An adaptive Interferometer Endpoint Algorithm for Compensationof Mask Thickness in Recess Etch Applications

摘要

Accurate interferometric endpoint control for fabrication of trench capacitor structures is challenging when thernmask layer, usually silicon nitride, can vary over a wide range in thickness from wafer to wafer. This is because thernnitride thickness strongly affects, first, the target etch depth, and second, the periodicity of the optical interferencernfringes used for etch depth determination by the endpoint system. What is needed is accurate mask thicknessrninformation coupled with the ability to adaptively compensate the endpoint control algorithm for both target etchrndepth and lack of fringe purity.rnA fringed endpoint signal represents the cyclic variation of reflectance for a single wavelength through time, arnresult of continual phase change between light reflected from the etch front and the etch mask as the etchrnprogresses. The reflections from the top and bottom of the mask layer can have any phase relationship; if these twornreflections are out of phase for a particular wavelength, then the mask is said to antireflective at that wavelength. Ifrnthe reflections are in phase, then the mask is reflective. Between these extremes there is continuous variation in thernrelative phase of the top and bottom reflections, and hence in relative reflectance.rnIn this work, initial efforts identified that the periodicity of the fringe period of any single wavelength early onrnin a recess etch is strongly affected by the relative reflectance of the mask. Our efforts also made clear that nonperiodicrnfringe signals can have origins in a non-constant etch rate with time; therefore it was important tornincorporate a means of separating these two effects to achieve optimal process control. Further analysis ofrnexperimental data taken from etching samples with widely varying mask thickness allowed us to draw thernconclusion that if a fringe counting wavelength is adaptively selected such that it has a consistent relativernreflectance for the particular mask, then fringe periodicity, and hence fringe counting accuracy, can be dramaticallyrnimproved. With these data, it was possible to construct an empirical functio n relating the mask thickness to thern"best" wavelength to use for fringe counting.rnTo test the approach, an adaptive endpoint algorithm incorporating the empirical function for wavelengthrnselection was developed. Then, a single test run was made on a prepared sample set spanning a wide range of maskrnthicknesses (152-190nm). The objective of the test was to recess the capacitor electrodes of all samples to 30nmrnbelow the bottom of the etch mask. Results reported below are for a number of wafer samples etched with the samernprocess recipe and same endpoint algorithm set within a single, automated run. An advanced feature endpointrnsystem (Applied Materials EyeD-IEP) was used in conjunction with a Silicon Etch DPS? Plus Centuraò systemrnfor the test. The endpoint system consists of a broadband light source2 and CCD spectrograph2 capable of analysisrnacross a 200-800nm range. The endpoint system software utilities2 enabled the measurement of blanket filmrnthickness, storing of the resulting thickness value in memory, and finally, recalling the thickness value fromrnmemory during the recess step in order to automatically modify fringe counter attributes at the appropriate time.rnIn the test, the mask nitride was measured both before and after an oxide removal (breakthrough) step. Thernpost-breakthrough nitride thickness measurement was taken and used with the above empirical relationship torndetermine the optimal light wavelength to use for fringe counting. The optimal wavelength was stored in memoryrnalong with the mask thickness, and then retrieved later on in the recess step to set both the fringe counter attributesrnand the total etch depth target (thicker mask = larger target etch depth). The test objective, to recess to a constantrnrecess depth below the bottom of the mask, was achieved. The approach is outlined pictorially in Figure 1.Summary etch results are presented in Figure 2. Details regarding the results and approaches will be presentedrnin the full paper.