Scanning Microwave Microscopy for Semiconductor Failure Analysis

By Wenhai Han

Table of Contents

Semiconductor Failure Analysis

Failure analysis is a vital process in the development of new products and/or the improvement of existing products in the semiconductor industry. Successful failure analysis can identify the root cause of a failed device and guide corrective actions.

Semiconductor failure analysis often involves a number of different techniques, such as curve tracing, scanning electron microscopy, transmission electron microscopy, microthermography, and focused ion beam analysis. Several techniques based on atomic force microscopy (AFM), such as scanning capacitance microscopy, conductive AFM, and scanning spreading resistance microscopy, have also been utilized for the analysis of various failed devices [1].

In this article, semiconductor failure analysis using Agilent Technologies’ recently developed scanning microwave microscopy (SMM) [2-5] is demonstrated.

Introduction to SMM

Scanning microwave microscopy combines the high spatial resolution afforded by an atomic force microscope (AFM) with the high-sensitivity electric measurement capabilities of a vector network analyzer (VNA). In SMM [2], an incident microwave signal generated from the VNA passes through a matched resonant circuit in the AFM and reaches the end of a conductive probe, which is in contact with a surface. The reflected microwave from the contact point is sensed by the probe and returned to the VNA.

By measuring the complex reflection coefficient, or S11 parameter, the capacitance/impedance at the contact point is obtained from the VNA. In practice, the mapped signals are the logarithm of the reflection coefficient amplitude, labeled as VNA amplitude, and the VNA phase. After proper calibration, the contact capacitance/impedance can be derived from the VNA amplitude and phase values. Prior to the introduction of SMM, only qualitative measurements had been used for this type of failure analysis work.

Agilent-exclusive SMM works on all major semiconductor types: Si, Ge, III-V (e.g., GaAs, InAs, GaN), and II-VI (e.g., CdTe, ZnSe). Unlike scanning-probe capacitive techniques, SMM does not require an oxide layer. It is also a superb choice for a wide range of biological and materials science applications, including the characterization of interfacial properties and contrast from molecular vibrational modes. In addition to its ability to work on semiconductors, glasses, polymers, ceramics, and metals, the technique lets Agilent AFM users perform high-sensitivity investigations of ferroelectric, dielectric, and PZT materials. Studies of organic films, membranes, biological samples, and ion channels can also benefit from the use of SMM.

SMM for Semiconductors

When a metal probe is in contact with a silicon surface, it forms a metal oxide-semiconductor capacitor — well studied in semiconductor physics [6]. In a simplified one-dimensional model, the total capacitance comes from the contributions of two capacitors connected in serial: the surface oxide dielectric layer with a fixed capacitance and the underneath depletion layer in the silicon substrate with a variable capacitance.

The capacitance variation of the depletion layer in response to an applied ac bias is determined by the depletion depth, which is in turn largely affected by the dopant concentration in the substrate. Therefore, by measuring the capacitance change induced by the applied ac bias, or dC/dV, the dopant concentration at each contact point can be mapped. Any failure due to abnormal dopant concentration can then be identified from the dC/dV image, simultaneously with the capacitance image measured from the VNA amplitude signal.

Experiment

The sample being tested was a depackaged 250nm static random access memory (SRAM) chip. A standard SRAM unit bit cell contains six field effect transistors (FETs): two p-type FETs in an n-doped well and four n-type FETs in a neighbor p well. Among thousands of bit cells on the chip, one was found failed. Inside the failed bit, one n-type FET was measured having an abnormal threshold voltage Vt. It was the 48th n-type FET on that row, as shown in Figure 1.

Figure 1. Optical image of a small section of the tested SRAM chip. The failed bit contains an n-type FET (the 48th on that row) with an abnormal Vt.

Scanning microwave microscopy was then utilized in an attempt to find any unusual properties of the transistor (i.e., properties that differed from those of the other, normal transistors). A pure Pt metal probe and a conductively coated silicon probe were both used. The Pt metal probe was 300 to 400 ìm long and made of solid platinum mounted on an aluminum substrate. Its spring constant was estimated as being from 0.3 to 0.8 nN/nm; its tip radius of curvature was approximately 10 to 20 nm. The silicon probe was 125 ìm long and coated with 20nm Ti and 10nm Pt. Its nominal spring constant was 5 nN/nm and it had a tip radius of about 40 nm.

Scanning was carried out under ambient conditions in contact mode. Selected microwave operation frequencies were between 2 and 5 GHz. The low frequency modulation was around 80 kHz. The scan rate was typically from 0.5 to 1 line/sec. Both types of probes showed consistent SMM results on the SRAM chip.

Results

To accurately identify any possible problem with a sufficient level of detail, every two pairs of FETs on the same row as the failed FET were scanned, from the 43rd/44th pair through the 51st/52nd pair, as shown in Figure 2. A total of four sets of images (A, B, C, D) was acquired under the same conditions. Each set contained topography (top), dC/dV (middle), and VNA amplitude (bottom) images, which were obtained simultaneously. For illustration purposes, the even number n-type FETs on the same row as the failed FET were outlined with squares in all images.

Figure 2. Four sets (A, B, C, D) of scanning microwave microscopy images on the failed SRAM chip. Each set contains topography (top), dC/dV (middle), and VNA amplitude (bottom) images, acquired simultaneously. The red squares outline the failed 48th n-type FET; the blue squares are normal n-type FETs on the same row.

The topography images of the 48th n-type FET, outlined by the red squares in Figures 2 B1 and C1, did not appear to have any structure difference compared to the other, normal n-type FETs (blue squares as well as the unmarked ones in all of the topography images). They were also very similar to those seen on other SRAM samples [2, 4].

In the dC/dV images, however, the difference was quite noticeable. Every normal n-type FET (blue square) consistently showed a dark area near the center. The dark (low) value in the dC/dV image represented p-type dopant of the well in the channel. The 48th n-type FET, on the other hand, was completely flat without any contrast, as shown in the red squares of both Figures 2 B2 and C2.

The missing p-dopant signal in the 48th n-type FET clearly indicated a change of dopant structure in the channel area of the transistor. The VNA amplitude images of the 48th n-type FET, if examined carefully in Figures 2 B3 and C3, also showed some different structure compared to the other, normal n-type FETs. This indicated a different capacitance/impedance value.

Conclusions

The utility of scanning microwave microscopy for semiconductor failure analysis has been demonstrated. Images of dopant concentration measured from the dC/dV signal on an SRAM chip clearly identified an unusual dopant structure in a failed n- type FET that differed from other, normal n-type FETs. Capacitance images measured from the VNA amplitude also showed a different contrast in the transistor.

This experiment has shown that scanning microwave microscopy can be a convenient direct-imaging tool for probing a variety of electrical failures in semiconductor devices at the micrometer to nanometer scale that are not visible from the surface topography structure.

References

[1] For example, P. Tangyunyong and C. Y. Nakakura, J. Vac. Sci. Technol. A 21, 1539 (2003); T. Tong and A. Erickson, ISTFA 2004, International Symposium for Testing and Failure Analysis, 42 (2004).

[2] Wenhai Han, Agilent Application Note 5989-8881EN, 2008.

[3] F. Michael Serry, Agilent Application Note 5989-8818EN, 2008.

[4] Wenhai Han and Hassan Tanbakuchi, ISPM 08, International Scanning Probe Microscopy Conference, Seattle, June 2008.

[5] Wenhai Han, Photonics Spectra, May 2008, p. 58.

[6] E. H. Nicollian and J. R. Brews, MOS (Metal Oxide Semiconductor) Physics and Technology, Wiley, New York, 1982.

About Agilent Technologies

Agilent Technologies nanotechnology instruments let you image, manipulate, and characterize a wide variety of nanoscale behaviors—electrical, chemical, biological, molecular, and atomic. Our growing collection of nanotechnology instruments, accessories, software, services and consumables can reveal clues you need to understand the nanoscale world.

Agilent Technologies offers a wide range of high-precision atomic force microscopes (AFM) to meet your unique research needs. Agilent's highly configurable instruments allow you to expand the system's capabilities as your needs occur. Agilent's industry-leading environmental/ temperature systems and fluid handling enables superior liquid and soft materials imaging. Applications include material science, electrochemistry, polymer and life-science applications.

Source: Agilent Technologies

For more information on this source please visit Agilent Technologies