緩衝材料在深層處置概念中，為放射性廢料桶與天然母岩間之緩衝介質；主要用途是包覆封存放射性廢料之廢料桶，並維持工程障壁系統之功能。一般來說緩衝材料需具備有良好熱傳導性質，以將放射性核種衰變過程中釋放之熱能快速傳遞至周圍母岩，而非將熱能積聚在廢料桶內，造成廢料桶毀損使核種釋出；目前各國對緩衝材料之選擇，依據其資源條件而有所不同，但研究結果發現，膨潤土與砂－膨潤土混合材料，具良好力學特性與熱學性質，故常被用來作為緩衝材料候選材料。 本研究旨在探討緩衝材料之熱傳導性質，藉由改良暫態熱探針法，發展出熱探針連續量測方法，消除ASTM規範D5334建議方法之熱探針與試體顆粒間接觸界面不均勻，以及劉俊志（2003）提出埋入式量測法試料需求量大且各試體間具變異性問題。連續量測方法將熱探針預先置入土體中，並在壓製膨潤土所用模具中預留通道，使導線可由通道延伸而出，藉由控制壓實膨潤土體之壓桿高度以達需求乾單位重。本試驗針對乾單位重在14kN/m3至18kN/m3間之夯實膨潤土－花崗岩碎石試體進行熱傳導係數之量測；由於利用同一試體進行不同乾單位重、含水量及花崗岩碎石含量下的連續量測，所以能消除材料間之變異性。試驗數據處理方面，利用等含水量法及等密度法成分體積比之概念，將試驗數據與SCS、DS及De Vries and Campbell model等熱傳導係數預測模式所得值比較，並提出Modified De Vries and Campbell model結合微觀力學模式預測方法，以評估各預測模式適用性。 Buffer materials are filled between radioactive wastes canisters and host rock vault. The main purpose is to enclose the canisters and to exhibit the performance of engineering barrier system. In general, buffer material should have well thermal conductivity to ensure the decay heat transported to the host rock instead of stock up in canisters result in erosion or corruption of it. Among the candidates of buffer materials in many other researches, we can found that mixture of bentonite and crushed granite has good mechanical properties and thermal characteristics. The study aims to investigate the thermal properties of buffer materials. An improved continuous heat probe method for thermal conductivity measurement is proposed. It avoided the surface contact problems inhibited in ASTM D5334 method and the individual differences of measurement from different specimens. In this new method, the probe is put in the center of the specimen with the signal cord laying out undergoing the compaction process. The dry density of the specimen is controlled by the piston of the mold in the dry density range of 14kN/m3 to 18kN/m3, and the thermal conductivity of different density can be measured. With concept of micromechanics applied, the results are analyzed to find the prediction model of the relationship between thermal conductivity and the properties including water content , dry density of clay, and the fraction of granite. It shows that the affection of water content on thermal conductivity could not be analyzed with micromechanics models, and should be predicted with regression. Buffer material is divided into two phases, one is bentonite with water content, the other is granite fraction. The prediction model for thermal conductivity in granite with bentonite is predicted by modified De Vries and Campbell model, then micromechanics model is applied to find the final prediction of the mixture of the two phases.