*The following is a guest post by Stephen Glasby.*

In July 2012 Tim Gowers wrote A new open-access venture from Cambridge University Press. The idea of arXiv overlay journals has been around for over 4 years, and there are very powerful ideas to support them. Why then does the mathematical community seem reluctant to support them?

Let me begin with a some history and context.

It is hard to overstate the importance of Mathematics Reviews (MR) and Zentralblatt (Zbl). Although Zbl had been reviewing journal articles since 1931, antisemitic pressures at that time led to the establishment of MR in 1940 (the last printing of MR was 2012). The electronic database MathSciNet, established in 1996, has made the printed versions of both MR and Zbl obsolete. We all know that MathSciNet is not merely a list of reviews: it contains links to journal articles, authors, references, citations, and related reviews. MR employs about 100 people in An Arbor, Michigan, and these few people have transformed the way we do mathematics. Mathematics is fortunate to have both MathSciNet and the arXiv. The natural sciences, by contrast, do not have an equivalent of MathSciNet!

While MathSciNet could be further improved and extended, I will change tack and now focus on the Open Access Problem: that it is not free to publish/access/reuse research papers. This larger problem affects all of Mathematics. The Open Access Problem applies more generally to the sciences and medicine and it is described, with some solutions, in this broad context in the illustrated video. (Who owns your personal data is a very large issue.) Let us focus on mathematical Open Access problems and solutions, for mathematics has both the arXiv and MathSciNet. When we locate relevant research using MathSciNet, increasingly often, clicking on “article” to download the journal article shows that the article is “owned” by a publishing company, or a learned society, and downloading can cost US$40, or an annual subscription! This is a growing and lasting problem for mathematics, retarding both future research and proper referencing of past research.

One solution to the Open Access Problem, suggested by Tim Gowers and others, is to use arXiv overlay journals. Why is this practical solution not being implemented broadly? Such journals exist, but there are relatively few. I argue below that the solutions are easily implemented, and require essentially no additional effort, and have huge potential benefits. But they do require the mathematical establishment to embrace the change. Let us first look at problems and solutions before addressing the resistance to change.

There are many models for peer-reviewed Open Access publication, these have names such as Platinum, Gold, Diamond, and Green, but non-experts forget the definitions or use non-standard definitions. It seems better say whether articles are free to Publish, Access, Reuse, Typeset and Edited, etc. We will focus on journals that are free to Access (i.e. read and print), and free (or very cheap) to Publish and Reuse (i.e. link to, or republish, content). Clearly Typesetting and Editing are lesser problems for mathematics than the accessibility of past research.

**Problem 0:** A small fraction of new mathematical research is put on the arXiv before it is published, and this will remain the case for some time.

**Solution:** Establish journals that do publish on the arXiv, see Problem 2. The arXiv was established in 1991. It has 1.2M papers up to 2016, most are physics, and about 280,000 are in mathematics. MathSciNet lists 2.1M reviews from 1991-2016 so I estimate (very crudely) that about 1/7 of mathematical research papers are on the arXiv. The number of arXiv papers was doubling every 4 years, but this rate is slowing down. It is possible that in a decade the majority of math research papers will have preprint versions posted on the arXiv.

**Problem 1:** Many of the prestigious journals are owed by publishing companies, or learned societies, which make our research effectively inaccessible because of copyright, and cost of access. The cost of an average mathematics journal is US$1,700/yr and some papers have indefinite copyright. Subscription costs for some journals with shorter copyright periods can cost over US$8,000/yr (e.g. Elsevier’s Nonlinear Analysis).

**Solution:** Establish new journals that publish for “free” on the arXiv, see Problem 4. The research will be accessible and free (to read) in perpetuity. Moreover, it will be clear that the paper has been peer-reviewed, and MathSciNet can link to the refereed arXiv paper.

**Problem 2:** New journals have lower impact factors than expensive established journals, and academics must publish in high impact journals to be promoted, to obtain grants, etc.

**Solution:** If all mathematicians put the refereed and corrected version of their paper on the arXiv (even if it lacked the journal formatting) then this problem would be solved. As remarked above, this is not likely to happen soon. Another solution is to ask the editors of Journal X to resign en masse and establish a new journal called Journal EX (for Electronic Journal of X). The editors will review to the same standard as before so Journal EX must have the same academic standing as Journal X, see Problem 5.

**Problem 3:** Journal X has Editorial Management software, the new Journal EX would require similar software to be developed and maintained.

**Solution:** Free Editorial Management software already exists. Learned societies, or grants, or nominal “Publication Fees” could subsidize the cost of maintaining and further developing the software, see Problem 4.

**Problem 4:** There are unavoidable costs that must be born by the arXiv and MathSciNet. Who will pay for these?

**Solution:** There are many solutions here. The major costs would be refereeing and editing, but we perform these gratis, and distribution via the arXiv is close to free, so that leaves copy editing and maintaining MathSciNet databases. A growing number of universities have pledged to support fees for Open Access publication for articles written by their faculty, see this link. A nominal Publication Fee could be levied from author(s), or author(s) may be required to donate a nominal sum to a fund to maintain the arXiv, to keep MathSciNet subscriptions low, and to develop and maintain Editorial Management software, and maybe fund some copy editing. (The 2017 pricing for MathSciNet Consortia ranges from US$339.00 per institution to US$11,887.00: a fraction of library subscriptions costs for mathematics journals.)

**Problem 5:** Editors of high impact journals need not be concerned about egalitarianism and Open Access. Why should they resign en masse from the board of Journal X and form a new Open Access Journal EX?

The three main problems are: critical mass, will, and inertia. The work load of a editor is independent of whether s/he works for a journal that: exploits mathematical creativity, or one that fosters creativity (by rapid, free, accessible publications). I argue that editor intransigence is a major impediment to Open Access publication. I know of cases of individual editors whose libraries do not carry key journals because of cost, and yet these editors do not support Open Access. Why? Gowers suggested that a number of mathematicians have an emotional objection to Open Access publication.

A minority of mathematicians are editors for Open Access journals. Is this minority more concerned about the health of mathematical research than the majority? or are there cogent reasons for maintaining the status quo? I have tried to make new arguments for Open Access publishing in mathematics. A good source for further reading is Tim Gowers blog.