Throughout this document the term alignment structure refers to any measurement mark used for identifying absolute or re- lative positions on a semiconductor wafer, e. g. alignment marks for aligning a wafer on a stage in an exposure tool or overlay targets for comparing the. deviation of two patterns in different layers or registration marks for measuring abso- lute positions on a wafer by comparison with reference data, which can originate from a library or a reference wafer.

Due to the rapidly decreasing minimum structure sizes of in- tegrated circuits, strong requirements are set to the minute exposure of a semiconductor wafer with a mask pattern. An ac- curate overlay is commonly accomplished by aligning a dedica- ted alignment mark that has been structured on the wafer sur- face in a previous exposure step, i. e. the recent layer, with another alignment mark being associated. with a mask pat- tern to be projected onto the wafer in the current step.

In an exposure tool, this alignment of the semiconductor wa- fer is performed prior to exposure by a position comparison of the structured alignment mark of the recent layer with the virtual, not yet projected alignment mark of the current lay- er.

After exposure, the accurateness of the performed alignment can be monitored in an overlay tool, where the position of associated alignment marks-both now being structured in subsequent layers-can be measured in x-and y-direction, and the difference between them then being compared with a threshold value.

Current overlay specifications require a maximum overlay to- lerance of 35-50 nm in the case of high-end products. The alignment marks, of which the position differences are measu- red during exposure alignment or overlay control, are common- ly positioned in the scribelines, or frames of each exposure field on the wafer. A common alignment procedure during expo- sure is to select a set of exposure fields already structured in the previous exposure step having a predefined number of position. In each of these, e. g. nine exposure fields, four alignment marks are measured in x-and y-direction and compa- red against their virtual counterpart through the optics.'Ha- ving determined the position differences, shift, scaling, and rotation of the current layer to be projected with respect to the previous layer is known and the position and movement of the wafer stage can be adjusted.

The information associated with the selected set of exposure fields determined for alignment is commonly tabulated and read out by the alignment control software, which automati- cally performs each of the alignment measurement steps.

A similar procedure is valid for the overlay control. In se- veral post-processing steps such as chemical-mechanical poli- shing (CMP) or etching, the alignment marks structured in a previous exposure step are often damaged or obscured, parti- cularly is some process change occurred. In these cases, the structure of the alignment marks has to be adapted to the process and changes therefore may lead deteriorated alignment marks. Thus, a sufficient alignment procedure, or overlay control, is in many instances not feasible, particularly if

the processes described above occur systematically, e. g. at the wafer edges. Therefore, the alignment quality decreases resulting in a lowered wafer yield.

To circumvent this, overlay or alignment control programs commonly flag other signals if preselected exposure fields and alignment marks cannot be detected or give two poor mea- surement signals. Unfortunately, the system then needs opera- tor input how to proceed with the present case, which consu- mes expensive time and requires larger operator staff.

It is therefore a primary objective of the present invention to decrease the time needed for an alignment measurement or overlay measurement of a semiconductor wafer and to increase the wafer yield.

The objective is solved by a method for performing an align- ment measurement of two patterns in different layers on a se- miconductor wafer, which comprises a set of exposure fields having a fixed number being selected for said alignment mea- surement, each of said exposure fields being provided with at least one alignment structure, comprising the steps of provi- ding the semiconductor wafer to a processing tool for perfor- ming said alignment measurement, selecting a first alignment structure in a first one of the set of exposure fields having a fixed number, performing an offset measurement using said first alignment structure, with issuing an error signal re- presenting the case that said offset measurement either is not feasible due to a poor alignment structure quality or provides an offset beyond a tolerance range, selecting a se- cond exposure field in response to said error signal at a substitute, which is not included in the set of exposure fields on said semiconductor wafer with the exception of the first exposure field, selecting a second alignment structure in said second exposure field, performing an offset measure- ment using said second alignment structure, and continuing with measuring a second relative offset of an alignment

structure in a next exposure field of said set of exposure fields.

According to the present invention, a different alignment structure, or alignment mark equivalently, is selected if the measurement of the position offset cannot be carried out. For example, this may be due to the obscuring or damaging effects as mentioned above. Instead of signalling an alarm to the operator for, e. g. interrupting the exposure process, a sub- stitute alignment structure is chosen, which provides a still sufficient amount of data for performing the adjustment de- termination during the alignment step, or a full characteri- sation of the overlay quality in a metrology inspection.

In order to retain the positions of the alignment mark across the wafer map as far as possible constant, the second align- ment structure selected can. be chosen from the same exposure field where the error occurred. Since the measurement of an additional alignment mark requires only a few seconds, while the overall quality is retained, a large amount of time is saved as compared with an operator intervention.

Most preferably, the alignment mark chosen from the same ex- posure field has a position in the vicinity of the damaged or obscured alignment structure. But the second alignment struc- ture may also be selected from a different exposure field, which is not part of the set of exposure fields being prese- lected. In this case, the total number of selected exposure fields is also retained holding the amount of positional in- formation constant that is necessary to perform the align- ment.

The present method is preferably implemented in overlay or alignment control programs selecting the necessary exposure fields and alignment marks positions from a stored table. If the error signal is issued which represents the problem of poor alignment structure quality or offset tolerance violai-

on, a substitute exposure field and alignment mark position which is preferably stored in the same table previously, thereby being associated with the initial fixed set of expo- sure fields and alignment marks. In such an example, each ex- posure field of the set of exposure fields has its own sub- stitute exposure field.

The present method is applicable to any alignment or overlay measurement of a process tool in semiconductor wafer manufac- turing. It is not restricted to exposure tools or overlay me- trology tools, but can also be used in, e. g. alignment procedures for other metrology tools like defect inspection, scanning electron microscopes, or processing tools needing an alignment because they affect distinct parts of the wafer surface.

The position of the second exposure field with respect to the first exposure field can advantageously be adapted to the problem that is to be expected by the damaging or obscuring process, for example it can be the next exposure field in the same row of the wafer map, or it can generally be chosen to be the next exposure field in the direction of the wafer cen- tre or vice-versa. It is also possible that the second expo- sure field is randomly chosen, thereby always considering that an exposure field already being a member of the fixed number of exposure fields being preselected must not be mea- sured twice. In a further aspect, the case that the second alignment mark also reveals problems with obscuration or da- mage is considered. According to the present invention, a third exposure field with a third alignment structure is then selected for performing the offset measurement of the align- ment marks from the different layers. The procedure of selec- ting even further exposure fields as a substitute can be con- tinued until a threshold number of substitute fields repre- senting the case that obviously any alignment mark is obscu- red, or perhaps a wrong mask pattern is projected onto the already structured wafer surface.

A method for refining the idea of the present invention con- cerning a self-learning system using the results of repeated- ly performing the method of the present invention as input to a neural network is considered. The neural network is trained using the steps according to the method of the present inven- tion. If, for example, selected exposure field repeatedly re- sult in a mismeasurement or flyer resulting in a misalign- ment, it reacts by altering the selection of substitution fields or even the predefined set of exposure fields. The sa- me is valid for the selection of the alignment marks in an exposure field. This idea can also be implemented by fuzzy logic.

Since an alignment and overlay quality can be retained or even enhanced due to the selection of exposure fields and their substitutes, the wafer yield is increased and the time spent in system maintenance or repair is advantageously redu- ced.

Further advantages and aspects are evident from the dependent claims.

The invention will be better understood by reference to the following description of embodiments of the invention taken in conjunction with the accompanying drawings, wherein figure 1 shows a first alignment structure with poor quality initiating a selection of a different alignment structure in the same field (lower left) or in another field (lower right), figure 2 shows an exposure field map of a semiconductor wa- fer with a 1 designating exposure fields with alignment marks being selected for a measurement, and a 2 or a 3 designating substitutes,

figure 3 shows a graph established for matching substitute targets with standard targets.

An alignment structure 20 embodied as an overlay target to be measured in a metrology tool, which has poor quality due to previous processing steps, is shown in the upper portion of figure 1. The alignment structure 20 is positioned in the up- per left corner of the scribeline of each exposure field 2 of a semiconductor wafer 1. In this example the alignment struc- ture 20 consists a) of 4 quadratically arranged twin-trenches formed in a previously structured deep trench layer of a 0.14 Um DRAM-pattern to form a reference target 31, and b) of four smaller equally structured bars in a resist layer, which is the gate contact layer, inside the square given in a) to pro- vide a resist target 30.

An overlay measurement, comprises measuring a deviation of the centre of the resist target 30 with respect to the already structured reference target 31 of the previously formed layer. While the resist pattern 30 is clearly visible on the SEM-image in the upper portion of figure 1, the reference target 31 of the previously structured layer reveals a poor structure contrast due to e. g. a local focus spot having oc- curred during a recent exposure or due to chemical mechanical polishing (CMP), thereby being obliterated.

If the alignment structure 20 is located an exposure field 2, which is part of the set of exposure fields 10 being selected for an alignment measurement, as shown in figure 2, the meas- urement of x-y-deviations in centre positions of the respec- tive layer patterns 30, 31 are hard to be carried out. In particular, the centre position of reference target 31 of the previously structured pattern is hard to be determined. The control unit of the metrology tool therefore generates an er- ror signal indicating that an offset measurement is not fea- sible since either the alignment structure provides poor quality or cannot be detected at all. The error signal is

then evaluated by the control program, which has access to a table comprising the position of a second alignment structure 21b in a substitute exposure field 11, which is uniquely as- sociated with the previous exposure field 10. The relative position inside the exposure field 10,11 is the same for both alignment structures 20,21b.

If particularly local phenomena like focus spots led to the problem of obliteration of the alignment structure 20, the probability that, an alignment measurement can be performed in the corresponding alignment structure 21b of an exposure field sufficiently far away is considerably larger than in the first exposure field. In the case of the alignment struc- ture 21b in the lower right corner of figure 1 both centre positions, i. e. the centre position of the resist target 30 and of the twin-trench reference target 31b, can easily be measured, and thus an accurate overlay determination can be achieved. Therefore, the inter-field change 101 in'figure 1 as performed by the metrology tool control program in re- sponse to the error signal provides a high degree of automa- tion, thereby saving time and improving product quality.

In figure 1 another embodiment of the present invention is shown in the lower left corner. The control program of the metrology tool recognizes from the error signal content, which originates from a detection of poor quality in the alignment structure 20, that the problem with the twin-trench reference target 31 is particularly due to effects of chemi- cal mechanical polishing (CMP). The problem can be circum- vented by switching from the alignment target 20 with twin- trench reference targets 31 having a structure width of 0,4 ym to another alignment structure 21a in an exposure field 11, which is essentially the same as the exposure field 10 that has a trench reference target 31a with a structure width of 1,2 Um (with a single-trench).

This may be advantageous, if larger structure sizes do not suffer such as strong as in the case of small structure sizes. In this embodiment, the second alignment structure 21a has a distance of approximately 60 hum from the first align- ment structure 20 having poor quality. The position of this second alignment structure 21a is also stored in a reference table used by the metrology tool control program to drive the optics to this position.

Thus, combined with the advantages of advanced process con- trol providing information of possible problems occurring during recent processes including CMP or exposure, the con- trol program of the metrology tool is able to determine whether to perform an inter-field change 101 or an intra- field change 100.

It is also possible to combine both changes 100,101. i. e. first to try out intra-field changes 100, and if this change provides no improvement, then second followed by an inter- field change 101.

A further improvement is achieved by applying corrections due to a measured reproducible offset between the intra-field substitute targets 21a and the standard targets 20 into the control overlay measurement. A corresponding measurement is shown in figure 3. There, an offset of-1 nm in overlay accu- racy between the gate contact alignment pattern 30a of the alignment structure 21a as a substitute and the corresponding. gate contact pattern 30 of the standard alignment structure 20 has been determined. The corresponding difference in y- direction amounts to +4 nm. These results may be used as in- put corrections for those alignment measurements, where the first alignment structure 20 cannot be measured due to poor quality and is therefore skipped in favour of the substitute alignment structure 21a. As expected, these offsets are con- siderably smaller than typical overlay specifications 200 amounting to +/-60 nm. The values provided above were meas-

ured for the same lot, where both alignment structures 20, 21a provided feasible measurements.

As is also shown in figure 2 the second substitute exposure fields 11 of the nine selected exposure fields 10 do not re- quire relative positions to the originating exposure field 10, which are arranged systematically, e. g. the next field to the right in all cases. Rather, they may individually be cho- sen by experience, or by implementing neural networks. This is particularly advantageous, if certain exposure fields re- peatedly deliver mismeasurements, or if a certain relative position of a substitute exposure field 11 repeatedly pro- vides high quality overlay measurements. The selected set may then be altered by incorporating the substitute exposure field 11 into the set, and by sorting out the exposure field 10 hitherto uniquely associated with it.

As a byproduct an operator is implicitly informed, that cer- tain exposure fields are inflicted with problems, which is due to previous processing. This helps to faster identify the cause, e. g. locally enhanced particle contamination, chuck problems, lens problems etc. of the exposure tool or any other process tool in the fab.

List of reference numerals 1 semiconductor wafer 2 exposure field 10 exposure field selected for alignment measurement, first said of exposure fields 11 substitute exposure fields, second exposure field 12 further substitute exposure field, third exposure field 20 alignment structure selected for alignment measurement 21a substitute alignment structure in same exposure field, second alignment structure 21b substitute alignment structure in other exposure field, second alignment structure 30 alignment pattern in first layer 31 alignment pattern in second layer 31a alignment pattern of substitute structure with different size 31b alignment pattern in second layer of substitute struc- ture with same size 100 intra-field change 101 inter-field change 200 overlay specification tolerance