超低介電常數材料熱傳導係數之研究

Abstract

隨著製程技術的進步，散熱問題將會嚴重影響積體電路的性能與可靠度。三倍頻法(3ω method)可以簡單、有效地量得等向性與非等向性的熱傳導係數。它不僅可以減少量測上等待熱平衡的時間，也不需要精確量出電阻就可由一倍頻與三倍頻的電壓粹取出待測物的上升溫度。本論即利用自行建立之三倍頻法(3ω method)測量系統，分析多種次世代與次次世代低介電常數材料之熱傳導性質，並提出物理模型。 實驗的結果顯示非孔隙的有機旋塗介電質(Organic Spin-On-Dielectric, Organic SOD)本身熱傳導特性不佳，不適合以摻入孔隙的方式降低介電質常數。碳摻雜氧化矽(Carbon Doped Oxide, CDO)因為較佳的熱傳導特性與低介電常數，是極有潛力的次世代介電質材料。對作為銅擴散障礙層來說，非晶質的碳化矽 (a-SiC:H) 的熱傳導係數遠低於晶體狀態的碳化矽，也比氮化矽低三倍以上，但是因為厚度較薄且熱傳導係數比主介電質高，對散熱應該不會造成太大的影響。 摻入的孔隙在高孔隙度的二氧化矽(silica)中傾向於水平排列，造成非等向性熱傳導係數與非等向性介電質常數。我們提出一個序列-平行混合的模型(Serial-Parallel Hybrid model)來模擬垂直跟水平方向的，孔隙度對熱傳導係數與孔隙度對介電質常數的關係。不均勻排列的孔隙使得傳統的三倍頻法二維模型不適用，本論文利用序列-平行混合模型提出一個擷取水平方向熱傳導係數以及介電質常數的方法。 本論文建立三倍頻測量系統，結合自行發展的序列-平行混合模型，可以正確的擷取水平與垂直方向的熱傳導係數以及介電質常數，並可以分析孔隙分佈情況，對低介電常數材料的基本性質分析極為重要，也是進一步分析多層導線散熱以及訊號延遲的重要基礎。

With the progress of integrated circuit process technology, thermal management becomes an important issue to improve performance and reliability of integrated circuits. The 3ω method is easy, effective, and useful for extracting the thermal conductivity including the isotropic and anisotropic characteristics. This method not only costs less time to reach the thermal balance but also can extract the temperature rise on the heater/thermometry without achieving its exact resistance. In this thesis, the 3ω method measurement system was set-up and the thermal conductivity of next generation low dielectric constant materials was investigated. Several dielectric films were investigated, including PECVD silicon dioxide, PECVD silicon nitride, Organic Spin-On-Dielectric (Organic SOD), Carbon Doped Oxide (CDO), amorphous SiC:H, and porous silica in view of the thermal conductivity and dielectric constant. The non-porous Organic SOD material has high dielectric constant and low thermal conductivity. Hence, it is not suitable to scale down the dielectric constant by introducing pores into the film. The CDO material is the best dielectric solution because of its relatively low dielectric constant and relatively high thermal conductivity. The thermal conductivity of a-SiC:H is much lower than that of crystalline SiC. Fortunately, because of the thinner thickness and relatively higher thermal conductivity compared with main IMD, the a-SiC:H film does not play the dominant role on the thermal management. However, it needs more simulation and experiment to understand the temperature contribution of the SiC film within the realistic multilevel structure with thermal vias. The porous silica material has strong anisotropic characteristic in both thermal conductivity and dielectric constant. We propose a Serial-Parallel Hybrid model to explain the correlation between porosity and thermal conductivity and that between porosity and dielectric constant in both in-plane and cross-plane components. The pores in the higher porosity silica film tend to distribute horizontally. This distribution of the pores in the dielectric film is the main factor inducing the anisotropic characteristic of both thermal conductivity and dielectric constant. The non-uniform distribution of pores also defeat the 3ω method. A novel method based on the hybrid model was proposed to extract the in-plane thermal conductivity and dielectric constant of such film. The anisotropic characteristic of the thermal conductivity may be accompanied by the anisotropic dielectric constant, which is a greatly complicated problem in the RC delay simulation of the circuits. Efficient thermal and structural design of future generation IC with porous low-K materials should take these factors into consideration.