Probing currents on semiconductors in the femtoamp range once required expensive special equipment and expertise. Even then, valuable data often was not collected. Now, improvements in instruments such as semiconductor parameter analyzers and in probing equipment are changing the situation—for the better.
Measuring leakage currents on semiconductor wafers and within devices has always been a challenge. But market demands for higher density ICs, lower power, improved reliability and features such as programmability have made this challenge even greater. Semiconductor makers continue to engineer device sizes and manufacturing processes to push leakages and device off-currents ever lower.
When these currents are less than a picoamp, direct measurement can be extremely difficult. Though large cross-section test patterns can scale up some currents, space limitations or special requirements, such as failure analysis and spot-defect investigations, often make this technique impractical. An alternative method, magnifying leakage via high temperatures, demands precise temperatures and a low noise environment.
Enhanced instruments and better probing equipment make it possible to achieve desirable results now when probing femtoamps at high temperatures. With the proper equipment and setup, there is little degradation in measurement capability up to chuck temperatures of 300° C.
Equipment
To make on-wafer femtoamp current measurements, you need an instrument capable of resolving suitably low currents and methods of maintaining a low noise environment on and around the wafer chuck. You can’t ignore capacitances: It can take 100 s for 10 fA to charge 1 pF through 1 V.
To achieve tolerable measurement times, use triaxial driven-guard methodologies to nullify the effects of interconnect and device capacitances. The equipment described here provides one way to achieve accurate, repeatable results.
Parameter Analyzer
The evaluations were made using a Hewlett-Packard 4156A Parameter Analyzer, which lists specifications such as a 10-pA measurement range with 1-fA resolution and +20-fA accuracy. The analyzer stayed well within these specifications during repeating and cross-checking measurements.
The manufacturer’s recommended operating procedures—including appropriate hold and delay time settings, integration periods and frequent offset cancellations—maximized performance.
Probe Station, Dark Box and Controller
The proper measurement environment was equipped with an analytic probe station mounted in a metal dark box (Figure 1). The box shields the wafer from ambient light, which dramatically promotes current leakage. It also extends the electrical shield to enclose the test site entirely.
Triaxial probes, developed expressly to address the challenges of low-current, high-temperature probing, were mounted to standard manipulators, with the probe cables connected to the internal side of triaxial bulkhead connectors located on the wall of the dark box. The triaxial chuck was mounted atop the probe station X-Y table, with all cabling and cooling tubes routed through a rear service port to an external temperature controller and recirculating chiller.
The chuck design incorporates an isolated wafer platen, with underlying driven-guard and shield layers. In use, the wafer platen is connected to the source-measure terminal of one high-resolution source-measure unit (HRSMU), while the triaxial chuck guard is tied to the driven guard of the same HRSMU. The chuck shield is grounded to the dark box frame, connecting it to the HRSMU shield via the through-wall connectors.
The analytical prober, dark box, QuieTemp controller, triaxial probes and chuck were provided by Signatone.
Setup and Measurement Tips
The first step is to eliminate as much light as possible. Even dim light increases leakage by orders of magnitude.
Inspect the light box and seal as many light leaks as possible.
Use tape-over-metal foil to block both light and electrical noise.
Close the door fully and engage light locks for each measurement.
Ensure that microscope lights are fully off; don’t just turn down the dimmer switch.
The second step is to provide stable, low-impedance grounds, since even millivolts of ground noise will degrade results. Pick one point—for instance, a through-wall connector lug—as a grounding point. Avoid ground loops.
Next, reduce or eliminate noise sources inside the dark box:
Use grounded shielding around hot-chuck power cables and thermocouples exiting the box.
Remove or shield insulator sleeves which could generate or store static electricity.
Avoid rubbing or bending cables or insulators which could induce static charges.
Test after the chuck temperature has stabilized; this minimizes heating-element noise.
Clean the chuck with alcohol periodically. Contamination can introduce leakages.
Using the Parameter Analyzer
Here are some key points from the user manuals and from lessons we learned while conducting the tests withthe HP 4156A:
Provide the recommended environmental conditions (23° C + 5° C, 5% to 60% relative humidity, and a warm-up period of at least 40 minutes to minimize drift).
To achieve 1-fA resolution, set the HRSMU to auto-range or fixed on the 10-pA range.
Set the integration period to long to average out noise and maximize accuracy.
Increase hold and delay times until sweeping in both directions achieves similar results. Hysteresis in double sweeps usually implies times were too short, or there was poor contact with the sample.
Perform a zero cancellation directly before each low-current measurement. Temporarily lift the probe tip to null out residual device leakages. The measurement should be made expediently to avoid introducing drift into the system.
Experiment to find minimum delay and hold times needed for good results. Use these settings to minimize measurement time, reducing the chances for instrument drift.
Device Issues
If the results are inconsistent, device- or wafer-level issues may be the root cause. Active (amplifying) devices must be biased to keep gains low, minimizing the possibilities of oscillation. Analyzer filtering and integration (averaging) functions could easily mask this condition. Overstressed or ESD-damaged devices can cause unpredictable leakages; again, filtering and integration could mask this problem.
Results
The triaxial probes were connected to the transistor source, drain, gate and substrate pads, with the substrate probe guard also connected to the triaxial chuck guard terminal. A -0.5-V substrate bias ensured that a resident protection device between gate and substrate would not interfere with subthreshold measurements.
With the drain-to-source voltage programmed to 0.2 V, the gate voltage was swept from -0.5 V to 1.5 V. Resulting drain current plots at wafer chuck temperatures of 20° C, 50° C, 100° C, 150° C, 200° C, 250° C and 300° C are presented in Figure 2.
Note that the measured 1- to 3-fA off-currents at 20° C are as expected by extrapolation of off-currents at 50° C, 100° C and 150° C. Also note that accurate measurements are made at all temperatures, but the device subthreshold current increases as predicted by device physics.
The results also show that the equipment made measurements as low as 12 fA, below the specified +20-fA accuracy limits of the analyzer.
About the Authors
Brian Dickson is Vice President of Marketing at the Signatone Probing Division of Lucas/Signatone. He has been with the company since receiving a B.S. degree in business marketing from Brigham Young University. Signatone, Probing Division, 393-J Tomkins Ct., Gilroy, CA 95020, (408) 848-2851.
Lee Branst, President of Caracal Communications, is a former Senior Editor of Lasers and Optronics magazine. He received a B.S. degree from Case Institute of Technology and an M.B.A. from Case Western Reserve University. Caracal Communications, P.O. Box 1513, Redondo Beach, CA 90278, (310) 542-4233.
Copyright 1996 Nelson Publishing Inc.
May 1996
|
|