Friday, December 09, 2005

NanoMetrology Improvement in Atomic Force Microscopy

12-09-2005 I will describe a novel method to improve nanoscale metrology using high aspect ratio scanning nanoprobes ( Atomic Force Microscopy )

Nanoscale metrology of very fine nano/ microfabricated features is extremely challenging. Accurate nanoscale metrology of structures formed in Integrated Circuit manufacture and development is important commercially.

Regrettably, Scanning Probe Microscopy metrology runs into what one might call an "probe artifact limited" resolution Barrier to metrology / imaging resolution of small "nanofabricated" structures, such as one finds in deep submicron Integrated Circuits and the like.

Seemingly desireable very high aspect ratio probe tips, bend flex and experience slip stick effects, which can limit usable metrology repeatablity and working "resolution" performance of the intended measurements. This occurs even with tapping mode imaging or other resonant imaging modes that are commonly associated with dramatic reductions in lateral / frictional slip stick.

The most troublesome of these artifacts arise from bending and flexing of extremely small and high aspect ratio probe tips as found with nanotube modified tips, electron beam deposited tips, or FIB - focused ion beam shaped / modified tips, where one has desirable high aspect ratio conceivably useful for imaging smaller high aspect ratio sample features, but the resultant images are often prone to undesireable step artifacts, limiting lateral dimensional metrology accuracy at the nanoscale.

This occurs even when resonant cantilever imaging modes such as Tapping ( RMS amplitude detection ) which already greatly reduces slip stick effects to a minimum, since nothing is dragging on the surface as with contact mode imaging.

CLICK READ MORE... for rest of article.

The interest in using nanotube modified like filametary like probe tips ( by modifying a more conventional microfabricated Atomic Force cantilever probe ) to attempt to scan / image / measure Integrated Circuit features often runs into a simple, yet challenging issue.

High aspect ratio "Filametary" extensions to the Apex of the probe tip, can be of usefully high aspect ratio and small enough diameter of tube/filament extension, to profile small structures such as sub 100nm wide gaps / vias / contact holes or similar.

BUT in using such a small filamentary nanotube like tip, the nanotube or other very high aspect ratio "filament" itself contributes substantially to image artifacts in the AFM image, due to the softness of the high aspect ratio slender nanotube ( or FIB shaped tip or electron beam depositied nanotip for examples), even when using resonant cantilever imaging modes.

If the end of the contacting tip moves laterally in a significant variable manner, relative to the probe cantilever body or larger conventional tip base, it can get a bit challenging to acheive desireable nanoscale accuracy in the measurement one is trying to undertake.

AFM microscopes controllably scan an XY image field pattern or some derivative ( like the predictive/adaptive scanning fields ) to form an "image data 2 D extent" not necessarily rectangular or square, but more often than not conventional 2D image XY scan field - rectangular or square.

The Z or height data of the sample being imaged, is most typically derived from constant force data obtained by servoing the probe cantilever for constant bend / force while scanning in 2 dimensions the lateral image extents.

( sample surface tracking can be done with contact mode, tapping or ostensibly "non-contact" / FM detection resonant modes - but the concept is basically the same - constant force servoing whether it be with RMS amplitude detection or phase / Frequency servoing ). Tapping type imaging tends to be preferred, with non-contact resonant imaging a useful alterative as both modes are largely free of simpler frictional lateral streaking type artifacts ( slip stick ).

One can produce constant height data from the Scanning Probe measurement and likewise lateral metrology of features with height defining edges setting the feature widths to be measured ( taking into account the ideal rigid body assumption of the probe tip shape ).
This is often called CD or critical dimension measurement in the terminology of IC fabrication.

It is often wrongly thought that this softness of the nanotube modified AFM tip, makes the potential application of nanotube modified tips nearly worthless for accurate lateral metrology, due to the buckling and bending the of the tube - mounted or grown at the probe tip apex, may experience.

But if one closely examines the actual behaviour of the slender nanotube while imaging, some key points become observable ( with a resonant cantilever probe imaging mode ).

Over relatively flat surfaces, few artifacts due to bending or buckling can be observed if the scan speed is kept to usefully low speeds - and the lateral bending can be made fairly consistent hence able to be "calibrated for" or compensated for in accurate metrology.

And obviously the probe force ( amplitude / force constant ) needs to be kept reasonably low to mitigate excessive buckling of the high aspect probe region, while scanning on relatively flat surfaces.

Usefully low forces can be obtained by lower resonance amplitudes for tapping imaging, or by using probe cantilevers of low enough force constant or combinations thereof.

Nanoscale metrology is often related to measurements of widths of fabricated structure lines ( eg. MOSFET gates, wiring interconnect for a few common examples ) or etched holes, as is found in fabricating electrical vias before interconnect wiring deposition as another interesting metrology example...observed in commercial IC fabrication.

These structures to be measured are NOT flat and do cause noticeable metrology error term artifacts when imaging with a nanotube or other high aspect ratio probe ( even seen with an FIB - focused ion beam shaped / modified probe tip, if the aspect ratio of the FIB filament tip is high enough / soft enough in lateral force constant )

Careful observation of the probe tip behavior while scanning, by understanding the image artifacts present when traversing noticeable step height topography, will see that much of the undesireable seemingly tricky behaviour of the slender high aspect ratio nanotube tip is mostly compression buckling of the nanotube when the probe is descending a step.

In contrast to the errors due to descending related buckling of the slender region of the high aspect ratio tip, when the nanotube or other slender filament is Ascending or climbing a step - the tube / filament is under TENSION and as such has pretty usefully accurate step placement needed for accurate lateral dimensional metrology.

Regrettably most all imaging control and analysis programs for Scanning Probe Microscopy (nano) metrology, do not provide any useful program code or routines to do the obvious improvement - reconstruct a synthesized image for metrology applications - of step edge position, where the image data is comprised mostly of relatively flat regions and steps.

How can a more accurate synthetic image of of AFM data using a nanotube or other high aspect ratio tip, be derived from artifact laden images that have tubes experiencing compressive buckling placement errors on descents of steps, be calculated?

Since Atomic / Scanning Probe Microscopes move the AFM probe tip cantilever typically in a raster pattern - with Trace and retrace ( as called in DI / Veeco Nanoscope instruments for example ), the probe goes back and forth across the mostly similar region at least 1 full cycle back and forth ( with slight offset in the slow scan direction between trace and retrace line data).

In traversing what we will call here the Forward Direction, most all features with have 1 side having the probe ascending ( rising ) in that "image" line data set, and the nanotube / high aspect ratio tip will be under tension on the rising edge of the step, with attendent benefits to the accuracy of the metrology of the rising step lateral placement ( low bending buckling tip artifacts ).

And the other step side of typical features ( still with the tip moving in what we are calling the "Forward scan direction" ), will have the probe tip descending "other" step, with the result that the "descending" features which the nanotube / high aspect ratio probe tip will likely experience noticable buckling related lateral errors, with somewhat random tip "placement" - very undesireable for accurate lateral nanoscale dimensional metrology.

We can generate data that is a subset of line data from each of two fast scan directions - to derive a data set which the nanotube / high aspect ratio tip is either quasi static ( nearly neutral / slight compression, or in tension), ie free of major tube/ tip related compressive buckling data errors, with the tip ascending larger steps, reconstructed by an data image line pair or two directions of scanning mostly over the same physical region, using rising tip data or flat regions taken selectively from the two opposite direction line data sets..

One determines from the first direction's data set, the compressively artifacted buckling ( descending tip ) regions - and replaces these error prone data subregions, selectively with the same physical region scanned in the opposite direction. The probe tip will have now scanned that previously descending step, in the opposite direction, by ascending the same step, eliminating or greatly reducing the magnitude of compressive buckling lateral error artifacts of the high aspect region of the modified AFM tip.

This method of scan data synthesis permits algorthmic elimination of major step descending buckling error artifacts from reducing metrology accuracy of lateral dimensions.

There are variants of the proposed method that can be applied to adaptive scanning for faster metrology data aquisition that are apparent from this innovation.

There are technical details of issue here - the opposite direction data sets have to be sufficiently well calibrated in correlation of physical data points to be of practical use. In one type of instrument this is partially accomplished with a variety of calibration constants in the scan calibration.

But the implementation of the required algorithm to derive a scan data set that is proincipally free of larger compressive buckling artifacts from a slender flexible probe descending a step is fairly straightforward to do.

One can easily envision that this can also be applied to novel scan fields - non-orthogonal and efficient scan field subsets that implement tracking scan fields and the like.

There is the possibility of applying this method of derived image data, mostly free of compressive probe tip buckling artifacts in rotated image scan fields and when the typical box averaging algorithms are employed for addressing the minor metrology error term seen with unaveraged data, where line roughness is not desired to be a contributor to the measurement (error ) of a feature width.


Post a Comment

Links to this post:

Create a Link

<< Home

Enter your email address:

Delivered by FeedBurner

Subscribe to Wendmans Views on Nanotech by Email