• Non ci sono risultati.

Development of a Lab On Chip platform for the early detection of neurodegenerative diseases

N/A
N/A
Protected

Academic year: 2022

Condividi "Development of a Lab On Chip platform for the early detection of neurodegenerative diseases"

Copied!
106
0
0

Testo completo

(1)

Philosophiae Doctor Thesis in

Physics & Nanoscience XXXII Cycle

Development of a Lab On Chip platform for the early detection of neurodegenerative diseases

By

Fausto Sirsi

Presented in Partial Fulfilment of the Requirements for the Degree of Doctor of Philosophy

Supervisor: Coordinator

Prof. Giuseppe Maruccio Prof. Rosaria Rinaldi

Co-supervisor:

Prof. Günter Reiss

Dott. Alessandro Paolo Bramanti

A.A. 2019-2020

(2)

Table of Contents

... 1

Introduction ... 4

Chapter 1: Spintronics theory and applications ... 6

1.1 Magnetoresistive structures theoretical overview ... 7

1.1.1 Giant Magneto Resistance ... 7

1.1.2 Tunnel Magneto Resistance ... 11

1.1.3 Planar Hall Effect ... 14

1.2 Applications ... 17

1.2.1 Data storage ... 17

1.2.2 Spin transistors ... 20

1.2.3 Biosensors ... 21

Chapter 2: Fabrication and characterization of Magneto-Resistive structures ... 30

2.1 Thin films deposition by sputtering: an overview ... 30

2.1 Magnetoresistive structures fabrication ... 32

2.2.1 Stack deposition and magnetic configuration settlement ... 32

2.2.2 Samples patterning ... 37

2.3 Magnetoresistive structures characterization ... 40

2.3.1 Samples characterization ... 41

2.1.4 Approaching particles’ sensing ... 47

Chapter 3: Magnetic particles’ sensing: a novel dynamic approach ... 50

2.2 Particles’ motion simulation ... 51

3.3 Magnetic particles’ sensing ... 59

(3)

3

3.3.1 Dropping test ... 59

3.3.2 Real time particles’ motion monitoring ... 63

Chapter 4: Sensing through Electrochemical Impedance Spectroscopy ... 65

4.1 Electrochemical Impedance Spectroscopy: an overview ... 65

4.2 EIS platform development ... 68

4.3 Tau protein sensing ... 70

4.4 Listeria monocytogenes sensing ... 74

Chapter 5: Readout system development ... 79

5.1 Printed circuit board design ... 80

5.2 Readout system automation ... 84

5.2.1 System description ... 84

5.2.2 The core: STM32 Nucleo L152RE ... 87

5.2.3 The Multiplexer: Maxim integrated MAX14661 ... 88

5.2.4 Building and first tests ... 89

5.2.5 System integration and signal shielding ... 92

5.2.6 Excitation and readout ... 93

Conclusions and future works ... 97

Bibliography ... 99

(4)

Introduction

The present thesis is mainly focused on the development, fabrication and characterization of spintronic devices, where the spin degree of freedom is also evaluated with the charge transport properties, to be applied for the dynamic sensing of magnetic nanoparticles and the realization of an electronic smart interface useful for interfacing Lab On Chip platforms with the measurement setup, not only for such a device but also for the characterization of sensing circuits based on different technologies.

Carried out om synergy with the European project MADIA (MAgnetic DIagnostic Assay for neurodegenerative diseases), the PhD research activities were focused in the field of magnetoresistive sensors, especially based on Giant Magneto-Resistance (GMR), Tunnel Magneto-Resistance (TMR) and Planar Hall Effect (PHE).The first three chapters are related to this kind of activities.

Chapter 1 is devoted to an overview about the theoretical standard treatment of the above- mentioned phenomena, exploiting a semi-classical and a quantum mechanics approach, until deriving the relationship between the structures’ response and the external magnetic field. Then, the possible applications of these technologies are discussed, in the field of data storage, development of spin transistor and, more in detail, the employment as sensing elements for biosensors and Lab On Chip platforms.

Chapter 2 concerns fabrication and characterization of magnetoresistive structures.

Specifically, the first part regards the activities performed during the 6 months of the secondment period, carried out at the Physics Department, Center for Spinelectronic Materials and Devices - Bielefeld University (Germany), in prof. G. Reiss’s group which is among the world leaders in the spintronics field and a partner of MADIA’s project. The fabrication consisted in the employment of laboratory techniques such as Magnetron Sputtering, X-Ray Reflectivity, Magneto Optics Kerr Effect measurements, UV lithography, Ion beam etching and thermal evaporation to obtain micron-sized arrays of magnetoresistive structures. Their

(5)

5 characterization was performed with electronic instrumentation for excitation and readout from DC to AC condition, until several tens of kHz, while Helmholtz coils were employed to provide a static magnetic field and the Oersted field generated by AC current lines, according to Ampere’s law, was instead used for alternated magnetic field generation.

Chapter 3 is centered on the key innovative concept of the European project MADIA: the development of a method for the dynamic sensing of magnetic nanoparticles diffusing on top of an array of magnetoresistive structures. First part is devoted to the choose of the fluidic approach, performed through the calculations with Finite Elements Method (FEM) of the expected diffusion or flow profiles of two different species in solution at different concentrations. The last part reports the dynamic sensing experiments performed within the project consortium.

As an alternative approach for the detection of neurodegenerative diseases markers, efforts were also dedicated to the evaluation of a different approach, based on Electrochemical Impedance Spectroscopy (EIS) devices, as reported in Chapter 4. Considering the versatility of this tool, similar devices have also been used for food pathogen detection and for monitoring in real team cancer cells cultures.

Finally, Chapter 5 reports about the development of an integrated readout system performed during this PhD research work, starting from the first Printed Circuit Boards (PCB) design until the final interface equipped with automation system and signal shielding.

(6)

Chapter 1: Spintronics theory and applications

Spintronics is a research field at the “boundary” between Electronics and Magnetism, with the aim to investigate and explain the phenomena related to spin-polarized currents flowing into a device, material or structure at the micro and nano scale.

A key phenomenon in spintronics is the so called magnetoresistance, a change in ohmic conductivity due to a magnetic field1. First observation of such a behavior dates to 1857, when W. Thomson measured in his works a small change in the resistivity of Nickel and Iron when varying the angle between the flowing current and an external magnetic field2. This behavior is today called Anisotropic Magneto-Resistance (AMR), namely a different conductivity for electrons with spin oriented parallelly or perpendicularly with respect to the flowing current1. More than a century later, another useful phenomenon was identified in the structures exhibiting the AMR by measuring the voltage drop in transversal direction with respect to the flowing current3. Ky found the dependence of this quantity on the angle between the flowing current and the magnetic layer magnetization.

From theoretical point of view, Mott can be considered the pioneer of Spintronics, with the development of its theory about transport of spin polarized current inside magnetic materials during his studies on the transition metals in the ‘30s of XX century4. The “two current model”

considers separated and not interacting channels for the electron transport having spin up and down when the distance travelled is less than the mean free path of electrons into the system, since this phenomenon can cause also spin flip. A confirmation of this assumption was obtained several decades later, through Julliere’s works on thin films made by 𝐹𝑒/𝐺𝑒𝑂/𝐶𝑜 at cryogenic temperatures, leading to the discovery of the Tunnel Magneto-Resistance (TMR) effect5, where the tunneling current through the insulator is proportional to the product of the densities of the states for any spin sub-band, so it depends on the relative orientation of the magnetizations of the two ferromagnetic layers.

A decade later, A. Fert6 and P. Grunberg7 groups discovered independently the phenomenon of Giant Magneto-Resistance (GMR) through their studies on the conduction behavior respectively on 𝐹𝑒/𝐶𝑟 and 𝐹𝑒/𝐶𝑟/𝐹𝑒, which are basically structures similar to the ones

(7)

7 showing tunnel magnetoresistance but with the replacement of the insulator with a metal. These systems exhibit a sharp change in the conductivity when varying a magnetic field acting on them. Fert and Grunberg were awarded with the Nobel Prize for Physics for these studies in 2007.

1.1 Magnetoresistive structures theoretical overview

In this section, a brief theoretical overview is provided about the working principles and main characteristics of the magnetoresistive structures investigated during this PhD research work.

As a figure of merit of the performances of a spintronic device, it’s useful to define the so called magnetoresistive ratio (MR), which represents an indicator of the change in electrical resistance of a system induced by a magnetic field 𝐵. MR can be considered the maximum excursion of a magnetoresistive signal normalized to its lower value as follows:

𝑀𝑅 = max𝑅(𝐵) − 𝑅

𝑅 =𝑅 − 𝑅

𝑅 . (1.1)

Although this definition can appear repetitive, the assertion is strong while formulated in this way because it takes into account any possible value of the magnetic field and shape of the curve.

1.1.1 Giant Magneto Resistance

The GMR effect in magnetic multilayers originates from the spin-dependence of the electrons scattering during charge transport in ferromagnetic materials, where the magnetization direction plays a key role for this phenomenon. The first theoretical model of this effect, based on a semi- classical approach, was developed by Camley and Barnas in 19898 after the experimental demonstration performed by Fert and Grunberg.

Let’s consider, for example, a Fe/Cr/Fe tri-layer where the material changes along z direction while the composition along x is uniform once z coordinate is fixed as sketched in Figure 1.1.

It’s possible to write the Boltzmann equation in any region as:

𝜕𝑔

𝜕𝑧+ 𝑔

𝜏𝑣 = 𝑒𝐸 𝑚𝑣

𝜕𝑓

𝜕𝑣 (1.2)

(8)

8 where 𝑓 is the equilibrium distribution function, 𝑔 its correction caused by scattering and an external electric field E along x direction and 𝜏 the average time between two scattering events.

Considering the planar extension of the multi-layer in 𝑥𝑦 plane, we can ignore what happens in 𝑦 direction since it is perpendicular to the electric field applied, so the net direction of the electrons flow into the structure.

The solving approach suggested by Camley and Barnas is to consider g in any region as the sum of four different contributions obtained by considering the electron spin ↑ and ↓ and the sign of their velocity with respect to z axis, which results in the expression8

𝑔 ± ↑(↓)(𝑣 , 𝑧) = 𝑒𝐸𝜏 𝑚

𝜕𝑓

𝜕𝑣 1 + 𝐴± ↑(↓)𝑒 | | (1.3) where the constant A can be determined by the boundary condition for any family of electrons.

Figure 1.1:Sketch of the system employed for Camley and Barnas calculations on GMR8 .

It’s possible to derive the values of the transmission coefficients during spin propagation. Once named 𝜃 the angle between the magnetization of the two iron electrodes, they are respectively8

𝑇↑↑= 𝑇↓↓= cos 𝜃

2 (1.4a)

𝑇↑↓ = 𝑇↓↑ = sin 𝜃

2 (1.4b)

where equation 1.4a represents the probability for an electron to be transmitted through the interface maintaining its spin unaltered, while equation 1.4b is the rate of spin flip during the

(9)

9 transmission. These results highlight the dependence of the electron transport on the reciprocal direction of the magnetization vector of the ferromagnetic layers.

Once obtained the values of the function g, which depend on the transmission coefficients8, for any case and region, it’s possible to derive the current density of the entire structure in the direction identified by the electric field by the expression:

𝐽(𝑧) = 𝑣 𝑔(𝑣 , 𝑧)𝑑 𝑣 (1.5)

A further integration with respect to the spatial coordinates of 𝐽 allow us to determine the total current flowing into the system for an applied voltage that causes the electric field 𝐸 considered, from which it’s possible to derive the ohmic resistance in dependence of the angle 𝜃.

Figure 1.2: GMR behavior Vs magnetic field correlated with the mutual direction of FM layers magnetization7. Specifically, when they are parallel, the structure has low resistance, while the opposite situation is present when their magnetic moments are anti-parallel (near zero field in this

case).

Two years later, Barthelemy and Fert were able to calculate the magnetoresistive ratio for the same structure described above (employing analytical procedures, but still in semi-classical approach) in dependence of geometrical structure’s parameters 9

(10)

10 𝑀𝑅 =3

2(𝑇− 𝑇) 𝑒 𝑡

𝜆 𝑡

𝜆

(1.6)

where 𝑇 and 𝑇 are the transmission coefficients for up or down spin, 𝑡 and 𝑡 represents the thickness of the Chromium and Iron layers respectively, while 𝜆 = 𝜏𝑣 is the typical length associated to the Fermi velocity of the system under study.

A quantum mechanical approach was proposed in the same period by Levy et al.10, who obtained more general results that can be reasonably applied to the specific case represented by Equation 1.6. The typical shape of a GMR effect is shown in Figure 1.2. In this case the structure7 is a trilayer 𝐹𝑒/𝐶𝑟/𝐹𝑒, where the ohmic resistance is measured during a magnetic field sweep (1 𝑂𝑒 = 10 𝑇). The high conductivity status (for negative fields) is a consequence of the parallel alignment of the ferromagnetic layers. Once one starts to rotate the layer magnetization due to the magnetic field torque, the signal increases and then saturates at its maximum value, when the electrode magnetizations are anti-parallel, obtaining a low conduction state. The following field increasing reduces again the angle between the iron layers magnetization (one already oriented in its direction), until they reach a low resistance state when they result parallel again.

Another interesting result can be predicted from RKKY interaction for a system consisting of two ferromagnetic layers separated by a non-magnetic conductor, originating the so-called interlayer exchange coupling. In this case, the lowest energy state for the system results periodically correlated with the spacer thickness11: some values for this parameter promotes a parallel orientation of the FM layers magnetization while, on the contrary, others facilitate the anti-parallel configuration, leading to different values of the magnetoresistive ratio. At the beginning of the ‘90s, Parkin et al. observed these oscillations finding a period several times larger than the prediction12. This unexpected behavior was then explained by Coehoorn’s work on Co/Cu and Fe/Cu both from theoretical and experimental point of view13. His model considered the non-magnetic layer as a group of molecular planes instead of a continuous and this discretization led him to a sort of “sampling” of the RKKY oscillations for real system, obtaining values in good agreement with the performed experiments.

(11)

11

1.1.2 Tunnel Magneto Resistance

Tunnel magnetoresistance effect arises in nanostructures where a thin insulating layer is sandwiched between two ferromagnetic materials. Since tunneling is a transition between quantum states belonging to the external electrodes, it can be treated by applying Fermi’s golden rule, which can be expressed for the current of such a tri-layer in the following form:

𝐼 =2𝜋𝑒

𝑓 𝐸 [1 − 𝑓(𝐸 + 𝑒𝑉)] − 𝑓(𝐸 + 𝑒𝑉) 1 − 𝑓 𝐸 𝑀 , 𝛿(𝐸 − 𝐸 )

,

(1.7)

where the indexes 𝜇, 𝜈 indicates the quantum status of the conductors 1 and 2, 𝑓(𝐸) is the Fermi-Dirac distribution calculated at the energy 𝐸, 𝑀 , is the element of the transfer matrix between initial and final state, while the term 𝛿(𝐸 − 𝐸 ) ensures energy conservation in the process.

In case of low temperature and 𝑒𝑉 << 𝐸 , the reverse tunneling contribution can be neglected and the Equation 1.7 can be reasonably approximated to14

𝐼 =2𝜋

ℏ 𝑒 𝑉 𝑀 , 𝛿 𝐸 − 𝐸 𝛿(𝐸 − 𝐸 )

,

(1.8)

so, the problem reduces to explicit the transfer matrix element value, which depends on the barrier features and the electronics status involved in the tunneling. According to Bardeen, 𝑀 , can be expressed as the flux of the current operator through any surface inside the barrier separating the conductors15

𝑀 , = ℏ

2𝑚 𝑑𝑆⃗ ⋅ (𝜓 ∇𝜓 − 𝜓 ∇𝜓 ) (1.9)

Considering Tersoff and Hamann theoretical studies16, it’s possible to consider the dependence of the current operator on square modulus of the second wave function calculated in the plane separating the barrier and the other electrode. Through plane wave expansion, it’s possible to derive the relationship14

𝐼 ∝ 𝑉𝐷 (𝐸 ) |𝜓 (𝑧 )| 𝛿(𝐸 − 𝐸 ) = 𝑉𝐷 (𝐸 )𝐷 (𝑧 , 𝐸 ) (1.10)

(12)

12 where z axis is directed perpendicularly to the junction and 𝑧 is the coordinate of the surface separating conductor 1 and the insulator.

If the assumption of 𝑒𝑉 << 𝐸 is no longer valid, it’s possible to obtain the current expression in integral form as follows14:

𝐼 = 𝐴 𝐷 (𝑒𝑉 − 𝐸)𝐷 (𝐸)𝑇(𝐸, 𝑧 = 𝑧 + 𝑑)𝑑𝐸 (1.11)

where A is a constant and 𝑇(𝐸, 𝑧) is the transmission coefficient for an electron having energy E (considering zero at the Fermi Energy of the first metal) through a rectangular potential barrier having thickness d and amplitude 𝜙 − 𝐸 (being 𝜙 the working function of the insulator).

From definition of conductance, it’s possible to obtain this parameter by differentiating Equation 1.11, leading to the relationship:

𝐺(𝑉) = 𝑑𝐼

𝑑𝑉≃ 𝑒𝐴𝐷 (0)𝐷 (𝑒𝑉)𝑇(𝑒𝑉, 𝑧 = 𝑧 + 𝑑) (1.12) where it has been considered that for voltages below some hundreds of 𝑚𝑉 other contribution affects this result for less than 10%17. In this case, G results to be proportional to the product of the density of the states of the two metals sandwiching the insulating barrier at energies which differs for the electric external bias 𝑒𝑉.

Considering Mott’s hypothesis of independence of the conductive channels for different spin orientation4, it’s possible to split the total conductivity obtained in Equation 1.12 as the sum of two different and separated contributions:

𝐺(𝑉) ∝ 𝐷 ,↑(0)𝐷 ,↑(𝑒𝑉) (1.13a)

𝐺(𝑉) ∝ 𝐷 ,↓(0)𝐷 ,↓(𝑒𝑉) (1.13b)

In a quantum mechanical treatment, an external magnetic field B causes an energy split 𝜇 𝐵 between the energy bands for spin up and down electrons18, originating a change in the density of the states at the Fermi Energy which can let a magnetoresistive effect to rise according to Equations 1.13. Notably, this separation exists spontaneously at zero field in ferromagnetic materials, originating from Heisenberg’s exchange interaction19.

(13)

13

Figure 1.3: Spin dependent density of states as a function of energy for two magnetic electrodes of a MTJ in parallel and anti-parallel configuration.

To evaluate more in detail the TMR effect, it’s useful to discriminate between majoritarian and minoritarian spin carriers considering the relative abundance in a specific ferromagnetic layer.

With reference to Figure 1.3, if two magnetizations are parallel, the majoritarian carriers on one side of the barrier will play the same role in the other electrode, having the chance to be “hosted”

in so many available electron states, specifically the same number if the composition and geometry are exactly the same.

On the contrary, if the two magnetizations are anti-parallel, majoritarian carriers in one side will be minoritarian in the other, having poor chance to perform the tunneling, which results inhibited in this case, so the system shows higher resistance with respect to the previous scenario. In any condition, the minoritarian carriers’ behavior can be neglected, since the transmission coefficient through the barrier is so smaller than unity at nm scale20. These two extremal conditions let us able to define the magnetoresistive ratio in terms of conductance (or equivalently resistance) of parallel and anti-parallel configurations, but also in dependence of the spin polarization of the electrodes as follows

𝑇𝑀𝑅 =𝐺↑↑ 𝐺↑↓

𝐺↑↓ =𝑅↑↓− 𝑅↑↑

𝑅↑↓ (1.14a)

𝑇𝑀𝑅 = 2𝑃 𝑃

1 − 𝑃 𝑃 (1.14b)

(14)

14 where 𝑃 = (𝑖 = 1,2). Equation 1.14b is also known as Julliere’s formula5. With more general calculations, it’s possible to derive also the generic dependence of the conductance on the angle 𝜃 between the magnetization of the two electrodes21

𝑇𝑀𝑅(𝜃) =𝐺↑↑+ 𝐺↑↓

2 +𝐺↑↑− 𝐺↑↓

2 cos(𝜃) (1.15)

A characterization in a magnetic field is shown in Figure 1.422. It’s possible to notice that the shape seems similar to GMR’s response, since also in this case it’s possible to identify low and high resistance states where the ferromagnetic electrodes magnetizations are respectively parallel and anti-parallel and an intermediate region where re-orientation takes place. However, the excursion between the two extremal conditions, i.e. the magnetoresistive ratio, is a couple of order of magnitude higher. Furthermore, Figure 1.4 reports only the effect of relatively small fields but moving to higher (positive) values the system reaches again a high conductive state when the electrodes magnetization are supposed to both follow the external field direction.

Figure 1.4: TMR Ohmic resistance and TMR Vs magnetic field obtained in22.

1.1.3 Planar Hall Effect

The concept of Planar Hall effect can be associated to magnetoresistance, although it’s not exactly the same, since it refers to the change in the voltage drop in perpendicular direction with respect to the flowing current and not the electron transport resistance in the direction of the applied electric field, as indicated by the term Hall inside the name. In any case, it is useful to define the Planar Hall Resistance (PHR) as the ratio of the measured voltage drop and the driving current23, to have a term of comparison with the other structures investigated.

(15)

15

Figure 1.5: Planar Hall effect typical layout.

This phenomenon exists in structures exhibiting AMR, where the Ohm’s law in local formulation can be written as24

𝐸⃗ = 𝑀 𝐽⃗ ⋅ 𝑀 [𝜌− 𝜌 ] + 𝜌𝐽⃗ + 𝜌 𝑀 × 𝐽⃗ (1.16) where 𝐽⃗ is the current density, 𝑀 a dimensionless unit vector parallel to magnetization and 𝜌, 𝜌 the electrical resistivity when magnetization is parallel or perpendicular to the current density respectively. Usually it results 𝜌> 𝜌 , and their difference normalized with their average value can reach the value of 3% at room temperature for Permalloy (i.e. a 𝑁𝑖𝐹𝑒 alloy)25. Considering a thin film of a ferromagnetic material with the layout shown in Figure 1.5 and supposing the thickness 𝑡 much smaller than the other dimensions, the magnetization vector lies in the xy plane due to the shape anisotropy1. If we flow a current 𝐼 = 𝐼 𝑥 in this system, the y component of the electric field according to Equation 1.16 is

𝐸 = (𝜌− 𝜌 )𝐽 𝑠𝑖𝑛𝜙𝑐𝑜𝑠𝜙 (1.17)

being 𝜙 the angle between magnetization and current density. Integration of Equation 1.17 over the space of the system leads to the evaluation of the PHE voltage as

𝑉 = 𝑉 =1

2Δ𝑅 𝐼 sin(2𝜙) (1.18)

(16)

16 where ΔR = . Equation 1.8 can be used to estimate the Planar Hall Resistance of the system, which is 𝑃𝐻𝑅(𝜙) = . In Equation 1.18, it is stated the dependence of this phenomenon on the angle 𝜙, that can be changed, for example through the torque exerted on the magnetization by an external magnetic field directed along y axis.

Figure 1.6: PHE resistance response calculated by Hung and coworkers23.

Considering a spin-valve system, so assuming that the ferromagnetic layer is pinned in a particular direction at zero field, this parameter can be calculated starting by the Stoner- Wohlfarth model, which is useful to determine the system energy in macro-spin approximation, so considering the behavior of the film as the one of a single magnetic domain, leading to the following expression26:

𝐸 = 𝐾 𝑡 sin 𝜙 − 𝑀 𝑡 𝐵𝑐𝑜𝑠(𝛼 − 𝜙) − 𝐽𝑐𝑜𝑠(𝛽 − 𝜙) (1.19) where 𝐾 , 𝑀 , 𝛼, 𝛽 are the anisotropy constant, the saturation magnetization and the angles between the easy axis and respectively the magnetic field and the exchange bias direction, while 𝐽 represents the interlayer coupling. The minimization of Equation 1.19 allow us to determine the angle dependence on the external magnetic field, which is23:

(17)

17 𝜙(𝐵) = 𝑛𝜋

2 +(−1)

2 arcsin 𝑉(𝐵)

𝑉 (1.20)

where 𝑛 = 0, ±1 to ensure existence and continuity of 𝜙. By inverting Equation 1.20 and taking into account Equation 1.18, it is possible to calculate the response of this structure with respect to an external magnetic field in terms of 𝑉 or 𝑃𝐻𝑅, as shown in Figure 1.6. It is possible to notice that the maximum and minimum values correspond to 𝜙 = ±45° since 𝑉 ∝ sin(2𝜙), while the signal saturates near the same value when the magnetization is oriented along the current axis, making an angle of 0° or 180°, which is a very different behavior with respect to the other structures investigated.

1.2 Applications

In this section, an overview of the principal applications of GMR, TMR and PHE phenomena is provided, starting from the basics employment and optimization for data storage, which boosted their diffusion in the last decade of the XX century to the usage in combination of semiconductors in hybrid structures for electronics. Finally, more attention will be paid to the application in biosensing devices, since the goal of this research work is the development of a magnetoresistive sensing platform.

1.2.1 Data storage

The high and low resistance states in magnetoresistive structures can be associated to 1 and 0 binary values for data storage used in Information and Communication Technologies (ICT) field.

The principle of operation of a magnetoresistance (MR) based read head is shown in Figure 1.7. The layer near the magnetic domain storing the information is sensitive to its magnetization direction, meaning that it couples in antiferromagnetic way with the bit to minimize the system energy. This orientation switch leads to the change from high to low resistance saturated status and vice versa without affecting the domain, so performing the reading, while the adjacent coil is deputed to “write” a bit exerting a torque through the application of a magnetic field due to the flowing current.

(18)

18

Figure 1.7: Scheme of a magnetoresistive read head.

At the beginning of ‘90s of the XX century, a mass production of AMR structures employed as read head for hard disk drives (HDD) led to a boost in the growth rate of storage areal density until 60% per year27, more than twice the trend associated to the improvement of the previous technologies based on thin films28 (Figure 1.8). Then, the trend of miniaturization in electronics drove research towards higher magnetoresistance in order to enable smaller bit sizes, since the signal to be measured is proportional to domains’ dimension. For this reason, starting from Fert’s and Grunberg’s group results, GMR based read head perfectly operating at room temperature were developed by IBM and reached the market in 1997.

Figure 1.8: Areal density of data storage respect to year of production.

(19)

19 The exploitation of AMR and GMR in magnetoresistive read heads permitted an impressive increase in areal memorization density from 0.1 to 100 𝑀𝑏𝑖𝑡 ⋅ 𝑖𝑛𝑐ℎ between 1991 and 2003.

Then, the optimization of TMR and the introduction of this technology provided a further boost, up to the impressive value of 600 𝐺𝑏𝑖𝑡 ⋅ 𝑖𝑛𝑐ℎ , corresponding to a bit surface of less than 2 ⋅ 10 𝑖𝑛𝑐ℎ 29.

A further application of Magnetic Tunnel Junctions concerns the development of Magnetic Random Access Memories, whose principle of operation in sketched in Figure 1.9. Basically, they consist in a matrix of these structures connected through perpendicular top and bottom electrodes, that can be employed to select a specific cell30.

Figure 1.9: Scheme of an MRAM: the memory cell is identified by the lines “bit” and “digit”, which

apply voltage or current for reading and writing, while the connected CMOS stabilizes the circuit.

In this case, there is no need of coils for writing, but the employment of spin-polarized current31 can exert the torque needed for one electrode switching32. Since this phenomenon has a threshold in terms of current magnitude, a reading stimulus useful for measuring the resistance must be below this critical value, which depends upon ferromagnetic electrode dimension33. Notably, MRAM has performances similar to standard RAM, but they are non-volatile29, being a good candidate for universal memories that can boost computers’ performances not presenting a bottle neck between the memories deputed for storage and calculation. Further optimization in the layout, by changing the cell from pillars to rings configuration, led to the development of Spin RAM, which performances are slightly better than MRAM29.

(20)

20

1.2.2 Spin transistors

Semiconductors play a key role in contemporary electronics for their versatility, since it is possible to engineer their electrons bands and to modify the Fermi’s level and the density of charge carriers (both electrons and holes) through the doping procedure.

Figure 1.10: Working principle of a Spin Transistor proposed by Datta and Das34

(https://www.nims.go.jp/mmu/tutorials/SpinFET.html).

In a transistor, it is possible to modulate the drain current by means of a gate voltage, obtaining high transmission speed and signal modulation especially after the introduction of Field Effect Transistors (FET) and, later, the Metal-Oxide-Semiconductor-FET (MOSFET), where the amplitude of the charge transport channel between source and drain can be precisely tuned by the application of the gate voltage, from saturation (maximum current) until the complete blockage.

(21)

21 In such devices, the introduction of spin degree of freedom opened new frontiers and applications, leading to the concept of Spin-FET by Datta et al. in 1990, as a prototype of a three terminal spintronic device34 which principle of operation is sketched in Figure 1.10. In this case, electron spins are injected from source and collected by drain after moving in the 2D Electron Gas (2DEG) at the interface between insulator and semiconductor, so having higher mobility with respect to the standard transport in semiconductors. The gate voltage determines spin precession through the Rashba effect generated by the associated electric field35, so the spin orientation of the carriers arriving to the drain is proportional to the gate voltage36. The magnetization of the drain terminal then acts as a filter, preferring magnetic moments with parallel direction to its magnetization, so at the end if the magnetization of source and drain forms an angle of 𝑛𝜋, with integer n, the resulting current flowing into the device results proportional to carrier spin orientation.

1.2.3 Biosensors

The employment of magnetic labels brings many advantages for biosensing: for example, it is possible to perform analyte separation from a solution followed by direct measurement and quantification. Considering that other biological cells and compounds are typically non- magnetic, this screening procedure is usually very effective and furthermore the magnetic background is negligible in such a measurement. This kind of particles are often synthetized with a silica coating, which let them be biocompatible for in vivo applications and more easy to functionalize with antibodies or aptamers able to bind a broad range of analytes.

Figure 1.11: Principle of detection for a MR based sensor. (a) DNA probe immobilization, (b) Hybridization of the analyte DNA, (c) binding of the marked magnetic particles 37.

(22)

22 A population of MNPs embedded in an external magnetic field tend to have a parallel magnetization with respect to its direction, exerting a stray field on a sensor in their vicinity, whose response will be influenced38. In this respect, for achieving a larger sensitivity, it is crucial to obtain a good compromise between the structure dynamic range and the particles’

magnetization, since it’s necessary to choose an external magnetic field value not so large, in order to avoid saturation of the response, but only not too small because in that case particles’

magnetizations results randomly oriented. Furthermore, it can be useful to add a superimposed AC magnetic field39 and to choose the right working point to have a better signal to noise ratio.

Regarding the latter, considering that the magnetic field generated by a dipole decays as the third power of the distance, it’s necessary to have the particles as close as possible to the sensor’s surface40, using, for example, a sandwich-like configuration between a ligand-acceptor pair as sketched in Figure 1.11. Several strategies have been developed for this purpose, starting for example37 with a sensors’ functionalization with a double strand DNA with biotin, able to bind streptavidin present on the MNP, in order to have the particles at a submicrometric distance from the sensing structure.

Figure 1.12: Bead Array Counter (BARC) developed and realized by Baselt et al. in 199841. The first idea to employ magneto-resistive structures also for particles’ sensing date back to 1998, when Baselt et al. developed a Bead Array Counter (BARC)41 based on multiple GMR junctions, in order to perform a fast and easy bio-recognition procedure. Their device is shown in Figure 1.12. Several years later (2004), Schotter et al. performed a comparative study between the performances of GMR-based sensors and standard fluorescent transduction,

(23)

23 showing a better sensitivity of the former in the concentration range of interest37. Specifically, the basic sensing element (there are 206 in the device) consists in a double spiral patterned on a GMR stack as shown in Figure 1.13 (a).

This promising result led to the development of a wide number of studies about spintronics devices for biosensing applications, in order to maximize their performances. Since the sensitivity of a GMR based sensor increases while decreasing its volume but, on the contrary, this leads to a smaller available signal, one good compromise could be a layout with a dense array of structures in the sub-micron range, as reported by Wood et al.42 in 2005. In this work, a proof of concept of magnetic particles’ detection was performed, where their magnetic moments were simulated with a magnetic SPM tip placed above the sensing surface at the typical distance of a sandwich assay (about 300 𝑛𝑚) and exerting a field comparable with 2 𝜇𝑚 sized commercial particles. However, to evaluate quantitatively bio-recognition events, the markers should be smaller of a couple of orders of magnitude, so that they can bind one or at least few analytes, but in this scenario the magnetic signal becomes smaller and it’s possible to reach the paramagnetic regime, where a stabilization is needed43. Li et al. demonstrated in 2006 the chance to measure quantitatively the presence of magnetic nanoparticles of few nanometers on top of sub-micron optimized GMR structures in the range of 20 − 200 𝑛𝑚, also integrating some permanent magnet in their device as fixed magnetic bias. Further evolution in the nanoscale range with highly dense arrays led to measure quantitatively a solution until the notable detection limit of 20 𝑧𝑚𝑜𝑙 44.

Figure 1.13: Sensing element of the device realized by Schotter et al.37 (a) and the results of comparative studies performed about sensitivity of their GMR based device and fluorescence-based

methods (b).

(a) (b)

(24)

24 Exploiting the sub-micron range for sensors dimensions can provide some disadvantages for industrial applications and mass production, because it requires nanofabrication techniques such as EBL which are expensive and slow. For this reason, in order to enhance the sensitivity with structures in the UV Lithography fabrication range, some alternatives were sought, for example by developing sensing circuits more complex than a single magneto-resistor. In this framework, Millen et al. organized a network of GMR structures arranged in a Wheatstone bridge geometry able to quantify the concentration of superparamagnetic particles in the picomolar range45. Another approach for optimizing DNA detection with this technology was developed by M. Koets et al. 46: by taking advantage of PCR amplification for the LamB gene of E. Coli, it has been possible to measure it quantitatively in the range 4 − 250 𝑝𝑀 with single big GMR sensors (3 𝜇𝑚 x 100 𝜇𝑚). In their devices, the magnetic tags were 300 nm superparamagnetic particles magnetized with on chip current lines, which could provide a bias field due to the Biot-Savart law. In this case, there was a baseline signal due to the current which needed to be subtracted from the particles’ one, obtaining a smooth sensitometric response in the range of interest. Furthermore, magnetic manipulation led to a very fast response compared to previous works, with few minutes needed for obtaining stable results. A further evolution of this kind of system was developed by Zhi et al.47, with a modular system consisting by a isothermal amplification module, microfluidic and GMR sensing part with a limit of detection of 10 𝑐𝑜𝑝𝑖𝑒𝑠/𝑚𝐿.

A very useful application of magnetic detection is in the mycotoxin field, considering that this kind of analyte is very small compared to others (< 300 𝑘𝐷), which is usually a problem for other detection methods. In this case, the signal is not provided by mass or volume of the analyte, but only by the magnetic tag: for this reason Mak et al. 48 developed an assay able to measure different kinds of mycotoxin in the 𝑝𝑔/𝑚𝐿 range, which is as low as in the case of more complex and expensive methods based, for example, on Surface Plasmon Resonance.

In 2015, Wang et al. 49 developed a probe station like system able to multiplex protein detection in a single cartridge in standard clinical concentration range, putting the basis for the application of this technology in mass-produced Point OF Care (POC) devices. A year later, Choi et al.50 further optimized and integrated a similar system, obtaining a portable device which can monitor IgG and IgM antibodies levels within one step for self-diagnosis that can be visualized and stored through a simple smartphone interface. The complete device is shown in Figure 1.14

(25)

25 and consists in a cartridge reader (similar to the equipment for the home monitoring of glycemia already available on the market) that can be employed with mono-usage cartridge.

Figure 1.14: Complete Point of Care platform for multiple immune-detection in a single cartridge developed by Choi et al.50 also equipped with a simple smartphone interface.

The first proof of concept of the possibility to employ Planar Hall Effect based sensors for magnetic particles’ sensing was performed in 2005 by Ejsing et al.51, whose approach is reported in Figure 1.15. They realized a simple but effective trilayer device able to detect few 2 𝜇𝑚 sized commercial magnetic beads, which were largely employed, after functionalization, in biological laboratory processes such as separation and purification protocols.

Later, it was also demonstrated the chance to detect a single 2.8 𝜇𝑚 particle simulated by the photolithography of a single dot in the center of a PHE based sensor surface52, which resulted in a shift in the response during the characterization in a static magnetic field. Also the effective detection of Dynabead M-280 coated with streptavidin was performed in 2007 by Thanh et al.

53 , employing 3 𝜇𝑚 x 3 𝜇𝑚 structures in a spin-valve configuration, while an improved device developed by Quang Hung et al. 54 in 2010 was able to get the impressive value of 1450 for the signal to noise ratio for beads detection.

(26)

26 Figure 1.15: Planar Hall Effect based device for particles sensing developed by Ejsing et al. in 200551. Systematic studies have been performed in order to optimize this kind of devices to pave the way for application in the biosensing field, for example regarding the influence of the thickness of free and pinned layer for the spin valve configuration55, obtaining that the larger is the former, the higher sensitivity shows the device, while the latter is correlated to this parameter in the opposite way. Furthermore, it has also been demonstrated that the employment of Wheatstone Bridge geometry can lead to the enhancement of the signal by a geometrical factor56. A further improvement in sensitivity has been achieved by Hung et al.57 (Figure 1.16), who got the impressive limit of detection value of 4 ⋅ 10 𝑒𝑚𝑢, about 3 orders of magnitude more powerful than expensive and complex systems based on the employment of superconductive quantum interference devices (SQUID).

For integrating Planar Hall effect-based sensors in a Lab On Chip platform, it is necessary to implement also a microfluidic module able to handle small volumes of fluid containing the magnetic nanoparticles (MNPs), specifically to drive them very close to the sensing surface, considering the third power trend of reduction of a dipole magnetic field. One example in this kind of applications is the work of Jeong et al. 58, who demonstrated the capability of detection of 2 ⋅ 10 𝑒𝑚𝑢 in presence of a thick passivation layer between sensing surface and particles, necessary to avoid fluid leakage which can lead to shortcut between terminals, so obtaining a perfectly reusable device.

(27)

27 Furthermore, by increasing the packaging of the structure with a meander-like shape arranged in Wheatstone bridge geometry, Hansen and Rizzi have demonstrated the possibility to measure particles’ concentration down to 4 𝑝𝑀 with this kind of sensing circuit, after an optimization of measurement parameters, especially the frequency59.

Figure 1.16: SEM image of the PHE ring realized by Hung et al.57, able to measure very low magnetic moment and resulted more sensitive than SQUID technology.

The first theoretical proof of concept of magnetic particles detection by Tunnel Magneto- Resistance junctions was performed by Schepper et al. in 200460. In this work, micromagnetic simulations technique was employed to study the influence of magnetite superparamagnetic beads on the response of tunnel magneto resistance structures consisting in a Permalloy free layer, an 𝐴𝑙𝑂 barrier and a 𝐶𝑜𝐹𝑒 reference layer in two configurations of magnetization: in plane and out of plane. They also studied the spin texture of the system in the first case. They showed, in principle, that the use of TMR sensors could lead to a limit of detection of some tens of particles, an impressive result for that period, especially considering the micrometer in plane dimensions of the structures. One year later (2005), Grancharov et al.61 performed the experimental detection of magnetic nanoparticles using Magnetic Tunnel Junctions (MTJs) above a functionalized 𝑆𝑖𝑂 passivation layer, with the observation of a change in the shape of the response Vs an applied magnetic field.

As in the previous cases, the optimization in the development of TMR based devices covered both the sensors performances and the realization of complex sensing circuits able to reduce thermal drift and to enhance sensitivity. For example, Shen et al.62 developed a Wheatstone bridge-based device, where each component consists in a series of 16 MTJs. In this condition,

(28)

28 they were able to get a limit of detection of few 𝑛𝑀, corresponding to about 30 𝑝𝑔/𝜇𝐿 of target DNA strand. MTJs’ employment in liver cancer detection showed promising results in Lei et al. work63, since they showed an excursion dependence of the response as a function of the magnetic field with different numbers of cancer cells immobilized onto the surface with antibodies.

An alternative sensing method, closer to fluorescent imaging and quantification than the simple measurement of magnetoresistance, was proposed by Chan et al.64 in 2011. Their readout technique was similar to the standard employment of TMR in hard drive read head, letting such a device scan a functionalized matrix able to immobilize MNPs bounded with an analyte to be revealed. In this case, any “pixel” shows a signal proportional to particles’ number, which can be integrated over a square of 100 𝜇𝑚 x 100 𝜇𝑚 area to obtain the map of the local concentration of the solution spotted. They obtained in this way a limit of detection of 10 𝑝𝑀.

An improvement of this method was developed by Vyas et al.65 through the employment of an Anderson Loop for the local magnetic field detection, which lets their system able to obtain a noise reduction in the signal acquisition, achiving a limit of detection for the particles stray field of about 0.006 %/𝑂𝑒. Another possible technique of detection was studied both theoretically and experimentally by Albisetti et al.39 in 2013 who derived the signal obtained by the superposition of an AC magnetic field and a DC: the latter was used to define the working point of the sensor, while the former was employed for magnetizing the particles. In this case, they found that the highest sensitivity (for MNPs detection) is at a working point where the product between the field and the second derivative of the resistance is maximum.

Figure 1.17: Flexible TMR based device realized by Chen et al.66 in 2017.

(29)

29 During the last years, one of the most important challenge in biosensing field is the development of wearable devices, in order to have the chance to monitor in real time people’s activities. In 2017, Chen et al.66 developed an innovative procedure able to obtain powerful TMRs’

structures over a flexible substrate (Figure 1.17), with high reusability (over 1000 bending cycles without losing in performance), proving the possible applications of MTJs also in this growing field.

(30)

Chapter 2: Fabrication and characterization of Magneto-Resistive structures

The fabrication of the magnetoresistive structures employed in this thesis started during the secondment period at the Physics Department, Center for Spinelectronic Materials and Devices - Bielefeld University, because the hosting group is among the world leaders in the field and also a partner of MADIA’s project.

Since Tunnel Magneto-Resistance junctions are the structures with the highest known magnetoresistive ratio, we decided to start with them to obtain first sensors, then other series of devices based on Giant Magneto-Resistance and Planar Hall Effect were realized by the research team to exploit different sensing strategies.

2.1 Thin films deposition by sputtering: an overview

The thin films needed for building the samples has been realized by sputtering, a technique belonging to the PVD (Physical Vapor Deposition) class. It consists on the extraction of atoms or molecules from a solid-state target made by the material of interest, in order to obtain their ordered deposition on a substrate. A sputtering technique causes the target sublimation by hitting its surface with particles with enough kinetic energy to be provided to break the binding.

Generally, ions are used as “bullets”, so it’s necessary to model the phenomenon considering Coulomb’s interaction. It can be shown that the Rutherford’s cross-section of a process where a particle with energy E can transfer a kinetic energy in the range [𝑇, 𝑇 + 𝑑𝑇] is67

𝑑𝜎(𝐸; 𝑇) = 𝜋𝑀

𝑀 𝑍 𝑍 𝑒 𝑑𝑇 𝐸𝑇

(2.1)

where the index 𝑖 = 1,2 is related to any particle involved and 𝑀 , 𝑍 stand for its mass and charge. It can also be shown that the maximum of energy transferrable in such elastic process is67

(31)

31

𝑇 = 4𝑀 𝑀

(𝑀 + 𝑀 ) 𝐸. (2.2)

It’s possible to have the sputtering when T is larger than the binding energy of the substrate.

Furthermore, there is the chance of energy transfer through vibrational modes to the

neighbors, so a single hit can lead to more particle’s expulsion. A visual representation of this phenomenon is shown in Figure 2.1 (a).

To obtain this phenomenon, it’s usual to generate some Argon plasma into the deposition chamber, in order to have a region where a consistent fraction of the gas is ionized exhibiting a conductive behavior68. If this system is between two electrodes, the free electrons are attracted from the anode and during the path they can ionize other atoms. On the contrary, the positive ions are accelerated to a lower potential region, acquiring kinetic energy and then striking the cathode, which can be a conductive target. The hit of a particle with well-defined mass at a particular energy have a defined probability to extract some surface atoms or molecules from the cathode and they will disperse in gaseous phase in the chamber depending on the plasma shape and the incidence angle. A fraction of them will deposit on the substrate with an order that depends on the temperature, the kinetic energy, the orientation and eventually on a rotational speed of the substrate. To be sure that this process is controlled and reproducible, it must take place in high vacuum, so that the contribution of other ionized gas can be considered negligible.

Figure 2.1: Generic sputtering representation (a) and DC magnetron Sputtering (b).

The employment of magnetic fields to confine the electrons inside the plasma, technique known as DC Magnetron Sputtering, can enhance and stabilize the process, since the probability of

(a) (b)

(32)

32 ionization is expressed per unity of path, the spiral “trajectories” which confines their motion can lead to an auto-sustainment of the plasma cloud and, furthermore, its shape results well defined and its density homogeneous, leading to a greater deposition accuracy than other methods, as represented in Figure 2.1 (b). It can be noticed that the ions have mass about thousand times larger than the electrons, so they are very poorly affected by this deviation and reaches the target about at the same rate as it was without the magnetic field.

This framework is not applicable to the deposition of insulating materials, since they cannot act as a cathode to which the Argon nucleus are accelerated. In such a case it’s necessary to proceed in a different way by implementing an RF sputtering method, whose typical realization is to embed a gas to be ionized in an electric field of about 10 𝑉 of amplitude and 10 𝑀𝐻𝑧 of frequency, since the electrons can follow its oscillation while the massive ions experience only its temporal mean value68. So, the system cathode/target/plasma can be considered as a capacitor and it’s possible to demonstrate that under certain conditions the ions can strike the target enabling again the sputtering deposition68. This method is more complex to be implemented than the previous one, since it’s necessary to have a finer control on the plasma density to obtain a constant deposition rate, furthermore it’s necessary to minimize the reflection of electromagnetic energy to sustain the cloud and there is mandatory that the insulating material covers the whole electrode, since if not the result is the deposition of its metal instead of the desired material.

2.1 Magnetoresistive structures fabrication

The fabrication of magnetoresistive devices consist in several steps, which one employs different techniques. The stack was deposited by Sputtering, the film thicknesses was measured by X-Ray Reflection, the magnetic configuration was set by an annealing procedure and the patterning was performed by UV Lithography and ion beam etching.

2.2.1 Stack deposition and magnetic configuration settlement

The instrumentation used for the stack deposition is shown in Figure 2.2 (a). It consists in a Leybold CLAB600 owned by the hosting institute and has different modules. The substrates are introduced in a loadlock (with 8 different positions), where the vacuum is generated after

(33)

33 loading. An intermediate zone is provided with a mechanical arm to move the sample into the deposition chamber, which can accommodate until 6 targets at the same time. A control system drives some valves to isolate the different regions of the instrument if the pressure between them is above a preset threshold: in this way, it’s possible to maintain the deposition chamber clean enough to assure the environmental condition in which the process takes place and to optimize the usage of the connected vacuum pumps.

The deposition is completely managed by software after setting the needed parameters, like target, sputtering and pre-sputtering time and Argon (and eventually oxygen) pressure.

First step of fabrication was to proceed with the target calibration. For this purpose, trial samples of any material involved in the stack have been deposited (Ta, Ru, MnIr, CoFe, CoFeB, MgO) for a well-known time. To determine their thickness, X-Ray Reflectivity (XRR) measurements were carried out. Considering Gibaud’s and Hazra’s model69, it’s possible to correlate, for small incidence angles, the XRR spectra to some characteristics of a thin film under study, as thickness and roughness, if the electronic density of the material is known. The procedure is not analytical but is based on the numerical simulation of the parameters inside the model to be compared with the experimental data through a fitting procedure, which ends when the difference between the dataset and the built curve is below a certain threshold. In case of multilayers, the superposition principle applies, so it has been possible to put some Ruthenium (the first material characterized) on top of the other metals to avoid their oxidation in air and to take into account the surface of the substrate.

Figure 2.2: Sputtering (a) and XRR (b) setup.

(a) (b)

(34)

34 Since the rate calculated for any materials was of the order of magnitude of about ten 𝑛𝑚/𝑚𝑖𝑛, the effective sputtering window has been calculated by considering also the shutter time of the system (0.65 s), to avoid significative deviation in the layers thickness with respect to the planned ones, especially regarding tunnel barrier.

Considering the ferromagnetic behavior of the materials employed, it was necessary to overcome the hysteresis phenomenon in the response, to let them be useful as sensors. There are several strategies to get this purpose, but we chose to perform it by the realization of a FM/Oxide/FM structure where the two electrodes have orthogonal magnetization at zero field, since this configuration is independent with respect to planar shape and dimension of the structures.

The deposition started with the following stack: Ru(4.0 nm)/Ta(4.0 nm)/MnIr(7.5 nm)/CoFe(3.0 nm)/Ru(x)/CoFeB(5.0 nm)/MgO(1.7 nm), where x=(0.80; 0.90; 1.00; 1.05; 1.10) nm from bottom to top, which is specular with respect to the higher half of the complete structure that will be discussed later, so from capping layer to tunneling barrier.

𝑅𝑢(𝑥) thickness determines the exchange coupling between the sandwiching 𝐶𝑜𝐹𝑒 and 𝐶𝑜𝐹𝑒𝐵 ferromagnetic layer after a process of magnetic annealing, which is performed in a vacuum tubular chamber (Figure 2.3) at 10 𝑚𝑏𝑎𝑟 and consists of heating up a sample in a furnace until 360 °C for 60 minutes and let it cool down for 30 minutes, between the poles of a permanent magnet with field of about 700 𝑚𝑇.

Figure 2.3: Annealing furnace after samples’ loading (a) and during heating (b).

(a) (b)

(35)

35 There are several techniques to measure the exchange bias between ferromagnetic layers, but the one employed during this work is based on Magneto-Optics Kerr Effect (MOKE)70. This phenomenon is based on the interaction of polarized light with the electrons’ spin and it’s appreciable only with ferromagnetic materials. In this case, the absorption is proportional to the magnetization of the probed area, obtaining and hysteresis-like curve, so the measurement of the reflected radiation can be used for determining the saturation magnetization and coercive field of a material.

The experimental setup is shown in Figure 2.4 (a): the system is arranged in longitudinal Kerr configuration70. The field is provided by a Helmholtz coil able to reach more than 1 T, while the probe consists in a red laser passing through a polarizer. The film is put between the poles through a holder that can rotate around z axis, so that the reflected beam can be directed to a photodetector that converts the received light power to current. A Python-based software controls the system by driving the coils and correct the measured values through some parameters set by the user, like dark current and pole distances. During the measurement, a black tarp has been put above the system to avoid the influence of the environmental source of light. After getting the data, the coercive fields were extrapolated for any of the samples prepared, which are related to the exchange bias energy between CoFeB and CoFe depending on the thickness of Ruthenium interlayer, as shown in Figure 2.4 (b).

Figure 2.4: MOKE experimental setup (a) and data obtained (b).

(a) (b)

(36)

36 Once chosen the desired pinning strength for the electrode, or equivalently the coercive field of about 50 𝑂𝑒, we started the complete device fabrication with the sputtering deposition of a complete TMR stack over a Silicon substrate, which consisted of 𝑇𝑎(5.0 𝑛𝑚)/𝑅𝑢(30.0 𝑛𝑚)/

𝑇𝑎(10.0 𝑛𝑚)/𝑅𝑢(10.0 𝑛𝑚)/𝑀𝑛𝐼𝑟(20.0 𝑛𝑚)/𝐶𝑜𝐹𝑒(3.0 𝑛𝑚)/𝑅𝑢(0.9 𝑛𝑚)/

𝐶𝑜𝐹𝑒𝐵(3.0 𝑛𝑚)/𝑀𝑔𝑂(1.8 𝑛𝑚)/𝐶𝑜𝐹𝑒𝐵(5.0 𝑛𝑚)/𝑅𝑢(1.05 𝑛𝑚)/𝐶𝑜𝐹𝑒(3.0 𝑛𝑚)/

𝑀𝑛𝐼𝑟(7.5 𝑛𝑚)/𝑇𝑎(4.0 𝑛𝑚)/𝑅𝑢(4.0 𝑛𝑚).

The first 𝑇𝑎(5.0 𝑛𝑚) act as sticking layer between the whole structure and the 𝑆𝑖/𝑆𝑖𝑂 (50 𝑛𝑚) substrate. The function of the bilayer 𝑅𝑢(30.0 𝑛𝑚)/𝑇𝑎(10.0 𝑛𝑚) will be fully clarified later, since we used the signal coming from them as a stopping feedback for the ion beam etching after photolithography. 𝑅𝑢(10.0 𝑛𝑚) act as a seeding layer for the above synthetic antiferromagnet, since it let the 𝑀𝑛𝐼𝑟(20.0 𝑛𝑚) have a better texture than it is when deposited on Tantalum. 𝑀𝑛𝐼𝑟(20.0 𝑛𝑚) adjacent molecular planes acquires through annealing anti-parallel magnetization and the last one strongly couples with 𝐶𝑜𝐹𝑒(3.0 𝑛𝑚), named in this case pinning layer. 𝑅𝑢(0.9 𝑛𝑚) determines a strong anti-ferromagnetic interaction between the underlying compound and 𝐶𝑜𝐹𝑒𝐵(5.0 𝑛𝑚), called pinned layer for this reason, which let this magnetic configuration to be stable for external magnetic field up to several hundred of Oe after annealing. 𝑀𝑔𝑂(1.8 𝑛𝑚) is the only insulating materials employed, which plays the role of the tunneling barrier.

The following layers 𝐶𝑜𝐹𝑒𝐵(5.0 𝑛𝑚)/𝑅𝑢(1.05 𝑛𝑚)/𝐶𝑜𝐹𝑒(3.0 𝑛𝑚)/𝑀𝑛𝐼𝑟(7.5 𝑛𝑚) are a specular structure of the one exposed above, with different thicknesses of 𝑅𝑢 and 𝑀𝑛𝐼𝑟 to obtain a smaller pinning strength to let this 𝐶𝑜𝐹𝑒𝐵(5.0 𝑛𝑚) at the upper side of the barrier to act as an almost free layer, so to be sensitive to small external magnetic fields. 𝑇𝑎(4.0 𝑛𝑚) is the seeding for the capping layer 𝑅𝑢(4.0 𝑛𝑚), which is highly conductive and its useful to have an ohmic contact for the future structure characterization.

Then magnetic annealing was carried out to set the strong pinning of the lower 𝐶𝑜𝐹𝑒𝐵, followed by MOKE measurements to check the magnetization switching of the two electrodes. Then, a second annealing was performed in perpendicular in-plane direction with respect to the first one at 260 °C for 40 minutes and same cooling down time and vacuum, in order to rotate the magnetization of the upper 𝐶𝑜𝐹𝑒𝐵, without affecting the other FM electrode, since its exchange bias was strong enough not to be affected at this temperature and field. In this way the stack reached the desired magnetic configuration able to reduce the hysteresis phenomenon, so the

(37)

37 electrodes at the edge of the barrier show perpendicular magnetization in absence of external fields.

2.2.2 Samples patterning

The samples have been patterned by using UV Lithography methods, since the details of the desired planar dimensions are of the order of some microns.

In general, the first step of a lithography technique is to uniformly cover (usually through spin coating) a sample with a polymeric substance, known as resist, which varies its properties if irradiated by a radiation source (e.g. UV light, laser, or electron beam). In case of positive lithography, the exposed regions will be removed after the development, while in negative lithography the irradiated zones will remain on the sample and the others detaches. An important parameter to be considered, which varies from resist to resist, is the dose, meaning the quantity of energy absorbed for mass unity by the substance, which determines the correct outcome of the process.

As already mentioned, during this work UV lithography technique has been employed, since it has some advantages with respect to EBL, especially simplicity and velocity of execution; in fact it is not necessary to put the samples in high vacuum and it can be implemented by using a mask which creates some shadow zones between the lamp and the sample, which protect below- positioned area from the radiation (so without designing the layout pixel by pixel by the radiation source). Furthermore, the time saving with respect to laser-based lithography and EBL increases rapidly with the area to be patterned, since its exposure time is almost always the same if the sample can be located in the lamp field. The limit of the resolution depends on the frequency of the light employed, since it is determined by diffraction, so generally this phenomenon becomes significant at sub-micron scale. On the other hand, this technique is less versatile than the other mentioned, since the mask should be fabricated a priori. For this reason, the first step of the procedure has been to realize a hard mask for UV Lithography on a laboratory glass by employing the DWL mask-less system owned by the hosting institution.

The layout was realized with the software Autocad 2018 and exported as .dxf, since this format is the only readable by the DWL. It consisted of two arrays made by 9 𝜇𝑚 sized squares, where there will be the MTJs pillars and some alignment crosses for fabricating the readout pads. All the lithography steps must take place in a clean room, where the atmosphere is controlled to

Riferimenti

Documenti correlati

This result strongly suggests that we are observing discrete shifts from part-time to full-time work, as conjectured by Zabalza et al 1980 and Baker and Benjamin 1999, rather

There- fore an important development of the present work would be represented by the implementation of the developed algorithms on GPU-base hardware, which would allow the

In this frame, since the Italian Pharmacovigilance System (Italian Medicines Agency—AIFA) only collects reports for registered drugs, the Italian National Institute of Health

We also show how GSMM can represent more recent modelling proposals: the triple stores, the BigTable model and Neo4j, a graph-based model for NoSQL data.. A prototype showing

(b) It is worthwhile noticing that, in case of the standard diffusion, the nodal set of the segregated minimal configurations shares the same measure theoretical features with the

Gli autori di «Prato pagano», piuttosto, adoperano liberamente l’antico come modello di frattura verso quella visione drammati- ca della Storia e soprattutto come modello

The expression “audiovisual translation” refers to all those translation techniques that aim to transfer the original dialogues of an audiovisual product into another

The original contribution that we will try to offer compared to the literature used will consist of a two-fold task: on the one hand , systematize diverse contributions in a