin polarized scalar relativistic calculations are performed
to obtain the total energy, density of states (DOS) and magnetic moment. 144
k-points in the irreducible part of the Brillouin Zone were used to calculate
the total and partial density of states by means of tetrahedron method (Stokes
et al 1987). The k and E convergence are checked by increasing the number of
K points and the energy convergence criteria. TB-LMTO’s have almost
universal decay and with ASA, The main advantage of this scheme is that it is
computationally effective.
2.3 EXPERIMENT
Synthesis of pure and doped GaN was carried out using a
experimental set-up as shown in Figure 2.2. Pure gallium metal (5N purity)
and transition metal namely cobalt was taken in the alumina boat and is kept
at the middle of the furnace (Senthil kumar and Kumar 2002) which consists
of a 80mm diameter quartz tube and of length 100 cm (with wall thickness of
3mm) .The assembly has an inlet and an outlet of diameter 10mm.
30
Figure 2.2 Experimental setup for the synthesis of pure and doped
Gallium nitride
Nitrogen gas is continuously flowed from the beginning of the
experiment for a clean ambient in the reactor and to remove the native oxide.
The reactor is kept inside a resistively heated furnace. A ceramic muffle of
length 100mm has been used for winding the heating element (Al Kanthal
wire: SWG 18 type).
The windings are made in such a way that the space between the
winding gradually becomes narrower at the ends of the tube compared to the
centre of the tube to compensate the heat loss at the ends of the reactor. The
windings are insulated by using a paste of corundum cement. A thermocouple
of K type is used to measure the reactor temperature. A Eurothrum
temperature controller Model 2604 with accuracy of 0.1 °C is used. Upon
attaining the growth temperature (~ 1223 K) ammonia gas of 3 slm is allowed
to flow for 8 hours and then the system is allowed to cool to room
temperature at an interval of 100 K/h under the flow of nitrogen. The
synthesized sample is etched with HF and HNO3 in 1:1 ratio in order to
remove unreacted metals, the final product is dried and then pulverized to a
fine powder. Structural investigations have been carried out using X-ray
31
diffraction (XRD) studies. The emission properties of the samples are studied
by Photoluminescence (PL) measurement with He-Cd (325nm) laser as the
excitation source at room temperature. Magnetization and temperature
dependent magnetic properties of the samples have been studied using
superconductor quantum interface device (SQUID) measurements.
2.4 RESULTS AND DISCUSSIONS
The schematic representations of the unit cell for wurtzite GaN
with vacancies used in our calculation are presented in Figure 2.3. The figure
represent Ga, N, Ga vacancy, N vacancy and Ni are represented by Violet,
Gray, Yellow, Blue and light Green balls.
(A) (B) (C)
(D) (E) (F)
Figure 2.3 Schematic diagram of (A) GaN Supercell with two Ni doped
to form dimer, (B) two Ni at different site (C) GaN Supercell
with Ga vacancy (D) Ni doped at Ga site with a Ga vacancy
(E) GaN Supercell with N vacancy, (F) Ni doped at Ga site
with N vacancy
32
2.4.1 Electronic and Magnetic Properties of Nickel Doped GaN
The electronic structure of GaN in the wurtzite structure is obtained
by means of TB-LMTO method. It is a direct band gap material as reported in
literature (Strite and Morkoç 1992). One and two atoms of gallium were
replaced by Ni to obtain a dopant concentration of 6.25 and 12.5 % in 2x2x2
supercell respectively. Since the substitution of Ni impurity in the GaN lattice
was done by replacing the Ga atom, the effect of the structural relaxation has
been neglected (Das et al 2003). When the transition metal (Ni) defect is
introduced in GaN, the five fold degenerate d-states split into two degenerate
d-states by the crystal field splitting (vanSchilfgaarde and Mryasov 2001).
One of the degenerate d-state involves in the bonding and antibonding with N
– p state, while the other act as a non-bonding state (eg and t2g) as shown in the
Figure 2.4, 2.5. As the t2g band lies above eg band, the tetrahedral bonding of
Ni atom in the GaN structure is confirmed (Das et al 2006). Thus the spin of
the defect state may give rise to the magnetic moment.
Figure 2.4 Electronic Band Structure of 6.25 % Ni doped GaN
33
Figure 2.5 Electronic Band Structure of 12.5 % Ni doped GaN
Figure 2.6 shows the total Density of State (DOS) of 6.25 % and
12.5 % Ni doped GaN. The half metallic nature is clearly seen in 6.25 %
when compared to 12.5 %. The partial density of states for Ni–3d and N-2p
which are given in Figure 2.7 shows the p-d hybridization of Ni-N in the
bonding state.
(A) (B)
Figure 2.6 (A) Density of states of 6.25 % Ni doped GaN (B) DOS of
12.5 % Ni doped GaN
34
(A) (B)
Figure 2.7 Partial density of states of Ni-3d and N-p in (A) Spin up
(B) Spin down state
The magnetic moment per Ni atom in 6.25 and 12.5 % are 1.13 B
and 1.32 B respectively, but the total magnetic moment in each case are
different because nickel atom polarizes the four nitrogen atoms bonding with
it. In 6.25 % concentration, the three equidistant Ni which bonds with N
atoms have a magnetic moment of 0.1808 B and the other N atom has 0.2151
B magnetic moment. For 12.5 % concentration, i.e., when two Ga atoms are
replaced by Ni, calculations are optimized by substituting the Ni atoms at two
different sites. Figure 1A shows the formation of the two Ni atoms which
forms a dimer with N atom and the Ni – Ni distance is 3.253 Å (Assadi et al
2009). The magnetic moment per Ni atom is found to be 1.13 B and the total
moment was also found to decrease when compared to 6.25 % concentration.
This is due to the interaction of Ni spins via N atom bonded to them. When Ni
is substituted as shown in Figure 2.3(B), the magnetic moment is found to
increase per Ni atom (1.32 B) and the system is not energetically favorable
when compared to the configuration shown in Figure 2.3(A). From this we
35
conclude that when Ni-Ni distance is 3.253 Å, the magnetic moment per Ni
atom is 1.13 B and the magnetic moment was less when compared to the
6.25 % concentration. For the other position, the magnetic moment is found to
increase at Ni site. For 12.5 % concentration the half metallic nature
disappears and becomes a ferromagnetic alloy. Hence, we conclude that only
for a small addition of Ni in GaN one can achieve spintronic in these systems.
In case of defect created by Ga vacancy as show in Figure 2.3(C).
The magnetic moment is due to the spin interaction of p-p electron of
surrounding nitrogen atom which is in the tetrahedral bonding state. The three
N atoms are at a distance of 3.689 Å from the Ga vacancy which give a
magnetic moment of 0.13 µB. The other N which is at a distance of 3.678 Å
from VGa contributes to a magnetic moment of 0.43 4µB. For Ni doped GaN
with Ga vacancy as shown in Figure 2.3(D). A strong Ni-N bond is formed by
charge transfer from d electron of Ni to acceptor level of VGa. Hence, the 3d
electron of Ni has no role in spin polarization and therefore, the observed
magnetic moment is due to the dangling bond electron of N which is 0.299
µB. The Nitrogen vacancy defect as shown in Figure (2.3E) has no role in
spin polarization. Nickel substituted in the gallium site with nitrogen vacancy
as shown in Figure 2.3(F) has a magnetic moment of 1 µB and is similar to the
6.25 % Ni dopant case but the moment is slightly less.
2.4.2 Electronic and Magnetic Properties of Cobalt Doped GaN
We calculated the magnetic moment for 1.56 % Co doped GaN
which corresponds to replacing one Ga atom by Co in a matrix of 32 Ga
atoms. The effects of structural relaxation due to impurities are neglected
because of small radial difference between Ga and Co atoms. Other forms of
defects such as interstitial or anti-site are not considered because they are
energetically strongly unfavorable when compared to Co occupying the
36
substitution site. The Co atom is in tetrahedral bonding with four nitrogen
atom. The bonding length of one of the nitrogen atom with Co is 1.946 Å with
a magnetic moment of 0.20 µB. The remaining nitrogen atoms show a
magnetic moment of 0.21 µB. The magnetic moment of Co atom per cell is
2.56 µB. This highly localized moment arises due to spin polarization from the
3d electron of Co which does not contribute to the p-d hybridization from 2p
and 3d electrons of N and Co atom. For 2.77 % and 4.16 % of Co dopants in
GaN, this is obtained by replacing Ga by Co in a matrix of 18 and 12 Ga
atoms. The calculated magnetic moment per cobalt atoms is same as
compared to 1.56 % but decrease to 1.48 µB for 4.16 %.
(A) (B)
Figure 2.8 Total density of state for (A) 2.77 % Co doped GaN (B) 4.16 %
Co doped GaN
Figure 2.8 (A, B) shows the total density of state of 2.77 % and
4.16 % Co doped GaN. It is observed form the density of states figure that
2.77 % concentration of Co in GaN is perfectly half metallic with almost 100
% spin polarization. For 4.16 % of Co dopant the system becomes metallic
but is still ferromagnetic, because the 3d electron of cobalt has a band in the
37
band gap of GaN. As a result, the contribution for spin polarization from 3d
electron is small, which in turn decreases the magnetic moment value of 1.08
µB for 4.16 % Co dopant.
2.5 CHARACTERISATION OF COBALT DOPED GaN
2.5.1 X-ray Diffraction Analysis
The most common use of powder (polycrystalline) X-ray
diffraction is chemical analysis. This can include phase identification.
Investigations on high/low temperature phases, solid solutions and
determination of unit cell parameters of a material. A typical diffraction
spectrum consists of a plot of reflected intensities versus the detector angle 2
or depending on the goniometer configuration. The 2-Theta values for the
peak depend on the wavelength of the anode material of the X-ray tube. It is
therefore customary to reduce a peak position to the interplanar spacing d that
corresponds to the h, k, l planes that caused the reflection. The value of the d-
spacing depends only on the shape of the unit cell. We get the d-spacing as a
function of 2-Theta from Bragg’s law d = n /2 sin . Once we know the d-
spacing and the corresponding indices h, k, l, we can calculate the lattice
parameter of the unit cell.
The synthesized GaN was characterized using D8 Brucker AXS X-
ray diffractometer with Cu K source. Figure 2.9 shows the XRD spectra of
synthesized GaN. The sample is in polycrystalline nature, and confirmed the
wurtzite structure of pure GaN. The lattice parameter calculated from the
XRD data is a =3.186Å and c =5.174Å, which are in good agreement with
the reported values (Senthil kumar and Kumar 2002).
38
Figure 2.9 XRD of synthesized pure Gallium Nitride powder sample at
1223 K
2.5.2 Photoluminescence
Photolumiscence (PL) is a non-destructive technique to characterize
the intrinsic properties of materials like band to band edge recombination to
evaluate the band gap energy, shallow defect levels like donor or acceptor
levels. When light (laser beam) of sufficient energy is incident on a material,
photons are absorbed and electronic excitations are created. Eventually, these
excitations relax and the electrons return to the ground state. If radiative
relaxation occurs, the emitted light is called Photoluminescence (PL). This
light can be collected and analyzed to yield a wealth of information about the
photoexcited material. The PL spectrum provides the transition energies,
which can be used to determine electronic energy levels. The PL intensity
gives a measure of the relative rates of radiative and nonradiative
recombination. Variation of the PL intensity with external parameters like
39
temperature and applied voltage can be used to characterize further the
underlying electronic states and bands. A monochromator is used to get a
single wavelength of light. The photo detector senses the radiative emission
and then it processes this weak signal for analysis. The PL setup used in our
experiment is shown in Figure 2.10.
Figure 2.10 Photoluminescence system
The emission properties of 1 % and 5 % Co doped GaN are
measured at room temperature using 325 nm He-Cd laser as excitation source
and PL spectra can be recorded for a wide range of 300-1500nm. It is
observed that there is a blue shift in the band edge emission as the cobalt
concentration increases. The broad band edge emission is due to the
polycrystalline nature of the synthesized samples. The emission at 3.24 eV is
mostly due to nitrogen vacancy defects (Keller et al 1995) followed by its first
phonon replica at 3.13 eV.
40
Figure 2.11 PL spectrum of Co doped GaN samples obtained with
325 nm excitation at room temperature
2.5.3 Magnetic Studies
One of the most sensitive forms of magnetometry is
superconducting quantum interference device SQUID magnetometer. This
uses the properties of electron-pair wave coherence and Josephson junctions
to detect very small magnetic fields with resolutions up to ~10-14 kilo Gauss
or greater. The central element of a SQUID is a ring of superconducting
material with one or more weak links called Josepheson’s Junctions. With
weak-links at two points, the critical current ic is much less than the critical
current of the main ring. This produces a very low current density making the
momentum of the electron-pairs small. The wavelength of the electron-pairs
is thus very long leading to little difference in phase between any parts of the
ring. The Schematic representation of a SQUID is shown in Figure 2.12.
41
Figure 2.12 Schematic diagram of a Superconducting Quantum
Interface Device (SQUID)
Most common use of SQUID is in magnetic property measurement
systems (MPMS). This is typically done over a temperature range from 4 K to
190 K. Figure 2.13 shows the magnetization curve of 1 % and 5 % Co doped
GaN as a function of magnetic field at 10 K. The observed magnetic moment
at 3000 Oe is 2.03 µB and 0.54 µB. From hysteresis loop it is clear that the
synthesized sample is a perfect ferromagnetic. A clear opening of hysteresis
cycle is observed only for 1 % Co doped GaN. But the trend is no longer
observed for higher percentage of dopant. Due to the solubility limit of
transition metal cobalt which may be less than 1% in GaN (Zaj c et al 2008),
so as the concentration of dopant increases, there is a good possibility for the
formation of other impurity phases such as CoGa which shows paramagnetic
behavior (Booth and Marshall 1970).
42
Figure 2.13 Magnetization curve of 1% and 5% Co doped GaN as a
function of applied magnetic field at 10 K
The other less possible impurity is CoOx because the synthesis of
GaN is done in nitrogen ambient to prevent oxidation, which may have
antiferromagnetic behavior. This may be the possible reason for the observed
decrease in magnetic moment values as the concentration of Cobalt dopant
increases. Figure 2.14 shows the temperature dependent magnetization (M-T
curve) of synthesized 1 % and 5 % Co doped GaN with a magnetic field of
3000 Oe. It is clear that there is no drastic decrease in magnetization value as
temperature increases till 300 K for 1 % Co doped GaN. This indicates that
the ferromagnetic transition temperature is above room temperature, since as
temperature increases the ferromagnetic order remains unchanged due to
strong alignment of 3d electron of cobalt in the applied field direction. For
higher percentages of cobalt, the magnetization decreases rapidly at very low
temperature.
43
Figure 2.14 Temperature dependent magnetization (M-T curve) of 1 %
and 5 % Co doped GaN under a magnetic field of 3000 Oe
The theoretical calculations have shown a large value of the
magnetic moment when compared with the experimental values. The reason
may be due to the structural defects, Co occupying the interstitial site and the
small strain produced due the ionic radii difference of Co which is not taken
into account while calculating the magnetic moment values and also due to
the variation in the percentage of dopant used while synthesizing the samples.
The origin of the observed ferromagnetism from SQUID measurement is not
clear and may be due to formation of isolated Co cluster in synthesized GaN.
2.6 CONCLUSIONS
Theoretical calculations have been carried out to investigate the
possibility of inducing ferromagnetism in GaN by the substitution of Ni and
Co. When the Ni concentration is 6.25% the system shows half metallic and
44
ferromagnetic behavior. The dopant site preference shows that Ni dimer via N
is energetically more stable when compared to the other position as discussed.
When a vacancy is created the system becomes metallic. When Ni is
substituted in Ga site with either a Ga or N vacancy the system shows
magnetism. In all these cases half metallic nature is absent. The half metallic
behavior is observed upto 2.77 % Co dopant, but for higher percentage of
dopant the system become metallic.
The structural and optical properties of doped GaN were studied
using X- ray diffractometer and photoluminescence respectively. The
magnetic properties were studied at 10 K using superconducting quantum
interface device (SQUID). From temperature dependent magnetization
measurements it is observed that there is no drastic change in the magnetic
moment up to room temperature for lesser cobalt dopant. However, we
assertively conclude from both experimental and theoretical observations that
only lower percentage of dopant in GaN is a suitable candidate for spintronic
devices operating at and above room temperature. The difference in magnetic
moment value between the experimental and theoretical results may be due to
the influence of structural defects on the ferromagnetic order and also the
position of cobalt occupancy in GaN matrix. To induce half metallic
ferromagnetism in GaN it is suggested that a low percentage of Ni or Co
without any vacancy in the system may be suitable to function as the good
candidate for spintronic applications.