Architectural design patterns (ADPs) are architectural solutions to common architectural design problems. They are an important concept in software architectures used for the design and analysis of architectures. An ADP usually constrains the design of an architecture and, in turn, guarantees some desired properties for architectures implementing it. Sometimes, however, the constraints imposed by an ADP do not lead to the claimed guarantee. Thus, applying such patterns for the design of architectures might result in architectures which do not fulfill their intended requirements. To address this problem, we propose an approach for the verification of ADPs, based on interactive theorem proving. To this end, we introduce a model for dynamic architectures and a language for the specification of ADPs over this model. Moreover, we propose a framework for the interactive verification of such specifications based on Isabelle/HOL. In addition we describe an algorithm to map a specifi cation to a corresponding Isabelle/HOL theory over our framework. To evaluate the approach, we implement it in Eclipse/EMF and use it for the verification of four ADPs: variants of the Singleton, the Publisher-Subscriber, the Blackboard pattern, and a pattern for Blockchain architectures. With our approach we complement traditional approaches for the verification of architectures, which are usually based on automatic verification techniques such as model checking.