Development of the spin-valve transistor

Abstract

As the easiest experimental approach, GMR (giant
magnetoresistance) is usually measured using the current in plane
(CIP)-GMR. The spin-valve transistor has previously been presented as a
spectroscopic tool to measure current perpendicular to the planes
(CPP)-GMR. Hot electrons cross the magnetic multilayer base
quasi-ballistically and the number reaching the collector depends
exponentially on the perpendicular hot electron mean free path.
Collector current changes of 390% at 77 K have already been measured.
Apart from the substantial fundamental value, such properties may be
useful for sensor applications. The electron energy range fills the gap
between the Fermi surface transport in resistance measurements and other
hot electron techniques such as spin polarised electron energy loss
spectroscopy (SPEELS). The preparation problem of the spin-valve
transistor and metal base transistor structures in general, the
deposition of a device quality semiconductor on top of a metal, has now
been tackled by bonding of two semiconductor substrates during vacuum
deposition of a metal: an excellent bond is achieved at room
temperature. TEM photos show a continuous buried metal film. Apart from
preparation of various metal base transistor like structures, many other
fields may benefit form this new technique

Full text

Page 1

IEEE TRANSACTIONS ON MAGNETICS, VOL. 33, NO. 5, SEPTEMBER 1991
3495
Development of the Spin-valve transistor
D.J. Monsma, R. Vlutters, T. Shimatsu, E.G. Keim*, R.H. Mollema, and J.C. Lodder
Information Storage Technology Group, MESA Research Institute, *Center of Materials Research,
University of Twente, 7500AE Enschede, The Netherlands
Abstract--As the easiest experimental approach, GMR (Giant
Magnetoresistance) is usually measured using the Current in
Plane (C1P)-GMR. The spin-valve transistor has previously been
presented as a spectroscopic tool to measure Current
Perpendicular to the Planes (CPP)-GMR. Hot electrons cross the
magnetic multilayer base quasi-ballistically and the number
reaching the collector depends exponentially on
perpendicular hot electron mean free path. Collector current
changes of 390% at 77K have already been measured. Apart
from the substantial fundamental value, such properties may be
useful for sensor applications. The electron energy range fills the
gap between the Fermi surface transport in resistance
measurements and other hot electron techniques such as spin
polarised electron energy loss spectroscopy (SPEELS). The
preparation problem of the spin-valve transistor and metal base
transistor structures in general, the deposition of a device quality
semiconductor on top of a metal, has now been tackled by
bonding of two semiconductor substrates during vacuum
deposition of a metal: an excellent bond is achieved at room
temperature. TEM photos show a continuous buried metal film.
Apart from preparation of various metal base transistor like
structures, many other fields may benefit from this new
technique.
the
INTRODUCTION
The discovery of giant magnetoresistance in magnetic
multilayers [I], more appropriately called the spin-valve
effect [a], has led to a large number of studies on metallic
systems, motivated both by the enormous potential of the
effect for magnetic sensors and the fundamental interest of the
underlying principles. These efforts have resulted in magnetic
multilayers that show considerable resistance change (of the
order of 10%) in fields useful for read head applications (20
Oe or 1.6 kA/m) and magnetic RAMS. Usually, the resistance
is measured with the current in the planes of the layers. From
a fundamental point of view however, the CIP configuration
suffers from several drawbacks: the CIP magnetoresistance
(MR) is diminished by shunting and channeling and diffusive
surface scattering reduces the MR for sandwiches and thin
multilayers.
Moreover, fundamental parameters of the effect, such as
the relative contributions of interface and bulk spin dependent
scatterings are difficult to obtain using the CIP geometry.
Measuring with the current perpendicular to the planes (CPP)
solves most of these problems, mainly because the electrons
cross all magnetic layers, but the perpendicular resistance of
the ultra thin multilayers is too small to be measured by
ordinary techniques. The first CPP-MR experiments were
Manuscript submitted April 24, 1997.
D.J Monsma 3 1-53-4893355 fax. -3343 D.J.Monsma@el.utwente.nl,
This work was supported by the European research program Brite Euram
“’7192, BRE2-0546.
reported on Co/Ag multilayers [3], where the multilayer was
sandwiched between superconducting Nb leads. The use of
microfabrication techniques for CPP measurements from
4.2K to 300K was firstly shown for FeKr multilayers [4],
where the multilayers were etched into micro pillars to obtain
a relatively large resistance (a few d).
measurements have confirmed the larger MR effect for the
CPP configuration. Experiments using electrodeposited
nanowires demonstrated CPP-MR at RT [5],[6] and recent
experiments with etched grooved substrates introduced the
so-called CAP (Current with an Angle to the Planes)-
configuration, for which also an increase of the spin-valve
effect has been observed as compared to its CIP counterpart
[7]-[9]. The spin-valve transistor has been presented as a new
magnetic field sensor and measurement tool based on
perpendicular hot electron transport in a spin-valve multilayer
[lo]. Here, a spin-valve multilayer serves as a base region of
an n-silicon metal base transistor (MBT) structure, showing a
very large change (215% at 77K) in collector current in 500
Oe (40 W m ) , using a (CO/CU)~
the electrons and the possibility to vary the electron energy
accurately in a range of about 0.2-3eV, raises possibilities for
fundamental research of the spin-valve effect and may lead to
unforeseen effects and new spin-valves. In contrast to usual
conduction electrons which are sensitive to the density of
states (DOS) at the Fermi energy (EF), the hot electrons are
sensitive to the DOS above EF.
Manipulation of electron energy may offer important
insights into the relative importance of the band structure
(DOS) and scattering potentials, which are understood to
form the basis of the scattering asymmetry between majority
and minority conduction electrons. Currently electron energy
spectroscopy in transport of magnetic systems
increased interest as an additional tool to study the
fundamentals of the spin-valve effect.
Both types of
multilayer. The hot nature of
receives
Fig. 1 SEM picture showing the collector, multilayers base region and
direct-bonded (in air) emitter. Aluminium wires are connected using
ultrasonic wire bonding.
0018-9464/97$10.00 0
1997 IEEE

Page 2

3496
Techniques such as spin polarised tunneling, scanning
electron microscopy with polarisation analysis (SEMPA),
spin polarised low electron energy diffraction (SPLEED) and
spin polarised electron energy loss spectroscopy (SPEELS)
are under investigation in various laboratories [111-[17].
PRINCIPLE
A SEM picture of one of the realised spin-valve transistor
structures in shown in Fig. 1. A (Cu 2nm/ColSnm)4
multilayer base is rf-sputtered onto the Si( 100) collector
substrate. The emitter has been direct-bonded in air to the
base multilayer. The emitter is negatively biased (forward)
using a DC current source, the collector substrate is in reverse
(positive voltage bias), in common base. The base area is 2.7
x 2.7 mm2. the emitter area 1.6 x 1.6 mm2.
HOT ELECTRON TRANSPORT
A schematic energy band diagram of the bonded Co/Cu
spin-valve transistor is shown in Fig. 2. The emitter
(collector) barrier height is about 0.7eV (0.6eV). The emitter
bias accelerates the electrons over the emitter barrier, after
which they constitute the hot, quasi-ballistic electrons in the
base. The probability of passing the collector barrier is
limited by collisions in the base, which affect their energy and
trajectory, by optical phonon scattering in the semiconductors
and by quantum mechanical reflections at the base-collector
interface. For a metal base transistor with a single metal base
film, this can be expressed by the common base current
transfer ratio or current gain ~=(Jc-J,e,k)lJ,=aeac,a,,
which a,, a, and aqm represent emitter efficiency, collector
efficiency and quantum
respectively. W is the base width and h is the hot electron
mean free path (MFP) in the base. The factor
the probability of transmission of the hot electrons through
the base. J, is the total collector current, Jleakage is the collector
leakage current, determined by the reverse-biased collector
Schottky barrier and J, is the injected emitter current. In the
spin-valve transistor under consideration, the collector current
of the CoKu spin-valve transistor depends exponentially on
the spin dependent hot electron mean free paths h~,?.,
base. As a result from the perpendicular trajectory, the
exponential mean free path (MFP) dependence and possibly
also the electron energy, the collector current variation is
large, as shown in Fig. 3.
e-w‘’, in
mechanical transmission,
represents
in the
I
I
I
I
I
I
I
I
-2.0
-1.5
-1.0
-0.5
0.0
0.5
1.0
1.5
2.0
Magnetic field (kOe)
FIG. 3. The collector current change of the Si-(Cu 2nm/Co1.5nm)4-Si spin-
valve transistor at 77K.
Because of the low barrier height and large area of the
collector, the leakage current is quite large (Ileak=30 FA) and
exceeds the hot electron collector current for an injection
current of 100 mA. Therefore collector current measurements
have been performed at 77K, reducing the leakage current to
very small values. Typical GMR characteristics of a second
peak Co/Cu multilayer, such as saturation field and hysteresis,
are observed. The value at coercive field is O.lpA, at +/-lo
kOe (796 W m ) 0.5pA, resulting in more than 390%
collector current change. The corresponding CIP-MR value
of the implemented multilayer was only 3% in 10 kOe (796
kA/m, 77K). The large values of CC(%) and JJJ, indicate a
short XJ,,~ (of the order of OS(1)nm).
In [ 181 a calculation has been made of the enhancement of
GMR as a function of the electron energy above the Fermi
level in Co/Cu(lOO) multilayers using a calculation within the
Kubo formalism in which the electronic bandstructure has
been taken into account. There is an asymmetry in the DOS
for majority and minority spin, which is typical for
ferromagnetic metals. The larger velocity of hot majority
electrons travelling increasingly in the sp band above EF
results in an increasing GMR with electron energy up to about
1 eV. Above 1 eV the asymmetry will diminish and GMR is
expected to decrease.
APPLICATION
PROSPECTS
As a consequence of the direct MFP dependence of the
transmission across the base, the spin-valve transistor allows
collector current quantification of spin dependent electron MFPs A & ( ? ) of the
individual layers and interfaces. The relation between the
change in collector current and the mean free path ratio hTlh~,
expected from the transport equations [lo] is illustrated in
Fig. 4. The relative change in collector current CC(%) and
the relative change in perpendicular hot-electron spin valve
effect in the base CPP-MR(%) are plotted versus the MFP
ratio h&. CC( %)=(Jp-JAp)/JAp can either be calculated using
W l h ~
[7] or using WILT:
hot-electron
base
I +-
FIG. 2. Schematic energy band diagram of the spin-valve transistor under
forward bias.

Page 3

3491
Since W/& represents maximum output, it is the
appropriate value for signal to noise calculations. As shown
in Fig. 4, the structure may amplify the intrinsic perpendicular
hot-electron MR effect strongly, depending on W/I of the
used multilayer.
10000
1
2
3
4
5
6 7 8 9 1 0
MFP ratio h,/h,
FIG. 4. Transmission characteristics of the spin-dependent transport. Notice
the difference with [lo].
The ratio Jleakage/Jcc and, for applications, the desired output
level of the spin valve transistor, set an upper limit to W/ht.
Since the collector current decreases with W/XT and the
relative collector current change increases with W / ~ T ,
of merit may be calculated by setting the relation between the
noise level, the output current level and the output current
change to an optimum. In Fig. 4 it is seen that even with a
very small attenuation factor (w/+1
collector current change is still very large, certainly in
comparison with the corresponding CIP value. This means
that with a suitable base multilayer or sandwich (with large
electron mean free path or small thickness) a large collector
current can be combined with a large change. From (1) it is
also seen that CC(%) does not depend on additional non-
magnetic layer thicknesses. The achievable sensitivity is
mainly determined by the signal to noise ratio SNR of the
device. The main noise source is partition noise and can be
represented by a collector shot noise generator i,=d(2qIGdf)
from which an indication of the SNR can be calculated. The
magnetic field dependent output is not a voltage as in a four
point measurement, but a high impedance collector current
source. Also, the noise sources of the SVT and the MR
resistor are different: in the SVT partition noise prevails, in
the resistor thermal noise. A more detailed analysis is
currently under study.
a figure
or e-'s0.4), the
PREPARATION OF THE SVT
crystal emitter [IO]. In direct bonding two flat, smooth and
clean materials can form a permanent electrical and
mechanical connection by means of spontaneous adhesion
[19]. In this way a sawn piece of silicon was, bonded in air
directly after sputtering the multilayer (See Fig. 1). This bond
was formed primarily by vanderwaals forces and hydrogen
bonds between adsorbed water molecules and was too weak
(= 5 kg/cm2) to allow further lithographical processing. For
the spin-valve transistor small sizes are required for room
temperature operation. Also, the current gain of non-magnetic
metal base transistors was still quite low (5% at room
temperature), partly due to the interface states at the
imperfectly bonded emitter Schottky barrier which caused
insufficient electron injection properties. We now introduce a
new technique, namely vacuum metal deposition bonding.
For the preparation Prime quality n(100) Si wafers and
n( 11 1) Ge wafers are used. The preparation of the silicon is
according to the following scheme (class 100 cleanroom
condition): Dry oxidation (30 nm), photoresist protection,
sawing of the substrates into 12x21 mm parts, HNO?
cleaning, DEMI rinse, HF 2% dip to remove native oxide
from sawing particles, 7 min. TMAH (Tetra Methyl
Ammonium Hydroxide) etching of Si particles at 9OoC, HF
50% strip of 30 nm oxide, DEMI rinse and dry Nz blow dry.
The Ge is cleaned as follows: photoresist protection, sawing,
HF (50%):H202(33%) 1:lO for Ge fragment etch, HN03
clean, DEMI rinse, HF dip, DEMI rinse and again blow dry.
A Teflon vacuum pipette is used to manipulate the samples
without contaminating the surface. Following this procedure
the substrates (a) and (b) are attached to the rotatable arms of
a bond tool using clips (Fig. 5). The bond tool is made mainly
from aluminum and stainless steel. Before loading the system
into the sputtering chamber, a spring is loaded. Via magnetic
coupling beam (c) is lifted and the spring is released.
Substrates (a) and (b) then move towards each other during
sputtering of the metal layer. A flywheel and gears which are
connected to the rotatable arms dampen the action and in
about 0.5 seconds after launch the substrates come into
contact. One of the two substrate holders is attached to the
arm in a flexible way to permit parallel alignment during
contacting. In Fig. 6, as an example, a cross section TEM
picture of two silicon substrates bonded with a 30 nm Ti layer
(2x15nm) sputtered using a background pressure ~=5*10-~,
In general, it is very difficult to grow device quality
semiconductor material on top of a metal. To circumvent this
problem we presented direct bonding in air to obtain a Single
Fig. 5 Spring driven substrate rotator, substrates (a) and (b) and beam (c).

Page 4

3498
Fig. 6 TEM cross section of Si-Ti30nm-Si
an argon pressure p,=7*10-’
can be seen, large crystalline regions traversing the entire Ti
film are observed. The (po1y)crystallinity is also confirmed by
electron diffraction patterns. An interface cannot be
distinguished, apparently the Ti atoms are still very mobile
during the bonding time and recrystallisation takes place
during bonding. More details about the metallurgical
properties of the bond will be reported upon later [20].
The basic law governing the bonding between the
substrates is the reduction of the surface free energy. Stress
created by roughness compensation during matching of the
surfaces counteracts this drive, but for microroughnesses
smaller than about 2 nm complete bonding proceeds
spontaneously, i.e. without force at room temperature. The
bonding strength was larger than the fracture strength of the
silicon. The yield of the bonding procedure is surprisingly
good, 100% under standard conditions. For the development
of spin-valve transistors into smaller devices, grinding,
polishing, lithography and wet chemical etching has been
performed successfully. Vacuum bonded Si-Au-Ge metal
base transistor structures with different device areas are
shown in Fig 7. Following this, the preparation of small area
( 1 Opm) room temperature operating spin-valve transistors is
mbar and Power P=30W. As
Fig. 7 SEM picture of lithographically processed Si-Au-Ge metal base
transistor structures.
Fig. 8 TEM picture Si-Aul4nm-Ge metal base transistor structure showing
indiffusion of gold into Ge.
projected, using both Si, Ge and GaAs. Metal base transistor
structures without magnetic layers are currently under
investigation to determine the transistor behaviour apart from
the perpendicular spin-valve transport.
A cross section TEM picture of the Si-Au-Ge MBT is
shown in Fig. 8. It reveals again crystals traversing the
interface, but also diffused regions of gold into the
germanium, presumably aided by the high energy of the
sputtered adatoms and diffusive nature of gold. Nevertheless,
electrical measurements have shown good emitter and
collector Schottky I-V characteristics (although 0.7 eV barrier
height instead of 0.8 eV which is expected for Au-Si) and
very reproducible hot electron collector currents.
An example of a transistor characteristic of the Si-
Aul4nm-Ge (350x350 pm) device is shown in Fig. 9. Ge is
used as collector material to obtain barrier height difference
between emitter and collector. The current gain is still quite
small, as may be expected from the diffused regions at the
interfaces (Fig. 8), but the versatility of the bonding technique
is expected to allow rapid improvement. Especially lower
energy deposition methods of the metal, as in molecular beam
epitaxy, or thermal- and e-beam evaporation may show major
0.8
2
E
I
”
c
2 0.6
L
3
0
I
”
- -
0
U
0.4
0
0
0.2
0.0
-0.50
-0.40
-0.30 -0.20 -0.10 0.00
0.10 0.20
0.30 0.40
0.50
Collector Voltage Vbc (V)
Fig. 9 Common base characteristic of lithographically processed vacuum
bonded Si-Aul4nm-Ge metal base transistor structure. IC is plotted versus
Vk for I,=O-100 mA in 10 mA steps.

Page 5

3499
improvements of the interface sharpness and current gain
For the spin valve transistor reproducibility is more
important. In the realised metal base transisitors the transfer
characteristics varied less than 10% from device to device,
which promises good prospects for future investigations. As
expected, the hot electron collector current is independent of
device area (0.1mm2-lmm2, see also Fig. 7), whereas the
Schottky leak current is proportional to this area. Many
injectors and collectors may now be tried for the development
of high gain metal base transistors and spin-valve transistors:
tunnel injection using a Si-Si02 structure, bandgap tailored
GaAlAs injectors, high mobility semiconductors, shallow
implants and so on. The advantages of semiconductor and
semiconductor-oxide injectors as compared to metal-oxide-
metal (MOM) injectors are the high maximum current density
(up to lo5 A/cm2), large breakdown voltage and reproducible
IV characteristics.
CONCLUSIONS
In conclusion, the spin-valve transistor obtains important
properties as a spectroscopic measurement tool of the CPP-
GMR: the electron energy dependence of the spin-valve
effect can be investigated in an important uncovered electron
energy range. The collector current is a direct measure of the
perpendicular hot electron mean free paths h&,) and is
strongly dependent on a magnetic field. For read head
applications soft spin valve sandwiches with long electron
mean free paths or small thicknesses should be implemented
for large collector current variation in low fields.
Full strength bonding at room temperature has been
demonstrated by mating of silicon substrates during sputter
deposition of a metallic film. TEM pictures show a buried
metal film without an interface. This new bonding technique
allows preparation of small devices and with it room
temperature operating spin-valve
applications of the new bonding technique lie in the field of
micromechanics, integration of optoelectronic integrated
circuits (OEICs) e.g. 111-V and Si wafers, and bonding in
general which may take advantage of the extremely strong
bond with its chemically resistant, thermally and electrically
conducting and, using magnetic materials, magnetic flux
preserving link.
transistors. Other
REFERENCES
[I] M.N. Baibich, J.M. Broto, A.Fert, F. Nguyen Van Dau, F. Petroff, P.
Etienne, G. Creuzet, A. Friederich, and
magnetoresistance of (001)Fe/(001)Cr magnetic superlattices” Phys. Rev.
Lett. vol. 61, pp. 2472-247.5, 1988.
[2] B. Dieny, “Giant magnetoresistance in spin-valvc multilayers,” ./. Mugn
Mugn, Muter, vol. 136, pp. 335-359, 1994.
[3] W.P. Pratt, Jr., S.F. Lee, J.M. Slaughter, R. Loloee, P.A. Schroeder, and
J. Bass, “Perpendicular giant magnetoresistance of Ag/Co multilayers,”
Phys. Rev. Lett. vol. 66, pp. 3060-3066, 1991.
[4] M.A.M. Gijs, S.K.J. Lenczowski, and J.B. Giesbers, “Perpendicular giant
magnetoresistance of microstructured FelCr magnetic multilayers from 4.2-
300K,” Phys. Rev. Lett, vol. 70, pp. 3343-3346, 1993.
[SI L. Piraux, J.M. George, J.F. Despres, C. Leroy, E. Ferain, R. Legras, K
Ounadjela and A. Fert, “Giant magnetoresistance in magnetic multilayered
nanowires,” Appl. Phys. Lett, vol. 65, pp. 2484-2486, 1994.
[6] A. Blondel, J.P. Meier, B. Doudin and J.P. Ansermet,” Giant
magnetoresistance of nanowires of multilayers,” Appl. Phys. Lett., vol. 65,
pp. 3019-21, 1994.
J.
Chazelas, “Giant
[7] Shinjo, T., Ono, T., and Yamamoto, H., Colloyium digest of the 14rh
internutionul Colloyium on magnetic ,films und .surfucc~s, Diisseldorf:
Heinrich Heine Universitat, 1994, pp.245-247.
[8] M.A.M. Gijs, M.T. Johnson, A. Reinders, P.E. Huisman, R.J.M.
Vandeveerdonk, S.K.J. Lenczowski
Perpendicular giant magnetoresistance of Co/Cu deposited under an angle
on grooved substrates,” Appl.Phys.Lett. vol. 66, pp. 1839-1841. 1995.
[9] M.A.M. Gijs and G.E.W.
Magnetoresistance of Magnetic Multilayers ,” (Review) , in press.
[IO] D.J.Monsma, J.C.Lodder, Th.J.A. Popma and B.Dieny, “Perpendicular
hot electron spin-valve effect in a new magnetic field sensor: The spin valve
Transistor,” Phys. Rev. Letters, vol. 74, pp. 5260-5263, 1995.
[I I] U. Feder Polorized Electrons in Su#uce Physics, Advanced Sei-ies in
Sufuce Science, vol. I, World Scientific, Singapore, 1985.
[ 121 R.J. Celotta, J. Unguris, and D.T. Pierce, “Hot electron spin-valve effect
in coupled magnetic layers,” J. Appl. Phys. vol. 75, pp. 6452-6455, 1994.
[I31 H.J.Drouhin, A.J. van der Sluijs, Y.Lassaily, and G. Lampel, “Spin
dependent transmission of free electrons through ultrathin cobalt layers,” J.
A[~pl. Phys. vol. 79, pp. 4734-4739, 1996.
[ 141 J.C.Grohli, D.Oberli, and F. Meier, “Crucial tests of spin filtering,”
Phys. Rev. B, vol. 52, pp. U13095-Rl3097, 1995,
[ 151 J.S. Moodera and L. R. Kinder, “Ferromagnet-insulator-ferromagnet
tunneling: spin-dependent tunneling and large magnetoresistance in trilayer
junctions,” J. Appl. Phys., vo1.79, pp. 4724-4729, 1996,
[ 161 H.C. Siegmann, “Surface and 2D magnetism with spin-polarized
cascade electrons,” Surface Science, vol. 307-309, pp 1076- 1086, 1994,
[I71 Tillmann D., “Very low energy spleed from Fe(OOI),” 2.Phys.B-
Condensed mutter, vol. 77, pp. 1-5, 1989,
[ 1 81 E.Y.Tsymba1 and D.G.Pettifor, “The influence of spin-independent
disorder on giant magnetoresistance.” J. Phy.s.: Condens. Mutrey, vol. 8, pp,
L569-LS75, 1996.
and
R.M.J. Vangansewinkel,”
Bauer, “Perpendicular Giant
[I91 U. Gosele, H. Stenzel, M. Reiche, T. Martini, H. Steinkirchner and
Q.Y. Tong, “History and future of semiconductor wafer bonding,” Solid
Sure Phenomenu vol. 47-48, pp. 33-44, 1996 and references therein.
[20] T. Shimatsu, D.J. Monsma, R. Mollema and J.C. Loddrr, unpublished.