Conference Paper

Measurements of effective thermal conductivity for advanced interconnect structures with various composite low-K dielectrics

IBM Microeletronics Div., Essex Junction, VT, USA
DOI: 10.1109/RELPHY.2004.1315303 Conference: Reliability Physics Symposium Proceedings, 2004. 42nd Annual. 2004 IEEE International
Source: IEEE Xplore


Accurate specification of design groundrules for interconnect systems requires knowledge of the thermal behavior of the systems. A key parameter that characterizes the thermal behavior is the thermal conductivity of the inter-level dielectric (ILD). In practical VLSI applications, the metal interconnects are fully embedded in a stacked, composite ILD media, which presents difficult challenges for the accurate determination of thermal conductivity. In this paper, we propose the concept of an "effective thermal conductivity" to model such complicated, composite media, and introduce a simple methodology to accurately measure effective and bulk thermal conductivities of various thin dielectric layers in integrated circuits. We present measured effective conductivities of several composite media, including various Cu/low-k dielectric configurations such as Cu/SiCOH, Cu/SiLK®, Cu/fluorinated silicate glass (FSG), and a hybrid stack with Cu lines in SiLK® and Cu vias in un-doped silicate glass (USG). Measurements were recorded in the temperature range from 30°C to 120°C using a unique combination of fully embedded Cu lines as heater/thermometers, wafer-level temperature vs. power (TVP) measurements, and the Harmon-Gill (H-G) quasi-analytical heat conduction model. The thermal conductivities of all the films studied here were observed to increase with rising substrate temperature.

23 Reads
  • [Show abstract] [Hide abstract]
    ABSTRACT: When measuring the best linear approximation of systems suffering from nonlinear distortions a specific class of periodic multiharmonic signals is normally used. These are signals with uniformly distributed random phases, termed random phase multisines. In this paper, it is shown that measurements of the best linear approximation of nonlinear systems can also be obtained by using a special type of low crest factor multisines. These signals are compared to random phase multisines and their properties are analysed in detail.
    No preview · Conference Paper · Feb 2002
  • [Show abstract] [Hide abstract]
    ABSTRACT: As dimensions continue to shrink and device densities increase, power and heat dissipation become an ever-increasing challenge. In this work, we investigate heat flow ramifications for a variety of very simple patterns. We start by describing the use of a finite element method (FEM) tool. This includes a brief description of the modeling procedure. The model results, for a given set of boundary conditions, are then compared to the physical measurements for that structure. Excellent agreement is demonstrated, thus calibrating the model. We then extend this model to structures for which we have no physical measurements. As importantly, we extend this model to structures and patterns that may be desirable for the next generation.
    No preview · Conference Paper · Nov 2004
  • [Show abstract] [Hide abstract]
    ABSTRACT: As the dimensions of IC structures shrink and dissipated power densities increase, thermal considerations have a growing importance in the development of advanced microelectronic components. Optimal thermal management requires the accurate knowledge of the thermal conductivities of their constitutive thin films. Actually, a precise knowledge of these material parameters is essential to predict the thermal behavior of the IC and then to take it into account in reliability issues. The present paper provides an analytical thermal resistance model used to extract the conductivities of fluorinated silicate glass (FSG), phosphorous silicate glass (PSG) and silicate carbide oxide (SiOCH). Joule heating measurements at 25°C performed on embedded copper lines have validated this model. Various dielectric stack configurations have been studied to isolate the contribution of each material in the thermal model. From these results, root mean square (rms) currents have been predicted to limit Joule heating in interconnects.
    No preview · Conference Paper · Nov 2004
Show more