(Chinese version here, below is the English version)

Congratulations! You are at the point to summarize your scientific findings and put them into a research paper. You might have thought that this is simple. As a student, you have written many study reports – lab reports, reading reports, essay for courses, … However, a research paper is more important, serious, tricky in many ways that you may not have thought of.

Some reasons are as follows:

  • Every paper affects your career. This is unlike any study report that you have written. If you wrote a bad report in one course, it will not affect your other courses. If you wrote a bad report when you were undergraduate, it will not affect your GPA for your postgraduate study. But if you write one bad research paper, the hurt in reputation carries along forever. So make sure to be more serious.

  • Nobody has the responsibility to read your research paper . This is unlike any study report that you have written, where the instructor/TA has to read it even it is not well-written. Thus you need to make it useful, attractive, and most importantly, not boring.

  • You have to give accurate credits to what have been done before by other researchers. This is unlike any study report that you have written, where nobody gets hurt if you do not cite references properly. But in a research paper, the way you write is part of the whole scientific evaluation process of all researchers. By improperly addressing (or missing) someone’s contributions, you are hurting the individual (though you may not have done it by purpose) .

♡: Except your supervisor. But clearly the purpose of writing a research paper is not for your supervisor’s reading, but other researchers.

♠: True, in the same improper process you may make someone unreasonably happy. But psychologically, the hate you get will be more than the love. So the effect will not average to zero. And more importantly, one should be responsible for making the scientific evaluation process as fair as possible.

Having understood the differences, let me expand the above points with detailed instructions.

Not just listing what you have done

This is the biggest issue of a typical new research paper writer. You may have written progress reports by directly listing what you have done, and got nice grades. But this time it is different. Remember that the final readers of your research paper will not be a course instructor or TA, but other researchers. They don’t have to read your paper. Your paper must be attractive to them. (Well, if your work is something that researchers in your field have to read, otherwise they are outdated, then forget what I said. But 99% of the research papers are not at this level.)

How to “sell” your work?

Think about (and discuss with your supervisor) the entry angle that you introduce the problem before start to write the paper. Practically and technically, you may have followed some references, and then generalized their method of calculation for your purpose. But typically, the paper should not be written following this logic – how you have followed and generalized your reference. Emphasize the physical motivation and result throughout the paper. You do need to give proper credit to the method that you have learned from the reference. But “generalizing the method” is usually not the key point of your new paper. Do not write the paper like a technical generalization in most cases, unless your work is nothing more than a technical generalization (but at that time likely you shouldn’t do the research at all). Discuss this point with your supervisor and get an agreement on the emphasize of writing before you start.

Big picture vs details

Although you may have spent 90% of your time in the computational and very technical details, most of your readers spend 90% of their reading time in learning the conceptional idea and the rest 10% for getting the rough method of calculation (unless the reader wants to follow your calculation to do some further work based on your work). Indeed, the technical details must be correct, and should be clear enough to follow. But never forget to show the big picture, the motivation, and how to put your work into context as clearly as you can. And as a physicist instead of an applied mathematician, you should indeed learn the motivation and the grand picture by reading and thinking. You will be sooner or later independent and your ideas and depth of thinking will then be the key difference between you and your peers with comparable technical strength.

Pay more attention to where readers care more

Indeed one should try our best everywhere in the paper. However, one always need to properly distribute one’s time in writing different parts of the paper. When distributing your efforts, you need to consider how likely/carefully different parts will be read by other researchers.

  • Title. It is read by almost everyone who read your paper, and most of them will judge if he/she wants to continue reading by the title. Make sure it will attract the correct set of readers.

  • Abstract. Remember that most (I would estimate 90%) readers of your paper only read the title and abstract. So properly summarize your idea and the importance of the work (not by adding the word “important” to sentences, but explain why it is important). The abstract should be written in a non-technical way, so (1) Non-experts can get roughly what you have done; and (2) Experts can get your idea. The abstract should not be too long. Otherwise it scares readers away. 100-150 words is a reasonable length. That means roughly 4-7 sentences. For example (this is not a strict rule, but just a suggestion if you don’t have better ideas. And you have to polish it instead of just “fill the template”):

    • Sentence 1: We study your topic.
    • Sentence 2: This solves your problem. You may combine it with sentence 1, or move it to the last sentence of the abstract.
    • Sentence 3: Your method is used to compute your observable. If the method is obvious for experts, you can omit it.
    • Sentence 4, 5, …: Highlights of your findings.

Two additional notes on abstract:

- Some new writers just use an expanded version of the above "Sentence 3" as the abstract. This is not good. Don't forget the other parts. They are in fact more important.
- Don't start every sentence by "We ...". Use active/passive voices, and other types of sentences in turn. 
  • Introduction and conclusion. When I was new in writing papers, my headache was: abstract, introduction and conclusion seems to be the same thing written three times. But they are absolutely not. Introduction typically has two parts: First, a story (in either a historical way or a logical way) about the current status of the research field. Second, introduce how your work fits to the story. Conclusion also typically has two parts: First, based on the information given in the paper, how the problem is solved (this differs from the introduction because in the introduction you do not assume one has read your paper. Here you can assume that and thus you can write something deeper). Second, what are the possible future directions opened up by your work. If you have anything not firmly proven by your calculations, but you still want to mention these possibilities, put them in the conclusion part as open issues.

  • Figures and tables. Even the readers do not read your main content (including introduction and conclusion), they may take a quick look in the content and be attracted by your figures and tables. So pay more attentions to them. For example:

    • Quantitative plots or tables of your result: Make them in a conventional way if similar figures exists in the literature. For example, there are standard formats for $n_s$-$r$ diagram, shape of non-Gaussianity, goodness of fit for the study of cosmology. The readers do not have to read anything additional to understand the figure quickly. More later for the format of the figures.

    • Qualitative sketches on the general idea. Design them such that people need minimal reading and can directly understand from the image.

  • The main content. You must write it in a correct and clear way to show how your claims in the other parts are supported. This is of course important, but considering that it is not too much different from your study report, you are more familiar to it.

    • Get to your work early. Do not spend more than one section to review what other people have done. The reader will lose patience before they come to anything interesting.

    • The paper is a selection of what you have done. You may have tried many approaches to the problem, some turn out to be not helpful. Don’t put them in the paper (you may put the ones with existing but not crucial relevance into appendices). Those content may contain your great efforts, but don’t present anything which are not useful to readers. (But for sure, if one approach shows contradiction to another approach, the contradiction must be resolved to make sure the correctness of the research. Do not try to hide problems). Remember that the purpose of the paper is not to show how hard you have worked, but rather the purpose is making it a useful building block of the whole human knowledge.

Show your seriousness in the writing

I trust that you have been serious but the result may not look the same way from the readers (or your supervisor). Many details matter. Let me summarize a few:

Use consistent styles

  • Consistent usage of capital letters and abbreviation. For example, don’t sometimes refer to an equation as “Equation (1)”, sometimes “equation (1)”, sometimes “eq. (1)” and sometimes “Eq. (1)”.
  • Write the same phase in the same way (even the phase may be written in another way) every time it appears. For example, “quasi-single field inflation” shouldn’t be written in a paper somewhere as “quasi-single field inflation”, somewhere as “quasi single field inflation”, and somewhere as “Quasi-Single field inflation”.

Pay attention to spelling/grammar

  • Use an editor with spell check. And avoid obvious grammar mistakes (trick: you can convert your PDF to MS Word, and Word has basic grammar checking. MS Word is the best grammar checker as far as I am aware of).
  • Use articles “a”, “an” and “the” properly. This is probably trivial for native speakers but pretty tricky for Chinese speakers (including myself). At least, the simple rules should be clear. See here for example. Especially, remember, if a word is countable, you must use an article.
  • If you have an incomplete sentence followed by an equation, put a comma at the end of the incomplete sentence only if you should put a comma anyway even the equation is considered as an ordinary word or phrase. For example, you should not write “The equation can be written into, another form”, so don’t put a comma even if the “another form” is replaced by an equation.

Use figures properly

  • Take a look at the properly produced figures (for example, here, here and here) and try to plot (draw) with a similarly professional standard.

  • Special attention should be paid to the way of presentation (plot which variables with which scales, natural choice of model parameters, etc), size of fonts, clearly visible ticks, and consistent line styles, colors & plot markers between figures.

  • Before turning to a more professional plotting software (which takes more efforts and not recommended), I’d like to remind that for plotting, all those details can be tuned in Mathematica.

  • For drawing illustrative sketch, it is helpful to be able to use a vector based drawing software, for example, inkscape.

  • When referring to the figure, use Fig.~\ ref{label} instead of Fig. 1. Note two differences: (1) the label should not be hard-coded, and (2) use ~ (non-breakable space) instead of an explicit space.

Do a favor to your supervisor/collaborators

  • Format the paper similarly to what your supervisor did. For example, add proper affiliations and the acknowledgment section according to the most recent paper by your supervisor.

  • Use the same conventions as your supervisor did unless there are clear advantages to use other conventions.

  • Do not let your supervisor debug your LaTeX. Take a look at the .log file, Make sure there is no compilation errors, and no obviously fixable compilation warnings (missing references, etc. There are some warnings such as overfull tex boxes which are usually fine to go with.)

  • Send files in a convenient format to collaborators (if you do not use Dropbox to share your progress).

    • When you expect your collaborators just to read & comment, send the PDF file only.
    • When you expect them to edit, send a PDF file together with a single .zip file containing all necessary files to compile (if the source file contains more than a single .tex file), no more (remove the .pdf, .aux, .bak, .bbl, .blg, .fdb_latexmk, .fls, .out, .synctex.gz, .log, etc), no less (make sure all the .sty, .bib files and figures, if any, are there, such that a complete installed system of LaTeX can compile the files once unzipped), and request them to edit explicitly. After making the .zip file, test to compile from it.

Cite papers properly

Citation is an important ingredient for a research paper. As a researcher, you feel delighted if your work is properly addressed by others and feel upset if not. The same happens for your readers. Some key points to note:

  • If you have not known the field well enough to properly “rank” the past papers, take a look at your supervisor’s previous papers on how to cite relevant work, or compare a few classic papers about how they cite the research direction. Sometimes those papers may make mistakes (for example Weinberg and the Higgs mechanism). But typically you will be fine. And you will have an excuse just in case that you are not fine.
  • If you cite several papers together, organize them in time order unless there is a particular reason.
  • Order the references such that Reference $[2]$ appears after Reference $[1]$ in the main text, etc. There are two ways to automatically do it: (1) Use BibTeX to manage your references. (2) Use external tools.
  • Use proper reference format. Use INSPIRE. Find a paper there, and use BibTeX or LaTeX(US). Keep the citation style consistent (for example whether to include the title of the paper into the bibliography).