I am trying to model a use case diagram with two different front end applications. I have a web-based front end with different use cases than my front end for a tablet app. Some use case might have the same name but a different scenario.
My question is if it is a good way to use two system blocks in the diagram for each front end, or should I combine all uses cases in one system block? How can I model it in a UML correct and clear way?
Setting the right boundaries is key to successful use-case modelling. According to the UML 2.5.1 specs:
A UseCase is a kind of BehavioredClassifier that represents a declaration of a set of offered Behaviors. Each UseCase specifies some behavior that a subject can perform in collaboration with one or more Actors. UseCases define the offered Behaviors of the subject without reference to its internal structure.
What matters is the system (i.e. the "subject" in use-case speak), the actors, and the behaviors that the subject offers to its actors (which, as QwertySO reminds us, supposed to provide value to the users, i.e. help to achieve the user's goal).
From the point of view of your users, the expected functionality is independent of the from the front-endend. According to your comment, today:
the web frontend is targeting the actor admin with a use case: create a product. On the other side, the tablet front end is targeting the user to buy a product but is not made for the admin.
But tomorrow, your admins might want to use a PC, or a smartphone for the urgent tasks. And maybe some normal users would expect a voice interface rather than a tablet. The use-cases shall help us to focus on theses needs and not the way we are going to address them.
So, in principle, you should model your use-case independently of the front-end. You should not either show the split between front-end and back-end.
Use-cases are not meant to do functional nor structural decomposition. This kind of distinction is best performed in component diagrams that decompose your system into components and in deployment diagrams that map the components to the servers and devices.
If you have already a component decomposition in mind, and your component are independent, you could think of modelling the use-cases of a sub-system or a component. This is valid, since UML specs tell us:
A subject of a UseCase could be a system or any other element that may have behavior, such as a Component or Class. Each UseCase specifies a unit of useful functionality that the subject provides to its users You can even model use-cases for a component, the actors being then the external actors and the other
But would this be helpful? You could have a use-case diagram for the admin application, with admin actors, and a system actor fo the back-end. But I'd not advise to go into that direction unless the app is a rich app with its own added value. If the app is just a front-end you'd just end-up with very redundant use-case diagrams (i.e. front-end UC and back-end UC would show the same use-cases) with a loss of end-to end relation to the real actor (i.e. the back-end UC would be associated with system actors corresponding to the front-end app, but you'd loose the direct link to the real actors).
In the end you might lose the main strength of UC, the focus on users and their goals. In exchange you'd have the burden to keep all the redundant models in sync but I don't quite see the advantage here. If you want to have more arguments about each alternative, you may also have a look here.