Self-Aligned Silicide Process

The invention relates to electronic semiconductor devices, and, more particularly, to fabrication methods for such devices.

Semiconductor integrated circuits with high device density require minimum size structures such as short gates and shallow source/drain junctions for field effect transistors ("FET"s), small area emitters for bipolar transistors, and narrow interconnection lines between devices. Short gates and narrow interconnection lines lead to higher resistance for the typical polysilicon gates and first level interconnection lines, and this increases RC time delays and slows circuit operation.

One approach to reduce the resistance of polysilicon gates and interconnection lines uses a self-aligned titanium silicide on top of the polysilicon. The silicidation process first deposits a blanket film of titanium metal and then reacts the titanium with any underlying silicon in a nitrogen atmosphere. Lastly, the process removes the titanium nitride formed from the titanium which did not become silicide. For example, see U.S. Pat. No. 4,821,085.

The removal of titanium nitride in this self-aligned silicide process requires a very high selectivity with respect to removal of the titanium silicide, otherwise the titanium silicide will be thinned and lose the advantages of siliciding. The removal of titanium nitride with the standard basic SCl solution (12.5% ammonium hydroxide solution, 12.5% hydrogen peroxide solution, and 75% water) at room temperature has a selectivity with respect to titanium silicide of roughly 7.8 to 1. This low selectively presents problems in the known silicide processes.

The present invention provides a highly selective etch of titanium nitride and titanium metal with respect to titanium silicide by the use of a solution with a large hydrogen peroxide to ammonium hydroxide ratio.

Further, the present invention provides a self-aligned silicide process which selectively removes titanium nitride with respect to titanium silicide with a hydrogen peroxide solution and then removes titanium silicide filaments after a phase conversion anneal with a hydrogen peroxide plus ammonium hydroxide solution.

Advantages of the invention include a simple method for removal of titanium nitride with high selectivity with respect to titanium silicide and a simple self-aligned silicide process.

The drawings are heuristic for clarity.

FIGS. 1a-1d illustrate in cross sectional elevation views steps of a first preferred embodiment method.

FIG. 2 is a component diagram showing various solutions.

The preferred embodiment methods strip titanium nitride (TiN) and titanium metal (Ti) selectivity (20-to-1) with respect to titanium silicide (TiSi.sub.2) with essentially a 30% hydrogen peroxide (H.sub.2 O.sub.2) solution in water. Stripping with solution temperatures of about 55.degree. C. gives practical etch rates.

Some etching of TiSi.sub.2 may be desired, such as removal of unwanted filaments which form during the siliciding process. Some preferred embodiments thus include a second strip with a solution containing both H.sub.2 O.sub.2 and ammonium hydroxide (NH.sub.4 OH) to be applied after the anneal which converts small grain C49 phase TiSi.sub.2 to large grain C54 phase TiSi.sub.2. Narrow filaments convert very slowly and the solution selectively strips C49 phase with respect to C54 phase.

FIGS. 1a-d illustrate in cross sectional elevation views the first preferred embodiment self-aligned titanium silicide method which includes the following steps.

(1) Start with silicon substrate 102 having polysilicon gate 104 on gate insulator silicon dioxide (gate oxide) 114 with gate 104 having sidewall oxide 124 for alignment of source/drains 134 and with field oxide 120 for transistor isolation. Also, polysilicon interconnect line 106 with sidewall oxide 126 may run over field oxide 120. Typically, the polysilicon gates and the field oxide each extend about 0.3 .mu.m above the substrate surface, so the top of poly interconnect 108 lies about 0.6 .mu.m above the substrate surface. Gate oxide 114 may be about 6 nm thick and source/drains 134 about 0.15 .mu.m deep. See FIG. 1a which illustrates a gate length and poly interconnect width of roughly 0.5 .mu.m.

(2) Sputter deposit titanium 150 to a thickness of about 60 nm. Titanium 150 coverage on the sloping sidewall oxides 124-126 and field oxide 120 is not crucial because this portion of titanium 150 will not form TiSi.sub.2 and will be stripped. See FIG. 1b.

(3) Insert the titanium coated substrate into a nitrogen (plus optional hydrogen) atmosphere and raise the temperature to roughly 600.degree. C. to drive the silicidation reaction for the portion of titanium 150 abutting silicon; namely, the surfaces of gate 104, poly interconnect 106, and source/drain 134:

Silicon diffuses through TiSi.sub.2, so the TiSi.sub.2 grows from gate 104, poly 106, and source/drain 134. (This TiSi.sub.2 is the metastable C49 phase and will be converted to the stable C54 /phase in a later anneal.) A competing reaction with the nitrogen from the atmosphere diffusing into the titanium limits the TiSi.sub.2 growth:

The titanium on oxides 120, 124, and 126 only slightly reduces the abutting oxide before nitrogen diffusing from the atmosphere will convert the titanium to TiN. Thus only a small amount of titanium oxide forms:

And some titanium remains unreacted as Ti metal. See FIG. 1c showing TiSi.sub.2 152 formed on source/drain 134, TiSi.sub.2 154-156 formed on gate 104 and poly 106, and TiN 160-164-166 formed on oxides 120-124-126. The top surfaces of TiSi.sub.2 152-154-156 typically contain some TiN which forms prior to the diffusing silicon arriving to form silicide. The TiSi.sub.2 will be roughly 80 nm thick and the TiN roughly 60 nm thick. Thus the selectivity of only 7.8 to 1 for SCl stripping of the TiN with a 100% overetch would remove roughly 15 nm of TiSi.sub.2, or about 20%.

(4) Immerse the silicided/nitrided substrate in a solution of 30% H.sub.2 O.sub.2 and 70% H.sub.2 O at 55.degree. C. for roughly 15 minutes. This solution has an etch rate for TiN of about 8 nm/min, so the 15 minute etch equals a 100% overetch. The selectivity with respect to TiSi.sub.2 roughly equals 35 to 1 (a TiSi.sub.2 etch rate of roughly 0.25 nm/min), so the TiN strip removes roughly 4 nm of TiSi.sub.2, or 5%. H.sub.2 O.sub.2 also strips any residual Ti metal and thin TiO.sub.2 on oxides 120-124-126. See FIG. 1d.

(5) Lastly, rinse the silicided substrate clean and then anneal in a forming gas (nitrogen plus hydrogen) atmosphere at about 750.degree. C. to convert the TiSi.sub.2 from metastable C49 phase to stable C54 phase and thereby lower the resistivity from roughly 75 microohm-cm to roughly 15 microohm-cm. This completes the siliciding, and further steps of forming metal level interconnections, passivation, wire bonding, and packaging complete the integrated circuit.

The preferred embodiment stripping solution may be modified while retaining high selectivity of TiN with respect to TiSi.sub.2. In particular, FIG. 2 is a component diagram graphically showing the portions of each of the three components NH.sub.4 OH, H.sub.2 O.sub.2,and H.sub.2 O) of possible stripping solutions. Recall that NH.sub.4 OH is typically provided as a 29% by weight solution with water (also expressed as 14% by weight NH.sub.3) and that H.sub.2 O.sub.2 is typically provided as a 30% by weight solution with water. Thus the point labelled SCl in FIG. 2 depicts a solution with a volume ratio of 1:1:6 of 29% NH.sub.4 OH solution to 30% H.sub.2 O.sub.2 solution to H.sub.2 O. Similarly, the other data points in FIG. 2 have labels reflecting their volume ratios of 29% NH.sub.4 OH solution to 30% H.sub.2 O.sub.2 solution to H.sub.2 O, respectively. The following table shows the TiN etch rate and selectivity with respect to TiSi.sub.2 (C49 phase) of these solutions.

The first two entries show a replacement of almost all of the NH.sub.4 OH by H.sub.2 O.sub.2 in SCl at room temperature decreases the TiN etch rate but increases selectivity somewhat. The second and third entries indicate a temperature independent selectivity but increased TiN etch rate for the NH.sub.4 OH mostly replaced by H.sub.2 O.sub.2 in SCl. The third and fourth entries show increasing H.sub.2 O.sub.2 (and minimal NH.sub.4 OH) concentration increases TiN etch rate plus nearly doubles the selectivity. The fourth and fifth entries shows a small temperature increase in selectivity but large increase in TiN etch rate. And the fourth plus sixth entries illustrate elimination of all NH.sub.4 OH decreases TiN etch rate by a factor of more than 3 but almost doubles selectivity. In summary, limiting or omitting the metal-complexing ammonium ion provides a tradeoff of decreased etch rates of both TiN and TiSi.sub.2 but increased selectivity of etching TiN over TiSi.sub.2. And a temperature increase can compensate for the decrease in etch rate of TiN. Thus a selectivity can be picked and the etch rate adjusted by temperature. For example, replacing almost all of the ammonia of SCl with peroxide (i.e., the 1:200:600 solution) yields better than a 10 to 1 selectivity, and then removing water from this solution (i.e., the 1:200:0 solution) raises the selectivity to better than 20 to 1 at 55.degree. C.

A second preferred embodiment method follows steps (1)-(5) of the first preferred embodiment but adds a TiSi.sub.2 filament strip with SCl after the annealling in step (5). That is, anneal to convert TiSi.sub.2 from higher resistivity metastable C49 phase to lower resistivity stable C54 phase, and then strip unconverted TiSi.sub.2 remaining in the C49 phase by immersing in a solution of SCl at room temperature. SCl readily etches C49 phase (about 0.73 nm/min) but does not detectibly etch the C54 phase TiSi.sub.2. TiSi.sub.2 152-154-156 on the gates, lines, and source/drains will generally be C54 phase, but narrow filaments (e.g., filament 158 in FIG. 1d) of TiSi.sub.2 on oxides 120, 124, or 126 would be too narrow to convert to C54 phase during the anneal and remain as C49 phase which is etched away by SCl in a few minutes. Indeed, the C49 grains may average about 0.1-0.2 .mu.m diameter; and the phase conversion time (time to convert half from C49 to C54) roughly depends upon g.sup.2 /(w-g) where g is the grain size and w is the filament width. So narrow filaments have a very long conversion time.

The delay of the TiSi.sub.2 filament etch until after the phase change anneal (rather than at the same time as the TiN strip) allows the use of the highly selective H.sub.2 O.sub.2 TiN strip and consequent minimal loss of TiSi.sub.2 on the gates, lines, and source/drains.

Because the selectivity of etching TiN relative to TiSi.sub.2 increases with decreasing NH.sub.4 OH and the etch rate increases with increasing H.sub.2 O.sub.2 concentration, the third preferred embodiment uses a solution of greater than 30% H.sub.2 O.sub.2 without any NH.sub.4 OH. Indeed, H.sub.2 O.sub.2 is a liquid with melting point about 0.degree. C. and boiling point about 150.degree. C. and miscible with water, so any concentration can be achieved.

The preferred embodiments may be varied in many ways while retaining one or more of the features of a TiN strip selective to TiSi.sub.2.

For example, the temperatures and concentrations of the strip solutions may be varied provided the desired selectivity is maintained, such as at least 10 to 1 or at least 20 to 1 in the titanium nitride strip; the selectivity of C49 phase to C54 phase etching is easily maintained. Further, the thicknesses and widths of the titanium layer and resulting reaction product TiN and TiSi.sub.2 may be varied; the reaction temperature may be varied and include rapid thermal annealing; the phase conversion temperature may be vaired.