Is it correct to represent a foreign key in a conceptual data model?

Experts share this. Here is my opinion.

The classic ER (Entity-Relationship) model identifies relationships but does not implement them as foreign keys. In an ER diagram, relationships are displayed as lines between boxes. Additional labels classify them as many-many or many-one, and as required or optional. The relational model always uses foreign keys to identify and implement relationships. The relational model diagram will use the lines between boxes for visual purposes, but the complete list of attributes will include these foreign keys.

Another difference between ER modeling and relational modeling is that in relational modeling, many relationships will appear as a box with two (or more) lines leading to it. In ER modeling, the binary relation of many, many models is modeled without its own box.

I've always used a conceptual model using ER modeling and left out foreign keys. Then, during logical design, I switched to the relational model and hooked up foreign keys. At this time, many, many relationships will receive additional boxes to hold a pair (or tuple) of foreign keys.

Many experts don't follow my practice. They believe that the ER model is overkill for conceptual modeling and that it is best to start with the relational model at the conceptual level while you are still doing the requirements analysis. Then you don't have to switch modeling conventions in the middle of the stream. From a theoretical point of view, they are right.

In cases where the ER model has worked well, the model for inexperienced customers is explained. Some of them have a faster intuitive understanding of the ER model than the relational model.