MgO Magnetic Tunnel Junction sensors in Full Wheatstone Bridge configuration for in-chip current field detection Raquel de Jesus Gandum Rato Gon¸ calves Flores Disserta¸ c˜ ao para a obten¸c˜ ao de Grau de Mestre em Engenharia F´ ısica Tecnol´ ogica J´ uri Presidente: Professor Jo˜ao Carlos Carvalho de S´a Seixas Orientador: Professora Susana Isabel Pinheiro Cardoso de Freitas Vogal: Professor Paulo Jorge Peixeiro de Freitas Outubro de 2010
95
Embed
MgO Magnetic Tunnel Junction sensors in Full Wheatstone Bridge ...
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
MgO Magnetic Tunnel Junction sensors in Full
Wheatstone Bridge configuration for in-chip current
field detection
Raquel de Jesus Gandum Rato Goncalves Flores
Dissertacao para a obtencao de Grau de Mestre em
Engenharia Fısica Tecnologica
Juri
Presidente: Professor Joao Carlos Carvalho de Sa Seixas
Orientador: Professora Susana Isabel Pinheiro Cardoso de Freitas
Vogal: Professor Paulo Jorge Peixeiro de Freitas
Outubro de 2010
Agradecimentos
Gostaria de agradecer em primeiro lugar a Professora Susana Cardoso, por me ter orientado e ajudado
durante este projecto, e ao Professor Paulo Freitas por me ter dado a oportunidade de trabalhar no INESC-
MN, com a sala limpa e todos os seus recursos ao meu dispor para este trabalho. Gostaria ainda de agradecer
ao Professor Candid Reig, do Departamento de Engenharia Electronica, da Universidade de Valencia, pela
sua inteira disponibilidade para me ajudar ao longo do projecto e ainda pela preciosa ajuda na realizacao e
discussao das medidas efectuadas em Valencia.
Um grande obrigado a todos os colegas e tecnicos do INESC-MN pela disponibilidade e ajudas prestadas
sempre que foi necessario, sem os quais nao teria sido possıvel chegar onde cheguei. Obrigado ainda a todos
aqueles que me proporcionaram um local de trabalho com muita alegria, boa disposicao e simpatia.
Agradecco aos meus colegas e amigos de mestrado pelo apoio e companheirismo e ainda a todos os meus
amigos que estiveram sempre presentes para dar animo e alegria a minha vida, em especial ao Ricardo por
ter estado la sempre que precisei.
Finalmente, ficarei para sempre agradecida e em dıvida para com a minha famılia pelo apoio incondicional
que me foi prestado em todos os aspectos que um curso acarreta para a vida. Um obrigada profundo as pessoas
mais importantes da minha vida: a minha mae Maria do Ceu, aos meus irmaos Manuel e Andre, a minha tia
Juquinha, a minha avo Catarina e a minha prima Monica. Agradeco ainda a famılia Raquel por todo o seu
apoio e carinho. E ainda ao Ze Maria pela sua companhia, que apesar de nao ser humano tambem faz parte
da famılia.
Obrigado ao meu avo Rato que me ensinou a ser uma pessoa melhor. Todos sentimos a sua falta!
iii
Resumo
Hoje em dia, e mais do que nunca, no que respeita a circuitos electronicos, a precisao e o controlo tornam-se
extremamente importantes, especialmente quando se trata de dispositivos relacionados com microeletronica,
onde medidas de corrente electrica, potencia e energia tem sido alvo de preocupacao. Por isso, dispositivos
precisos, facilmente integraveis, de baixo consumo energetico e baixo custo sao exigidos e sao essencialmente
necessarios em circuitos integrados, sistemas micro-eletromecanicos (MEMS), entre outros.
Juncoes de efeito de tunel magneticas (MTJ) tem sido utilizados como sensores de correntes electricas
em circuitos integrados. Alem disso, estes sensores tambem tem sido utilizados em configuracao de ponte de
Wheatstone para medicoes de baixas correntes.
Assim sendo, neste trabalho sao apresentados sensores MTJ em ponte de Wheatstone para deteccao de
corrente electrica em circuitos integrados com uma nova configuracao: cada elemento resistivo da ponte
consiste num conjunto de sensores MTJ ligados em serie, em vez de um unico elemento. O objetivo e medir
o campo criado por uma corrente electrica, aplicada em pistas incorporadas no chip durante o processo de
microfabricacao, reduzindo a separacao entre a fonte do campo e os elementos sensitivos, levando a uma
maior sensibilidade. Para obter uma ponte de Wheatstone completa e balancada, as pistas de corrente tem
que ser apropriadamente desenhadas.
Utilizando series de 18 MTJ elementos, alimentados por uma corrente de 100μA, obteve-se uma ponte
de Wheastotne com uma sensibilidade de 0.267 mVmAVb
= 1.334 mVOeVb
, e com um regiao linear de 40Oe, onde a
resistencia de entrada obtida para este dispositivo foi 544.96Ω. A tensao de “offset” obtida foi de −1.27mV .
Obteve-se uma R × A para a juncao de MgO de 1.82kΩ · μm2. Quando comparados com os dispositivos ja
existentes anteriormente, este novo tipo de pontes mantem o seu funcionamento ate tensoes de alimentacao
de 40V.
Palavras-chave: Sensores magnetoresistivos, Sensores MTJ de MgO, Pontes de Wheatstone, Sensores
de corrente
v
vi
Abstract
Nowadays and more than ever, control and precision regarding electronic circuits becomes very important,
especially concerning microelectronic devices where electrical current, power and energy measurements have
been a matter of concern. New precise, integrable, low power consumption and low cost devices are demanded
and they are mainly needed in integrated circuits (IC), systems-on-chip (SOC), micro-electromechanical
systems (MEMS), among others.
Magnetic tunnel junctions (MTJ) have been currently used as sensors for electrical currents measurements
in IC. In addition, these sensors have also been used in Wheatstone bridge configuration for low current
measurements.
So, Full Wheatstone bridges magnetic-tunnel-junction-based sensors for electrical current sensing at the
IC level are presented, with a new configuration: the resistive elements of the Wheatstone bridges consist
in arrays of MgO MTJ’s elements connected in series, instead of bridges with one single MTJ has resistive
element. The goal is to measure the field created by an electrical current, driven through paths incorporated
in the chip during the microfabrication process, reducing the separation to the sensing elements, leading to
improved sensitivity. And in order to get a full balanced Wheatstone bridge, the current paths need to be
properly designed.
Using series of 18 MTJs elements, biased with 100μA, it was obtained a bridge presenting a sensitivity of
0.267 mVmAVb
= 1.334 mVOeVb
, with linear behavior in the range of 40 Oe, where the bridge’s input resistance was
544.96Ω. The offset voltage of the transfer curve was −1.27mV and the R×A of the junction was determined
as 1.82kΩ · μm2. As a great improvement comparing to the previously existing devices, these new type of
bridges can hold up to voltages of 40V without breaking down.
Keywords: Tunneling magnetoresistance (TMR) sensors, MgO MTJ, Wheatstone bridge, Current
Module 4 DC Power Pressure Gas Flow Deposition Rate
Al 2.0 kW 3.0 mTorr 50 sccm Ar ∼37.5 A/s
The four modules have different purposes, which are:
1. Module one consists in an array of lamps and it’s used for flash annealing.
21
2. Module two has a magnetron, which allows to use it to soft sputter etching processes. The soft etch is
used before the deposition of T iW (N) and AlSi1%Cu0.5%, in order to remove the oxide layer that is
naturally formed in metallic layers.
3. Module three is used for reactive sputtering deposition of T iW (N), where the material target is of
T iW and the N is obtained from the plasma, which is made of Ar and N2. T iW (N) is a material that
provides physical damage’s protection to the micro-structures and also protects Aluminum layers from
chemical substances. Moreover, it works as a anti-reflective coating in the optical lithography.
4. Module four is for AlSi1%Cu0.5% sputter deposition. This metal alloy is used as tunnel junction’s top
contact and as current paths, usually 3000A thicks.
3.1.3 UHV II
Figure 3.4: UHVII: front view picture (left) and side view schematic (right).
UHV II is a manual sputtering system used for oxide deposition, built in INESC-MN and installed
in a class 10000 clean room. This tool consists in a single deposition chamber, with direct access to the
atmosphere, that needs to be vented each time a deposition is made to place and remove the samples.
Therefore, in order to reach a base pressure of 3 × 10−7 Torr it is required a period of ∼ 10 hours. The
oxide is deposited from a Al2O3 ceramic target placed facing down under a φ6′′ magnetron. An Ar plasma
is created with a RF power supply. The samples are placed facing up in a cooled φ6′′ table under the target.
The deposition rate is maintained constant by balancing the gas inlet (constant flow of 45 sccm) and the
pumping speed (the turbo pump is set to operate in cruise speed). The oxide films deposited by this tool
were used as insulator layers between two metal layers during the tunnel junction fabrication.
Table 3.3: Deposition conditions for UHVII
UHVII RF Pressure Gas Flow Deposition Rate
AL2O3 200W 3.0 mTorr 45 sccm Ar 11.43A/min
3.1.4 Alcatel SCM450
Alcatel SCM 450 is a commercial sputtering tool, with only one chamber. This chamber has three φ4′′
magnetrons in the bottom under the three targets facing up, and four substrate holders on the top facing
22
down. The substrate table can be static during a deposition or can be set to rotate at a speed of 4 rpm.
The substrate holders can be connected to a RF power supply to perform sputter etch and the magnetrons
can be supplied with either RF or DC power. As the UHV II system, Alcatel doesn’t have a loadlock which
means that 10 hours are required to reach a base pressure of 10−7 Torr.
Figure 3.5: Alcatel picture and chamber schematic view.
This facility is installed in a class 10000 clean room and was used for SiO2 deposition as an insulator
layer in the tunnel junction fabrication.
Table 3.4: SiO2 deposition setpoints in Alcatel SCM450 system
Alcatel RF Power Ar Flow Deposition Pressure Target rotation Deposition rate
SiO2 140W 20 sccm 3.0 mTorr 4 rpm 22.2 A/min
3.2 Ion Beam Deposition Systems
A Ion Beam Deposition (IBD) systems can be used for deposition of thin films and also for ion beam
milling, which consists in a non selective dry etch process. In IBD systems, a highly energetic ion beam is
used to remove material from a target that will deposit on a substrate a thin film. The ion beam is created by
an ion source, called deposition gun, within which is created a plasma by an RF power supply. The ions are
extracted from the gun by a set of three charged grids, that pull out the ions from the plasma and accelerate
them towards the vacuum chamber as a uniform collimated beam that will collide with the target. In this
process, the ions that hit the target are less energetic than in the sputter process, implying a lower deposition
rate [14], [17], [18].
The IBD system has also another ion source −the assist gun, where the beam points directly to the
substrate and can be used to ion milling process. Besides assist and deposition guns, there are two neutral-
izer guns inside the chamber, that emit electrons to neutralize the accumulated ions on the target and on
the insulating substrates.The assist and deposition guns, the substrate and the target are disposed in a Z
configuration.
The substrate table has a permanent magnet array mounted around it, producing a 40 Oe magnetic field
that defines the easy axis during the deposition. This table can be rotated depending on whether a deposition
or an etch is being made. The substrate holder also rotates during deposition and milling processes in order
23
Figure 3.6: Schematic view of the Z configuration of the IBD systems.
to achieve a better uniformity throughout the sample. The target holder has a hexagonal prism shape with
one different target in each face, and rotates according to the material that will be deposited, letting a specific
target exposed to the ion beam, while the others targets remain protect from contamination by a shutter.
There is also another shutter protecting the sample until assist and deposition guns have the ion beams
stable, accordingly to the set parameters. The chamber vacuum is obtained with a turbo-molecular pump
and a cryogenic pump, achieving a working base pressure of 10−8Torr.
There are two IBD systems in INESC, the Nordiko 3600 and the Nordiko 3000, both installed in a class
100 cleanroom. They are both very similar to each other, where the main difference is that N3600 is a much
larger system being capable to produce φ8′′ substrates, instead of the φ6′′ provided by N3000. During this
work both facilities were only used as ion milling tools.
Table 3.5: IBD systems set values for the assist gun
N3000 RF V+ V- I (Beam Current) Gas Flow
Ion milling 58W 500V 200V 30mA 8 sccm Ar
N3600 RF V+ V- I (Beam Current) Gas Flow
Ion milling 160W 735V 350V 105 mA 10 sccm Ar
3.3 Reactive Ion Etch
3.3.1 LAM Research Rainbow Plasma Etcher
The LAM Research Rainbow Plasma Etcher is a facility used to perform reactive ion etching (RIE). When
compared to sputter etch or to ion milling, the major advantage of RIE is its higher selectivity because it
uses a chemically reactive plasma to remove the material. Like in the sputter etch process, in the RIE process
the sample is placed inside a grounded vacuum chamber, within which the plasma is created, and negatively
biased attracting towards the sample the plasma ions. The choice of the plasma species is made according to
the material to be etched, so that the ions react chemically with the materials on the surface of the samples.
24
For instance, sulfur hexafluoride is a common choice for silicon etch. Due to the kinetic energy of ions, they
also extract some material from the sample by transfer of moment to the sample, as in the sputter etch
process. The etching rate is much higher in RIE than in sputter etch or milling.
Figure 3.7: LAM front view picture.
In this work LAM was used for SiO2 etching, since it was deposited on top of aluminum stripes and only
then etched opening paths to elements pads, thereby protecting the metal from the chemical compounds used
in the process.
Table 3.6: Setpoints for SiO2 etching in the LAM tool
LAM RF Power Gas Flow Deposition Pressure Electrodes Gap He Clamp Etching rate
SiO2 ecth 100W200 sccm Ar
140 mTorr 1.3 cm 14 Torr ∼8.3 A/s100 sccm CF4
3.4 Pattern Transfer and Lithography: Direct Write Laser Optical
Lithography (DWL)
During the tunnel junction fabrication process, it’s necessary to pattern micron sized devices. To do so
it’s required to selectively deposit or remove material from a substrate, which is possible either by etching
or by lift-off. In both processes, a mask must be transfered to the substrate, which can be done by optical
lithography. This last process includes mask design, vapor prime, photo-resist (a photo sensitive polymer)
coating, lithography exposure, photo-resist develop and mask transfer from photo-resist to film by lift-off or
etching. The masks are designed using a CAD software and then converted to a set of binary files located
in the lithography system hard disk. These masks can be converted as inverted or non-inverted for etching
process or for lift-off process, respectively.
25
3.4.1 Vapor Prime
Vapor prime improves photo-resist adhesion on sample surface and it’s made before coating the sample.
Basically, the vapor prime system consists in an vacuum oven within which the samples are placed and
submitted to a cycle (program 0):
1. Wafer dehydration and purge oxygen from the chamber: Vacuum, 10 Torr, 2 minutes
N2 inlet, 760 Torr, 3 minutes (3 times)
2. Priming: Vacuum, 1 Torr, 3 minutes
hexamethyldisilizane (HDMS), 6 Torr, 5 minutes
3. Purge prime exhaust: Vacuum, 4 Torr, 1 minute
N2 inlet, 500 Torr, 2 minutes
Vacuum, 4 Torr, 2 minutes
3.4.2 Photo-resist coating and developing
Figure 3.8: SVG tracks.
After the vapor prime, samples need to be coated with the photo-resist. There are two types of photo-
resist: positive (used during this work) and negative. The positive photo-resist is made of a resin and a
photoactive compound dissolved in a solvent that is also a dissolution inhibitor. This type of photo-resist has
a photo-reactive component which becomes unstable when exposed to light with a determined wavelength,
allowing to dissolve it by developing.
So, in the coating track, the samples are coated with 1.5μm positive photo-resist (PFR7790G27cP) and
after lithographic laser exposure, developed in the developer-track with Ethyl lactate (Pth70eg). The coating
and developing procedures are:
1. Coating - Recipe 6/2 :
• Dispense of photo-resist at 800rpm for 5s
• Spin at 2.8 krpm for 40s, then at 1.6 krpm for 5s
• Clean photo-resist from the wafer border at 1krpm for 2s
• Spin at 1.5 krpm for 15s
26
• Bake at 110◦C for 60s
2. Developing - Recipe 6/2 :
• Bake at 90◦C for 60s
• Cooling for 30s
• Water spray and rinse at 500rpm for 1s
• Dispense of developer at 500rpm for 5s
• Development for 60 s with the wafer stopped
• Rinse with DI water at 1 krpm for 20s
• Drying with wafer rotating at 3.5 krpm for 30s
3.4.3 Optical Lithography Exposure
Figure 3.9: Photo-resist mask designed by optical lithography.
The optical lithography exposure is performed with a direct-write laser system DWL 2.0 by Lasarray, using
a 442nm Helium-Cadmium Laser of 120mW. This power can be adjusted for each exposure with respect to
the reflectivity of the substrate material. The write lens is focused on the sample surface, by an air pressure
auto-focus system.
Figure 3.10: Picture of DWL system.
The samples are mounted on a mechanical x− y stage and fixed by vacuum, and only then aligned. For
alignment a dual CCD camera system (macro/micro) is used, with a measurement accuracy of 70 nm. The
sample is then exposed, meaning that the laser sweeps the sample according to the mask designed in the
27
AutoCAD software. Laser scans samples by stripes with a width of 200 μm. Each stripe corresponds to
several scans performed by successively writing pixels from left to right (pixel grid pitch = 200 nm), in the
x direction. After finishing one scan, the sample is moved one step in the y-direction and the next scan is
done. At the end of a stripe exposure, the stage moves to the origin in the y-direction and 200μm to the right
in the x-direction, starting a next stripe. The minimum feature achieved by this system is 0.8μm. DWL is
placed in a class 10 clean-room.
3.4.4 Pattern transfer processes
Lift-off
In a lift-off process, the material to be patterned is deposited on top of the patterned photo-resist. Then,
in a wet bench in the grey room the sample is immersed into a microstrip solution, which acts as a photo-resist
solvent when heated at 65◦C (resist strip). As the microstrip reaches the photo-resist under the deposited
material, this material is also lifted-off. To improve the process, samples are submitted to ultrasounds,
facilitating the microstrip penetration until the photo-resist. After the stripping, the sample is rinsed with
IPA (isopropilic alcohol), and then with DI water. The sample is finally blow dried with an N2 gun.
Figure 3.11: Illustration of the lift-off process.
Etching
In this process, the material to be patterned is already deposited at the start. Next, it’s covered with
a patterned mask and then the unprotected material is removed. As in lift-off process, the substrate is
submitted to resist strip leaving only on the substrate the material covered previously by the mask. During
this work, the etching was carried out by using ion milling or by reactive ion etch processes.
Figure 3.12: Illustration of the etch process.
The reactive ion etch is done in LAM and it’s used during this work to etch SiO2. The ion milling etch is
performed either in N3000 and N3600 to etch metals, in order to define the junction pillar and the electrical
contacts.
28
3.5 Characterization
3.5.1 Magnetic Thermal Annealing
Before any transport and magnetic measurements, the tunnel junctions are submitted to a magnetic
thermal treatment, that promotes the material re-crystallization, because the crystalline structure of the
material can have some defects as-deposited. Besides, the annealing enhances the TMR, by re-enforcement
of the magnetic layers magnetization, optimizing the spin polarization at the barrier interfaces. Once the
annealing temperature for MTJs is higher than the blocking temperature of the antiferromagnetic layer,which
drops the exchange field to zero, it’s necessary to apply a magnetic field during the annealing to set the pinning
layer and improving its magnetization, which improves the TMR value.
In the annealing setup, the substrate is mounted on a copper block, heated by a halogen lamp (100W,
12V) placed inside the copper block. A high vacuum grease is used between the sample and the copper holder
to improve the thermal contact.
Figure 3.13: Annealing setup.
In the annealing process, the sample is submitted to a thermal cycle: first the temperature rises from the
room temperature to the annealing temperature, taking one hour; then for another hour, the temperature is
kept constant at 320◦C (annealing temperature); finally, the sample will cool down to room temperature, by
its own. During this cycle, the sample is kept inside a removable glass vacuum chamber, while an external
magnetic field of 4kOe is applied. This field is produced by two water-cooled magnets, where the distance
between those can be adjusted to achieve the desired magnetic field value. The maximum achieved field is
of 5 kOe for a certain gap between the magnets.
Figure 3.14: Temperature cycle during magnetic thermal annealing.
29
3.5.2 Vibrating Sample Magnetometer (VSM)
The VSM (DMS model 880) is a commercial system used to measure the magnetization of unpatterned
samples as a function of the applied magnetic field. This field is generated by two electromagnets and
can reach a maximum of 13kOe. The sample is mounted on a glass rod, that vibrates horizontally at a
given frequency along with the sample. This motion induces a variation of the magnetic flux, generating an
electromotive force in the coils around the sample. As this force is proportional to the magnetic moment of
the sample, this last can thus be determined. The VSM has a field resolution of 0.1 Oe and a sensitivity
down to 10−5emu.
Figure 3.15: DMS model 880 VSM picture.
Figure 3.16: Example of a VSM measurement for a bulk structure annealed at 240◦C.
3.5.3 Profilometer
The Dektak profilometer is a tool that allows to measure the topography of a sample through a piezo-
resistive sensor. This sensor sweeps the sample surface, within a certain range, detecting any topographic
changes such as pillars or holes in the film. This technique is very useful to measure thickness films, for
instance, oxide films, ensuring the thickness deposited is the desired one, also allowing to determine deposition
rates. To measure the oxide thickness it’s used a calibration sample, where some stripes of ink are made
before the deposition. After deposition, the ink is removed with acetone, which leaves steps between the oxide
30
film. These steps height can be easily determined by the profilometer. This tool has a vertical resolution of
5Aand has 400Aof thickness as lower detection limit .
Figure 3.17: Profilometer’s picture.
3.5.4 Ellipsometer
The ellipsometry is a technique used to determine the properties of thin films, such as thickness and
refractive index, where a collimated beam of monochromatic light is focused in a sample at a given angle -
the RUDOLPH EL ellipsometer used in this work uses a 632.8nm light wavelength, and before the sample,
the beam is linearly polarized.
Figure 3.18: Ellipsometer’s picture.
After being reflected in the sample and through refraction and reflection mechanisms, the beam has a
different state of polarization, which compared to the incident polarization state allows to determine two
angles: δ and Ψ. Through a numerical model based on Fresnel equations, given these two angle it’s possible
to infer the refractive index and the thickness of the film. This method requires that the dielectric is deposited
31
on top of a fully reflective material, such as silicon, Si. As the profilometer, this method allows to calibrate
the thickness of the deposited oxide (Al2O3 and SiO2) but further, it also helps to ensure the oxide target is
not contaminated through the determination of the refractive index, which is well known for each material:
1.62 for Al2O3 and 1.47 for SiO2.
32
Chapter 4
TMR based sensors: Wheatstone Bridge for
Electrical Current Sensing
TMR sensors applications can be classified into two major groups: magnetic storage systems (hard disks
read heads and magnetic random access memories) and magnetic field sensing. Focusing in the magnetic
field sensing, for sensors where a linear sensor output is required, TMR elements can be arranged in a four
elements bridge configuration due to inherent linearity and the null output in the absence of a magnetic field.
TMR based sensors are nowadays used in very different fields such as positioning control devices in robotic
and related systems, geomagnetism, biotechnology applications and electrical current measurements.
Specifically, this work is focused on the electrical current sensing by TMR based sensors in full Wheatstone
bridge configuration, where these sensors can be used as amp-meters in integrated circuits (ICs).
4.1 Wheatstone Bridge
Figure 4.1: Wheatstone bridge circuit.
A Wheatstone Bridge is an electrical circuit first described by Samuel Hunter Christie (1784-1865) in
1833, being however popularized by Sir Charles Wheatstone, whom invented many uses for this circuit once
he found the description in 1843. It is commonly used in electronic devices to measure an unknown electrical
resistance. The fundamental concept of the Wheatstone Bridge is two voltage dividers, both fed by the same
input, where the circuit output is taken from both voltage divider outputs, and is given by the following
expression:
33
Vo =
(R2
R1 +R2− R4
R3 +R4
)Vi (4.1)
In its classic form, a galvanometer (a very sensitive dc current meter) is connected between the output
terminals, and is used to monitor the current flowing from one voltage divider to the other. If the two voltage
dividers have exactly the same ratio R1/R3 = R2/R4, then the bridge is said to be balanced and no current
flows in either direction through the galvanometer, being the output null. If one of the resistors changes even
a little bit in value, the bridge will become unbalanced and current will flow through the galvanometer.
Even though a unique resistance can be used as sensing element, a Wheatstone bridge setup is always a
good recommendation as the starting step in the design of resistive sensors, since it provides a differential
output as a function of the resistance variation. In fact, using MTJs as resist elements in a Wheatstone
Bridge allows to have a linear magnetic field sensor with an offset-free signal.
Depending on the considered case or the particular requirements, there are several bridge configurations
that can be used [20], [21], [22].
4.1.1 Quarter Bridge Configuration
In this configuration, four resistances of identical nominal value are used, being only one of them active:
R1 = R2 = R3 = R and R4 = R+ΔR. In this case, the output voltage is given by:
Vo =ΔRR
2(2 + ΔR
R
)Vb (4.2)
The relationship of the bridge output with the variation of resistance is not linear. Besides, to obtain a
output value of the order of the ΔRR ratio, a high Vb is required. This configuration is only used when small
variations of ΔR occur.
4.1.2 Half Bridge Configuration
In the half bridge configuration, two opposite resistances in the bridge are active: R2 = R3 = R and
R1 = R4 = R+ΔR. The output voltage is:
Vo =ΔRR
2 + ΔRR
Vb (4.3)
As in the previous configuration the output is non-linear, although the output value is now the double of
the previous.
4.1.3 Full Bridge Configuration
In order to get a full bridge response, paired resistances need to be made active: R1 = R4 = R+ΔR and
R2 = R3 = R−ΔR, and the output is given by:
Vo =ΔR
RVb (4.4)
This configurations shows to be the preferable one, since the output is linear and it has the higher output
signal, when compared to previous configurations.
34
Figure 4.2: Different Wheatstone bridge configurations.
4.2 Application: Electrical current sensing
The electrical current can be measured by different means:
• Shunt resistances, based on Ohm’s law: V = R · I• Coils, based on Faraday’s law of induction,
• Solid state sensors, being the TMR based sensors used in this work.
In the third case, the magnetic field created by a electrical current flow is detected by the MTJs sensors,
where the electrical will be driven along a conductive strap, microfabricated on top of the MTJs sensors. To
calculate the magnetic field created by the current [26], the Biot-Savart law is taken into account:
H (r) =1
4π
∫V
J(r′)× r − r′
|r − r′|3 d3r′ (4.5)
The created magnetic field is being calculated in an arbitrary point r′ around the conducting line, where
this line has a rectangular cross section ω×h and the current flow is assumed to be uniform over the conductor,
so J = I/(ω × h), where:
• J(r′)
= (0, 0, J), meaning the current flows out of the plane, in the z-direction, with the x and y
components being zero.
• r − r′ is the vector that links the point at which the field is being calculated,r, and all the points,r′,where J is not null. This vector lays in the plane and can be described as (rx, ry, 0), where rx = x− x′
and ry = y − y′.
• |r − r′| is the distance between r and r′ and can be replaced by r =√
r2x + r2y + r2z .
• J(r′)× (r − r′) is therefore equal to (Jry,−Jrx, 0).
Finally, the field generated in the x-direction (that due to the design coincides with the easy axis of the
magnetic sensors) can be expressed as:
Hx =1
4π
∫V
Jryr3
dx′dy′dz′ (4.6)
35
where the integration limits are −ω/2 < x < ω/2, −h/2 < y < h/2 and −l/2 < z < l/2, and l is the total
length of the line. When for a current line the following can be considered: ω h, a simpler expression can
be obtained for the field:
Hx(x, d) =1
π
I
2ω
[arctan
(ω/2 + x
d
)+ arctan
(ω/2− x
d
)](4.7)
where d is the distance from the line surface and x is the position across the line. The calculated magnetic
field has units of A/m using SI system, and can be converted to Oe multiplying by a factor of 4π1000 . By way
of example, using equation 4.7, two simulations of the created field Hx created by a current flow, as function
of the x are presented in figures 4.3 and 5.2.1.
Figure 4.3: Magnetic field generated by an 160μm wide and 0.3μm thick current line.
Figure 4.4: Magnetic field generated by an 25μm wide and 0.3μm thick current line.
36
4.3 Noise and detectivity of an array of tunnel junctions
The minimum magnetic field magnitude that can be detected for a sensor (for a certain bandwidth and
applied field) is defined as the detectivity of the sensor, D (in T/√Hz), and is expressed by:
D =SV
ΔV/ΔH(4.8)
where SV (V/√Hz) is the output noise of the sensor and ΔV/ΔH is the sensor sensitivity, both measured
under an external magnetic field, and applying as bias voltage, V, in each junction of 5 to 10 mV, assuring
the maximum TMR ratio is obtained.
The voltage noise in a individual magnetic tunnel junction is given by:
S2V (f) = 2eIr2 coth
(eV
2kBT
)+
αV 2
A
1
f(4.9)
The first term represents the white noise and the second one the 1/f noise, being the white noise composed
by the thermal and the shot noises, and the 1/f term given by the Hooge model like expression, where e is
the electron charge, I the biasing current of the sensor, r the sensor resistance, V the voltage between the
sensor electrodes, kB the Boltzmann constant, T the temperature, α the Hooge like parameter, A the area
of the junction, and f the frequency.
Considering now a device with N individual MTJ sensors in series, the noise spectral density will be N
times the noise spectral density for a single junction:
S2V (f) = N
(2eIr2 coth
(eV
2kBT
)+
αV 2
A
1
f
)(4.10)
Also for the N elements, the device sensitivity is given by:
ΔV/ΔH =
(ΔR
RΔH
)RI =
(ΔR
RΔH
)NrI = γNrI (4.11)
Therefore, the detectivity squared for such devices is:
D2 =S2V
(ΔV/ΔH)2=
N
γ2(NrI)2
(2eIr2 coth
(eV
2kBT
)+
αV 2
A
1
f
)(4.12)
Assuming that V � kBT , then the detectivity equation becomes:
D =1√N
1
γ
√4rkBT
V 2+
α
A
1
f(4.13)
So trough equation 4.13, it’s established that for N MTJs elements connected in series, the noise density
increases with√N while the sensitivity increases with N . Therefore, the detectivity increases proportionally
to 1√N
[27]. Then it was for this reason that the series of MTJs were chosen as resistance elements: in order
to improve the Wheatstone bridges’ detectivity.
37
4.4 Device design
For this work a full bridge configuration was chosen. As resistance elements of the bridge are used MTJs
elements connected in series. In this section follows a detailed explanation of the design of the devices used
throughout this work and its fabrication process.
Bridge design
There are several mechanisms that can be used in order to get a full bridge behavior, i.e., for a certain
input the resistances need to be made active in pairs, as noted previously in 4.1.3. Since the resistance
elements are MTJ’s sensors, a full bridge behavior implies that for the same conditions two sensors be in
the maximum resistance state and the others two in the minimum resistance state, where maximum and
minimum resistances correspond to the anti-parallel and parallel states of free and pinned magnetic layers of
the structure.
There are two main approaches: manipulate the reference layer to set two different and symmetric ori-
entations in adjacent arms of the bridge or set locally different orientations of the free layer. Regarding the
first, the same direction of the reference layer (for both bridge’s arms) is defined by annealing, and only then
the pinning direction is reversed locally through a local current heating. The opposite pinning directions
can also be defined right at the annealing, by rotating half of the bridge of 90◦, although this implies that
electrical contacts between bridge’s arms need to be done externally to the chip.
Concerning the second option, the one chose for this work, the pinning direction of both arms is defined
simultaneously by annealing, and then locally, a magnetic field is applied in opposite directions, which rotates
the free layer locally. This can be done, for instance, by an electrical current flowing in different directions
above or below the sensors, meaning the device output can only be measured by applying an electrical current,
instead of an external magnetic field.
Figure 4.5: Current lines configurations.
So, in the chosen configuration, a current path is fabricated on top of sensors, where the current flows
above R1 and R4 in one direction an in opposite direction above R2 and R3. The magnetic field generated by
the current flow is parallel or anti-parallel to the pinned layer, respectively, so the free layer magnetization
rotates accordingly to the external field orientation. A current is driven through terminals A and B. This
way, R1 and R4 increase/decrease their values and R2 and R3 decrease/increase their values, thus obtaining
a full Wheatstone behavior. The sensor is fed through terminals a and b, and the output is taken between
terminals c and d. Moreover, two different configurations for the current path are used: series, where the
current flowing between A and B is the same that flows above the sensor, and parallel, where the current path
divides into two paths, so the value of the current above the sensors is half of the current inputed between A
and B.
38
4.5 Microfabrication
4.5.1 Stack Deposition
The first step in this microfabrication process is to deposit the layers of the MTJ structure’s stack using
N200 facility. The stack is deposited on top of a 1′′× substrate, usually Si coated with Al2O3 oxide. On top
of the MTJ stack it’s deposited 150Aof TiW(N) in the N7000.
Figure 4.6: MTJ’s stack deposited in N2000.
4.5.2 Bottom Electrode Definition
In this step the bottom electrode is defined. First, a photo-resist mask is deposited on top of the structure
Figure 4.7: Scheme of bottom electrode definition.
according to the 1st layer of the mask, defining the bottom electrodes. After the optical lithography is
performed and photo-resist is developed, the sample is etched at an angle of 70◦ by ion milling in N3000 or
N3600, removing the entire stack until the substrate, leaving pillars of 8μm×76μm (where the smallest length
coincides with the magnetic easy axis). Finally, the sample is submitted to lift-off, removing the photo-resist.
Figure 4.8: Picture of bottom electrodes.
39
4.5.3 Junction Pillar Definition
Figure 4.9: Junction pillar definition.
The junction pillars are defined by performing a 2nd lithography, followed by ion milling etch, defining
areas of 2μm × 30μm. In this etching, the material is removed till it reaches the CoFe layer, of the SAF
structure, i.e., the entire pinned layer needs to have the same geometry of the junction pillar, in order to get
the desired output. If in the etching the barrier isn’t overcame, the sensor will be shorted, and if the etch is to
excessive, the entire bottom electrode can be over-etched. Since this is a very critic step, a calibration sample
is used, having deposited the exactly same layers that have to be etched. To avoid material redeposition
around the oxide barrier, the sample is etched at 70◦ until the barrier, and at 40◦ from that one.
Figure 4.10: Picture of junction pillars.
4.5.4 Electrode Insulation
Now with the bottom electrodes and the junction pillars defined, an insulating layer of 800AAl2O3 is
deposited in the UHVII system. This layer aims to insulate the pillars and the bottom electrodes from each
others, guaranteeing that the electrical current will only flow through the barrier and perpendicular to it.
After the deposition follows the lift-off of the photo-resist and the oxide protecting the top of the pillars,
letting them uncovered so that the top electrodes can have electrical contact with the bottom electrodes.
Figure 4.11: Bottom electrodes and junction pillars insulation.
40
4.5.5 Top Electrode Definition
Figure 4.12: Top electrodes definition.
With the top of pillars uncovered, a 3rd lithography is made defining the junction top electrodes,
followed by the deposition in N7000 of 1500Aof Al, plus 150Aof T iW (N) as protective layer. Finally the
lift-off is done, leaving only the metal of the contacts. Top electrodes have an area of 16μm× 34μm.
Figure 4.13: Picture of top electrodes.
4.5.6 Sensors Insulation
As in this process, it’s necessary to fabricate current paths on top of the sensors, an insulation layer
between both is imperative, but leaving the sensors electrodes pads (420× 420μm2) uncovered by the oxide.
There are two ways of doing so: by lift-off or by RIE.
Figure 4.14: Insulating layer between sensors and current lines.
By lift-off, the samples are first submitted to the 4th lithography, after which 5000Aof oxide are deposited,
and only then submitted to lift-off, leaving the electrodes pads opened. The disadvantage of this method is
the fact that the chemicals used to develop the photo-resist have direct contact with the leads of Al, corroding
and oxidizing this metallic layer. However, this method allows to use Al2O3 instead of SiO2, once that Al2O3
is a better insulating layer being less porous and more compact.
41
In the alternative method, immediately after the top electrode definition, 5000Aof SiO2 are deposited in
the Alcatel system. Only then, the 4th lithography is made (using a non-inverted layer), protecting the entire
surface with photo-resist except for the contact pads. This way, when the RIE is performed in the LAM,
only the oxide on top of the pads will be etched. In the end, only the photo-resist lift-off has to be done.
The SiO2 oxide is chosen because it’s not possible to etch Al2O3 in the LAM tool. However, the aluminum
oxide can be etched in the IBD systems, but since it’s a very compact oxide, the etch rate is very small (less
than 1A/min), which would take many hours to etch 5000A. In LAM, the SiO2 etch rate is about ∼ 8A/s,
taking about 6 minutes to perform this step.
4.5.7 Current Lines Definition
The current lines are defined by a 5th lithography, after which 3000Aof Al and 150Aof TiW(N) are
deposited in the N7000. After deposition, the unwanted metal is lifted-off, leaving only the current paths of
aluminum well defined.
Figure 4.15: Current paths definition: scheme (left) and picture (right).
4.5.8 Final Passivation
The goal of this step is to protect the current lines from oxidation and physical damage depositing an
oxide layer, leaving open pathways to the pads of the current lines and to the pads of the sensors electrodes.
As in the sensors insulation step, this final passivation can be done by the same two methods, where the
oxide thickness deposited is 1000A.
4.5.9 Annealing
The magnetic tunnel junctions need to be annealed under an applied magnetic field, at an annealing
temperature of 320◦C during one hour, in order to promote the re-crystallization in the as-deposited material
and re-enforcing the magnetization of the magnetic layers.
4.5.10 Encapsulation
Finally, samples are diced according to its several dies and each one is mounted onto chip carriers sepa-
rately. Then, the samples are wirebonded and get completely ready to be characterized.
42
Figure 4.16: Picture of a die mounted on a chip carrier and wirebonded.
4.6 Characterization
The characterization and measurements techniques are a very important part when dealing with MTJ
sensors. For this work in question, the bridge behavior must be characterized by the use of an electrical
current driven through the paths, which gives a I-V curve (V-output voltage and I-current from lines), where
the V-I curves can be get for DC or for AC electrical current measurements. Furthermore, DC measurements
can be made at room temperature or inside a climatic chamber, allowing to study the bridge’s parameters
dependence with temperature. AC measurements aim is to measure bridge’s output as function of the
frequency of the electrical current in the paths. Concerning each bridge element and the test structures,
these can be characterized by applying an external magnetic field, which allows to obtain their MR curves.
4.6.1 MR Curves
The Manual Transport Measurement Setup (MMS) is used to measure transfer curves of the patterned
devices, namely the MR transfer curves of the MTJ elements (resistance as function of applied magnetic field),
providing direct information about the magnetic configuration of the MTJ. The transfer curve expresses the
magnetic configuration of the structure which minimizes the energy at each field point. The curve shape
depends on the coercive field (HC) and Neel coupling field, HF of the free layer, and also on the demagnetizing
fields and interlayer fields affecting the free layer which are related to the area and aspect ratio of the device.
Figure 4.17: Manual transport measurement setup picture.
This setup uses TiW needles in micropositioning probes, to make electrical contact with the devices pads.
These needles have a resolution of 10μm, which allows to establish contact with small area pads. To obtain
a MR transfer curve, an external magnetic field created by two Helmholtz coils is applied along the sample
43
easy axis, in the range of ±140Oe, while the MTJ are fed by certain bias current, so that each MTJ’s element
has a biased voltage of 5mV, meaning the value of the bias current depends on the number of elements, and
on the resistance of elements. The current source (KEITHLEY 220 Programmable current source), which
supplies the electrical current to bias the MTJ’s elements is connected to bottom and top contacts, as well
as the voltmeter (KEITHLEY 182 Sensitive Digital Voltmeter) that will read the output. The coils that
create the magnetic field are fed by a second current source. The control and read of all of this instruments
is made through a GPIB bus connected to a computer by USB interface. In order to protect the sensors from
the charge accumulations at the instruments’ terminals (which can yield to a current discharge, disrupting
the magnetic materials), two switches are connected to the current source and the voltmeter, so the circuit
remains shorted while the contact between probes and pads is being established, and only after the circuit is
unshorted, letting the current flow through the MTJ.
Regarding individual MTJ’s, the measurements are mainly done with a four probe arrangement, but for
MTJ’s connected in series, due to their design (only one input and one output), the measurements need to be
made with only two probes, which means that the current source and the voltmeter terminals are connected
to the same pads.
In a two probe configuration, we have the follow equivalent electric circuit:
Figure 4.18: Electric circuit for two probe measurements.
V = (2Rcontact +RMTJ)× Ibias (4.14)
So the TMR magnitude will be:
TMR =(V/I)Max − (V/I)min
(V/I)min× 100 (4.15)
TMR =2Rcontact +RMax
MTJ − 2Rcontact −RMinMTJ
2Rcontact +RMTJMin
× 100 (4.16)
TMR =RMax
MTJ −RMinMTJ
2Rcontact +RMinMTJ
× 100 (4.17)
By equation 4.17, it can be seen that in the two probe measurement the TMR obtained is not the real
TMR value, but smaller due to the term of the contact’s resistance.
In the four probe configuration, since the current source and the voltmeter are not connected to same
pads, the equivalent electrical circuit is different from the two probes one. Analyzing the circuit, it can be
assumed that there is no current flowing through the voltmeter, since the input resistance of the voltmeter
44
is much higher than the resistance of the MTJ’s elements. So the current flowing through the MTJs is IBias
and the voltage measured is only the MTJ’s voltage. Thereby, this configuration allows to measure the MTJ’s
resistance, providing a real measurement of the TMR.
V
I=
VMTJ
IMTJ= RMTJ (4.18)
The acquisition software was designed at INESC-MN in Visual Basic language. It automatically controls
each instrument, applying the desired bias current, sweeping the field between ± 140Oe twice (hysteresis
loop), and reading the voltage for each point, presenting at the end of the measurement a curve of resistance
as function of the applied field. The software allows to choose the range of the applied field, and also the bias
current. The desired bias current can be directly inputed or the user can choose the desired voltage output
(at zero field) and through three measurements, the software extrapolates the bias current corresponding to
the desired voltage.
4.6.2 I-V curves
To measure the DC I-V curves of the bridges and test structures is used a setup pretty similar with
the manual transport measurement setup, except a few modifications, since it’s now required an additional
current source instead of an external magnetic field source. This setup uses a current source (KEITHLEY
220 Programmable current source) to fed the MTJ’s elements of the bridge and a voltmeter (KEITHLEY 182
Sensitive Digital Voltmeter) to read the bridge output. If this measure is performed with the micropositioning
probes, it requires 6 of them, which becomes very difficult in question of space arrangement. So instead, the
V-I curves are measured with the device encapsulated and wirebonded, and placed in a appropriated adapter,
that allows to established the electrical connections through coaxial cables.
Figure 4.19: I-V measurement setup picture.
The electrical current, that flows above the sensors creating the magnetic field is provided by a second
and additional current source, also an KEITHLEY 220 Programmable current source. As seen previously in
section 4.4, the biasing current source is connected to a and b contacts, the voltmeter between c and d, and
the field current source to A and B, as shown in figure 4.20.
The acquisition software was specifically developed to this work in Visual Basic language, from the version
of the MMS software. The communication between the program and the instruments is also done through a
GPIB bus with an USB interface. In this setup, the bias current value can be chosen as in the MMS software,
and it also allows to choose the desired range of the current (the one that creates the magnetic field). So the
45
current will be sweep within the desired range, while the voltage output is read for each point, and in the
end it’s presented the V-I graphic.
Figure 4.20: I-V curves measurement: equivalent electric circuit (left) and picture with electrical scheme(right).
4.6.3 AC Characterization
The AC characterization setup is used to apply an AC current in the current lines, creating an AC
magnetic field, and to measure the voltage output of the bridges as a function of the frequency. This kind
of measurements can be done in two different experimental conditions: with a biasing current feeding the
bridge, allowing to measure the bridge AC transfer curve; or without biasing, in order to measure coupling
effects. The AC characterization setup includes a signal source (HP 33120A), a power supply (GW Instek,
GPC-3030DQ), an oscilloscope (Tektronix TDS3034), a current probe (Tektronix TCP202), and a differential
probe (Tektronix ADA400A).
Figure 4.21: Experimental setup for AC characterization.
The Wheatstone bridge is fed by a DC voltage source while the output is taken by the differential probe,
between the pins c and d. This differential probe is connected to an oscilloscope, with a detection limit of
2-3 VRMS , allowing to see and determined the RMS value of the output. The field’s current is an sinusoidal
current supplied by the HP source, with a frequency’s range until 10MHz, and where the peak value can
be inputed. To guarantee that the applied current has the correct shape and value, it’s used a current
probe connected to the oscilloscope. So keeping the bias value (in the case of the biased measurement), the
frequency is swept between 100Hz and 10MHz, while the RMS value of the output voltage is taken for each
frequency point.
46
Figure 4.22: Electric scheme of the AC measurements.
4.6.4 Thermal Characterization
The thermal characterization of a sensor is very important, specially when electrical current sensors are
used, which usually implies a high and changing environmental temperature. Besides the self heating effects,
inherent to the flow of a large electrical current, need to be quantified, in particular, when the current lines
are incorporated in the chip. The thermal characterization setup includes a signal source (Keithley 220), a
power supply (Agilent E3631A), a voltmeter (Agilent E34401A), a resistance thermometer (PT100) and a