KR20050087428A - Deposition of tungsten nitride - Google Patents

Abstract

A tungsten nitride layer deposition method is disclosed. The method forms a tungsten nitride layer using a controlled deposition technique such as a pulsed nucleation layer (PNL). Initially a tungsten layer is formed on the substrate surface. The tungsten layer is then exposed to a nitriding agent to form a tungsten nitride layer. Repetitive cycles of contact with the reducing agent, tungsten precursor and nitriding agent are used to form a relatively thick layer. In one aspect the cycle comprises contacting a dopant precursor, such as phosphine or arsine.

Description

Deposition of tungsten nitride

Reference of related patent application

This application is described in Fair et al. This invention is related to US Provisional Application No. 60 / 441,843, filed Jan. 21, 2003, the entirety of which is incorporated herein by reference. This application also discloses the invention "Ultra-thin tungsten layer of improved step coverage" (by Lee et al) and US patent application Ser. No. 09 / 975,074 (US Pat. No. 6,635,965). / 649,351, filed Aug. 26, 2003, relates to a method for reducing the roughness of a tungsten thin film and improving step coverage by Lee et al, which is incorporated herein by reference for all purposes. .

The present invention relates to a pulsed nuclation layer (PNL) method for tungsten nitride deposition. In particular, the present invention relates to a method for depositing tungsten nitride in a partially assembled semiconductor device. The present invention is particularly useful for applications requiring metal or metal nitride deposition on dielectrics, metals, silicon compounds, silicon with good adhesion, good step coverage, and low processing temperatures (eg, 400 ° C. or below).

Tungsten nitride is used for various purposes in the assembly of semiconductor devices. By being deposited by traditional means such as PVD and PECVD, tungsten nitride provides a relatively low resistance, good adhesion and good diffusion barrier to dielectric thin films. In the past, an important limitation that prevented widespread application of WNs was poor step coverage at high aspect ratio trenches, vias, and contacts.

For successful nanometer scale applications, tungsten nitride must be deposited thinly and conforamally at high aspect ratio properties. Traditional PVD (physical vapor deposition) techniques cannot achieve this criterion. For thin on-angle coverage, the CVD (chemical vapor deposition) method is generally considered. Traditional CVD processes include a gas phase reactant comprising a tungsten precursor (typically tungsten hexafluoride (WF ₆ )) and a nitrogen containing gas (eg N ₂ ) near the heated wafer surface while steam is applied to the system. It is related to introducing them simultaneously. This reaction is accelerated by the energy provided from the heated wafer and the free energy change of the chemical reaction. Growth of the tungsten nitride thin film continues as long as the reactants and energy sources are supplied.

Although standard tungsten nitride CVD technology provides good step coverage and satisfies low aspect ratio characteristics (e.g. <5: 1 aspect ratio), semiconductor fabrication techniques are approaching nanometer scale, so step coverage and gap fill The demands are becoming more stringent and CVD may not be suitable for this purpose. Conventional tungsten nitride plasma-enhanced CVD has relatively poor step coverage in the CVD process. (<50% SC at columnar splice ratio 5: 1). This is not suitable for current and future semiconductor technology, where the aspect ratio exceeds 10 to 1 and the critical dimension is less than 100 nanometers. CVD and especially PNL or ALD tungsten processes may provide the very high step coverage and on-angle deposition required in modern semiconductor devices (as opposed to tungsten nitride processes) but will not adhere directly to the dielectric surface. Tungsten is now required as an adhesive layer, such as TiN, before deposition on the dielectric surface. As a result, the high deposition temperatures required by many TiN deposition techniques (eg, PECVD-TiN from TiCl ₄ ) cannot be applied to low-K dielectrics or nickel silicon compounds.

Therefore, there is a need for an improved method for tungsten nitride deposition.

The present invention provides a method for depositing a tungsten nitride layer on a substrate. Here the method provides good adhesion of tungsten nitride to the substrate, good control over the deposition thickness, and good step coverage for the substrate high aspect ratio region. To achieve this, the present invention uses at least the following work. (Performed in various orders): (iii) providing a reducing agent layer on the substrate surface (ii) contacting the substrate surface with precursor-containing tungsten to form a tungsten layer on the substrate, and (iii) nitriding the tungsten layer To form tungsten nitride.

In many cases, the substrate is a semiconductor wafer or a partially assembled semiconductor wafer. Uses of the present invention include tungsten nitride (or partially) diffusion of copper, gate electrodes, capacitor electrodes, and diffusion barriers and / or adhesive layers of tungsten plugs, double damascene copper. Sacrificial hard masks for wiring formation, including use as light shields for CCD devices. In many of these applications tungsten nitride is at least partially deposited on the dielectric material.

In a preferred embodiment, the reducing agent is a boron-containing agent, more preferably diborane (B ₂ H ₆ ). This borane reducing agent can be introduced as a gaseous reactant, which decomposes at the surface of the substrate to form a boron-containing "sacrificial layer". Preferably, the sacrificial layer is about 3 to 20 Angstroms thick and is deposited at a temperature of about 200 to 400 ° C.

In another embodiment, the reducing agent is a silane or other non-boron containing reducing agent. In this embodiment, the reducing agent is introduced to the substrate prior to introducing the tungsten containing precursor

Form an adsorbed or saturated layer. Otherwise, tungsten-containing precursors may be introduced to the substrate before the reducing agent. If a tungsten-containing precursor is introduced before the reducing agent, a thin thin film of tungsten-containing precursor is formed on the substrate and subsequently decomposed upon contact with the reducing agent to form tungsten.

In one embodiment, the tungsten-containing precursor is WF ₆ , WCl ₆ or W (CO) ₆ . Of course, other tungsten-containing precursors suitable for reducing to elemental tungsten may also be used. This includes combinations of tungsten-containing precursors. (Gas, liquid or solid). In addition, other suitable nitriding agents may also be used. Examples include N ₂ , NH ₃ , NF ₃ , N ₂ H _6, and combinations thereof.

In some embodiments of the invention, a gas purge is applied after exposure to one or more reactants (eg, reducing agent, tungsten containing precursor, and nitriding agent). In many cases, the atmospheric gas is applied after each reactant is introduced. Atmospheric gas purifies the residual gaseous reactants around the substrate surface to react with the new gaseous reactants for the next step process. According to one embodiment of the invention, the tungsten layer is treated with hydrogen or Ar—H 2 plasma prior to the nitriding step to remove halogen by-products, unreacted halogen reactants, or other unnecessary gases prior to introduction of the nitrifier.

In another aspect of the invention, tungsten nitride is provided to the tungsten nitride module in one or more deposition stations. The tungsten nitride module includes a wafer preheating device and a substrate preclean station. The preclean module provides reactive precleaning properties, enabling the use of clean chemical reactions based on fluorine produced by cleaving fluorine-containing agents using inductively coupled plasma. Moreover, the wafer precleaning device or other device in the tungsten nitride deposition module retains the feature to passivate the substrate after precleaning the substrate. Preferably, the module for tungsten nitride deposition is in vacuum connection with a module provided for pulsed nucleation of tungsten or CVD of tungsten.

In one embodiment, the method also covers the substrate by one or more of the following means. (a) hydrogen exposure; (b) exposure to distant H / H ₂ plasma; (c) exposure directly to H / H _2, Ar / H _2, or RF plasma; (d) exposure to WF ₆ ; (e) continuous or simultaneous exposure to H ₂ or H / H ₂ plasma and NH ₃ ; And (f) exposure to oxygen.

Another aspect of the invention provides a method of forming a tungsten nitride layer on a substrate. This method is characterized by the following steps; (a) placing the substrate in the deposition chamber; (b) depositing one or more pulse deposited tungsten layers on the semiconductor wafer; (c) depositing a pulse deposition layer of one or more layers of tungsten nitride on one or more layers of tungsten; And (d) optionally repeating (b)-(c) to create a multilayer structure of multiple tungsten and tungsten nitride layers or a bilayer of W-WN. The resulting composite thin film can have various layer structures. In one embodiment, the base layer of the W-WN composite membrane is a tungsten layer. In another embodiment, the base layer of the W-WN composite membrane is a WN layer. In another particular embodiment, the ratio of W and N atoms is present in the W and WN layers at a ratio of about 2 to 1. Therefore stoichiometrically, W ₂ N is formed by heat treatment directly or indirectly.

The present invention relates in certain aspects to depositing tungsten nitride on an intermetallic dielectric material to act as a copper diffusion barrier. In this use, the tungsten layer can be deposited over the tungsten nitride layer. From this, copper is deposited on tungsten nitride or W-WN barriers to form a thin film stack suitable for forming single or double damascene copper interconnects.

If a metal tungsten layer is applied as part of the WN-W barrier system, the pulse deposition layer (PNL) method, atomic layer deposition method or CVD method (e.g., WF ₆ and H ₂ or SiH ₄ or other combinations thereof) Will be deposited using).

In traditional damascene processes, the copper layer is first provided as a copper seed layer, which can be deposited using sputter deposition or electroless plating methods. After the copper seed layer is deposited, a bulk copper layer can be deposited over the seed layer using electrolytic plating methods.

Another aspect of the invention relates to the use of tungsten nitride. It can be used in combination with a tungsten nitride tungsten layer to form a gate electrode. In other applications, a tungsten nitride layer or tungsten nitride-tungsten bilayer may be used for the formation of capacitor electrodes for DRAM or other storage devices. These and other features and advantages of the present invention are described in more detail in conjunction with the accompanying drawings below.

detailed description

As mentioned above, the present invention provides a method for depositing a tungsten nitride layer. In particular, there is provided a method for applications in which tungsten nitride is deposited on a dielectric surface, where a thin, angled and adhesive film is required. Preferred methods relate to pulsed nucleation film (PNL) deposition techniques, which are described in detail below.

A preferred approach to the PNL process involves first depositing diborane (or other boron-containing precursor) on the surface of the substrate to form a "sacrifice" boron-containing layer. This sacrificial layer then reacts with the tungsten precursor to form tungsten. This diborane deposition process is not a traditional self-limiting ALD type deposition process. Rather, diborane reacts on the dielectric surface and breaks down into a thin film of boron. This reaction can proceed as long as the substrate is continuously exposed to diborane. However, in order to ensure the limiting capacity of tungsten substantially formed in the subsequent steps, diborane deposition is preferably limited to about 3 to 10 angstroms in thickness. This corresponds to about one or two monolayers of boron.

It will be noted that the terms boron layer (or thin film) and boron-containing layer encompass boron hydride and various boron compounds such as mixtures or mutual combinations of these compounds and / or elemental boron as well as pure elemental boron. The term " element boron " here also encompasses pure boron as well as a combination of certain doses of elemental boron with other materials such as many boron compounds.

In the second operation of this process, the boron layer is exposed to a tungsten precursor, which is reduced by boron to form tungsten. In a third operation, the tungsten layer is converted to tungsten nitride by contact with the nitriding agent. Preferred products are stoichiometric W ₂ N, other WN, WN ₂ and various other stoichiometric materials are also included in the present invention. The final thin film also contains various hydrides and / or amines for example. If a certain amount of tungsten is left unconverted, this may limit the adhesion to the underlying dielectric. Therefore, it is important that the amount of tungsten present in any of the circuits of the PNL is small enough that all of it is converted to nitride in the nitriding step. The capacity of the produced tungsten is limited by the amount of boron previously formed as a sacrificial layer on the underlying substrate. Therefore, the amount of diborane deposited and reacted to form boron effectively controls the amount of tungsten. This in turn ensures that all tungsten can be converted to nitride in a single operation.

After these three operations are completed, a very thin film of tungsten nitride is produced. Thereafter, these three operations are repeated until the tungsten nitride required as multiple circuits is formed. In the accompanying circuits, other reactants may be used. In particular, diborane may be replaced with silane or other reducing agents such as hydrogen and silicon.

The conditions of the process used in the various stages of the circulating process are wide. Relevant process conditions include pressure, temperature, dose, concentration, and time. The nitriding process provides the most fluid choice of these conditions. The diborane process is preferably performed at a relatively high temperature (eg, at least 250 degrees Celsius, preferably about 200 to 400 ° C.) to ensure that a sufficient amount of boron is formed on the substrate in a suitable amount of time. With respect to other parameters, the process step is generally carried out in a dose which is determined by the pressure of about 0.1 to 300 torr and the flow rate and exposure time. These parameters are described in more detail below.

The PNL process is preceded by various substrate pretreatments such as degas processes, anneal, and / or preclean (eg mild sputter etch in argon and / or hydrogen). These pretreatments have various purposes, such as absorption of water vapor and removal of surface oxididation from the electrically active region of the semiconductor device. This is explained in more detail below. The PNL process may also include various post-treatments, such as the deposition of a metallic tungsten layer on top of tungsten nitride to form a WN-W bilayer.

In a typical scenario, the PNL-WL deposition process is preceded by wafer degassing / preheating and preclean. Traditional wafer degassing is carried out in an independent high-vacuum chamber, where the wafer is heated and the gas species are ejected. Existing and well-known wafer preclean strategies include physically removing contaminants and surface oxidation with direct Ar and Ar-H ₂ sputter etch prior to deposition of the barrier.

Implementation of the PNL-WN involves preheating in situ and precleaning the reactive wafers. Wafer preheating is used to blow moisture and remove other contaminants from the wafer surface. Wafer precleaning is performed by brief exposure of the wafer to tightly controlled capacities of elemental or molecular fluorine. In some cases, fluorine is produced by decomposing NF ₃ into F and N ₂ by means of inductively coupled plasma (ICP). Other powdered source gases (eg, F ₂ , CF _4, C ₂ F ₆ , ClF ₄ , etc.) and other decomposition techniques may equally be used within the scope of the present invention. Fluorine species produced during precleaning of the wafer react with the native oxides and other residues to form volatile products, which are absorbed or released from the wafer surface.

In one embodiment, the wafer preclean device comprises an NF ₃ phone line, which is deposited directly from the mass flow controller (NFC ₃₎ until the flow is fully stabilized before it is introduced into the chamber's wafer preclean device. Allow the process to be switched to the front line of the chamber. The outlet valve of the parallel allows the controlled release of NF ₃ is to be passed from the beginning of the NF ₃ passes into the deposition chamber.

As mentioned above, all gas flows, valves, and plasma source commands proceed according to the input I / O sequence, so these commands are delivered in packets to the IOC, with a +/- 10ms command. It is performed with a sequence of time control precision. This tight fluorine dose control during wafer precleaning is suitable for precleaning inherent oxides and other contaminants from the wafer surface, but is sensitive from silicon wafer surfaces (silicon or silicon raw material-drain connections, polysilicon connections). , Or high K gate or capacitor dielectric).

If fluorine-based wafer precleaning results in a fluorine saturated semiconductor wafer surface after cleaning, various post-clean wafer processing is used to remove or adsorb fluorine and thus promote efficient and uniform PNL-WN growth involved. Post-cleaning protection strategies include:

1. Expose the wafer surface to atomic and / or molecular hydrogen. Atomic hydrogen can be prepared using ICP raw materials by the same or other means as the precleaning techniques of the above described preferred practice. Atomic or molecular hydrogen may be delivered with, for example, an Ar or N ₂ carrier gas.

2. Expose the fluorine and fluorine compounds on the surface by directly exposing the wafer surface to Ar or hydrogen ions.

3. The wafer surface is exposed to WF ₆ prior to exposure to B ₂ H ₆ or other reducing agent. This WF ₆ is substituted with surface fluorine to promote nucleation on the cleaned surface.

4. The surface is simultaneously exposed to WF ₆ and B ₂ H ₆ for the CDD-W nuclii.

5. Expose the wafer surface to NH ₃ .

6. Combine 1 and 5 above. Atomic hydrogen exposure converts surface fluorine to HF and NH ₃ will then produce NH ₄ F. Note that NH ₄ F is volatilized above about 100 ° C. Preferred performance of this process sequence includes heating the wafer to about 250 ° C. or higher during preheating of the wafer, precleaning, and H ₂ / NH ₃ passivation.

In some PNL applications, it is desirable to provide a four-working circuit. The fourth operation here is preserved for the introduction of dopants into tungsten nitride. Therefore, in addition to exposing the substrate to the reducing agent, tungsten precursor and nitriding agent, the process includes exposing the substrate to the raw material of the dopant. Examples of such dopant raw materials include phosphine and arsine. Introducing them separately is more suitable than introducing them with oxidizing or nitriding agents because they are potentially explosive or incompatible in combination with other reactants.

Doping is suitable for changing the characteristics of tungsten nitride, in particular their working function. The working function is an electrical property which affects the charge distribution of adjacent layers. Doping will therefore be suitable for use in the present invention in connection with the creation of gate electrodes or battery electrodes. For other applications such as plug filling and barrier layer formation, doping is not particularly considered. In these applications the barrier properties, conformity, thickness, conductivity and adhesion of the tungsten nitride layer are of paramount importance.

Note that if a tungsten-tungsten nitride bilayer is formed, the metallic tungsten composition gate of this bilayer may be deposited by a rotary PNL process, a bulk CVD process, or other suitable process. CVD has the advantage of producing low resistive thin films, and PNL has the advantage of making more right angles without protrusions.

It will be noted that the present invention is not limited to PNL processes using diborane or other boron-containing materials. More generally, any suitable reducing agent can be used. In some cases, the PNL working order can be changed and therefore does not necessarily begin with exposure to a reducing agent. In other cases, for example, the first PNL operation is tungsten precursor adsorption. This is accompanied by the contact of a reducing agent and the contact of a nitriding agent. Moreover, the reducing agent and nitriding agent can be provided simultaneously in a single operation. In addition, exposure to reducing agents, tungsten precursors and nitriding agents may be staggered or redundant.

There are many uses for this PNL-WN technology. Only some are described here, and many other processing equipment are also used. These include multi-device and single device deposition chambers. If a single device deposition chamber is used, different precursor gases are provided from the same chamber tubing. If multiple device chambers are used, the device carries the substrate from the deposition apparatus to the deposition apparatus, each chamber being provided for a different single reactant. Moreover, some deposition apparatus may be used for PNL reactions only, others may be used for tungsten CVD or other deposition reactions.

Process flow

The flow of a general process for the production of tungsten nitride according to the invention is shown in the flowchart of FIG. 1. Firstly the substrate surface is provided as indicated at 101. In many embodiments of the present invention, the substrate is a semiconductor wafer composed of partially assembled circuits. This may include trenches and / or vias or metal lines for wiring. Otherwise, they will include thin dielectric layers that act as gate or battery dielectrics.

On the substrate surface provided at 101, this process forms a tungsten metal layer. This can be done using either method. In the first method, the reducing agent is initially introduced to the substrate surface (block 103) so that the reducing agent (or a portion of the reducing agent) is absorbed or otherwise remains on the surface of the substrate. See FIG. 105. As above, some embodiments produce a "sacrifice layer" of reducing agent, especially when the deposited layer is a boron layer.

More generally, the reducing agent may be any process-compatible compound that can effectively reduce the tungsten precursor to produce a layer of metallic tungsten. Suitable reducing agents include various boron-containing reducing agents, preferably borane (BH ₃ ), diborane (B ₂ H ₆ ), triborane and the like. Examples of other boron containing reducing agents include boron halides (eg BF ₃ , BCl ₃ ) and hydrogen. Other reducing agents include silicon hydrides such as silanes and organic derivatives thereof.

After the reducing agent layer is created on the chamber surface, tungsten containing precursor gas is introduced to the substrate surface. See block 107. This tungsten-containing precursor is reduced upon contact with the sacrificial layer on the adsorbed substrate surface to form a tungsten metal layer. Any suitable tungsten-containing precursor may be used which can be reduced by the reducing agent of the sacrificial layer to produce a layer of tungsten metal in the present invention. Such suitable tungsten containing precursors include WF ₆ , WCl ₆ , W (C0) ₆ , and combinations thereof. WF ₆ has been found to work particularly well with the boron sacrificial layer. Various other tungsten precursors known to those skilled in the art can be used. Some of these are RG Gordon, J, Barton, and Seigi Suh's 'Materials, Technologies, and Reliability for Advanced Interconnects and Low-K Dielectrics II' (edited y S, Lahiri, Mat, Res, Soc, Proc, 714E, Pittsburgh PA, 2001). It is hereby incorporated by reference for all purposes.

In another method of forming the metallic tungsten layer, the tungsten containing precursor is introduced to the substrate surface before the reducing agent. See block 111. This tungsten-containing precursor may be adsorbed onto the substrate surface to form a tungsten-containing precursor thin film (or portion of the tungsten-containing precursor). See block 113, the diversity of this process sequence is particularly effective for the following wafer precleans based on fluorine.

Suitable tungsten-containing precursors are those which can saturate the substrate surface by adsorption and are reduced by a reducing agent to form a tungsten metal layer. In general, the aforementioned precursors (WF ₆ , WCl ₆ , W (C0) ₆ ) and combinations thereof work well.

After providing a tungsten-containing precursor to the substrate surface, a reducing agent is introduced to the wafer surface. This reduces the adsorbed tungsten containing precursor layer (block 115) to form a tungsten metal layer. Again, any suitable reducing agent can be used that can reduce the adsorbed tungsten-containing precursor layer to produce the tungsten metal layer according to the present invention. As an example, a boron-containing reducing agent, preferably borane, more preferably diborane (B ₂ H ₆ ) is included. Other examples include silanes and derivatives of silanes.

Regardless of whether the process uses operations 111. 113 and 115 or 103, 105 and 107, the resulting product is essentially the same tungsten layer. In this regard, as shown in FIG. 1, this process is concentrated in a single set of operations.

As in block 117. Optional hydrogen plasma treatment may be performed to remove excess halogen by-products and unreacted halogen reactants. This hydrogen plasma treatment chemically removes these compounds that sometimes adsorb to the surface of the substrate and / or the walls of the reactor. These may be due to wafer precleaning or tungsten precursor dose steps.

Once the tungsten metal layer is formed and optionally plasma treated, a nitrogen containing gas is introduced to the substrate surface to convert at least the top layer of tungsten metal to tungsten nitride. See block 119. Examples of suitable nitrogen containing gases include N ₂ , NH ₃ , NF ₃ , and N ₂ H ₄ . In particular, this operation completely converts the tungsten layer into tungsten nitride. If the resulting nitride layer is not thick enough for the intended use, the tungsten deposition and nitriding steps described above can be repeated. If tungsten nitride is thick enough, this process is complete. See results.

Certain changes to the procedures described above will be apparent. First, the substrate can be pretreated by wafer pre-cleaning (Ar or Ar-H ₂ ) or reactive cleaning (F, F ₂ , NF ₃ , CF ₄ , etc.) by heat cooling or sputter etch. This precleaning is used to remove inherent oxides and other contaminants from electrical connections, or to remove etch residue from connections or vias. In some cases, it is possible to promote PNL-WL adhesion to the dielectric substrate by preconditioning with light sputter etch or etch based on reactive fluorine.

In addition, the PNL circuits 111, 113, and 115 or 103, 105, and 107 may include dopant introduction operations. Dopant precursors may be introduced with one of the other three reactants or may be introduced without the other reactants. In other cases, the PNL circuit includes a separate operation for the dopant process, and the entire circuit includes four separate operations. Examples of dopants include phosphorus, arsenic, antimony, bismuth, boron, aluminum, gallium, indium, nitrogen, thallium and combinations thereof.

Finally, it is advantageous to introduce atmospheric gases between reaction contacting operations, for example between 103 and 105, and / or between 105 and 107, and / or 107 and 103. Moreover, atmospheric gases can flow continuously in any of these operations and during and between operations. Atmospheric gases are generally inert. Examples include noble gases (eg argon) and nitrogen.

Hydrogen gas may also be used as the atmosphere gas or the carrier gas. Hydrogen effectively neutralizes the residual fluorine from WF ₆ and reduces the level of fluorine or other halogens in the final PNL-WL thin film. Hydrogen may be required as a co-reactant with some of the dopant precursors described in the foregoing paragraphs.

Deposition unit work

As above, the method of the present invention uses pulse nucleation (PNL) and related deposition techniques. These techniques are used to form tungsten nitride. Hereinafter, a tungsten nitride deposition technique according to the present invention will be described.

PNL deposition is generally a method of depositing a plurality of atomic-scale layers on the wafer surface by sequential injection and removal of reactants from the chamber. PNL deposition differs in that traditional CVD techniques and chemical reactant gases are injected independently, sometimes in pulsed form, instead of reactant gases injected simultaneously, and therefore they are generally not mixed in the chamber. For example, when using gases A and B, gas A is first injected into the chamber and molecules of gas A are chemically or physically adsorbed on the surface of the substrate to form a saturated layer of A. In general, the gas A remaining in the chamber is purified using an inert gas. Thereafter, gas B is injected to contact and react with the adsorbed layer of A to form a reaction product of A and B. Since this reaction is limited by the amount of A adsorbed, and A is distributed evenly on the substrate surface, good step coverage is obtained. B is allowed to flow on the substrate for a time sufficient to react between A and B. That is, all adsorbed A is consumed in the reaction. After this point, the remaining B and byproducts of the reaction are purged from the chamber. This process is repeated for multiple layers of material to be deposited.

PNL deposition is similar to atomic layer thickening (ALD). This also includes sequential injection of reactants into the wafer surface. The present invention preferably uses PNL technology. Generally, however, any ALD or atomic layer epitaxy (ALE) technique can be used. Various chemical vapor deposition (CVD) can also be used. Both ALD and PNL can use long or short reactant exposure dose times. The microstructure of the result is generally a function of the purity of the material, which in turn is adjustable in both ALD and PNL. Although the definition is slightly different, there is certainly a significant overlap between ALD and PNL. One important difference between PNL-W and ALD-W is that the preferred PNL-W process typically deposits 3-4 atomic layers of tungsten per deposition cycle, which is a single atomic layer (or more academically biased ALD process). Less than that). The operational difference is that pure ALD is generally performed at significantly lower pressures than PNL. (mTorr vs. 1-100 Torr) Pure ALD will use low pressure reaction pumping rather than relying on inert carrier atmosphere gas to sweep away the remaining reactants from the deposition chamber in the reactant purge step. As a result, pure ALD generally cannot contain reactive carrier gases (such as H ₂ ) in the DOS process, as planned for PNL-W and PNL-WN. In the specific case of PNL-WN deposition based on boron, B ₂ H ₆ is not a self-limiting process (unlike SiH ₄ or WF ₆ adsorption), but rather a true CVD process.

As above, tungsten nitride PNL deposition uses a reducing agent, tungsten precursor, and nitriding agent introduced at different times on the wafer surface. Nitriding agents are generally introduced after the elemental tungsten layer is formed. This layer is formed by contacting the reducing agent with the adsorbed tungsten precursor layer or by contacting the tungsten precursor with the reducing agent adsorbed layer or sacrificial layer. By either method, the chamber-purification operation separates the pulses of the individual reactants.

One example of a tungsten nitride PNL deposition technique is shown in FIGS. 2A-2D. This figure sequentially shows the cross section in which a tungsten nitride layer is formed on the substrate surface at the molecular level. Note that this figure is presented in a cartoon to enhance understanding of PNL deposition. These are not intended to limit the invention but are shown schematically.

In FIG. 2A, substrate 201 is exposed to diborane 203. Diborane species are adsorbed in the form of BHx species 205 on the surface of the substrate. This BHx species 205 interacts with the substrate surface to form boron species 207 of the element. Boron has been found to be a particularly good reducing agent for producing tungsten nitride by PNL. If possible, it works with the hydroxyl groups on the dielectric surface and strips them out to promote adhesion and nucleation. Boron also removes or draws off the donor fluorine from other processes or from low-k dielectrics. Perhaps this will form an anchor point and promote good adhesion therefrom to the underlying dielectric layer of tungsten.

At temperatures above about 250 degrees Celsius, boron species rapidly decompose to form elemental boron. This boron forming process continues as long as the boron precursor is present and as long as there is an inducing force for this decomposition reaction. This reaction proceeds well in dielectrics such as, for example, silicon dioxide and low k dielectrics. As above, the resulting boron layer of the element is sometimes referred to as the "sacrificial layer".

The formation of the boron sacrificial layer is not a self-limiting reaction. This boron can accumulate well to a thickness in excess of one atomic layer. Therefore, exposing the diborane 203 should be limited to the point where a boron layer of the required thickness is obtained. Generally the required thickness is about 1 to 2 monolayers of boron species. Preferably, the thickness is less than about 10 Angstroms, more preferably about 3 to 10 Angstroms.

During generation of the boron sacrificial layer, a suitable diborane gas pressure ranges from about 0.1 to 300 Torr, more preferably about 1 to 100 Torr and most preferably about 1 to 40 Torr. This applies to diborane reducing agents than others. Preferably, the temperature of the substrate is at least about 300 ° C. More preferably, it is about 300-500 degreeC, More preferably, it is 350-400 degreeC. In certain embodiments, diborane is supplied for about 0.5 to 3 seconds at a rate of about 200 sccm (per device). For example a two second purge is then applied to the bottom of the introduction of the tungsten precursor.

In FIG. 2B, once the normal boron sacrificial layer 211 is formed to the required thickness, the substrate surface 201 is exposed to tungsten hexafluoride (WF ₆ ) gas 209 or other suitable precursor. This tungsten hexafluoride (WF ₆ ) 209 species is reduced in contact with the boron sacrificial layer, thereby forming a thin tungsten metal film. See 215 layer of FIG. 2C. The thickness of this tungsten metal layer is limited to the amount that can be produced by the reaction of the boron sacrificial layer. Note that it is desirable to limit the dose of WF ₆ to prevent excessive exposure to this fluorinated formulation. Because fluorine can attack the properties of exposed silicon oxide or other substrates. Therefore, the thickness of boron is limited as above. Inert gases, such as argon, are used to dilute by-products or to remove species on the flowable gas in the reactor.

As noted above, in some embodiments, the surface of the newly formed tungsten metal layer is treated with hydrogen plasma to help remove chemical byproducts. The time of exposure to the hydrogen plasma is sufficient time to remove halogen by-products and precursors that remain unreacted and react with or not react with the reactants introduced in the next deposition cycle. If hydrogen plasma treatment is used, the hydrogen plasma pulse is preferably between about 0.1 to 30 seconds in duration.

During the reduction reaction of the tungsten containing precursor with the boron sacrificial layer, the appropriate gas pressure is determined as above for the deposition of the reducing agent. Suitable temperatures of the process are about 200 to 450 ° C, more preferably about 250 to 350 ° C. In a specific embodiment, tungsten hexafluoride is supplied at a flow rate of about 150 sccm (per device), with a feed time of about 0.25 to 3 seconds. This is accompanied by, for example, atmospheric gas for about 2 seconds.

Again in FIG. 2C, once the tungsten metal layer is formed and the by-products have been sufficiently purged, the nitriding agent is introduced to the substrate surface in the form of NH ₃ 213. In this embodiment, the tungsten metal layer is about one monolayer thick and the resulting tungsten nitride is therefore also about one monolayer thick. Tungsten nitride is shown in layer 217 of FIG. 2D. If the tungsten metal layer is more than one single layer thick, the reaction is controlled so that only the top of the tungsten metal is converted to tungsten nitride. In one embodiment, it is preferred that tungsten is converted to tungsten nitride. This is because tungsten nitride adheres to the dielectric more strongly than elemental tungsten. This provides another need to limit the thickness of tungsten formed during the B ₂ H ₆ -WF ₆ reaction. As mentioned above, if thicker tungsten nitride is required, the process of FIGS. 2A-2D can be repeated until the desired thickness is achieved.

During the nitriding process, the appropriate gas pressure is as described above for the reducing agent. Suitable process temperatures are about 200 to 450 ° C, more preferably about 300 to 400 ° C. The nitrogenous gas is converted, at least in part, into a plasma, which in one embodiment is introduced into the substrate by biasing the substrate together with the electrodes of the PF power supply. Moreover, this process uses an ultraviolet source or other energy stimulus to promote the nitriding reaction. In certain embodiments, ammonia is supplied for about 1 to 10 seconds at a flow rate of about 250 sccm (per device). This is accompanied by, for example, an atmospheric gas of about 2 seconds.

Dopant deposition unit operations are not shown in FIGS. 2A-2D. This process is optional and depends on the use of the tungsten nitride layer. If a separate dopant deposition operation is used, a pressure that matches one of those specified above for the reducing agent will be used. In addition, suitable temperatures will be used to proceed with the conversion and conversion of the dopant precursor to the elemental dopant. Examples of precursor compounds include B ₂ H ₆ , NH ₃ , SiH ₄ , phosphine and arsine.

As noted above, one aspect of the present invention provides the inclusion of a thin film of tungsten deposited prior to the first deposition of PNL-WN by PNL or other means. In one implementation, this is done by simply inhibiting the nitride sequence of the PNL-WN process during the first few deposition cycles. After the W seed layer is deposited, the PNL-WN process continues without vacuum break to deposit the required total capacity of PNL-WN and possibly PNL-W nucleation and CVD-W.

The present invention has found that they can achieve significant enhancement of adhesion to the dielectric by introducing a thin film (<50 ms and preferably about 10-50 ms) of tungsten seed layer prior to PNL-WN deposition. Possible mechanisms for this improvement include, but are not limited to, changing the stoichiometry of the WN thin film at the WN-dielectric interface if the native PNL-WN process is nitrogen rich for a more complete W2N composite. X-ray diffraction analysis studies show that this film is an apparent W2N microstructure with better adhesion.

Thin tungsten seeds may also be used to react with metal silicon compounds or silicon in connection, bitline, or other structures. The resulting tungsten silicon compound reduces the electrical resistance of WN-silicon compounds (NiSx, CoSix, TiSix) or WN-Si (N +, P +, poly-Si) interfades, as titanium acts in the current tungsten metallization scheme. You can. This tungsten seed reacts with silicon at a temperature of about 500 ° C. or more (preferably about 550-650 ° C.) to produce silicon compounds. Therefore, the interlayer of this thin silicon compound is suitable for the silicon or metal silicon compound of tungsten and tungsten nitride connection when the traditional Ti-TiN-W thin film stack is replaced by W-seed / PNL-WN / PNL-W / CVD-W thin film. There is a very advantageous advantage of reducing the connection resistance. These thin tungsten seeds are easily converted to tungsten silicon compounds at temperatures above about 550 ° C, which dramatically reduces the resistance of the WN-Si interface. (PNL-WN, on the other hand, is resistant to reaction with silicon until it reaches about 750 ° C.) This tungsten seed embodiment forms a thin and easily controlled tungsten silicon compound layer for reduced connection resistance. Let's do it.

In another aspect, the present invention provides a modified layer of PNL-WM and tungsten. If this layer is very thin (preferably less than about 10 mm 3) the structure of the layer is annealed to a more tungsten-rich form of tungsten nitride. By using a wide range of stoichiometric controls provided by layered WN-W deposition it is possible to regulate the working function of the composite material. This is very important for both the use of gate electrodes and capacitor electrodes. It is also beneficial to control the excess nitrogen of the final thin film by stoichiometric chemical bonding with stable tungsten nitride, such as W2N.

It is also desirable to form a layer structure starting with PNL-WN or PNL-W. In a preferred embodiment, the layered structure starts and ends with tungsten.

It is advantageous to carry out PNL-WN deposition in a hydrogen-argon environment or a fluorine removal environment. Creating PNL-WN thin films with very low fluorine content is important for many successful applications. One strategy where this object can be achieved is to use an Ar—H ₂ carrier gas mixture. H ₂ reacts with free fluorine to produce HF, which can be effectively removed from the semiconductor wafer. In a preferred embodiment, Ar-H ₂

The carrier gas mixture is approximately 50% hydrogen.

Applications

Various applications of the present invention are presented below.

Application 1: Copper Diffusion Barrier

The damascene process uses tungsten nitride as the copper diffusion barrier. It has also been found that the tungsten nitride layer acts as an adhesive layer to the copper metallization layer.

The copper metallization layer may comprise two components, a first seed thin layer and a second electrolyte bulk copper layer. The seed layer may be deposited by physical vapor deposition (PVD), CVD, ALD or electroless deposition. Electroless deposition is accomplished with a plating solution comprising a copper ion source, a reducing agent, and a base that renders the solution alkaline. PVD processes using hollow cathode electron tubes are preferred, but other PVD devices may be used.

Tungsten nitride is deposited by PNL or ALD. The copper seed layer can be formed indirectly through the metal tungsten layer or directly to the nitride. If a metal tungsten layer is used, a stack consisting of dielectric, tungsten nitride, metal tungsten and copper is formed. Metal tungsten can be formed by PNL, ALD, CVD or a combination thereof.

The following presents several process options.

1.PNL WNx-PNLW-CVDW (Two Stage Tungsten Deposition) -HCM-Cu Seed (HCM = Hollow Cathode Electron Tube) -Electrolytic Copper Plating

2. PNL WNx-PNLW-HCM-Cu Seed-Electrolytic Copper Plating

3. PNL WNx-CVDW-HCM-Cu seed-electrolytic copper plating

4.PNL WNx-HCM-Cu Seed-Electrolytic Copper Plating

5. PNL WNx-PNLW-CVDW-electroless copper seed and electrolytic copper filling

6. PNL WNx-PNLW-electroless copper seed and electrolytic copper filling

7. PNL WNx-electroless copper seed and electrolytic copper filling

In each case PNL WNx acts as an adhesive layer. PNL WNx or PNL-W both act as copper barriers. Advantages of the WN-W barrier over traditional barriers such as Ta-TaN include conformality (100% at> 10: 1 AR contacts) and improved electrophoresis resistance due to improved tungsten-copper bonding.

If the WN-W is not integrated with copper seed deposition (in a single vacuum integration tool), oxides may be removed from the W surface by argon or argon-hydrogen ion bombardment. For non-vacuum integrated W-Cu, W-Cu adhesion is greatly improved after cleaning.

Options 1 and 5 include a two stage tungsten deposition process including both CVD and PNL deposition processes. This can be done because the CVD deposited film has a low resistivity and the PNL deposited film has a good step cover.

For option 5-7, direct plating is more effective on tungsten surfaces than tungsten nitride, so the PNL-WN / PNL-W stack is preferred. Direct plating is performed on the tungsten / tungsten nitride barrier stack in basic plating solution to facilitate removal of the initial tungsten oxide from the Cu—W interface by solvation of tungsten oxide.

Application 2: tungsten plugfill

The PNL tungsten nitride deposition method of the present invention is used in a tungsten plug generation process for contact or via filling in IC wafer fabrication. The tungsten nitride layer acts as an adhesion layer to the diffusion barrier and the tungsten contact. The adhesive layer for the PNL WN barrier or tungsten plugfill is in contact with tungsten, metal silicides (TiSi _x , CoSi _x , NiSi _x , WSi _x ), silicon (N +, P +) or other electrically conductive material. It can be combined in an integrated contact plug film stack including PNL-W (seed layer) / PNL-WN (barrier-adhesive layer) / PNL-W (nucleation layer) / CVD-W (primary conductor and bulk plug fill). . This tungsten plugfill integration is intended to replace traditional Ti / TiN / W tungsten plugfill film stacks.

In current semiconductor metallization for device geometries over 90nm, tungsten is used as the primary conductor for contact or via applications. Typical film stacks used these days are PVD-Ti / CVD-TiN / W nucleation / CVD-W. PNL-WN can be used to simplify the film stack to PNL-WN / PNL-W (nucleation) / CVD-W. The application of PNL-W has many advantages over the art. First, eliminating the need for Ti-TiN semiconductors, manufacturers can eliminate the entire set of tools, greatly simplifying the process flow and reducing the cost of manufacturing semiconductor devices. The process temperature of PNL-WN and PNL-W is less than 400 ° C., in particular less than 300 ° C., compared to more than 450 ° C. for CVD-TiN of TDMAT and 500 ° C. or more for CVD-TiN produced from TiCl. Degradation of the process temperature allows logic and memory device manufacturers to replace titanium and cobalt silicide metal contacts with nickel silicides that phase change into new phases with much higher electrical resistance above 450 ° C. Removal of PVD-Ti in the tungsten metallization stack eliminates the protruding problems associated with the PVD process at the contacts or inlets of the vias. This creates inherent features that result in seam formation concentrations during subsequent filling of cylindrical contacts or vias with CVD-W. Replacing the Ti-TiN with PNL-WN to remove the protrusion results in no protrusion at the feature hole prior to CVD-W, thus reducing the center seam of the result. This seam can be opened during CMP of the tungsten film to complete tungsten metallization. Finally, by reducing the total thickness of the liner barrier needed inside the contacts or vias prior to CVD-W, PNL-WN reduces contact or via resistance compared to traditional PVD-Ti / CVD-TiN / CVD-W film stacks. This reduction in resistance is due to the fact that a larger proportion of contacts or via cross-sections can be used for lower resistivity tungsten deposition compared to higher resistivity Ti and TiN.

The advantages of tungsten nitride plug fill integration over traditional Ti-TiN integration are:

1) The PNL-WN / PNL-W process can fill vias and contacts with aspect ratios greater than 20: 1. This is a big improvement over the CVD-TiN / PVD-Ti step cover.

2) Elimination of Ti-TiN facilities and processing steps. This simplification reduces the cost of the tungsten plugfill process. In one aspect, precleaned wafers, PNL-WN, PNL-W and CVD-W are completed in a single wafer pass through an integrated cluster tool. Eliminating Ti-TiN deposition tools reduces wafer movement from tool to tool, reduces semiconductor clean room space, and reduces investment in semiconductor manufacturing facilities.

3) Reduced semiconductor wafer maximum processing temperature. Modern semiconductor wafer processing heat burden requirements require a maximum processing temperature of less than 450 ° C for contact metallization and less than 350 ° C for vias in low-K dielectrics. Both PNL-WN and PNL-W are deposited at wafer temperatures below 300 ° C.

4) Post-treatment of tungsten plugs-reduction of CMP center seam opening (coring). As tungsten grows centered on the contact or via side, a thin seam is left on the plug centerline. In the case of Ti-TiN liner barrier stacks the relatively poor step cover of (PVD) Ti and (CVD) TiN results in protruding features at the contact inlet. This forms smaller diameter features at the top than at the plug center. This protruding problem creates an open seam inside the plug because the tungsten film growing before the center of the feature is fully filled seals the top of the feature. These seams are exposed during CMP, resulting in wafer defects. In the case of the PNL-WN / PNL-W liner barrier film stacks, both materials are virtually 100% step covered and thus there is no seam during the CVD plug fill.

5) The thickness of the liner barrier layer is reduced from 200 μs for typical Ti-TiN to less than 50 μs for PNL-WN and is a primary conductor and CVD- with a lower resistivity than PVD-Ti, CVD-TiN or PNL-WN. W can be filled in a significant amount of contact or via. This reduces the electrical resistance of the small diameter contacts or vias where the liner barrier layer thickness is a significant proportion of the feature radius.

In the present invention, the PNL WN may be formed in the via or contact hole directly on the dielectric or with the PNL-W seed layer included. Tungsten is deposited by PNL, ALD, CVD or a combination thereof. The process can also be integrated with degassing or precleaning steps (eg plasma etching) prior to PNL-WN deposition. It may be advantageous to create a combined TiN / WN barrier layer.

The reactor used supports single wafer processing or multi-station sequential deposition using WN PNL and W CVD integrated into a single module. In one aspect wafer preheating, preclean and PNL-WN deposition are combined in one multi station process module and the second process module is dedicated to PNL-W and CVD-W deposition. In one aspect, wafer preheating / degassing and wafer pre-cleaning are performed as separate process modules in an integrated cluster tool for improved process flexibility.

The flow chart is as follows:

PNL-WN-PNLW nucleation layer-CVDW plug fill

2.PNL-W Seed / PNL-WN Barrier Adhesion / PNLW Nucleation / CVD-W Bulk Plug Fill

3. Preheat / Wafer Pre-cleaning / PNL-W Seed / PNL-WN Barrier Adhesion / PNLW Nucleation / CVD-W Plug Fill

4.PNL WN-CVD-W Plug Fill

5. WN-W in combination with degassing and precleaning (DFE or reactive cleaning)

6. WN-W and HCM-Ti (thin titanium layer deposited through hollow cathode electron tube) in combination with degassing and precleaning (DFE or reactive cleaning)

DFE is Dual Frequency Etch. For example, Novellus INOVA wafer precleaning (Novellus Systems, Inc., San Jose, Calif.) Uses argon ions from a dual frequency induced plasma to provide controllable ion acceleration (low frequency components) independently of high density plasma (high frequency components). do.

Application 3: Capacitor Electrode

Current DRAM capacitor electrodes using TiN / polysilicon electrodes both have high deposition temperature problems with both CVD-TiN (above 500 ° C, TiCl ₄ ) and polysilicon CVD (above 600 ° C, SiH ₄ ). These high deposition temperatures introduce reaction byproducts into the capacitor dielectric, reducing the dielectric constant of the dielectric and the capacitance of the resulting storage capacitor. High step cover is an essential requirement for advanced memory cell electrode formation and nearly 100% step cover in features with aspect ratios of 15: 1 or greater is superior to previous CVD and PVD WN processes. The work function of WN and W is 4 to 4,5 eV, which is higher than that of CVD-TiN (2.2 eV). The high work function of the electrode is known to reduce leakage of memory cell capacitors.

PNL deposited WN layers are used alone or as metal electrodes in a PNL-WN / PNL-W film stack. More generally, the WN layer at the top or bottom capacitor electrode can function as an adhesive layer, barrier layer, or primary electrical conductor. Tungsten is deposited by PNL, ALD, CVD or a combination thereof. Degassing or preliminary cleaning may be performed. Single wafer processing or multi-station sequential deposition can be used.

Integrated circuit capacitor electrodes are currently manufactured from film stacks consisting of CVD-TiN and highly doped polysilicon. The deposition temperatures of TiCl ₄ based CVD-TiN and polysilicon are at least 550 ° C. and at least 600 ° C., respectively. This high temperature introduces impurities (e.g., Cl) into the capacitor dielectric and oxidizes the TiN barrier layer to reduce capacitance and increase capacitor leakage. WN-W capacitor electrodes significantly reduce the fabrication temperature to reduce leakage and, in the case of comparable leakage, to improve post-anneal capacitance. The following process is used to deposit the upper and lower capacitor electrodes.

1.PNL-WN / PNLW nucleation layer / CVD-W

2.PNL- W / PNL-WN / PNL-W / CVD-W

3.PNL WN-CVD-W Plug Fill

4. WN-W in combination with degassing and precleaning (DFE or reactive cleaning)

Capacitors are trench capacitors, pin capacitors, plate capacitors, or other structures suitable for ICs. In the case of a stack capacitor, a bottom electrode is deposited on the polysilicon bottom electrode to facilitate structure formation. The high step cover of the PNL-WN and PNL-W enables the features needed to implement the PNL-WN for modern semiconductor memory cell electrodes.

Application 4: Gate Electrode

In this case, the WN functions as an adhesive layer, a barrier layer, or a primary electrical conductor at the gate electrode. Tungsten nitride is applied directly to the gate dielectric or to the polysilicon electrode to reduce the polysilicon line thickness.

Requirements for transistor gates are adjustable work function, thermal stability, and oxidation resistance, such as W / N stoichiometry of deposited films or boron (such as diborane), silicon (such as silane), or nitrogen (such as ammonia). The addition of a dopant can control the work function of PNL WN. In addition to boron and nitrogen, group IIIV dopants such as Al, Ga, P, As may be used. Another effective way to control the work function of PNL-WN is to create stratified PNL-WN and PNL-W structures. The number, thickness, and order of layers are variable, but very thin (less than 10 μs), with PNL-WN and PNL-W alternating, preferably starting and ending with PNL-W.

PNL-WN or WN / W film stacks as gate electrodes provide metal gates free of charge depletion found in non-metal gate electrodes made of polysilicon. Charge depletion increases the gate dielectric thickness. WN / W gate electrodes may be formed over the polysilicon gate electrodes to mitigate polysilicon gate requirements without changing the gate dielectric / polysilicon interface. For mixed N + and P + transistor devices, layered PNL-W / PNL-WN gate electrode structures can be fabricated to facilitate work function regulation.

Various processes are possible.

1.PNL- WN-PNL-W-CVD-W Bulk Deposition and Interconnection

2. Layered PNL- WN-PNL-W-CVD-W Bulk Deposition and Interconnection

3. PNL-WN / PNL-W Gate Electrode with Poly-Si Plug and Interconnect

4. PNL-WN / PNL-W / CVD-W on thin poly-Si gate electrode for poly-Si thickness reduction

5.PNL- WN-CVD-W Plug Fill

Tool configuration options are as follows:

1) WN-W in combination with degassing and precleaning (DFE or reactive cleaning)

2) Single wafer processing or multi-station sequential deposition using integrated WN, PNL, and WCVD in a single module

Application 5: Other Applications

PNL-WN acts as an adhesive layer and barrier for depositing bitline or wordline local interconnects in DRAM devices. It also serves as an adhesive layer for W deposition on oxides in semiconductor applications, such as light shielding of CCD devices. It also acts as a hard mask for low-K dielectric patterning. WN-W stacks can be deposited with less exposure of photoresist to hazardous amines at lower temperatures than traditional PECVD SiN hardmasks.

Deposition apparatus

The method can be carried out in several process chambers. For example Novellus Systems Concept2 Altus chambers, Novellus Systems Concept3 Altus chambers or various commercial CVD tools. The reactor may comprise a plurality of deposition stations in which parallel reactions occur (US Pat. No. 6,143,082). In one aspect pulsed nucleation takes place in a first station, which is one of multiple deposition stations equipped in a single deposition chamber. Reducing gases, tungsten-containing precursors and nitride gases are alternately introduced to the semiconductor substrate surface at the first station using separate gas supply systems that create a local pressure on the substrate surface. During the exposure to the reaction gas, an inert gas or a combination of hydrogen and inert gas is injected into the chamber. This process takes place simultaneously in multiple deposition stations. Or some stations do PNL deposition of WN and some do PNL or CVD deposition of WN. In this case, the PNL deposition of the WN may be at one or more stations. After the required PNL cycle is terminated to deposit the entire nitride thickness, the substrate is moved to another station and metal tungsten is deposited on the new WN layer. In one aspect, PNL-WN (including tungsten seed layer formation) is deposited in a dedicated process module and PNL-W and CVD-W are deposited in another process module in a vacuum integrated cluster tool.

In another aspect, the semiconductor substrate is moved between different deposition stations during a single PNL deposition cycle. In this case different stations are dedicated to different processes in the cycle. For example, one or two stations provide a reducing agent, one or two other stations provide tungsten precursors, and another station provides a nitriding agent. In other cases, a particular station provides a doping precursor. Various stations may be provided for simultaneous delivery of dedicated gas and inert gas. The WN is deposited by repeatedly moving the wafer between stations so that the wafer is repeatedly exposed to reducing gas, tungsten-containing precursor gas, and nitride gas until the required tungsten thickness is achieved.

In this scenario, the wafer is indexed from one deposition station to the next for parallel wafer processing. Indexing continues until all substrates are coated to the required thickness. Several deposition stations, each having a local pressure separate from an adjacent station, are possible in a single chamber.

Each station has a heated substrate support that heats and maintains the substrate at a predetermined temperature. Additionally there is a back gas distribution system that prevents deposition of the W film on the back of the substrate and a vacuum holding branch that holds the substrate to the support. Finally, the deposition chamber is equipped with a system for moving the substrate or wafer between and between deposition stations.

Another aspect of the invention provides a module having the following elements for alternating deposition of a WN or WN / W stack:

A plurality of deposition stations comprising a showerhead or a dispersion plate for uniform gas introduction in pairs with a heated structure holding the wafer under the showerhead;

A plurality of deposition stations having a showerhead mounted flush with the top of the module vacuum chamber to minimize gas recirculation in the module and to facilitate effective gas injection between alternate deposition steps;

A full gas injected top plate of the modular vacuum chamber consisting of an injection gas column covering the top plate area not occupied by the deposition showerhead for improved inter-station separation and shorter gas injection time between deposition cycles;

Wherein the heated structure of each deposition station is completely or partially separated from each other by an annular pumping ring connected to the chamber outlet. This feature further improves the separation between stations and allows different processes to be performed simultaneously in alternating stations in the same module.

The module further includes an apparatus provided for each showerhead and generating an RF plasma between the showerhead and the substrate support. Such means include RF energy sources, match networks, electrical connections. In another case the module comprises means for generating a remote plasma in the chamber. Some modules provide a line directly connected to the process vent exhaust so that process gas passes through the deposition chamber immediately after each material flow controller (MFCs) is turned on. In addition, a device may be provided for adjusting the line fill volume in the gas delivery system. This is important for the timely delivery of nitriding agent, tungsten precursor (WF ₆ ) or reducing agent (SiH ₄ , B ₂ H ₆ ). With this hardware, gas that is turned on and off during the PNL can be delivered.

The invention can be practiced using a gas branch pipe system that provides line charges to various gas distribution lines as shown in FIG. Branch 304 has an inlet 302 from a tungsten-containing precursor gas source, branch 311 has an inlet 309 from a diborane or other reducing gas source and branch 319 is an inlet from a nitriding source. Has 321. Branch tubes 304, 311 and 319 provide tungsten containing precursor gas, reducing gas and nitriding agent to the deposition chamber via valve distribution lines 305, 313 and 325. Various valves are opened and closed to fill the line and pressurize the distribution line. For example, valve 306 is closed against vacuum and valve 308 is closed to pressurize dispense line 305. After the appropriate time has passed, valve 308 is opened, valve 315 is closed and tungsten-containing precursor gas is delivered to the chamber. After a suitable time, the valve 308 is closed. The valve 306 is then opened to the true curve so that the chamber is vacuumed.

Similar procedures are used for reducing gas and nitriding delivery. Dispensing line 313 is filled by closing valve 315 and closing valve 317 against vacuum for introduction of reducing gas. Opening of the valve 315 delivers the reducing gas to the chamber. Similarly, dispensing line 325 is filled by closing valve 327 and closing valve 323 against vacuum for nitriding introduction. Opening of valve 327 delivers ammonia or other nitriding agent to the chamber. The time it takes to fill the line changes the initial delivery timing and amount of gas.

Figure 3 shows a vacuum pump in which valves 306, 317 and 323 are opened to purge the system. The supply of gas through the various distribution lines is controlled by controllers such as material flow regulators controlled by microprocessors, flow durations, flow rates, and sequenced digital signal processors.

If the process uses both a boron containing reducing agent and a silane reducing agent, there may be two subsystems for the reducing agent.

The PNL process requires accurate timing of the valves and material flow regulators to provide reagent pulses to the semiconductor substrate during PNL-WN deposition. In one approach, valve and MFC commands are delivered to the digital input / output controller (IOC) in discrete information packets containing instructions for time-dependent commands during PNL deposition. The Novellus C2 and C3 ALTUS system provides at least one IOC sequence. IOCs are physically located at various points in the device, such as in a process module or stand-power rack away from it. Each module has multiple IOCs (eg three per module). All instructions to control the valve and set the MFC flow (carrier and reactant gas) for the correct instructions contained in the sequence are contained in a single IOC sequence. Thus, the timing of all devices can be adjusted absolutely and relatively. There are usually a plurality of IOC sequences at any given moment. Thus, PNL is performed at stations 1-2 and timing is adjusted for all hardware components required for PNL-WN deposition at these stations. The second sequence may simultaneously deposit CVD-W in different deposition stations in the same module. The relative timing of the devices that control the delivery of reagents to stations 3-5 is important in this family of devices but the relative timing of the PNL process at stations 1-2 is offset from the relative timing of CVD at stations 3-5. The IOC translates information into a packetized sequence and delivers digital or analog command signals directly to a pneumatic solenoid bank or MFC that controls the valve. This reduces command execution delay in the valve or MFC by 5ms. A typical control system, one commanded to the IOC, experiences communication delays between computer control module operation and the IOC. In single command mode, the delay can exceed 250ms.

Nitriding agents are introduced by first flowing through a nitriding material flow controller (MFC) for good response and reproduction and then in a vacuum chamber to stabilize the flow before reaching the chamber. Valve 327 is closed and valve 323 is open to stabilize the nitriding flow. The distribution line system pressurizes the distribution line 325 to close the valve 323 to ensure a controlled initial release of the nitriding agent and the valve 327 is closed for about 0.10 to 3.33 seconds. The system then opens the valve 327 to the chamber WHr and closes the valve toward the vacuum so that the nitriding agent is delivered to the chamber during deposition. All valve timings are controlled using IOC command sequences. The above procedure can be applied to deposit tungsten nucleation layer, bulk layer, cap layer using PNL or CVD.

One branching system sequence for delivering boron containing gas (eg diborane) to the chamber is as follows. First during a period of MFC or other flow regulator stabilization, the system introduces a diborane-carrier gas mixture into the vacuum chamber. The nitrogen carrier gas takes 0.5-5 seconds for a 5% by volume diborane mixture. By closing the valve and the outlet of the chamber to the next vacuum, the system pressurizes the deborane delivery branch. This takes 0.1-5 seconds. This results in initial reagent ejection when the outlet to the chamber is opened. The outlet valve opens for 0.1 to 10 seconds. A suitable carrier gas is then used to release the diborane from the chamber.

The pulse of the tungsten containing gas is generated as follows. Initially, the system introduces WF ₆ into the vacuum chamber for a period of time during MFC or other flow regulator stabilization. This takes 0.5-5 seconds. By closing the outlet 306 and the outlet 308 to the chamber, the system then pressurizes the tungsten gas delivery branch. This takes 0.1-5 seconds. This results in initial reagent ejection when the outlet to the chamber is opened. The outlet valve 308 opens for 0.1 to 10 seconds. A suitable purge gas is then used to release the tungsten containing gas from the chamber.

The valves 317 and 315 are adjusted to effect pulsed flow of silane or other reducing agent. The process proceeds similarly to tungsten containing gas. After pulsing the reducing gas for 0.1-10 seconds, the valve 315 is closed and a purging gas is injected into the chamber.

The hardware elements used are:

1) PNL-WN process that is sufficiently heated and insulated so that all internal surfaces are maintained above 100 ° C to prevent condensation of NH4F, a by-product of B2H6-WF6-NH3 reaction.

2) A system of gas manifolds such that a single reaction MFC (paired with inert carrier gas MFC) is disrupted to supply the reactant to a multiple deposition station. This embodiment effectively shares components across multiple deposition stations, effectively reducing hardware costs. It also provides multiple stations from a single source, reducing the variety of wafers to wafers and stations to stations.

3) The system with the (# 2) shared reaction manifold wherein each outlet of the manifold to the deposition station is individually valved. As a result, the user can select whether a given reaction is delivered for a particular deposition station during a given deposition cycle. Multiple deposition also improves the ability of the mechanism to deliver layered films. For example, the ammonia outlet at the first deposition station can be closed for one or more deposition cycles to promote the growth of thin tungsten seeds (from the B 2 H 6 -WF 6 reaction).

4) The stoichiometric control is facilitated over the entire surface of the wafer by the presence of a receiving backside gas delivery hardware and backside ammonia for transferring Ar, H2 and NH3 to the backside of the wafer via the heated wafer susceptor. Without back side reaction control, a number of reactants are depleted at the wafer edge.

5) Inclusion of wafer station in PNL-WN precipitation module.

6) Inclusion of wafer preclean station in PNL-WN precipitation module.

7) Inclusion of reactive wafer preclean molecules using atomic and molecular fluorine to remove natural oxygen, etch residues and other contaminants from the surface of semiconductor wafers.

a) Use of inductively coupled plasma sources that dissociate NF3, CF4, C2F6 or other fluorine-containing gases as sources for atomic and molecular fluorine.

b) Handling the diver and line fill gas above to precisely control the fluorine precursor reaching the wafer for preliminary cleaning. Excessive etching is undesirable as it results in delicate fluorine attack, such as shallow silicide junctions, in direct plug-fill applications.

8) The incorporation of a chamber purge allows the majority of the module top that is not occupied by the showerhead to actively deliver reagents to the semiconductor being actively purged by an inert carrier. The showerhead is embedded in the purge gas column at the same height. This configuration can eliminate gas recirculation inside the chamber and shorten reagent release time. Gas curtains improve separation of multiple station chambers.

9) Time-dependent command control to valves and MFCs via the IOC allows valve timing sequences to be delivered in packets to the IOC and all commands for deposition are made without reading or writing data to the deposition module control computer. By eliminating slow read and write communications, the IOC can adjust valve and other device timing within 10-20ms, which is sufficient for PNL-WN processing due to short line charge, dosing, and purging times.

1 shows a flow diagram illustrating the related work used in the present invention. A thin, regular tungsten metal layer is formed on the substrate, followed by tungsten nitride to form tungsten nitride.

2A-2D of FIG. 2 show sequential cross-sectional views of the pulse nucleation layer deposition mechanism.

3 shows a schematic diagram illustrating the basic characteristics of a device suitable for the practice of the invention.

Claims (50)

 (a) depositing a gaseous boron-containing agent on the substrate to form a boron-containing sacrificial layer on the substrate; (b) exposing the boron-containing sacrificial layer to a tungsten-containing precursor to form a tungsten layer; (c) exposing the tungsten layer to a nitriding agent to form a first tungsten nitride layer; (d) performing tungsten nitride deposition with one or more additional cycles to complete the formation of the tungsten nitride layer, and forming a tungsten nitride layer on the substrate comprising contacting a reducing agent, a tungsten-containing precursor, and a nitriding agent in each additional cycle. How to The method of claim 1 wherein the substrate is a partially fabricated semiconductor device. The method of claim 1 wherein the tungsten nitride layer is deposited on at least a portion of the exposed dielectric of the partially fabricated semiconductor device. The method of claim 1 wherein the boron containing agent is borane. The method of claim 1, wherein the boron-containing sacrificial layer is a thickness of 3 ~ 20Å. The method of claim 1 wherein the tungsten-containing precursor is WF ₆ , WCl ₆ , W (CO) ₆ or a combination thereof. The method of claim 1 wherein the nitriding agent is N ₂ , NH ₃ , NF ₃ , N ₂ H _6, or a combination thereof. The method of claim 1 further comprising injecting a gas after at least one of steps (a), (b) and (c). The method of claim 1 further comprising forming a metal tungsten layer over the tungsten nitride layer after step (d). 10. The method of claim 9, wherein the metal tungsten layer is deposited by CVD. 10. The method of claim 9, wherein the metal tungsten layer is deposited by a pulsed nucleation layer process. 10. The method of claim 9, further comprising depositing a copper layer over the metal tungsten layer. 13. The method of claim 12, wherein the first copper layer portion is deposited on the metal tungsten layer using sputter deposition. The method of claim 13, wherein the second copper layer portion is deposited on the first copper layer portion by electrolytic plating. 13. The method of claim 12, wherein the copper layer is a copper seed layer. The method of claim 15, wherein the copper seed layer is deposited from the electroless plating solution. The method of claim 1 further comprising treating the tungsten layer produced in (b) with hydrogen or argon-hydrogen plasma prior to exposure to nitriding agent in (c). The method of claim 1 further comprising providing a dopant to the tungsten nitride layer. 19. The method of claim 18, wherein the dopant is phosphorus, arsenic, antimony, bismuth, boron, aluminum, gallium, indium, nitrogen or thallium. The method of claim 1, further comprising pretreating the substrate by an annealing process or plasma etching prior to step (a). The method of claim 1, wherein the at least one additional cycle of tungsten nitride deposition comprises exposing to a reducing agent that does not include a boron containing agent. The method of claim 21 wherein the reducing agent comprises silicon hydride. 2. The method of claim 1 including forming a metal tungsten layer on the tungsten nitride layer to form a gate electrode consisting of a tungsten nitride layer and a metal tungsten layer. 2. The method of claim 1 including forming a metal tungsten layer on the tungsten nitride layer to form a capacitor electrode consisting of a tungsten nitride layer and a metal tungsten layer. 2. The method of claim 1, wherein a metal tungsten plug is formed on the tungsten nitride layer to form a tungsten interconnect and the tungsten nitride layer acts as an adhesive layer, a diffusion barrier layer, or a nucleation layer for subsequent tungsten deposition. 27. The method of claim 25, further comprising forming a titanium layer prior to forming the tungsten nitride layer. The method of claim 1, wherein the boron-containing agent, tungsten-containing precursor and nitriding agent are carried in an inert carrier gas or a mixture of inert gas and nitrogen or hydrogen molecules. (a) placing the substrate in a deposition chamber; (b) depositing a gaseous reducing agent on the substrate to form a reducing agent layer on the substrate; (c) exposing the reducing agent layer to a tungsten-containing precursor to form a tungsten layer; (d) exposing the tungsten layer to a nitriding agent to form a first tungsten nitride layer; (e) A method of forming a tungsten nitride layer on a substrate comprising repeating steps (b) to (d) to complete tungsten nitride layer formation for one or more additional cycles. 29. The method of claim 28, wherein the reducing agent is borane. 29. The method of claim 28, wherein the reducing agent is silane. 29. The method of claim 28, wherein (c) is performed sequentially in steps (b) and (d) prior to (c). 29. The method of claim 28, wherein before step (b), step (c) is performed sequentially. 29. The method of claim 28, wherein the reducing agent, tungsten-containing precursor or nitriding agent comprises different compounds when used in forming the first tungsten nitride layer and when used in (e). 34. The method of claim 33, wherein the reducing agent comprises borane in step (b) and silicon hydride in step (e). 29. The method of claim 28 including forming a metal tungsten layer on the tungsten nitride layer to form a gate electrode consisting of a tungsten nitride layer and a metal tungsten layer. 29. The method of claim 28 including forming a metal tungsten layer on the tungsten nitride layer to form a capacitor electrode consisting of the tungsten nitride layer and the metal tungsten layer. 29. The method of claim 28, wherein a metal tungsten plug is formed on the tungsten nitride layer to form a tungsten interconnect and the tungsten nitride layer serves as an adhesion layer, a diffusion barrier layer, or a nucleation layer for subsequent tungsten deposition. 29. The method of claim 28, further comprising forming a copper seed layer over the tungsten nitride layer, wherein the tungsten nitride layer comprises a diffusion barrier. 29. The method of claim 28, wherein at least one of steps (b), (c), (d), and (e) is performed at a station different from (a) in a multi-station apparatus. 29. The method of claim 28 comprising the step of exposing the plasma after (b), (c), (d) or a combination thereof. 41. The method of claim 40, wherein the plasma is an RF plasma comprising Ar, N ₂ , H ₂ , NH _3, or a combination thereof. The method of claim 40, wherein the plasma is remote from the reaction chamber and comprises Ar, N ₂ , H ₂ , NH _3, or a combination thereof. 29. The method of claim 28, wherein tungsten nitride deposition is performed in a multi-station reaction chamber to (a) deposit tungsten nitride in one or more deposition stations of the chamber; (b) pulsed nucleation layer (PNL) tungsten is deposited at zero or one or more stations of the reactor; (c) CVD tungsten is deposited at zero or one or more stations of the reactor; (d) the substrate is moved from one deposition station to another to form a layered film consisting of WN, PNL-W and CVD-W; (e) The layered film of (d) is combined in any order and in any number of layers. 29. The method of claim 28, wherein tungsten nitride is deposited in a dedicated tungsten nitride module having one or more deposition stations and the tungsten nitride module comprises a wafer preheating station and a preclean station; The preclean module provides features for reactive precleans using fluorine based cleaning chemicals generated by decomposition of fluorine containing reagents using an inductive pair plasma; Wherein a wafer preclean station or another station in the tungsten nitride deposition module provides a feature to passivate the substrate after substrate preclean The method of claim 28, wherein the passivation of the substrate comprises: (a) hydrogen exposure; (b) remote H / H ₂ plasma exposure; (c) direct H / H ₂ or Ar / H / H ₂ or RF plasma exposure; (d) WF ₆ exposure; (e) H / H ₂ or H / H ₂ plasma and NH _{3 in} series or simultaneously; (f) by means of exposure to oxygen; 29. The method of claim 28, wherein the tungsten nitride deposition module is vacuum integrated with a tungsten CVD or pulsed nucleation dedicated module. (a) placing the substrate in a deposition chamber; (b) depositing one or more pulsed deposition tungsten layers on the semiconductor wafer; (c) depositing one or more pulsed deposition tungsten nitride layers on the one or more tungsten layers; (d) repeating (b) to (c) to form a tungsten nitride layer on the substrate, including the step of creating a W-WN bilayer or a multi-layer structure composed of multiple tungsten and multiple tungsten nitride layers 48. The method of claim 47, wherein the bottom layer of the W-WN composite film is a tungsten layer. 48. The method of claim 47, wherein the lower layer of the W-WN composite film is a WN layer. 48. The method of claim 47, wherein the atomic ratios of W and N in the W and WN layers are present in a ratio of 2 to 1 so that stoichiometric W ₂ N is formed directly or indirectly by heat treatment.