Limitation in article references

When working with single journal articles, I noticed that the number of references in the CrossRef article metadata may in some cases be incomplete as compared to the published article.

For example the published article DOI:10.1103/PhysRevB.56.6100 has 18 references, but the Crossref metadata will only list 16. Missing are references to 2 handbooks which could be listed with a proper ISBN.

Not all reference DOIs in the CrossRef article metadata may be resolved. For example this article DOI:10.1063/1.358185 should reference in CrossRef metadata key->r2 this DOI:10.1103/PhysRevB.56.6100, but is unable. The article DOI is in the CrossRef dataset though.

When conducting automated bibliographic research, solving the inaccuracies may be relevant.

Hi Dave,

Thanks for your post.

It’s up to each publisher to determine whether they supply references with their publications’ metadata records, and if they do opt to supply them, how they do it.

Obviously, we strongly encourage our members to provide references as comprehensively as possible, and we do our best to support them in doing that and to communicate the benefits to them and the whole scholarly community of doing so. But, nonetheless, it isn’t strictly required.

So, unfortunately, the only answer we have for why there are sometimes discrepancies or omissions is: that’s what the publisher (or a service provider working on their behalf, such as a hosting platform) submitted.

With regard to your first example, we do sometimes see a pattern where only references that the publisher is able to match to registered DOIs are submitted. That’s not best practice, and not what we would recommend. But, like any other metadata quality issue, it can only be handled on a case-by-case basis. So, we’d encourage you to report that issue to American Physical Society - as the publisher of 10.1103/PhysRevB.56.6100 - and if you’d like us to forward your concerns or comments along to our contacts at the publisher, we can do that as well.

I’m not sure what you mean here. Could you clarify? Which DOI is the citing item and which is the item being cited? And for the citing item, can you copy in the relevant reference from its reference list? I’ll be happy to take a closer look.


Thank you Shayn,

The metadata of the article with this DOI:10.1063/1.358185 should reference this DOI: 1103/PhysRevB.56.6100. The metadata from the article (first DOI) does not include the DOI and so the reference cannot be found with automation.

The REST API will give:
{“key”:“r2”,“first-page”:“12”,“volume”:“45”,“author”:“Qi Q.”,“year”:“1992”,“journal-title”:“Phys. Rev. B”,“ISSN”:“”,“issn-type”:“print”}, but no DOI.



Thanks for those details, Dave.

References don’t have to include the DOI. They can be formatted as structured or unstructured citations, without the DOI included, even if a DOI is registered for that item.

You can see the options for reference metadata detailed here.

All that said, 10.1103/PhysRevB.56.6100 is registered for this publication:
Vandormael, D., Grandjean, F., Briois, V., Middleton, D. P., Buschow, K. H. J., & Long, G. J. (1997). X-ray-absorption spectral study of theR2Fe17−xMxsolid solutions (R=Ce, Nd andM=Al, Si) Physical Review B, 56(10), 6100–6106.

The reference you’re pointing to in the metadata for 10.1063/1.358185 seems to referring to a completely different item.

<citation key="r2">
<journal_title>Phys. Rev. B</journal_title>
<author>Qi Q.</author>

The journal is the same, but it’s a completely different author, publication year, volume number, etc.

I can’t access the reference list on the landing page for 10.1063/1.358185 because it’s behind a paywall, so I don’t have a way to check whether the reference that’s been deposited to Crossref reflects the reference in the publications reference list accurately.


Thank you again Shayn,

You are correct with the incorrect reference DOI in my post. My apologies, there were too many DOIs in my script notes and I copied the wrong one.

The DOI should be: 10.1103/PhysRevB.45.12278 which is in CrossRef. Phys rev B is a renowned journal in physics.

The reference you showed in the XML code does not include a DOI and is also incomplete, therefore extremely hard to find again.

0163-1829 Phys. Rev. B Qi Q. 12 45 1992

The “classic” print reference associated with the XML code is:
Q. Qi, H. Sun, R. Skomski, and J. M. D. Coey, Phys. Rev. B 45, 12 278 (1992)
Now one can find the reference again.

Would it be great if the CrossRef algorithms would match each work with a unique identifier e.g. DOIs for articles and some books or ISBNs for older books?

In that case I can use scripts to resolve the metadata of references without any manual intervention through Scopus or WoS.



Thank you Shayn,

we’d encourage you to report that issue to American Physical Society - as the publisher of 10.1103/PhysRevB.56.6100 - and if you’d like us to forward your concerns or comments along to our contacts at the publisher, we can do that as well

If you would be so kind to forward my concerns to APS and AIP, see text below:

DOI: 10.1103/PhysRevB.56.6100 (Phys. Rev. B)
Publisher: APS
Total references

  • in print: 18
  • in CrossRef metadata: 16

missing two references to handbooks

DOI: 10.1063/1.358185 (J. Appl. Phys.)
Publisher: AIP
Total references

  • in print: 25
  • in CrossRef metadata: 20

missing 5 references in CrossRef
Second reference in CrossRef is incomplete i.e.
0163-1829 Phys. Rev. B Qi Q. 12 45 1992
and cannot be resolved without going to the print publication i.e.:
Q. Qi, H. Sun, R. Skomski, and J. M. D. Coey, Phys. Rev. B 45 , 12 278 (1992)

These were ad-hoc selections of articles with missing reference metadata. Imagine how many I may miss out when I gather reference metadata of hundreds of articles through the REST API.

Thanks again Shayn,


As far as we’re concerned, that reference is not incomplete. We don’t require a DOI for a reference to be complete. It’s optional. While including the DOI might be ideal, the structured or unstructured citation is sufficient.

We do match references with DOIs where possible. That’s how our Cited-by service supplies citation data to our members for their own publications, and that’s how we derive the is-referenced-by counts for each item. But those matches are not inserted into the public metadata record. The public record only consists of the metadata supplied by the item’s publisher directly.

Sure, I’ll be happy to pass that along. You may want to contact them directly as well for the fastest possible response.

1 Like

Hi Shayn,

Thank you so much.

I will do the same and pass on the concerns to AIP and APS.

have a great day.



1 Like