Optical beam–induced current �OBIC� imaging iswidely employed for failure or defect detection of pnjunctions, interlevel shorts, transistor states, etc., inintegrated circuits �IC�.1–9 An OBIC image is a mapof the current magnitudes that are induced when a�focused� optical beam is scanned across an IC sam-ple. Scanning confocal microscopy, with its focusedprobe beam, is readily combined with OBIC imagingto produce a pair of confocal reflectance image andOBIC images of the sample from the same beam scan.
The one-photon-absorption OBIC �1P-OBIC� is pro-duced by an illuminated semiconductor material ifthe probe photon energy exceeds the semiconductorbandgap Eb, i.e., �1P � hc�Eb, where �1P is a single-photon wavelength, h is the Planck’s constant, and cis the speed of light in vacuum. The 1P-OBIC isproportional to the probe-beam intensity, and themeasured 1P-OBIC signal is an integrated effectalong the optical beam path. Unlike confocal im-
The authors are with the National Institute of Physics, Univer-sity of the Philippines, Diliman, Quezon City, Philippines 1101.The e-mail address of C. Saloma is [email protected].
Received 10 December 2001; revised manuscript received 26March 2002.
ages that are high-contrast displays of the reflectanceof a three-dimensional sample,10,11 the corresponding1P-OBIC image of the same sample has low contrastand lacks vertical resolution.
A two-photon OBIC �2P-OBIC� has been demon-strated to generate high-contrast images of semicon-ductor sites in an IC.12–14 The 2P-OBIC utilizes anexcitation beam with a wavelength �2P � hc�Eb.The 2P-OBIC is proportional to the square of thebeam intensity and is highly localized within the focalvolume of the excitation beam. Another techniquefor obtaining high-contrast 1P-OBIC images is vianear-field microscopy with a subwavelength fibertip.15 A major drawback of the 2P-OBIC is the highcost of a femtosecond laser source. Image genera-tion in near-field microscopy is slow and unsuitablefor generating large image fields. It is also sensitiveto ambient experimental conditions.
Here we present a computationally efficient proce-dure for generating high-contrast images of semicon-ductor sites in the IC from their 1P-OBIC andconfocal reflectance images that are obtained fromthe same focused beam. The procedure utilizes thefollowing properties: �1� Only semiconductor mate-rials produce an OBIC signal, and �2� confocal reflec-tance images are optically sectioned images of bothmetallic and semiconductor surfaces. We show thatthe product of the low-contrast 1P-OBIC image andthe confocal image results in a high-contrast �axial-
dependent� map that reveals only the semiconductorsites in the confocal image. Similarly, the product ofthe complementary to the 1P-OBIC image and theconfocal image yields an optically sectioned imageexclusively of the nonsemiconductor sites in the ICsample.
Another advantage of 2P-OBIC imaging over 1P-OBIC imaging is realized when observations aremade in the presence of an intervening highly scat-tering medium between the focusing lens and thesemiconductor material.5,12 Because the scatteredintensity is inversely proportional to a power of theincident wavelength16 and because �2P � 2�1P, amuch greater percentage of the 2P excitation photonsis delivered at the focal volume of a 2P excitationbeam than their 1P counterparts for the same scat-tering medium and numerical aperture �NA� of thefocusing lens.17 The scatter-induced broadening ofthe axial distribution of the 2P-OBIC signal is lesssevere than that of the 1P-OBIC signal. In one-photon �1P� fluorescence excitation microscopy with alarge-area photodetector, the effect of scattering is todegrade the signal-to-noise ratio of the generated im-ages.18
It is worth noting that confocal microscopy is alsorobust against the unwanted effects of scattering byan intervening medium. The photodetector pinholeacts a spatial filter that permits the detection of pho-tons emanating only from the focal volume of theprobe beam.18 The undesirable image contributionof the photons from the out-of-focus planes can beminimized through careful choice of the pinhole size.
1P-excitation confocal microscopy can be done withobjectives of relatively low NA values but long work-ing distances—an advantage that is of practical im-portance for wide-field observation and when dealingwith thick samples.18 In contrast, 2P-excitation im-aging requires objectives with large NA values togenerate sufficiently high intensities at the focal spotbecause the 2P-excitation absorption cross section ismuch smaller than its 1P-excitation counterpart.Such objectives, however, normally have short work-ing distances that limit our ability to scan axiallythick samples at long depths. Aberration-free highNA objectives with long working distances are quiteexpensive to manufacture.
Our presentation proceeds as follows. In Section2 we present the principle that we employed in theimage-processing procedure. The experimental re-sults are presented and discussed in Sections 3 and 4,respectively.
2. Basis
The amplitude point-spread function �PSF� of a con-focal microscope �with a point light source and a pointdetector� is given by10,11 h1�x, y, z�h2�x, y, z� � h1h2,where h1 and h2 are the PSFs of the focusing andcollector lens, respectively. In a confocal reflectancemicroscope, h1 � h2 � h. The confocal intensityimage ir�x, y, z� that is generated from a three-
dimensional object with a reflection amplitude distri-bution or�x, y, z� is described by
ir� x, y, z� � �or� x, y, z� � h2� x, y, z��2, (1)
where R represents a convolution operation. Metalsand semiconductors surfaces have relatively high�or�x, y, z��2 values. The intensity PSF, which is theconfocal image that is produced by a point object, is�h2�x, y, z��2. The optical sectioning capability of aconfocal microscope that permits the generation ofhigh-contrast images is a consequence of the �h2�x, y,z��2 behavior of the PSF.
The same focused beam of the confocal microscopegenerates the 1P-OBIC signal whose strength is pro-portional to the beam intensity and depends not onlyon the 1P absorption cross section and the incidentbeam power but also on the optical path length.This implies that the measured 1P-OBIC signal isdoes not exhibit a z dependence and is calculated as
is� x, y� � ���
�
os� x, y, z� � �h� x, y, z��2dz, (2)
where os�x, y, z� represents the distribution of thesemiconductor material in the sample and is�x, y� �0. For nonsemiconductor materials �e.g., metalsand dielectrics�, os�x, y, z� � 0. Because is�x, y� hasno axial dependence, a 1P-OBIC image has low con-trast and contains no information about the depthdistribution of the semiconductor sites in the sample.It has already been reported that the 1P-OBIC imagelacks vertical resolution.3,12
However, an exclusive high-contrast image of semi-conductor sites can be derived from ir�x, y, z� and is�x,y� by taking their image product: s�x, y, z� � ir�x, y,z�is�x, y�, where s�x, y, z� � 0. From the propertiesof or�x, y, z� and os�x, y, z�, it is evident that s�x, y, z�is nonzero only for semiconductor materials. FromEqs. �1� and �2�, the associated PSF for the productimage is given by h4�x, y, z� � h2�x, y, z�dz � h6�x, y�h4�z�, where we have assumed that h�x, y, z� � h�x,y�h�z�, in the final expression. Therefore s�x, y, z�provides an exclusive map of the semiconductor sitesand exhibits the vertical resolution of ir�x, y, z�.
An exclusive high-contrast image of the metallicsites is obtained from the product: m�x, y, z� � ir�x,y, z�im�x, y�, where im�x, y� � � is�x, y� and is aconstant that represents the highest s�x, y, z� valuethat is possible for a given optical setup.
In practice, the sample is scanned by the focusedbeam at a sampling interval that takes into accountthe central spot size of h�x, y, z� and the Rayleighresolution criterion.19 The scanned confocal and 1P-OBIC images are represented by ir�i, j, k�� and is�i,j, k��, respectively, where x � i�x, y � j�y, and z �k�z; i, j � 1, 2, . . . J and k � 1, 2, . . . K. The sam-pling intervals are given by �x, �y, and �z. In ourexperiments, an 8-bit analog-to-digital converters areutilized for both ir�x, y, z� and is�x, y� so that 0 � ir�i,j, k� � 255 and 0 � is�i, j, k� � � 255.
scanned product image, s�i, j, k�� � ir�i, j, k�is�i, j, k��,has a computational complexity of order 1. The al-gorithm can be implemented quickly. The s�i, j, k�values are also not susceptible to rounding-off errorsthat are attendant in iterative reconstruction algo-rithms with high computational complexity.20
3. Experiments
A beam-scanning reflectance microscope was con-structed for both 1P-OBIC and confocal imaging �Fig.1�. Via a beam splitter BS, the output beam of theargon-ion laser �514 nm, Omnichrome� is directed toa scanning mirror system that is composed of twogalvanometer mirrors �General Scanning ModelG115� for x �G1� and y �G2� scanning and two lenses�L1, L2; focal length f � 60 mm� that constitute a 4ftransfer lens. Another pair of lenses L3 � f � 18 mm�and L4 � f � 200 mm� expands and collimates thescanned beam and inputs it to a conventional in-verted microscope assembly �Olympus IMT-2�. Aninfinity-corrected objective lens O focuses the beaminto the exposed top surface of the IC �Erasable Ran-dom Read Only Memory, National Semiconductors27C32�. Precise two-dimensional scan control of thefocused beam is achieved via a pair of 12-bit digital-to-analog converters �DAC�.
The reflected light is collected back by the sameobjective lens and focused by lens L5 � f � 40 mm�toward a pinhole that is placed in front of photode-tector PD �Hamamatsu H5784�. The 1P-OBIC ismeasured by inputting the output of the IC pin thatis nearest to the probe surface area to a current-to-voltage converter composed of an operational ampli-fier � A 741� and a feedback resistor R � 2 M�. Theother converter input is the common reference �GND�for the electronic circuits including the IC sample. A1P-OBIC signal is induced because the siliconbandgap Eb � 1.1 eV is less than the excitation pho-ton energy of 2.417 eV. Typical 1P-OBIC valueswere in the microampere range �optical excitationpower at 514 nm: �5 mW�. Both the PD and the
converter outputs are digitized via 12-bit analog-to-digital converters �ADC� into a personal computer�PC�.
Figure 2�a� presents a series of confocal images�NA � 0.5, 20� magnification, �z � 1 m, �x � �y �0.23 m, sampling duration per image � 32 s, pin-hole diameter � 25 m, J � 128�. The 20� objectiveprovided a good balance between spatial resolution�high NA� and axial depth scan �long working dis-tance�. In general, metals have higher reflectancesthan semiconductor surfaces, but these two materialsare still quite difficult to distinguish from each otherin a confocal image. Figure 2�b� shows the corre-sponding 1P-OBIC images for the axial locations.As expected, the raw 1P-OBIC images have low con-trast and do not exhibit any significant axial depen-dence. The dark portions of a 1P-OBIC imagerepresent the nonsemiconductor �metal� sites of thesample.
Figure 3�a� shows the processed images of the
Fig. 1. Optical setup of beam-scanning microscope for simulta-neous confocal reflectance and 1P-OBIC imaging. The optical ex-citation power is controlled via a neutral density filter �ND�.
Fig. 2. Comparison of �a� confocal and �b� 1P-OBIC images atvarious axial locations ��z � 1 m, image size is 30 m � 30 m,128 by 128 pixels�.
semiconductor sites at the same axial locations as inFig. 2�a�. The images were generated via the proce-dure described in Section 2. Consistent with ourtheoretical analysis, the processed images exhibit ahigh contrast that is typical of the confocal images inFig. 2�a�. Figure 3�b� presents the correspondingimages of the metal sites at the same axial locationsas in Fig. 2�a�.
4. Discussion
Our experiments showed that s�x, y, z� is an exclusivehigh-contrast image of the semiconductor sites in theconfocal image. It is obtained quickly and efficientlyfrom the 1P-OBIC and confocal reflectance imagesthat are generated by the same focused excitationbeam. The noniterative image-processing algo-rithm has low complexity and could be implementedquickly. The laser-scanning microscope that isneeded for the experiment is much less expensive
than that required by the 2P-OBIC. s�x, y, z� per-mits for a more accurate determination of the locationof a circuit defect in an IC—a task that is difficult toaccomplish with optical images.
A possible source of ambiguity in is�x, y� are thecontributions of other non-OBIC signals that are alsoinduced when a strongly focused beam irradiates asample surface. These non-OBIC signals may comefrom Raman scattering, photoluminescence, andother residual 1P absorption because of free carriers,which becomes more likely at high dopant levels.The photoelectric effect can also become significant asthe excitation wavelength becomes shorter. Thesenon-OBIC signals produce an error in s�x, y, z�.Such errors are practically nonexistent in 2P-OBICimaging. Note, however, that these errors could beminimized by sufficient control of the beam excitationpower and excitation wavelength in a confocal micro-scope.
5. Conclusion
An efficient image-processing procedure has beendemonstrated to obtain exclusive high-contrast im-ages of semiconductor sites in a confocal image by useof the 1P-OBIC and confocal reflectance images thatare generated by use of the same focused beam. Thehardware support for the procedure is significantlycheaper than that required by 2P-OBIC imaging,which also generates high-contrast images of semi-conductor sites.
This research is supported by grants from the Of-fice of the Vice Chancellor for Research and Develop-ment, University of the Philippines in Diliman andthe Philippine Council for Advanced Science andTechnology Research and Development.
References1. B. P. Richards and P. K. Footner, The Role of Microscopy in
Semiconductor Failure Analysis �Oxford U. Press, New York,1992�.
2. S. Takasu, “Application of OBIC�OBIRCH�OBHIC �Semicon-ductor Failure Analysis�,” JEOL News 36E, 60–63 �2001�.
3. A. Ribes, S. Damaskinos, H. Tiedje, A. Dixon, and D. Brodie,“Reflected-light, photoluminescence and OBIC imaging of so-lar cells using a confocal scanning MACROscope�microscope,”Sol. Energy Mater. Sol. Cells 44, 439–450 �1996�.
4. T. Koyama, K. Sonoda, J. Komori, Y. Mashiko, and M. Umeno,“Detection of defects in metal interconnects by the nonbias-optical beam induced current technique,” J. Appl. Phys. 86,5949–5956 �1999�.
5. C. Xu and W. Denk, “Two-photon optical beam induced currentimaging through the backside of integrated circuits,” Appl.Phys. Lett. 71, 2578–2580 �1997�.
6. K. Isoird, M. Lazar, L. Ottaviani, M. L. Locatelli, C. Raynaud,D. Planson, and J. P. Chante, “Study of 6H-SiC high voltagebipolar diodes under reverse biases,” Appl. Surf. Sci. 184, 477–482 �2001�.
7. T. Koyama, M. Umeno, J. Komori, and Y. Mashiko, “Evalua-tion of sliced morphology by near-infrared-laser optical-beam-induced-current technique,” Jpn. J. Appl. Phys. 40, 6446–6452�2001�.
8. D. Pogany, K. Esmark, M. Litzenberger, C. Furbock, H. Goss-ner, and E. Gornik, “Bulk and surface degradation mode in
Fig. 3. Exclusive images of �a� semiconductor sites and �b� metalsites at various axial locations ��z � 1 m, 128 by 128 pixels,image size, 30 m � 30 m�. The images are derived with theconfocal and 1P-OBIC images in Fig. 2.
9. G. Soelkner, J. Kreutle, J. Quincke, W. Kaindl, and G.Wachutka, “Back side optical beam induced current method forthe localization of electric field enhancements in edge termi-nation structures of power semiconductor devices,” Microelec-tron. Reliab. 40, 1641–1645 �2000�.
10. T. Wilson and C. J. R. Sheppard, Theory and Practice of Scan-ning Optical Microscopy �Academic, New York, 1984�.
11. T. Wilson, Confocal Microscopy �Academic, London, 1990�.12. C. Xu and W. Denk, “Comparison of one- and two-photon op-
tical beam-induced current imaging,” J. Appl. Phys. 86, 2226–2231 �1999�.
13. F. J. Kao, M. K. Huang, Y. S. Wang, S. L. Huang, M. K. Lee,and C. K. Sun, “Two-photon optical-beam-induced current im-aging of indium gallium nitride blue light-emitting diodes,”Opt. Lett. 24, 1407–1409 �1999�.
14. M. Unlu, B. Goldberg, W. Herzog, D. Sun, and E. Towe, “Near-
field optical beam induced current measurements on hetero-structures,” Appl. Phys. Lett. 67, 1862–1864 �1995�.
15. R. Newton, Scattering Theory of Waves and Particles�McGraw-Hill, New York, 1966�.
16. C. Blanca and C. Saloma, “Monte-Carlo analysis of two-photonimaging through a scattering medium,” Appl. Opt. 37, 8092–8102 �1998�.
17. C. Saloma, C. Palmes-Saloma, and H. Kondoh, “Site-specificconfocal fluorescence imaging of biological microstructures inturbid medium,” Phys. Med. Biol. 43, 1741–1759 �1998�.
18. C. Palmes-Saloma and C. Saloma, “Long-depth imaging ofspecific gene expressions in wholemount mouse embryos withsingle photon excitation confocal fluorescence microscope andFISH,” J. Struct. Biol. 131, 56–66 �2000�.
19. M. Born and E. Wolf, Principles of Optics, 7th ed. �CambridgeU. Press, Cambridge, UK, 1999�.
20. C. Saloma, “Computational complexity and observation ofphysical signals,” J. Appl. Phys. 74, 5314–5319 �1993�.
