Fabrication Test Circuit and Method for Signalling Out-of-Spec Resistance in Integrated Circuit Structure

This invention relates in general to fabrication and testing of integrated circuits and, in particular, to a test circuit and method for monitoring an integrated circuit fabrication process for possible out-of-spec resistance shift or other conductivity defect in a given current carrying structure within the integrated circuit.

Reliability and performance of integrated circuits (ICs) are significant considerations both in terms of fabrication and subsequent use of the circuits. Attempts are continually being made to ensure high reliability at all stages of the fabrication process. Many IC current carrying structures, such as contact pads, vias, and diffusion lines, can not be readily isolated for reliability and performance testing. In such cases, special test sites, for example, in the kerf region of an integrated circuit chip, are provided for reporting fabrication reliability information on one or more particular type of structure forming part of the active integrated circuitry.

As a specific example, reliability testing of silicided polysilicon field-effect transistor (FET) gates is discussed herein. Testing of such structures conventionally relies on a long, serpentine test structure representative of the polysilicon resistance (PCRS) of a large number of such gate structures. Monitoring of this serpentine structure, which may approach 1000 microns in length, can detect discrete or catastrophic type defects resulting from the fabrication process. Unfortunately, minor out-of-spec shifts in PCRS are often masked and therefore missed. This is because the conventional practice is to determine a total resistance for the serpentine structure and from that calculate an average resistance value for an individual silicided polysilicon gate by dividing the total resistance of the serpentine structure by the corresponding number of gates-whose dimensions comprise such a structure, which can be several hundred. Thus, an individual gate or even several gate structures can be out-of-spec without the resultant "average" resistance value obtained from monitoring the long serpentine test structure reflecting such a condition.

Thus, although the conventional approach works well with discrete defects such as particle or dirt induced defects, it often fails to adequately identify random process induced resistance shifts such as occurring during conversion of polysilicon to silicided polysilicon. Research has shown that silicidation of polysilicon occurs in two stages. Upon application of heat to appropriately doped polysilicon, a first type of silicide is formed having a sheet resistance of approximately 16-20 ohms per square. When this first type of silicide is again heat treated, a second type of silicide is produced having significantly lower sheet resistance, e.g., approximately 4-5 ohms per square. The process of forming the silicide, referred to in the art as transformation, is sensitive-both to overheating and to insufficient uniform heating. If overheated, the silicide will melt, thereby forming a discrete defect. Conversely, if insufficiently uniformly heated, for example, to transform the first type of silicide (16-20 ohms per square) into the second type of silicide (4-5 ohms per square), then some areas of the integrated circuit will have "out-of-spec" PCRS. Such resistance variations may degrade performance or even result in circuit failure if, for example, the out-of-spec structure is disposed in a critical timing path. Depending upon the dimensions of the silicided polysilicon gates, the probability of the first type of silicide remaining upon completion of fabrication processing can be high, resulting in a bi-modal sheet resistance distribution for the silicided polysilicon gates across the integrated circuit.

Because the difference between non-shifted PCRS and shifted PCRS is relatively small, e.g., 12-15 ohms per square, the conventional averaging approach is ineffective in identifying limited numbers of such out-of-spec structures. Thus, a need exists for a more sensitive test circuit and test method for signalling possible out-of-spec resistance shifts, or other type of defect, in a current carrying structure of an integrated circuit.

Briefly summarized, the invention comprises in a first aspect a fabrication test circuit for signalling possible out-of-spec resistance in a current carrying structure of an associated integrated circuit. The fabrication test circuit includes a plurality of monitor structures electrically connected in parallel. Each monitor structure has an associated test structure and threshold sensitive device. Each test structure is predesigned relative to the current carrying structure of the integrated circuit such that an out-of-spec resistance in the test structure signals a possible out-of-spec resistance in the current carrying structure. The threshold sensitive device of each monitor structure outputs a fail signal when a resistance proportional signal from the associated test structure exceeds a predefined level, the fail signal being representative of out-of-spec resistance in the associated test structure and possible out-of-spec resistance in the current carrying structure of the integrated circuit. Monitor means are provided for monitoring the plurality of monitor structures for a fail signal output from at least one monitor structure.

In another aspect, an automated fabrication test device is provided comprising a plurality of monitor structures and an automated count means for monitoring thereof. The monitor structures are electrically connected in parallel with each structure comprising an associated test structure and a threshold sensitive switch device. Each test structure is predesigned relative to a current carrying structure of an associated integrated circuit such that an out-of-spec resistance in the test structure signals a possible out-of-spec resistance in the current carrying structure. The threshold sensitive switch device of each monitor structure outputs a fail signal when resistance of the associated test structure is above a predefined level. The fail signal is representative of an out-of-spec resistance in the associated test structure and possible out-of-spec resistance in the current carrying structure of the associated integrated circuit. The automated count means monitors the plurality of monitor structures for a fail signal output from at least one of said monitor structures, and accumulates a total count based on each fail signal output from the monitor structures.

In still another aspect, a method is provided for fabricating a test device for indicating possible out-of-spec resistance in a current carrying structure of an integrated circuit. The method includes the steps of: predesigning a plurality of monitor structures such that each monitor structure has an associated test structure and threshold sensitive switch device, each test structure being predesigned relative to the current carrying structure of the integrated circuit such that an out-of-spec resistance in the test structure signals a possible our-of-spec resistance in the current carrying structure, the threshold sensitive switch device of each monitor structure being designed to output a fail signal when resistance of the associated test structure is above a predefined level, the fail signal being representative of an out-of-spec resistance in the associated test structure; and fabricating the plurality of monitor structures such that the monitor structures are electrically connected in parallel and are disposed adjacent to the integrated circuit.

To restate, a test circuit and test method are provided for integrated circuit fabrication monitoring which provide enhanced detection resolution, particularly of certain types of out-of-spec resistance shifts, such as PCRS shifts. Along with identifying the occurrence of out-of-spec fails in one or more monitor structures, the location thereof about the associated integrated circuit can be readily isolated. If desired, out-of-spec fails in the test circuitry can be automatically counted and a total count digitally output, along with an indication of location of failed test structures. In addition to detecting out-of-spec fails, the techniques presented can also be employed to monitor the occurrence of discrete defects in polysilicon, as well as in other types of current carrying structures within the integrated circuit. For example, the method could be used to detect high contact resistance at contact pads, or degraded conductance within a diffusion line or metal via.

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:

FIG. 1 is a schematic of one embodiment of a test circuit in accordance with the present invention;

FIG. 2 is a plot of current versus voltage characteristic for a field-effect transistor device employed in the test circuit embodiment of FIG. 1;

FIG. 3 is a schematic of another embodiment of a test circuit in accordance with the present invention;

FIG. 4 is a plot of current versus voltage characteristic for a diode device employed in the test circuit embodiment of FIG. 3;

FIG. 5 is a block diagram of an automated test device in accordance with the present invention;

FIG. 6 is a schematic of one embodiment of one detection circuit repeated thirty-two times in the PCRS detection circuits of the automated test device of FIG. 5; and

FIG. 7 is a timing diagram of the ENABLE and DATA OUTPUT signals of the automated test device of

In the embodiments discussed below, various test circuits and methods are presented for signalling possible out-of-spec resistance in a current carrying structure of an associated integrated circuit. Each test circuit and method employs a threshold sensitive device in association with a test structure predesigned to have a sheet resistance corresponding to the sheet resistance of the current carrying structure to be monitored. As a specific example, the current carrying structure is assumed to comprise one or more field-effect transistor (FET) gates composed of polysicilicon and having a silicided polysilicon sheet resistance (PCRS) to be monitored. However, those skilled in the art should recognize that the principles presented herein are equally applicable to monitoring of a continuum of resistance values, and not just to monitoring a bi-modal resistance distribution such as that of PCRS. Further, and unless otherwise apparent, the term "out-of-spec" as used herein encompasses any resistance shift or conductance degradation, for example, resulting from incomplete silicidation, as well as discrete defects.

In the embodiments presented, the test circuits include complementary metal-oxide semiconductor (CMOS) circuits with P-channel field-effect transistors (PFETs) indicated in the drawings by a rectangle with a diagonal line formed therein and a control element or gate electrode arranged adjacent thereto and with N-channel field-effect transistors (NFETs) indicated by a rectangle without a diagonal line formed therein but with a control element or gate electrode arranged adjacent thereto.

FIG. 1 presents a first embodiment of a test circuit, generally denoted 10, in accordance with the present invention. Circuit 10 includes a plurality of test structures 12 each of which has an associated non-linear, threshold sensitive device, i.e., FETs T1, T2, T3 , . . . , Tn. Each test structure and associated non-linear device comprises a monitor structure, whereby circuit 10 includes a plurality of monitor structures each having a test structure and an associated non-linear, threshold sensitive FET. As shown, each test structure 12 connects to the gate `G` of the associated FET T1, T2, T3 . . . Tn. These FETs comprise non-linear devices having commonly coupled drains `D` and commonly coupled sources `S`. The drains `D` are connected to a supply voltage Vdd, while an output test signal I.sub.test is provided through the respective FET sources `S`.

Each monitor structure is driven by a current input I.sub.in. In accordance with the invention, the input current nodes may be electrically tied together for simultaneous testing of all "n" monitor structures, or multiplexed together through a multiplexer 15 (shown in phantom). Through multiplexing, any number of test structures in the "n" monitor structures connected in parallel may be simultaneously monitored. Further, a histogram of current versus test structure resistance can be generated by sequentially multiplexing current input I.sub.in to the various nodes.

The electrically parallel-connected monitor structures are disposed between a first common node A and a second common node B. In this embodiment, node A is grounded, while node B outputs the test current I.sub.test from the various monitor structures. Each test structure 12 (R.sub.1, R.sub.2, R.sub.3 , . . . R.sub.n) is predesigned with a sheet resistance corresponding to that of the current carrying structure, e.g., FET gate, in the associated integrated circuit undergoing process monitoring.

Each test structure 12 preferably has similar if not identical dimensions to the current carrying structure being monitored. If size variation in the current carrying structure exists, then test structures 12 might have a dimension comprising to a narrowest width of the current carrying structure and an average length thereof. The narrowest width is critical since the narrower the structure, the more sensitive polysilicon is to improper silicidation. By way of specific example, for monitoring silicided polysilicon, each test structure 12 can be structured as follows: ##EQU1## wherein: R.sub.nom =nominal resistance of a test structure,

V.sub.T =threshold voltage of the associated FET,

I.sub.in =test current input to test circuit,

.rho..sub.CS4 =resistivity of the polysilicon test circuit,

l=test structure length,

t=test structure thickness, and

w=test structure width.

By so defining the nominal sheet resistance of the test structure, the field-effect transistor's subthreshold slope (discussed below) can be employed to amplify slight variations in resistance of the test structures.

FIG. 2 depicts in log scale a conventional current versus voltage characteristic for a field-effect transistor, such as NFETs T1, T2, T3, etc. Preferably, input current I.sub.in to NFETs T1, T2, T3 . . . Tn is set in the sub-threshold region of each transistor's characteristic. In the sub-threshold region, output current I.sub.test increases exponentially with small increases in control voltage V.sub.in at the gate. This significant current amplification is employed in this embodiment of the invention to "detect" an out-of-spec test structure.

In the sub-threshold operational range, good sensitivity to variations in resistance is attained. A resistance change from, for example, 4-5 ohms per square to 16-20 ohms per square will create a small voltage V.sub.in change at the gate but a significant current I.sub.test change at the source of the respective non-linear device T1, T2, T3, etc. This exponential current I.sub.test change at the FET output comprises a flag that the associated test structure is out-of-spec. Because current does increase exponentially, a single out-of-spec structure can be identified notwithstanding simultaneous monitoring of a plurality of test structures. For example, as shown in FIG. 2, in the sub-threshold range a 0.1 to 0.2 variation in gate voltage V.sub.in produces a 20-40.times. change in test current I.sub.test output from the FET.

Since silicided polysilicon sheet resistance (PCRS) has a bi-modal resistance distribution between in spec and out-of-spec, the number of test structures 12 having out-of-spec resistance can be readily determined from current I.sub.test. However, even if the structure under test has a continuum type distribution when out-of-spec, it is still possible to detect that at least one test structure is out-of-spec, and through empirical data, it may even be possible to deconvolve a test current I.sub.test output to determine within a certain probability the number of out-of-spec test structures. Typically, however, the manufacturing sector is interested in simply knowing whether or not a process has completely produced a desired structure. Thus, if only one test structure is out-of-spec, detailed analysis would typically be conducted to determine the extent and location of failure.

Ideally, the monitor structures are disposed on a semiconductor wafer in close proximity to the integrated circuit(s) 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 of an integrated circuit or conceivably, in a designated test chip on the wafer. 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. Again, a primary consideration is that the test structures be located in close proximity to the current carrying structure in order to be subject to similar process variations. If desired, the test circuit can be designed in a robust manner to be insensitive to certain process defects, yet still sensitive to PCRS problems.

FIG. 3 depicts another embodiment of a fabrication test circuit, generally denoted 20, in accordance with the present invention. As shown, multiple monitor structures are again electrically connected in parallel within the test circuit. Each monitor structure is composed of two series connected test structures 22 disposed between a first common node A' coupled to a high voltage V.sub.HIGH, and a second common node B', coupled to a low voltage. Input to circuit 20 is a test signal I.sub.test fed in parallel across respective forward-biased diodes D1, D2 . . . Dn to the common connect nodes between test structures 22 within the monitor structures.

The circuit includes a first plurality of test structures in an upper portion of the test circuit, labeled R.sub.1U, R.sub.2U . . . , R.sub.nU, and a second plurality of test structures in the lower portion of the circuit, labeled R.sub.1L, R.sub.2L . . . , R.sub.nL. Again, each test structure 22 is predesigned to have a sheet resistance equivalent to that of the current carrying structure (not shown) being monitored. Dimensional considerations for structures 22 are similar to those discussed above in connection with the test structures of FIG. 1. Diodes D1, D2 , . . . , Dn isolate the input node voltage V.sub.in from the common connect node voltages V.sub.M1, V.sub.M2 . . . V.sub.Mn. This, in this embodiment a monitor structure consists of two resistive legs R.sub.1U & R.sub.1L, R.sub.2U & R.sub.2L , . . . R.sub.nU & R.sub.nL and an associated threshold sensitive diode D1, D2 , . . . , Dn, respectively.

Operationally, the voltage V.sub.M at the common connect node within each monitor structure should equal V.sub.HIGH divided by two if the corresponding test structures 22 are both in spec. Device and voltage values are selected such when in spec that the corresponding diode (D1, D2 . . . , Dn) conducts a known amount of current I.sub.test for a given input voltage V.sub.in. If the voltage required to force a test current I.sub.test through the associated diode(s) D1, D2 , . . . Dn increases, then an out-of-spec resistance is identified.

By way of example, if three volts is applied across a series connection of resistors in one monitor structure, and if both resistors are in spec, the voltage at the common connect node V.sub.M will be 1.5 volts. Therefore, in order to drive a test current to the common connect node, an input voltage V.sub.in at the current test node of approximately 2.1 volts will be needed. In this regard, reference the typical current-voltage characteristic for a diode of FIG. 4, wherein the diode begins conducting at approximately 0.6 volts. In contrast, if one resistor of the voltage divider is out-of-spec, with a resistance of 16-20 .OMEGA.s/.quadrature., then the common connect node voltage V.sub.M will be approximately 2.5 volts and the input voltage V.sub.in necessary to drive current to the common connect node will increase to approximately 3.1 volts. These numbers assume that the individual test structures comprise silicided polysilicon having the bi-modal in spec/out-of-spec distribution discussed above. The remaining in-spec monitor structures will begin conducting current at the lower 2.1 volt level.

If desired, the input voltage V.sub.in and test current I.sub.test could be multiplexed in a manner similar to that shown in connection with the input current I.sub.in of FIG. 1. If multiplexed, then detailed information on each monitor structure can be obtained. Further, note that if simultaneously measured it is possible to determine the number of test structures that have failed, at least in a bi-modal PCRS implementation of test structures 22, but location information will require separate analysis. Advantageously, fabrication test circuit 20 of FIG. 3 is easy to implement early in the fabrication process.

In order to reduce the amount of manual testing, an automated test device is disclosed below with reference to FIGS. 5-7.

FIG. 5 depicts one embodiment of an automated fabrication test device, generally denoted 40. Principal components of device 40 include a plurality of PCRS detection circuits 42, a corresponding plurality of latches 44, and combinational counter logic 46. In the embodiment shown, thirty-two separate detection circuits and corresponding latches are assumed. Data from latches 44 is counted in combinational counter logic 46, the output of which comprises a binary signal representative of the total number of out-of-spec test structures in PCRS detection circuits 42. The device is controlled by an input "ENABLE" signal, which is buffered by an inverter 41 before being fed to PCRS detection circuits 42 and to latches 44.

One embodiment of a single detection circuit 30 is shown in FIG. 6. This circuit is implemented thirty-two times within PCRS detection circuits 42 of FIG. 5. An NFET Q1 of detection circuit 30 and silicided polysilicon resistor R.sub.PC are designed such that the voltage level at node `X` is below the threshold voltage V.sub.T of an NFET Q3 comprising part of a voltage amplifier 32, which is responsive to the voltage at node `X`. Amplifier 32 also includes a PFET Q2 connected as shown with the output node `Y` at the commonly coupled drains `D` of NFET Q3 and PFET Q2. NFET Q3 is sized larger than PFET Q2 so that it will pull node `Y` low when turned on. The signal at node `Y` is fed to a buffer 34 comprising a PFET Q4 and an NFET Q5 before being output to latches 44 (FIG. 5).

Operationally, the voltage at node `X` is maintained below a threshold voltage V.sub.T of NFET Q3 when silicided polysilicon resistance R.sub.PC is in spec, and hence the output voltage at node `Y` remains high. If there is a defect or process variation resulting in an out-of-spec resistance at R.sub.PC sufficient, for example, to cause PCRS R.sub.PC to increase by about 4 times, node `X` voltage will increase above threshold voltage V.sub.T of NFET Q3, thereby pulling voltage at node `Y` low. (Again, note that device Q2 is designed to be small relative to device Q3 so that NFET Q3 can pull down node `Y` whenever node `X` voltage is higher than the threshold voltage V.sub.T of that device.) The output voltage at node `Y` is buffered by an inverter 34 before being fed to latches 44 (FIG. 5).

As numbered in FIG. 5, latches 44 also comprises thirty-two latches implemented in parallel, each latch receiving an output from a respective detection circuit of the PCRS detection circuits 42. By using multiple PCRS detection circuits and latches, simultaneous PCRS detection is accomplished simply by bringing the "ENABLE" signal high. After a specified hold time the number of PCRS fails can be read in binary number through the use of combinational counter circuit 46 connected to receive the output of latches 44. Note that the five bit binary output from logic 46 can represent a total number of fails or, alternatively, the location of a single failed test structure within the thirty-two detection circuits of the device, i.e., assuming that the "ENABLE" signal is multiplexed through the detection circuits.

FIG. 7 is a timing diagram showing the brief delay interval between activation of the "ENABLE" signal and the occurrence of valid data at the data output of the device. The time delay, for example, less than 20 nanoseconds, between the "ENABLE" signal going high and the latching of valid data will vary with technology.

As an enhancement, a fabrication test circuit may be disposed about a given integrated circuit on a wafer having multiple fabrication test circuits present, each of which monitors a different integrated circuit on the wafer. Thus, resistance monitoring can be readily accomplished by separately testing the fabrication test devices for a total number of failures on a wafer, and if present, the general location of any fails on the wafer will be known by the location of the test circuit signalling the failure. As an example, 1000 or more test structures may be employed on a wafer to provide a sufficient level of confidence in a fabrication process.

From the above discussion, those skilled in the art will note that a test circuit and test method have been provided for integrated circuit fabrication monitoring which have enhanced detection resolution, particularly of certain types of out-of-spec resistance shifts, such as PCRS shifts. Along with identifying the occurrence of out-of-spec fails in one or more monitor structures, the location thereof about the associated integrated circuit can be readily isolated. If desired, out-of-spec fails in the test circuitry can be automatically counted and a total count digitally output, along with an indication of location of failed test structures. In addition to detecting out-of-spec fails, the techniques presented can also be employed to monitor the occurrence of discrete defects in polysilicon, as well as in other types of current carrying structures within the integrated circuit. For example, the method could be used to detect high contact resistance at contact pads, or degraded conductance within a diffusion line or metal via. Currently, testing contact pad resistance is similar to the approach described initially herein for screening silicide resistance problems.

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. 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.