JP2003524888A - Method for depositing nanolaminate thin films on sensitive surfaces - Google Patents

Abstract

SUMMARY OF THE INVENTION The present invention provides means for growing conformal metal nitrides, metal carbides and metal films from aggressive chemicals, and nanolaminate structures incorporating these films. The amount of corrosive compounds (eg, hydrogen halide) is reduced during the deposition of transition metal, transition metal carbide, and transition metal nitride thin films on various surfaces (eg, metals and oxides). The getter compound protects surfaces that are sensitive to hydrogen and ammonium halides (eg, aluminum, copper, silicon oxide, and deposited layers) from corrosion. Also disclosed are nanolaminate structures (20) incorporating metal nitrides (eg, titanium nitride (30) and tungsten nitride (40)) and metal carbides, and methods for forming the same.

Description

Detailed Description of the Invention

FIELD OF THE INVENTION The present invention generally relates to alternating self-saturating chemical reactions.
hemistries) for depositing thin films on substrates. More particularly, the present invention relates to using corrosive species during thin film formation while preventing corrosion of materials on a substrate.

[0002] First BACKGROUND OF THE INVENTION Atomic Layer Epitaxy (Atomic Layer Epitaxy), known as (ALE), atomic layer deposition (Atomic Layer Deposition) (ALD) is a advanced variation of CVD. The ALD process is a series of self-saturating surface reactions (sequen
tial self-saturated surface reactions). Examples of these processes are described in detail in US Pat. Nos. 4,058,430 and 5,711,811. The deposition process described benefits from the use of inert carriers and purging gases, which speeds up the system. Due to the self-saturating nature of the process, ALD enables near-perfect conformal deposition of atomically thin levels of film.

The technique was first developed for the conformal coating of substrates for flat panel electroluminescent displays, which desirably exhibit a high surface area. More recently, ALD has found use in the manufacture of integrated circuits. The extraordinary conformality enabled by the technology
) And control are well suited for the increasingly demanded scaled-down dimensions of the state of the art semiconductor processing.

While ALD has many potential applications in semiconductor manufacturing, integrating these new processes into established process flows is
It can lead to many new problems. Therefore, there is a need for an improved ALD process.

[0005] According to the gist one aspect of the present invention, a method for forming nano laminate structure (nanolaminate structure) by atomic layer deposition (ALD) type processes on a substrate in the reaction space is provided. The nanolaminate structure has at least two adjacent thin film layers that include at least one metal compound layer. Each thin film layer is in a different phase than the adjacent thin film layer.

According to another aspect of the invention, a nanolaminate structure having at least three thin film layers is provided. Each layer has a thickness of less than about 10 nm. At least one of the layers is selected from the group consisting of metal carbides and metal nitrides.

According to another aspect of the invention, there is provided a method for depositing a material on a substrate in a reaction space. The substrate has a surface that is susceptible to a halide attack. The method involves alternating pulses of reactants in multiple deposition cycles.
pulses), where each cycle includes: supplying a first reactant to deposit a halide-terminate species on the surface.
chemisorbing a monolayer of about 1 or less of d species; removing excess first reactant and reaction byproducts from the reaction space; and halogen from the monolayer before repeating the cycle. Gettering compounds.

Detailed Description of the Preferred Embodiments The present disclosure teaches a method for protecting sensitive surfaces during ALD deposition. Those skilled in the art understand that while applicable to nanolaminate construction, protection of sensitive surfaces from corrosion has applications in other contexts, and vice versa.

DEFINITIONS For purposes of this description, the "ALD process" refers to a sequence of alternating and self-saturating surpluses of material deposition on a surface.
face reaction). The general principles of ALD are disclosed, for example, in US Pat. Nos. 4,058,430 and 5,711,811, the disclosures of which are incorporated herein by reference.

“Reaction space” is used to mean a reactor or reaction chamber, or any defined volume therein, where conditions can be adjusted by ALD to effect thin film growth.

“Adsorption” is used to mean the attachment of atoms or molecules on a substrate.

“Surface” is used to mean the boundary between the reaction space and the features of the substrate.

“Getter”, “gettering agent” or “gettering agent”
A scavenger "is a volatilization that can form new volatile compounds from halogen or halide species adsorbed on the surface or from halides in the reaction space, such as hydrogen halides or ammonium halides. Used to refer to a sex chemical species, typically new halogen compounds are more preferred than hydrogen or ammonium halides.
A) less corrosive to exposed features.

The symbols “-” and “=” attached at one terminus to an atom refer to the number of bonds to the unspecified atom or ion.

The subscript “x” in metal nitrides (eg WN _x or TiN _x ) means transition metal nitrides that have a wide range of phases with varying metal / nitrogen ratios and are not necessarily stoichiometric. Used to

The subscript “x” in a metal carbide (eg, WC _x or TiN _x ) means a transition metal carbide, not necessarily stoichiometric, having a wide range of phases with different metal / carbon ratios. Used for.

“Nanolaminate structure” means a layered structure comprising stacked thin film layers of different phases with respect to the growth direction of the nanolaminate. "Alternating" or "stacked" means that adjacent thin film layers are different from each other. In nanolaminate structures, there are always at least two phases of the molecule. Preferably, there are at least 3 adjacent phases. A single phase exists when molecules or atoms are evenly mixed in space so that differences cannot be detected by analytical methods between different parts of space. Different phases are any differences recognized in the art, such as the phase interface (ph
The crystal structure, crystallite lattice parameter, crystallization stage, electrical conductivity and / or chemical composition of the thin film on either side of the ase interface).

Desirably, each phase or layer in the stack is thin,
Preferably it is less than about 20 nm thick, more preferably less than about 10 nm, and most preferably less than about 5 nm. A "thin film" is a vacuum from a source to a substrate,
Means a film grown from an element or compound that is transported as discrete ions, atoms or molecules through the gas or liquid phase. The thickness of the membrane depends on the application and can vary widely, preferably over one atomic layer to 1,000 nm.

“Metallic thin film” means a thin film consisting essentially of metal. Depending on the reducing agent,
The metal film may include some metal carbide and / or metal boride in an amount that does not negatively affect the characteristic metal properties of the film, or the characteristic properties of the nanolaminate.

Integration issues halides in general, and transition metal halides in particular, are attractive source chemistries for ALD due to their high volatility and resistance to thermal decomposition. Source chemicals. Of these halides, compounds that are liquids or gases near room temperature (eg, TiCl ₄ and WF ₆ ) are preferred because they do not generate solid particles in the source vessel. In addition to their volatility,
Many such halide compounds are particularly useful for ALD processing because they are of the species of interest (eg, metal-containing species).
It enables the chemisorption of
) Because it leaves one or less monolayers of the terminated species. The halide tail prevents further chemisorption or reaction of the species of interest, so that the process is self-saturating and self-limiing.
ting).

Metal halides can be used, for example, in the formation of metal, metal nitride and metal carbide thin films by the ALD process. However, these processes did not produce the desired perfectly conformal deposition of ALD. The discussion of FIG. 2 and Examples 1, 2 and 4 is sustained, for example, by "exposed" copper during ALD formation of metal nitrides and carbides using alternating metal halides with ammonia. Demonstrate the corrosive damage caused. In fact, Example 4
Demonstrates that such damage can be sustained even when copper is covered by 5 nm of tungsten metal.

ALD Using Metal Halide and Source Chemistry with High Hydrogen Content
The process may release hydrogen halides (eg, HF, HCl) as reaction byproducts. These reactive byproducts can destroy certain metal surfaces, leaving deep pits in the metal or even removing all metal. Silicon dioxide is also susceptible to corrosion due to the formation of volatile silicon halides. These hydrogen halides may also react with other reactants during the ALD phase (eg, excess N 2 during the nitrogen phase).
H ₃ and) combined with further harmful species exacerbate corrosion problems (e.g., ammonium halide (e.g. NH ₄ F) may form). Thus, by-products from alternating halide- and hydrogen-bearing reactants tend to corrode exposed materials (eg, aluminum, copper and silicon dioxide) in partially manufactured integrated circuits.

Preferred Workpiece A preferred embodiment involves deposition of metal, metal carbide and metal nitride thin films by ALD on the surface of a substrate. In one embodiment, the thin film forms a nanolaminate. More specifically, embodiments include deposition on "sensitive" surfaces that are susceptible to corrosion in the presence of halides and especially hydrogen halides. Such sensitive surfaces include, for example, metals such as aluminum and copper, and silicon compounds such as silicon oxide and silicon nitride.

As described in more detail below, such sensitive surfaces generally have a negative or near-zero Gibbs free energy (, about the reaction between the surface and the hydrogen halide or ammonium halide). ΔG _f ).

FIG. 6 illustrates a dual damascene context in which deposition is desired over multiple such materials at the same time. The structure is a plasma enhanced CVD (plasma) using a tetraethyl orthosilicate (TEOS) precursor.
It includes a first or lower insulating layer 50, in the form of silicon oxide, specifically deposited by enhanced CVD (PECVD). Insulation layer 50 is formed over barrier layer 51 (typically silicon nitride), which in turn overlies conductive elements 52. Conductive element 52 typically consists of a highly conductive wiring metal, and most preferably copper, in a dual damascene context. An etch stop (etc.) formed over the first insulating layer 50 from a material having a significantly different etch rate compared to the underlying insulator 50.
h stop) 54. The etch stop layer 54 (typically silicon nitride) includes a plurality of openings 55 across the workpiece that serve as hard masks in defining contact vias. A second or top insulating layer 56 (also PECVD TEOS) is formed over the etch stop 54, and a polishing shield 58 is formed in a subsequent chemical mechanical planarizing (CMP) step. Stop. Polishing shield 58 typically comprises a relatively hard material (eg, silicon nitride or silicon oxynitride).

As will be appreciated by those skilled in the art, the dual damascene structure is formed by a photolithography and etch process that defines a plurality of trenches 60 having contact vias 62 extending from the trench floor in discrete arrangements. Trenches 60 help define wiring patterns for interconnection of electrical devices according to integrated circuit designs. The contact vias 62 define the layout in which electrical connection to the bottom electrical element or wiring layer is desired according to the circuit design.

Those skilled in the art will appreciate that a variety of alternating materials and structures can be used to achieve these ends. For example, the preferred insulating layers 50, 56 are PECVD TE.
While including the OS, in other arrangements, the materials for these layers may include any of a number of other suitable dielectrics. For example, dielectrics have recently been developed that exhibit a low dielectric constant (low k) when compared to conventional oxides. These low k dielectrics are
Includes polymeric materials, porous materials and fluoride-doped oxides. Similarly, the barrier 51, etch stop 54 and shield 58 may include any of numerous other materials suitable for their aforementioned functions. Moreover, any or all layers 51, 54 and 58 may be omitted in other schemes for manufacturing dual damascene structures.

The dual damascene trenches 60 and vias 62 are then lined with a thin film 150, as shown in FIG. The thin film 150 may be selectively formed over a particularly desired surface of the structure, but most preferably, according to a preferred embodiment, a blanket, conformal deposition by ALD.
al deposition). In the illustrated embodiment, the thin film is conductive, allowing electrical signals to flow therethrough.

Integrated circuits include interconnects, typically made of aluminum. Recently, copper has become an attractive material in the field. However, copper easily diffuses into the surrounding material. Diffusion affects the electrical properties of the circuit and can cause active components to malfunction. Diffusion can be prevented by an electrically conductive diffusion barrier layer. Amorphous films are believed to enhance the properties of diffusion barriers because ion diffusion prefers thin film grain boundaries. Preferred diffusion barriers are transition metal nitrides (eg TiN _x , TaN _x and WN _x ). We also found that metal carbides (eg, WC _x ) are good conductive diffusion barriers.
usion barriers).

Conventionally, a thin lining film in a dual damascene structure is used.
lm) is a conductive adhesion sub-layer (eg, tungsten metal), a barrier sub-layer (eg, titanium nitride) and a seed sub-layer (eg, titanium nitride). For example, P
VD copper). The preferred thin film 150 may include one or more of these sub-layers formed by ALD, and may also include one or more sub-layers formed by other methods. Preferred embodiments include, for example, a method of forming tungsten metal by ALD over oxide and copper structures without etching. However, in general, minimizing the thickness of the lining layers, which results in a highly conductive metal (preferably copper) that is subsequently deposited.
It is desirable to maximize the volume of the structure occupied by. To this end, the preferred embodiment also provides for depositing a barrier layer directly over both oxide and copper surfaces (or other sensitive surfaces) without etching the sensitive surfaces, or for corrosion. It provides a means for depositing a barrier layer over a very thin adhesion layer.

As will be appreciated by those skilled in the art, formation of thin film 150 is followed by trench 60.
And vias 62 may be filled with a highly conductive material such as electroplated copper. A polishing process then ensures that the individual lines are separated within trench 60.

Nanolaminate Structure Nanolaminates have enhanced diffusion barrier properties.
es) is a layered structure. Nanolaminates consist of multiple thin films and are constructed to create very complex diffusion paths for impurities by the usual disruption of crystal growth during deposition. Thus, the nanolaminate comprises alternating thin film layers of different phases, for example having different crystallite structures and different crystallite lattice parameters.

According to a preferred embodiment of the present invention, the nanolaminate structure is formed on the substrate. The nanolaminate structure is preferably composed of at least one transition metal compound thin film layer, which is desirably conductive and serves a diffusion barrier function. The metal compound can be a metal nitride or a metal carbide. The nanolaminate structure can also include one or more elemental metal thin film layers.

The nanolaminate structure is preferably a layered structure comprising alternating stacked thin film layers of materials having different phases with respect to the growth direction of the nanolaminate. The nanolaminate structure preferably comprises materials with at least two different phases. Therefore, at least two adjacent thin film layers preferably have different phases.
For example, they can have different structures, compositions or electrical resistivities from each other. In a nanolaminate with three layers, at least one of the layers preferably has a different phase than the other two layers.

The nanolaminate structure preferably comprises at least two thin film layers. More preferably, they include at least 3 thin film layers. 3 nano-laminate structure
When including two membrane layers, it is preferably a "sandwich" structure in which the middle layer has a different phase than the two outer layers.

Preferably, the nanolaminate is grown such that the phases alternate with the layers. Therefore, each other layer is preferably in phase. However, all thin films of one nanolaminate structure can be in different phases, for example when each thin film layer is made of a different material. This structure impairs the diffusion of ions in the structure (im
pair) to obtain a large number of phase interfaces.

An example of a nanolaminate structure is shown in FIG. 9, which is a schematic representation of the first four thin film layers of metal nitride nanolaminate made by the ALD type process of the present invention. Pulsing sequence to obtain each layer
Are shown in FIG. The layers are not proportional to a certain proportion in the figure (not to scale
), And the subscripts x, y, a and b are integers.

FIG. 10 shows uniform growth and sharp interfaces between layers in the preferred nanolaminate. FIG. 10 shows 1.8 n of clearly distinct titanium nitride 30 (light gray).
3 is a transmission electron microscope (TEM) photograph of a nitride nanolaminate structure showing an m thin film layer and a 4.5 nm thin film layer of tungsten nitride 40 (dark gray).

The number of layers stacked in the nanolaminate structure can vary, but from 2 to 500,
Preferably it can vary from 3 to 300, more preferably 4 to 250, and even more preferably 4 to 20. The thickness of the nanolaminate structure is preferably from 2 molecular layers to
1,000 nm, more preferably 5 nm to 200 nm, and even more preferably 10 nm to 100 nm. Desirably, each layer is thin, preferably 20 each.
Sub-nm thickness, more preferably less than about 10 nm, and most preferably less than about 5 nm.

The thin film layers making up the nanolaminate structure of the present invention preferably have different phases and properties from each adjacent layer. These differences can be in the following properties, but those skilled in the art will recognize that other properties are contemplated and that the properties will vary depending on the type of thin film in the laminate structure: Crystallite Structure The crystallite structure changes according to the deposited species as well as according to the metal / nitrogen ratio of the nitride thin film layer. Variations in crystal structure are space groups, unit cell dimensions
And can occur in a number of details including the orientation of crystallites in the thin film layers.

There are 230 space groups such as face-centered, cubic and hexagonal. Thus, nanolaminate structures can be made by depositing alternating thin film layers on a substrate, each having hexagonal and cubic space groups of crystallites. Variations in space groups can change unit cell size.

The unit cell is the smallest repeating atomic arrangement within the crystallite, and the size of the unit cell can vary. For example, a nanolaminate can be made by depositing alternating thin film layers that include materials with small and large unit cells.

The crystal orientation in the thin film layer according to the Miller index may also change. For example, the nanolaminate can have the following structure: (100) / (111) / (100) /
(111) /. ． ． .

[0044]   2.composition  The composition of the metals in the illustrated nanolaminate includes metal nitrides.
/ Nitrogen ratio, or metal / charcoal in an exemplary nanolaminate containing metal carbide
It refers to atomic make-up, such as the prime ratio. Different due to metal / nitrogen ratio
An example of a nanolaminate structure containing the following phases is: Ta₃N_Five/ TaN / Ta ₃ N_Five/ TaN /. ． ． . Another example is that some thin film layers contain nitrogen and others contain.
Structure, for example, W / WN / W / WN /. ． ． Is.

3. Electrical resistance Electrical resistance also varies according to the metal / nitrogen ratio. Amorphous or near-amorphous structures may have distinctly different resistivities when compared to each other. In general, the more nitrogen present in the thin film layer, the higher the resistivity. An example of a possible nanolaminate structure is one that includes alternating thin film layers of materials that have low and very low resistivity.

The nanolaminate of the present invention can be used, for example, as a diffusion barrier in integrated circuits.
n barriers). They are also reflectors for x-rays (
can be used as a reflector). Nanolaminate structures suitable for such applications preferably include thin film layers of high atomic number transition metals or high atomic number transition metal nitrides and low atomic number elements or nitrides. In the context of this invention, an atomic number is considered "high" when it is at least about 15 or higher and "low" when it is about 14 or lower. High atomic number nitrides are preferably prepared using source materials that include tungsten or tantalum. The low atomic number nitrides are preferably inorganic nitrides, especially beryllium, boron, magnesium, aluminum and silicon nitrides. Preferably, the thin film layers are arranged in the nanolaminate such that layers with high atomic number nitride alternate with layers with low atomic number nitride.

The nanolaminate structures described herein, including metal nitrides or carbides on other conduction barrier layers, are particularly suitable for interconnect barriers, as described with respect to FIG. Moreover, these materials are susceptible to attack from hydrogen and ammonium halides in the process of deposition. Therefore, the method of deposition described below enables good quality nanolaminate structures.

Preferred ALD Method The method presented herein enables the deposition of conformal thin films and nanolaminates from aggressive chemicals on chemosensitive surfaces. Geometrically difficult applications are possible due to the use of self-limited surface reactions.

According to a preferred embodiment, the thin film, in particular the nanolaminate structure, has an atomic layer deposition (
Atomic Layer Deposition (ALD) type process allows integrated circuit workpieces containing surfaces susceptible to halide attack.
) Or formed on a substrate. Such sensitive surfaces can take a variety of forms. Examples include silicon, silicon oxide (SiO ₂ ), coated silicon.
), Low-k materials, metals (eg, copper and aluminum), alloys, metal oxides and various nitrides (eg, transition metal nitrides and silicon nitrides) or combinations of these materials. As described above with reference to FIGS. 6 and 7, the preferred damascene and dual damascene contexts include a silicon oxide based insulator and exposed copper lines under the contact vias.

The substrate or workpiece placed in the reaction chamber is subjected to alternating surface reactions of source chemicals to grow a thin film. In particular, thin films are formed by a cyclic process in which each cycle deposits, reacts or adsorbs a layer on a workpiece in a self-limiting manner. Preferably, each cycle comprises at least two different phases, where each phase is self-liming.
It is a saturative reaction with an iting effect. Thus, the reactants are determined, under favorable conditions, by the number of available sites for the amount of reactants that can be bound to the surface and incidentally by the physical size of the chemisorbed species (including the ligand). To be selected. The layer left by the pulse is self-terminat at a surface that is non-reactive with the remaining chemistry of the pulse.
ed). This phenomenon is referred to herein as "self-saturation."

Maximum step coverage on the surface of the workpiece
) May be obtained when about 1 or less monolayers of source chemical molecules are chemisorbed in each pulse. Each next pulse is the surface left by the previous pulse,
It also reacts in a self-limiting or self-terminating manner. The pulsing sequence is repeated until a thin film of desired thickness, or a nanolaminate with the desired structure, is grown.

According to a preferred embodiment, the pulsed reactants are selected to avoid etch damage to the workpiece surface. Example 8 below provides one embodiment in which the reactants do not significantly etch the surface.

More preferably, the reactant comprises a chemical species that is detrimental to the substrate. But each AL
The getter phase during the D cycle scavenges harmful species, thereby allowing the use of advantageous volatile reactants to aid self-saturation in each phase while protecting sensitive surfaces. . For example, Examples 3 and 5-7
Scavenging or gettering during each cycle
A deposition process including a phase is disclosed. For metal thin film deposition (Example 3), at least two different source chemistries are used in alternation, one gettering a halide from the other chemistry. For metal nitride thin film depositions (Examples 5-7), at least three different source chemistries are used in alternation: forming a monolayer of about 1 or less terminating with a halogen ligand and as the layers are deposited A first reactant containing a desired chemical species for: a getter for scavenging a halide from a monolayer; and a second reactant containing another chemical species (especially nitrogen) desired when the layer is deposited.

FIG. 8 generally illustrates a three-phase cycle for depositing a binary material. However, one of ordinary skill in the art will readily appreciate that the principles disclosed herein can be readily applied to depositing ternary or higher complex materials by ALD.

The semiconductor workpiece containing the sensitive surface is loaded into the semiconductor processing reactor. An exemplary reactor specifically designed to enhance the ALD process is commercially available from ASM Microchemistry of Finland under the trade name Pulsar 2000 ^™ .

If necessary, exposed surfaces of the workpiece (eg, trench and via sidewall surfaces and metal floor, shown in FIG. 6) are terminated (t).
erminated), reacting with the first phase of the ALD process. The first phase of the preferred embodiment is, for example, hydroxyl (OH) or ammonia (NH ₃ ) terminati
on). In the examples discussed below, the dual damascene structured silicon oxide and silicon nitride surfaces do not require separate termination. Certain metal surfaces, such as the bottom of via 61 (FIG. 9A), may be terminated, for example, by ammonia treatment.

After initial surface termination, if necessary, a first reactant pulse is then delivered 102 to the workpiece. According to a preferred embodiment, the first reactant pulse comprises a volatile halide species reactive with the carrier gas flow and the workpiece surface of interest and further forms part of the deposited layer. Including chemical species. Therefore, halogen-containing species are adsorbed on the workpiece surface. In the illustrated embodiment, the first reactant is a metal halide and the thin film formed comprises a metallic material, preferably a metal nitride. The first reactant pulse self-saturates the workpiece surface such that none of the excess components of the first reactant pulse further react with the monolayer formed by this process. Self-saturation results in termination of the monolayer and protection of the layer from further reactions due to the halide tail.

The first reactant pulse is preferably supplied in gaseous form and is therefore referred to as a halide source gas. In some cases, the reactive species is
May have a melting point above the process temperature (eg CuCl melts at 430 ° C.,
On the other hand, the process is performed at about 350 ° C). Nevertheless, a halide source gas if the species exhibits a vapor pressure sufficient to carry the species to the workpiece under process conditions at a concentration sufficient to saturate the exposed surface. Are considered "volatile" for the purposes of this specification.

The first reactant is then removed 104 from the reaction space. Preferably step 1
04 preferably diffuses or purges excess reactants and reactant by-products from the reaction space with a reaction chamber volume of greater than about 2 purge gas, and more preferably with a chamber volume greater than about 3. It is only necessary to stop the flow of the first chemical reaction while continuing to flow the carrier gas for a sufficient time. In the illustrated embodiment, the removal 102 includes continuing to flow the purge gas for about 0.1 to 20 seconds after stopping the flow of the first reactant pulse. Inter-pulse purging has a serial number of 09 / 392,371 and 199
Filed on September 8, 9th and IMPROVED APPARATUS AND METHOD FOR GROWTH OF
It is described in a co-pending US patent application entitled A THIN FILM, the disclosure of which is incorporated herein by reference. In other arrangements, the chamber may be fully evacuated between alternating chemical reactions. For example, METHOD AND APPARAT
PC released on June 6, 1996 entitled US FOR GROWING THIN FILMS
See T Publication No. WO 96/17107. This disclosure is incorporated herein by reference. Adsorption 102 and reactant removal 104 together represent the first phase 105 in the ALD cycle. The first phase may also be referred to as the halide phase.

When the reactants of the first reactant pulse are removed from the chamber 104, getter pulses are delivered to the workpiece. The getter pulse captures or removes 106 the ligand termination (eg, by ligand-exchange, sublimation or reduction) of the adsorbed composite monolayer formed in step 102. The getter species saturate the workpiece surface, preferably with carrier flow, to ensure removal of halide tails before further pulsing. Temperature and pressure conditions are preferably arranged to avoid diffusion of the getter through the monolayer into the underlying material.

As will be better understood from the more detailed discussion below, the reaction between the halide tail and the getter species in the adsorbed monolayer is thermodynamically favored.
More specifically, the reaction between the getter and the halide-terminated monolayer is generally
Characterized by the negative Gibbs free energy. Therefore, the halide species is more easily transferred to the rest of the adsorbed complex formed in the first phase 105,
It binds to the getter species (or to its reaction byproduct in the case of scavenging by ligand-exchange). Similarly, getters can bind to free halides in the reaction space.

The getter-halide complex (desirably also volatile) is then also removed 108 from the reaction space, preferably by a purge gas pulse. Removal may be as described for step 104. Scavenger pulse 106
And removal 108 together form the second phase 109 of the illustrated ALD process.
, Which may also be referred to as the scavenger or getter phase.

The first two phases are sufficient to form a metal film, such as a metal film layer in a nanolaminate structure. However, for the formation of binary metal layers (eg metal nitride layers) one additional phase is preferably used. In other arrangements,
The getter may leave the component in place of the halide. For example, a triethylboron getter can leave carbon when scavenging fluorine from a tungsten composite.

In the illustrated embodiment, a second reactant pulse is then delivered 110 to the workpiece. The second chemical reaction desirably reacts with or is adsorbed on the monolayer left by the getter phase 109. The getter phase is particularly useful when the second reactant contains a hydrogen-bearing compound that tends to form hydrogen halides. In the illustrated embodiment, the second reactant pulse 110 includes supplying a carrier gas with a hydrogen-bearing nitrogen (eg, NH ₃ ) source gas to the workpiece. The nitrogen or nitrogen-containing species from the second reactant preferably reacts with the previously adsorbed monolayer, leaving a nitrogen compound. In particular, when the first reactant comprises a metal halide, the second reactant leaves a monolayer of about 1 or less of metal nitride.
The second reactant pulse 110 also limits deposition in the saturated reaction phase.
Leave a surface termination that behaves like. The nitrogen and NH _x tails that terminate the metal nitride monolayer are non-reactive with the NH ₃ of the second reactant pulse.

After a time sufficient to fully saturate and react the monolayer with the second reactant pulse 110, the second reactant is removed 112 from the workpiece. Similar to the removal of the first reactant 104 and the removal of the getter species 108, this step 112 is preferably a second step.
Stopping the flow of chemistry and keeping the carrier gas flowing for a time sufficient for the excess reactants and reaction byproducts from the second reactant pulse to diffuse out of the reaction space and be purged. Including. Second reactant pulse 110 and removal 1
12 together represent the third phase 113 in the illustrated process, and may also be considered the nitrogen or hydrogen phase because nitrogen reacts with and forms part of the grown film. At the same time, hydrogen is released in the reaction.

In the illustrated embodiment in which the three phases alternate, once the excess reactants and byproducts of the second chemical reaction have been removed from the reaction space, the first phase of the ALD process is repeated. . Therefore, again supplying the first reactant pulse to the workpiece 1
02 forms another self-terminating monolayer.

Thus, the three phases 105, 109, 113 together represent one cycle 115 that is repeated to form a metal nitride monolayer in an ALD process. The first reactant pulse 102 generally reacts with the termination left by the second reaction pulse 110 in the previous cycle. This cycle 11
5 is repeated a sufficient number of times to create a film that is thick enough to perform its desired function.

Although illustrated in FIG. 8 with only the first and second reactants with an intermediate getter phase, it is understood that in other arrangements additional chemistries may also be included in each cycle. For example, if desired, cycle 115 can be extended to include different surface preparations. Moreover, the second getter phase may be performed in each cycle after the nitrogen phase 112. The cycle 115 then continues with steps 102-112. Furthermore, although illustrated in the examples below with a first metal phase and a second nitrogen phase, it is understood that the cycle can be initiated in the nitrogen phase depending on the exposed substrate surface and phase chemistry.

In the production of the nanolaminate, after the first monolayer of metal, metal carbide or metal nitride is deposited, the starting materials, pulsing parameters and cycles are preferably different for the phases of the next monolayer and The phase interface is altered so that it forms between any two membrane layers. For example, alternating two-phase and three-phase cycles creates a nanolaminate structure with alternating metal and metal nitride layers. In another example, the metal source chemistries are alternated in each iteration of a three phase cycle to create alternating layers of metal nitride.

In the illustrated metal nitride embodiments (Examples 5-7), the first reactant comprises a metal halide (eg, WF ₆ or TiCl ₄ ) that supplies the metal to the growth layer; It contains triethylboron (TEB); and the second reactant contains ammonia (NH ₃ ) which provides nitrogen to the growth layer.

The examples presented below demonstrate the advantages of using a halogen-getter for thin film deposition. Examples 1, 2 and 4 illustrate where corrosion of the copper metal surface was found, and other examples illustrate where corrosion was eliminated according to the preferred embodiment. The degree of corrosion was not quantified. Corrosion, either present or absent, was measured by optical and SEM images. In practice, the resistance to corrosion depends on the application.

Source Material In general, the source material (eg, metal source material, halogen getter and nitrogen source material) is preferably at a sufficient vapor pressure, substrate temperature, of the compound for performing ALD deposition. It is selected to provide sufficient thermal stability and sufficient reactivity. "Sufficient vapor pressure" supplies sufficient source chemical molecules in the gas phase to the surface of the substrate to allow self-saturation reaction at the surface at the desired rate. “Sufficient thermal stability” means that the source chemical itself has growth-condensation (growth-distu
rbing condensable) phase, or does not leave harmful levels of impurities on the surface of the substrate due to thermal decomposition. One purpose is to avoid uncontrolled condensation of molecules on the substrate. "Sufficient reactivity" causes self-saturation in pulses short enough to allow commercially acceptable throughput times. Further selection criteria include availability of high purity chemicals and ease of chemical handling.

The thin film transition metal nitride layer is preferably from a metal source material, and more preferably 3, 4, 5, 6, 7, 8, 9, 10, 11 and / or 1 of the Periodic Table of Elements.
Prepared from volatile or gaseous compounds of Group 2 transition metals. The elemental metal thin film layer is also preferably made from these compounds or Cu, Ru, Pt, Pd, Ag.
, Au and / or Ir. More preferably, the metal and metal nitride source materials include transition metal halides.

1. The halide source material first reactant preferably comprises a species that is corrosive to the surface of the workpiece exposed during deposition, especially when combined with the second reactant. In the illustrated embodiment, the corrosive species of the first reactant is advantageous in that it provides a volatile source gas for delivering the desired deposition species. Moreover, the corrosive species promotes self-limited deposition by forming part of the ligand that inhibits further growth during the first pulse.

In particular, the first reactant of the preferred embodiment comprises a halide, and more preferably a metal halide. As mentioned above, metal halides are volatile,
And thus excellent vehicles for the delivery of metal to the workpiece.
Is. Moreover, the halogen tail terminates the surface of the chemisorbed monolayer and prevents further reaction. The surface is thus self-saturated, promoting uniform film growth.

In the illustrated embodiment (see Examples 3 and 5-7 below),
Each of the halide source materials comprises a metal halide that tends to induce etching or corrosion during the ALD reaction. For example, Examples 1, 2 and 4
Exposure to ALD process containing TiCl ₄ or WF ₆ pulse, respectively
Shows corrosion of copper from.

However, as shown by Example 8, TaF ₅ does not etch copper during tantalum nitride deposition. Thermodynamic calculations (see Figure 5) support the experimental results, and also show that hafnium bromide and niobium fluoride do not corrode copper when depositing metal nitrides (see Figure 11). That). Low valent metal halides can be expected to have fewer halogen atoms to donate and corrode less sensitive surfaces than high valent metal halides. The metal halide source chemical can be transferred onto the reducing agent prior to the substrate space to reduce the valence or oxidation state of the metal in the metal halide, thus reducing the halide content of the metal halide. And it reduces the potential for corrosion of the substrate surface. The method of using a solid or liquid reducing agent before the substrate space is described in our pending Finnish patent application FI 19992235. Therefore, metal sources such as TaF ₅ , hafnium bromide and niobium fluoride are not considered corrosive in the ALD process in question. Therefore, such metal source materials can be used without the gettering methods disclosed below.

The gettering method is carried out by using a transition metal halide, particularly IV of the periodic table of elements.
Group (Ti, Zr and Hf), Group V (V, Nb and Ta) and Group VI (Cr
, Mo and W) have been used successfully with halides of elements selected from The name of the family follows the system recommended by IUPAC. Transition metal fluorides, chlorides, bromides and iodides may be used depending on the particular metal. Some metal-halogen compounds, such as ZrF _4, are not sufficiently volatile for the ALD process.

2. Gettering or Scavenging Agent 2.1 Boron Compound In the examples, the gettering agent triethylboron (TEB) was used to protect the copper surface from corrosion. Among the possible reaction products, the following are advantageous due to gettering effects: Halogen (for example from metal halides, hydrogen halides or ammonium halides) and the central boron of the TEB molecule. Boron halide formed by reaction with an atom; Ethyl halide formed by reaction of a halogen (eg, from a metal halide, hydrogen halide or ammonium halide) with an ethyl group of a TEB molecule; or hydrogen Ethane, formed by the reaction of (eg, from a hydrogen halide molecule) with the ethyl group of a TEB molecule.

It will be appreciated by those skilled in the art that the gettering effect shown here is not limited to TEB. One class of boron compounds is borane (B _x H _y ).

Volatile boron compounds having at least one boron-carbon bond are more preferred for certain metals, and more preferred are hydrocarbon groups attached to boron.
Very long or bulky groups bonded to boron can shield the central atom of the molecule, so that the preferred reaction is very time consuming or the substrate temperature is very high. Unacceptable process conditions such as Therefore, the getter compound is preferably selected from volatile boron compounds having at least one boron-carbon bond.

2.2 Silicon Compounds Silicon compounds, for example having a silicon-bonded alkyl group, are represented by reaction formula R1
And as shown in R2 can be used to getter halogens or hydrogen halides. Each reaction with a hydrogen halide molecule is believed to consume one silicon-carbon bond. Therefore, the getter compound is at least 1
It may be selected from volatile silicon compounds having one silicon-carbon bond.

[0083]

[Equation 1]

2.3 Germanium and Tin Compounds Germanium compounds having an alkyl group attached to germanium, as well as alkyltin compounds are within the possible range if it is necessary to getter halogen or hydrogen halide. Thus, the getter compound may be selected from volatile germanium and tin compounds having at least one metal-carbon bond.

2.4 Aluminum, gallium and indium compounds In the case of alkylaluminium, gallium or indium compounds, the reaction shows some detrimental complexity. As an example, trimethylaluminum (TMA) decomposes in the presence of metal halides, leaving carbon on the surface. The use of these compounds for gettering halogens or hydrogen halides is described in A
Requires careful setup of LD process parameters. However, in a less preferred arrangement, the getter compound may be selected from volatile aluminum, gallium or indium compounds having at least one metal-carbon bond.

2.5 carbon compound In the case of a carbon compound, when a double- or triple-bonded carbon is present in the molecule,
It is possible to find binding sites for hydrogen halides (R3 and R4)
. It is difficult to calculate the thermodynamic favorability of a reaction,
This is because surface chemistry differs from gas phase chemistry, eg due to absorption and desorption energies. For getter compounds selected from volatile carbon compounds, the compounds are preferably at least 1 carbon atom between carbon atoms.
It has two double or triple bonds.

[0087]

[Equation 2]

[0088]   2.6Nitrogen compounds   For nitrogen compounds, the problem is that nitrogen halides are usually thermally unstable.
is there. Alkyl-nitrogen and hydrogen halide combination to form any nitrogen halide
The reaction between things is probably unfavorable. However, chloride from alkylamine
Alkyl formation is theoretically possible (R5). Free Gibbs energy (ΔG _f ) Was calculated. Kinetic factors that affect the reaction rate are
, Not understood. Getter compounds selected from volatile amines are preferred
Are amine- and halogen-bearing species (eg,
, Hydrogen halide or ammonium halide or free halogen)
It has a negative or near-zero free Gibbs energy for the reaction.

[0089]

[Equation 3]

Some amines are stronger bases than ammonia (NH ₃ ). Such amines are salt-like with acidic hydrogen halide molecules without destroying them.
) Compounds may be formed. The bond enhances the removal of hydrogen halide from the copper metal surface before corrosion occurs. The getter compound selected from volatile amines preferably forms a sufficiently stable salt with hydrogen halides or a reaction between volatile amines and hydrogen halides leading to the formation of volatile amine-hydrogen chloride salts. Has a free Gibbs energy of negative or near zero value.

2.7 Phosphorus Compounds Phosphorus halides are very stable, and it is possible to use organophosphorus compounds for gettering halogens or hydrogen halides. The formation of metal phosphide is a competitive reaction, and depending on the application, phosphorus compounds are unacceptable. The getter compound selected from phosphorus compounds preferably has at least one phosphorus-carbon bond.

2.8 Zinc Compounds Alkyl zinc compounds are commercially available. Currently, zinc is not compatible with the state of the art process flow for integrated circuits. The getter compound may be selected from zinc compounds having at least one zinc-carbon bond in an environment in which zinc exposure is allowed.

2.9 Iron and Lead Compounds Organic-iron and organic-lead compounds form volatile metal halides. The getter compound may be selected from iron or lead compounds having at least one metal-carbon bond.

2.10 Metallocene Compounds The getter compounds are selected from volatile metallocenes (eg ferrocene, dicyclopentadienyl iron) or volatile derivatives of metallocenes (eg 1,1′-di (trimethylsilyl) ferrocene). The metal may form a volatile metal halide.

2.11 Boron-Silicon Compounds Getter compounds may also be selected from volatile boron-silicon compounds having at least one boron-silicon bond, such as tris (trimethylsilyl) borane. Both silicon and boron can form volatile halides.

2.12 Metal Carbonyl Compounds The getter compound may be selected from volatile metal carbonyls or volatile derivatives of metal carbonyls such as cyclohexadiene iron tricarbonyl, where such metals are volatile metal halogens. A compound can be formed.

2.13 General Reaction Scheme for Organic Gettering Agents A general reaction scheme for halogen gettering with volatile E (—CL ₃ ) _m G _n compounds is shown in R6. E is an element in the periodic table; L is a molecule attached to carbon C; X is a halogen; G is an unspecified molecule or atom attached to E; and m and n Is an integer, where the sum of m and n depends on the valence of E. There is a chemical bond between E and C.

[0098]

[Equation 4]

A general scheme for gettering hydrogen halides with volatile E (-CL ₃ ) _m G _n compounds is shown in R7. There is a chemical bond between E and C. E is an element in the periodic table; L is a molecule attached to carbon C; X is a halogen; G is an unspecified molecule or atom attached to E; and m
And n are integers, where the sum of m and n depends on the valence of E. The reaction formula is simplified. In fact, there is an additional reaction between the surface and the chemisorbing E compound.

[0100]

[Equation 5]

The getter compound E (-CL ₃ ) _m G _n may bind halogens or hydrogen halides or dissociate the hydrogen halides or ammonium halides (diss).
ociate) to form non-corrosive volatile halogen compounds.

2.14 Silane, Borane and Germanium Compounds R8 to R10 for silane (Si _x H _y ) and borane (B _m H _n ) where x, y, m and n are positive integers. Shows a thermodynamically favorable reaction capable of binding hydrogen halide to a less corrosive compound.

[0103]

[Equation 6]

Ammonium halides react with silane and borane (R11-R14).
), But they can also interfere with the growth of transition metal nitrides by forming silicon nitride or boron (R15-R18). The reactivity of ammonium halides is based on the well-known fact that when heated they begin to dissociate into ammonia (NH ₃ ) and hydrogen halides.

[0105]

[Equation 7]

When ammonium halide molecules (NH ₄ F, NH ₄ Cl, NH ₄ Br, NH ₄ I) are present on the reaction chamber surface, to the extent possible to prevent the formation of non-volatile silicon nitride or boron nitride. It is advantageous to use less silane or borane. When molecular hydrogen halides (HF, HCl, HBr, HI) are present on the reaction chamber surface, the dose of silane or borane is such that the acidic hydrogen halide forms silicon halide or boron halide. However, it is adjusted so that there are substantially no extra silane or borane molecules that can bind onto the metal or metal nitride surface and interfere with the metal or metal nitride growth.

Germanes (Ge _r H _t , where r and t are positive integers) are
In particular, hydrogen halides can be used to form volatile germanium halides.

Although there are just pure silicon-hydrogen, boron-hydrogen and germanium-hydrogen compounds in the examples, those skilled in the art will recognize that there is a series of similar compounds useful as gettering agents. It will be easily found in the literature. In silane (Si _x H _y), borane (B _m H _n) and germane (Ge _r H _t), hydrogen atoms, one may be substituted by a halogen atom. For example, SiH ₄ →
SiH ₃ F → SiH ₂ F ₂ → SiHF ₃ . Mixed halogen compounds such as SiH ₂ FCl are also possible. These compounds can serve as gettering agents, as long as there is at least one hydrogen atom bonded to silicon, boron or germanium.

In principle, the getter compound may be selected from silane, borane or germane having at least one hydrogen atom bonded to silicon, boron or germanium.

3. Source Material for the Second Reactant The second reactant also generally corrosive species on the surface of the workpiece exposed during deposition, especially when combined with the second reactant. including. In the illustrated embodiment, the corrosive species of the first reactant is advantageous in that it provides a volatile source gas for delivering the desired deposition species.

In "pure" metal deposition by ALD, the second reactant is replaced with another pulse of the first reactant. For example, in Example 3, both the first and second reactant pulses include WF ₆ . Therefore, only one reactant is alternated in the getter phase. Each WF ₆ pulse can potentially produce a volatile halide compound or a free excited halide species that can corrode sensitive surfaces such as copper, aluminum or silicon oxide. For example, a halide - subjecting the end metal to ammonia, tend to produce hydrofluoric acid (HF) and ammonium fluoride (NH ₄ F).

In order to form binary, ternary or higher complex materials, the following reactants preferably include hydrogen-containing compounds and, in the illustrated case of metal nitride deposition, also metal Providing nitrogen to the nitride deposition process. The second reactant used as the nitrogen source material is preferably volatile or gaseous. For example, ammonia is both volatile and highly reactive, facilitating rapid reaction with chemisorbed species from the first reactant. Preferably, the second reactant is selected from the following group: Ammonia (NH ₃ ); A salt of ammonia, preferably a halide salt, especially ammonium fluoride or ammonium chloride; ₃ ) and alkyl derivatives of the compound (eg CH ₃ N ₃
); Hydrazine (N ₂ H ₄ ) and salts of hydrazine (eg, hydrazine hydrochloride); ・ Organic derivatives of hydrazine (eg, dimethylhydrazine); ・ Nitrogen fluoride (NF ₃ ); ・ Primary, secondary Secondary and tertiary amines (eg methylamine, diethylamine and triethylamine); nitrogen radicals (eg NH ₂ ^* , NH ^** and N ^*** , where “ ^* ” is a free electron capable of forming a bond) And other excited species including nitrogen (N).

While the getter phase is particularly useful in combination with hydrogen-bearing reactants, it is used prior to other reactants such as NF ₃ and hydrogen-free nitrogen radicals listed. It remains informative in case.

Alternatively, the second reactant may provide carbon and form metal carbide. For example, after a WF ₆ pulse, TEB not only getters a halide tail,
However, it was rather found to leave some carbon in the ligand exchange reaction. Metal carbides serve as excellent barrier materials in place of or in addition to metal nitrides in nanolaminates.

4. Selection Criteria for Source Materials Metal corrosion is expected when the Gibbs energy (ΔG _f ) is negative or near zero for reactions between: metal halides and metals; hydrogen halides and metals; Or-ammonium halide and metal. Here, the metal presents a sensitive surface during the reaction, and hydrogen halide and / or
Alternatively, ammonium halides are formed as a by-product of surface reactions.

Silicon compound (eg, silicon oxide or silicon nitride) corrosion occurs when the Gibbs free energy (ΔG _f ) is negative or near zero for reactions between:
Expected at the surface: -hydrogen halides and silicon compounds; -ammonium halides and silicon compounds. Here, the silicon compound exhibits a sensitive surface during the reaction, and the hydrogen halide and / or ammonium halide are formed as by-products of the surface reaction.

If the theoretical calculations suggest that corrosion is possible, it is recommended to add a getter to the process. Getter molecules mix with corrosive molecules and prevent corrosion of sensitive surfaces.

The selection of getter compounds can be based on molecular simulations. An exemplary simulation program is available from Hypercube Inc. , Florida, U
HyperChem release 4.5, commercially available from SA. The program visualizes the physical appearance and electrostatic potential geometry of getter molecule candidates, and
For example, triethylboron) serves to assess whether it has areas accessible for reaction with corrosive molecules. Molecules that have a structure physically or electrostatically shielded from reaction with potentially harmful chemicals will give weak getters as they increase the number of reactions and the throughput of the reactor experiences. Create. Simulation of reactions between molecules and surfaces requires more complex software. Molecular Simulation Inc. (
Cerius ² commercially available from MSI), USA is an example of a program that can predict the outcome of a chemical reaction.

Chemical Reactions Several examples are provided here to further illustrate the chemical reactions in transition metal nitride thin film growth. Generally, a conformal and uniform thickness of metal nitride is desirable. ALD allows metal monolayers to be reacted with nitrogen in alternating pulses.

In the first scheme, titanium tetrachloride (TiCl ₄ ) is considered an example of a metal source material, and ammonia (NH ₃ ) is an example of a nitrogen-containing compound. The substrate is a silicon wafer with a native oxide on the surface. T
iCl ₄ reacts with OH-containing surface sites on the substrate.

[0121]

[Equation 8]

The reducing agent R is used to reduce TiCl ₃ to TiCl ₂ .

[0123]

[Equation 9]

The possible reaction mechanisms between nitrogen-containing compounds (NH ₃ in this example) are numerous and complex. For example:

[0125]

[Equation 10]

[0126]   Reactions R21-R28 refer to unreduced titanium.

The next TiCl ₄ pulse reacts with the active site as by reaction 29 or 30.

[0128]

[Equation 11]

The most preferred nitride surface sites for chemisorption of metal-containing components (particularly metal halides) are those with ═NH or —NH ₂ groups. = Surface density of the NH and -NH ₂ groups may vary according to the nitrogen source chemicals used.

If a lower resistivity is sought, titanium with three bonds is preferred,
This is because TiN has a lower resistivity than nitrogen-rich titanium nitride. In the final nitride crystal lattice, the bond states are more complex than in the simplified scheme above, because the ions occupy different types of sites, and the bonds are ionic or covalent. This is because they can have properties and possible dangling bonds near crystal defects and grain boundaries. Each pulsing cycle adds up to a molecular layer of titanium or nitrogen containing the species that form the nitride lattice. However, due to the bulky ligands or a small number of active surface sites around the adsorbed metal atoms, the growth rate can be less than 1 molecular layer per cycle.

In the second scheme, tungsten hexafluoride (WF ₆ ) is considered an example of a metal source material, and ammonia (NH ₃ ) is an example of a hydrogen-rich nitrogen compound. The substrate is a silicon wafer with a silicon dioxide (SiO ₂ ) coating. There are surface sites with —OH groups on SiO ₂ . In the metal pulse, these sites react with the WF ₆ molecule (R19). The next ammonia pulse will generate even more HF gas (R20). It should be noted that the presence of W-N bonds is a simplification. In fact, W
And N form a lattice, and they share electrons with some neighboring atoms.

[0132]

[Equation 12]

Due to the high HF production, corrosive side reactions can occur at the surface (R21).
All reaction products are very volatile and they leave a substrate. As a result, the SiO ₂ is etched. As a generalization, incompatibility problems are possible when metal fluorides and hydrogen-rich nitrogen compounds come into contact with silicon oxide.

[0134]

[Equation 13]

In the third scheme, there is a copper metal coating on the surface of the substrate. Titanium tetrachloride (TiCl ₄ ) is considered an example of a metal chloride source chemical, and ammonia (NH ₃ ) is an example of a hydrogen-rich nitrogen compound.

[0136]

[Equation 14]

[0137]   As for Example 1, corrosion of the copper surface is observed, as discussed below.

[0138]

[Equation 15]

EXAMPLES In carrying out the preferred embodiment, the conditions in the reaction space are preferably arranged to minimize gas phase reactions that can lead to the formation of condensed material. The reaction between the species chemisorbed on the surface and the gaseous reactant is self-saturating. The reaction between the by-products and the gaseous getter forms volatile compounds.

Deposition can be performed under a wide range of pressure conditions, but it is preferred to operate the method at reduced pressure. The pressure in the reactor is preferably about 0.01 mbar to 50 mbar.
r, more preferably about 0.1 mbar to 10 mbar.

The substrate temperature is kept low enough to keep the bonds between the subsurface thin film atoms intact and to prevent thermal decomposition of the gaseous source chemicals. On the other hand, the temperature of the substrate is: to provide an activation energy barrier for surface reactions
It is kept high enough to prevent phsisorption of the source material and minimize condensation of gaseous reactants in the reaction space. Depending on the reactants, the temperature of the substrate is typically 100 ° C to 700 ° C, preferably about 250 ° C to 400 ° C.

The source temperature is preferably set below the substrate temperature. This is when the partial pressure of the source chemical vapor exceeds the condensation limit at the substrate temperature,
It is based on the fact that the controlled layer-by-layer growth of thin films is compromised.

Since the growth reaction is based on a self-saturating surface reaction, it is not necessary to set tight boundaries for pulse and purge times. The amount of time available for the pulsing cycle is limited primarily by economic factors such as the desired throughput of product from the reactor. Very thin membrane layers can be formed with relatively few pulsing cycles, and in some cases this allows the use of low vapor pressure source materials with relatively long pulse times.

Example 1: Deposition of TiN from TiCl ₄ and NH ₃ PVD copper coated 200-mm silicon wafers are commercially available from ASM Microchemistry Oy of Espoo, Finland.
Loaded into a Pulsar 2000 ^™ ALD reactor. The substrate was heated to 400 ° C in a flowing nitrogen atmosphere. The reactor pressure was adjusted to about 5 mbar by means of a mass flow controller in the nitrogen line and a vacuum pump. Next, a TiN _x layer was grown by ALD from successive pulses of TiCl ₄ and NH ₃ separated by inert nitrogen gas.

One deposition cycle consisted of the following steps: TiCl ₄ pulse, 0.05 s N ₂ purge, 1.0 s NH ₃ pulse, 0.75 s N ₂ purge, 1.0 s.

This cycle was repeated 300 times to form an about 5-nm TiN _x film. The growth rate of the TiN _x film was about 0.17Å / cycle. The wafer was then unloaded from the reactor for analysis. Four-point probe and Energy Dispersive Spectroscopy (
A resistivity of 150 μΩcm was obtained by EDS) measurement.

[0147]

[Equation 16]

Formula R37 is a simplified presentation of the reaction. On the surface, the reaction sites to attract TiCl ₄ molecules (e.g., -NH and = NH) is assumed to exist. T
After the iCl ₄ pulse, there are probably —TiCl ₃ and ═TiCl ₂ groups on the surface, which can react with the NH ₃ molecule of the next pulse.

The theoretical result of formula R37 is a TiN _x film of uniform thickness over the copper surface.
However, FIG. 2 shows that there was pitting corrosion on the copper film.
Corrosion is initiated when the HCl formed as a by-product in nitride growth (R37) reacts with copper. It is also possible that HCl easily reacts with excess NH ₃ to form ammonium chloride (NH ₄ Cl), with NH ₄ Cl acting as a vapor phase carrier for copper chloride.

Example 2: Deposition of WN from WF ₆ and NH ₃ 200-mm silicon wafers coated with PVD copper were placed in Pulsa.
r Loaded into 2000 ALD reactor. The substrate was heated to 400 ° C in a flowing nitrogen atmosphere. The reactor pressure was adjusted to about 5 mbar by means of a mass flow controller in the nitrogen line and a vacuum pump. next,
The WN _x layer was grown by ALD from successive pulses of WF ₆ and NH ₃ separated by inert nitrogen gas.

One deposition cycle consisted of the following steps: WF ₆ pulse, 0.25 s N ₂ purge, 1.0 s NH ₃ pulse, 0.75 s N ₂ purge, 1.0 s.

This cycle was repeated 70 times to form an approximately 5-nm WN _x film. The growth rate of the WN _x film was about 0.6Å / cycle. The wafer was then removed from the reactor for analysis.

Etch damage to the copper film was visible even under the light microscope due to the nitride process. A large amount of HF was derived from the process (R38). HF can attack the copper surface (R39). Copper corrosion was not expected because the vapor pressure of copper fluoride was low at the substrate temperature. However, HF is also readily react with NH ₃ extra during ammonia pulse to form ammonium fluoride. Therefore, NH ₄ F can act as a vapor phase carrier for CuF and cause corrosion.

[0154]

[Equation 17]

Example 3 WC _x Deposition Using Gettering Compounds 200-mm silicon wafers coated with PVD copper were placed in Pulsa.
r 2000 ^™ ALD reactor. The substrate was heated to about 400 ° C in a flowing nitrogen atmosphere. The reactor pressure was adjusted to about 5 mbar by means of a mass flow controller in the nitrogen line and a vacuum pump. Thin films rich in tungsten metal were grown by ALD from successive pulses of WF ₆ and triethylboron (TEB) separated by inert nitrogen gas.

One deposition cycle consisted of the following steps: WF ₆ pulse, 0.25 s N ₂ purge, 1.0 s TEB pulse, 0.05 s N ₂ purge, 1.0 s.

[0157]   This cycle was repeated 70 times to give about 5-nm W-rich carbonized tongue.
A stainless film was formed. The growth rate of the thin film was about 0.6Å / cycle. Next
At, the wafer was removed from the reactor for analysis. Copper corrosion is scanning type
It was not observed by electron microscopy (hereinafter referred to as SEM). WF ₆ The exact reaction mechanism between the monolayer left by the pulse and the TEB pulse is known.
It is not. TEB acts as a halogen getter, and boron fluoride and fluorine
It appears to form ethyl iodide gas, leaving some carbon in the film.

Example 4 Deposition of W / TiN on Copper Metal A 200-mm silicon wafer coated with PVD copper was placed on a Pulsa.
Load into r 2000 ^™ ALD reactor. The substrate is heated to 350 ° C in a flowing nitrogen atmosphere. The pressure of the reactor is adjusted to about 5 mbar to 10 mbar by means of a mass flow controller in the nitrogen line and a vacuum pump. Thin films rich in tungsten metal are grown by ALD from successive pulses of WF ₆ and nido-pentaborane (B ₅ H ₉ ) separated by inert nitrogen gas.

One deposition cycle consists of the following steps: WF ₆ pulse, 1.0 s N ₂ purge, 1.0 s B ₅ H ₉ pulse, 3.0 s N ₂ purge, 1.0 s.

This deposition cycle is repeated a sufficient number of times to form an approximately 5-nm W-rich film. Then the TiN _x layer is separated by TiCl ₄ and N separated by an inert nitrogen gas.
Grow by ALD from successive pulses of H ₃ .

[0161] One deposition cycle consisted of the following steps: · TiCl ₄ pulses, 0.05 s · N ₂ purge, 1.0 s · NH ₃ pulse, 0.75 s · N ₂ purge, 1.0 s.

This deposition cycle was repeated 200 times, approximately 5-n over the tungsten film.
m TiN _x film is formed. Finally, the wafer is removed from the reactor for analysis. Copper corrosion is observed by optical microscopy. Therefore, W of 5 nm is A
Not sufficient to protect the copper surface from corrosive reactions during TiN _x deposition by LD.

Example 5: Deposition of WN with gettering compounds on copper metal . 200-mm silicon wafers coated with PVD copper were pulsed onto Pulsa.
r 2000 ^™ ALD reactor. The substrate was heated to 400 ° C in a flowing nitrogen atmosphere. The reactor pressure was adjusted to about 5 mbar by means of a mass flow controller in the nitrogen line and a vacuum pump. Tungsten nitride thin films were grown by ALD from successive pulses of WF ₆ , TEB and NH ₃ separated by inert nitrogen gas.

One deposition cycle consisted of the following steps: WF ₆ pulse, 0.25 s N ₂ purge, 1.0 s TEB pulse, 0.05 s N ₂ purge, 1.0 s NH ₃ purge, 0.75s · N ₂ purge, 1.0s.

This cycle was repeated 70 times to form an approximately 5-nm W-rich film. The growth rate of the thin film was about 0.6Å / cycle. The wafer was then removed from the reactor for analysis.

No copper corrosion was observed by SEM. The exact reaction mechanism between WF ₆ and TEB is unknown. It appears that TEB forms boron fluoride and ethyl fluoride gas, leaving a negligible residue on the surface.

Example 6: Deposition of WN / TiN nanolaminates with gettering compounds Two different types of 200-mm wafers were used for this experiment.
One wafer had a PVD copper coating, while the other wafer had an electrochemically deposited (ECD) copper film. Pulsar 2000 ^™ AL for each copper-coated wafer
Loaded into D reactor. The substrate was heated to 400 ° C in a flowing nitrogen atmosphere. The reactor pressure was adjusted to about 5 mbar by means of a mass flow controller in the nitrogen line and a vacuum pump.

First, a WN _x layer was grown by ALD from successive pulses of WF ₆ , triethylboron (TEB) and NH ₃ separated by an inert nitrogen gas pulse.

One deposition cycle consisted of the following steps: WF ₆ pulse, 0.25 s N ₂ purge, 1.0 s TEB pulse, 0.05 s N ₂ purge, 0.3 s NH ₃ pulse, 0.75s · N ₂ purge, 1.0s.

TEB functions as a getter compound capable of removing halogen from the surface. The deposition cycle was repeated 70 times to form an approximately 5-nm WN _x layer. The growth rate of WN _x was about 0.6Å / cycle.

A TiN _x layer was then grown by ALD on the WN _x layer from successive pulses of TiCl ₄ and NH ₃ separated by an inert nitrogen gas pulse. One deposition cycle consisted of the following steps: · TiCl ₄ pulses, 0.05 s · N ₂ purge, 1.0 s · NH ₃ pulse, 0.75 s · N ₂ purge, 1.0 s.

This cycle was repeated 300 times to form an approximately 5-nm TiN _x film over the WN _x film. The growth rate of the TiN _x film was about 0.17Å / cycle. The wafer was then removed from the reactor for analysis. A four point probe and energy dispersive spectroscopy (EDS) measurements gave a resistivity of about 140 μΩcm.

The same deposition program was used for both types of copper coated silicon. Figures 3 and 4 show that there was no copper pitting or corrosion during deposition. Therefore, 5 nm WN _x was sufficient to protect the underlying PVD or ECD copper from corrosion during TiN _x deposition.

Example 7: Deposition of TiN using gettering compounds on copper metal . 200-mm silicon wafers coated with PVD copper were pulsed onto Pulsa.
r 2000 ^™ ALD reactor. The substrate was heated to 400 ° C in a flowing nitrogen atmosphere. The reactor pressure was adjusted to about 5 mbar by means of a mass flow controller in the nitrogen line and a vacuum pump. TiN layer,
ALD was grown from successive pulses of TiCl ₄ , TEB and NH ₃ separated by a pulse of inert nitrogen gas.

[0175] One deposition cycle consisted of the following steps: · TiCl ₄ pulses, 0.05 s · N ₂ purge, 1.0 s · TEB pulse, 0.05 s · N ₂ purge, 1.0 s · NH ₃ pulse, 0.75s · N ₂ purge, 1.0s.

This cycle was repeated 300 times to form an about 5-nm TiN _x film. The growth rate of the TiN _x film was about 0.17Å / cycle. The wafer was removed from the reactor for inspection. Using an optical microscope at × 40 magnification,
It was revealed that there were no signs of corrosion. The exact nature of the surface reaction involved is unknown. Without wishing to be bound by theory, the TiCl ₄ molecule is
Preferably believed to bind to a = NH and -NH ₂ groups on the surface. Some feasible reactions to release HCl are shown in (R28) and (R29). It seems that TEB scavenged the released HCl.

[0177]

[Equation 18]

A further process improvement is the addition of a TEB pulse followed by an NH ₃ pulse to absorb the remaining HCl molecules that may form in the surface reaction. Some possible reactions that liberate more HCl are shown in formulas R42 and R43.

[0179]

[Formula 19]

Example 8: Tantalum Nitride Deposition A 50 mm x 50 mm piece of copper coated silicon wafer is commercially available from ASM Microchemistry, Oy, Espoo, Finland, F
Loaded into -120 ^™ ALD reactor. The substrate is placed in a flowing nitrogen atmosphere at 400
Heated to ° C. The pressure of the reactor was adjusted by a nitrogen mass flow controller and a vacuum pump to about 5 mbar. The tantalum nitride layer was grown by ALD from successive pulses of TaF ₅ and NH ₃ separated by inert nitrogen gas.

[0181] One deposition cycle consisted of the following steps: · TaF ₅ pulse, 0.2 s · N ₂ purge, 1.0 s · NH ₃ pulse, 1.0 s · N ₂ purge, 2.0s.

This cycle was repeated 2000 times to form an approximately 16-nm Ta _x N _y film. The growth rate of the film was about 0.08Å / cycle. The wafer was removed from the reactor for inspection. Neither light microscopy nor SEM showed any signs of copper corrosion.

Example 9: Depositing a nanolaminate structured silicon substrate on ASM Microchemis, Espoo, Finland.
Loaded into F-200 ^™ ALD reactor, commercially available from try. The reactor pressure was adjusted to 5 mbar absolute by a vacuum pump and flowing nitrogen. The substrate was heated to 360 ° C. First, a tantalum nitride film was grown on the substrate by repeating the pulsing sequence. Inert nitrogen gas is
Titanium tetrachloride vapor was carried into the reaction chamber. Excess TiCl ₄ and reaction byproducts were purged off with N ₂ gas. After purging, N ₂ gas carried ammonia vapor to the reaction chamber. Extra NH ₃ and reaction by-products, was purged removed by N ₂ gas: · TiCl ₄ pulses, 0.05 s · N ₂ purge 1.0 s · NH ₃ pulse, 0.75 s · N ₂ purge, 1. 0s.

A tungsten nitride thin film was grown on top of the titanium nitride film by repeating another pulsing sequence: • WF ₆ pulse, 0.25 s · N ₂ purge, 1.0 s · TEB pulse, 0. .05s · N ₂ purge, 0.3 s · NH ₃ pulse, 0.75 s · N ₂ purge, 1.0 s.

Processing was continued by depositing alternating thin film layers of titanium and tungsten nitride. A total of 6-18 nitride thin film layers were deposited, depending on the sample. The total thickness of the ALD nanolaminate was about 70 nm. The film appeared as a dark, light reflecting mirror. The color was slightly reddish, unlike titanium or tungsten nitride. Membranes were analyzed by transmission electron microscopy (TEM), energy dispersive spectroscopy (EDS) and 4-point probe measurements. The TEM picture (FIG. 10) showed a clear nanolaminate structure with separate titanium and tungsten nitride thin film layers. According to EDS, titanium, tungsten and nitrogen molecules were present in the film. The amount of impurities was estimated to be less than 1 at .-%. The membrane was electrically conductive. Resistivity was calculated by combining thickness (EDS) and 4-point probe results. The resistivity of the non-optimized sample was about 400 μΩ-cm.

[0186] Example 10: deposition nanolaminates metal / metal nitride Nanoramine over bets using two transition metal source, using the above-described ALD process, alternating thin layers of metals and metal nitrides Was manufactured using. Two different transition metal sources were used for the thin film layers.

[0187]   Thin film layer 4: tantalum nitride   Thin film layer 3: Tungsten metal   Thin film layer 2: tantalum nitride   Thin film layer 1: Tungsten metal   Substrate.

An odd number of thin film layers (1, 3, 5, etc.) were deposited from tungsten source chemicals and reducing source chemicals. Even number of thin film layers (2, 4, 6, etc ...)
Was deposited from tantalum source chemistry, any reducing source chemistry and nitrogen source chemistry. All source chemical pulses were separated from each other using an inert purge gas.

Example 11: Deposition of metal / metal nitride nanolaminates using one transition metal Nanolaminates were made with alternating thin film layers of metal nitride and metal.
One transition metal source was used for the thin film layers.

[0190]   Thin film layer 4: Tungsten metal   Thin film layer 3: Tungsten nitride   Thin film layer 2: Tungsten metal   Thin film layer 1: tungsten nitride   Substrate.

An odd number of thin film layers (1, 3, 5, etc ...) Was deposited from a tungsten source chemistry, any reduction source chemistry and a nitrogen source chemistry. An even number of thin film layers (2, 4, 6, etc.) were deposited from the tungsten source and reducing chemistries. All source chemical pulses were separated from each other using an inert purge gas.

Although the above invention has been described by means of certain preferred embodiments, other embodiments will be apparent to those skilled in the art in view of the disclosure herein. Therefore, the present invention is not intended to be limited by the description of the preferred embodiments, but rather is defined by reference to the appended claims.

[Brief description of drawings]

These and other aspects of the invention will be readily apparent to those of ordinary skill in the art in light of the above description and accompanying drawings, which are meant to illustrate and not limit the invention.

FIG. 1 is a scanning electron micrograph (SEM) taken from a copper film formed by physical vapor deposition (PVD). Measuring voltage is 10
It was kV.

FIG. 2 is an SEM taken from TiN covered PVD copper according to the ALD process without the use of getter or scavenger pulses. The black areas in the photo show the areas of copper that were etched during the TiN processing.

FIG. 3 is an SEM taken from PVD copper covered first with WN and then with TiN according to a preferred embodiment of the present invention (Example 6).

FIG. 4 is an electrochemically depo covered first with WN and then with TiN according to a preferred embodiment of the present invention (Example 6).
sited) (ECD) SEM imaged from copper.

FIG. 5 is a graph depicting the equilibrium state of compounds between tantalum, fluorine and copper as a function of temperature. Source chemical for calculation is 10 molar TaF ₅
And 1 mol Cu.

FIG. 6 illustrates an exemplary dual damascene structure in a partially fabricated integrated circuit having exposed copper and insulating oxide surfaces over which metal and metal compound deposition is desired. It is sectional drawing of a workpiece.

7 shows the workpiece of FIG. 6 after lining the dual damascene trench and contact vias with a conformal thin film according to a preferred embodiment.

FIG. 8 is a flow chart generally illustrating a method for forming binary compounds by atomic layer deposition (ALD) according to some of the preferred embodiments.

FIG. 9 is a schematic diagram of the first four thin film layers of the ALD nitride nanolaminate and the pulsing sequence for each thin film layer.

FIG. 10 is a transmission electron microscope (TEM) photograph of a nitride nanolaminate structure.

FIG. 11 shows a reaction (X) or a reaction (○) with refractory (refrac) of copper metal.
tive) is a table showing metal halides. (X) means lack of sufficient data to draw conclusions regarding the ability to use these reactants without gettering.
Metals are in their highest possible oxidation state. The Gibbs free energy of reaction can be calculated using a computer program (HSC Chemistry, version
3.02, Outokumpu Research Oy, Pori, Finra
nd).

─────────────────────────────────────────────────── ─── Continued front page    (31) Priority claim number 19992235 (32) Priority date October 15, 1999 (October 15, 1999) (33) Priority claim country Finland (FI) (31) Priority claim number 60 / 175,799 (32) Priority date October 15, 1999 (October 15, 1999) (33) Priority claiming countries United States (US) (31) Priority claim number 60 / 176,948 (32) Priority date January 18, 2000 (January 18, 2000) (33) Priority claiming countries United States (US) (31) Priority claim number 20000564 (32) Priority date March 10, 2000 (March 10, 2000) (33) Priority claim country Finland (FI) (81) Designated countries EP (AT, BE, CH, CY, DE, DK, ES, FI, FR, GB, GR, IE, I T, LU, MC, NL, PT, SE), OA (BF, BJ , CF, CG, CI, CM, GA, GN, GW, ML, MR, NE, SN, TD, TG), AP (GH, GM, K E, LS, MW, MZ, SD, SL, SZ, TZ, UG , ZW), EA (AM, AZ, BY, KG, KZ, MD, RU, TJ, TM), AE, AG, AL, AM, AT, AU, AZ, BA, BB, BG, BR, BY, BZ, C A, CH, CN, CR, CU, CZ, DE, DK, DM , DZ, EE, ES, FI, GB, GD, GE, GH, GM, HR, HU, ID, IL, IN, IS, JP, K E, KG, KP, KR, KZ, LC, LK, LR, LS , LT, LU, LV, MA, MD, MG, MK, MN, MW, MX, MZ, NO, NZ, PL, PT, RO, R U, SD, SE, SG, SI, SK, SL, TJ, TM , TR, TT, TZ, UA, UG, US, UZ, VN, YU, ZA, ZW (72) Inventor Sanilla Bille Antero             Finland FI-00570             Helsinki Tsu Pubori 6 Sae 27 (72) Inventor Kaipiosari Johanna             Finland FI-04400             Yarvenpa Satakunnankatsu 13 Ar 6 (72) Inventor Soininen Pekka Juha             Finnish country IFN-02200             Espoo Hultyaton Tsunche 21 a             -1 F-term (reference) 4K030 AA03 AA04 AA05 AA13 AA24                       BA38 BB13 CA04 EA03 FA10                       HA01 HA15 LA15                 5F045 AA00 AB40 AC08 AC09 AC12                       AC19

Claims (36)

[Claims] 1. A method for forming a conductive nanolaminate structure on a substrate in a reaction space by an atomic layer deposition (ALD) type process, the nanolaminate structure comprising at least one metal compound layer. A method comprising at least two adjacent thin film layers, wherein each thin film layer is in a different phase than the adjacent thin film layers. 2. The method of claim 1, wherein the nanolaminate structure comprises at least 3 layers. 3. The method of claim 2, wherein each of said layers has a different composition than an adjacent layer. 4. The method of claim 1, wherein the metal composite layer comprises metal carbide. 5. The method of claim 1, wherein the metal composite layer comprises metal nitride. 6. The method of claim 1, wherein at least one layer comprises elemental metal. 7. The method of claim 1, wherein the nanolaminate structure is a diffusion barrier in integrated circuits. 8. The method of claim 1, wherein the nanolaminate structure is formed on a substrate susceptible to halide attack. 9. The ALD-type process comprises providing alternating pulses of reactants in multiple deposition cycles, each cycle comprising: supplying a first reactant to a halide termination chemistry over a surface. Chemisorption of less than about 1 monolayer of species; removing excess first reactant from the reaction space; and gettering halide from the monolayer before repeating the cycle, 9. The method of claim 8, comprising: 10. The method of claim 9, wherein each cycle further comprises feeding a hydrogen-bearing second reactant. 11. The hydrogen-bearing second reactant comprises a source of nitrogen.
The method described in 0. 12. The method of claim 11, wherein the source of nitrogen comprises ammonia. 13. The method of claim 11, wherein the first reactant comprises a metal halide. 14. The method of claim 9, wherein the surface comprises copper. 15. The method of claim 14, wherein the surface further comprises a form of silicon oxide. 16. The method of claim 9, wherein the surface is formed of a material less than 5 nm thick on copper. 17. The method of claim 9, wherein the surface comprises a metal. 18. The method of claim 9, wherein gettering comprises reduction. 19. The method of claim 18, wherein gettering comprises exposing the halide terminating species to a boron compound. 20. The boron compound comprises triethylboron (TEB).
The method according to claim 18. 21. The method of claim 1, wherein the nanolaminate is a conductive diffusion barrier. 22. A nanolaminate structure comprising at least three thin film layers each having a thickness of less than about 10 nm, at least one of said layers being selected from the group consisting of metal carbides and metal nitrides. Laminated structure. 23. The nanolaminate structure of claim 22, wherein each of the thin film layers has a thickness of less than about 5 nm. 24. The nanolaminate structure of claim 23, comprising 4 to 250 alternating thin film layers. 25. The nanolaminate structure of claim 24, comprising 4-20 thin film layers. 26. The laminate structure of claim 24, wherein each layer comprises a different phase metal compound than an adjacent layer. 27. The nanolaminate structure of claim 26, wherein each layer comprises a different composition than an adjacent layer. 28. The nanolaminate structure of claim 27, wherein two different metal nitrides are alternated. 29. The nanolaminate structure of claim 26, comprising at least one metal layer. 30. The nanolaminate structure of claim 26, comprising at least one metal carbide layer. 31. A method of depositing material on a substrate in a reaction space, the substrate including a surface susceptible to halide attack, the method providing alternating pulses of reactant in a plurality of deposition cycles. Each cycle comprising: supplying a first reactant to chemisorb less than about 1 monolayer of halide-terminated species over the surface; excess first from the reaction space. Removing reactants and reaction byproducts; and gettering halide from the monolayer prior to repeating the cycle. 32. The method of claim 31, wherein the first reactant comprises a metal halide. 33. The method of claim 31, further comprising providing a second reactant to react with the chemical species after gettering the halide. 34. The method of claim 33, wherein the second reactant comprises a source of nitrogen and the material comprises a transition metal nitride. 35. The method of claim 31, further comprising providing a carbon source to react with the chemical species, and the material comprises transition metal carbon. 36. The method of claim 31, wherein the material comprises a thin film within a nanolaminate stack.