Photoconductance in magnetic tunnel junctions

Abstract

The photoconductance of magnetic tunnel junctions has been studied in order to obtain directly determined values of the tunnel barrier height. Experiments for different types of junc-tions show that the presence of an aluminum layer close to the oxide barrier is crucial for the observation of a large photocur-rent. For Ni 80 Fe 20 -based magnetic tunnel junctions, the effective barrier height increases with increasing magnetic layer thickness between the aluminum source layer and the aluminum oxide bar-rier. This trend is also visible in the barrier height determined from a Simmons fit to the current–voltage characteristics.

Full text

Page 1

2712IEEE TRANSACTIONS ON MAGNETICS, VOL. 38, NO. 5, SEPTEMBER 2002
Photoconductance in Magnetic Tunnel Junctions
P. H. P. Koller, F. W. M. Vanhelmont, R. Coehoorn, and W. J. M. de Jonge
Abstract—The photoconductance of magnetic tunnel junctions
has been studied in order to obtain directly determined values of
the tunnel barrier height. Experiments for different types of junc-
tions show that the presence of an aluminum layer close to the
oxide barrier is crucial for the observation of a large photocur-
rent. For Ni??Fe??-based magnetic tunnel junctions, the effective
barrier height increases with increasing magnetic layer thickness
between the aluminum source layer and the aluminum oxide bar-
rier.Thistrendisalsovisibleinthebarrierheightdeterminedfrom
a Simmons fit to the current–voltage characteristics.
Index Terms—Magnetoresistance, magnetoresistive devices,
photoconducting materials, photoconductivity, tunneling.
I. INTRODUCTION
F
knowledge of the height of the tunnel barrier is of crucial
importance. Within the simple Simmons [1] or Brinkman [2]
models, the energy barrier shape is assumed to be square or
trapezoidal, respectively. The barrier height parameter
the barrier thickness
can be determined in an indirect way,
namely by a Simmons or Brinkman fit to the current–voltage
characteristics of the junction. It should be noted, however, that
the barrier height and thickness cannot be determined inde-
pendently within these models. This is an important drawback
if, for example, the thickness is nonuniform over the junction
area. It has been shown that, in such a case, the barrier height
obtained from a fit to the –
curve is too high [3].
Methods to measure the barrier height directly include
ballistic-electron-emission-microscopy (BEEM) experiments
where a scanning-tunneling-microscope tip is used as a source
of hot electrons [4], [5]. However, with BEEM only local
barrier heights can be determined. These can be different from
the integral barrier height [5], which is important for device
applications. Also, restrictions apply to the thickness of the
top-electrode, because a significant part of the hot electrons
must be able to traverse this thin layer without being scattered.
In this paper, we present results from a photoconductance
technique, with which the transport properties across the insu-
lating layercan be investigated.For thefirst time, this technique
OR THE understanding of the resistance and magne-
toresistance (MR) of magnetic tunnel junctions (MTJs),
and
Manuscript received February 11, 2002; revised May 22, 2002. This work
was supported by the Dutch Foundation for Fundamental Research on Matter
(FOM).
P. H. P. Koller and W. J. M. de Jonge are with the Department of
Applied Physics, Eindhoven University of Technology, NL-5600 MB
Eindhoven, The Netherlands (e-mail: koller@natlab.research.philips.com;
W.J.M.d.Jonge@tue.nl).
F. W. M. Vanhelmont and R. Coehoorn are with the Philips Research
Laboratories, NL-5656 AA Eindhoven, The Netherlands (e-mail: frederik.van-
helmont@philips.com; reinder.coehoorn@philips.com).
Digital Object Identifier 10.1109/TMAG.2002.803171.
TABLE I
STUDIED TUNNEL JUNCTION SYSTEMS
has been applied to MTJs to determine the potential step at the
barrier/electrode interfaces in a direct way.
II. SAMPLE PREPARATION
The investigated tunnel junctions were all fabricated with the
useofaninsitushadowmasksystemresultinginacross-barge-
ometry. The substrates used were Corning 1737 glass slides that
were chemically cleaned in a boiling ammonia/ethanol solution
and exposed to an oxygen plasma prior to deposition to remove
all organic material. The layers are deposited by dc magnetron
sputtering in a Kurt J. Lesker ultrahigh vacuum sputter machine
with a base pressure of 10
is transferred under vacuum to a separate oxidation chamber,
where an oxygen plasma can be ignited.
Through a metallic shadow mask in contact with the sub-
strate, the bottom electrode is first sputter deposited. Then the
maskisremoved,anda1.7-nmaluminumlayerisuniformlyde-
posited across the whole substrate. This layer is oxidized in an
oxygen plasma for 90 s (pressure
After oxidation, the shadow mask is rotated 90 and placed
again in contact with the substrate. The top electrode is then de-
posited. This fabrication technique results in 200 m
tunnel junctions in a cross-bar geometry. No anneal steps were
carried out afterward. Five different material systems, as shown
in Table I, were studied. MTJs prepared in a similar way with
the same machine have previously been shown to give MR ra-
tios of up to 20% [6].
Pa. For oxidation, the sample
9.3 Pa and power35 W).
200 m
III. EXPERIMENTAL PROCEDURE
For the conductance, magnetoconductance, and photocon-
ductance measurements, all carried out at room temperature,
the setup shown schematically in Fig. 1 is used. All electrical
measurementswereperformedusingafour-pointconfiguration.
For the optical excitation studies, a 75-W xenon lamp in
conjunction with a 1/4 m Jobin–Yvon grating monochromator
is used as a monochromatic light source. The absolute light in-
tensity is measured using a calibrated photodiode at the sample
position before each measurement. A mechanical chopper is
used in conjunction with a lock-in amplifier to separate the
0018-9464/02$17.00 © 2002 IEEE

Page 2

KOLLER et al.: PHOTOCONDUCTANCE IN MTJs2713
Fig. 1.
experiments on different tunnel junction structures. ? represents a lens, ?? a
beam-splitter, and ?/? a compensation circuit to eliminate the photovoltage
contribution.
Schematic view of the experimental setup for photoconductance
photocurrent from the dark current. A bias voltage can be
applied to the junction structure. The dc tunnel current as well
as the ac photocurrent are measured with the use of a personal
computer. Due to the significance of the lead resistances com-
pared to a junction resistance also a photovoltage will develop
across the junction. An active compensation circuit is used to
stabilize the bias voltage and thus eliminate the photovoltage
contribution. Finally, a camera is used to position the sample
under the light spot.
All measurements were carried out at zero applied bias
voltage and with the light incident through the top electrode.
The light spot always completely covered the junction area.
IV. RESULTS
When a monochromatic light beam is incident on a junction
structure, photons that penetrate the metal can excite electrons
to a higher energy state. These hot electrons have, to a first
approximation, an isotropic momentum distribution. Due to
inelastic scattering processes, only a portion of the electrons
will reach the barrier interface without energy loss. When the
energy of the hot electrons is higher than the oxide potential
barrier, these electrons can enter the conduction band of the
oxide layer. When collected at the other electrode, they will
contribute to the measured photocurrent. If the top electrode is
thin enough also, electrons excited in the bottom electrode can
contribute tothephotocurrent. Thetotal photocurrent is thesum
of these two currents. In a simple analysis, the photocurrent
can, for photon energies
the Fowler equation (see [7] and the references therein), from
which the photocurrent is proportional to (
Since the 1970s, it is already known that Al–AlO –Al junc-
tions show a strong photoconductance [7]. Our results for type
I samples (see Fig. 2) show indeed a large signal, from which
the effective barrier height can be deduced (see below). We
wanted to extend these measurements to our “standard” MTJs
(type V) so that an easy and direct method for barrier height
determination is obtained. However, these tantalum-topped ex-
change-biased tunnel junctions, showing magnetoresistance ra-
tios of 20% and more, did not show any photocurrent. We at-
tributethispresentlytoaquiteshorthot-electronmean-freepath
in Ni Fe
[8] and Ta [9], at energies above the expected en-
ergy barrier of
2 eV. To investigate the role of aluminum
, be represented by
.
Fig. 2.
was always incident through the top electrode. A positive photocurrent means
thatthenetelectronflowisfromthetoptothebottomelectrode.Nophotocurrent
could be measured on samples of type V (glass–M–AlO –M–Ta). Note the
different scales for samples III and IV.
Photocurrent curves for different tunnel junction structures. The light
TABLE II
BARRIER HEIGHTS DETERMINED FROM PHOTOCONDUCTANCE EXPERIMENTS
in photoconductance measurements, we made several different
tunnel junction structures with Al present in one or both of the
electrodes (types I–IV). It is clear from Fig. 2 that when Al is
placed close to the barrier interface, a large photocurrent is vis-
ible. When Al is placed below thebarrier (type III), the net elec-
tronflowis,forphotonenergiesbetween
from the bottom to the top electrode. These results clearly show
the large efficiency of Al in forming a source of hot electrons
that eventually contribute to the measured photocurrent.
Using the Fowler equation, the effective barrier height can
be determined by plotting the square root of the photocurrent
versus the photon energy [7]. In Table II, the determined ef-
fective barrier heights and the resistance–area product (
are presented. Except for samples of types I and III, all con-
ductance curves could be fitted well with the Simmons equa-
tion. Types I and III showed a significant asymmetry in their
current–voltage characteristics, which could only be fitted well
withtheBrinkmanequationrevealingalargebarrierasymmetry
(
2 to 3 eV). They also have a higher
than the other junction types. This can be attributed to the fact
2.3eVand 2.8eV,
)
product

Page 3

2714IEEE TRANSACTIONS ON MAGNETICS, VOL. 38, NO. 5, SEPTEMBER 2002
Fig. 3.
Simmons fits (?) as well as MR ratios ( ) for type IV MTJs with varying top
electrode Ni
Fethicknesses.
Barrier heights obtained from photoconductance curves ( ) and
that the oxidation process does not stop at a less reactive bottom
electrode, whereas for the other junctions the Co Fe
electrode layer at least slows down the oxidation process [6].
This results then in a thicker and nonisotropic barrier layer,
leadingtoalarger
productandasignificantbarrierheight
asymmetry.
For type II and IV junctions, the composition is the same ex-
cept for the top electrode, which apparently results in a lower
effective barrier height when Al is situated at the interface. To
investigatethisfurther,aseriesofsamplesoftypeIVweremade
with varying thicknesses for the Ni Fe
trode between the AlO barrier and the Al source layer. The
barrier heights and the MR ratios are shown in Fig. 3. A clear
increase in barrier height with increasing Ni Fe
visible,whichisalsoaccompaniedbyanincreaseinMR.Fitting
the Simmons equation to the current–voltage characteristics of
thedifferentjunctionscorroboratesthisdependence:onlythein-
creaseismuchstronger,andthebarrierthicknessdecreaseswith
increasing Ni Fe
thickness (not shown). However, when the
barrier height determined from photoconductance experiments
is used as a fixed fit parameter in the Simmons fit, the barrier
thickness remains roughly constant (not shown).
Apparently, the Al forms a lower effective barrier height with
AlO ascomparedtoNi Fe .PollackandMorris[10]showed
similar results on a Ni electrode. The Ni Fe
the maximum TMR is reached (15–20 nm) is, however, much
larger than previous results on Co Fe
bottom
layer in the top elec-
thickness is
thickness where
(4 nm)/Ta [11]. This
suggests that the Al layer somehow changes the barrier/elec-
trode interface, even at relatively large Ni Fe
resulting in a lower MR. The exact cause of this influence is at
present not clear.
Currently, we are investigating different materials to see
whether material combinations other than Ni Fe –Al allow
one to obtain a significant photoconductance as well as main-
tain a high MR ratio.
thicknesses,
V. CONCLUSION
For the first time, photoconductance experiments have been
carriedoutonmagnetictunneljunctions.TheroleofAlasanef-
fective source and transport medium for hot electrons has been
demonstrated. From a Fowler analysis, the barrier height for
samples with different magnetic and nonmagnetic electrodes
has been measured. It was found that the effective barrier height
increases with increasing thickness of the Ni Fe
tween the barrier and the Al source layer.
layer be-
REFERENCES
[1] J. G. Simmons, “Generalized formula for the electric tunnel effect be-
tween similar electrodes separated by a thin insulating film,” J. Appl.
Phys., vol. 34, pp. 1793–1803, 1963.
[2] W.F.Brinkman,R.C.Dynes,andJ.M.Rowell,“Tunnelingconductance
of asymmetrical barriers,” J. Appl. Phys., vol. 41, pp. 1915–1921, 1970.
[3] W. Oepts, “Spin dependent transport in structured magnetic layers,”
Ph.D. dissertation, Dept. Physics, Eindhoven Univ. Technology,
Eindhoven, The Netherlands, 1999.
[4] W. H. Rippard, A. C. Perrella, and R. A. Buhrman, “Ballistic electron
microscopy study of ultrathin oxidized aluminum barriers for magnetic
tunnel junctions,” Appl. Phys. Lett., vol. 78, pp. 1601–1603, 2001.
[5] O.Kurnosikov,J.E.A.deJong,H.J.M.Swagten,andW.J.M.deJonge,
“DirectobservationoflocalhotelectrontransportthroughAl O tunnel
junctions,” Appl. Phys. Lett., vol. 80, pp. 1076–1078, 2002.
[6] A. E. T. Kuiper, M. F. Gillies, V. Kottler, G. W. ‘t Hooft, J. G. M. van
Berkum, C.vanderMarel, Y.Tamminga, andJ.H. M.Snijders,“Plasma
oxidation of thin aluminum layers for magnetic spin-tunnel junctions,”
J. Appl. Phys., vol. 89, pp. 1965–1972, 2001.
[7] J.KadlecandK.H.Gundlach,“Resultsandproblemsofinternalphotoe-
mission in sandwich structures,” Phys. Stat. Sol. (a), vol. 37, pp. 11–28,
1976.
[8] R. Vlutters, R. Jansen, O. M. J. van ’t Erve, S. D. Kim, and J. C. Lodder,
“Hot-electron transport through Ni
Appl. Phys., vol. 89, pp. 7305–7307, 2001.
[9] M. Aeschlimann,S. Pawlik, andM.Bauer, “Femtosecond time-resolved
measurement of electron relaxation at metal surfaces,” Ber. Bunsenges.
Phys. Chem., vol. 99, pp. 1504–1508, 1995.
[10] S. R. Pollack and C. E. Morris, “Electron tunneling through asymmetric
filmsofthermallygrownAl O ,”J.Appl.Phys.,vol.35,pp.1503–1512,
1964.
[11] J. J. Sun and P. P. Freitas, “Dependence of tunneling magnetoresistance
on ferromagnetic electrode thickness and on the thickness of a Cu
layer inserted at the Al O /CoFe interface,” J. Appl. Phys., vol. 85, pp.
5264–5266, 1999.
Fein a spin-valve transistor,” J.