Mechanical probing, long a mainstay of semiconductor device failure analysis, has been limited by the need to position a probe using an optical microscope. As semiconductor line widths shrink below the optical microscope’s resolution limit, positioning probes goes from painfully tedious to nearly impossible.
Using mechanical probes in conjunction with ion-beam milling to perform surgery on ailing microcircuits also forces you into lengthy cycles that involve moving the device or wafer back and forth between the prober and the focused ion-beam (FIB) system. Yet, mechanical probing has advantages that the failure analyst just can’t live without.
The ideal solution is to move the mechanical prober into the FIB vacuum chamber and use the ion beam in the imaging mode (when the ion-beam current is so low that it removes an insignificant amount of material) to position the probes. This integration of the prober and FIB is particularly advantageous since complex, advanced-technology ICs already require FIB milling of holes through overlying passivation and metallization layers to reach signals buried beneath.
Ordinary mechanical probing stages also have limited resolution, making them difficult to use in probing submicron ICs. An advanced high-resolution motorized stage, such as is needed for FIB positioning, is required.
This new probing concept allows the FIB to make up the deficiencies of optical microscopes and the prober to help solve some difficulties for the FIB system as well. The two techniques combine to create a technology that works better than either alone or in tandem, making it possible to mechanically probe advanced, high-value submicron ICs.
The Trouble With Mechanical Probing
When integrated e-beam probers were introduced in 1987, the demise of mechanical probing for analyzing devices was predicted. Mechanical probing was tedious and required that you peer through an optical microscope to align needlelike mechanical probes with their targets, then carefully touch the probe tip down onto the target.
The process was tedious because, with optical microscopes of the required magnification, the depth of focus is on the order of a micron or even smaller. This means you cannot keep both the probe tip and the target in focus at the same time.
Positioning a probe starts with focusing the microscope on the device surface at low magnification. Then you have to find the area where the feature of interest lies and position that area at the center of the field of view. Next, you shift to higher magnification (so you can actually see the feature you want to probe), refocus on that feature, and recenter it in the field with the greater precision afforded by the higher magnification.
Now you have to find the probe tip. You can’t just move it under the microscope lens because, with the microscope focused on the feature of interest, the probe tip is no more than a diffused shadow. You have to back the focus plane up to the level of the probe tip. Then you move the probe tip into position, focus on it, and carefully position it in the center of the field.
When you’re convinced that the alignment is as perfect as you can make it, you refocus on the target feature, making sure that it is still centered, and back the focus up slightly. The focus plane is a few microns above the IC surface. You can carefully lower the needle until it comes into focus just above the surface.
Finally, you refocus on the surface and gently put the needle down. If everything is just right, the probe-tip image will begin to come into focus as the tip touches down right where you want it.
But the probe will never come completely into focus. Because the probe’s physical dimensions are larger than the microscope’s depth of field, the only part that will be in focus is the point where it actually touches the surface. Then you start over again with the next probe.
As semiconductor features approached the 1-micron size, concern arose that it might not be possible to resolve them with an optical microscope at any magnification. Mechanical probes also can damage circuit features, and they add undesirable electrical loads (especially for high-speed signals) that affect circuit performance.
Overcoming the Limits
But mechanical probing didn’t die. Even when the advent of FIB milling made it possible to selectively remove passivation and overlying layers to expose the feature of interest to e-beam probes, engineers still returned to mechanical probing for certain measurements.
Mechanical probing, making electrical contact with the feature of interest through physical contact, can do some things better than any other technique. Multiple probes can source and measure voltages and currents more precisely than any other technology. They work reliably through a wide voltage range, can be used over a frequency range from DC to RF, can pick up transient and one-shot signals, cost much less than other probing options and make use of established, well-understood technology.
But there is a drawback. It is difficult to align probes through an optical microscope. And the two things that restrict the usability of the optical microscope for submicron-geometry devices are its limited depth of field and resolving power.
What is needed is a microscope with a much larger depth of field and much greater resolving power. The FIB used in the imaging mode meets these requirements and provides a long working distance as a bonus.
Figure 1 compares the aperture angles, working distances and depths of focus of three imaging techniques: the optical microscope, the thermionic e-beam microscope and the FIB. The FIB system has a depth of focus two orders of magnitude larger than that of the optical microscope and a resolving power that is effectively unlimited.
In the absence of optical flaws, such as spherical aberration, an optical system’s depth of field is limited by its aperture angle. Optical microscopes must have relatively wide aperture angles to collect enough light to create a well-illuminated image.
FIB systems, however, have no such limitation. Because of the properties of the ion beam, they work best with narrow beams, small aperture angles and large working distances. These properties make them ideal for positioning mechanical microprobes.
Unlike optical microscopes, FIB systems must work in a vacuum environment; to use an FIB system to position probes, you must put the prober into a high-vacuum chamber.
Moving the Probes
By necessity, each probe must be moved in three dimensions. You have to move it along the X and Y axes to position it over the feature of interest and then carefully move it vertically along the Z axis to touch it down on the surface. And you have to control each probe’s position along each axis to within a fraction of a micron.
Conventional probers use one three-axis stage for each probe. Each axis of each stage is typically actuated by a micrometer screw or permanent-magnet stepping motor. Such stages are bulky, covered with lubricating oils and generally not what you want to put into the confines of a vacuum chamber.
To accommodate such bulky components would require a chamber with a large volume. Large vacuum chambers take a long time to pump down. The lubricants on the stage evaporate and contaminate the chamber walls and the microcircuit-under-test.
One solution uses piezoelectric stepper motors as shown in Figure 2. Placing a voltage across a piezoelectric crystal causes it to flex by an amount proportional to the applied voltage. In a piezoelectric stepper motor, four crystals are actuated in sequence to “walk” the stage along the axis in tiny steps. Since the steps are small, high resolution is possible. Resolutions on the order of 0.2 to 0.3 microns have been achieved.
Since the piezoelectric effect operates very quickly, the stage can move rapidly. Piezoelectric steppers can move the stage at speeds in excess of 10 mm per second.
A stage resolution of 0.2 microns is acceptable when probing older devices with an optical microscope. The resolution limit for optical microscopes is roughly 0.3 microns. When your line width is less than half a micron, as it is with the most advanced semiconductors today, and the resolution of the FIB microscope is much finer, you need still better stage-position resolution.
Piezoelectrics come to the rescue again. By surmounting the “coarse” stepper-actuated stage with a “fine” vernier stage actuated by three crystals, each one moving the probe along one of three axes, you can get fine control. Such stages are used to provide subatomic-size control in atomic-force microscopes.
The Combined System
Figure 3 shows a combined FIB/microprober. This embodiment makes provision for three mechanical probes, for a gas injector used when depositing new material on the circuit via the FIB and for selectively milling using halogen gas.
The FIB column has its own X-Y stage so you can position it over the feature of interest. Used in the imaging mode, the FIB is a high-resolution microscope with a very large depth of field that allows positioning the probes via their individual three-axis stages.
With the probes in position, the operator can perform microsurgery on the DUT. The operator can mill holes, cut conductors, locally depassivate selected areas, cut windows in power planes, deposit conductors or insulators, and generally perform all the functions that have made FIB systems useful for analyzing semiconductor failures. At the same time, this architecture allows for full high pin-count stimulus, using the mechanical probes to source and measure voltages and currents in the active microcircuit.
Figure 4 compares the diagnostic flow when using a separate mechanical prober in tandem with a FIB system to using a combined FIB/microprober. When using separate systems in tandem, the diagnostician must plan a set of modifications of the DUT, use the FIB system to make those modifications, remove the DUT from the FIB system, set it up in the prober and, finally, test the DUT to find the effects of those modifications.
To minimize the time wasted by moving the DUT between the two machines, several modifications must be done at several nodes in one loop at the FIB system. Then the modifications must be separately tested in another loop at the microprober.
With the combined system, modifications can be made and tested in one loop. In fact, it is possible to keep the device active and under test even while modifications are being made. Not only does combining the systems speed up the process by removing the time-consuming need to move the DUT between machines and reposition probes, it also makes what-if operations quick and easy.
Advantages for FIB Milling
Putting the prober into the FIB vacuum chamber makes it easier to position the probes. It also has definite advantages when using the FIB as a micromilling machine. The capability to source and measure voltages and currents at the feature you are milling improves control of surface charging, fine-tune cutting and deposition operations.
Fine tuning of milling and deposition operations on individual circuit elements has obvious advantages. For example, suppose you are trying to add a 233-W resistor between two metal lines. Having a microprober in the vacuum chamber allows you to place a probe on each end of the conductor and measure the resistance of the new resistor while depositing it with the FIB and gas injector. Then you can continue the deposition until the resistance between the probes reaches 233 W , and stop.
FIB operations that benefit from having microprobes include:
Measuring deposition resistance.
Verifying deposited connections.
Assessing and eliminating leakage currents in redepositions and gallium implantation.
Verifying cut resistance, especially for large cuts such as power buses and high isolation resistance.
Measuring the capacitance of a deposition, such as when tuning propagation delays.
Iteratively repairing margin failures.
The following example illustrates how a combined FIB/microprober system can improve fault diagnosis. A major manufacturer of SRAMs had a problem of intermittent single-bit errors. During functional memory testing, when the tester would write data into the memory and then immediately read it back out, the SRAMs would test perfectly. When the memory tried to hold data for some time, leakage currents would slowly redistribute the charge stored in individual memory elements until single bits would flip. The time constants of these bit errors would vary (depending on the leakage resistance at each given memory location on a particular SRAM IC) from milliseconds to 10 seconds or more.
While trying to find the failure using separated FIB and microprober systems in tandem, failure analysis engineers were able to dissect and test only two or three chips in a day. The time to go through one cycle from FIB to the microprobe and back exceeded two hours. They were trying to cut conductors with high isolation resistance, which took days to characterize.
They also had problems with charging, which caused damage that induced errors in subsequent leakage-current measurements. The high-aspect ratio cuts they used made it difficult to detect end-points for milling operations. Finally, they found that the delicate FIB operations were extremely operator dependent.
Combining FIB and the microprober into one machine would ease these problems. Shortened cycle times would make it possible to increase the number of modify/test cycles from two or three per day to 100 or more per day. Similarly, pad contact resistances could be characterized on-line.
Grounding key contact features would reduce or eliminate charge damage. On-line monitoring of key resistances would ease end-point detection and reduce the operator dependency of results.
Adding a mechanical probing system into the vacuum chamber of an FIB system extends the capabilities of mechanical probing techniques. The FIB, used in the imaging mode, becomes a positioning microscope with resolution and depth of field making it possible to probe submicron-geometry semiconductor devices.
The microprober, in turn, solves problems that have plagued the use of FIB milling systems since their introduction. Having both capabilities in one machine shortens diagnostic cycle times and makes possible operations that were unthinkable otherwise.
About the Authors
Christopher Talbot, Ph.D., who has been affiliated with Schlumberger Technologies since 1984, is the Product Manager for the IDS ProbePoint eXtension. Dr. Talbot received a B.S. degree in physics and electronics and a doctorate in physics from the University of Reading, U.K. He also has taught physics at Western Australian Institute of Technology and was a software engineer for Systems Technology in the Netherlands.
Douglas Masnaghetti is the Project Manager for the ProbePoint eXtension at Schlumberger Technologies. He has been with the company for 11 years, including 8 years in the Diagnostic Systems Group. Mr. Masnaghetti received a B.S. degree in electrical engineering from San Jose State University and an M.S. degree in electrical engineering from Santa Clara University.
Mariel Stoops is the E-Beam Applications Supervisor at the ATE Division of the Schlumberger Diagnostic Systems Group. Before joining the company five years ago, she was a Product Engineer for National Semiconductor. Ms. Stoops received a B.S. degree in electrical engineering from the University of the Pacific.
Schlumberger Technologies, ATE Division, 1601 Technology Dr., San Jose, CA 95110-1397, (408) 437-5162.
Copyright 1995 Nelson Publishing Inc.
May 1995
|