Large grain copper interconnect lines for MRAM.
Laser annealing of copper bit lines in MTJ memory devices forms large-grain structures that reduce resistance and scattering, addressing the copper wire size effect, thereby improving access speed and array size in MRAM devices.
Patent Information
- Application Number
- JP2022544275
- Authority / Receiving Office
- JP · JP
- Patent Type
- Patents
- Current Assignee / Owner
- Priority Date
- 2020-01-27
- Filing Date
- 2021-01-05
- Publication Date
- 2024-11-19
- Estimated Expiration
- 2041-01-05
AI Technical Summary
The challenge in forming copper interconnect lines for magnetic tunnel junction (MTJ) memory devices is the nonlinear increase in resistance due to the copper wire size effect, which affects the access speed and size of memory arrays, particularly in MRAM devices, as copper wires thinner than 40 nm experience significant electron scattering at grain boundaries.
The formation of large-grain copper bit lines is achieved through laser annealing, which melts and recrystallizes the copper bit lines, aligning grains perpendicularly to vertical pillars, primarily located between MTJ pillars, to reduce grain boundaries and scattering, thereby reducing resistance.
This method effectively lowers the effective resistivity of copper lines to near bulk copper values, enhancing access speed and array size by reducing the copper wire size effect, maintaining the magnetic properties of MTJ stacks.
Smart Images

Figure 0007586610000001 
Figure 0007586610000002 
Figure 0007586610000003
Abstract
Description
[Technical field]
[0001] The present invention relates generally to electrical, electronic and computer technologies, and more particularly to copper interconnect lines for arrays of magnetic tunnel junction (MTJ) memory devices. [Background technology]
[0002] Magnetic tunnel junction stacks are suitable for use in a variety of electronic applications, such as non-volatile memory devices and magnetic field sensors. Magnetic random access memories (MRAMs) can provide faster operating speeds than, for example, flash memories. MRAM devices may even be able to replace dynamic random access memory (DRAM) devices in some applications.
[0003] A magnetic tunnel junction includes two magnetic layers and a tunnel barrier layer disposed between them. The magnetic layers can be characterized as a "reference layer" and a "free layer," respectively, and the tunnel barrier can be a thin tunnel oxide layer. The magnetization direction of one layer of the junction is fixed, so it serves as the reference layer. The magnetization of the free layer can be determined by an electrical input. The MTJ contains two stable resistance states for digital memory applications and multiple states for neuromorphic applications. Charge flow from the reference layer to the free layer switches between the states of the MTJ by overcoming an energy barrier.
[0004] Silicon-based devices typically include multiple interconnect metallization layers above the device (front-end of line / FEOL) layers that contain field effect transistors (FETs) or other structures. FEOL processing includes high temperature steps to manipulate the conductivity of semiconductors. Middle-of-line (MOL) processing includes steps typically used in the fabrication of metal contacts for logic circuit components such as field effect transistors (FETs), resistors, diodes, and capacitors. MOL processes include intermediate temperature steps to form semiconductor-metal compounds (silicides, germanosilicides) for electrical contacts. Back-end-of-line (BEOL) processing involves the formation of metal interconnect wires to connect devices formed in the FEOL processes to form electrical circuits. Metal lines including interconnect wires are sequentially deposited (e.g., M1, M2, M3, etc.) on top of the FEOL layers, including dielectric layers that contain trenches and vias that are filled with conductors. Trench openings are conventionally formed in the dielectric layers using, for example, known damascene or dual damascene techniques. The M2 and M3 lines may have a wider pitch than the M1 lines. The interconnect wires in each metal line are electrically connected to the interconnect wires in other metal lines and to the devices in the FEOL layer. BEOL processing includes low temperature steps to form the metal wires and preserve the temperature sensitive FEOL and MOL structures. In the BEOL layer for memory and neuromorphic applications, high density MTJ structures (MRAM bits) are formed. Low temperature BEOL processes preserve the delicate MTJ devices where both the magnetic and tunnel layers are highly susceptible to excessive temperature processing during post processing. These high density MTJ memory cells require thin metal lines for access. The thin wires extend above and below the MTJ memory cells. The access speed and the maximum size of a single memory bank or array depend on the resistance of the thin wires; the lower the wire resistance, the faster the access speed and the larger the single array.
[0005] Damascene copper wires are chosen to reduce the wire resistance because copper has good electrical conductivity. However, when the copper wires are narrowed below 40 nm in the lateral and vertical directions, a nonlinearization of the copper wire resistance occurs, known as the interconnect size effect or copper wire size effect. The physical mechanism behind the size effect is the electron diffusive scattering from the grain boundaries and interfaces of the wire. The average grain size of copper decreases with the wire dimensions, typically becoming comparable to the wire width. The interconnect size effect is the dominant constraint in achieving low resistance with narrow copper wires. Summary of the Invention
[0006] Large-grain copper interconnect lines for MRAM are formed without significantly affecting the magnetic properties of the MTJ stacks in the nonvolatile memory device.
[0007] A method for forming large grain copper bit lines for electronic structures includes obtaining an MRAM structure including a device wafer including an electronic device, a metallization layer overlying the device wafer and including bottom electrodes therein, the bottom electrodes being electrically connected to the electronic device, and vertical pillars having sidewalls over the metallization layer, the vertical pillars each including a magnetic tunnel junction and a top electrode. Copper bit lines exhibiting a copper wire size effect are formed over the top electrodes. The copper bit lines are 40 nanometers wide or less and 3000 square nanometers or less in area. Cut off Area and width of copper bit lines Compared to The copper bit line has a small copper grain size. The copper bit line is melted by laser annealing. This method allows the copper bit line to recrystallize during the cooling period after the laser anneal, thereby forming copper particles therein. Compared to the width of the copper bit lines The method further includes forming large grains to reduce copper wire size effects in copper bit lines formed over the top electrode.
[0008] A further method of forming large grain copper bit lines for electronic structures includes obtaining an MRAM structure including vertical pillars having a pitch of 60 nanometers or greater, each vertical pillar including a magnetic tunnel junction and a top electrode, a plurality of spaces between the vertical pillars, each space separating each of the vertical pillars, and a dielectric layer filling each of the plurality of spaces between the vertical pillars. A copper bit line is formed along the pitch of the vertical pillars over the top electrode. The copper bit line extends over the spaces between the vertical pillars and has a width of 40 nanometers or less and an area of 3000 square nanometers or less. Cut off Area and width of copper bit lines Compared to The copper bit lines have small copper grain sizes. The method further includes laser annealing the copper bit lines, thereby melting the copper bit lines, and recrystallizing the copper bit lines during a cooling period after the laser annealing. Compared to the width of the copper bit lines Large grains are formed that are aligned perpendicular to each of the vertical pillars, with the large grains having grain boundaries that are primarily located above the dielectric layer that occupies the spaces between the vertical pillars.
[0009] In a further aspect of the invention, a magnetoresistive random access memory structure includes a plurality of vertical pillars having sidewalls. Each of the plurality of vertical pillars includes a magnetic tunnel junction and a top electrode. A pillar pitch of the plurality of pillars is equal to or greater than 60 nanometers. The memory structure further includes a plurality of spaces, each of the plurality of spaces being between an opposing pair of sidewalls of adjacent vertical pillars. A dielectric layer fills each of the plurality of spaces. A copper bit line is above and electrically connected to each of the top electrodes. The copper bit lines have a line width equal to or less than 40 nanometers and grains having a grain size greater than twice the line width and an average grain size equal to or greater than 0.06 μm. The pillar pitch is along the copper bit lines.
[0010] The techniques and structures disclosed herein can provide substantial beneficial technical effects. By way of example only and not limitation, one or more embodiments can provide one or more of the following advantages: · Reduce copper grain boundaries in the copper lines above the MTJ stack. Reducing the electrical resistance of the copper lines above the MTJ stack without affecting the magnetic properties of the MTJ stack. -Allows high-speed access to the MRAM array.
[0011] These and other features and advantages will become apparent from the following detailed description of illustrative embodiments thereof, which is to be read in connection with the accompanying drawings. [Brief description of the drawings]
[0012] The following drawings are presented by way of example only, and not by way of limitation, in which like reference numerals (where used) indicate corresponding elements throughout the several views.
[0013] [Figure 1] FIG. 2 is a schematic cross-sectional view showing a device wafer containing FEOL devices / circuitry, a first level metal layer on the device wafer, and a second level metal layer including a bottom electrode. [Diagram 2] FIG. 2 is a schematic cross-sectional view illustrating an example MTJ stack, a dielectric cap, and an electrode layer above a second level metal layer. [Diagram 3] 3 is a schematic cross-sectional view of the structure of FIG. 2 after formation of an MTJ pillar from the layers of the MTJ stack. [Figure 4] FIG. 2 is a schematic cross-sectional view of the MTJ pillar, associated caps and electrodes after nitride encapsulation. [Diagram 5] FIG. 2 is a schematic cross-sectional view after deposition and planarization of an interlevel dielectric layer. [Figure 6] FIG. 2 is a schematic cross-sectional view showing a copper bitline having relatively small grains and numerous grain boundaries. [Figure 7] FIG. 1 is a schematic cross-sectional view showing the melting of the top copper line by laser annealing and its recrystallization to form a top copper line with relatively large grains and improved interfaces. [Figure 8A]6 is a graph showing the tunneling magnetoresistance effect as a function of laser annealing temperature after forming a top copper line on the structure shown in FIG. 5 and depositing a top capping layer on the top copper line. [Figure 8B] 1 is a graph showing resistance area product (RA) as a function of laser annealing temperature.
[0014] It should be understood that elements in the figures are illustrated for simplicity and clarity: common but well-understood elements that may be useful or necessary in a commercially feasible embodiment may not be shown to improve visibility of the illustrated embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0015] The principles of the invention are described herein in the context of exemplary embodiments. It will be understood, however, that the specific embodiments and / or methods illustratively shown and described herein are to be regarded as exemplary and not limiting. Moreover, it will be apparent to one skilled in the art given the teachings herein that numerous modifications can be made to the illustrated embodiments that are within the scope of the claims. That is, no limitations with respect to the embodiments shown and described herein are intended or should be inferred.
[0016] Incorporating memory devices near the FEOL layers, rather than placing them at special locations on the chip or using separate dedicated memory chips, may improve access times and overall circuit performance. MRAM is one type of memory device that can be incorporated by BEOL CMOS processing just above the FEOL layers, for example between the M2 and M3 lines or layers. The monolithic structure 20 shown in FIG. 1 is obtained following FEOL processing and formation of metal lines M1 and M2. The FEOL layers contain electronic devices such as field effect transistors (FETs) that comprise the integrated circuit. The first metal layer M1 contains contact and via conductors (not shown) electrically connected to the electronic devices in the FEOL layer 22. The second metal layer M2 contains copper lines electrically connected to the M1 layer. The memory cell structure may contain one access transistor or may be of the cross-point type with all active elements (transistors) located at the periphery of the array. In either case, the MTJ memory element has two connections, one through the bottom M2 metal line and one through the top M3 metal line. The top M2 line of the MTJ MRAM cell connects to the source of the NFET access transistor, and the top M3 line of the MTJ MRAM cell connects to the array periphery circuitry. Both the M3 and M2 lines are sometimes referred to as bit lines. For cross-point MRAM memory cells, both the M2 and M3 metal lines connect to the array periphery circuitry. Large memory arrays or banks result in long M3 connection metal lines. In some embodiments, as many as 64,000 individual MTJ cells or bits are connected to a single M3 bit line. High density memory arrays require thin M2 and M3 metal lines and tight M2 and M3 pitch. Long, thin M3 lines come at the cost of substantial memory access times, which in turn limits the maximum array or memory bank size. Increasing the number of memory arrays or banks per memory die sacrifices overall memory density because peripheral circuitry is duplicated.Thus, there is a continuing desire to reduce the resistance of long, narrow bitlines, thereby enabling faster access times or larger individual memory arrays, or both.
[0017] The trench openings are conventionally formed in the ILD layer 24 using, for example, known damascene techniques. The deposition of the ILD layer is followed by photolithography and etching steps. In particular, a photoresist (not shown) is applied over the ILD layer. The photoresist can be applied by any suitable technique, including, but not limited to, coating or spin-on techniques. A mask (not shown) patterned with shapes defining the trench openings (and possibly contact holes) to be formed is provided over the photoresist, and a photolithography process is used to transfer the mask pattern to the photoresist and form recesses in the uncovered areas of the photoresist. The patterned photoresist is then used to form recesses of the same pattern in the ILD dielectric layer by etching, which is typically used to form trenches and contact holes. Dry etching (e.g., reactive ion etching) can be used to form such trenches and contact holes. The etching selectively removes portions of the ILD layer 24. The depth of the trench openings can be controlled by using a timed etching process. Alternatively, the dielectric layer may include multiple layers that can be selectively etched. In such cases, the etching process selectively removes the top layer of the ILD layer and stops on the underlying layer which forms an etch stop. After formation of the trench opening, the photoresist may be stripped from the ILD layer by ashing or other suitable process. The resulting structure may be subjected to a wet clean.
[0018] A further step in the manufacturing process involves depositing a conformal layer of liner material. The conformal layer of liner material lines the sidewalls and bottom of the trench or other opening in the ILD layer 24. The liner material may include one or more thin layers of material including, for example, tantalum (Ta), tantalum nitride (TaN), titanium nitride (TiN), cobalt (Co), ruthenium (Ru), tungsten (W), tungsten nitride (WN), manganese (Mn), manganese nitride (MnN), or other liner materials (or combinations of liner materials) such as RuTaN, Ta / TaN, CoWP, NiMoP, NiMoB, etc., suitable for a given application. The thin liner acts as a barrier diffusion layer and an adhesion layer. The conformal layer of liner material is deposited using known techniques such as CVD, ALD, PVD, etc. A thin conformal copper (Cu) seed layer can be deposited on the surface of the liner using PVD, and then Cu can be electroplated to fill the damascene (or dual damascene) openings in the ILD layer. Electroplating is followed by a thermal anneal step.
[0019] A three-step chemical mechanical polishing process (CMP) is then performed to remove overburden liner, seed, and metallization materials to planarize the surface of the semiconductor structure down to the ILD layer 24. A metal cap layer (not shown) may be selectively deposited on the exposed metal interconnect layer in the trench. For example, metals such as cobalt, ruthenium, or manganese may be deposited using chemical vapor deposition or atomic layer deposition to form the metal cap layer. A post-deposition clean may be required to ensure there is no leakage or degradation due to metal residues that may be present on the resulting structure.
[0020] The M2 layer may alternatively be referred to as the bottom or lower line of the memory element or cell, or the bottom portion of the bit line. A SiCOH dielectric film having a dielectric constant (k) of about 2.7-2.8 may be used to form the M2 layer. Such a dielectric film may be deposited using PECVD. This film is patterned as described above to form a trench which is later filled with copper to form the bottom electrode 26. The copper electrode preferably has a relatively low resistance. A low resistance electrode may be obtained by annealing and recrystallization of the deposited metal.
[0021] Referring to FIG. 2, the MTJ stack film is deposited using, for example, physical vapor deposition (PVD) or ion beam deposition (IBD). The tunnel barrier layer 28 can be formed by oxidizing a metal layer without oxidizing the underlying reference layer 25. A metal capping layer 29 is formed on the free layer 27. The capping layer can be composed of, for example, Nb, NbN, W, WN, Ta, TaN, Ti, TiN, Ru, Mo, Cr, V, Pd, Pt, Rh, Sc, Al or other refractory metals or conductive metal nitrides. The capping layer 29 can have a thickness of 2 nm-25 nm. A top electrode layer 30 is formed on the capping layer. The top electrode 30 can be composed of Ta, TaN, Ti, TiN, Ru, RuN, RuTa, RuTaN, Co, CoWP, CoN, W, WN or any combination thereof. The top electrode 30 can have a thickness of 2 nm-25 nm. Patterning of the MTJ stack film to form pillars 32, some of which are shown in FIG. 3, is accomplished by a stack etching process. Ion milling (ion beam etching or IBE) is an effective MTJ stack etching technique. A dielectric cap (not shown) is formed from a hard mask (not shown) adjacent to the upper surface of the top electrode layer 30. The dielectric cap may comprise silicon nitride or other compositions suitable for ion beam etching of the MTJ stack film. After ion beam etching, the dielectric cap is removed.
[0022] Referring back to the exemplary structure shown in FIG. 3, the MTJ stack is shown after deposition and patterning of the hard mask, ion beam etching of the MTJ stack film, and removal of the hard mask. Etching of the MTJ stack film results in vertical pillars 32 being formed from the stack film, each of the pillars including a magnetic tunnel junction consisting of a tunnel barrier layer 28 between the reference layer 25 and the free layer 27. Compounds that can form the tunnel barrier layer include aluminum oxide and magnesium oxide, among others. The tunnel barrier layer 28 is made as thin as possible, typically only 3 nm or less, to allow for large tunneling currents. The ultra-thin tunnel barrier layer 28 makes the MTJ structure susceptible to high temperature heat treatments, since such ultra-thin layers allow adjacent metal elements to diffuse through them and change their properties. Annealing MTJs with ultra-thin tunnel barriers 28 formed at temperatures above 500° C. is typically avoided. The free layer 27 can be formed of magnetically active metals such as Fe and / or Co, FeCoB, and combinations thereof, as well as other intermediate layers known in the art, and may include multiple layers. The reference layer 25 is adjacent to a bottom electrode 26 formed in an electrically insulating (ILD) interlayer 24 and may include multiple layers including a pinned layer, a ferromagnetic layer, and a spacer layer between the ferromagnetic layers. Exemplary materials for forming the magnetic reference layer include iron, nickel, cobalt, chromium, boron, manganese, and alloys thereof. A large single array of MTJ pillars (MRAM bits) can be formed (64 kb to 256 Mb or larger). The large single array of MTJ pillars can be replicated several times to reach a target capacity (e.g., up to 1 Gb) per product die. The MTJ pillars can form a "checkerboard" pattern in some embodiments with a pillar spacing (pitch) of 60 nanometers or more along the subsequently formed top bit line.
[0023] The intermediate layer 24 may be formed of a low-k dielectric material, such as silicon oxide or SiCOH as described above, and may include multiple dielectric sublayers. Chemical vapor deposition (CVD), including plasma-assisted CVD, may be used to deposit low-k (k less than 4.0) dielectric materials, such as porous SiCOH. The bottom electrode 26 is electrically connected to a metal contact via Q1 / V1, which is electrically connected to the M1 layer. A portion of the metal capping layer 29 is adjacent to the top surface of the free layer of each exemplary pillar 32. A top electrode layer 30 is adjacent to the top surface of the metal capping layer 29 of each pillar, as shown in FIG. 3, to form a top electrode. The layers of the pillars 32 are not necessarily drawn to scale.
[0024] An encapsulation layer 34 is formed on the MTJ pillar 32. The encapsulation layer may include multiple layers. For example, a manganese oxide or other oxide (e.g., metal silicate or metal oxynitride) compound may be formed on the MTJ pillar, followed by deposition of a conformal dielectric layer such as silicon nitride. The pillar and optional oxygen-containing barrier layer may be encapsulated with a silicon nitride layer to improve barrier properties. In embodiments where the encapsulation layer is a conformal layer of silicon nitride, a layer thickness of 12 to 15 nanometers is formed in one or more exemplary embodiments. The silicon nitride layer extends over the pillar 32 as well as the top surface of the M2 layer. A silicon nitride layer having a thickness of 5 to 6 nanometers is sufficient to provide an acceptable copper and oxygen diffusion barrier when used in combination with, for example, an underlying manganese silicate layer of 2 to 3 nanometers thickness adjacent the sidewall of the pillar. The manganese silicate / silicon nitride diffusion barrier combination will have a lower average dielectric constant (k) than a barrier consisting of silicon nitride alone. The relatively thin copper / oxygen diffusion barrier layer enables the fabrication of nano-MTJ devices that are compatible with advanced CMOS scaling.
[0025] In some embodiments, MnSiO xA liner is deposited on the layer. For example, a manganese nitride (MnN) liner having a thickness of one to five nanometers (1-5 nm) can be used. Such a liner can be deposited using atomic layer deposition (ALD) or chemical vapor deposition (CVD). MnSiO x The / MnN barrier layer provides an effective oxidation barrier as well as a barrier to copper diffusion. In some embodiments, 2-3 nm of MnSiO x A barrier is formed on the MJT pillars 32 using a 5 nm MnN liner and a 5 nm MnN liner. In the embodiment including the MnN liner, the Mn is preferred because excess Mn acts as an oxygen scavenger and thus results in a good / better oxidation barrier. 3 N 2 phase or Mn 4 Stable phases such as MnN x The preferred material is MnO. x In embodiments where x is formed or deposited on the sidewalls of pillars 32, it is preferred that the value of x is less than 2 in order to function as an oxygen scavenger and serve as an oxygen barrier. <x<1.0である。
[0026] The technique of encapsulating the MTJ pillars in a diffusion / oxidation barrier and oxidizing the residue can be, for example, by applying MnO x or MnSiO x This involves either directly depositing Mn or selectively depositing Mn and then oxidizing it. The oxidation barrier material can be deposited using atomic layer deposition (ALD) to encapsulate the MTJ pillar 32 and extend onto the M2 wiring layer. The oxygen in the deposited compound can oxidize any residues that may be present on the MTJ pillar. Manganese oxides are represented by several empirical formulas, e.g., MnO, MnO 2 , MnO 3 , Mn 3 O 4 , Mn 2 O 3In some embodiments, a manganese oxide having a relatively high oxygen content is deposited immediately adjacent to the MTJ pillar, while a manganese oxide compound having a low oxygen content is deposited later in the deposition process to form a graded diffusion barrier. The inner portion of the diffusion barrier layer can eventually reach a relatively low oxygen concentration once the oxygen therein is used to oxidize the metal residue. ALD deposited MnSiO x The oxygen content of the MnSiO may be similarly graded, with at least a higher initial oxygen content in the regions adjacent to the pillars 32 where metal residues requiring oxidation are expected. x In an embodiment where Mn(SiO ) is deposited on the MTJ stack using plasma-assisted atomic layer deposition, the atomic percent composition of manganese, silicon, and oxygen are, in some embodiments, 12.8%, 28.5%, and 58.7%, respectively. The composition of the manganese silicate film adjacent to the MTJ stack in some embodiments is accordingly Mn(SiO ). 2 ) 2 As used herein, MnSiO x It will be understood that the layer need not be, and is not required to be, perfectly uniform in composition. For example, manganese silicate (Mn(SiO 2 ) 2 The manganese silicate described herein has a dielectric constant of about 3.7, compared to 6.8-7.0 for silicon nitride.
[0027] An interlevel dielectric (ILD) layer 36 is deposited on the encapsulation layer 34 of the structure 40. The ILD layer 36, like the intermediate layer 24, can be formed of a low-k material such as silicon oxide or SiCOH as described above and may include multiple dielectric sublayers. The deposition process for the ILD layer 36 should have the ability to fill gaps due to the underlying topography, for example, to fill between the MTJ pillars. For this reason, deposition of a flowable ILD material is preferred. Chemical vapor deposition (CVD), including plasma-assisted CVD, can be used to deposit a low-k (k less than 4.0) dielectric material such as porous SiCOH. Alternatively, spin-on deposition of a flowable low-k material can be used. The ILD layer 36 fills the spaces between the pillars 32 and is adjacent to the encapsulation (barrier) layer 34. The ILD layer 36 can be optionally cured using a variety of known curing techniques, including UV light-assisted curing at less than 400° C. The resulting structure is then subjected to chemical mechanical planarization (CMP) down to the upper surface of the top electrode 30 to obtain a structure 50 as shown diagrammatically in FIG.
[0028] A top copper bit line 38, or in other words an MRAM bit line, is formed using a top metallization layer such as M3. The M3 layer is formed similarly to the M2 layer using a damascene process. An ILD layer (not shown) is first deposited over the MTJ pillars 32, the encapsulation layer 34, and the ILD material 36. A trench opening is conventionally formed in the ILD layer using a known damascene process. Specifically, a photoresist (not shown) is applied over the ILD layer. The photoresist can be applied by any suitable technique, including but not limited to coating or spin-on techniques. A mask (not shown) patterned with shapes defining the trench openings (and possibly contact holes) to be formed is provided over the photoresist, and a photolithography process is used to transfer the mask pattern into the photoresist and form recesses in the uncovered areas of the photoresist. The patterned photoresist is then used to form recesses of the same pattern in the ILD dielectric layer by etching typically used to form trenches and contact holes. Dry etching (e.g., reactive ion etching) can be used to form such trenches and contact holes. The etch selectively removes a portion of the ILD layer. The depth of the trench opening can be controlled by using a time-controlled etching process. Alternatively, the dielectric layer may include multiple layers that can be selectively etched. In such a case, the etching process selectively removes the top layer of the ILD layer and stops at the underlying layer forming an etch stop. After formation of the trench opening, the photoresist can be stripped from the ILD layer by ashing or other suitable process. The resulting structure may be subjected to wet cleaning. This M3 level contact via is in direct contact with the top electrode 30 of the MTJ pillar 32 to form the MRAM bitline. In some embodiments, the width of the M3 bitline trench is 15 nm to 40 nm and its depth is 30 nm to 80 nm, i.e., approximately twice its width. In some embodiments, the M3 bitline pitch is two or four times the M3 bitline trench width, i.e., 30 nm to 160 nm.
[0029] A further step in the fabrication process includes depositing a conformal conductive layer of liner material 39. The conformal layer of liner material lines the sidewalls and bottom of a trench or other opening in the ILD in which the top copper bit line will be formed. The liner material 39 may include one or more thin layers of materials such as, for example, tantalum (Ta) or ruthenium (Ru) or both. Other liner materials (or combinations of liner materials) suitable for a given application may alternatively be used. The thin liner serves as a barrier diffusion layer and an adhesion layer. The conformal layer of liner material 39 is deposited using known techniques such as CVD, ALD, or PVD. In some embodiments, an adhesion layer or copper wetting layer is deposited on the barrier material (e.g., TaN) and includes a material such as Ta or Ru that is immiscible in liquid copper. Materials such as cobalt that are miscible with liquid copper are not used as liner materials in preferred embodiments. The total thickness of the liner 39 is between 3 nm and 6 nm. A thin copper (Cu) seed layer can be deposited on the surface of the liner using PVD, and then Cu (MRAM bitlines 38) can be electroplated to fill the damascene (or dual damascene) openings in the M3 level ILD layer of the structure. Electroplating is followed by a low temperature thermal anneal (<200° C.).
[0030] A three-step chemical mechanical polishing process (CMP) is then performed to remove excess liner, seed, and metallization materials to planarize the surface of the semiconductor structure to the top surface of the M3 ILD layer (not shown). A metal cap layer (not shown) may be selectively deposited on the exposed metal bit line layer in the trench. For example, a metal such as cobalt, ruthenium, or manganese may be deposited using chemical vapor deposition or atomic layer deposition to form the metal cap layer. In some embodiments, the metal cap layer comprises a material such as Ta or Ru that is immiscible in liquid copper. A post-deposition clean may be required to ensure there is no leakage or degradation due to metal residues that may be present on the resulting structure. In some embodiments, the cross-sectional copper area of the M3 bit lines 38 is 300 nm 2 From 3000 nm 2 The M3 metal bit lines 38 typically run perpendicular to the M2 metal lines, forming cross points at their intersections.
[0031] A top dielectric capping layer 42 is subsequently deposited over the formed M3 metal bitlines and associated vias electrically connecting the M3 bitlines to the top electrodes 30. The top dielectric capping layer comprises an essentially non-porous dielectric material in a preferred embodiment of the present invention. A low-k silicon carbide material such as NBLoK™ may be deposited on the M3 metal bitlines 38 using plasma-assisted chemical vapor deposition at 375° C. This material contains Si, C, O, H, and N and is available from Applied Materials, Inc. Silicon nitride and silicon oxynitride are non-limiting examples of other materials that may be deposited on the top copper lines 38. The top capping layer 42 may comprise multiple layers in some embodiments.
[0032] The deposition of the dielectric capping layer 42 provides a desirable heat treatment of the formed copper metal bitlines 38 and associated vias at 375° C. As a result of this treatment, the average grain size of the copper increases slightly but remains relatively small, typically less than the width of the copper portion of the M3 metal bitlines 38. Specifically, the average grain size of the copper is less than 40 nm for copper line widths of 40 nm or less. The resulting copper bitlines 38 therefore have relatively small grains and numerous grain boundaries, such as those shown in FIG. 6. Despite the described layer 39 and its diffusion barrier, adhesion or wetting layers, and metal cap, the M3 metal bitlines 38 are essentially elemental copper structures with low levels of impurities unintentionally introduced during the deposition process, or alternatively, copper alloys that melt in a similar temperature range as elemental copper. Elemental copper melts at approximately 1090° C. Thus, the term “copper bitlines” encompasses elemental copper bitlines, which may contain impurities, as well as copper alloy bitlines that melt at approximately 1150° C. or less.
[0033] The resistance of a copper line per unit length is determined by two factors: the resistivity of copper and the cross-sectional area of the copper wire. For standard macro-sized wires, the resistivity of copper is a material property independent of the cross-sectional area of the wire, and the wire resistance per unit length is the resistivity of copper divided by the cross-sectional area of the copper. For nano-sized wires, such as the case of M3 metal bitlines, this relationship breaks down and the wire resistivity becomes dependent on the wire geometry. This is because electrons in copper nanowires collide with and transfer momentum to wire walls and grain boundaries more often than to copper lattice vibrations and copper crystal defects. In bulk copper, the average distance between electron collisions, known as the electron mean free path, is about 40 nm. As the cross-sectional dimensions of the wire become comparable to or smaller than the bulk copper electron mean free path of ~40 nm, the resistivity of the wire increases due to electron collisions at interfaces and grain boundaries. This effect is called the interconnect size effect, and the resulting resistivity of the nanowire is known as the effective resistivity of a copper nanowire, defined as the product of the measured copper wire resistance per unit length and the cross-sectional area of the copper. For example, the effective resistivity of a copper wire in an 80 nm wide line is close to the resistivity of bulk copper, about 1.7 μΩ cm, whereas the effective resistivity of a copper nanowire with a cross-sectional area of about 300 nm 2 The effective resistivity of the 18 nm wide lines is more than five times higher, at about ~9 μΩ cm.
[0034] The growth of large copper grains in thin copper wires can lower the effective resistivity of copper, partially reversing the interconnect size effect. By increasing the average copper grain size to greater than about 80 nm, the grain boundary scattering component of the interconnect size effect can be substantially eliminated. Furthermore, electron scattering from wire interfaces can be diffuse or specular. Diffuse scattering causes electron momentum to be transferred to the wire sidewalls, adversely affecting the effective resistivity of copper. Specular scattering does not transfer electron momentum and does not affect the effective resistivity of copper. The design of the copper wire interface is directed at increasing the specular scattering component over the diffuse scattering component. This can further suppress the interconnect size effect and bring the effective resistivity of copper closer to the bulk value.
[0035] The top (M3) copper bit line 38, electrically connected by via conductors to the top electrode 30 of the underlying MTJ pillar, is laser annealed and then cooled. The laser annealing process melts the top copper line for a very short time while the liner 39, top electrode 30, metal capping layer 29, and free layer 27 remain solid. The magnetic properties of the MTJ pillar 32 are essentially unaffected by the laser annealing and recrystallization of the M3 copper bit line 38. The recrystallization of the M3 copper bit line 38 begins at the liner portion 39 at or above the top surface of the top electrode 30. The grain size of the metal or single-phase metal alloy is an estimate of the average grain size. As shown diagrammatically in FIG. 7, a recrystallized top (M3) copper bit line 38' is formed with relatively large grains (e.g., larger than 0.06 μm), increased specular scattering from the wire interface, and, as a result, lower electrical resistance and effective resistivity. In some embodiments, an optically opaque sacrificial absorber layer (not shown) is optionally deposited on the top dielectric capping layer 42 prior to the laser anneal. The purpose of such a sacrificial layer is to uniformly absorb the laser radiation regardless of the metal line material and their pattern density, including any of the capping layer and liner materials. The sacrificial absorber layer can be a conductive film such as tantalum nitride or titanium nitride having a thickness sufficient to absorb all or most of the laser radiation. The sacrificial absorber film is selectively removed down to the top dielectric capping layer 42 after the laser anneal.
[0036] "Laser annealing" is an annealing method that uses a laser to induce heat in the metal interconnects being processed. A laser is an electro-optical device that emits coherent radiation 44. In some embodiments, a typical laser emits light at a defined wavelength in a narrow, low divergence beam. One advantage of using laser irradiation for the annealing process is that very high radiation intensities or energy densities can be achieved in short exposure times because the light of the laser can be easily shaped and focused to specific areas of the annealing surface.
[0037] In some embodiments, the short exposure duration is achieved by raster scanning a focused laser beam over the substrate surface including the dielectric passivation layer 42, the top (M3) copper bit line 38, and the MTJ pillars 32. In this case, the exposure duration measured at the incident intensity full width at half maximum (FWHM) is the beam width in the scan direction divided by the scan speed. In an alternative embodiment, the short exposure duration is achieved by using a pulsed laser. In this case, the laser beam is shaped to achieve the required peak intensity over a selected substrate area, such as over an entire product die or a portion thereof, and the laser is operated in a pulsed mode, such as in the case of a Q-switched laser. The pulse duration of the Q-switched laser at the intensity FWHM determines the exposure time of the substrate. The exposure process is repeated over the entire wafer surface in a step-and-repeat manner with some target beam overlap for adjacent exposures. In some embodiments, the exposure of the wafer surface to the laser beam includes exposures of 1 to 100 pulses. For multiple laser pulses, the total cumulative duration (defined by multiplying the duration of each at full width at half maximum by the number of pulses) is between 100 and 3,000 nanoseconds.
[0038] In some embodiments, the laser type used in the laser annealing method of the present application is selected from solid-state Nd:YAG lasers emitting at 1064 nm and frequency doubled or tripled to emit at 532 nm or 355 nm, respectively, or excimer lasers emitting below 400 nm. Excimer lasers can be driven by chemical reactions involving excited dimers, or excimers, which are short-lived dimers or heterodimers formed from two species (atoms) where at least one is in an excited electronic state. Commonly used excimer molecules include F2 (fluorine, emitting at 157 nm) and rare gas compounds such as ArF (193 nm), KrCl (222 nm), KrF (248 nm), XeCl (308 nm), and XeF (351 nm). Excimer lasers are typically operated in a Q-switched pulsed mode suitable for step-and-repeat pulsed wafer exposure. Solid-state Nd:YAG lasers offer an alternative to excimer lasers due to their stable high output at 1,064 nm that can be efficiently wavelength doubled or tripled to generate radiation at 532 nm or 355 nm. Solid-state lasers can be configured in continuous, pulsed, or Q-switched pulsed modes, suitable for both raster scanning and step-and-repeat pulsed operation. The choice of laser wavelength is important to couple the laser radiation into a suitable absorbing material. Common dielectric materials do not absorb or only slightly absorb (e.g., absorb less than 5% of the coupled radiation) even at short wavelengths from about 350 nm to about 250 nm. On the other hand, metals and metal compounds absorb radiation at wavelengths of about 600 nm or shorter, so shorter wavelength lasers are more preferable for structures with metal layers, such as the copper bit lines described herein. In some embodiments, a XeCl laser (308 nm) can be used, which couples about 30-70% of its incident radiation into the top (M3) copper bitline 38 and the MTJ pillar 32.
[0039] When the substrate including the dielectric passivation layer 42, the top copper bit line 38, and the MTJ pillar 32 is exposed to laser radiation, either through a raster scan or a laser pulse, the temperature of the top copper bit line 38 and the MTJ pillar 32 begins to rise from its base value and then quickly drops. As the temperature of the top copper bit line 38 and the MTJ pillar 32 rises, thermal energy or heat flows into the adjacent structures including the dielectric passivation layers 34, 36, 42 and the underlying bottom electrode structure 26, causing their temperatures to rise in lockstep with the temperature of the top copper bit line 38 and the MTJ pillar 32. A typical temperature-time trace of a nanosecond laser annealing process includes four distinct temperature regions: an initial or base substrate temperature, a heat-up portion, a temperature peak point, and a cool-down portion. Such a triangular temperature-time profile or trace is often referred to as a spike anneal or laser spike anneal. The initial or base substrate temperature can be in the range of 23° C. (room temperature) to 400° C. This temperature is typically set by a hot plate on which the substrate rests. Other heating devices can be used to keep the substrate at the base temperature. Such alternative or additional heating devices can include backside or frontside lamp annealing, microwave heating, and a secondary millisecond-scale preheat laser beam. A laser beam having a wavelength near 308 nm raises the surface temperature of the substrate, including the top (M3) copper bit line 38 and the MTJ pillar 32, from the base temperature to a peak value at a ramp up rate of about 1,000,000,000° C. / sec to about 100,000,000,000° C. / sec. The temperature of the adjacent structures, including the bottom electrode structure 26, rises in lockstep with the temperature of the copper line 38 and the MTJ pillar 32, but at a slightly slower rate, so that its peak value is 20° C. to 100° C. lower than the peak temperature of the copper bit line 38. After laser irradiation, the surface temperature, including the top copper bit line 38 and the MTJ pillar 32, rapidly returns to the base temperature at a temperature ramp down rate of about 300,000,000° C. / sec to about 30,000,000,000° C. / sec.Anneal durations measured near the temperature peak point, typically at a level 50° C. below the peak point, range from about 1 ns to about 500 ns, with about 10 ns to about 100 ns being more typical. Laser anneal process durations are typically specified in terms of radiation exposure duration at FWHM, rather than anneal duration at a temperature level 50° C. below the peak temperature. These duration quantities are related to each other, and in some embodiments, the anneal duration is a fraction (e.g., about 1 / 3) of the radiation exposure duration.
[0040] The laser-induced surface temperature rise is set by the laser incident radiation intensity, the laser pulse or exposure duration, and the optical and thermal properties of the dielectric passivation layer 42, the top copper bit line 38, the MTJ pillar 32, and the underlying substrate structures. A short front-side laser exposure results in non-uniform heating of these substrate structures. The heat penetration depth for nanosecond-scale laser pulses is about 1 micron to about 5 microns for uniform metal structures such as copper lines, and about 100 nanometers to about 500 nanometers for typical interconnect dielectrics. Furthermore, the temperature drop across the non-uniform interfaces is typically less than 10° C. at metal-metal interfaces and less than 100° C. at dielectric interfaces. Thus, the top copper bit line 38 and the MTJ pillar 32 heat with a small temperature gradient from top to bottom, about 0.1-0.3° C. / nm, with the top copper bit line 38 having a higher peak temperature than the bottom electrode 26. The temperature gradient in the dielectric structure is several times larger, typically 0.3-3° C. / nm.
[0041] It will be appreciated that the radiation intensity required for the M3 copper bit-line 38 to reach a target annealing temperature range above the melting point of copper (1090° C.) will depend on the particular underlying substrate structure, laser pulse time, and selected laser wavelength. However, for a particular substrate structure and annealing parameters, the required incident radiation intensity can be determined experimentally by observing the copper bit-line 38 melting at approximately 1090° C. In some embodiments, the nanosecond laser wavelength is 308 nm, the substrate base temperature is 250° C., the pulse duration at FWHM is 160 nanoseconds, and the combined laser energy density at which the top copper bit-line 38 at or above the M3 level begins to melt is 0.1 J / cm. 2 (This is the incident laser energy density of 0.16 J / cm 2 ) The exact metal level of the top copper bit line 38, i.e., the amount of thermal insulation from the FEOL substrate 22, may change this threshold. The presence of metal caps, barriers, and wetting layers such as TaN, Ru, Co, Ta, etc. in the copper bit line 38 may also change this threshold. The pattern density of the top copper bit line 38, the MTJ pillars 32, and the bottom interconnect structure may also change this threshold. Substrate structure and laser parameters may vary, thereby affecting the selection of incident laser energy density, but is generally around 0.05 J / cm. 2 from about 3 J / cm 2 Incident laser energy densities ranging from about 100 nanoseconds to about 3,000 nanoseconds and laser radiation exposure durations of about 5 to about 200 nanoseconds can be used to melt the top copper bit lines 38. In some embodiments, the top copper bit lines 38 are melted multiple times by exposing them to multiple laser pulses with cumulative durations of about 100 nanoseconds to about 3,000 nanoseconds.
[0042] After the laser irradiation, the temperature of the top copper bit line 38 begins to drop rapidly, and the liquid copper is below its melting / freezing point. The liquid copper initiates the solidification process by nucleating solid copper seeds and laterally growing copper grains from the seeds. Excessive random nucleation of solid copper seeds can result in small copper grains as a result of competing lateral growth from adjacent closely spaced copper seeds. The presence of the MTJ pillars 32 below the copper bit line 38 provides an effective local heat sink, creating a relatively cool spot in the portion of the copper bit line 38 directly above the MTJ pillar, facilitating solid copper seed nucleation at that cool location. This effectively programs the position of the solid copper seed in the top copper bit line 38.
[0043] Separating adjacent MTJ pillars along the top copper bit line 38 by about 60 nm or more results in a large grain copper line 38' with a structure 70 as shown diagrammatically in FIG. 7. Also, impurities that are not miscible in liquid copper are pushed out of the copper to the surface and grain boundaries, resulting in a purer copper line with better electrical properties. Furthermore, the solidification of the liquid copper modifies the copper / liner interface, increasing the specular scattering component. The use of materials immiscible in liquid copper for the liner 39 and bit line 38 cap (not shown) immediately adjacent to the elemental copper ensures that these materials are not incorporated within the copper wire, and that the wire interior contains essentially pure elemental copper (or copper alloy, if an alloy is used) with excellent electrical conductivity. Multiple successive copper melt anneals are beneficial to both increase the average grain size and refine the copper from immiscible impurities. 7, the resulting bit line grains 38A', 38B', 38C' upon completion of the laser annealing / recrystallization cycle described herein are each vertically aligned with the three illustrated underlying MTJ pillars 32. The grain boundaries are located primarily within the bit line regions between the MTJ pillars and above the spaces that contain the encapsulation and dielectric layers 34, 36.
[0044] The end result of melting and recrystallization of the copper lines is that the effective copper resistivity is reduced to near the resistivity of bulk copper, ~1.7 μΩcm. In some embodiments, the size effect of the copper wire is reduced by half or more, meaning that the relative increase in effective copper wire resistivity to the bulk copper resistivity per given wire cross-sectional area is reduced by half or more. For example, if the cross-sectional area of copper is 2000 nm 2 The effective resistance of the copper bit lines is reduced from about 3.7 to about 2.7, i.e., by more than 25%; in other words, the increase in effective bulk resistivity relative to the bulk resistivity (~2 μΩcm) is halved to 1 μΩcm.
[0045] Such ultrashort laser pulses avoid damage to the dielectric materials of the layers including the dielectric cap layer 42 and the ILD layer 36 as well as the MTJ pillar 32. The feasibility of the laser annealing technique as described above for a top copper bit line formed on an MRAM pillar is demonstrated in FIG. 8A and FIG. 8B. The graphs in both figures further show the temperatures at which copper and CoFeB melt and the threshold for MTJ function degradation. The exemplary MTJ pillar 32 (shown in the previous figure) includes a functional material including a CoFeB alloy that has a relatively low melting point and is therefore susceptible to high temperature (above 400° C.) annealing. Furthermore, a low temperature (below 400° C.) post-treatment is typically required to prevent diffusion of metal elements across the tunnel barrier layer of the MTJ pillar 32, which may lead to shorting of the device. For an exemplary blanket MTJ stack and exemplary laser annealing parameters, the melting points and corresponding damage thresholds of the MTJ functional materials at the incident laser energy densities were experimentally determined by observing the electrical function of the MTJ stack before and after laser annealing. An exemplary MTJ stack included a copper top electrode capped with a ruthenium (Ru) cap. Different optical reflectivities of Ru at 308 nm and different pattern densities (blanket films) increased the melting threshold of copper to 0.37 J / cm. 2 Thus, in this example, the target annealing temperature range of about 1000 °C to about 1300 °C is achieved with an incident laser energy density of 0.33 J / cm 2from 0.46 J / cm 2 It is necessary that
[0046] Figure 8A shows the change in an MTJ parameter known as the tunnel magnetoresistance effect, plotted against J / cm 2 The results are shown for a range of incident laser energy densities, measured in units, and the resulting calibrated peak annealing temperatures. A high magnetoresistance effect is desirable. As shown in FIG. 8A, the magnetoresistance effect does not change substantially below 1150° C., so laser annealing of the top copper bit line (element 38 in FIG. 6) can be performed up to this temperature without material degradation of the magnetic tunnel junction. The magnetoresistance effect increases at higher annealing temperatures and at approximately 0.4 J / cm. 2 Similarly, in Figure 8B, another parameter of the MTJ, known as the resistance-area (RA) product, is shown before and after laser annealing, which is 0.4 J / cm 2 The results show no degradation up to 1000 . For a given magnetoresistance value, a relatively low RA is desirable. RA increases / degrades at higher laser annealing energy densities. This demonstrates that the copper bit line 38 can be melted for durations of tens to hundreds of nanoseconds without causing degradation of the MTJ pillar 32 located underneath the copper line 38 / 38'.
[0047] The figures above show exemplary process steps / stages in the fabrication of an exemplary structure. While the fabrication method as a whole and the structures formed thereby are entirely novel, the specific individual process steps required to carry out the method may utilize conventional semiconductor fabrication techniques and conventional semiconductor fabrication tools. Given the teachings herein, one of ordinary skill in the relevant art will already be familiar with these techniques and tools. Moreover, one or more of the process steps and tools used to fabricate semiconductor devices are described in many readily available publications, such as, for example, James D. Plummer et al., Silicon VLSI Technology: Fundamentals, Practice, and Modeling 1st Edition, Prentice Hall, 2001, which is incorporated herein by reference. It is emphasized that while several individual process steps are described herein, they are merely exemplary and one of ordinary skill in the art will be familiar with several equally suitable alternatives that may be applicable.
[0048] It should be understood that the various layers and / or regions illustrated in the accompanying figures may not necessarily be drawn to scale. Furthermore, one or more semiconductor or other layers of the type commonly used in such integrated circuit devices may not be explicitly shown in certain figures for ease of illustration. This does not imply that these not explicitly shown semiconductor or other layers are omitted in the actual integrated circuit device.
[0049] In light of the preceding discussion, it will be appreciated that, in general, a method for forming an MRAM structure including improved large grain copper top interconnect lines (bit lines) includes obtaining a first structure including a device wafer 22 including electronic devices and a metallization layer M2 overlying the device wafer and including bottom electrodes 26 electrically connected to the electronic devices. Vertical pillars 32 having sidewalls are formed on the metallization layer, each including a magnetic tunnel junction and a top electrode 30. A barrier (encapsulation) layer 34 is formed on the sidewalls of the pillars 32 to initially encapsulate the pillars, as shown in FIG. 4. A dielectric layer 36 is deposited between the sidewalls of the vertical pillars. A top copper bit line 38 is formed on the top electrode, having a relatively small grain, a width of 40 nm or less, and a thickness of 3,000 nm, as shown diagrammatically in FIG. 6. 2 A capping layer 42 is deposited over the top copper bit line, resulting in an exemplary structure 60 as shown in FIG. 6. The top or M3 copper bit line is melted one or more times by laser irradiation without significantly affecting the magnetic properties of the magnetic tunnel junction in the pillar 32. The method further includes recrystallizing the top copper bit line 38 during a cooling period after the laser irradiation to form relatively large grains and an improved interface. The recrystallization of the top copper bit line proceeds upward from the top electrode 32. The laser annealing parameters are selected to form large grains in the bit line without damaging the MTJ pillar. Exemplary parameters include a laser wavelength of less than 600 nm, a laser exposure time range of 1-500 nanoseconds, and a cumulative laser exposure time of 100-3,000 nanoseconds. The structure can be subjected to 1-100 laser anneal and cool down cycles at temperatures below 1150° C., for example.
[0050] Adjacent MTJ pillars 32 and their top electrodes 30 are spaced apart by 60 nm or more along the bitline 38', resulting in an average copper grain size greater than 60 nm in one or more embodiments. The copper grain boundaries are located primarily between adjacent MTJ pillars, as shown in FIG. 7, while the copper grains are aligned perpendicular to the vertical MTJ pillars. A pitch of MTJ pillars greater than 60 nanometers along the bitline 38' still allows for a high density of MTJ cells, with individual MTJ cells having an area of 0.01 μm in one or more embodiments. 2 The effective resistivity of bitline 38' is reduced by more than 25% compared to bitline 38 described above with respect to Figure 6, and the increase in effective bulk resistivity relative to the bulk resistivity is reduced by half or more. Thus, a structure 70 as shown diagrammatically in Figure 7 can be obtained.
[0051] The resulting memory array according to the inventive principles described herein includes MTJs and 3,000 nm 2 The large grain top copper bitline has a cross-sectional area of less than 0.01 μm, low access resistance, grains aligned vertically to the vertically extending MTJ pillars, and grain boundaries located primarily between adjacent pillars. In one or more embodiments, the pitch of the MTJ pillars along the bitline 38′ is sixty nanometers (60 nm) or greater, which allows for the formation of relatively large grains. In an exemplary memory array, the pitch of the MTJ pillars along the bitline 38′ is sixty nanometers (60 nm) or greater, which allows for the formation of relatively large grains. 2 An individual MRAM cell area of less than 100 μm is provided. The relatively large grains of the recrystallized copper bit lines have an average grain size of greater than 0.06 μm (or greater than twice the bit line width) after the copper bit lines are recrystallized.
[0052] At least some of the techniques described above may be implemented in integrated circuits. In forming integrated circuits, identical dies are typically fabricated in a repeating pattern on the surface of a semiconductor wafer. Each die includes the devices described herein and may include other structures or circuits or both. The individual dies are cut or diced from the wafer and then packaged as integrated circuits. Those skilled in the art will know how to dice wafers and package dies to produce integrated circuits.
[0053] Those skilled in the art will appreciate that the exemplary structures discussed above can be distributed in raw form (i.e., a single wafer having multiple unpackaged chips), as bare die, or in packaged form, or incorporated as part of an intermediate or final product that would benefit from having a structure including magnetic tunnel junctions and associated bit lines formed in accordance with one or more exemplary embodiments.
[0054] The diagrams of the embodiments described herein are intended to provide a general understanding of the various embodiments and are not intended to serve as a complete description of all elements and features of devices and systems that may utilize the circuits and techniques described herein. Many other embodiments will be apparent to those skilled in the art upon receipt of the teachings herein. Other embodiments may be utilized and derived therefrom, such that structural and logical substitutions and modifications may be made without departing from the scope of the present invention. It should also be noted that in some alternative implementations, some of the steps of the exemplary method may be performed in a different order than that depicted in the figures. For example, two steps shown in succession may in fact be performed substantially simultaneously, or certain steps may sometimes be performed in reverse order, depending on the functionality involved. Also, the drawings are merely representational and are not drawn to scale. Thus, the present specification and drawings are to be considered in an illustrative and not restrictive sense.
[0055] In this specification, embodiments may be referred to individually or collectively or both by the term "embodiment" merely for convenience, and are not intended to limit the scope of the present application to any single embodiment or inventive concept when in fact multiple embodiments are shown. Thus, while specific embodiments have been shown and described herein, it should be understood that one or more of the specific embodiments shown may be substituted with mechanisms that achieve the same purpose. Combinations of the above embodiments, as well as other embodiments not specifically described herein, will be apparent to those of ordinary skill in the art given the teachings herein.
[0056] The terms used herein are for the purpose of describing particular embodiments only and are not intended to be limiting of the invention. As used herein, the singular forms "a," "an," and "the" are intended to include the plural unless the context clearly indicates otherwise. Furthermore, as used herein, it will be understood that the terms "comprises," "including," "comprising," or both, specify the presence of the stated features, steps, operations, elements, or components, or combinations thereof, but do not exclude the presence or addition of one or more other features, steps, operations, elements, components, or groups or combinations thereof. Terms such as "top," "bottom," and "vertical" are used to indicate the relative positional relationship of elements or structures to one another, but not to relative heights.
[0057] The corresponding structures, materials, acts, and equivalents of all "means or step plus function" elements in the following claims are intended to include any structures, materials, or acts for performing that function in combination with other claim elements as specifically claimed. The description of the present invention has been presented for purposes of illustration and description, but is not intended to be exhaustive or to limit the invention to the disclosed form. Many modifications and variations that do not depart from the scope and spirit of the invention will be apparent to those skilled in the art. The embodiments have been selected and described to best explain the principles and practical applications of the invention and to enable those skilled in the art to understand the invention in terms of various embodiments with various modifications as suited to the particular applications contemplated.
[0058] The Abstract is provided to comply with 37 CFR 1.72(b). It is submitted with the understanding that it will not be used to interpret or limit the scope or meaning of the claims. Moreover, in the foregoing Detailed Description, it will be seen that various features are grouped together in a single embodiment for the purpose of streamlining the disclosure. This method of disclosure is not to be interpreted as reflecting an intention that the claimed embodiments require more features than are expressly recited in each claim. Rather, as the appended claims reflect, claimed subject matter may reside in less than all features of a single embodiment. Accordingly, the following claims are hereby incorporated into this specification with each claim standing on its own as separately claimed subject matter.
[0059] Given the teachings provided herein, one skilled in the art will be able to contemplate other implementations and applications of the present technology and the disclosed embodiments. Although exemplary embodiments have been described herein with reference to the accompanying drawings, it will be understood that the exemplary embodiments are not limited to those precise embodiments, and that various other changes and modifications may be made thereto by those skilled in the art without departing from the scope of the appended claims.
[0060] In a preferred embodiment of the invention described herein, a method is provided for forming large grain copper bit lines for electronic structures, the method comprising obtaining an MRAM structure including vertical pillars having a pitch of 60 nanometers or greater, each vertical pillar including a magnetic tunnel junction and a top electrode, a plurality of spaces, each space being between and separating each of the vertical pillars, and a dielectric layer filling each of the plurality of spaces between the vertical pillars, and forming a copper bit line on top of the top electrode, extending along the pitch of the vertical pillars and over the spaces between the vertical pillars, the copper bit line having a width of 40 nanometers or less and an area of 3000 square nanometers or less. Cut off Area and width of copper bit lines Compared to Small stomach forming copper bit lines having copper grains of size 100; laser annealing the copper bit lines, thereby melting the copper bit lines; and recrystallizing the copper bit lines during a cooling period after the laser anneal, thereby having grain boundaries that are vertically aligned with the vertical pillars and that are primarily located above the dielectric layer in the spaces between the vertical pillars. Compared to the width of the copper bit lines and causing large grains to form in the copper bit line. Preferably, laser annealing the copper bit line comprises multiple successive copper melt anneals and recrystallizing the copper bit line comprises multiple cooling periods between the successive copper melt anneals. The laser annealing preferably has a laser power of 0.05-3 J / cm 2 The laser annealing is performed at a laser annealing density of 1000 nm to 1000 nm. Preferably, the MRAM structure further includes a metal liner that is immiscible with liquid copper and extends over the top electrode and the dielectric layer, and further includes forming the copper bit lines directly on the metal liner. Laser annealing the copper bit lines preferably includes causing the annealing to occur at or below 1150° C., and the metal liner remains in a solid state during the laser annealing.
Claims
1. 1. A method for forming large grain copper bit lines for electronic structures comprising: a device wafer including an electronic device; a metallization layer overlying the device wafer and including a bottom electrode therein, the bottom electrode being electrically connected to the electronic device; vertical pillars having sidewalls on the metallization layer, each of the vertical pillars including a magnetic tunnel junction and a top electrode; and obtaining an MRAM structure comprising: forming a copper bit line on the top electrode that exhibits a copper wire size effect, the copper bit line having a width of 40 nanometers or less, a cross-sectional area of 3000 square nanometers or less, and a copper grain size that is small compared to the width of the copper bit line; laser annealing the copper bit lines, thereby melting the copper bit lines; recrystallizing the copper bit lines during a cooling period after the laser anneal, thereby forming grains in the copper bit lines that are large compared to a width of the copper bit lines and reducing the copper wire size effect of the copper bit lines formed on the top electrode; A method comprising:
2. 2. The method of claim 1 , wherein the large grains formed by recrystallizing the copper bit lines are aligned vertically to the vertical pillars and include grain boundaries primarily above spaces between the vertical pillars, and further wherein the vertical pillars have a pitch along the copper bit lines of 60 nanometers or greater.
3. 2. The method of claim 1 , wherein after recrystallizing the copper bit lines, the large grains have an average grain size greater than 0.06 μm, and further wherein recrystallizing the copper bit lines comprises reducing the copper wire size effect by half or more.
4. 4. The method of claim 3, wherein the MRAM structure further comprises a dielectric layer between each of the vertical pillars, and the large grains include grain boundaries that extend primarily above the dielectric layer.
5. The laser annealing may be performed at a temperature of from 0.05 to 3 J / cm 2 The method of claim 4 , wherein the laser annealing density is
6. 6. The method of claim 5, wherein the tunneling magnetoresistance effect of each magnetic tunnel junction is substantially the same before and after recrystallizing the copper bit lines.
7. 6. The method of claim 5, wherein the MRAM structure further comprises a metal liner that is immiscible with liquid copper and extends over the top electrode and the dielectric layer, and further wherein the copper bit lines are formed directly over the metal liner.
8. 10. The method of claim 1, wherein the vertical pillars have a pitch along the copper bitlines of at least 60 nanometers.
9. 9. The method of claim 8, wherein laser annealing the copper bit lines comprises multiple successive copper melt anneals.
10. 1. A magnetoresistive random access memory structure comprising: a plurality of vertical pillars having sidewalls, each of the plurality of vertical pillars including a magnetic tunnel junction and a top electrode, the plurality of vertical pillars having a pillar pitch of 60 nanometers or greater; a plurality of spaces, each of the plurality of spaces being between an opposing pair of the sidewalls of adjacent vertical pillars; a dielectric layer filling each of the plurality of spaces; a copper bit line overlying and electrically connecting each of the top electrodes, the copper bit line having a line width of 40 nanometers or less and grains having a grain size greater than twice the line width and an average grain size of 0.06 μm or more, the pillar pitch being along the copper bit line; A magnetoresistive random access memory structure comprising:
11. 11. The magnetoresistive random access memory structure of claim 10, wherein the grains of the copper bit lines are aligned perpendicular to the vertical pillars, the grains including grain boundaries that lie primarily above the plurality of spaces.
12. 11. The magnetoresistive random access memory structure of claim 10 further comprising a metal liner comprising a material immiscible with liquid copper, said metal liner directly contacting said top electrode and extending over said dielectric layer.
13. The copper bit line is 3,000 nm 2 13. The magnetoresistive random access memory structure of claim 12 having a cross-sectional area of less than .
14. The vertical pillar includes one or more MRAM cells, each of the one or more MRAM cells being 0.01 μm thick. 2 13. The magnetoresistive random access memory structure of claim 12 having an area of:
Citation Information
Patent Citations
Method for manufacturing a metal-containing thin layer with low electrical resistance
JP2005532675A
Magnetic tunnel junction devices and fabrication
JP2012527129A
Semiconductor device, semiconductor storage device and method of manufacturing the semiconductor device
US20150371999A1
Co-fabrication of magnetic device structures with electrical interconnects having reduced resistance through increased conductor grain size
US20180366640A1
Semiconductor integrated circuit device and method for producing the same
WO2010064732A1
