The invention relates generally to silicon integrated circuit design and process technology. In particular, the invention pertains to trench isolation process technology.
Implementing electronic circuits involves connecting isolated devices through specific electronic paths. In silicon integrated circuit fabrication it is necessary to isolate devices from one another which are built into the same silicon matrix. They are subsequently interconnected to create the desired circuit configuration. In the continuing trend toward higher device densities, parasitic interdevice currents become more problematic, thus isolation technology has become one of the most critical aspects of contemporary integrated circuit fabrication.
Over the last few decades a variety of successful isolation technologies have been developed to address the requirements of different integrated circuit types such as NMOS, CMOS and bipolar. In general, the various isolation technologies exhibit different attributes with respect to such characteristics as minimum isolation spacing, surface planarity, process complexity and defect density generated during isolation processing. Moreover, it is common to trade off some of these characteristics when developing an isolation process for a particular integrated circuit application.
In metal-oxide-semiconductor (MOS) technology it is necessary to provide an isolation structure that prevents parasitic channel formation between adjacent devices, such devices being primarily NMOS or PMOS transistors or CMOS circuits. The most widely used isolation technology for MOS circuits has been that of LOCOS isolation, an acronym for LOCal Oxidation of Silicon. LOCOS isolation essentially involves the growth of a recessed or semirecessed oxide in unmasked non-active or field regions of the silicon substrate. This so-called field oxide is generally grown thick enough to lower any parasitic capacitance occurring over these regions, but not so thick as to cause step coverage problems. The great success of LOCOS isolation technology is to a large extent attributed to its inherent simplicity in MOS process integration, cost effectiveness and adaptability.
In spite of its success, several limitations of LOCOS technology have driven the development of alternative isolation structures. A well-known limitation in LOCOS isolation is that of oxide undergrowth at the edge of the mask which defines the active regions of the substrate. This so-called bird's beak (as it appears) poses a limitation to device density, since that portion of the oxide adversely influences device performance while not significantly contributing to device isolation. Another problem associated with the LOCOS process is the resulting circuit planarity or lack thereof. For submicron devices, planarity becomes an important issue, often posing problems with subsequent layer conformality and photolithography.
Trench isolation technology has been developed in part to overcome the aforementioned limitations of LOCOS isolation for submicron devices. Refilled trench structures essentially comprise a recess formed in the silicon substrate which is refilled with a dielectric material. Such structures are fabricated by first forming submicron-sized trenches in the silicon substrate, usually by a dry anisotropic etching process. The resulting trenches typically display a steep sidewall profile as compared to LOCOS oxidation. The trenches are subsequently refilled with a dielectric such as chemical vapor deposited (CVD) silicon dioxide (SiO2). They are then planarized by an etchback process so that the dielectric remains only in the trench, its top surface level with that of the silicon substrate. The etchback process is often performed by etching photoresist and the deposited silicon dioxide at the same rate. The top surface of the resist layer is highly planarized prior to etchback through application of two layers of resist, and flowing the first of these layers. Active regions wherein devices are fabricated are those that were protected from etch when the trenches were created. The resulting structure functions as a device isolator having excellent planarity and potentially high aspect ratio beneficial for device isolation. Refilled trench isolation can take a variety of forms depending upon the specific application; they are generally categorized in terms of the trench dimensions: shallow trenches (<1 μm), moderate depth trenches (1–3 μm), and deep, narrow trenches (>3 μm deep, <2 μm wide). Shallow Trench Isolation (STI) is used primarily for isolating devices of the same type and is often considered an alternative to LOCOS isolation. Shallow trench isolation has the advantages of eliminating the birds beak of LOCOS and providing a high degree of surface planarity.
The basic trench isolation process is, however, subject to drawbacks, one of these being void formation in the trench during dielectric refill. Such voids are formed when the refilling dielectric material forms a constriction near the top of a trench, preventing flow of the material into the trench interior. Such voids compromise device isolation as well as the overall structural integrity. Unfortunately, preventing void formation during trench refill often places minimum size constraints on the trenches themselves, which may compromise device packing density or device isolation. For example, a key parameter measuring device isolation is the field threshold voltage between adjacent devices, that is, the voltage necessary to create a parasitic channel beneath a field oxide region linking adjacent devices. The field threshold voltage is influenced by a number of physical and material properties of the trench isolator such as insulator thickness, dielectric constant ε, substrate doping, field implant dose and substrate bias. Thus, a principal difficulty in decreasing the trench depth is the compromise in device isolation. Clearly, it is highly desirably to develop a shallow trench isolation process which overcomes the problem of void formation while providing effective device isolation.
It is an object of the present invention to provide a trench isolation process which alleviates the problem of void formation during dielectric refill. It is another object of the present invention to provide a trench isolator having reduced dimensions, advantageous for device density and wafer planarity. It is a further object of the present invention to provide a shallow trench isolator having enhanced device isolation characteristics.
In accordance with one aspect of the present invention, a process for isolating devices on a semiconductor substrate comprises first removing portions of the semiconductor substrate, thereby forming recesses preferably having a trench profile. The trenches are then refilled with a material having a dielectric constant lower than the dielectric constant of silicon dioxide which is about 3.9. Using a low-ε dielectric material allows the trench dimensions to be reduced while still providing effective device isolation characteristics. Preferably, the dielectric material comprises a halide-doped glass such as Fluorine-doped SiO2. To insure against device contamination, the invention further comprises forming a barrier layer over the trenches prior to refilling them with the low-ε dielectric material.
In accordance with another aspect of the present invention, an isolation structure in a semiconductor substrate comprises a recessed trench formed in the semiconductor substrate and a material having a low dielectric constant filling the trench. The trench structure preferably has a depth less than 250 nm, and furthermore comprises a barrier layer disposed between the interior trench surface and the dielectric material. The dielectric material preferably has a dielectric constant lower than about 3.9, and may comprise a Fluoride-doped silicon dioxide composition.
In accordance with yet another aspect of the present invention, a method of reducing the formation of voids in a refilled trench isolation process comprises forming trenches having an aspect ratio less than about 2:1, and then refilling the trenches with a material having a dielectric constant less than the dielectric constant of silicon dioxide. The trenches preferably have a depth of less than 200 nm, and are refilled with a material comprising a Fluorine-doped silicon dioxide composition.
These and other aspect and attributes of the present invention will become more fully apparent with the following detailed description and accompanying figures.
In accordance with the principles of the present invention, an improved shallow trench isolation technology utilizes a trench that is shallower than prior art trenches, and yet provides the same degree of device isolation. The shallower trench helps prevent the formation of voids during dielectric refill. However, despite the smaller dimensions of the present inventive trench, equivalent device isolation is achieved through use of a dielectric refill having a lower dielectric constant ε than in prior art isolation trenches. To better illustrate these inventive principles, a brief description of an exemplary STI process is provided first hereinbelow.
An exemplary STI process may comprise first a masking, patterning and dry etch process, producing trenches in the silicon substrate as shown in
After the trenches 14 are formed, the mask 12 is removed by selective etching or chemical mechanical polishing and the trenches are refilled with a dielectric material 16, as shown in
Following the trench refill 16, the top surface of the substrate 10 is planarized by an etchback process, typically also performed using a chemical/mechanical polish. Prior to etchback, the substrate 10 may be coated with a layer of photoresist (not shown) in order to provide a planar surface with which to begin the etchback. The etchback itself provides a planarized substrate surface 18, having dielectric material 16 filling the trenches 14 up to and level with the top surface 18.
As shown in
In accordance with the principles of the present invention, void formation is alleviated by decreasing trench depth. Utilizing shallower trenches decreases the possibility of void formation and favorably increases surface planarity of the final refilled trench structure. For example in a typical DRAM application, a trench in accordance with the present invention may have dimension of approximately 200 nm deep and 250 nm wide while prior art trenches typically have dimensions of approximately 275 nm deep and 350 nm wide. However, as is well known in the art, a key parameter measuring device isolation is the field threshold voltage between adjacent devices, that is, the voltage necessary to create a parasitic channel beneath a field oxide region linking adjacent devices. The field threshold voltage is influenced by a number of physical and material properties of the trench isolator such as insulator thickness, dielectric constant ε, substrate doping, field implant dose and substrate bias. Thus, a principal difficulty in decreasing the trench depth is the compromise in device isolation.
To circumvent this problem, the shallow trench isolation of the present invention maintains effective device isolation in a shallower trench by utilizing dielectric materials having a lower dielectric constant than used in the prior art. For a given trench geometry, the field threshold voltage is a decreasing function of the field dielectric constant. Thus, to compensate for smaller trench dimensions, the present invention utilizes dielectric materials having lower dielectric constant. A possible dielectric material is a low index glass such as a halide-doped silicon dioxide, deposited by introducing the halide during CVD of silicon dioxide. For example, F:SiO2 possesses a dielectric constant of approximately 3.2, while typical CVD-SiO2 has a dielectric constant of about 3.9. Use of such materials allows a relative decrease in trench depth by about 20%.
Fluorine or other elements comprising a reduced dielectric constant material may however cause deleterious effects on neighboring devices if they diffuse into adjacent active areas. Therefore, a preferred embodiment of the present invention also incorporates a diffusion barrier layer lining the trench so as to prevent dopant migration into the silicon substrate. Use of a preferred barrier layer in the form of a grown oxide or nitride film, or a deposited stoichiometric or non-stoichiometric oxide or nitride film inhibits contamination of the isolation field-effect transistor, thereby preserving desirable characteristics such as a high threshold voltage. The integrated devices subject to isolation are also protected by the barrier layer from contamination.
In accordance with the aforementioned principles, a preferred shallow trench isolation may for example comprise the following process steps illustrated in
As mentioned previously, to avoid contamination of substrate regions adjacent to the trenches 22, it is preferable to form a barrier layer 24 over the trenches 22 prior to dielectric refill as shown schematically in
As shown in
To complete the trench structure, a planarizing step is performed as shown in
Thus, the present invention provides several advantages over the prior art by avoiding cavities in the trenches, providing more effective device isolation using low-ε materials and having a greater degree of planarity in the final trench structure.
Although described above with reference to the preferred embodiments, modifications within the scope of the invention may be apparent to those skilled in the art, all such modifications are intended to be within the scope of the appended claims.
This application is a continuation of and claims priority under 35 U.S.C. § 120 to U.S. application Ser. No. 08/932,228, filed on Sep. 17, 1997, now U.S. Pat. No. 6,831,347, which is a divisional of U.S. application Ser. No. 08/547,620, filed on Oct. 24, 1995, now U.S. Pat. No. 5,702,976, both of which are entitled SHALLOW TRENCH ISOLATION USING LOW DIELECTRIC CONSTANT INSULATOR, and each of which applications is incorporated herein by reference in its entirety.
Number | Name | Date | Kind |
---|---|---|---|
4109030 | Briska et al. | Aug 1978 | A |
4502914 | Trumpp et al. | Mar 1985 | A |
4567061 | Hayashi et al. | Jan 1986 | A |
4624046 | Shideler et al. | Nov 1986 | A |
4748134 | Holland et al. | May 1988 | A |
4959325 | Lee et al. | Sep 1990 | A |
5087586 | Chan et al. | Feb 1992 | A |
5139967 | Sandhu et al. | Aug 1992 | A |
5159428 | Rao et al. | Oct 1992 | A |
5169491 | Doan | Dec 1992 | A |
5191509 | Wen | Mar 1993 | A |
5260229 | Hodges et al. | Nov 1993 | A |
5294563 | Rao | Mar 1994 | A |
5298451 | Rao et al. | Mar 1994 | A |
5302233 | Kim et al. | Apr 1994 | A |
5338968 | Hodges et al. | Aug 1994 | A |
5356828 | Swan et al. | Oct 1994 | A |
5358892 | Rolfson | Oct 1994 | A |
5358894 | Fazan et al. | Oct 1994 | A |
5366590 | Kadomura | Nov 1994 | A |
5369051 | Rao et al. | Nov 1994 | A |
5372951 | Anjum et al. | Dec 1994 | A |
5429995 | Nishiyama et al. | Jul 1995 | A |
5459096 | Venkatesan et al. | Oct 1995 | A |
5470783 | Chiu et al. | Nov 1995 | A |
5492736 | Laxman et al. | Feb 1996 | A |
5492858 | Bose et al. | Feb 1996 | A |
5530293 | Cohen et al. | Jun 1996 | A |
5567553 | Hsu et al. | Oct 1996 | A |
5571576 | Qian et al. | Nov 1996 | A |
5637528 | Higashitani et al. | Jun 1997 | A |
5661335 | Anjum et al. | Aug 1997 | A |
5672539 | Thakur et al. | Sep 1997 | A |
5712186 | Thakur et al. | Jan 1998 | A |
5846888 | Chapek et al. | Dec 1998 | A |
6072226 | Thakur et al. | Jun 2000 | A |
Number | Date | Country |
---|---|---|
02077127 | Mar 1990 | JP |
02077127 | Mar 1990 | JP |
04074425 | Mar 1992 | JP |
153717 | Jun 1996 | JP |
Number | Date | Country | |
---|---|---|---|
20050087833 A1 | Apr 2005 | US |
Number | Date | Country | |
---|---|---|---|
Parent | 08547620 | Oct 1995 | US |
Child | 08932228 | US |
Number | Date | Country | |
---|---|---|---|
Parent | 08932228 | Sep 1997 | US |
Child | 11012012 | US |