Separating one micron from another: an accurate probe station lets you count the atoms inside a device--well, almost.
[FIGURE 1 OMITTED]
Production probing emphasizes high reliability to ensure low down time, minimal-contact scrub-mark damage, and maximum throughput. These factors are affected by the prober as well as the probe card, the amount of overtravel used to ensure good electrical contact, and the number of times a pad is probed. Positioning accuracy is measured in single-digit microns, which underscores the need for high mechanical stability.
Shrinking bonding pad size has only made probing more difficult. The bonding pad pitch has reduced from 90 [micro]m in the mid '90s to under 50-[micro]m today and will continue to decrease slowly over time. However, pad size must remain in the 10s of microns to allow wire bonding. The present goal of ball-bonder manufacturers is development of reliable sub-45-[micro]m pitch processes. (1)
As an example of typical prober specifications, the Wentworth Laboratories Pegasus[TM] 300S semiautomatic 300-mm machine lists X-Y positioning resolution as 0.625 [micro]m, accuracy of [+ or -] 8 [micro]m over the 300-mm range, and repeatability of [+ or -] 4 [micro]m. Because the positioning uncertainty is small compared to the size of a bonding pad, travel from pad to pad and device to device can be programmed by the operator.
To put these numbers into perspective, the diameter of a human hair ranges from about 17 to 181 [micro]m. Black hair is thicker than red hair. Diameter also depends on a person's genetic makeup and age, the weather, and if the measurement is made at the end or close to the hair root, according to Brian Ley in The Physics Factbook. So, in simple terms, the 300S is capable of positioning a probe on the finest human hair after traversing 300 mm.
The probe station is constructed with high-precision lead screws and guide ways. But to achieve this degree of accuracy, Wentworth Laboratories has used vision pattern recognition to measure the X-Y stage's intrinsic positional errors over the entire probing area.
"Once known," explained John Fitzpatrick, the company's engineering manager, "unique algorithms dynamically compensate for these errors during stage movement. For example, in a typical movement of 100 mm, the stage actually may move 100.01 mm. Our algorithms will apply a correction factor to compensate for the error by instructing the system to move 99.99 mm."
Taking a different approach to high-accuracy positioning, Electroglas has developed a closed-loop, linear motor-driven, air bearing-supported X-Y stage for the EG5|300 Production Prober. Jeff Hintzke, the company's vice president of marketing, explained, "This type of stage is more like those used in front-end lithography tools and delivers better than 1.0-[micro]m accuracy. We found that using ball screws and traditional bearings could not consistently deliver this level of performance, especially over time as the parts wear."
In contrast, analytical probe stations may be required to position a probe to contact a 0.1-[micro]m feature. These tools are used by engineers performing failure analysis or researchers examining detailed device performance. With such small dimensions come changes to the prober's construction and its operation.
Because deep submicron structures are too small to be resolved with visible light, some probe stations use scanning electron microscopy (SEM). Others use atomic force microscopes (AFM), although currently this technology has a small following.
As shown in Figure 2, an AFM contacts the DUT surface with a very small probe tip. The tip's vertical deflection is measured by a built-in laser system. The probe also can provide electrical contact to the DUT. Very fine height resolution is possible over a limited X-Y range.
[FIGURE 2 OMITTED]
Accurate Z-axis positioning to within a micron takes on more importance because failure analysis requires access to the inner layers of a device. Special culling probes and lasers are used to expose the desired areas.
And without enhanced mechanical stability, it would be impossible to maintain long term contact of a probe on a very small feature. Consequently, the probe station's mechanical stiffness and insensitivity to temperature changes become critical.
Signatone's Director of Technical Sales Mike Hathaway said, "We've concentrated on reducing probe-station flexure and increasing vibration isolation and load-bearing capability of the optical platform. We've also developed hands-off control of these tools. To achieve 50-nm positioning resolution, an optically encoded thumbwheel control system is linked to a computer-aided positioner.
"We use 2" thick machined aircraft aluminum bases on the 8" probe stations and 3" thick bases on the 12" machines," he continued. "It's relatively simple to increase leadscrew pitch, stepper-motor increments, and encoder scales, but without a stable platform to operate from, these improvements are meaningless."
The mechanical stiffness D of a flat plate of thickness h is given by
D = E[h.sup.3]/(12 (1-[v.sup.2]))
Where: v = Poisson's Ratio
E = Young's Modulus
Clearly, increasing the thickness of a prober base can greatly improve stiffness because of the [h.sup.3] term. Another approach uses a different material.
For granite, Young's modulus equals 100 GPa, compared to about 69 for aluminum. In addition to low moisture absorption and high abrasion resistance, granite is used for surface plates because it doesn't distort easily under heavy loads. Also, granite's coefficient of thermal expansion is approximately three times smaller than that of aluminum. Some probers do use granite bases, but the material more often shows up in massive air-supported vibration isolation tables.
Although aluminum and granite have nearly identical densities of 0.1 lb/[in.sup.3] or 165 lb/[ft.sup.3], because of its lower stiffness, an aluminum slab the same size as a granite one would not work as well in this application. The stiffness of the granite ensures that any vibration transmitted to the slab will equally affect all parts mounted to it.
The vibration-isolated Cascade Microtech Model S300 Probing Station incorporates a large granite base-plate supported on air springs built into its legs. According to Larry Dangremond, the company's product manager for DC and CV applications, this construction provides a very rigid wafer-chuck mounting.
Probes are mounted on the platen above the chuck, and the stiffness of the heavy steel platen ensures repeatable probe positioning to within a few microns. This level of performance is more than adequate for customers who probe pads rather than actual device structures. However, for failure analysis and circuit debug work, a thicker platen is available with submicron stability.
Given probe stations with the necessary mechanical stiffness to examine submicron features, a related set of requirements soon becomes apparent. A list of small-geometry probing challenges was provided by Clint Waggoner, a product manager at The Micromanipulator Company.
* Designing a probe needle with less than a 100-nm point radius that is strong enough to withstand multiple probe touchdowns.
* Maintaining a low-contact resistance to the device being probed.
* Providing an imaging system that facilitates viewing of the small target while not damaging or altering the circuit characteristics of the devices under test.
Probes with 0.1-[micro]m radius tips are available from a number of sources. although their performance may not be equal. Because of its high strength and hardness, tungsten or a rhenium-tungsten alloy is used virtually in all of these very small probes.
Such a small tip has a high resistance associated with it. Theoretically, if the tip tapered to a perfect point, the contact area would be zero and the resistance would be infinite. A tip radius of 0.1 [micro]m corresponds to a very small cross-sectional area near the probe tip and a resistance hundreds of times higher than in the much larger diameter shank.
For example, the Probing Solutions Model 407X Disposable Probe is 1.4" long, although the tapered tungsten tip accounts for only 0.013". The 0.003" diameter tungsten cat whisker is bonded to a bendable 0.020" dia nickel shank that makes up most of the length. An inch length of 0.20" dia nickel wire contributes only 11 m[ohm] to the total probe resistance, but the very small tungsten tip adds more than 1.5 [ohm].
In addition to the high resistance caused by the small-radius tip, the actual contact of the tip to the device feature being probed also may add significant contact resistance (CRES). The preparation of the wafer, the surface roughness and cleanliness of the tip, and the contact pressure all affect CRES. A thorough discussion of the effects of oxidation and probe roughness can be found in Reference 2.
Gold plating is a partial solution to controlling contamination and CRES. but in the very thin layer that would be necessary to maintain a submicron tip radius, it can't have much effect on probe resistance. As an example of the amount of plating necessary to make a substantial reduction in resistance, Advanced Probing Systems deposits a 5-[micro]m thick silver plating on some models of larger tungsten probes to improve the current-carrying capability. This amount of silver reduces resistance by less than 50%.
Mr. Waggoner described some of his company's solutions to probe-tip problems. "The NANO-100 probe-needle contact resistance is consistently below 10 [ohm] for 50-nm to 250-nm point radii. We coat the probe tips with materials that do not alter the tip size but do resist oxide formation. We also have become proficient at cleaning samples prior to probing to remove oxide on the probing targets. Because the prober operates in a [10.sup.-6] Torr vacuum, further oxide formation is eliminated."
Of course, you also need to see the probe and guide it to the desired landing spot. And, it needs to stay there. Pad-to-pad movements can be programmed after the initial contact has been made, but to establish the starting position, some probe stations use a point-and-shoot system. The station's software will drive the probe to the point defined by a cursor you position with the controller's mouse.
In the production oriented Electroglas EG5|300, the stated positioning accuracy now includes probe-card tolerances. "We use fiducials patterned on the probe card as our main alignment targets instead of the individual probe tips," said Mr. Hintzke. "The offset between fiducials and probe tips is supplied by the probe card supplier or measured on the prober. The technique eliminates inaccuracies inherent in trying to image very small probe tips and the drift that normally occurs to the probe tips with use."
Maintaining reliable contact for a long time is a common problem, especially when a thermal chuck is being used for high-temperature research. Some applications, such as investigations of time-dependant dielectric breakdown (TDDB) or negative device temperature instability (NDTI), involve positioning probes and leaving them there for three hours. These tests may be run at 300[degrees]C.
To avoid the misalignment that would result from thermal expansion, a vision system can be used to maintain the original probe-to-device feature relationship. For tests taking much less time but involving devices on different wafers, a vision system may eliminate the delay required to tallow all paris of the probe station to reach thermal equilibrium. Taking a different approach, Signatone has added temperature control of the platen and chuck support structure to speed up thermal stabilization of the system.
Also, with thermal chucks come more sources of electrical noise. Driving the heating elements from a DC source rather than AC is part of the solution as is attention to shielding and power supply design.
Guarding and shielding techniques are available that ensure stable measurements below 1 fA and 1 fF. These low levels are achievable, but doing so takes care. Cabling connected to the chuck flexes each time the chuck moves. Special insulation and cable construction have been developed to minimize noise and leakage currents. Also, the need for wafer cleanliness cannot be stressed too greatly: one fingerprint or some solder flux will add orders of magnitude to the leakage values.
Portability is yet another factor. Years ago, 3" and 4" wafer-probing stations could easily be carried from your desk to a colleague's at the other end of the lab. Today's 300-mm probe stations are more permanent installations. For example, the Cascade S300 weighs 1,720 lb. It must be connected to a vacuum line, sources of dry air and compressed air, and both 115-V and 230-V power. Nevertheless, because a vibration isolation table is built in, moving the trail may be easier than for some other models.
Less integrated solutions still would require isolation tables for very fine work, so the combined weight of the probe station and table could be similar to that of the S300. And, the separate parts would have to be reintegrated after the move. The point is that 300-mm probers and probe stations require more planning with regard to both their use and maintenance than smaller machines.
The list of application-related considerations includes additional things such as elimination of moisture or frost build-up when using a cold chuck and a dark environment for making light-sensitive measurements. These requirements and others associated with your specialized research should be discussed with potential vendors before deciding on any purchase.
For example, several manufacturers offer light-tight enclosures, but you also may need remote control of probe positioning. In general, you need to determine all the likely uses for a new prober or probe station and then decide which products crime closest to your requirements. A very large number of variations are possible if the range of probe positioners, chucks, enclosures, microscopes, shielding, and size is taken into account.
(1.) Klossner, M., et al, "An Integrated Approach to Solving Sub-45-[micro]m Wire Bond Process Challenges," Kulicke & Sofia, SEMICON Singapore, 2001.
(2.) Broz, J., and Rincon, R., "Understanding Probe-Contact a-Spot Oxidation During Elevated-Temperature Wafer Test," EE-Evaluation Engineering, September 1999, pp. 58-69.