Patent Application
20110236704
The present invention relates to a process for fabricating antiferromagnetic layers, and more particularly those that are used in spintronics.
The materials owe their magnetic properties to the fact that certain atoms have one or more atomic sublayers having a single electron whose magnetic spin is not cancelled out by the opposed spin of another electron. Most of these materials have several single electrons, for which the algebraic sum of the elementary magnetic moments is not zero.
Four main categories of magnetic materials may be distinguished:
The first category is formed by ferromagnetic and ferrimagnetic materials. The latter are characterized in that the magnetic moment of an atom is strongly coupled with the magnetic moment of neighboring atoms by exchange coupling, which tends to align in a same direction the magnetic moments of all the atoms inside a same magnetic domain (called Weiss domain). For ferromagnetic materials, each of these atoms magnetized in a same direction has the same magnetization intensity. The magnetic behavior of ferrimagnetic materials is very close to that of ferromagnetic materials. Here also, the magnetic moments of atoms of a same domain are in a same direction, but in ferrimagnetism, the peripheral electrons are distributed differently between the two spins when one passes from one atom to another, such that the magnetization intensity varies according to each atom. However, in the two cases (ferrimagnetic and ferromagnetic), the existence of magnetic domains and their formation are governed by the same laws: Consequently, either ferrimagnetic materials or ferromagnetic materials will be referred to in the rest of the description.
When they are not saturated but are in a disordered state or are weakly magnetized, the ferromagnetic materials are thus constituted of a plurality of magnetic domains (Weiss domains) separated between each other by magnetic walls (for example, Bloch walls): A magnetic domain is a magnetic microstructure in which the magnetic moments are all oriented in a same direction. Magnetic domains have irregular shapes, whose dimensions are on the order of some hundreds of nanometers, or even a micron, and the magnetization is very intense. The magnetic orientations of two juxtaposed domains are initially poorly coupled, which causes magnetic noise when a spin current flows through the material. In fact, each electron traversing a magnetic domain undergoes a spin transfer depending on the difference between its magnetic orientation and that of the domain under consideration.
For hard layers of ferromagnetic material, the algebraic sum of magnetic moments of all domains has a fixed non-zero value determining its macroscopic magnetization. Subjected to an external magnetic field, these materials align their magnetic domains in the direction of the external field. The more intense this field, the more numerous the magnetic domains that orient themselves along its direction, until saturation, that corresponds to the alignment of all magnetic domains in the direction of the external field. Hard ferromagnetic materials have an atomic structure that makes a random reorientation of magnetic domain magnetizations after removal of the external magnetic field difficult. All of these magnetic properties reversibly disappear under the effect of thermal agitation beyond the Curie temperature. It will be noted that the stability of these hard layers may be ensured by its form and/or by exchange coupling with an antiferromagnetic layer.
The second category of magnetic materials is constituted of diamagnetic materials characterized in that almost all of the atoms do not have an atomic sublayer with a single electron; For each sublayer, the magnetic moment created by an electron is thus cancelled out by the magnetic moment of the electron matching it. The resulting magnetic moment for each atom has an initially random direction, but zero intensity. No magnetic coupling exists between two neighboring atoms. However, when such a material is subjected to an external magnetic field, the magnetic moment of each atom tends to very slightly orient itself in the opposite direction from this field, progressively forming, as the field intensity increases, magnetic domains. Their magnetization intensity remains much less than the magnetization of a ferromagnetic material; moreover, it is not possible to reach saturation.
The third category relates to paramagnetic materials that are characterized in that their atoms have atomic sublayers with at least one single electron. However, no coupling between two neighboring atoms or long distance magnetic order exists. When they are subjected to an external magnetic field, the magnetic moment of each atom tends to very slightly orient itself in the direction of this field, progressively forming, as the field intensity increases, magnetic domains. Their magnetization intensity remains much less than the magnetization of a ferromagnetic material and no remanence is observed after exposure to an external field. Again, reaching saturation is thus not at all possible.
The fourth category of magnetic materials is that of antiferromagnetic materials. Their atoms have saturated layers, whose spin magnetic moments cancel themselves two by two. Their magnetic moment has a completely ineffective intensity, to the point of cancelling any interaction with an external magnetic field. Nevertheless, they have an antiferromagnetic structure characterized by the ordering into two subnetworks with opposed magnetization, whose result is zero. Nevertheless, the subnetworks are organized into magnetic domains, called. Néel domains, that separate the regions where the antiferromagnetic order has nucleated according to crystallographic orientations that are different as well as equivalent in symmetry. Without intervention other than the growth of the material, these domains are naturally expected to be smaller (by one to two orders of magnitude) than the Weiss domains of ferromagnetic materials. In the particular case of thin layers (from 1 to 100 nm), these domains are delimited between each other by the fact that a same domain presents at its outer surface one of the magnetization subnetworks, oriented in a certain direction, the atomic layer immediately inside this domain being clearly constituted of the subnetwork with opposed magnetization (same direction and opposite direction). This ordering exists below the Néel temperature and reversibly disappears above this temperature to give way to a slight paramagnetism or absence of magnetic order. These Néel domains are at the origin of magnetic noise when a spin current flows through the material, in a comparable manner to Weiss domains for ferro- or ferrimagnetic materials.
Some known techniques enable antiferromagnetic layers to be obtained in which the atomic layer per unit of area (external) corresponds to a magnetized network in one direction, the atomic layer immediately deeper (internal) clearly corresponds to the magnetized network in the opposite direction. Such an arrangement enables the atoms from the external layer per unit of area to establish a magnetic exchange action with the atoms from a material placed in immediate contact. By placing a ferromagnetic layer in contact, one may impose, by exchange coupling, the magnetic direction of this ferromagnetic layer. Since the antiferromagnetic layer is totally insensitive to the external magnetic field, it thus locks the orientation of the ferromagnetic layer. In this way is obtained the ferromagnetic/antiferromagnetic coupling used to produce the “hard layers” mentioned above, at a fixed magnetization direction, in giant magneto resistance elements, spin valves, magnetic storage and, more generally, any spintronics.
Whatever the magnetic material, the presence of magnetic domains (Weiss, Néel, etc.) separated by walls is observed; these magnetic microstructures will subsequently be designated by the generic term magnetic domain.
As already mentioned above, the magnetic domains, whatever they are, are at the origin of a noise (Barkhausen noise) induced by the displacement of walls of these domains. Consequently, having magnetic layers whose magnetic domains are as big as possible is useful in spintronics, in order to limit this noise. One way to reduce the number of small domains consists of applying a magnetic field to the magnetic material that is sufficiently strong such that the material contains practically no more walls and is monodomain. However, this solution is not applicable to antiferromagnetic material layers; their lack of sensitivity to the external magnetic field does not allow them to act on domain dimensions.
One known solution to enlarge the Néel domains of an antiferromagnetic layer consists of using the following process:
placing this antiferromagnetic layer in very close contact with a ferromagnetic layer;
Nevertheless, this process presents the following disadvantages:
In addition, control of this process on magnetic domains remains limited.
In conclusion, the known process described above has an extremely limited utilization and is hardly applicable to many spintronics circuits.
In this context, the object of the present invention is to provide a process for fabricating an antiferromagnetic layer allowing small size Néel domains to be eliminated and to significantly increase the size of the remaining Néel domains while getting rid of the limitations mentioned above (interdiffusion, inapplicability of the process with antiferromagnetic materials having a too-high Néel temperature, recrystallizations, homogeneity of the treatment, inapplicability to substrates such as silicon).
For this purpose, the invention proposes a process for fabricating an antiferromagnetic layer comprising the following steps:
Antiferromagnetic material in which at least one of the components of material of said first layer may be integrated by diffusion during growth is understood to refer to:
It is noted that the external magnetic field applied must have a certain amplitude to obtain the shifting of domains; Typically, during magnetization of a ferro- or ferrimagnetic material, when a magnetic field applied has a too-low amplitude, the response of the material may be reversible. In this case, the spins may follow, at least partially, the external field applied but the domain walls do not move. When the magnetic field is cut, the spins return to their initial state and nothing has changed. Consequently, according to the invention, it is necessary to apply a magnetic field exceeding this phenomenon to obtain shifting of the walls.
Just as with the magnetic field, time necessary for switching domains is understood to refer to the time necessary so that the shifting of walls endures in a stable position after elimination of the magnetic field. If a magnetic field is applied for a too-short time and/or with a too-weak magnetic field, the modifications will be reversible. Wall shifting is typically done at the millisecond scale, a magnetic field with higher amplitude tending to slightly reduce this value. Therefore, one must leave the present field for the time necessary so that the walls are effectively shifted, and in the case of a short pulse followed by a Larmor precession, add the time necessary for stabilizing the electronic spins to the wall shifting time.
The amplitude of the field and the application time of this field depend on the material. In general, the reversible magnetization zone must be overcome.
Thanks to the invention, the antiferromagnetic layer will repeat the statistical distribution of domains from the first magnetic layer, which will have enlarged magnetic domains at the time of deposition of the first atomic layers of antiferromagnetic layer, these first layers being in a sufficient number so as to establish the ferromagnetic order. The antiferromagnetic order is established over great distances with relation to other magnetic (several nanometers) or structural (less than the nanometer) orders. The growth of the antiferromagnetic layer is carried out from a first layer, either ferrimagnetic or ferromagnetic, or paramagnetic or diamagnetic: The antiferromagnetic order is established following a sufficient thickness of ferri, ferro, para or diamagnetic material. In all cases, this first layer must have a sufficient thickness so that the ferri, ferro, para or diamagnetic order is established. This order generally corresponds to the thickness of at least three or four atomic layers (typically on the order of a nanometer). The process from the invention consists of intervening at a stage that is sufficiently early in the growth of the antiferromagnetic layer in order to avoid problems from the prior art. When the antiferromagnetic order is not yet established, it may be manipulated by modifying the magnetic domains from the first layer (initial layer) by application of an external magnetic field (permanent or not permanent). In fact, the applicant had the surprise of observing that the statistical distribution of the antiferromagnetic layer repeats the statistical distribution of the initial layer. Consequently, the antiferromagnetic state of the antiferromagnetic layer is modified by applying a magnetic field before the transition to the antiferromagnetic order and by modifying the domains of the initial layer. Thus, the size, shape or statistical distribution of the antiferromagnetic Néel domains may be controlled without resorting to annealings or post-processing methods. The process according to the invention thus enables having recourse in spintronics to antiferromagnetic materials with a high Néel temperature and to ferromagnetic materials coupled by exchange with Curie temperatures lower than the Néel temperature of the antiferromagnetic layer.
It should be noted that the magnetic field is only applied from the time when the ferro, ferri, para or diamagnetic order is established. Thus, the process according to the invention totally differs from known processes to influence the formation of metallic films, magnetic or not, by using a magnetic field in a plane parallel to the surface of the substrate aiming to prevent high-energy electrons coming from a plasma source from bombarding and thus altering the surface of the film during its development: These processes absolutely do not aim to influence the distribution of magnetic domains by application of a magnetic field to an initial layer in which the magnetic order is established. These processes contribute even less to enabling an antiferromagnetic layer to grow on the initial layer and to repeating the statistical distribution of magnetic domains of the initial layer.
The method according to the invention may also present one or more of the characteristics below, considered individually or according to all technically possible combinations:
the growth of said first layer is carried out on a substrate cleaned of any contamination;
Another object of the present invention is a magnetic structure comprising at least one antiferromagnetic layer obtained by the process according to the invention.
Advantageously, the magnetic structure according to the invention comprises at least one ferromagnetic layer deposited on said antiferromagnetic layer and in which the configuration of magnetic domains is identical to that of said antiferromagnetic layer.
Other characteristics and advantages of the invention will clearly emerge from the description given below, for indicative and in no way limiting purposes, with reference to the attached figures, among which:
In all figures, common elements bear the same reference numbers.
The process according to the invention advantageously utilizes the surprising observation by the applicant that the statistical distribution of magnetic domains is identical in a ferrimagnetic film of γ-Fe2O3, with a thickness of less than 3 nm, and in an antiferromagnetic film of α-Fe2O3 with a thickness greater than 3 nm.
This phenomenon is first illustrated by
The person skilled in the art observes in the two images from
This surprising phenomenon is confirmed by
The statistical distribution of domain sizes obeys, in the two cases, the statistical laws of a random field Ising model, typical of ferromagnetic materials. The fractal dimension obtained from domain images is 1.89±0.02 and the roughness coefficient is 0.60±0.04, which corresponds to the exponents expected in the hypothesis of a ferromagnetic domain propagation equation (governed more precisely by a Kardar-Parisi-Zhang type equation). That said, seeing boundaries of antiferromagnetic domains responding to a model designed for physical propagation phenomena was not expected.
The process according to the invention advantageously utilizes identical statistical distributions between the antiferromagnetic layer and the initial layer on which it grows.
This process thus comprises a first step 1 consisting of depositing, on a substrate, a first magnetic layer (ferri, ferro, para or diamagnetic).
The second step 2 consists, after depositing with a thickness sufficient so that the magnetic order (ferri, ferro, para or diamagnetic) of the material of the first layer is established, i.e., in practice at least three or four atomic layers, of applying an external magnetic field with sufficient amplitude to cause the shifting of magnetic domain walls of the first layer for a time at least equal to the switching time of these domains. In other words, a magnetic field is applied with sufficient amplitude and duration to shift the walls of the magnetic domains of the first layer from a first statistical distribution to a second statistical distribution, the second statistical distribution presenting:
According to the third step 3, on the first ferri, ferro, para or diamagnetic layer in which the magnetic domains are modified, an antiferromagnetic layer is caused to grow of a material in which at least one of the components of the material of the first layer may be integrated by diffusion during growth; This second antiferromagnetic layer, that may advantageously be of the same chemical composition as the first layer, forms a magnetic structure in which the Néel domains repeat the shape and dimensions of the Weiss domains of the first layer.
According to a preferential embodiment of the invention, the material utilized for the first layer is a ferrimagnetic material; the invention finds a particularly interesting application in the case of the ferrimagnetic γ-Fe2O3 material.
According to a first embodiment of the process according to the invention, the first layer (initial layer) is deposited in a thin film on a substrate in an environment that is free from contamination and without chemical reaction facing the deposited material, preferably under ultra-high vacuum (typically a residual vacuum of less than 10−9 mbar).
In the case of Fe2O3, a substrate of α-Al2O3(0001) or Pt(111) and a growth chamber with a residual vacuum of 5.10−10 mbar may be utilized. The Pt substrate prevents the presence of charge effects for certain measures. The growth of Fe2O3 films is carried out on a substrate cleaned of any contamination by using atomic oxygen plasma and Fe atom evaporation from an MBE (Molecular Beam Epitaxy) source. The evaporants have a high purity (99.999% for the Fe here) and are evaporated with flux on the order of 0.1 nm/min. The pressure during deposition remains better than 10−8 mbar for an oxygen plasma source that dissociates approximately 10% of the oxygen atoms. The Fe2O3 layer may be made in a wide temperature range going from ambient temperature up to 450° C.
As mentioned above, at an early stage of growth, for a thickness such that the ferrimagnetic order is established (γ-Fe2O3) but not yet the antiferromagnetic order (α-Fe2O3), a saturating magnetic field is applied. In the example given above, the growth was stopped for a thickness of 2 nm and the sample was subjected to magnetic induction of 2 Tesla for 30 seconds with a high field strength speed on the order of 5 minutes and a low field strength also on the order of 5 minutes. The magnetic field may be applied in any direction but a particularly effective result in magnetic anisotropy will be obtained when it is applied in an easy magnetization direction, in particular for materials presenting high magnetocrystalline anisotropy. For γ-Fe2O3, this magnetocrystalline anisotropy is weak; Consequently, the orientation of the sample could be any orientation. As we will see, application of the magnetic field enables the statistical distribution of magnetic domains to be modified.
The growth is then continued up to a sufficient thickness so that the antiferromagnetic order is established to carry out the growth of an antiferromagnetic layer (second layer) of α-Fe2O3. The final thickness may be chosen according to the application, the remaining antiferromagnetic domains are subsequently fixed.
For the material cited as an example, thicknesses up to 30 nm have been tested.
The result of this process is illustrated by
The round dots relate to a layer directly deposited in a magnetic field. The square dots relate to a layer of Fe2O3 obtained by the process according to the invention whose growth has been stopped for a thickness of 2 nm where magnetic induction of 2 Tesla has been applied for 30 seconds with a high and low field strength speed on the order of 5 min. The growth was then continued up to a thickness of 10 nm.
It is observed that the statistical universality class of the second curve is kept but this second curb is vertically displaced and the smallest domains have been eliminated. In other words, two effects linked to the magnetic field application are observed:
In the example illustrated in
This double effect that consists not only of eliminating the smallest domains but also of increasing already large domains is explained by the fact that the antiferromagnetic layer repeats the statistical distribution of the domains of the ferrimagnetic layer on which it grows; Consequently, by acting on the statistical distribution of ferrimagnetic layer domains, the statistical distribution of antiferromagnetic layer domains is acted on. The statistical distribution in α-Fe2O3 is thus modified when a magnetic field is applied to γ-Fe2O3 before finishing the growth, this modification being attainable without requiring thermal treatment. In other words, the antiferromagnetic state of the α-Fe2O3 layer is modified by applying a magnetic field before the transition to the antiferromagnetic order. Thus, the size, shape or statistical distribution of the antiferromagnetic domains may be controlled without resorting to thermal annealings or post-processing methods. By using this process, a specific magnetic anisotropy may be “imprinted” in the antiferromagnetic material thanks to the action of a magnetic field in the early phase that exists for a thickness of less than the appearance of the antiferromagnetic order. According to the preferential embodiment where a magnetic field is applied for the time necessary for shifting domains, the fields to be applied will typically be on the order of 0.01 Tesla for times of at least some milliseconds.
Of course, the process according to the invention applies to other materials such as Fe2O3. Thus,
Of course, the process according to the invention is not limited to the embodiments that have just been described for indicative and in no way limiting purposes with reference to
In particular, the invention was more particularly described in the case of a first ferrimagnetic layer; As we have already mentioned, as the ferro and ferrimagnetic properties of the layers are very close, the invention also applies to an initial ferromagnetic layer. In addition, the process is also applicable to first paramagnetic or diamagnetic initial layers. Thus, by way of example, a critical nanometric size exists for which particles of Ni—Mn transit from a paramagnetic order to an antiferromagnetic order [see in particular Ladwig et al. Journal of Electronic Materials 32 (2003) pp 1155-1159]; Implementing the process with an initial (first layer) paramagnetic layer of Ni—Mn that transits to a second antiferromagnetic layer of Ni—Mn may thus be considered. Moreover, thin films of Cr are often diamagnetic while bulk chromium adopts an antiferromagnetic order [see in particular K. Schrôder and S, Nayak, Physica Status Solidi (b) 172 (1992) pp 679-686]. Implementing the process with an initial (first layer) diamagnetic layer of Cr and a second antiferromagnetic layer of Cr may thus also be considered.
As with the process described previously, a specific magnetic anisotropy may be “imprinted” in the antiferromagnetic material thanks to the action of a magnetic field in the early phase that exists for a thickness of less than the appearance of the antiferromagnetic order. For paramagnetic or diamagnetic initial layers, the application of the magnetic field is done until the thickness is sufficient so that the antiferromagnetic order is established. For an early paramagnetic phase, a moderate field of some 0.01 T to some 0.1 T will be sufficient. In the case of the utilization of a first diamagnetic layer, the limited magnetic susceptibility of the diamagnetic materials requires applying higher amplitude magnetic fields to influence the latter, typically from 1 to several Tesla, or even more.
In addition, in the example described, the growth was stopped and the sample was taken out of the growth chamber to be subjected to a magnetic field. Of course, it is also possible to apply the magnetic field directly in the growth chamber. By designating the term “magnetic means” to refer to the assembly of devices enabling a magnetic field to be applied at the location where the MBE (or other) deposition will be carried out, these magnetic means may be constituted either by at least one permanent magnet or by at least one vacuum coil arranged directly in the chamber.
The process according to the invention finds an immediate application in spin electronics, also designated by the term spintronics. Spintronics is a growing discipline that consists of utilizing the spin of the electron as an additional degree of freedom with relation to conventional electronics on silicon that only utilize its charge. In fact, spin has a significant effect on the transport properties in ferromagnetic materials. Many spintronics applications, in particular memories or logic elements, utilize stacks of magnetoresistive layers comprising at least two ferromagnetic layers separated by a non-magnetic layer. One of the ferromagnetic layers is trapped in a fixed direction and acts as a reference layer while the magnetization of the other layer may be switched relatively easily by the application of a magnetic moment by a magnetic field or a spin polarized current.
These stacks may be magnetic tunnel junctions when the spacer layer is insulating or structures known as spin valves when the spacer layer is metallic. In these structures, the resistance varies according to the relative orientation of magnetizations of the two ferromagnetic layers.
As we have already mentioned, the magnetization of one of these ferromagnetic layers (called hard layer, HL) is fixed. The stability of this layer may be ensured by its shape and/or by exchange coupling with an antiferromagnetic layer. This exchange coupling necessitates the deposition of a ferromagnetic layer on an antiferromagnetic layer, the latter may be an antiferromagnetic synthesis layer. Here all the interest of the process according to the invention may be seen since the magnetic jig created via this process may be utilized to propagate in magnetic junction layers of the spin valve or tunnel junctions type. The magnetic domains of the antiferromagnetic layer are in fact also repeated by the ferromagnetic layer that will be grown on it. This phenomenon is illustrated by
The images are made by spectromicroscopy from a source of circularly polarized monoenergetic photons with energy close to absorption thresholds L2 or L3 of Co for the left image and Fe for the right image; The image results from the weighted difference of images observed for horizontal and vertical linear polarizations for observation of antiferromagnetic domains and circularly left and right polarizations for observation of ferromagnetic domains. The direction of incident photons is indicated by an arrow in the image. For the image of the antiferromagnetic layer, the white zones represent the magnetic domains with magnetic moments parallel or antiparallel to the direction of incident photons (represented by a double black arrow). The grey or black zones represent magnetic domains with magnetic moments substantially perpendicular to the direction of the incident photons (represented by a double white arrow). For the image of the ferromagnetic layer, the black zones represent magnetic domains with magnetic moments opposed to the direction of the incident photons (represented by a white arrow). The white zones represent magnetic domains in which the direction of magnetic moments is situated following the direction of the incident photons (represented by a black arrow).
As may be observed in these two images, the Co layer reproduces the same magnetic domain configuration as the underlying antiferromagnetic layer of Fe2O3. Thus, with a magnetic field that is sufficiently intense during the early phase, it is completely possible to obtain monodomain layers (or in any case, to eliminate reduced size domains and to increase the size of the remaining domains) in order to reduce the noise linked to reduced size magnetic domains. The process according to the invention opens the way to spintronics applications allowing the utilization of antiferromagnetic materials with a high Néel temperature (this is particularly the case with Fe2O3 whose Néel temperature is about equal to 650° C.) and the utilization of ferromagnetic materials with lower Curie temperatures (free from the requirement for a high Curie temperature via the absence of thermal treatment). In the case of an application aiming to obtain larger size domains (or even a monodomain), macroscopic magnetic anisotropy is imprinted to the assembly of materials utilized by using a macroscopic magnetic field. However, it will be noted that it is also possible to utilize a magnetic field that is applied locally, for example via an MFM (Magnetic Force Microscopy) tip, by patterning magnetic domains and by thus creating a jig in domain form and by then depositing the ferromagnetic material on the antiferromagnetic layer obtained by the process according to the invention, for example for the development of magnetic sensors of the spin valve or tunnel junction type that will return to the form of domains impregnated at the start in the initial ferrimagnetic layer.
| Number | Date | Country | Kind |
|---|---|---|---|
| 0857646 | Nov 2008 | FR | national |
| Filing Document | Filing Date | Country | Kind | 371c Date |
|---|---|---|---|---|
| PCT/FR09/51950 | 10/13/2009 | WO | 00 | 6/10/2011 |
