Abstract
Achieving sustainability in production systems requires the integration of environmental, economic, and social objectives across their entire lifecycle. This is one of the main aspirations that this research considers as driving the adoption of Digital Twins (DTs). However, while DTs are increasingly adopted to support sustainability goals, most implementations remain focused on local optimization within specific lifecycle phases; with limited integration across phases and sustainability pillars. In this context, the emerging concept of the Cognitive Digital Twin (CDT) offers the potential to enable more adaptive and context-aware decision-making; yet its cognitive dimension remains under-structured in sustainability-oriented applications. This study first develops a structured conceptualization of CDT cognition through a capability-feature distinction differentiating internal cognitive processes from observable intelligent behaviors. Building on this construct, a Systematic Literature Review analyzes DT applications exhibiting cognitive features in production systems, mapping lifecycle coverage, sustainability pillars addressed, enabling technologies, and the evolutionary integration of cognitive features. The literature findings reveal a predominance of Middle of Life applications and environmental-economic coupling, with limited lifecycle continuity and only partial cognitive integration. To address these limitations, the paper proposes an integrative framework for sustainability-oriented CDTs, linking enabling technologies, cognitive capabilities, cognitive features, and sustainability services across lifecycle phases and multiple pillars. An empirical application roadmap further clarifies how the framework can be operationalized in industrial contexts. The framework posits how structured cognition within DT frameworks can support multi-pillar and lifecycle-spanning sustainability. Overall, this study reframes sustainability-oriented CDT development as an architectural and cognitive design challenge rather than a purely technological one, and identifies key open challenges with associated managerial implications and research directions.