IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 48, NO. 1, FEBRUARY 1999 69 Temperature Measurements of Metal Lines Under Current Stress by High-Resolution Laser Probing V´ eronique Quintard, Stefan Dilhaire, Tam Phan, and Wilfrid Claeys Abstract— Accelerated reliability tests of aluminum intercon- nections with respect to electromigration are often performed by high current density stressing of particular structures designed to promote failures of this kind. In order to derive predictions of the interconnect lifetime under normal operating conditions, it is essential to know the temperature of the test structure when high current-density stressing conditions are applied. In this paper, we present the first experimental temperature measurements with two high spatial resolution laser probes of metal lines. The two optical probes provide an indirect access to the temperature of the metal line. The calibration of the probes for temperature mea- surements is performed by comparison to electrical temperature measurements. The probes have a lateral resolution of 1 m. Two types of test structures, used in accelerated electromigration analysis, are studied. The measurements are compared with two theoretical simulations. Good agreement is obtained only when a particular value of the thermal conductivity of the submicrometer oxide insulation layer of the metal line is used. This is not surprising, since the thermal properties of submicrometer layers are size dependant. Our experimental method provides an elegant way to determine these parameters. Index Terms—Interconnections, interferometry, laser applica- tions, reliability, temperature measurement, thermal variables measurement, thermoresistometry. I. INTRODUCTION W ITH THE increasing complexity of integrated circuits and also with the reduction in size of the components, the limiting factor for circuit lifetime is the aluminum intercon- nects failure. The principal cause of this is electromigration, which is a mechanism of mass transport induced by electric current. This degradation process can be accelerated by the increase of temperature or current density in the metal line. In order to establish a fabrication process, quick evaluation techniques and test structures for electromigration have been developed. If we want to predict the interconnect lifetime under normal operating conditions from such accelerated tests, it is essential to know the temperature of the test structure when high current-density stressing conditions are applied. The lack of a correct determination of these temperatures is one of the principal limitations in the evaluation of component lifetime with accelerated tests. Until now, only modeling was available to estimate the temperature of metal lines with nonuniform geometry [1], [2]. To our knowledge, no Manuscript received January 27, 1996; revised January 11, 1999. The authors are with the University of Bordeaux, 33405 Talence, France (e-mail: wclaeys@frbdx11.cribx1.u-bordeaux.fr). Publisher Item Identifier S 0018-9456(99)02846-6. experimental verifications of the theoretical calculations of temperature of nonuniform micrometric structures as SWEAT, have been done. In this work, we present an original experimental method for the determination of the temperature profile along thin metal lines under current density stress. The measurement is provided by two laser probes we have built in the laboratory: one is a high resolution interferometer, the other one is a reflectometer. Optical measurements can be performed on lines as narrow as two micrometers because the highly focused laser beam has a beam diameter smaller than one micrometer. In this paper we explain how optical measurements allow the determination of temperature for a given current. Because our measurements are indirect temperature measurements, they need to be calibrated. After a description of our two electromigration test struc- tures, we present reflectometric and interferometric measure- ments of test structures of uniform geometry. Then, we present electrical measurements that we use as temperature calibration for these lines. Finally, we apply our method to determine the temperature of structures with nonuniform geometry. All measurements are correlated to simulations. Furthermore, we raise the problem of using macroscopic parameter values for microscopic structures in simulations. II. ELECTROMIGRATION TEST STRUCTURES Patterns under test are ASTM (Standard guide F1259 M-96; Annual book of ASTM Standards, vol. 10.04) and SWEAT (Standard wafer level electromigration accelerated test) struc- tures. Both devices are made of non passivated Al (99.5%) Cu (0.5%) metallization on a SiO insulator layer. The ASTM test structure is a uniform long straight line (800 m), 3 m wide, often used in long-term lifetime testing [see Fig 1(a)]. In contrast, the SWEAT test structure, as shown in Fig. 1(b), contains several alternate narrow and wide regions connected in series by tapered sections at 45 with respect to the main axis. This type of sample has been proposed as a rapid wafer-level production monitor of metallization quality [3]. The metallization structures are stressed at high current density (greater than A/cm ) and high temperature generated by joule heating (typically 100 C). The particu- lar geometry is supposed to accelerate the electromigration mechanism by allowing the formation of high temperature and current density gradients, which should promote the formation of voids and hillocks during lifetime testing. 0018–9456/99$10.00  1999 IEEE