Good writing is vital to getting credit for the work you have done. It can easily make the difference between publishing in a high impact journal and a low impact journal.

Software

The group’s preferred tool for writing is overleaf, which is an online latex editor. Mark has a subscription for this, which allows multiple users and tracked changes if he created the initial project. Pretty much all journals accept latex these days. More likely than not, we already have a suitable template, but you can almost certainly find one on the website of the relevant journal if needed. We do use word from time-to-time, which is easy to use and has good tools for tracking changes, but is not as good at handling re-ordered text and references. For figures, we suggest the python package matplotlib.

Suggested steps for preparing a manuscript

Do not imagine that you can just write everything perfectly and then submit it after checking for typos. Almost everyone needs to revise and re-order their initial text. Writing a good paper is hard! We suggest the following sequence

Each paper should make one, maybe two, key points. Write these down in order to focus on them and consider changing them as you tighten up exactly the best way to convey the value of your work.
Construct the figures together with an ordered sequence of roughly ten bullet points telling the story you want to get across. Think about what order is best.
Draft the figure captions. The first line of the caption should give a general sense what the figure shows. After this, you can define the more detailed information. Tell the reader what they should pay attention to: Don’t just refer to the data and expect them to work it out for themselves. Try your best to convey the overall story with the figures and captions in isolation. Ask Mark to check the structure of the manuscript.
I usually like to organize the citations of the relevant literature at this point. The manuscript needs to put new discoveries made in this work front and center, while being careful to do justice to what has been done before.
Draft the body of the paper. Here, you will need to think carefully about which points should be made immediately after the data is presented and which are better left as discussion points later in the text.
Draft the conclusions. Shorter is usually better here. It is often tempting to make a generic call more more experiments. Unless this invokes a particular plan or particular goal this can be very nebulous.
The introduction is probably the hardest part of the paper to write, which is why we suggest writing it last after you are very familiar with the material in question. A good introduction starts from a general conceptual problem and logically argues that there is an important unknown question, which is the question that the paper solves. A bad introduction often just makes a case for the research field in general. For high impact journals, you will usually want to make the introduction as short as possible as this makes the work more exciting and keeps the reader focused on your data. If there is some boilerplate information you need the reader to know, it is usually best locate this in the body of the paper while discussing the data. One usually wants to avoid general literature review – keep the discussion in the context of how the manuscript goes beyond what was done previously.
Repeat the previous steps and consider whether you can make better choices for the order of the argument, the points to emphasize, and the way that the work is presented in the context of the literature.
Do not go overboard trying to meet word or page limits in the first draft. If you write a slightly longer initial draft and then diligently revise the text, the final result will almost certain be better.
After you have a complete draft, it can be helpful remind yourself of the overall structure of the document. It should break-down into a sequence of elements each with a specific and readily-apparent purpose. Does the text have the necessary subheadings or skeletal sentences to convey this?
Almost all journals provide an opportunity to include additional material beyond what is in the main manuscript itself. This has various names: supplemental material, supplementary information, appendices etc. Journals are often very vague about what should and should not be in these addenda. Our advice is centered around the idea that it is very disruptive to force the reader to stop reading and refer to another document. We suggest that you aim that the main manuscript provides a complete logical argument for its conclusions. If you cannot do this, make sure that when the main text relies on something in an appendix, that the main text specifies what logical step is provided there. The reader should only feel the need to read the appendix if they are not willing to trust the assertion in the main text – you do not want to give them a feeling that they must dredge through a long separate document in case they might miss an important part of the argument.

Things to check for

Readers will often try to understand the paper by just reading the figures and captions. Go through the paper reading the captions and looking at the figure and try think whether the reader will follow the general story.
Remember that the audience is not a mind reader. If a sentence is a little difficult for you to parse, it will likely be missed completely by the audience. For better or worse, the onus is on you to convey your ideas even if the reader is lazy! Anyone who was read more than one or two referee reports will have seen that ideas are often missed. Try to put yourself in the shoes of a potential PhD student or a rushed reviewer.
Be judicious about using acronyms. If the text in question is short and/or the text is only used a couple of times it is probably not worth using an acronym – it will just make the text harder to read. The main exception to this is when the acronym is unambiguous and very well known. Acronyms should, of course, always be defined in the manuscript.
Activate voice is usually better than passive voice. Save the passive voice for special cases, such as when you want to stress the object more than the subject.
Many students are taught that papers should be written exclusively in the third person past tense. This will make the work unbearably boring to read! Ignore this rule. In science, we are aiming for statements that are generally true: now, in the past and in the future. The simple present tense is appropriate to convey this: use it as your “go to” sentence construction. As with almost everything in English, there will be some exceptions, such as the methods section, which is often better written in past tense.
It is best to use a variety of first, second, third person as appropriate. Try reading papers in the journal you are aiming for to get a sense of what is best.
On occasion, one sees multiple different terms used to refer to a similar thing e.g. charge density wave, charge order, and stripes. If these really refer to the same thing, it is better to just use one term. If you want to convey a distinction, be sure to include clear explanations or your meaning may well be lost or worse still you may actively confuse people.
It is usually better to put action in verbs: “we analyzed the data” tends to be better than “we performed analysis on the data”. Some examples of verbs reformed into nouns are: regulation, analysis, occurrence, understanding, investigation, delineation, and performance. Check for these and consider replacing them.
Short sentences are usually easier and clearer to read than long ones. Tend towards writing in short sentences and avoid sequences of many long sentences. (This is not a hard-and-fast rule.) Think of paragraph breaks in a similar way. They provide a charge to release focus on a prior idea and increase emphasis on a new idea.
Sentences with multiple clauses or other complex constructions can be grammatical, or even beautiful, but keep in mind that you are making life hard for yourself! You are probably better off with a simple sentence.
Multiple run on sentences with the same construction usually reflects bad or lazy writing. Excessive use of the same word can reflect the same problem ‘show’, for instance, can be replaced by ‘feature’, ‘reveal’, ‘display’ etc.
English is blessed with a huge vocabulary, so, when choosing words, lean heavily towards simpler, common words. Using fancy words can be good, but make sure you are either using them to pass on a specific meaning or to call special emphasis to something particularly important.
Be cautious about stating your data ‘clearly’ shows something. If your data are especially clear, the feature in question will leap out at the reader without special emphasis in the text. ‘Clearly’ often conveys very little and the sentence will probably be better without it.
Journals sometimes insist that you do not use terms like “new”, “novel”, “unique”, “for the first time”. We do not dislike these words per se, but we urge care here. Make sure that you are not just stating that your work is “novel” as a lazy way to avoid explaining why your work is novel. Consider modifying the prior sentence to set up a problem that you can solve in the following sentence. If you have a clear explanation of novelty, you may find that stating “novelty” is unnecessary.
Take care not to make assumptions about the gender identity of researchers. The Chicago Manual of Style has a useful section about this.

Other resources

Improving Your Scientific Writing provides excellent advice.
This essay on writing by George Orwell is a classic. Read it while drinking tea.
For technical matters, Appendix A of the Rev. Mod. Phys. Style Guide is highly recommended. The amsmath user guide is a helpful document for the use of latex and in particular handling multiline equations.