Adhesion between dielectric layers in an integrated circuit

Abstract

A method for improved adhesion between dielectric material layers at their interface during the manufacture of a semiconductor device, comprising operations for forming a first layer (1) of a dielectric material, specifically silicon oxynitride or silicon nitride, on a circuit structure (7) defined on a substrate of a semiconductor material (6) and subsequently forming a second layer (3) of dielectric material (silicon oxynitride or silicon nitride particularly) overlying the first layer (1). Between the first dielectric material layer and the second, a thin oxide layer (2), silicon dioxide in the preferred embodiment, is formed in contact therewith. This interposed oxide (2) serves an adhesion layer function between two superimposed layers (1,3).

Description

BACKGROUND AND SUMMARY OF THE INVENTION

This invention relates to processes for manufacturing monolithic integrated circuits which include overlaid layers of dielectric materials, in particular layers of silicon oxynitride and silicon nitride.

The adjacency of dielectric material layers having the same or different compositions is a common occurrence in today's technology at several manufacturing stages of integrated circuits. In such processes, a second layer is subsequently formed over a first layer of dielectric material. That second layer may overlap the first at a next stage so as to completely cover it, as is the case with certain passivating steps, for example. However, before forming the second layer, operations are often carried out such as the formation of a sacrificial layer which is then etched away, either fully or partially, or such as the partial removal of the first layer itself without using a sacrificial layer. In any case, the second dielectric layer will have bottom surface portions in direct contact with top surface portions of the first.

Dielectric material layers usually function to provide electric insulation between conductive layers and to protect underlying structures of the integrated circuit against contaminants (impurities, moisture) or impacts. Additionally, the provision of successive layers ensures protection for the device even when any one of the layers becomes damaged, e.g., by the formation of fine cracks. Therefore it is important that no regions be allowed to have less than perfect adhesion between the overlaid layers in the areas of contact. Nonetheless, some materials develop peculiar adhesion problems at the interface which arc not fully understood by those skilled in the art.

These problems already exist on account of certain inherent properties of the materials. In fact, when these are placed in layer form over another material, they develop by inherent stresses a more or less marked tendency to deform the underlying structure into a more or less curved shape, which may be concave or convex, according to whether the stress is a tensile or a compressive one. Where the adjacent underlying structure is also formed of another layer of dielectric material, the magnitude of the force acting on the two layers may, even when the stress is of the same type and perhaps of equal value, be different due to fact that the thicknesses are generally different. At high values of that stress, the layers tend to separate and possibly "delaminate". In general, spontaneous delamination occurs where the stress at the interface exceeds the molecular adhesion forces acting between the two layers.

This problem is aggravated by the application of external mechanical loads which add further stress, e.g., while dicing the wafer on which the circuit is formed. Other significant influences may include changes in temperature during the circuit manufacturing cycle, and environmental chemical attacks; these effects are apt to degrade the interlayer bonds.

The difficulty of achieving adhesion of the layers is further enhanced by certain methods, which are commonly applied during the manufacture of an integrated circuit.

In the first place, current deposition techniques generate layers with uneven regions especially near interfaces. As those skilled in the art are well aware, the dielectric material layers are customarily formed (excepting the first, which may be grown thermally) by deposition within a reactor, using a chemical vapor deposition (CVD) technique. The chemical precursors of the elements to be deposited are reacted while in a gaseous state. The chemico-physical properties of the layer and the reaction rate are controlled by process parameters such as pressure and temperature, and by the possible ignition of plasma induced by an RF drive voltage. However, a dielectric deposited by a CVD technique is never truly uniform throughout its thickness: the first tens of nanometers have different composition characteristics from the remainder of the layer because the surface electron states of the underlying layer, where the reaction takes place, affect its properties. Especially where the plasma-assisted CVD technique is used, from plasma ignition to the moment that the reactor enters a steady state of operation, a short settling and stabilizing period must be expected before the deposition rate itself attains steady state and becomes uniform. This may introduce increased uneveness in the characteristics of the deposited layer.

On the other hand, it is possible for an interruption in the deposition of a layer to cause saturation of the surface bonds, making the adhesion of the next layer more difficult still. In addition, where the layer on which the deposition is carried out has been subjected to a previous processing step, the surface to be deposited will in general be altered. For example, in the event of a previous plasma etching step by ion bombardment (a process used for removing a sacrificial layer, for example), damage is inflicted to the surface under the mechanical and physical aspects, possibly resulting in a change of density. Further, a surface alteration under the chemical aspect may occur due to the incorporation of residual spurious chemical elements from an etching process (etching chemical elements or residues from an attached layer).

The problem is of special importance because the trend in the art is nowadays toward assembling integrated circuits into inexpensive plastics packages. A characteristic feature of plastics materials is that they are pervious and, therefore, provide no barrier against moisture, while in addition, the circuit encapsulating operation--including resin heating and subsequent cooling, with attendant release of moisture from the resin--may easily admit water into the package. This water can eventually lead to corrosion of the conductive layers, or in any event, make the integrated circuit unstable, impairing its performance.

Thus, the technology for making such devices provides for the use of silicon oxynitrides and nitrides as dielectrics. These materials do provide a barrier against migrating water, and they can also effectively prevent contamination from sodium.

On the other hand, overlaid silicon nitride and oxynitride layers--whether in the nitride/nitride or oxynitride/oxynitride or nitride/oxynitride combination--have a marked tendency to separate after deposition, because they are highly rigid materials inherently likely to build up internal stresses.

In order to improve adhesion between dielectric materials, some techniques have been developed which provide treatments of the surface of an underlying layer prior to forming the next.

One prior solution consists of bombarding the surface of a layer to be deposited with ions of an inert gas, such as argon or nitrogen, under plasma (sputtering process). In this way, the roughness of the surface is enhanced mechanically to provide increased gripping area for the reactants to be deposited, which improves adhesion. That technique applies to the instance of oxide layers deposited using tetraethylorthosilicate as the precursor (known as TEOS).

Another solution which can be adopted where the surface of the underlying layer includes spurious chemical elements is the provision of a cleaning step by plasma etching, using an oxygen or mixed N.sub.2 and NH.sub.3 plasma, or by a wet etch. In this way, it becomes possible to provide improved chemical uniformity at the interface.

All these prior approaches may also be used in succession or simultaneously for improved efficiency, and have proved adequate with certain materials. However, particularly with silicon oxynitride and silicon nitride layers, they are ineffective to prevent separation of two successive layers, which not infrequently occurs some time after deposition, as can be verified experimentally.

The present invention advantageously provides a method for causing improved adhesion at the interface between layers of dielectric materials during the manufacture of integrated circuits having multilayered structures. Specifically, the method is also effective with layers of silicon nitride and oxynitride in the nitride/nitride or oxynitride/oxynitride or nitride/oxynitride combination, thereby improving the reliability of integrated circuit devices encapsulated in plastics packages.

BRIEF DESCRIPTION OF THE DRAWINGS

The invention and its advantages can be better understood by having reference to the following description of a specific process based on the inventive principle, given by way of non-limitative example, and to the accompanying drawings, wherein:

FIG. 1 shows a sectional view, not drawn to scale, of a circuit structure implementing the method of this invention;

FIG. 2 shows a sectional view of an application of the invention to a specific process.

DETAILED DESCRIPTION

In accordance with the invention, a thin layer of an oxide (silicon dioxide in the preferred embodiment) is formed at the interface between a first layer of a dielectric material and a successive layer of dielectric material, specifically silicon oxynitride and/or silicon nitride. This intermediate oxide acts as an adhesion layer between the two superimposed layers.

The method can be applied to the construction of semiconductor devices during discrete steps, e.g. passivation steps as are generally carried out in all processes.

As shown in the drawing, a method according to the invention comprises conventional initial operations in the manufacture of integrated circuits, such as the formation of a substrate 6 of a semiconductor material, e.g. monocrystalline silicon, of the circuit structure 7 sought. Then, a protective dielectric material layer 1 is usually formed over that structure, which may have conductive material layers at the top, the layer 1 conforming to the underlying structure. Thereafter, one or more dielectric material layers are formed in succession. Shown at 3 in FIG. 1 is just one of these layers. The dielectric layers 1 and 3 usually comprise silicon oxynitride or silicon nitride.

The top surface of the layer 1 may be treated, according to prior art methods and as previously noted, to promote adhesion of the layers 1 and 3.

In the preferred embodiment of the method according to this invention a thin oxide layer 2, specifically doped or non-doped silicon dioxide (SiO.sub.2), is interposed before the second dielectric layer 3 is formed.

The deposition of the dielectric layer 3 is followed by optional steps, not shown in the drawing, to complete the formation of the device, in accordance with conventional methods.

The layer 2 provides an adhesive layer between the layers 1 and 3, and serves no specific passivation function. It should be thin, its thickness dimension being preferably on the order of a few tens of nanometers, specifically within the range of 5 to 50 nm.

The decision to have the silicon dioxide doped, usually with boron or phosphorous, or left undoped will depend on the particular process that includes this adhesion step. Doping with boron to yield BSG (BoroSilicate Glass) improves the reflow capability of the layer by application of heat; silicon dioxide doped with phosphorous (PSG or PhosphoSilicate Glass) serves a barrier function against contaminants (by gettering). In any event, the amount of dopant used is normally kept fairly low.

The formation of silicon dioxide is preferably obtained by deposition using a CVD technique, either under a low pressure (LPCVD) or atmospheric pressure (APCVD) or a subatmospheric pressure (SACVD), or alternatively by plasma activation (PECVD=Plasma-Enhanced CVD). The chemical precursors used may also be varied: tetraethylorthosilicate (TEOS) and silane (SiH.sub.4) are the silicon sources most widely in use. Molecular oxygen (O.sub.2), optionally together with ozone (O.sub.3) and occasionally with nitrogen oxides (NO.sub.2, NO), are instead used for oxygen. The chemico-physical characteristics of the resultant layer depend on the deposition conditions, which should take the process into account to which it is subjected, and which dictate the precursors to be used. A major advantage, well recognized by the skilled ones, of deposition from silane is the low temperature that can be used during the process (below 500.degree. C.). The silicon dioxide deposited from TEOS has better layer properties than that from silane (excellent uniformity, close conformation with the underlying structure).

Doped (with boron, phosphorous, or both) silicon dioxide is commonly obtained directly upon CVD deposition of the layer, by also loading dopant-supplying molecules as reactants into the reactor. Specifically, diborane (B.sub.2 H.sub.6) or trimethylborate (TMB) are used for borosilicates, and phosphine (PH.sub.3) or trimethylphosphite (TMPi) or triethylphosphate (TEP) or trimethylphosphate (TMPa) for phosphosilicates. Of course, boron and phosphorous doping may be combined.

The adhesive properties of the oxide layer, which is interposed between two dielectric layers in accordance with this invention, have been verified experimentally. Using scanning or transmission electron microscopy (SEM, TEM), it is found that the presence of the layer of silicon dioxide is especially effective to prevent the two dielectric layers from separating. The effectiveness of the material has been verified both for doped dioxide and undoped dioxide. The exact reason why that thin layer improves the adhesion of dielectrics is yet not understood from the physical standpoint.

The adhesion function can be generally served by a thin layer of a not strictly stoichiometric silicon oxide (i.e., Si.sub.x O.sub.2-x not SiO.sub.2), and experimental tests point to the likelihood that a thin layer of an oxide other than of silicon, e.g. of titanium, may also be used.

It should also be considered that, while the materials to which the method described is preferably applied are silicon oxynitrides and silicon nitride, it may also prove effective with other dielectrics.

In a typical example, an application of the process according to the present invention will be described in relation to a final passivation process, as used for the manufacture of an EPROM memory encapsulated in a plastic package (i.e. flash EPROM or OTP EPROM). FIG. 2 shows schematically a cross section of such a device, as obtained at the end of the process (the elements identical to the elements of FIG. 1 are identically designated). The final passivation stack of FIG. 2, according to the invention, includes an oxide adhesion thin layer 2 to improve adhesion between two dielectric layers 1 and 3, where they are in contact. The dielectric layers 1 and 3, in this example, are silicon oxynitride layers. The formation of the passivation layers over a structure 7 like EPROM cells (including metal strips on the top) starts with a first oxynitride layer 1 (e.g. 0.4 .mu.m thick) completely covering the structure. A material with a low viscosity is then used to planarize the upper surface, as well known by the skilled in the art. Typically a SOG, a silane precursor in an organic liquid, is spun onto the entire surface, and then cured. A partial etch-back of the SOG leaves it only in the deeper zones (shown as 4 in FIG. 2). An oxide adhesion layer 2 is then formed over the portion of the first oxynitride layer not covered by the SOG, and over the residual SOG, before the formation of a second oxynitride layer 3 (e.g. 1.2 .mu.m thick). In this example, the oxide adhesion layer 2 is an undoped silicon dioxide deposited by PECVD using TEOS as precursor. Finally, a top PSG layer 5 is a standard requirement in a typical passivation process.

The final passivation scheme has therefore the following sequence of steps:

1) 0.4 .mu.m oxynitride deposition; 2) 0.4 .mu.m SOG spinning+SOG curing (425.degree. C., 45 min in N.sub.2); 3) SOG mini-bake for 60 seconds at 390.degree. C. (set point temperature) to dry SOG just before deposition; 4) cooling for 90 seconds; 5) SOG partial etch-back for 1750 A (69 sec total etch time+20 sec Ar sputter); 6) TEOS undoped (USG) oxide adhesion thin layer deposition (about 350 A thick, 3 sec of deposition), between the two oxynitride films; 7) 1.2 .mu.m oxynitride+0.4 .mu.m TEOS PSG depositions.

All the deposition on the process are in this example plasma enhanced. It has to be noticed that an Ar sputter step is maintained (also if it does not solve the decohesion problems, as already discussed) to clean the surface before the deposition of the thin oxide layer and to smooth the top corners of the first oxynitride layers.

The specific process conditions and film characteristics for the formation of the oxide adhesion thin layer are described in the following tables, for forming an adhesion layer of 350A:

TABLE 1__________________________________________________________________________ TEOS (close loop plasma dep. temp. spacing pressure control) O2film power (W) (.degree.C.) (mils) (Torr) sccm sccm__________________________________________________________________________PE-TEOS 440 390 190 9.0 930 470USG__________________________________________________________________________ TABLE 2__________________________________________________________________________ thick unif. % depos. moisture (ww & stress step-cov rate resistance densityfilm wtow) RI (dyne/cm.sup.2) (1:1 A.R.) (A/min) (%) (g/cm.sup.3)__________________________________________________________________________PE-TEOS 2 & 3 1.45 2.3 E8 50% 7600 1.7 2.12USG 3.sigma. .+-.0.1 compr.__________________________________________________________________________

Table 1 gives the parameters of the process. The deposition of the TEOS silicon dioxide has a duration of about 3 seconds and is preceded by 16 second of gas stabilization. The characteristics of the film are shown in table 2.

This has experimentally demonstrated that the introduction of the interface adhesion layer is successful both from the morphological (adhesion improvement) and electrical (no charge loss) points of view, if compared with samples without the silicon dioxide adhesion layer. Construction analysis can be used to confirm the degree of adhesion of the second oxynitride layer to the underlying surface, and therefore to the first oxynitride layer. Moreover, additional electrical tests show no significant difference between a process without the thin oxide and the process described. Therefore no negative effect on the passivation quality nor on the dielectrics reliability can be found. Killer integrity test at wafer level can be done to assure passivation integrity. No problems arising from a "stufa" test (4 hours at 425.degree. C.) can be found.

Besides, the addition of this cohesion step does not affect the net throughput of the system.

A thin thermal TEOS-O.sub.3 thin oxide gives also the same positive results as adhesion layer.

Thus, the method of this invention leads to an advantageous improvement of the adhesion between layers of dielectric materials, as is observed experimentally. It may be used to seal together layers in succession, thereby indirectly improving their insulative and protective effect on devices.

This effect is specially noticeable with silicon oxynitrides and silicon nitrides, which were at the root of the major problems encountered with other techniques.

It will be appreciated that the above embodiment has been described by way of illustration and not of limitation, and that many changes may be made thereunto without departing from the invention scope.

In fact, the embodiment of the invention described may be applied to all those processes wherein portions of dielectric material layers are to be superimposed. Additionally, the dissimilar characters of the CVD techniques used to form the adhesion oxide layer bear no significant influence on the adhesive property of the thin oxide layer.

Processing may continue with conventional step to complete fabrication, e.g. deposition and planarization of further interlevel dielectric, contact etch, metal deposition and etch, via patterning, protective overcoat deposition, or only etching contact pad apertures, etc. See generally VLSI TECHNOLOGY (2nd ed. Sze 1988); G. Anner, PLANAR PROCESSING PRIMER (1990); R. Castellano, SEMICONDUCTOR DEVICE PROCESSING (1991); W. Runyan & K. Bean, SEMICONDUCTOR INTEGRATED CIRCUIT PROCESSING TECHNOLOGY (1990); and the annual proceedings of the IEDM conferences.

Claims

1. An integrated circuit, comprising:

a substrate of a semiconductor material;

a first patterned conductor layer formed over said substrate;

a dielectric; and

a second upper patterned conductor layer;

wherein said dielectric comprises:

a first layer of a first silicon nitride compound;

a thin silicon dioxide layer in contact with said first layer of said first silicon nitride compound at least in a region; and

a second layer of a second silicon nitride compound overlaying said first layer of said first silicon nitride compound at least in said region and in contact with said thin silicon dioxide layer at least in said region;

the thickness of said thin silicon dioxide being less than both the thicknesses of said first and second layers of said first and second silicon nitride compounds

wherein the thickness of said thin silicon dioxide layer is in the range between 5 and 50 nm.

2. An integrated circuit according to claim 1, wherein said silicon dioxide is undoped and is formed by means of a CVD technique using TEOS.

3. An integrated circuit according to claim 1, wherein said silicon dioxide is doped with phosphorus.

4. An integrated circuit according to claim 1, wherein said silicon dioxide is doped with boron.

5. An integrated circuit according to claim 1, wherein said silicon dioxide is formed by thermal process using TEOS-O.sub.3.

6. An integrated circuit according to claim 1, wherein said first and second silicon nitride compounds are constituted by either a silicon oxynitride or stoichiometric silicon nitride.

7. An integrated circuit, comprising:

a substrate of a semiconductor material;

a patterned conductor layer formed over said substrate; and

a protective overcoat;

wherein said protective overcoat comprises:

a first layer of a first silicon nitride compound;

a thin silicon dioxide layer in contact with said first layer of said first silicon nitride compound at least in a region; and

a second layer of a second silicon nitride compound over-laying said first layer of said first silicon nitride compound at least in said region and in contact with said thin silicon dioxide layer at least in said region;

the thickness of said thin silicon dioxide being less than both the thicknesses of said first and second layers of said first and second silicon nitride compounds

wherein the thickness of said thin silicon dioxide layer is in the range between 5 and 50 nm.

8. An integrated circuit according to claim 7, wherein said silicon dioxide is undoped and is formed by means of a CVD technique using TEOS.

9. An integrated circuit according to claim 7, wherein said silicon dioxide is doped with phosphorus.

10. An integrated circuit according to claim 7, wherein said silicon dioxide is doped with boron.

11. An integrated circuit according to claim 7, wherein said silicon dioxide is formed by thermal process using TEOS-O.sub.3.

12. An integrated circuit according to claim 7, wherein said first and second silicon nitride compounds are constituted by either a silicon oxynitride or stoichiometric silicon nitride.