Integrated Circuit With Improved Tub Tie

1. Field of the Invention

The present invention relates to an improved technique for connecting integrated circuit transistors to the doped semiconductor regions in which they are formed, and the integrated circuits formed thereby.

2. Description of the Prior Art

In the design of integrated circuits employing field effect transistors, it is frequently necessary to electrically connect a source electrode to the doped semiconductor region in which it is formed. A doped semiconductor region in which transistors are formed, and having a different doping level or type than the semiconductor substrate (or epitaxial layer thereon) in which it is formed, is referred to as a "tub". For example, in complementary field effect device technology, (e.g., CMOS), the p-channel transistors may be formed in a n-tube connected to the positive power supply voltage (V.sub.DD), and the n-channel transistors may be formed in a p-tub connected to the negative power supply voltage (V.sub.SS). The sources of the p- and n-channel transistors are then electrically connected to their respective tubs, in order to receive their respective power supply voltages. One process for forming both p-tubs and n-tubs is disclosed in U.S. Pat. No. 4,435,896 coassigned with the present invention, with others also being known in the art. Furthermore, in some processes, only a single type of tub region is present, being of the opposite conductivity type as the substrate region in which it is formed. Then, the transistors of one conductivity type are formed in the tubs, and the transistors of the opposite conductivity type are formed in the surrounding substrate region. In either the twin tub or the single tub case, the electrical connection between the source and the tub in which it is formed is referred to as a "tub tie".

In order to provide a tub-tie, the prior art provides for opening two contact windows in the dielectric layers overlying both the source region and an adjacent portion of the tub. A low-resistance conductive link, typically aluminum, is then deposited so as to overlie the dielectric material between the windows, and extends into the windows to contact both the source and tub regions. However, this technique requires that space on the integrated circuit be provided for the two contact windows, which is typically greater than the space otherwise required for the device fabrication. An alternative prior art technique utilizes one large contact window, rather than two normal size windows. The large single window provides an opening over both the source and tub regions. A single conductive plug, typically aluminum, is then deposited into the window, thereby electrically connecting the two regions at the bottom of the window. However, this has the disadvantage that the window must be made large enough to ensure that it overlaps both the source and substrate. Typically more area is again required for layout of the circuit than if the tub tie were not present. Furthermore, the nonstandard size window complicates the use of computer aided design layout techniques. Therefore, the use of a large single contact window is usually reserved for hand-packed circuit layouts. In the foregoing techniques, to reduce the contact resistance, a more highly doped tub contact region is formed in the tub, before forming the tub tie. For example, a n+ doped region is provided in an n tub, and the contact window opened over the tub contact region, with the tub tie conductor then being deposited.

It is known to decrease the contact resistance to source/drain regions, as well as to the gate electrode, by forming a silicide surface layer thereon. This may be accomplished in a single operation by the so-called "salicide" process, by reacting a refractory metal (e.g., titanium) with the exposed silicon of the source/drain regions, as well as the polysilicon gate electrode; see, for example, "Titanium disilicide self-aligned source/drain+gate technology", C. K. Lau et al, IEDM Technical Digest, pp. 714-717 (1982).

We have invented a technique for forming an electrical connection between a field effect transistor source region and the doped silicon semiconductor tub region in which it is formed. A metal silicide layer is formed on both the source region and a tub contact region during the same process that forms the metal silicide on the drain and gate. A dielectric layer may be deposited onto the silicide layer, and a window may be formed therein for connecting the source and tub regions to a power supply conductor.

FIG. 1 shows a cross section of an integrated circuit substrate utilizing the inventive tub tie.

FIG. 2 shows schematically the tub tie connections in a CMOS inverter.

The following detailed description relates to an improved technique for forming an integrated circuit tub tie. Referring to FIG. 1, a cross section of a CMOS inverter utilizing the present technique is illustratively shown. A n tub 101 and p tub 102 are located in a silicon semiconductor body 100. As used herein, the term "tub" means a region formed in the silicon body that has a different conductivity type, or doping level (or both) than the bulk of the silicon body. The body 100 may be a single crystal silicon substrate, or silicon epitaxial layer, with polycrystalline silicon bodies also being possible. Formed in the n tub is a p channel transistor having p+ source region 103 and drain region 104. Adjacent to source region 103 is a n+ tub contact region 105. In the present technique a metal silicide layer 106 electrically connects source region 103 to the tub contact region 105, and hence to the n tub 101. The metal silicide layer is contacted by a conductor 128, typically aluminum, which connects to a positive power supply voltage, V.sub.DD. Hence, both the n tub 101 and the source are placed at V.sub.DD in operation. Metal silicide layers 107 and 108 are also formed on the drain region 104 and polysilicon gate electrode 109, respectively. The gate electrode is insulated from the channel portion of the n tub by gate insulator 110, typically silicon dioxide, with sidewall spacers being optionally included when the lightly-doped drain technique is used to improve transistor reliability in the face of electric field induced degradation.

In the illustrative embodiment, the source 111 of the n channel transistor formed in p tub 102 is similarly connected by metal silicide layer 112 to the p+ tub contact region 113. Conductor 114 is connected to the negative power supply voltage V.sub.SS, thereby placing both the source region 111 and the tub 102 at V.sub.SS in operation. Metal silicide layers 116 and 117 are formed on drain region 115 and polysilicon gate electrode 118, respectively. A conductor 119 connects the drain regions 104 and 115 of the p and n channel transistors, respectively. Isolation is provided by field oxide regions 120-122, and the conductors 128, 114, and 119 are insulated from underlying regions by glass layers 123-127.

A typical process sequence for forming the foregoing regions is as follows: The p and n tubs (101, 102) are formed in a silicon wafer, and the field oxide regions (120-122) formed. The gate oxide regions (110, 118) and n+ polysilicon gate electrodes (109, 117) are patterned. A n- lightly doped drain (LDD) implant is then accomplished for the n-channel transistor, and a p- LDD implant for the p-channel transistor. A silicon dioxide layer is then conformably deposited, and anisotropically etched to form the sidewall spacers (129, 130, 132, 133). The n+ and p+ source/drain implants for the n-channel and p-channel transistors are then accomplished, followed by a high- temperature anneal. The metal silicide layers 106, 107, 108, 112, 116 and 117 are then formed, by reacting the exposed silicon surfaces of the corresponding regions with a metal. The metal is typically refractory, and in the preferred embodiment is titanium, with others being possible, including tantalum, molybdenum, tungsten, and cobalt.

A typical process known in the art for forming the metal silicide is as follows: A 1000 angstrom (0.1 micrometer) thick layer of titanium was firstly blanket deposited over the regions previously formed on the substrate. A silicon ion beam implant was then accomplished at a 110 keV beam energy, 1.times.10.sup.15 ions/cm.sup.3 dose. This mixes silicon atoms with the native oxide on the titanium/silicon substrate interface, thereby providing for more consistent formation of the silicide. The titanium was then rapidly thermal annealed at 625 degrees C. for 60 seconds in a nitrogen atmosphere, by exposure to a flashlamp. This formed titanium silicide in those regions where the deposited titanium contacted exposed silicon, being the source/drain regions, the tub contact regions, and the gates. The titanium was selectively removed from the other regions (e.g., from over the field oxide and the gate sidewall spacers) by etching in a solution of phosphoric acid and hydrogen peroxide for 9 minutes at 80 degrees C. The silicide was again rapidly thermal annealed in nitrogen, at 900 degrees C. for 20 seconds, for improved resistivity. Following the silicide formation, a glass layer was deposited, and windows formed therein, producing regions 123-127. An aluminum layer was then deposited and patterned, forming conductor regions 114, 119 and 128. Still other layers, for example additional dielectric and conductor layers, and oxide or nitride cap layers, may be deposited as desired.

A schematic of the complementary inverter formed as above is shown in FIG. 2. However, the present technique may be applied to transistors in any circuit. Furthermore, it is not necessary in some cases to connect the sources of both n and p channel transistors to their respective tubs. For example, in a single tub process wherein the p-channel devices are located in n tubs formed in a p type substrate, only the sources of the p-channel devices may need be connected to their tubs. Even then, it is not necessary in some cases to connect all of the p-channel devices to their tubs. Note also that the present technique does not require opening a contact window at all in order to form the tub tie. For example, a tub tie may be provided for a given number of transistors, and contact windows opened over a fewer number of tub ties, in order to save space. Finally, although the foregoing has been illustrated using insulated gate (e.g., CMOS) technology, it may also be used for other types, including junction field effect transistor technology.