Integrated Circuit Contact Technique

As the size of active impurity regions in semiconductor integrated circuit substrates shrink the size of the electrical contacts to those regions likewise shrink. A limiting factor is the minimum line that can be produced in a given technology. When the impurity region is reduced to that minimum dimension the contact "window" that is made prior to forming the contact cannot, as is customary, be made smaller than the region to be contacted. As a result both features are made with the minimum feature size and the contact will almost certainly be partially non-aligned or offset with the region to be contacted. This situation is commonly referred to as "line-on-line" features, i.e. features that require minimum line on minimum line registration. Whether the alignment is visual or automatic it is not possible to make precise line-on-line features.

This difficulty can be overcome by implanting an appropriate impurity into the contact window prior to forming the contact. This produces a self-aligned contact with the region to be contacted.

As useful as this tool evidently is, it suffers when applied to CMOS technology. In that case two implants and additional masks are required because p-type impurities are required for the p-channel device windows and n-type impurities for the n-channel device windows.

A need remains for a technique for producing line-on-line contacts that can be applied to CMOS as well as other technologies.

One such technique is described and claimed in application of Ser. No. 597,924, filed Apr. 9, 1984. This application is directed to an alternative solution. It is based on two important recognitions. One, in many CMOS circuits the number of devices of each conductivity type is quite different. For example, VLSI memory devices may combine n-channel memory cells with CMOS peripheral drive and buffer circuits. Such devices may have 90% n-channel gates and only 10% p-channel. Secondly, with careful choice and control of the doping levels in the device it is possible to expose non-selectively all the source-drain windows on the chip to a single implant step, and convert the exposed channel stop in the offset portion of the windows of one conductivity type without converting the source-drain region in the window of the other conductivity type. A key to achieving this is to have the source-drain region more heavily doped than the channel stop. Having recognized these two potential properties of CMOS devices it is possible to apply a single non-selective or blanket implant to devices of both conductivity types and correct for the offset problem in the windows of devices of one conductivity type, in the most advantageous case, devices of the prevalent type. The source-drain regions of the other type of device are made large enough to avoid the offset problem.

FIG. 1 is a schematic representation of a cross section of portions of an integrated CMOS chip at the point in processing showing the problem introduced by offset line-on-line contacts;

FIG. 2 is a schematic view similar to that of FIG. 1 further illustrating the proposed solution according to the invention;

FIG. 3 shows the device of FIG. 2 with the contact layer in place.

The problem that the invention addresses is illustrated in connection with FIG. 1. Reference is made to copending application of Ser. No. 597,924, filed Apr. 9, 1984 for a more detailed description of the background of the offset contact window problem that arises as a consequence of line- on-line contacts. In FIG. 1, sections of an exemplary CMOS integrated circuit chip are shown. The device on the left of the broken chip section is exemplary of an n-channel device. Typically this device may appear on the chip as part of the memory array, or as one of a CMOS pair of devices used, for example, in the peripheral memory cell support circuits or in a wide variety of logic devices. This n-channel source-drain contact is made line-on-line, which means that the width W of the source-drain region 21 is approximately the same as the width W' of contact window 22. As shown, the contact window is offset by an amount .DELTA., which corresponds in this exemplary showing to the alignment tolerance set by the lithographic design rules. The contact may be misaligned less than .DELTA., so shown here is the worst case. The consequence of the offset is the potential short between the contact layer that will subsequently be applied (see FIG. 3) and the channel stop region, shown at 23. That results in a short from the contact to the substrate. In a conventional CMOS pair of devices, the other device of the pair, the p-channel device, is of comparable size, and will exhibit a similar short if the source-drain contact is made line-on-line. It is evident that a non-selective implant using impurities of one conductivity type, or the other, will cure one shorting problem but not the other.

The solution according to this invention is to design one type of device, in this case the p-channel devices, with source-drain contact widths of the order of W+2.DELTA.. This insures that the offset that creates the shorting problem with the n-channel devices will not create the same problem with the p-channel devices. Shown in FIG. 1 is source-drain region 21' with width W+2.DELTA.. The contact window, having a width approximately equal to W, will not expose the n-channel stop 23' and no shorting problem exists. At this point a non-selective ion implant, as shown in FIG. 2, converts the exposed channel stop in the n-channel device contact window to n-type material, thus removing the potential for shorting of the contact to the substrate. Due to careful selection of the respective doping levels between the p-type channel stop 23 and the p-type source-drain region 21', and the non-selective implant dose applied at the step represented by FIG. 2, the exposed portion of the channel stop is converted to n-type as desired, but the source-drain contact region remains p-type as desired.

Other elements shown in FIGS. 1-3 are essentially equivalent to those shown in the copending application referred to earlier and the description of that application is incorporated by reference herein. However, since the doping levels of certain of the regions shown in FIGS. 1-3 are especially important to the efficacy of the techniques of this invention several of the manufacturing details will be repeated here.

In FIG. 1, the substrate 10 is typically silicon having a resistivity of the order of 3-100 O cm. It is designated p- with respect to the other two regions shown, p-type channel stop 23 which is relatively more heavily doped, and p-type source-drain 21' which is more heavily doped still and is designated p+. The doping level for channel stop 23 can be 10.sup.16 -10.sup.19 /cm.sup.3, typically 10.sup.16 -10.sup.17, and the doping level for source drain region 21' can be 10.sup.18 -10.sup.21 /cm.sup.3 typically 10.sup.19 -10.sup.20 /cm.sup.3. As indicated earlier the source-drain region is more heavily doped than the channel stop to allow the latter to be converted to n-type by the non-selective n-implant without changing the conductivity type of the source-drain region. For the purposes of this invention it is recommended that this difference be of the order of factor of >10x, and preferably of the order of 100x. Referring back to FIG. 1 the conventional insulating means 11,12 and 13, i.e. field oxide 11, regrown window oxide 12 and doped or deposited glass layer 13, are typically present as shown. Doping levels for the section containing the p-channel device are conventional. The region 21' is exposed to the non-selective n-type implant which forms region 25 (FIG. 2) and which remains p-type. To insure this, a typical implant dosage for the n-type implant would be 10.sup.13 -10.sup.14 /cm.sup.2. The dose required in a given case will depend on a variety of factors all well known to those skilled in the art and it is not practical here to deal specifically with all variables. The objective is straightforward, i.e. apply a dose sufficient to convert the conductivity type of the channel stop but insufficient to convert the conductivity type of source-drain region 21'. If the initial doping difference between these regions is 10.sup.2 a practical result would be to use a dose that would produce a doping level in region 24 (FIG. 2) that is of the order of ten times that of 23 (FIG. 1) resulting in a p-type doping level in region 25 that is of the order of ten times less than that of region 21'. Typically that implant will be arsenic or phosphorus.

FIG. 3 shows the structure of FIG. 2 with metal contact layer 14 in place.

It will be evident to those skilled in the art that structures complementary to those shown here may be treated in similar fashion. It will also be evident that while this description uses as an example source-drain contacts in a CMOS structure the need for making contacts to both n-type and p-type regions using a single mask level may arise in a variety of other contexts. Although the approach described here is especially attractive if the number of one type of contact is considerably smaller than the number of the other type, it is nevertheless useful in a device where both types are equally or nearly equally represented.

The foregoing description sets forth in detail a solution to problems arising from line-on-line contacts. In the specific case illustrated both the contact and the region to be contacted are of comparable size. A directly analogous offset problem arises when the contact region is large but circumstances dictate that the contact be placed as near as possible to the edge of that region. Therefore for the purposes of this invention line-on-line is defined as if to mean edge-on-edge, i.e. the edge of one feature, e.g. the left edge of contact window 22 in FIG. 1, aligned to the left edge of diffused region 21. The remaining edge or edges may or may not be interdependent.

Various additional modifications and deviations of the techniques described herein will occur to those skilled in the art. All variations and extensions of those techniques that rely basically on the teachings through which this invention has advanced the art are properly considered to be within the scope and spirit of this invention.