OSA's Digital Library

Optics Express

Optics Express

  • Editor: Michael Duncan
  • Vol. 14, Iss. 3 — Feb. 6, 2006
  • pp: 1021–1026
« Show journal navigation

Spectral microthermography for component discrimination and hot spot identification in integrated circuits

Godofredo Bautista, Carlo Mar Blanca, Serafin Delica, Bernardino Buenaobra, and Caesar Saloma  »View Author Affiliations


Optics Express, Vol. 14, Issue 3, pp. 1021-1026 (2006)
http://dx.doi.org/10.1364/OE.14.001021


View Full Text Article

Acrobat PDF (354 KB)





Browse Journals / Lookup Meetings

Browse by Journal and Year


   


Lookup Conference Papers

Close Browse Journals / Lookup Meetings

Article Tools

Share
Citations

Abstract

We demonstrate an efficient and versatile spectral microthermography technique for identifying hot and cold spots in the active layer of a biased integrated circuit. Hot (cold) spots are regions where heat accumulates more rapidly (slowly) than the average rate of the entire active layer. Knowledge of the hot and cold spot locations is crucial in assessing the thermal integrity of a layer structure because hot spots are locations were defects are more likely to develop. The active layer is uniformly illuminated with light from a tungsten lamp and its reflectance image r(x, y) is scanned across (x-direction) the entrance slit of a grating-prism pair (GRISM) spectrometer to produce a spectral map R(λ; x, y) where λ is the wavelength [450 ≤ λ(nm) ≤ 650]. For a particular slit position x = x1 , the GRISM spectrometer outputs a one-dimensional spectral map R(λ; x1, y). A pair of maps Rub(λ; x, y) and Rb(λ; x, y) are obtained from the active layer in the absence and presence of voltage bias, respectively. A reflectance gradient map ∆R(λ; x, y) = Rb(λ; x, y) - Rub(λ; x, y), is derived and used to locate possible hot and cold spots because ∆R(λ; x, y) is proportional to the temperature gradient ∆T(λ; x, y). We use the technique to generate gradient maps of a photodiode array and the emitting surface of a biased light emitting diode. Two different semiconductor materials could be distinguished easily from their dissimilar reflectance spectra.

© 2006 Optical Society of America

1. Introduction

Spectral microscopy is a very promising technique for probing the optical characteristics of metal and semiconductor material structures in integrated circuits (IC) with high spatial resolution. Non-invasive methods have been devised for quantifying the thermal behavior of fabricated structures by monitoring changes in material reflectance with temperature variation (thermoreflectance) [1

1. G. Tessier, G. Jerosolimski, S. Holé, D. Fournier, and C. Filloy, “Measuring and predicting the thermoreflectance sensitivity as a function of wavelength on encapsulated materials,” Rev. Sci. Instrum. 74, 495–499, (2003). [CrossRef]

, 2

2. G. Tessier, S. Holé, and D. Fournier, “Ultraviolet illumination thermoreflectance for the temperature mapping of integrated circuits,” Opt. Lett. 28, 875–877, (2003). [CrossRef] [PubMed]

]. Other methods for mapping thermal gradients utilize reagents like liquid crystals, phosphors and fluorescent dyes deposited on the IC surface [3

3. J. Kolzer, E. Oesterschulze, and G. Deboy, “Thermal Imaging and Measurement Techniques for Electronic Materials & Devices,” Microelectron. Eng 31, 251–270, (1996). [CrossRef]

]. The methods however, are chemically invasive and severely limited by the thermal phase transition of the reagent.

The effective imaging resolution and detector requirements of imaging thermographs are set by the spectrum of the illumination source. Infrared-based techniques [5

5. J. Christofferson and A. Shakouri, “Thermal measurements of active semiconductor micro-structures acquired through the substrate using near IR thermoreflectance,” Microelectron. J. 35, 791–796, (2004). [CrossRef]

] are effective for multilayered structures but they require costly IR cameras for detection. On the other hand, ultraviolet light sources permit high-resolution imaging but they can hasten component failure or induce thermal stresses due to their high photon energies [2

2. G. Tessier, S. Holé, and D. Fournier, “Ultraviolet illumination thermoreflectance for the temperature mapping of integrated circuits,” Opt. Lett. 28, 875–877, (2003). [CrossRef] [PubMed]

]. Recently, visible light sources have been used which allow wavelength selection to maximize the sample reflectance signal and enhance measurement sensitivity [1

1. G. Tessier, G. Jerosolimski, S. Holé, D. Fournier, and C. Filloy, “Measuring and predicting the thermoreflectance sensitivity as a function of wavelength on encapsulated materials,” Rev. Sci. Instrum. 74, 495–499, (2003). [CrossRef]

]. However, the method could only provide one-dimensional spectral maps thereby limiting its utility to simple periodic structures.

Here, we demonstrate a versatile and cost-effective microthermography technique for measuring the spectral response of an arbitrary location on the active layer of an IC sample. The optical system employs broadband illumination and can currently derive spectral information within the 450 – 650 nm-range. The technique is applied in the following tasks which are important in IC failure analysis: a) To distinguish subtle differences between semiconductor edifices, b) To detect possible hot and cold spots in the active layer, and c) To measure the thermoreflectance of biased light emitting diode (LED) in the presence of a strong electroluminescence background. Hot spots are locations where defects are more likely to develop in the active layer. We also derive the thermal map of the LED to demonstrate the efficacy of our technique even to light emitting devices.

The discrimination of semiconductor and metal substructures in the active layer has already been previously demonstrated using a combination of confocal reflectance microscopy and optical beam-induced current imaging [6 – 8

6. V. Daria, J. Miranda, and C. Saloma, “High-contrast images of semiconductor sites via one-photon optical beam-induced current imaging and confocal reflectance microscopy,” Appl. Opt. 41, 4157 (2002). [CrossRef] [PubMed]

]. A thermographic method based on visible-wavelength optical-feedback absorption microscopy has also been developed [9

9. CM. Blanca, V.J. Cemine, V.M. Sastine, and C. Saloma, “High-resolution differential thermography of integrated circuits with optical feedback laser scanning microscopy”, Appl. Phys. Lett. 87, 231104-(1–3) (2005). [CrossRef]

]. The spectral protocol presented here, allows us to determine subtle performance differences among semiconductor structures and to conduct rapid thermal analysis using only the information that is contained in the reflectance spectra. Infrared emission spectral microscopy has also recently been utilized to measure differences in the infrared emission spectra due to subsurface leaky and good transistors in a biased microprocessor [10

10. C. Saloma, A. Tarun, M. Bailon, and M. Soriano, “Rapid subsurface detection of nanoscale defects in live microprocessors by functional infrared emission spectral microscopy” Appl. Opt. 447302–7306 (2005). [CrossRef] [PubMed]

]. Working with visible illumination improves the imaging resolution of the microthermograph and allows us to select the spectral band that optimizes the reflectance signal thereby improving the detection efficiency and dynamic range of measurement.

2. Methodology

Figure 1(a) depicts the optical configuration of our spectral microthermograph that can image an IC active layer at diffraction-limited resolution. The detected response RO(λ; x, y) from the active layer location (x, y) is: RO(λ; x, y) = I(λ)C(λ)D(λ)O(λ; x, y), where O(λ; x, y) is the reflectance spectrum, I(λ) the spectral emission of the tungsten lamp, C(λ) the collective reflectance of the filters, beamsplitter and lenses, D(λ) the detector response and λ is the illumination wavelength. We calibrated the thermograph using the control response RM(λ; x, y) = I(λ)C(λ)D(λ)M(λ; x, y), where M(λ; x, y) is the spectral reflectance of a silver mirror (New Focus) with a 95% uniform reflectance in the 450 – 700 nm-range.

The normalized spectral reflectance map R(λ; x, y) = RO(λ; x, y) / RM(λ; x, y) = O(λ; x, y) / M(λ; x, y), only contains spectral information about the IC sample. To ensure detection linearity, the signal intensity at the detector plane is regulated to fall within the linear response range of a black and white CCD camera (Hamamatsu C2400-77, 768 × 493 pixels).

The active layer is uniformly illuminated with light from a power-regulated tungsten lamp and a reflectance image r(x, y) is produced by an optical microscope [Olympus IX71, infinity-corrected 0.5 NA (20X) objective lens (UplanF1)]. The IC sample is mounted on a tri-axis, closed-loop scanning stage (Thorlabs PT3-Z6, displacement resolution = 50 nm). The image r(x, y) of the active layer is spatially filtered by the entrance slit of a grating-prism pair (GRISM) spectrometer (Imspector V9). At the slit location x = x1, the spectrometer produces a one-dimensional map RO(λ; x1, y) that is captured by the CCD camera. Figure 1(b) shows a one-dimensional map RM(λ; x1, y) that is produced by the silver mirror as sample.

A reflectance map RO(λ; x, y) of the entire active layer is generated by scanning the image r(x, y) across the slit in the horizontal x-direction at increments of 1.05 μm which is the step size value that yields the correct aspect ratio of a circular test object (Fig. 2(a)). One RO(λ; x, y) spanning the 450 – 650 nm-range, takes approximately three minutes to generate (size of map: 400 × 400 pixels).

Fig. 1. (a) Microthermography setup, (b) Control reflectance spectrum from flat mirror, and (c) Orientation of line scans and corresponding set of reflectance spectra.

3. Results

3.1 Discrimination of Semiconductor Sites

We acquire a pair of normalized R(λ; x, y)’s from two different sites (indicated by arrows) of a fabricated structure consisting of a gold (Au) ring (outer diameter = 250 μm) on gallium arsenide (GaAs) substrate. Figure 2(a) presents the wide-field reflectance image of the structure and Fig. 2(b) shows that Au and GaAs exhibit distinct spectral responses with different rates of increase in the long wavelengths.

Fig. 2. (a) Reflectance image (300 × 300 μm2) of Au-GaAs structure, and (b) measured reflectance spectra (resolution = 10 nm) from two different sites.

Figure 3(a)–(c) present the R(λ; x, y)’s of a photodiode array sample (Matsushita MN80905017) at λ = 478 (blue image), 559 (green) and 641(red) nm, respectively. The structures in the active layer exhibit spectral responses that change with λ. We are restricting our analysis to three wavelengths to simplify image representation and to explain more clearly the reflectance properties of the sample. Inclusion of other wavelengths is straightforward but would require more complicated image representation.

The image in Fig. 3(e) is the sum of the images in Fig. 3(a)–(c). The composite image delineates clearly the various substructures in the active layer. The white (P1) regions are described by a uniform reflectance spectrum that is a distinct characteristic of metals. The red (P3) semiconductor sites reflect only the red component and are absorbing at λ < 641 nm, while the yellow (P2) sites are absorbing at λ > 641 nm.

Figure 3(f) shows the R(λ; x, y)’s from the three selected sites (P1, P2, P3) in Fig. 3(e). The spectra exhibit different features that can be used to classify different structures in the active layer. Figure 3(e)–(f) demonstrate that our technique could distinguish two different semiconductor structures (P2, P3) from their dissimilar reflectance spectra. Defective components can also be distinguished from their characteristic reflectance spectra.

We note that red and yellow semiconductor sites are difficult to distinguish in the corresponding wide-field image [Fig. 3(d)] which is the sum of the images taken at all reflected wavelengths.

3.2 Thermographic detection of hot and cold spots

We also determine the thermal properties of the photodiode array sample by first obtaining its spectral map Rub(λ; x, y) in the absence of a voltage bias. The sample is then heated by applying a voltage bias and the corresponding map Rb(λ; x, y) is measured. Next, we calculated ∆R(λ; x, y) = Rb(λ; x, y) - Rub(λ; x, y) = ∆R, which describes gradients of thermal activity across the active layer. The effect of increasing ambient temperature is also investigated by mounting the array on a thermoelectric cooler.

Differences in ∆R can be positive or negative depending on the sign of dR/dT which depends on λ and on the material. We separated the ∆R values according to their algebraic signs. In Fig. 4(a)–(b) are the thermoreflectance maps taken at an IC package temperature T = 25°C. Biasing causes the active layer to respond differently. Regions with nonzero +∆R values are characterized by high photon absorption rates. They accumulate heat more rapidly due to the localization of non-radiative carriers by redistribution and are the primary hot spot candidates. On the other hand, regions with -∆R values are cold spots where the absorption of incident photons is impeded by the migration of energetic free carriers. The formation of hot and cold regions become more likely at T = 47°C [Fig. 4(c)–(d)].

Fig. 3. Photodiode array sample. Spectral maps (480 × 480 μm2) at λ = 478 nm (a), 559 (b), and 641 (c). In (d) is the wide-field reflectance image and in (e) is the superposition of images in (a)–(c). Shown in (f) are the reflectance spectra from selected locations of the composite image in (e). Image intensity scale: black = 0.
Fig. 4. Maps of reflectance gradient +∆R [(a), (c)] and -∆R [(b), (d)] at λ = 613 nm and T = 25°C [(a), (b)] and 47°C [(c), (d)]. Image field of view is same as in Fig. 3.

3.3 Thermoreflectance measurement in a strong electroluminescence background

The thermal differential ±∆R maps are not straightforward to derive in biased light emitting devices like semiconductor lasers or LED’s where the reflectance signal is accompanied by a strong electroluminescence signal. We now demonstrate a simple spectral unmixing procedure for separating the weak reflectance signal from the electroluminescence background allowing us to calculate the ±∆R maps of the biased LED sample.

We measure the reflectance map Rub(λ; x, y) of the uniformly-illuminated LED sample in the absence of a voltage bias. We also determine the electroluminescence map E(λ; x, y) which is obtained from a biased LED sample in absence of tungsten lamp illumination. Finally we acquire the mixed map Rmix(λ; x, y) = Rb(λ; x, y) + E(λ; x, y) from a uniformly-illuminated biased LED sample. We then calculate the background-free LED reflectance spectrum Rb(λ; x, y) using the relation: Rb(λ; x, y) = Rmix(λ; x, y) - E(λ; x, y). The thermoreflectance change ∆Runmixed that occurs when the LED becomes biased is: ∆Runmixed = Rb(λ; x, y) - Rub(λ; x, y).

The surface-mounted LED sample was operated with an injection current of 60 mA at 25°C. Its electroluminescence peaked at 605 nm. Figures 5(a)–(c) show a set of spectral maps E(λ; x ,y) that illustrate the weakening of the luminescence signal at λ < 605 nm. At 570 nm, the non-uniformity of the emitting structure is made more apparent by the weakly emitting p-n junction (lower right quadrant of image). We calculate the ±∆Runmixed maps [Figs. 5(d)–(e)] to determine the thermoreflectance behavior of the sample. The reflectance spectrum peaks at 538 nm which is different from the electroluminescence peak. The +∆Runmixed maps reveal regions of non-uniform thermal activity that become more distinguishable at 506 nm. The corresponding -∆Runmixed maps display no detectable thermal change. Spectral unmixing has extended the dynamic range of thermoreflectance measurements even to currents beyond the luminescence threshold. It enabled us to investigate the thermal properties of a live LED sample.

Fig. 5. [(a)–(c)] Luminescence map (480 × 480 μm2) of surface emitting LED at different wavelengths. [(d)–(f)] Corresponding unmixed +∆R maps of the same field of view.

4. Conclusions

We have demonstrated a versatile and inexpensive spectral microthermography technique for identifying hot and cold spots in the active layer of a biased IC sample using the information that is contained in their thermoreflectance signal R(λ; x, y). Two semiconductor sites of different material composition can be distinguished by their dissimilar R(λ; x, y)’s. Two thermoreflectance gradient maps ±∆R were generated from a pair of R(λ; x, y)’s taken under unbiased and biased conditions of the IC sample. The ±∆R maps were used to locate hot (cold) spots where heat accumulates more rapidly (slowly) than the average rate of the entire active layer. Knowledge of the hot and cold spot locations is crucial in assessing the thermal integrity of a layer structure because defects are more likely to develop at hot spots [7

7. J. Miranda and C. Saloma, “Four-dimensional microscopy of defects in integrated circuits,” Appl. Opt. 42, 6520 (2003). [CrossRef] [PubMed]

]. We have also utilized the technique to measure the thermal gradient maps of the surface of a biased LED sample demonstrating the efficacy of our technique to light emitting devices where the reflectance signal is accompanied by a strong electroluminescence background.

Acknowledgments

Grants from OVCRD-University of the Philippines Diliman and PCASTRD-Dept of Science and Technology Philippines. Condensed Matter Physics Laboratory (NIP) for the Au-GaAs sample.

References and links

1.

G. Tessier, G. Jerosolimski, S. Holé, D. Fournier, and C. Filloy, “Measuring and predicting the thermoreflectance sensitivity as a function of wavelength on encapsulated materials,” Rev. Sci. Instrum. 74, 495–499, (2003). [CrossRef]

2.

G. Tessier, S. Holé, and D. Fournier, “Ultraviolet illumination thermoreflectance for the temperature mapping of integrated circuits,” Opt. Lett. 28, 875–877, (2003). [CrossRef] [PubMed]

3.

J. Kolzer, E. Oesterschulze, and G. Deboy, “Thermal Imaging and Measurement Techniques for Electronic Materials & Devices,” Microelectron. Eng 31, 251–270, (1996). [CrossRef]

4.

G. Gosch, ed., Handbook of Thermo-Optic Coefficients of Optical Materials with Applications (Academic, San Diego, Calif, 1998).

5.

J. Christofferson and A. Shakouri, “Thermal measurements of active semiconductor micro-structures acquired through the substrate using near IR thermoreflectance,” Microelectron. J. 35, 791–796, (2004). [CrossRef]

6.

V. Daria, J. Miranda, and C. Saloma, “High-contrast images of semiconductor sites via one-photon optical beam-induced current imaging and confocal reflectance microscopy,” Appl. Opt. 41, 4157 (2002). [CrossRef] [PubMed]

7.

J. Miranda and C. Saloma, “Four-dimensional microscopy of defects in integrated circuits,” Appl. Opt. 42, 6520 (2003). [CrossRef] [PubMed]

8.

V. J. Cemine, B. Buenaobra, CM. Blanca, and C. Saloma, “High contrast microscopy of semiconductor and metals sites in integrated circuits via optical feedback detection,” Opt. Lett. 29, 2479–2481 (2004). [CrossRef] [PubMed]

9.

CM. Blanca, V.J. Cemine, V.M. Sastine, and C. Saloma, “High-resolution differential thermography of integrated circuits with optical feedback laser scanning microscopy”, Appl. Phys. Lett. 87, 231104-(1–3) (2005). [CrossRef]

10.

C. Saloma, A. Tarun, M. Bailon, and M. Soriano, “Rapid subsurface detection of nanoscale defects in live microprocessors by functional infrared emission spectral microscopy” Appl. Opt. 447302–7306 (2005). [CrossRef] [PubMed]

OCIS Codes
(110.6820) Imaging systems : Thermal imaging
(120.6200) Instrumentation, measurement, and metrology : Spectrometers and spectroscopic instrumentation
(180.0180) Microscopy : Microscopy

ToC Category:
Imaging Systems

History
Original Manuscript: October 24, 2005
Revised Manuscript: February 1, 2006
Manuscript Accepted: February 1, 2006
Published: February 6, 2006

Citation
Godofredo Bautista, Carlo Mar Blanca, Serafin Delica, Bernardino Buenaobra, and Caesar Saloma, "Spectral microthermography for component discrimination and hot spot identification in integrated circuits," Opt. Express 14, 1021-1026 (2006)
http://www.opticsinfobase.org/oe/abstract.cfm?URI=oe-14-3-1021


Sort:  Author  |  Year  |  Journal  |  Reset  

References

  1. G. Tessier, G. Jerosolimski, S. Holé, D. Fournier, and C. Filloy, "Measuring and predicting the thermoreflectance sensitivity as a function of wavelength on encapsulated materials," Rev. Sci. Instrum. 74, 495-499, (2003). [CrossRef]
  2. G. Tessier, S. Holé, and D. Fournier, "Ultraviolet illumination thermoreflectance for the temperature mapping of integrated circuits," Opt. Lett. 28, 875-877, (2003). [CrossRef] [PubMed]
  3. J. Kolzer, E. Oesterschulze, and G. Deboy, "Thermal Imaging and Measurement Techniques for Electronic Materials & Devices," Microelectron. Eng 31, 251-270, (1996). [CrossRef]
  4. G. Gosch, ed., Handbook of Thermo-Optic Coefficients of Optical Materials with Applications (Academic, San Diego, Calif., 1998).
  5. J. Christofferson and A. Shakouri, "Thermal measurements of active semiconductor micro-structures acquired through the substrate using near IR thermoreflectance," Microelectron. J. 35, 791-796, (2004). [CrossRef]
  6. V. Daria, J. Miranda, and C. Saloma, "High-contrast images of semiconductor sites via one-photon optical beam-induced current imaging and confocal reflectance microscopy," Appl. Opt. 41, 4157 (2002). [CrossRef] [PubMed]
  7. J. Miranda and C. Saloma, "Four-dimensional microscopy of defects in integrated circuits," Appl. Opt. 42, 6520 (2003). [CrossRef] [PubMed]
  8. V. J. Cemine, B. Buenaobra, C.M. Blanca and C. Saloma, "High contrast microscopy of semiconductor and metals sites in integrated circuits via optical feedback detection," Opt. Lett. 29, 2479-2481 (2004). [CrossRef] [PubMed]
  9. C.M. Blanca, V.J. Cemine, V.M. Sastine, C. Saloma, "High-resolution differential thermography of integrated circuits with optical feedback laser scanning microscopy", Appl. Phys. Lett. 87, 231104 (2005). [CrossRef]
  10. C. Saloma, A. Tarun, M. Bailon and M. Soriano, "Rapid subsurface detection of nanoscale defects in live microprocessors by functional infrared emission spectral microscopy" Appl. Opt. 44, 7302-7306 (2005). [CrossRef] [PubMed]

Cited By

Alert me when this paper is cited

OSA is able to provide readers links to articles that cite this paper by participating in CrossRef's Cited-By Linking service. CrossRef includes content from more than 3000 publishers and societies. In addition to listing OSA journal articles that cite this paper, citing articles from other participating publishers will also be listed.


« Previous Article  |  Next Article »

OSA is a member of CrossRef.

CrossCheck Deposited