Test Device and Method for Signalling Metal Failure of Semiconductor Wafer

This invention relates in general to fabrication and testing of interconnection networks for integrated circuits and, in particular, to devices and methods for testing a semiconductor wafer for possible metal failure of integrated circuit interconnect thereon.

Reliability of integrated circuits is a significant consideration both in terms of fabrication and subsequent use. Attempts are continually being made to ensure high reliability at all stages of the fabrication process. For example, where feasible electromigration test structures are incorporated on semiconductor wafers with the formation of integrated circuits therein. Each integrated circuit can then individually be electrically bonded to the test structures to determine whether electromigration may cause failure of the integrated circuit.

Very-large-scale integration (VLSI) aluminum-based metallizations can be subject to failure due to various non-electrical forces, both during fabrication and more gradually, in field use. As presented herein, these two types of non-electrical failures are respectively referenced as a "quality" problem and a "reliability" problem. The present invention is particularly advantageous in connection with addressing the "quality" issue. This is because fabrication-resultant failures must necessarily be identified from a single test during or subsequent to semiconductor fabrication.

Today's metallization practice is to layer aluminum (Al) interconnect with thin films of refractory materials, such as titanium or tungsten, so that if a void in the aluminum forms, electrical continuity through the interconnect is maintained, albeit at increased resistance. Unfortunately, the increased resistance is typically insufficient to identify the failed interconnect conductor. This is because the additional resistance due to the aluminum void is often insufficient to distinguish the interconnect line from a normal resistance sampling after accounting for measurement variations.

The principal solution presently in use to identify fabrication-type metal failures is to perform a physical construction analysis of interconnect conductors of interest. For example, conductors can be physically delayered and examined with a scanning electron microscope. Such an approach, however, is clearly time consuming and destructive.

Another approach to detecting metal failures during fabrication is to employ laser technology to evaluate reflectivity responses of the interconnect conductors. Unfortunately, such an approach has proven difficult to calibrate, has limitations concerning distinguishing voids in narrow lines and near tungsten interconnect studs, and is clearly more laborious than a direct electrical measurement approach.

A further option might be the use of some variation of "burn-out testing," described in the literature as "SWEAT" (Standard Wafer-level Electromigration Acceleration Test). This test provides for passing a much larger current than used for the "long-term" test through a test structure that is optimized for the accelerated test. The large current provides for joule heating of the test conductor only, which avoids the necessity for heating the entire wafer. The SWEAT technique is described in "Wafer Level Electromigration Test For A Production Monitoring," B. J. Root et al., in the Proceedings of the International Reliability Physics Symposium, pp. 100-107 (1985). The disadvantage to this technique is that the tests are time consuming and destructive to the tested structure, and would be less directly correlatable to non-electrical metal failures such as the metal voids principally addressed herein.

A still further option would be to employ electrical measurements more sensitive to metallization degradation than simple resistance measurements, for example, 1/frequency noise. The method of applying such techniques to wafer monitoring has been described in European Patent document 0484013 A2, entitled "Method for Manufacturing An Integrated Circuit". The drawbacks to this approach, however, are the questionable adaptability of the technique in practice, the sophisticated test equipment required to perform the technique, and the long test time required.

Multiple "reliability" type measurements are less problematic because precise measurements as a function of time can be compared to previous measurements to identify resistance increases resulting, for example, from metal failure. Even in such cases, however, other factors such as temperature variations can confound measurements. The devices and methods described herein assist in minimizing these confounding affects and, thus, in obtaining a more accurate indication of metal failure growth for evaluating "reliability".

Thus, a need exists in the integrated circuit fabrication art for a simple, reliable technique for electrically identifying a semiconductor wafer as possibly having non-electrically induced, metal failure in the integrated circuit interconnect on the wafer.

Briefly described, the invention comprises in one aspect a test device for signalling possible metal failure of a semiconductor wafer. The test device includes a metal monitor structure and a metal control structure. The metal monitor structure is prone to metal failure, while the metal control structure is designed to resist metal failure. The structures are further designed with identical resistances with no metal failure of the metal monitor structure and a resistance difference upon metal failure of the metal monitor structure. The resistance difference signals possible metal failure at an active region of the semiconductor wafer.

In a specific embodiment, the metal monitor structure of the test device comprises a first plurality of metal segments electrically connected together, with at least one metal segment of the first plurality of metal segments prone to metal failure. The control structure comprises a second plurality of metal segments electrically connected together such that no metal segment of the second plurality of metal segments is prone to metal failure. Again, the monitor structure and the control structure have substantially identical resistances without metal failure of the at least one metal segment prone to metal failure, and a resistance difference upon metal failure of the at least one segment prone to metal failure. This resistance difference signals possible metal failure above an active region of the semiconductor wafer.

As a further enhancement, a plurality of test devices, each structured as outlined above, can be disposed in an inactive region of the semiconductor wafer. If desired, different test devices can be configured to experience different mechanical forces.

In still another aspect, a method is provided for fabricating a test device for flagging possible metal failure on a semiconductor wafer. The method includes the steps of: predesigning a metal monitor structure and a metal control structure such that the metal monitor structure is prone to metal failure and the metal control structure is resistant to metal failure, and such that the metal monitor structure and the metal control structure have substantially identical resistances without metal failure of the monitor structure and a resistance difference upon metal failure of the monitor structure, the resistance difference indicating possible metal failure on the semiconductor wafer; and fabricating the metal monitor structure and the metal control structure at an inactive region of the semiconductor wafer.

In all aspects, the present invention embodies a simple electrical technique for monitoring non-electrically induced metal failure to assure quality and reliability of integrated circuit interconnect wiring against metal failures, such as metal voids. The technique can employ various different device configurations according to numerous factors, such as space available, measurement sensitivity, extent of voiding deemed critical, etc. The monitor and control structures of the test device can be designed in series with a voltage tap in the middle to allow, for example, Kelvin measurements of both structures with a common current source. Further, a series of monitors and control structures could also be implemented to increase the likelihood of detecting voids. An optimization of these options is possible based on the assumptions of void nucleation and resistance increase statistics.

These and other objects, advantages and features of the present invention will be more readily understood from the following detailed description of certain preferred embodiments of the invention, when considered in conjunction with the accompanying drawings in which:

FIGS. 1a & 1b are elevational and plan views, respectively, of one embodiment of a metal monitor structure in accordance with the present invention, shown implemented in two metallization levels;

FIGS. 2a & 2b are elevational and plan views, respectively, of one embodiment of a metal control structure in accordance with the present invention, again shown implemented in two metallization levels;

FIGS. 3a & 3b are elevational and plan views, respectively, of an alternate embodiment of a metal monitor structure in accordance with the present invention, shown implemented in three metallization layers;

FIG. 4 is an elevational view of an alternate embodiment of a metal control structure for use in connection with monitoring two metallization levels simultaneously;

FIG. 5 is an elevational view of an alternate embodiment of a portion of a metal monitor structure in accordance with the present invention for use in connection with monitoring two metallization levels simultaneously; and

FIG. 6 is a partial plan view of a semiconductor wafer with multiple integrated circuits, and shown with multiple test devices in accordance with the present invention disposed therebetween.

As used herein "metal failure" encompasses any degradation in interconnect conductance of metallization layers of an integrated circuit. In addition to metal voiding, the phrase "metal failure" encompasses, e.g., any constriction of metal interconnect that reduces conductance of the interconnect.

Metal failure detection in accordance with this invention is based on certain empirical observations regarding the occurrence of metal interconnect quality problems during fabrication. A first observation is that metal failure of a semiconductor wafer sporadically occurs and that voiding is present only on certain wafers. Further, when present, voiding typically occurs at multiple locations throughout the wafer. Thus, a test structure can be created for the wafer to signal or flag a wafer having a high probability that metal failure has occurred at other locations on the wafer. Another observation is that metal failure through voiding principally occurs in the aluminum layer of today's composite interconnect wiring, and in particular, in aluminum wiring segments extending beyond a certain minimum length. Closed lengths of short aluminum metal have not been observed to accommodate non-electrical formation of metal voids. Physical evaluation has shown that the typical metal void varies with grain size and length of metal conductors, but generally has the same width as the line conductor and a length of approximately 1 micron.

The critical minimum length of a metal segment below which voiding will not occur can be empirically determined by one of ordinary skill in the art. Non-electrical voiding is believed to be thermally induced during fabrication and may occur from a merging of submicroscopic vacancies within a metal segment. For example, after metallization of a semiconductor wafer, temperatures in the range of 450.degree.-500.degree. C. may be sufficient to initiate metal failure. Again, it is important to the present invention to note that if a metal segment is small enough, there are believed to be insufficient vacancies in the metal structure which can merge together to form a void large enough to cause a perceptible resistance increase. Thus, lengths of segments shorter than the empirically determined minimum length are referred to herein as "resistant" to failure, while those longer than this minimum length are referred to as being "prone" to failure.

In one aspect, this invention comprises a test device consisting of a pair of test structures, referred to as a "monitor structure" and a "control structure," each of which has a plurality of the metal segments. Variations in the lengths of metal segments are employed such that electrical measurements can be made with high sensitivity to the presence of metal failure in the test device. FIGS. 1a & 1b depict one embodiment of a two metallization level (M1, M2) monitor structure in accordance with the present invention. In this embodiment, length `A` denotes a short M1 segment, length `B` a long M1 segment, and length `C` a standard M2 segment. Segment `A` length and segment `C` length are chosen sufficiently short to resist voiding within the segment.

Segment `C` can be chosen as small as possible since the purpose of transitioning between metallization levels is to form discrete, closed metal segments of controlled length. Metal studs 20 interconnect the metal segments on the two metallization levels. As one example, the studs may comprise tungsten studs. In this example, segment `B` is chosen from empirical data sufficiently long to be prone to metal failure.

As is common with today's interconnect, each metal segment is assumed to comprise a thin film composition of conductive materials. As one specific example, each segment can principally comprise a layer of aluminum metal, along with a single layer of refractory material, such as titanium or tungsten disposed underneath the aluminum layer. A wide variety of alternate composition configurations are possible without departing from the scope of the present invention. For example, a second layer of refractory material could be disposed above the aluminum layer. Further, metal studs 20 could be removed such that the metallization levels physically contact. The only qualification to such a structure would be that the aluminum layers of different segments remain physically isolated such that the length of the aluminum layers in the various segments is known.

FIGS. 2a & 2b depict one example of a control structure in accordance with the present invention, which is similar to the monitor structure presented in FIGS. 1a & 1b. Principal differences are that segment length `D` is somewhat longer than segment length `A` in the monitor structure, and that the control structure lacks a long segment, such as segment `B` of FIGS. 1a & 1b. Segment length `E` of metallization level M2 of the control structure may, if desired, be the same length as segment length `C` of the monitor structure. As one specific example, segments `A`, `B`, `C`, `D` and `E` may respectively have lengths of 5 microns, 510 microns, 2 microns, 10 microns, and 2 microns. Again, the second metallization level allows fabrication of closed metal segments of controlled length. Metal vias or studs 20, e.g., fabricated of tungsten, electrically interconnect the metallization levels. The thin film composition of the metal segments in the control structure is the same as that of the monitor structure.

The control structure and the monitor structure are predesigned such that at designated measurement pads, there is an identical resistance between the two structures absent metal failure in that segment of the monitor structure prone to metal failure, e.g., segment `B` in the monitor structure of FIGS. 1a & 1b. This is achieved by equating the total length of metal segments in the control structure with the total length of metal segments in the monitor structure, in addition to equating the total length of metal segments in respective metallization levels. Further, equal numbers of transitions are employed between metallization levels, such that an equal number of metal studs (if present) are encountered in each structure.

Resistance can be accurately measured via a four point (Kelvin contact) probe which could connect to the monitor and control structures as shown in FIGS. 1b & 2b. The applied force is assumed to comprise a constant current force that is applied to the monitor Structure and to the control structure. Voltage would be sensed across each structure as shown. From the known current and voltage, accurate resistance values are thus readily obtained.

Ideally, the monitor and control structures are disposed on a semiconductor wafer in close proximity so that they experience the same process variations and mechanical forces. The test structures can be disposed over inactive portions of the wafer, for example, in a kerf region or in a designated test chip on the wafer. FIG. 6 depicts a portion of a semiconductor wafer wherein various monitor and control structures 30 & 32, respectively, are shown disposed in the kerf region 40 between integrated circuit chips 42. Structures 30 & 32 are located in various arrangements. A common consideration in each arrangement is the disposition of the monitor structure in close proximity to the control structure. As a further variation, the test structures could be disposed over active chip portions of the wafer, i.e., assuming that there is available space.

By way of example, the monitor and control structures are shown in parallel, in series, and in spaced series alignment. If desired, a group of monitor and control structure pairings can be implemented to increase the likelihood of accurately flagging metal failure on a wafer. For example, a first monitor and control structure pairing can be constructed for signalling metallization level M1 failures, while a second monitor and control structure pairing can be fabricated for detecting metallization level M2 failures, etc. Further, the configuration of the monitor and control structures themselves can vary, e.g., the individual segment lengths and number of segments can change, according to various factors such as space available, measurement sensitivity desired, extent of metal voiding deemed critical, etc.

As noted, various monitor and control structure configurations may be desirable to simulate different physical forces on the structures. FIGS. 3a & 3b depict a modified monitor structure wherein a third metallization level M3 crosses and physically and electrically connects to second metallization level M2. The metal segment prone to failure in this monitor structure is assumed to comprise segment length `F` in metallization level M2. In this embodiment, a downward physical force could be generated in the region of the tungsten studs interconnecting metallization level M3 and metallization level M2, particularly with thermal cycling of the wafer. As shown, such a force is directly applied to that metal segment of the monitor structure having a predesigned disposition to failure.

As a further variation, a single monitor and control structure pairing could be configured with multiple metal segments of the monitor structure prone to failure. For example, a monitor structure may be fabricated with a first metal segment prone to failure in metallization level M1 and a second metal segment prone to failure in metallization level M2. Assuming that the length of the two segments prone to failure are the same, then a control structure such as depicted in FIG. 4 may be employed. Specifically, the metal segment length `G` is the same in both metallization levels. Again, segment length `G` is sufficiently short so as to resist metal failure. FIG. 5 depicts one embodiment of the short metal segments of the corresponding monitor structure. As shown, segment length `H` is the same in both metallization level M1 and metallization level M2. The two metal segments prone to metal failure are omitted from the figure.

To summarize, the present invention comprises test structures and a simple electrical technique for flagging possible metal failure of a wafer to assure quality and reliability of integrated circuit interconnect wiring. The technique can be implemented in various different structure configurations according to numerous factors, such as space available, measurement sensitivity, extent of voiding deemed critical, etc. The monitor and control structures could be designed in series with a voltage tap in the middle to allow, for example, for Kelvin measurements of both structures with a common current source. A grouping or a series of monitors and control structures could be implemented to increase the likelihood of accurately signalling voids. An optimization of the various options is possible based on the assumptions of void nucleation and resistance increase statistics.

While the invention has been described in detail herein in accordance with certain preferred embodiments thereof, many modifications and changes therein may be effected by those skilled in the art. For example, the monitor and control structures presented herein could be readily adapted for use in a Wheatstone bridge. Accordingly, it is intended by the appended claims to cover all such modifications and changes as fall within the true spirit and scope of the invention.