What is the minimum number of research articles that must be cited to write a review article?

Your article needs to be interesting, the number of articles is secondary. If a field is very new than a review of a few articles could be interesting. However, if there are only 20 artcles it could also mean it is just too early to take stock, and your sub-sub-field is not ready for a review article yet. The purpose of a review article is to find partial answers on which there is some form of consensus, and identify open questions and new directions of research. Is your field sufficiently ripe for that?

Also look outside your discipline. It may be that there is a longstanding research tradition dealing with that issue in another discipline.

The minimum papers you need to review will depend on two things:

• Your research the field (which might be relatively small or large)
• The depth vs breadth of your review

In computer science, there have been quality review papers in highly respected venues which only focus on around 10 papers, but where the review lacked in breadth, it made up for in depth, which can be more important if you are looking to make a valuable contribution to the field.

An example is this paper published in TSE.

Asides the depth of what would be discussed, the area and its level of maturity, the quality of the survey and the definition of its scope also matters. A number of recent and well known literature can be considered while highlighting major strengths and limitations of each proposal as well as new interesting insights that could be considered in the future. The point is that the contribution of the paper should be glaring to the reader of the paper. Also, it may be highly required to properly define the criterion for choosing the articles you are considering if you have chosen a sub-set of literature from a particular scope. This should also reflect in the title and (or) abstract of your paper. Another very important point is the style of organizing your survey paper. For instance with 20 main references you can still write a survey with over 80 references. The other 60 references could be used to strengthen specific points within the paper as well as other secondary issues of importance within the paper. In this case, it is highly recommended you show the readers what value can be derived from those 20 references by showing a summary table with strengths and limitations, salient differences, metrics, qualitative/quantitative comparison etc. or illustrations and charts, all leading to more intuitive questions to be answered in the future.