Purpose
Cross sections serve two main functions. Cuts through representative structures within an IC show relationships of layers and features, such as step coverage, interfaces between layers, and possibly embedded defects or voids. Precision cross sections through specific defects often lead to the process step or mechanism which produced the defect.
Background and Theory
When applicable, non-encapsulated cross sectioning is simpler and faster than the classical encapsulated method. Results of non-encapsulated cross sectioning are more suitable for viewing in a scanning electron microscope.
The passivation layer of the IC provides sufficient encapsulation of malleable metallization to prevent smearing. Rounding of the front edge is avoided by using hard, flat grinding and polishing surfaces. Originally, 600 grit SiC paper was used to grind to within 20 – 30 µm of the target. Polishing from that point was accomplished solely and slowly on a glass disc. (Hammond and Vogel, p 221) However, the need for faster throughput and the introduction of hard tungsten plugs necessitated new polishing media and techniques.
Today, the many possible choices for grinding and polishing media can be confusing. The non-encapsulated process is still easy to understand and execute. The procedure explained here is basic, simple and fast. It produces excellent results which can be imaged on either standard or field emission SEMs.
Equipment
Three variable speed 8 inch (200 mm) polisher/grinders are recommended for this process. One variable speed polisher, 10 – 150 RPM, is sufficient; but, frequent wheel changes with a single polisher will slow down the process. Required facilities are 120 or 220 volts power, a clean water supply, and a drain.
Dry nitrogen (or a supply of aerosol cans) is also required to blow samples dry.
A stereo microscope and a metallurgical microscope are required to check progress.
The only specialty equipment required is a polishing fixture and sample mount. The process given here is designed specifically for Accelerated Analysis’ PF101. However, with some increase in grinding times, the lighter fixtures sold by Technology Associates can be used.
Regardless of the number of polishers available, several interchangeable 8 inch (200 mm) wheels are recommended for this procedure. Suggested allocation of nylon or aluminum wheels is suggested below:
» SiC paper (600, 800, 1000, 1200 grit)
» 1.0 µm Al2O3 on Mylar (or 1.0 µm diamond on Mylar)
» 0.3 µm Al2O3 on Mylar (or 0.5 µm diamond on Mylar)
» 0.1 µm diamond on Mylar
» 0.05 µm Al2O3 on Mylar
» 0.1 µm and 0.05 µm diamond suspension on low nap cloth
Procedure
The general cross sectioning procedure has been broken down into four main steps: cutting and mounting, grinding, fine grinding / rough polishing, and polishing.
Cutting and Mounting
The result of cutting is a piece of silicon roughly 5 mm (0.25 inches) square. The desired cross section target should be within 50 µm of one edge. The silicon piece is then attached to a sample mount such that the edge containing the target is parallel to and extending over the edge of the sample mount, as shown in Fig. 1.
Figure (1) Silicon piece attached to sample mount with wax. Note: Target cross section line must extend beyond end of mount.
Variations involving diamond saws can be substituted for the cleaving technique described below. When the structure to be cross sectioned is large, the break can be into the structure itself, eliminating all grinding steps. An excellent example of using a laser to direct the break through a specific structure is shown by Gajda (p 13).
Determine the desired cross section line. Establish visible landmarks to use as navigation references. Photograph if necessary. If suitable landmarks do not exist, create visible marks with a laser or mechanical probe.
Cleave or break the wafer parallel to and within 50 µm of the desired cross section line. This first break creates the edge which will be ground and polished.
Break away a strip about 5 mm or 0.25 inches wide including the intended target. Wider pieces can be polished, but the additional area reduces effective pressure, and severely decreases removal rates.
Finally break away about 5 mm or 0.25 inches.
Heat a sample mount to 125°C. Apply a dot of wax and mount the sample in cantilever fashion on the sample mount. (See Fig. 1 above.)
Grinding
The purpose of grinding is to rapidly achieve a surface, one micron away from and parallel to, the desired cross section line. The surface to be polished should be flat and scratches should be no greater than those caused by 1 µm Al2O3 or diamond particles on Mylar. The grinding procedure typically requires less than 15 minutes including inspection time.
The procedure given below is applicable to all silicon samples, regardless of composition. However, variations in grinding media can be entirely satisfactory, as long as the final surface is flat and contains only very small scratches. The optional diamond polishing media on Mylar may cut faster, but are not better than Al2O3 unless the sample contains a high percentage of very hard material such as tungsten.
If more than 40 µm of material must be removed to reach the final cross section line, begin grinding with 600 grit SiC paper. Insert sample mount with sample into the polishing fixture. Start the wheel rotation at 60 – 100 RPM and adjust water to a low but continuous flow. Hold the polishing fixture by its sides near the back (Teflon block end) between thumb and finger. Lower the fixture and sample, Teflon end first, to the rotating wheel. The wheel rotation is into the sample. (See Fig. 2.) Allow the fixture to exert pressure of its weight onto the sample. The SiC paper will show a dark strip behind the sample as material is removed. Wait 20 – 30 seconds, then check progress.
Figure (2) Hold polishing fixture between thumb and forefinger such that wheel rotates into the sample. Allow weight of the fixture to determine pressure on the sample.
Adjust angle of the polishing line with adjustment screw on the PF101 fixture as desired. Continue until the ground surface of the sample is within 20 µm of the target line. (20 µm is about 1/5 of an IC bonding pad.) Clean the sample with running water before proceeding to the next step.
Figure (3) Rinse sample and fixture with water between each step. Grinding technique is the same for every step.
Set up the grinder with 800 grit SiC paper. Maintain wheel rotation at 60 – 100 RPM and keep slow, continuous water flow. If 600 grit SiC was used prior to this step, remove at least 10 µm of material to eliminate the subsurface damage created by the 600 grit grinding. Continue until the ground surface is within 10 µm of the target line. (See Fig. 3.) Again, adjust angle of the polishing line with adjustment screw on the PF101 fixture. Typical time for this step is 30 – 60 seconds depending on actual size of the sample.
Set up the grinder with 1200 grit SiC paper. Maintain wheel rotation at 60 – 100 RPM and keep slow, continuous water flow. If 800 grit SiC was used prior to this step, remove at least 5 µm of material to eliminate the subsurface damage created by the 800 grit grinding. Continue until the ground surface is within 6 µm of the target line. Typical time for this step is 30 – 60 seconds depending on actual size of the sample.
Set up the grinder with 1 µm Al2O3 on Mylar. Maintain wheel rotation at 60 – 100 RPM and keep slow, continuous water flow. If 1200 grit SiC was used prior to this step, remove at least 4 microns of material to eliminate the subsurface damage created by the 800 grit grinding. Continue until the ground surface is within 2 µm of the target line. Typical time for this step is 60 seconds depending on actual size of the sample.
Fine Grinding / Rough Polishing
The purpose of fine grinding / rough polishing is to remove damage due to 1 µm grinding and to enter the geometry of interest. For many, if not most applications, the finish after fine grinding is perfectly adequate to reveal details of semiconductor structure. This fact makes non-encapsulated cross sectioning less intimidating. If an error in judgement of removal rates does not permit further polishing the cross section can reasonably be declared complete after this “rough polishing” step. If the cross section is to be viewed on a standard tungsten filament or LaB6 SEM, the chemical etching to delineate features can, at the same time, remove fine details. For example, note that the procedure given by Hammond and Danyew (p 162) produces excellent results using 0.3 µm Al2O3 as the final polish. The additional step with 0.1 µm diamond on Mylar achieves a superior finish without danger of smearing soft metals.
Set up the grinder with 0.3 µm Al2O3 on Mylar. Decrease wheel rotation to 30 RPM and keep slow, continuous water flow. Remove at least 1 µm of material. Continue until the ground surface is within 2 µm of the target line. Typical time for this step is 60 seconds depending on actual size of the sample.
Set up the grinder with 0.1 µm diamond on Mylar. (Diamond is superior to Al2O3 . See NOTE below.) Again, use slow wheel rotation of 30 RPM and keep slow, continuous water flow. Continue until the ground surface is within 1.0 – 1.5 µm of the target line. Typical time for this step is 60 seconds.
NOTE: Until this step, SiC, Al2O3, or diamond could be used interchangeably with little effect on the end result. However, for this final approach into the defect or geometry of interest, diamond is preferable because of its sharp, clean cutting properties.
Polishing
The purpose of polishing is to remove all scratches and to leave the exposed cross section without artifacts from the grinding and polishing sequence. Additionally, the completed cross section should be exactly centered on the contact, via, or other feature in the target. Polishing will remove very little material (2 µm or less) and can not remove any deep scratches or damage created by grinding too close to the desired finish with course abrasives. However, polishing will remove surface scratches from the 0.1 µm diamond. Polishing can expose very subtle details otherwise lost in polishing artifacts.
Polishing is most important when the cross section is to be viewed in a high-resolution field emission SEM (FESEM). High resolution and low accelerating voltage FESEMs can image cross sections as polished or with very light delineation etches.
The details of polishing depend on the composition of the sample:
Si, SiO2, and Al with/without thin barrier metal layers
Set up the grinder/polisher with 0.05 µm Al2O3 on Mylar. Adjust wheel rotation to 30 RPM and keep slow, continuous water flow. Polish for 30 seconds or until 0.1 µm diamond scratches are removed. Repeat if necessary and material permits.
The final polishing step uses the same 0.05 µm Al2O3 set up to remove the streaks which may trail from edges of metal or polysilicon lines. For this step hold the polishing fixture between thumb and finger such that the wheel rotates into the Teflon block end of the fixture. Lower the sample to the wheel, Teflon end first, and polish for 30 seconds. (See Fig. 4.)
Figure (4) Final polish, if necessary, with fixture held in reverse direction on 0.05 µm AL2O3. (Never grind in reverse direction!)
Inspect, clean by rolling a cotton tip saturated with soapy water over the polished surface and top. Rinse with copious flow of water and blow dry.
Si, SiO2, and metallization including tungsten plugs or layers
Set up the grinder/polisher with low nap cloth. Place a circular track of 0.1 or 0.05 µm diamond suspension on a damp wheel. Adjust wheel rotation to 70 – 100 RPM. Do not use water. Polish for 30 seconds. Repeat as necessary.
Inspect, clean by rolling a cotton tip saturated with soapy water over the polished surface and top. Rinse with copious flow of water and blow dry.
General Comments
The grinding procedure given involves several short steps. Certainly, satisfactory results can be obtained by grinding longer with fewer steps. However, the procedure will be tedious if overall grind time is increased by using fine grit to remove material slowly. Additionally, if grinding time is excessive, material removal rate may decrease as media wears and be difficult to judge.
Cross sections wider than the recommended 0.25 inches width will require significantly longer grinding and polishing times.
Do not use worn out grinding media. Be aware of increased removal rate when going from worn to fresh media. Surprisingly, even fresh, finer grit polishing media may have a faster removal rate than coarser, worn media.
Even if a cleaved edge is close to the desired final line such that no grinding is required, a brief grind on 1200 grit SiC is recommended to make a flat surface. Grind only as necessary to create a flat surface, even if a portion of the surface remains untouched. Jagged edges on a cleaved sample, if not removed, will scratch away the Al2O3 or diamond.
End Notes
* JJ Gajda, FG Trudeau, JA Wade, “Semiconductor Structure Enhancement for SEM Analysis,” Proceedings of the International Symposium for Testing and Failure Analysis, 1981.
* BR Hammond and RR Danyew, “A Microfinishing Technique for Semiconductor Failure Analysis,” Proceedings of the International Symposium for Testing and Failure Analysis, 1988.
* BR Hammond and TR Vogel, “Non-encapsulated Microsectioning as a Construction and Failure Analysis Tool,” Proceedings of the 20th International Reliability Physics Symposium, 1982.
* P Kasuba, “A Technique for Achieving Ultra-Smooth Chip Cross Sections,” Proceedings of the 21st International Symposium for Testing and Failure Analysis, 1995.
Accelerated Analysis
PMB Q-310
Half Moon Bay, CA 94019
DavidBurgess@AcceleratedAnalysis.com
Phone: 650-867-8443 | Fax: 408-241-0547