Magnetic device and corresponding method

Abstract

One aspect of the invention relates to a method of manufacturing a magnetic device (4) comprising a dielectric layer (3) having a via, so-called "via", comprising: forming a tunnel junction (41) on the dielectric layer (3), the tunnel junction comprising a first magnetic layer intended to form said "free layer" (411); and a second magnetic layer intended to form the "reference layer" (412), an insulator layer intended to form a "tunnel barrier" (413); a free layer (411) or a reference layer (412) is located in the via and surrounded by the dielectric layer (3).

Description

Technical Field

[0001] The technical field of the present invention relates to magnetic tunnel junctions, such as those implemented in magnetic random access memories or magnetic sensors, and more particularly to magnetic tunnel junctions having an out-of-plane orientation of magnetization. Background Art

[0002] Spin-transfer torque magnetic random-access memory (STT-MRAM) is one of the most promising emerging nonvolatile memory technologies. It combines nonvolatility, near-infinite write endurance, high speed, low power consumption, and scalability. These characteristics have spurred the commercialization of STT-MRAM for a variety of standalone and embedded applications, particularly NOR flash replacement and last-level cache (static RAM, or SRAM) replacement.

[0003] STT-RAM is based on a magnetic tunnel junction (or "junction" or "MTJ (magnetic tunnel junction)"), which includes a reference layer and a free layer separated by a tunnel layer. The reference layer has a magnetization with a fixed orientation, and the free layer has a magnetization with a switchable orientation. Today, the junction relies on the out-of-plane orientation of the magnetization in the free and reference layers (the out-of-plane orientation is considered to be relative to the plane of the free layer of the junction). It turns out that this leads to better scalability of the thermal stability factor Δ,

[0004]

[0005] Among them, E B is the energy barrier required to reverse the magnetization from the out-of-plane orientation to the opposite orientation, k B is the Boltzmann constant and T is the operating temperature. The thermal stability factor determines the storage retention time and affects the switching current required to switch the magnetization from one orientation to another.

[0006] A particular device, called a perpendicular magnetic tunnel junction (or "perpendicular junction" or "p-MTJ"), relies on exploiting the interfacial magnetic anisotropy present at the interface between a free layer (typically based on FeCo(B)) and an MgO layer (such as a tunnel barrier). The interfacial magnetic anisotropy tends to overcome the shape and orient the magnetization out of the plane of the free layer.

[0007] Despite this, further scaling these perpendicular junctions to below sub-20nm diameters presents fundamental challenges. As the junction diameter decreases, the thermal stability factor Δ, attributed to the perpendicular magnetic anisotropy at the interface, decreases. This can be understood as the perpendicular magnetic anisotropy being roughly proportional to the surface area of the interface between the MgO layer and the free layer.

[0008] A proposal to address this decreased stability is to add FeCoB / MgO interfaces to increase the total surface area providing perpendicular magnetic anisotropy. However, this approach is quite restrictive in terms of material selection, relies on the quality of several interfaces, and has a strong temperature dependence.

[0009] In order to maintain a high thermal stability factor at low technology nodes, the document [N. Perrissin & al., "A highly thermally stable sub-20nm magnetic random-access memory based on perpendicular shape anisotropy", Nanoscale 10, 12187-12195 (2018)] suggests taking full advantage of the shape anisotropy of the free layer and increasing its thickness to a value approximately equal to or greater than the diameter of the free layer. As a result, the shape anisotropy of the free layer no longer promotes easy-plane orientation, but further stabilizes the magnetization along the perpendicular orientation (relative to the plane of the free layer). This can be understood by considering the magnetostatic energy of the magnetized column, as expressed in [M. Beleggia et al., "Demagnetizing factors for cylindrical shells and related shapes", Journal of Magnetism and Magnetic Materials 321, 1306-1315 (2009)]:

[0010]

[0011] With a ratio of τ (L is the thickness of the free layer and r is its radius), is a hypergeometric function and m z is the defined orientation of the free layer magnetization with saturation magnetization M0 and vacuum permeability μ0.

[0012] Figure 1a shows an example of the magnetostatic energy of a free layer as a function of its aspect ratio (thickness relative to diameter), considering two different orientations of the magnetization (in the plane of the free layer and perpendicular to the plane). The results were obtained considering a uniform magnetization of 1 MA / m. Figure 1a shows a clear crossover between the in-plane orientation (black arrows and cross-hatched areas) and the out-of-plane orientation (white arrows and solid black areas). Furthermore, this crossover is geometry dependent, and the magnitude of the magnetization only plays a role in the magnetostatic energy. The total energy barrier can be expanded by taking into account the previously proposed magnetostatic energy and introducing the contribution of the interface anisotropy, which is given by the surface anisotropy term k s Definition, usually the value is 1.4mJ / m 2 In this way, sufficiently high thermal stability factors can be maintained for junctions with diameters of sub-10 nm. Figure 1b shows the results for a magnetization saturation of 1 MA / m and a surface anisotropy term k at an operating temperature of 300 K. s 1.4mJ / m 2 The thermal stability factor Δ of the free layer for different aspect ratios is shown in Figure 3. The results show that, for a fixed diameter, the thermal stability factor Δ can be increased by increasing the thickness of the free layer. This increase is even stronger with the addition of different anisotropy sources, opening up different possibilities for engineering such free layers.

[0013] The perpendicular shape anisotropy (also called “shape anisotropy” or “PSA”) provided by a thick free layer has attracted attention due to its scalability and wide temperature operating range. Figure 2A An example of a magnetoresistive device including a perpendicular junction 1 a is shown. Figure 2B An example of a magnetoresistive device comprising a shape anisotropic junction 1b (also referred to as a "PSA-MTJ") is shown. In both cases, the devices 1a, 1b are composed of a seed layer 11, a synthetic antiferromagnet 12, a subsequent texture breaker 13, and an out-of-plane magnetized reference layer 14. In each magnetic junction 1a, 1b, the reference layer 14 is separated from a thin free layer 16a by a tunnel barrier 15. Each thin free layer 16a exhibits an out-of-plane magnetization due to the interface effect between the thin free layer 16a and the tunnel barrier 15 in which it resides.

[0014] Figure 2B The knot 1b and Figure 2AThe junction 1a in FIG1 is different in that the junction 1b includes a ticker free layer 16b stacked on top of a thin layer 16a to promote shape anisotropy that allows for higher stability. Such additional layer 16b may be referred to as a "perpendicular shape anisotropy layer," "shape anisotropic layer," or "PSA (perpendicular shape anisotropy) layer," while the thin free layer 16a may be referred to as a "perpendicular interface anisotropic layer," "interface anisotropic layer," or "IA (interfacial anisotropy) layer."

[0015] Figure 3 A side view of two magnetic stacks 1c and 1d according to the prior art is shown, each of which includes a tunnel junction 1c and 1d. The stacks 1c and 1d rest on a substrate 191. Each tunnel junction in the stacks 1c and 1d includes at least a thick free layer (or shape anisotropy layer) to induce perpendicular shape anisotropy. The figure shows the steps for manufacturing a magnetic device (such as a magnetic memory or a magnetic sensor) that includes a tunnel junction. One of the tunnel junctions (corresponding to the stack 1d) is "bottom pinned" because the reference layer (and synthetic antiferromagnet) is located below the thick free layer. In other words, the reference layer is located between the tunnel layer and the substrate 191. The other tunnel junction (corresponding to the stack 1c) is "top pinned" because the reference layer (and synthetic antiferromagnet) is located above the thick free layer. This means that the thick free layer is located between the tunnel barrier and the substrate 191.

[0016] At this step in the manufacturing process, each magnetic stack 1c, 1d is also connected to a metal layer 192 that will serve as the bottom electrode. Figure 3 In the embodiment of the present invention, the metal layer 192 is a metal via that passes through the substrate 191 to connect to, for example, a circuit. At this step in the manufacturing process, each magnetic stack 1c, 1d also includes a hard mask 193 on its top portion. The hard mask 193 is used in the previous step to etch the layer to define the outline of the stack 1c, 1d and align the stack relative to the metal via 192. The hard mask 193 can also be used in further manufacturing steps to produce the final device (such as a memory for a sensor). Due to the thick free layer and hard mask 193, the magnetic stack 1c, 1d is relatively tall, for example, greater than 200nm.

[0017] They may also exhibit a high aspect ratio (measured as height divided by diameter), for example higher than 15. Consequently, the stacks 1c, 1d tend to fall, tilt or collapse during subsequent manufacturing steps (such as the trimming steps described below). This results in a large proportion of non-functioning tunnel junctions.

[0018] exist Figure 3, a complementary step of trimming the sides of the stacks 1c, 1d is shown. Trimming is performed to further reduce the diameter of the stack and increase the aspect ratio of the thick free layer (and improve its shape anisotropy). This allows tunnel junctions with a diameter of less than 20nm to be obtained, as shown in [N.Perrissin&al.,"A highly thermally stable sub-20nm magnetic random-access memory based on perpendicular shape anisotropy", Nanoscale 10, 12187-12195 (2018)]. Trimming can be performed by exposing the sides of the magnetic stacks 1c, 1d to a grazing ion beam 194. The grazing beam 194 is tilted at an angle of about 10° to 15° relative to the surface of the substrate 191. However, the increased height of the stacks 1c, 1d will produce large shadow areas 195 around each stack 1c, 1d because they will block the grazing ion beam 194. No other magnetic stack 1c, 1d can be made in the shaded area, as it will not be trimmed correctly. This implies a large point-to-point distance between adjacent stacks, which severely degrades the density of the final magnetic device obtained from the stacks 1c, 1d.

[0019] Therefore, there is a need to provide a robust magnetic stack that can be used in further manufacturing steps to produce a magnetoresistive device and that allows reducing the point-to-point distance between immediately adjacent stacks. Summary of the Invention

[0020] One aspect of the present invention relates to a method of manufacturing a magnetic device, the magnetic device comprising:

[0021] - a conductive electrode, the so-called "bottom electrode"; and

[0022] - a dielectric layer which is situated on the bottom electrode and has a through-hole (so-called "via") which exposes part of the bottom electrode.

[0023] The method comprises the following steps:

[0024] - forming a tunnel junction stack on the dielectric layer, the forming comprising the following sub-steps:

[0025] - forming a first magnetic layer intended to form said "free layer" having a switchable magnetization;

[0026] - forming a second magnetic layer intended to form said "reference layer" having a fixed magnetization; and

[0027] - forming an insulator layer intended to form a "tunnel barrier" that separates the first magnetic layer and the second magnetic layer from each other and is configured to allow spin transfer torque between the first magnetic layer and the second magnetic layer;

[0028] The first magnetic layer or the second magnetic layer is located in the via hole and is surrounded by a dielectric layer;

[0029] - A magnetic device is defined by a tunnel stack.

[0030] Switchable means that the magnetization of the free layer can be switched between different orientations, eg, reversed between two opposite orientations.

[0031] Fixed magnetization means that the magnetization orientation of the reference is set and does not switch between different orientations.

[0032] Such methods allow the fabrication of magnetic devices with magnetic tunnel junctions electrically connected to the bottom electrode. By depositing the magnetic layer of the junction using vias, the device ( Figure 3 Compared to the example of the present invention (e.g., a 1.5-μm CMOS process), the overall height of the magnetic device, measured from the dielectric layer, is reduced. The magnetic layer, and in particular the free layer, can exhibit a larger aspect ratio without increasing the height of the magnetic device. The magnetic device is also less likely to tilt or fall completely. Furthermore, due to the reduced height, the trimming step results in a narrower shadow zone around the device compared to prior art devices, and the point-to-point distance between adjacent devices can be reduced, thereby enabling high-density integration of magnetic devices.

[0033] Fabricating magnetic devices using vias is also compatible with complementary metal oxide semiconductor (CMOS) fabrication processes.

[0034] Advantageously, the step of defining the magnetic device by the tunnel stack comprises the following sub-steps:

[0035] - forming a hard mask on the tunnel stack, the hard mask being aligned with the via; and

[0036] - Etching the magnetic devices in the tunnel stack down to the dielectric layer and in line with the hard mask.

[0037] Advantageously, the formation of the tunnel junction stack includes depositing a conductive layer, for example made of TaN, on the inner wall of the via hole before forming the first magnetic layer or the second magnetic layer.

[0038] Advantageously, deposition of the conductive layer may also be performed on the exposed portions of the bottom electrode.

[0039] Advantageously, before depositing the conductive layer on the inner wall of the via hole, a non-conductive layer is deposited on the inner wall of the via hole, and the conductive layer is deposited on the non-conductive layer.

[0040] Advantageously, the layer in the first magnetic layer or the second magnetic layer located in the via hole is formed to reach the top of the via hole. The top of the via hole corresponds to the top surface of the dielectric layer, for example.

[0041] Advantageously, after the step of defining the magnetic device from the tunnel junction stack, the first magnetic layer has a magnetic anisotropy induced by its shape, which causes the magnetization of the first magnetic layer to spontaneously point out of the plane of the layer, for example, the plane of the top surface of the dielectric layer.

[0042] Advantageously, the step of forming the tunnel junction stack includes a sub-step of forming a third magnetic layer in contact with the insulator layer and having a magnetic anisotropy at the interface between the third magnetic layer and the insulator layer, which magnetic anisotropy causes the magnetization intensity of the third magnetic layer to spontaneously point out of the plane of the layer.

[0043] Advantageously, the method comprises the step of forming a ferromagnetic shell, said forming of the ferromagnetic shell being performed before the forming of the magnetic tunnel junction stack, the ferromagnetic shell being formed against the inner wall of the via and being embedded with a non-magnetic material.

[0044] A second aspect of the present invention relates to a magnetic device comprising:

[0045] - a conductive electrode, the so-called "bottom electrode";

[0046] a dielectric layer located on the bottom electrode and having a through hole (a so-called "via") exposing part of the bottom electrode; and

[0047] - a magnetic tunnel junction, the magnetic tunnel junction comprising:

[0048] - a magnetically free layer having a switchable magnetization;

[0049] a magnetic reference layer having a fixed magnetization; and

[0050] a tunnel barrier separating the free layer and the reference layer from each other and configured to allow spin transfer torque between the free layer and the reference layer;

[0051] A free layer or reference layer is located in the via and is surrounded by a dielectric layer.

[0052] "In a via" means located in a via.

[0053] Advantageously, the device comprises a conductive layer, for example made of TaN, on the inner wall of the via, which conductive layer separates the free layer or reference layer from the dielectric layer.

[0054] The present invention and its various applications can be better understood by reading the following description and examining the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] These figures are for reference only and are not intended to limit the present invention. Unless otherwise specified, identical elements appearing in different figures have unique reference numerals.

[0056] Figure 1A Examples of magnetostatic energy of the free layer for several aspect ratios of a magnetic tunnel junction according to the prior art are shown.

[0057] Figure 1B Thermal stability factors of the free layer for different aspect ratios of a magnetic tunnel junction according to the prior art are shown.

[0058] Figures 2A to 2B An example of a magnetic tunnel junction according to the prior art is shown.

[0059] Figure 3 A side view showing the steps of trimming the side surfaces of two magnetic stacks according to the prior art.

[0060] Figure 4 A first embodiment of a method of manufacturing a magnetic device according to the invention is shown.

[0061] Figure 5 Shows that it can be executed Figure 4 A base device of a first embodiment of the method.

[0062] Figures 6 and 7 Shows the formation Figure 5 Implementation method of the steps of the basic equipment.

[0063] Figures 8 to 14 A first implementation of the method according to the invention is shown for obtaining a first embodiment of the magnetic device according to the invention.

[0064] Figure 15 A second embodiment of a magnetic device is shown that can be obtained following a second embodiment of the method according to the invention.

[0065] Figures 16 to 21 A third implementation of the method according to the invention is shown for obtaining a magnetic device of the third embodiment.

[0066] Figure 22 A fourth embodiment of a magnetic device according to the invention is shown.

[0067] Figure 23 A fourth embodiment of a fabrication according to the invention is shown.

[0068] Figure 24 A fifth embodiment of a magnetic device according to the invention is shown.

[0069] Figure 25 A fifth embodiment of a fabrication according to the invention is shown.

[0070] Figure 26 A sixth embodiment of a magnetic device according to the invention is shown.

[0071] Figures 27 to 31 The calculation of the magnetic stray field is shown.

[0072] Figures 32 to 39 Shows the obtained Figure 26 A sixth embodiment of a magnetic device according to a sixth embodiment of the present invention is manufactured. DETAILED DESCRIPTION

[0073] Figure 4 shows the manufacture or production of Figures 13 to 15 An embodiment of the method of the magnetic device 4 is shown. Figures 5 to 12 Different steps or sub-steps of the method are shown.

[0074] Figure 5 A basic device capable of carrying out the method is shown. The basic device comprises a conductive electrode 2 and a dielectric layer 3. The conductive electrode 2 may be referred to as a "bottom electrode" because it is intended to be electrically connected to a magnetic tunnel junction 41 of a magnetic device 4. It is also intended to serve as an electrode of a final magnetoresistive device (such as a magnetic memory or a magnetic sensor) which may be obtained from the magnetic device 4 (e.g. after being connected to the top electrode). The bottom electrode 2 may be a metal wire made of, for example, W. It may also be electrically connected to a circuit or component (such as a transistor or a selector) located below the bottom electrode 2. The bottom electrode 2 may be located on a substrate or in a groove in said substrate (e.g. Figure 5 As shown in FIG, only the top surface thereof is exposed. It may also be a via hole filled with a metal material.

[0075] Dielectric layer 3 is located on bottom electrode 2. Dielectric layer 3 can be embedded in bottom electrode 2 (when bottom electrode 2 is located on the substrate) or can be located on the top surface of bottom electrode 2 (when bottom electrode 2 is located in a trench in the substrate). In the latter case, dielectric layer 3 can be located on the substrate, embedding a portion of bottom electrode 2. Dielectric layer 3 is a non-magnetic, insulating material. It can be made of SiO2 or SiN. It can also be made of a polymer that can withstand chemical and mechanical polishing.

[0076] The dielectric layer 3 has a top surface 32 which is for example a free surface where the manufacturing steps will be carried out. The top surface 32 can also define a plane P called "layer plane".

[0077] The dielectric layer 3 includes a through hole 31, called a "via". The via 31 passes through the entire dielectric layer 3 from the top surface 32 to the bottom electrode 2. The via 31 exposes a portion 21 of the bottom electrode 2, in this case a portion 21 of the top surface of the electrode 2. The via 31 can have a straight shape or a truncated flared shape. The via 31 has an inner wall 310, also called a side wall. In the straight shape, the inner wall 310 is perpendicular (+ / - 10°) relative to the plane P of the layer. In the truncated flared shape, as shown in Figure 1, the inner wall 311 is inclined relative to the direction perpendicular to the plane P of the layer. For example, the inner wall is inclined at an angle greater than or equal to 10°, for example equal to 15° or 20°. This specific shape can be provided by the drilling method adopted, in particular for vias 31 with a diameter of less than 50 nm.

[0078] Figure 6 and Figure 7 Shows the formation Figure 5 , which is an embodiment of the steps of the basic device shown in . Starting from a bottom electrode 2 located in a trench in the substrate, the formation S1 comprises a sub-step of depositing S11 a dielectric layer 3, for example using chemical vapor deposition (also called "CVD (chemical vapor deposition)") or physical vapor deposition (also called "PVD (physical vapor deposition)"). The deposition can be performed in a blanket manner, and chemical and mechanical polishing (also called "CMP (chemical and mechanical polishing)") can be performed to obtain a dielectric layer 3 with a well-controlled thickness Z3 (measured perpendicular to the plane P of the layer). The thickness Z3 of the dielectric layer can be between 10 nm and 100 nm, for example equal to 30 nm.

[0079] After the deposition of the dielectric layer 3, a drilling step S12 may be performed, which creates the via 31. An embodiment of the drilling step S12 may include forming a sacrificial protruding before the deposition S11 of the dielectric layer 3, and performing a specific etching on the sacrificial protruding after exposing the sacrificial protruding by CMP.

[0080] The vias 31 may also be drilled S12 using, for example, electron beam lithography after deposition S11 of the dielectric layer 3. The drilling step S12 may also rely on self-assembly or any other process providing organized vias 31. Electron beam lithography typically provides the vias 31 with a flared or truncated flared shape.

[0081] Figures 8 to 12 Shown by Figure 5 The basic device shown in FIG. 3 includes a step S31 of forming a tunnel stack 410 on a dielectric layer 3. The forming step S3 includes a sub-step S31 of forming a first magnetic layer 4110. Figure 8 As shown in FIG. In this figure, the first magnetic layer 4110 fills the via 31 and reaches its top (corresponding to the top surface 32 of the dielectric layer 3). Thus, the magnetic layer 4110 is bounded by the inner walls of the via to form a "free layer" 411 (or "storage layer" if the magnetic device 4 is intended to form a magnetic storage device). A notable feature of the free layer 411 is that it has magnetization; however, the orientation of the magnetization is not fixed. It can be reversed.

[0082] At this stage of the manufacturing process, the magnetization of the resulting free layer 411 is still free to behave. However, it is preferred to constrain the magnetization of the free layer 411 to be aligned in a predetermined direction or at least out of plane (such as the plane P of the layer). Such constraint is obtained by adjusting the anisotropy induced in the free layer 411. For example, the diameter D of the first magnetic layer can be adjusted. 411 and thickness Z 411 , to provide perpendicular shape anisotropy, which causes the magnetization of layer 411 to point out of the plane P of the layer. The deposition method used to form free layer 411 may also add some bulk contribution to the anisotropy, such as magnetocrystalline anisotropy. If the anisotropy favors the magnetization, the magnetization will preferably point out of the plane spontaneously. Due to the weak in-plane shape anisotropy contribution, the magnetization may also be slightly tilted and exhibit a weak in-plane vector.

[0083] The first magnetic layer 411 has a thickness Z measured perpendicular to the plane P of the layer. 411 , which is at most equal to the thickness Z3 of the dielectric layer 3. It can also have a diameter D at most equal to the diameter of the via 31 31 Diameter D 411 The diameter D 31 、D 411 It may depend on the height considered in the via 31 (especially when the via has a flared shape).

[0084] exist Figure 8In the embodiment, a free layer 411 is formed in the via 31. For example, the free layer is obtained by depositing the first magnetic layer 4110 by CVD, PVD, or electrochemical deposition. The magnetic material used can be Fe, Co, Ni, or any alloy thereof. Non-magnetic materials can also be added to form a magnetic multilayer structure, such as a magnetic / non-magnetic / magnetic multilayer structure.

[0085] The magnetic material is deposited to completely fill the via hole 31 and overcome the top of the via hole 31, as shown in FIG. Figure 10 In other words, the thickness of the deposited magnetic material measured from the bottom of the via hole 31 is greater than the thickness Z3 of the via hole 31 .

[0086] Figure 9 and Figure 10 An embodiment of the sub-step of forming S31 the first magnetic layer 4110 is shown. A conductive layer 3111 may be deposited on the inner wall 310 of the via 31 before depositing the first magnetic layer 4110. The conductive layer 3111 may be conformally deposited on the inner wall 310 of the via 31. The conductive layer 3111 may be deposited by CVD, PVD, or any method that allows for conformal deposition, such as atomic layer deposition (ALD) or electroplating.

[0087] It can be made of a non-magnetic material such as TaN. If the conductive layer 3111 is non-magnetic, it will reduce the diameter D of the resulting free layer 411. 411 (measured parallel to the plane P of the layer). Thus, it helps to increase the aspect ratio (thickness Z 411 Divide by the diameter D 411 The reduction in diameter can be controlled by the thickness (measured perpendicularly relative to the surface after deposition) of the conformally deposited material (the material used to form the conductive layer 3111).

[0088] If the conductive layer 3111 is magnetic, for example, made of Co, the resulting diameter D of the free magnetic layer 411 is 411 Equal to the diameter D of the via 31 31 In other words, the resulting free magnetic layer 411 includes the magnetic conductive layer 3111 .

[0089] exist Figures 4 to 14 In the embodiment of FIG, a conductive layer 3111 is also deposited on the exposed portion 21 of the bottom electrode 2, which is located at the bottom end of the via 31. This can improve the electrical contact with the bottom electrode 2. However, in this case, adjusting the thickness of the conductive layer 3111 may not change the aspect ratio of the resulting free layer 411, because by reducing the diameter, the height will be reduced in the same manner.

[0090] The first magnetic material 4110 may be deposited on the conductive layer 3111, such as Figure 10 Preferably, the via 31 is filled, for example, by deposition in a blanket manner. CMP can be used to remove excess magnetic material and reach the top 32 of the via 31. CMP can also provide a clean surface without magnetic material between adjacent magnetic devices.

[0091] CMP, which may be performed after deposition of the first magnetic layer 411, tends to provide little roughness and thus good texture for further deposited layers 4140, 4130, 4120. However, CMP may induce some dishing of the free layer 411, which may reduce its thickness Z. 411 , and especially its magnetic thickness (which may be smaller than the geometric thickness). The thickness Z of the via 3 should be adjusted accordingly 31 and the thickness of dielectric layer 3 to balance this recessing effect.

[0092] Figure 11 Two sub-steps are shown of forming S32 a second magnetic layer 4120 and forming S33 an insulator layer 4130 on top of the free layer 411. The second magnetic layer 4120 is intended to form a "reference layer" 412 (which is effectively formed after its definition). The reference layer 412 has a magnetization with a fixed orientation. The insulator layer 4130 is intended to form a "tunnel barrier" 413. For example, the insulator layer 4130 is configured to allow spin transfer torque between the free layer 411 and the second magnetic layer 4120. To this end, the insulator layer 4130 is located between the free layer 411 and the second magnetic layer 4120, thereby completely separating these layers. Therefore, in Figure 11 In this embodiment, the sub-step of forming the insulator layer 4130 is performed before the sub-step of forming the reference layer 4120. The insulator layer 4130 can be deposited S33 on the free layer 411 and preferably extends away from the top surface of the free layer 411. The insulator layer can be deposited in a blanket manner to cover a large area around the free layer 411. It can be deposited by CVD or PVD. It can be made of a nitride (such as TiN) or an oxide (such as MgO, Al2O3 or TiO2). The thickness of the insulator layer 413 is preferably between 0.6nm and 6nm and more preferably between 1nm and 2nm.

[0093] The thickness of the insulator layer 4130 allows control of a product called the R×A product, which is the resistance of the junction 41 multiplied by the area of the tunnel barrier 413 obtained by the insulator layer 4130 (e.g., after the definition step). For small diameter magnetic junctions 41, the R×A product is selected to be less than 10Ω.μm. 2 Therefore, the thickness of the insulator layer 4130 is selected to provide a sub-10Ω.μm resistance of the junction 41.2 R×A product.

[0094] To achieve a high tunnel magnetic response (so-called "tunnel magnetic response") (e.g., about 200%), the magnetic device 4 may include a ferromagnetic layer 414, a so-called "interface magnetic layer," which may promote a specific crystal structure of the tunnel barrier 413, such as a body-centered cubic lattice with a (100) texture. The interface magnetic layer 414 is, for example, in direct contact with the reference layer 412 on one side and with the interface magnetic layer 414 on the other side.

[0095] The interface layer 414 can be made of an FeCo alloy or FeCo(B) alloy that is amorphous when deposited. However, the interface layer 414 can crystallize into a body-centered cubic lattice with a (100) texture when annealed, typically in a range between 250°C and 400°C. The crystallization of the interface layer 414 tends to force the tunnel barrier 413 to crystallize according to the same structure. The annealing can be performed after the device 4 is fabricated or at least before the hard mask is deposited.

[0096] The interface layer 414 is as follows Figure 13 shown. Figure 11 and Figure 12 A method of forming the interface layer 414 is shown. For example, a third magnetic layer 4140, intended to form the interface layer 414, can be deposited in direct contact with the insulator layer 4130. The third magnetic layer 4140 can be deposited first on the free layer 414, and then the insulator barrier 4130 can be deposited. The insulator layer 4130 thus separates the third magnetic layer 4140 from the second magnetic layer 4120.

[0097] When the tunnel barrier 413 is made of MgO, the interface layer 414 can also induce interface anisotropy at the interface between the interface layer 414 and the tunnel barrier 413, especially when the interface layer 414 is made of FeCo or FeCo(B).

[0098] The free layer 411 and the interface layer 414 can be stacked directly one above the other to promote strong coupling between their respective magnetizations. This ensures that the generated magnetization exhibits a unique and rigid magnetization (also known as a "macrospin"). However, the free layer 411 and the interface layer 414 can also be separated by a thin layer that is suitable for the crystallographic structure of the magnetic layers 411n414. However, the thickness is preferably selected to maintain the macrospin behavior of the generated magnetization.

[0099] When amorphous elements are removed from the interface between the interface layer 414 and the tunnel barrier 413, the crystallization of the interface layer 414 and the tunnel barrier 413 can be assisted. To this end, a structural transition layer (also called an "amorphizing-getter layer" or "B-getter layer" because it tends to capture B elements from the magnetic layer, which is a common amorphous element) can be inserted between the free layer 411 and the interface layer 414. For example, the interface layer 414 can be made of FeCo(B). In this case, the B-getter layer is located between the free layer 411 and the interface layer 414. The B-getter layer is in direct contact with at least the magnetic layer made of FeCo(B). The B-getter layer can be made of W, Ta, Mo or Hf. The thickness of the B-getter layer is preferably between 0.1 nm and 0.3 nm.

[0100] The second magnetic layer 4120 is deposited S32 on top of the insulator layer 4130 by CVD, PVD.

[0101] Figure 12 and 13 The step S5 of defining the magnetic device 4 from the tunnel stack 410 is shown. This step includes, for example, the sub-step of forming S51 a hard mask 42 on the tunnel stack 410 (e.g., Figure 12 As shown), and a sub-step of etching S52 the tunnel junction stack 410 not hidden by the hard mask 42 to the dielectric layer 3.

[0102] Hard mask 42 can be a metal protrusion formed on top of tunnel junction stack 410. It can be made of SiN, Ta, TaN, or any other conductive material that can withstand ion etching. Its shape induces a columnar shape that is aligned with the hard mask 42, which is caused by etching the material not hidden under the hard mask 42. Therefore, the hard mask is preferably aligned with the free layer 411 and, therefore, with the via 31. The remaining material of the tunnel junction stack 410 and the free layer 411 form a tunnel junction 41.

[0103] Etching S52 can be performed using a reactive ion etching method and is preferably performed in a direction perpendicular to the plane P of the layer (also referred to as the "vertical direction"). Etching is preferably performed until it reaches the top surface 32 of the dielectric layer 3. As a result, redeposition of the etched material may occur, thereby trimming the side of the portion of the magnetic device 4 that is exposed from the dielectric layer 3. The trimming can be performed using ion beam trimming at a grazing angle.

[0104] The hard mask 41 is preferably a conductive layer so that it can be connected to a conductive electrode, such as a top electrode.

[0105] Figure 14A further manufacturing step is shown in which a complementary dielectric material 5 is deposited so that it surrounds the portion of the magnetic device 4 where the dielectric layer 3 is exposed. This complementary dielectric material ensures that no direct electrical connection occurs between the conductive layers. The complementary dielectric material 5 can be made of SiN or a polymer. The complementary dielectric material can also be made of the same material from which the dielectric layer 3 with the via 31 is formed, if this same material allows for planarization.

[0106] Figure 15 Another embodiment of a magnetic device 4 is shown, which can be obtained by following another embodiment of the method of the present invention. This embodiment differs from the previous embodiment in that the conductive layer 3111 is conformally deposited only on the inner wall 310 of the via 31. The exposed portion 21 of the bottom electrode does not contain the conductive layer 3111. Therefore, the free layer 411 is in direct contact with the bottom electrode 2. In addition, the conductive layer 3111 on the inner wall 310 has been deposited with a greater thickness (measured perpendicular to the inner wall 310). This further reduces the diameter D of the free layer 411. 411 and increases its aspect ratio (Z 411 / D 411 ).

[0107] In order to induce perpendicular anisotropy, the free layer 411 may have magnetic anisotropy induced by its shape. The free layer is made of, for example, Fe, Ni, or Co or any alloy thereof. The diameter of the free layer may be determined by the diameter D of the via hole 31. 31 Thus, the free layer may have a thickness such that it exhibits an aspect ratio that tends to spontaneously orient its magnetization out of the plane P of the layer. The aspect ratio of the first sublayer is, for example, greater than 0.5 and preferably comprised between 0.7 and 1.5.

[0108] Figures 16 to 21 Another embodiment of a method of manufacturing or producing a magnetic device 4 is shown. Figures 16 to 18 The steps shown differ from the previous two embodiments in that a non-conductive layer 3112 is deposited on the inner wall 310 of the via hole 31 before depositing a conductive layer 3111 on the inner wall 310 of the via hole 31. The conductive layer 3111 is then deposited on the non-conductive layer 3112.

[0109] The non-conductive layer 3112 is formed by, for example, CVD, PVD, ALD or electroplating conformally (obtained Figure 16 Preferably, the non-conductive material 3112 at the bottom of the via 31 is removed in order to achieve good electrical contact with the bottom electrode 2. To this end, the non-conductive material 3112 can be subjected to vertical etching to remove its portion parallel to the plane P of the layer, in particular its portion at the bottom of the via 31 (as shown). Figure 17Then, the conductive layer 3111 is conformally deposited on the non-conductive layer 3111 (as shown). Figure 18 shown).

[0110] The non-conductive layer 3112 further reduces the diameter D of the via hole. 31 , which helps to reduce the diameter D of the free layer 411 411 and increase its aspect ratio. The reduction in diameter is related to the thickness of the non-conductive layer 3112 (measured perpendicular to the surface on which it is located). The thickness of the conformal layer of non-conductive material (and therefore the reduction in diameter) is primarily controlled by the conformal deposition process used, and can even be controlled by the vertical etch used to remove the bottom.

[0111] After forming the non-conductive layer 3112, the first magnetic layer 4110 may be formed on the conductive layer 3111 as described above. Figures 19 to 20 shown. Figure 21 The magnetic device 4 shown and produced in Figure 14 The magnetic device in is different in that the free layer 411 is narrower and, therefore, can exhibit a larger aspect ratio (and higher shape anisotropy).

[0112] Figure 22 An embodiment of a magnetic device 4 that can be obtained by following another embodiment of the method is shown. This embodiment differs from the previous embodiment in that the second magnetic layer 4120 is formed first and the via 31 is filled. Thus, the via 31 defines a reference layer 412 (rather than the free layer 411), which is therefore located in the via 31. The reference layer 412 is surrounded by the dielectric layer 3. Since the reference layer 412 is located below the free layer 411, the magnetic device 4 is "bottom-pinned."

[0113] The resulting magnetization of the reference layer 412 should exhibit a fixed orientation. To this end, the thickness of the dielectric layer 3 can be greater than that of the dielectric layer in the previous embodiment. The resulting high aspect ratio of the reference layer 412 (calculated as Z3 / D 31 ) will induce a strong shape anisotropy, which can pin the magnetization of the reference layer 412 in a fixed direction (perpendicular to the plane P of the layer). The aspect ratio of the reference layer 412 is preferably tailored so that the magnetization of the free layer 411 will flip before the magnetization of the reference layer 412. The aspect ratio of the reference layer 412 is, for example, higher than that of the free layer 411. If the free layer 411 and the reference layer 412 exhibit the same diameter, the thickness Z3 of the dielectric layer 3 should be greater than the thickness Z 411 For example, the reference layer 412 can be made of Co. It can also exhibit an aspect ratio between 0.8 and 3 (calculated as Z3 / D 31 ).

[0114] To reduce the thickness Z3 of dielectric layer 3, a third magnetic material 4140 can be deposited between second magnetic layer 4120 and insulator layer 4130. This can induce interface anisotropy (relative to the plane P of the layer), which can enhance the overall anisotropy of the interface and reference layers 414, 412. For example, interface layer 414 can be made of an Fe alloy (such as FeCo or FeCo(B)), and when the interface layer is made of MgO, it is in direct contact with tunnel barrier 413. Interface anisotropy can help reduce the restrictions on the aspect ratio of second magnetic layer 412 and, therefore, help reduce the thickness of dielectric layer 3.

[0115] Figure 23 Another embodiment of the manufacturing method is shown, which helps to reduce the thickness Z3 of the dielectric layer 3 even further, as Figure 24 In this embodiment, the reference layer 412 is strongly coupled to the synthetic antiferromagnet 43. The synthetic antiferromagnet 43 is, for example, located in the via 31 and the reference layer 412. For example, the synthetic antiferromagnet is located between the reference layer 412 and the bottom electrode 2. The synthetic antiferromagnet 43 may include two [Co / Pt]n multilayers (where n is the number of Co / Pt layers) that are exchange-coupled, thanks to the Ru layer. Thanks to the presence of the Ru layer, the synthetic antiferromagnet 43 can also be coupled to the reference layer 412.

[0116] Figure 23 An embodiment of the method includes, before the sub-step S32 of forming the second magnetic layer 4120, the sub-step of forming S34 a synthetic antiferromagnet 43 in the via 31 when the reference layer 412 is intended to be located in the via 31. When the synthetic antiferromagnet is a [Co / Pt]n multilayer, the synthetic antiferromagnet 43 can be deposited using electrodeposition.

[0117] Figure 22 The embodiment of exhibits a larger thickness Z3 of the dielectric layer 3. However, the reference layer 412 does not rely on the synthetic antiferromagnet 43. It thus allows a departure from the use of scarce materials such as Co, Pt or Pd.

[0118] like Figure 25 As shown, another embodiment of the manufacturing method allows forming Figure 26 The magnetic device 4 of the embodiment, wherein the stray magnetic field by the reference layer 412 is reduced, as Figures 27 to 31 shown.

[0119] exist Figure 26In an embodiment, a ferromagnetic shell 44 surrounds a reference layer 412 located in the via 31. To reduce stray fields, the ferromagnetic shell 44 is antiferromagnetically coupled to the reference layer 412. To this end, the ferromagnetic shell 44 and the reference layer 412 are separated from each other by a non-magnetic layer 3113 (also referred to as a "spacer"). The non-magnetic spacer 3113 can be an insulator or a metal layer. When the non-magnetic spacer 3113 is an insulator, the coupling between the reference layer 412 and the ferromagnetic shell 44 is purely magnetostatic. When the non-magnetic spacer 3113 is a metal, particularly Ru, the coupling between the reference layer 412 and the ferromagnetic shell 44 can be exchange coupling and magnetostatic coupling. The thickness of the non-magnetic spacer 3113 (measured perpendicular to the surface on which it is located) can define the coupling strength and / or coupling properties (i.e., the magnetizations of the reference layer 412 and the shell 44 are parallel ferromagnetically coupled, or the magnetizations are antiparallel antiferromagnetically coupled). Therefore, the thickness is chosen such that an antiferromagnetic coupling is obtained between the ferromagnetic housing 44 and the magnetization of the reference 412 .

[0120] To balance the stray fields of the reference layer 412, the ferromagnetic shell 44 should exhibit a magnetic volume (defined as the volume of its magnetic material multiplied by the saturation magnetization of the magnetic material) that is as close as possible to the magnetic volume of the reference layer 412, or at least between 50% and 150% of the magnetic volume of the reference layer 412. For example, the thickness and / or width of the ferromagnetic shell 44 can be selected so that its magnetic volume is between 50% and 150% of the magnetic volume of the reference layer 412.

[0121] The field compensation is also better if a ferromagnetic shell 44 surrounds the reference layer 412 in the via 31 over most of the thickness of said reference layer 412. The thickness of the ferromagnetic shell 44 is, for example, equal to the thickness of the reference layer 412. For example, it can be equal to the thickness Z3 of the dielectric layer 3.

[0122] In an alternative embodiment, the free layer 411 may be located in the via 31 instead of the reference layer 412 , and the ferromagnetic housing 44 thereby compensates for the stray fields of the free layer 411 .

[0123] Figures 27 to 31 The high aspect ratio reference layer 412 (eg, Figure 22 Calculated stray field lines (shown). Figure 27 Only the stray field of the reference layer 412 is shown in FIG. Figure 28 Only the stray fields of the ferromagnetic housing 44 are shown, and Figure 29 , the stray field of the reference layer 412 surrounded by the ferromagnetic housing 44 is shown. Figure 29 In FIG. 4 , the stray field lines generated by the housing 44 and the reference layer 412 are shown in opposite directions, which can mitigate the total stray field.

[0124] Figure 30 and Figure 31 The reference layer 412 is shown without the shell ( Figure 30 ) and the reference layer 412 with the shell ( Figure 31 ) in the free layer 411 (black square in line with the reference layer 412). In this example, the reference layer 412 has a thickness of 30 nm and a diameter of 5 nm, with a magnetization saturation of 1 MA / m. The reference layer 412 does not contain Pt / Pd and has a fixed magnetization due to its high aspect ratio shape anisotropy. The ferromagnetic shell 44 is separated from the reference layer 412 by 1 nm and has a total width of 10 nm, with a magnetization saturation of 1.446 MA / m. Since the two objects are dipole-coupled, they act as a single object, thereby further improving stability. Iso-fields of 100 mT and 200 mT are also shown. It can be observed that the stray fields are completely mitigated by using the ferromagnetic shell 44. The stray field from the reference layer 412 with the shell is less than 100 mT at the free layer 411. This result also demonstrates that, thanks to the ferromagnetic shell 44 , a Pt-free or Pd-free magnetic tunnel junction can exhibit low stray fields, thus allowing for further high-density integration.

[0125] Figures 32 to 38 Shows the obtained Figure 26 The steps of another embodiment of the method for manufacturing an embodiment of the magnetic device 4 are as follows.

[0126] First, the via 31 is preferably wider to allow enough space to form the ferromagnetic shell 44 therein. The diameter of the via 31 is, for example, 20 nm or even wider. Conformal deposition of the dielectric material 3111 can be achieved using chemical-based techniques such as PVD, CVD, ALD, or electroplating. Figure 32 ). The conformal deposition of the dielectric material 3111 is performed with a thickness (measured perpendicular to the surface 311 on which it is located) that enables the diameter of the ferromagnetic shell 44 and the magnetic layers 411 , 412 surrounded by the shell 44 to be controlled.

[0127] Conformal deposition of the ferromagnetic layer 440 is achieved (e.g. Figure 33 As shown). The ferromagnetic material can be conductive or non-conductive. It can also utilize the chemical-based techniques mentioned above. The thickness of the layer 440 is selected to optimize stray field mitigation. The ferromagnetic material used is, for example, Ni or Fe or any alloy thereof, such as Ni 20 Fe 80 or Ni 80 Fe 20 .

[0128] A vertical etching is performed to remove a portion of the ferromagnetic layer 440 and the dielectric layer 3111 at the bottom of the via hole 31 (eg, Figure 34 ). It exposes the bottom electrode 2 in contact with the tunnel junction 41. It also forms a ferromagnetic shell 44 having, for example, a cylindrical shape.

[0129] The non-magnetic spacer 3113 is conformally deposited on the ferromagnetic shell 44 ( Figure 35 ). The spacer 3113 may be an insulator material such as ZnO, HfO2, Al2O3 or TiN. It may also be a metal material that allows exchange coupling between the housing 44 and the layer intended to be located in the via 31. The metal material is, for example, Ru. In the case where the non-magnetic spacer 3113 is an insulator, vertical etching ( Figure 36 ) to expose the bottom electrode 2 and allow good electrical contact.

[0130] The thickness of the nonmagnetic spacer 3113 allows control of the coupling strength between the housing 44 and the magnetic layers 411, 412 in the via 31. It also allows control of the diameter reduction in the via 31, which can increase the aspect ratio of the magnetic layers 411, 412 in the via 31.

[0131] exist Figure 37 In the example of FIG. 4 , a second magnetic layer 4120 is deposited to fill the remaining empty space in the via 31. Chemical and mechanical polishing ( Figure 38 ) to form the reference layer 412 and allow further formation of the remaining layers of the magnetic device 4 ( Figure 39 ).

[0132] In the case where the non-magnetic spacer 3113 is metal, no vertical etching is required, and the reference layer 412 can be deposited in the remaining space of the via 31, as shown in FIG. Figure 39 In this figure, the metal spacer 3113 is conductive and allows electrical connection between the bottom electrode 2 and the reference layer 412.

Claims

1. A method for manufacturing a magnetic device (4), the magnetic device comprising: - a conductive electrode (2), the so-called "bottom electrode"; a dielectric layer (3) situated above the bottom electrode and having a through hole (31), a so-called "via", exposing a portion (21) of the bottom electrode (2), The method comprises the following steps: - forming (S3) a tunnel junction stack (410) on the dielectric layer (3), said forming comprising the following sub-steps: - forming (S31) a first magnetic layer (4110) intended to form said "free layer" (411) having a switchable magnetization; - forming (S32) a second magnetic layer (4120) intended to form said "reference layer" (412) having a fixed magnetization; and - forming (S33) an insulator layer (4130) intended to form a "tunnel barrier" (413) separating the first magnetic layer (4110) and the second magnetic layer (4120) from each other and configured to allow spin transfer torque between the first magnetic layer (4110) and the second magnetic layer (4120); The first magnetic layer (4110) or the second magnetic layer (4120) is located in the via hole (31) and is surrounded by the dielectric layer (3); - Delimiting (S5) the magnetic device (4) from the tunnel stack (410).

2. The method according to claim 1, wherein Defining the magnetic device (4) from the tunnel stack (410) comprises the following sub-steps: - forming (S51) a hard mask (42) on the tunnel stack (410), the hard mask (42) being aligned with the via (31); as well as - etching (S52) the magnetic device (4) in the tunnel stack (410) down to the dielectric layer (3) and in line with the hard mask (42).

3. A method according to any one of the preceding claims, wherein The formation (S3) of the tunnel junction stack (41) includes depositing a conductive layer (3110) on the inner wall (310) of the via hole (31) before forming the first magnetic layer (4110) or the second magnetic layer (4120).

4. The method according to claim 3, the deposition of the conductive layer (3111) being also performed on the exposed portion (21) of the bottom electrode (2).

5. The method according to claim 3 or 4, wherein: Before depositing the conductive layer (3111) on the inner wall (310) of the via hole (31), a non-conductive layer (3112) is deposited on the inner wall (310) of the via hole (31), and the conductive layer (3111) is deposited on the non-conductive layer (3112).

6. A method according to any one of the preceding claims, wherein The layer of the first magnetic layer (4110) or the second magnetic layer (4120) located in the via hole (31) is formed to reach the top (32) of the via hole (31).

7. A method according to any one of the preceding claims, wherein After the step of defining the magnetic device from the tunnel junction stack, the first magnetic layer (4110) has a magnetic anisotropy induced by its shape, which causes the magnetization of the first magnetic layer to spontaneously point out of the plane (P) of the layer.

8. A method according to any one of the preceding claims, wherein The step of forming the tunnel junction includes a sub-step of forming a third magnetic layer (4140), wherein the third magnetic layer is in contact with the insulator layer (4130) and has magnetic anisotropy at the interface between the third magnetic layer and the insulator layer (4130), and the magnetic anisotropy causes the magnetization intensity of the third magnetic layer to spontaneously point outward from the plane (P) of the layer.

9. The method according to any one of the preceding claims, comprising the step of forming (S2) a ferromagnetic shell (44), said forming (S2) of said ferromagnetic shell being performed before said forming (S3) of said tunnel junction (41), said ferromagnetic shell being formed against said inner wall (311) of said via (31) and being embedded with non-magnetic material.

10. A magnetic device (4), comprising: - a conductive electrode (2), the so-called "bottom electrode"; a dielectric layer (3) situated above the bottom electrode and having a through hole (31), a so-called "via", exposing a portion (21) of the bottom electrode (2); as well as - a magnetic tunnel junction (41), said magnetic tunnel junction comprising: - a magnetically free layer (411) having a switchable magnetization; - a magnetic reference layer (412) having a fixed magnetization; and a tunnel barrier (413) separating the free layer (411) and the reference layer (412) from each other and configured to allow spin transfer torque between the free layer (411) and the reference layer (412); The free layer (411) or the reference layer (412) is located in the via hole (31) and is surrounded by the dielectric layer (3).

11. The magnetic device (4) according to claim 10, comprising a conductive layer (3111) located on an inner wall of the via (31), the conductive layer separating the free layer (411) or the reference layer (412) from the dielectric layer (3).

Images