We have investigated the characteristics of THz emissions from p/n junctions with metallic lines under non-bias conditions. The waveforms, spectra, and polarizations depend on the length and shape of the lines. This indicates that the transient photocurrents from p/n junctions flow into the metallic lines that emit THz waves and act as an antenna. We have successfully demonstrated the non-contact inspection of open defects of multi-layered interconnects in a large-scale integrated circuit using the laser THz emission microscope (LTEM). The p/n junctions connected to the defective interconnects can be identified by comparing the LTEM images of normal and defective circuits.
© 2011 OSA
Over the past two decades, there has been a tremendous increase in research interest on THz science and technology, especially in THz spectroscopic imaging [1–7]. THz technology has many applications; for example, hidden objects can be inspected using THz transmission images obtained using nondestructive and non-ionizing diagnostics [8,9]. Imaging and sensing techniques using THz emissions from samples that are excited by ultrafast laser pulses have been investigated and applied to the inspection of semiconductor devices and sensing of the pH value of chemical solutions [10–14]. Recently, we have proposed a laser THz emission microscope as a novel tool for inspecting electrical failures in semiconductor devices under non-bias conditions; we succeeded in identifying metal oxide semiconducting field effect transistor (MOSFET) devices with defective interconnections without using any electrical probes [15,16]. In the failure analysis of large-scale integrated (LSI) circuits, it is important to make a distinction between acceptable LSIs and defective ones; moreover, the localization of defective areas is also indispensable for clarifying the causes of failures using physical analysis techniques such as transmission electron microscopy and focused ion beam technique. Conventional defect localization techniques employ infrared optical beam induced resistance change (IR-OBIRCH), infrared emission microscopy, and electron beam testing, but they require the application of bias voltages to the electrical contacts and reading of electrical signals from the LSI chips; this makes it difficult to perform defect localizations in LSIs during the manufacturing process without using electrode pads .
In this study, we have investigated techniques to obtain and analyze laser THz emission microscope (LTEM) images of semiconductor devices; in particular, we have investigated the THz emission characteristics from photo-excited p/n junctions connected to metallic lines under non-external bias voltage conditions. Further, by comparing the LTEM images of normal and defective circuits, we were able to successfully detect electrical failures in LSI interconnects without using any electrical probes.
2. Experimental setup
Figure 1 shows our experimental setup of the LTEM for noncontact and nondestructive inspections of LSIs. Pulses from a 1.058-μm wavelength fs mode-locked Nd:glass laser were used to excite the p/n junctions in LSIs. Since most LSI chips have multi-layered interconnections that prevent the excitation of p/n junctions with fs laser pulses from the circuit side, the excitation through the silicon substrate can be performed in the measurement of LSI chips . Transient photocurrents from the photo-excited p/n junctions result in the generation of THz waves through the silicon substrate. In order to enhance the sensitivity, the low-temperature grown GaAs photoconductive antenna detector was excited by the second harmonic of the 1.058μm wavelength. The THz waveforms emitted from the samples were obtained by sweeping the time delay position. The signals that were synchronized to the modulation frequency of the excitation laser by an acousto-optic modulator (20 kHz) were
lock-in detected. LTEM images with a fixed time delay position can be acquired by scanning the laser on the sample using the Galvano mirror; this enables rapid acquisition of laser scanning images without lock-in detections (typically 2 s for an image of 512 × 512 pixels). The acquisition time of the LTEM image (512 × 512 pixels) was approximately 5 min when the lock-in time constant was 1 ms. In order to measure the THz waveforms from the p/n junctions in the LSI, the position of the laser spot was monitored through the silicon substrate by using infrared illuminations and a charge coupled device (CCD) camera. The spatial resolution of the LTEM image depends on the laser spot size. A solid immersion lens was used to measure the LTEM image with the high spatial resolution .
3. THz emission from photo-excited p/n junctions connected to metallic lines
Here, we investigated the influence of interconnects on the THz emission from semiconductor devices; for this investigation, we measured the THz emission waveforms from simple test samples that consisted of p/n junctions with various metallic line structures. Figure 2 shows the test samples fabricated on p-type silicon substrates. In this experiment, the fs laser irradiated the area close to the contact region at normal incidence from the metal line side and THz waves emitted backward were detected. In this configuration, THz waves cannot be detected directly from the p/n junction because the photocurrent from the p-n junctions flowing perpendicular to the sample surface emits THz waves in parallel to the surface of the Si substrate, which is similar to the THz emission from the surface electric field in semiconductors . Figure 3(a) shows the THz waveforms emitted from different lengths of the line along the y-direction, as shown in Fig. 2(b). The polarization of the detected THz waveforms was parallel to the line along the y-direction. The polarization parallel to the x-direction could not be detected. The amplitude of the THz wave depends on the line length, and the maximum amplitude emission was obtained for a length of 300 μm. The line length of the strongest emission depends on the sensitivity of the photoconductive detectors. In this experiment, we used the photoconductive detector which had a bow-tie antenna with 6 mm length. Figure 3(b) shows the THz waveforms from samples that have straight lines along the (–y) direction, as shown in Fig. 2(c). The polarity of the THz waveforms in Fig. 3(b) is inverted compared to that in Fig. 3(a), which suggests a change in the current flow direction along the metal lines. Wang et al., reported that THz waves can propagate over 24 cm along with the single metal wire with 0.9 mm in diameter without large dispersion and the radiation
loss arisen from diffractive spreading of the propagating mode was dominant below 0.3 THz . In our case, the radiation loss of the metal line is expected to be much larger because the metal line width is 0.5 μm which is much smaller than the wavelength. In this point of view, the detected THz wave from the test sample can be considered as the radiation loss from the metal line which the photocurrent flows into from p-n junction. These indicate that the metal line can work as an antenna for THz emission. Figure 3(c) shows the Fourier spectra of the THz emission waveforms of Fig. 3(a). The emission frequency and bandwidth decrease with increasing metal line length which indicates that THz waves are emitted as a result of strong coupling of the photocurrent and the metal line. The resonant frequency f r of a dipole antenna on a thick dielectric substrate can be approximated as
where c is the light velocity in vacuum, λ r is the resonance wavelength, l e is the effective antenna length, ε e and ε d are the effective dielectric constant defined by ε e = (1 + ε d)/2 and the dielectric constant of the substrate, respectively. The Eq. (1) shows that the emission frequency increases with decreasing the effective antenna length. As regarding the bandwidth, Smith et al., who investigated the emission properties of photoconductive dipole antennas with length of 50-200 μm (fabricated on a radiation damaged silicon-on sapphire substrate), reported the absence of resonance peaks at expected frequencies for dipoles shorter than 100 μm which might be explained by the larger antenna width [19,20]. Our results qualitatively agree with these results. Figure 3(d) plots the dependence of the negative peak position of the THz waveforms on the line length which reflects the broadening of THz pulses as a result of the decrease of the resonant frequency. The observed peak delay suggests that the THz emission signals from p/n junctions with different line lengths appear in the LTEM images at different time delay positions. In addition, we confirmed that the THz emission signal disappear when there is no contact between the metal line and the n+ layer in the test structure, which indicates that the generated photocurrent cannot flow into the metal line because of the impedance mismatch between the metal line and the small capacitance gap at THz frequencies.
The LTEM image of the sample with 300-μm line lengths is shown in Fig. 3(e) which was acquired at the positive peak position. In Fig. 3 (e), the white area and the dotted red square indicate the positive signals of the THz emission and the n+ layer shape, respectively. The THz emission area is limited to a small spot that is located under the contact region between the metal line and n+ type layer, which indicates that photocurrent from the area except the contact region can hardly flow into the metal line. On the other hand, Fig. 3(f) shows the LTEM image of a sample with the same line length. The surface of the n+ layer of the test sample is covered by a silicide layer in this case. The area of the THz emission corresponds to the n+ layer shape. The image in Fig. 3(f) was measured by moving the sample stage and the white horizontal line is an artifact due to the low reproducibility of the sample stage in high speed scan. The propagation delay of the photocurrent from the laser excitation spot to the contact area can be roughly estimated below tens of fs that are considerably smaller than the THz pulse width, which agrees with the uniformity of the THz amplitude image in the Fig. 3(f). The difference of the THz emission area between Figs. 3(e) and (f) might be explained by the difference of the conductivity because the conductivity of the silicide layer is higher than that of n+ layer and almost identical to that of the metal.
Figures 4(a) and (b) show the dependence of the THz waveform and peak position, respectively, on the gap position. The amplitude of the THz waveform increases while the peak position delay increases with increasing L2; this trend is in agreement with the results for the line length dependence (Fig. 3). This result indicates that the transient photocurrents cannot propagate over the gap. We attribute this to the small capacitance of the gap which increases the impedance mismatch and the photocurrent is reflected at the end of the line. Treizebre et al. , who investigated the THz Planar Goubau Line (PGL) for biological
characterization, reported that the THz field propagating along the main PGL was strongly coupled with the line separated by a gap whose capacitance is much larger than our case, resulting in the resonant frequency change of THz transmission. It is expected that the photocurrent from the p-n junction can propagate the interrupted line with the gap structure which decreases the impedance mismatch for the THz transmission.
Figures 5(a) and (b) show the THz waveforms from p/n junctions with a metal line bent at right angles, as shown in Fig. 2(e). The detected polarizations in Figs. 5(a) and (b) are parallel to the x- and y-directions, respectively. The amplitude of the waveform increases with the line length parallel to the polarization of the THz electric field; this agrees with the results for the line length dependence shown in Fig. 3. Figure 5(c) shows the dependence of the maximum peak position of the THz waveforms on the bending position L3. The peak delay positions of both polarizations (Ex, Ey) increase with L3. Interestingly, the delay in the peak position of Ex is larger than that in the peak position of Ey, which corresponds to the photocurrent path in the right-angled line; photocurrent from p-n junction flows in y-direction and then in x-direction.
We also investigated the influence of the structure of the p/n junctions on the THz emission. Figure 6(a) shows the sample structure. The samples have three types of junctions—n+/p, n+/n/p, and p+/p—and are connected to a metal line with a length of 100 μm. Here, the n+ and p+ type layers have higher concentrations of donor and acceptor impurities than the n and p type layers, respectively. Figure 6(b) shows the THz waveforms obtained by exciting the
junctions with fs laser pulses. The THz emission can be observed even from the p+/p junction. The junction between the layers of the same type, but with different carrier concentrations, has a depletion layer with an internal electric field that can accelerate the photo-excited carriers, thereby leading to THz emissions. The polarity of the THz waveform from the p+/p structure is opposite to that of the n+/p and n+/n/p structures, and this corresponds to the direction of the transient photocurrent from the junctions. The peak amplitude of the THz waveform from the n+/p structure is larger than the amplitudes of the waveforms from the n+/n/p and p+/p structures; this is consistent with the electric field strength in the depletion layer and corresponds to the difference in the Fermi energy in the diffusion layers.
In this section, the THz emission characteristics from the p/n junctions with metal lines have been investigated by measuring the THz emission waveforms. The THz waveforms are strongly affected by the metal line structure because transient photocurrents from p/n junctions flow into the metal line and the metal line plays the role of a THz emission antenna. To illustrate this, Fig. 7 shows the differences between the THz emission signal of a normal semiconductor device and one with an electrical failure in the interconnection. When a p/n junction is excited by fs laser pulses, transient photocurrents start flowing into the interconnection, thereby causing a THz emission. The experimental results demonstrate that the emitted THz waveforms are significantly affected by the interconnection structure. Therefore, it is expected that LTEM can be used to detect the interconnection defects that cause the electrical failures in semiconductor devices because changes in the photocurrent paths in semiconductor devices induce changes in the THz emission signal. Furthermore, it is useful to analyze the THz emission waveform from a p/n junction connected to the defective interconnection for the localization of the electrical failures in semiconductor devices.
4. Non-bias inspection of defective interconnections in LSI circuits
In this experiment, we used a logic LSI circuit (c7552 ISCAS’85 benchmark circuit) as a test sample for demonstrating the inspections of defective interconnections using LTEM under non-bias conditions. We cut the interconnections of a normal LSI using the focused ion beam (FIB) technique to prepare a defective LSI. Figure 8(a) shows an image of the c7552 circuit recorded by laser scanning. Figure 8(b) shows the THz emission waveforms obtained by focusing the laser irradiation at the area indicated by a red arrow in Fig. 8(a). Figures 8(c) and (d) show LTEM images obtained by fixing the delay stage at the maximum peak position indicated by the arrow in Fig. 8(b). The detected polarizations of Figs. 8(c) and (d) are parallel to the x- and y-directions, respectively. The difference between the two images reflects the anisotropy of the interconnection structure of the LSI. As shown in section 3, the polarization of the THz emission depends on the path of the photocurrent that flows through the interconnection structure. Figure 8(c) has a striped pattern that corresponds to the structure of the voltage common drain FET (VDD) and Ground (GND) line, which is shown in Fig. 8(e). However, the THz signal around the central area is weaker than that near the LSI edge. This can be explained by the cancellation effect of the THz emissions from the transient photocurrent flowing in the VDD and GND lines along the x and (–x) directions. On the other hand, the LTEM image in Fig. 8(d) is extremely different from that in Fig. 8(c). In addition to the THz emission signal from the VDD and GND lines, many THz signals from other interconnection structures can be seen. Since the THz emission waveforms are strongly affected by the interconnection structure, it is reasonable to expect that the LTEM image depends on the time delay position. It is necessary to measure LTEM images with multi-time delay positions to determine the THz emission signals from all interconnections.
We also conducted experiments to demonstrate the detection of the p/n junction connected to a defective GND line in the LSI. Figures 9(a) and (b) show the LTEM images in the normal and defective samples, respectively. The observed area is indicated by a red box in Fig. 8(a).
Both the images were acquired using a solid immersion lens that enables improved spatial resolution . Furthermore, the solid immersion lens increases the THz emission efficiency from the chip because it reduces the total reflection loss in the Si substrate. Figure 9(c) shows the CAD layout around the observed area. Strong THz emission areas superimposed on the stripe pattern can be observed at the contact regions between the interconnection and diffusion layers. On comparing the LTEM images, it can be observed that one of the three THz signals, corresponding to the contact regions within the yellow box in the normal LSI image, is absent in the defective one, as indicated by a red arrow in Fig. 9(b). This result indicates that transient photocurrents from the p/n junction cannot flow into the GND line due to the open defect, and moreover, the THz emission signal decreases when the GND line length decreases.
We have investigated the THz emission characteristics from p/n junctions connected to various metallic line structures. It is found that the THz emissions from p/n junctions are strongly affected by the interconnection structure where transient photocurrents flow from the p/n junctions. Furthermore, the THz waveform depends on the open defect position in the metal line. These results indicate that the analysis of the THz waveforms is useful to the localization of the defect position in the interconnection in LSIs. We demonstrated this non-bias inspection technique for LSI interconnection using LTEM for the first time. The system provides rapid data acquisition using the rapid laser scanning with Galvano mirrors; the acquisition time for an LTEM image is below 5 minutes. In the comparison of the LTEM images between a normal LSI and a defective one, the p/n junction connected to the interconnection with an open defect can be determined without any electrical probes. The development of a defect localization technique in LSI interconnection based on the numerical simulation of the THz emission waveform change due to the defect is currently underway.
The test samples and LSI chips in this study were fabricated in the chip fabrication program of VLSI Design and Education Center, University of Tokyo. The authors wish to thank Mr. Inoue (TDIPS) for preparing the defective LSI chip. This work was supported by SENTAN, Japan Science and Technology Agency, Japan.
References and links
1. M. van Exter, C. Fattinger, and D. Grischkowsky, “Terahertz time-domain spectroscopy of water vapor,” Opt. Lett. 14(20), 1128–1130 (1989). [CrossRef]
3. D. M. Mittleman, M. Gupta, R. Neelamani, R. G. Baraniuk, J. V. Rudd, and M. Koch, “Recent advances in terahertz imaging,” Appl. Phys. B 68(6), 1085–1094 (1999). [CrossRef]
4. B. Ferguson and X.-C. Zhang, “Materials for terahertz science and technology,” Nat. Mater. 1(1), 26–33 (2002). [CrossRef]
5. M. Tonouchi, “Cutting-edge terahertz technology,” Nat. Photonics 1(2), 97–105 (2007). [CrossRef]
6. A. J. Huber, F. Keilmann, J. Wittborn, J. Aizpurua, and R. Hillenbrand, “Terahertz near-field nanoscopy of mobile carriers in single semiconductor nanodevices,” Nano Lett. 8(11), 3766–3770 (2008). [CrossRef] [PubMed]
9. J. F. Federici, B. Schulkin, F. Huang, D. Gary, R. Barat, F. Oliveira, and D. Zimdars, “THz imaging and sensing for security applications-explosives, weapons and drugs,” Semicond. Sci. Technol. 20(7), S266–S280 (2005). [CrossRef]
10. T. Kiwa, M. Tonouchi, M. Yamashita, and K. Kawase, “Laser terahertz-emission microscope for inspecting electrical faults in integrated circuits,” Opt. Lett. 28(21), 2058–2060 (2003). [CrossRef] [PubMed]
11. M. Yamashita, K. Kawase, C. Otani, T. Kiwa, and M. Tonouchi, “Imaging of large-scale integrated circuits using laser-terahertz emission microscopy,” Opt. Express 13(1), 115–120 (2005). [CrossRef] [PubMed]
12. S. Kim, H. Murakami, and M. Tonouchi, “Transmission-type laser THz emission microscope using a solid immersion lens,” IEEE J. Sel. Top. Quantum Electron. 14(2), 498–504 (2008). [CrossRef]
13. T. Kiwa, K. Tsukada, M. Suzuki, M. Tonouchi, S. Migitaka, and K. Yokosawa, “Laser terahertz emission system to investigate hydrogen gas sensors,” Appl. Phys. Lett. 86(26), 261102 (2005). [CrossRef]
14. T. Kiwa, J. Kondo, S. Oka, I. Kawayama, H. Yamada, M. Tonouchi, and K. Tsukada, “Chemical sensing plate with a laser-terahertz monitoring system,” Appl. Opt. 47(18), 3324–3327 (2008). [CrossRef] [PubMed]
15. M. Yamashita, C. Otani, K. Kawase, K. Nikawa, and M. Tonouchi, “Non-contact inspection technique for electrical failures in semiconductor devices using a laser terahertz emission microscope,” Appl. Phys. Lett. 93(4), 041117 (2008). [CrossRef]
16. M. Yamashita, C. Otani, K. Kawase, T. Matsumoto, K. Nikawa, S. Kim, H. Murakami, and M. Tonouchi, “Backside observation of large-scale integrated circuits with multilayered interconnections using laser terahertz emission microscope,” Appl. Phys. Lett. 94(19), 191104 (2009). [CrossRef]
17. R. Inoue, K. Takayama, and M. Tonouchi, “Angular dependence of terahertz emission from semiconductor surfaces photoexcited by femtosecond optical pulses,” J. Opt. Soc. Am. B 26(9), A14–A22 (2009). [CrossRef]
19. P. R. Smith, D. H. Auston, and M. C. Nuss, “Subpicosecond photoconducting dipole antennas,” IEEE J. Quantum Electron. 24(2), 255–260 (1988). [CrossRef]
20. M. Tani, S. Matsuura, K. Sakai, and S. Nakashima, “Emission characteristics of photoconductive antennas based on low-temperature-grown GaAs and semi-insulating GaAs,” Appl. Opt. 36(30), 7853–7859 (1997). [CrossRef]
21. A. Treizebre, M. Hofman, and B. Bocquet, “Terahertz spiral planar Goubau line rejectors for biological characterization,” Prog. Electromagn. Res. M 14, 163–176 (2010). [CrossRef]