Identifying chromatin domains (CDs) from Hi-C data is currently a central problem in genome research. Here we present Multi-CD (https://github.com/multi-cd), a unified method to discover CDs at various genomic scales. Integrating approaches from polymer physics, financial market fluctuation analysis and Bayesian inference, Multi-CD identifies the CDs that best represent the global pattern of correlation manifested in Hi-C, and reveals the multi-scale structure of chromosome. At each scale, the CDs are consistent with the results of existing methods, as well as with biological data from independent sources. CD solutions compared across different scales and cell types allow us to quantify the hierarchy between four major families of CDs, and to glean the principles of chromatin organization: (i) Sub-TADs, TADs, and meta-TADs constitute a robust hierarchical structure. (ii) The assemblies of compartments and TAD-based domains are governed by distinct organizational principles. (iii) Sub-TADs are the common building blocks of chromosome architecture. CDs acquired from our interpretation of Hi-C data using Multi-CD not only provide new insights into chromatin organization, but also offer a quantitative account for its cell-type-dependence and function.with γ ij = (σ ii + σ jj − 2σ ij )/2, where σ ij (= δr i · δr j ) is the positional covariance, determined by the topology of polymer network [42]. Indeed, distance distributions measured using fluorescence measurement between chromatin loci justifies this hypothesis (see Fig. S1). Importantly, our interpretation of chromosome conformation as a locally equilibrated, quasi-stable polymer network enables a one-to-one mapping of the contact probability p ij = rc 0 dxP (x; γ ij ) to the positional covariance σ ij , and hence to the cross-correlation matrix, (C) ij = σ ij / √ σ ii σ jj (see Methods). The cross-correlation matrix C normalizes the wide numer-