How To Write a Scientific Paper and Format it Using L
A
T
E
X
Jennifer E. Hoffman
1, 2, ∗
1
Department of Physics, Harvard University, Cambridge, Massachusetts 02138, USA
2
School of Engineering & Applied Sciences, Harvard University, Cambridge, Massachusetts 02138, USA
(Dated: June 7, 2024)
The goal of this document is to demonstrate how to write a scientific paper. We walk through the
process of outlining, writing, formatting in L
A
T
E
X, making figures, referencing, and checking style and
content. Source files are available at: http://hoffman.physics.harvard.edu/example-paper/.
I. GETTING STARTED
1 You should start writing your paper
while you are
working on your experiment. Prof. George Whitesides
says: “A paper is not just an archival device for stor-
ing a completed research program; it is also a structure
for planning your research in progress. If you clearly
understand the purpose and form of a paper, it can be
immensely useful to you in organizing and conducting
your research. A good outline for the paper is also a
good plan for the research program. You should write
and rewrite these plans/outlines throughout the course
of the research. At the beginning, you will have mostly
plan; at the end, mostly outline. The continuous effort
to understand, analyze, summarize, and reformulate hy-
potheses on paper will be immensely more efficient for
you than a process in which you collect data and start to
organize them only when their collection is ‘complete’ .”
Here are some concrete steps to get started.
1. Read George Whitesides’ “How to Write a Paper” [1].
2. Read through at least one full paper in your target
journal, to get a sense of the content and writing style.
3. To clarify in your own head the purpose of your paper,
start by drafting your abstract [2].
4. Before you tackle the body of the paper, sketch block
outlines of the figures. Decide what images and plots
you will put in the paper, and how the panels will
be arranged. An excellent example of an outline with
sketched figures is shown in Ref. [3], and can be com-
pared to the final published paper in Ref. [4].
5. Outline at the paragraph level before you write. Look
at how many paragraphs there will be in the style of
paper you are trying to write. For example, for a
standard 4-page scientific letter, aim for 13 paragraphs
(generally, you can estimate about 200 words per para-
graph). Figure out how to tell your entire story (not
numbers, just story!) in 13 stand-alone sentences.
6. Make each of those sentences into the first sentence
of a paragraph, and fill into each paragraph only de-
tails that are relevant to that first sentence. If you
find yourself writing details about the figures, cut and
paste them into the captions.
7. If you think of references as you go, you can include the
∗
jhoff[email protected]ard.edu
minimal identifying information in parentheses to trig-
ger your memory later, e.g. “(WhitesidesAdvMat)”, so
all of the information is compact.
8. Dig into the existing literature to write the intro para-
graphs. A thorough literature search may take a full
focused week for each intro paragraph. Use an or-
ganized, three-pass approach to keep a good balance
between depth and breadth of your search [5].
9. Rewrite your abstract, taking into account what you
have learned from the process of writing the paper. As
you fine-tune your abstract, refer again to Nature’s
instructions for writing an abstract [2] and for clear
communication more generally [6].
2 As you contemplate the paper you have just writ-
ten, put yourself in the shoes of the reviewers (includ-
ing your collaborators). You already work many, many
hours/week, and you don’t really want to spend more
time reading this paper. So you’re going to be very happy
if the figures are pretty, the text flows logically, the ref-
erences are hyperlinked for easy access, and you can un-
derstand the paper quickly. But you’re going to be very
grumpy if you can’t get the main points of the paper from
scanning through the figures & captions. You’re going to
be even grumpier if you invest time in reading the pa-
per but you still can’t get it. Your evaluation of this
paper is likely to be swayed by your ease of understand-
ing, regardless of the scientific merits of the work. (See
Ref. 7 for more information on how formatting, even as
simple as font choice, will influence the reader’s “cogni-
tive ease” and ultimately their judgment of the report.)
Down the road, consider a reader who might cite the
paper and launch you to fame and glory: the potential
citer’s decision will be influenced by their ability to easily
understand your paper.
3 Your paper should be fractal. Somebody with
one minute to look at it should be able to get the main
idea just from reading the abstract. Somebody with 5
minutes should be able to look at the figures and get
more out of it. Somebody with 10 minutes should be
able to get the story from the introduction, first sentence
of each paragraph, and conclusion.
II. L
A
T
E
X
4 L
A
T
E
X is a formatting language that allows profes-
2
sional, publication-quality presentation of scientific text,
equations, figures, and hyperlinked references. REV-
TeX is a specific set of L
A
T
E
X macro packages assem-
bled by the American Physical Society (APS) to stan-
dardize manuscript formatting for the various Physical
Review journals. The APS author website [8] lists ac-
ceptable submission formats as: REVTeX (preferred),
L
A
T
E
X, or MSWord. It is sensible to draft your paper
in L
A
T
E
X/REVTeX from the beginning. You will need a
reference manager, a text editor, and a L
A
T
E
X compiler.
Typically, you will be working closely with other authors,
so you should pick a platform that enables simple sharing
and collaborative workflow.
5 Here we recommend the online L
A
T
E
X compiler Over-
leaf (http://www.overleaf.com/) for text editing and
compiling. Overleaf (formerly known as ShareL
A
T
E
X) is
an online platform that is great for collaborative writing:
it tracks changes and even has a real-time chat function.
Note that LyX is a WYSIWYG (what you see is what
you get) L
A
T
E
X editor that may seem tempting if you ini-
tially find tex formatting daunting, but LyX may come
back to bite you in the long run, because it hides the guts
of the tex code from the user, making it harder to control
formatting details.
1. Choose & install a reference management software
such as Mendeley: http://www.mendeley.com/
(See appendix A for usage tips.)
2. Make an account on Overleaf:
http://www.overleaf.com/
3. Download the source files for this example paper [9].
4. Copy the tex and bib files into a new overleaf
project, and share with your collaborators.
6 Alternatively, you may prefer to run L
A
T
E
X directly
on your own laptop. In that case, follow the steps
below:
1. Choose & install a reference management software:
• Mendeley: http://www.mendeley.com/
• Papers: http://www.papersapp.com/
• Endnote: http://www.endnote.com/
2. Download & install a L
A
T
E
X compiler such as MikTex:
http://miktex.org/
3. Download & install REVTeX 4-2 (see appendix B):
http://authors.aps.org/revtex4/
4. Pick an editor such as WinEdt:
http://www.winedt.com/
(You may need to request a university license; may
require ∼ 24 hour turnaround.)
5. Download the source files for this example paper [9],
open the tex and bib files in WinEdt, and get started
writing!
7 It’s important not to lose sight of your outline, as
you fill in the details of your paper. This L
A
T
E
X tem-
plate file allows you to title each paragraph using the
\ptitle{} command. You should keep these titles in
place throughout the entire paper-writing process; they
will serve as a constant reminder to keep each paragraph
focused on a single point. You should be able to skim
through these bold paragraph titles, without reading any
of the intervening sentences, and still understand the ba-
sic logical flow of the paper. At the final step before
submission, comment out the line \ptitletrue in the
header, to hide the paragraph titles. But do not delete
the paragraph titles, because they will remain useful to
you in the inevitable paper revision process down the
road!
III. FORMATTING
8 Whether you are using a compiler on your computer
or online, please use the latest version of REVTeX, and
check your formatting carefully.
□ Check math & symbolic formatting, as in Table I.
TABLE I. Formatting mathematical symbols.
Incorrect Correct
cosθ cos θ
T
sample
T
sample
V
rms
, V (rms) V
rms
E
x
, x direction E
x
, x direction
⃗
B
app
⃗
B
app
Sb
2
T e
3
, Sb2Te3 Sb
2
Te
3
Sb
2−x
V
x
Te
3
Sb
2−x
V
x
Te
3
dI/dV dI/dV
B = 5T , B=5T B = 5 T
x direction, X direction x direction
1st, 1
st
, 2nd, 2
nd
1
st
, 2
nd
□ Use \label{tab:name} and \ref{tab:name} to refer
to tables. Ref. [10] is an automatic table generator.
□ Check hyphenation. Sometimes L
A
T
E
X likes to di-
vide a single letter off the beginning or end of a
word, for line wrapping. The default settings for
L
A
T
E
X’s hyphenation of English-language words are
\righthyphenmin=3 and \lefthyphenmin=2, but ap-
parently they can be mysteriously reset to allow single
dangling letters.
□ Check spacing. When a period falls in the middle of
a sentence, use a \ (backslash) to prevent L
A
T
E
X from
thinking it’s the end of the sentence and thus adding
extra space, as shown in Table II. If you want to pre-
vent a linebreak, you can use ~ instead of \.
□ Number all equations. But do not separately number
each line of a single multi-line equation.
□ Use \label{eqn:name} and \ref{eqn:name} to refer
to equations.
□ If the equation is mid-paragraph, use \noindent at
the beginning of the first line following the equation.
9 Here is an improperly-labeled equation in the middle
of a paragraph.
1 + 1 = 2
3
TABLE II. Spacing.
L
A
T
E
X Output
Incorrect e.g. incorrect e.g. incorrect
Incorrect Fig. 2 Fig. 2
Correct e.g.\ correct e.g. correct
Correct Fig.\ 2 Fig. 2
Correct Fig.~2 Fig. 2
Use \noindent to prevent indentation mid-paragraph.
Here is a more interesting example of a properly labeled
equation: the Pythagorean theorem relates the 3 sides of
a right triangle according to Eqn. 1,
a
2
+ b
2
= c
2
. (1)
Eqn. 2 shows one more example of a multi-line equation
extending the Pythagorean theorem to find the diagonal
d of a rectangular prism of sides a = 3, b = 4, and c = 12.
d =
p
a
2
+ b
2
+ c
2
=
p
3
2
+ 4
2
+ 12
2
= 13 (2)
IV. STYLE & GRAMMAR
□ Use the active voice. If you use the passive voice, it
is very hard to tell the difference between what you
personally worked hard to do in your experiment, vs.
what you are citing as a prior result. Use of the ac-
tive voice is an important component of ethical credit
attribution.
□ Avoid pronouns if at all possible (e.g. not “that” but
“that voltage signal”).
□ Use adjectives if there is any doubt (e.g. not “the mod-
ulation” but “the z modulation”).
□ Use parallel list structure: ensure that each item in a
list (or an “and” statement) is the same part of speech
and the same category of thing. Not “we demon-
strated apples and the observation of bananas” but
“we demonstrated apples and bananas”.
□ Define all acronyms and symbols at first use; then use
the acronym consistently from that point on.
□ Don’t capitalize the words you are about to turn
into an acronym. Not“Scanning Tunneling Microscopy
(STM)” but“scanning tunneling microscopy (STM)”.
Not “liquid Mercury (Hg)” but “liquid mercury (Hg)”.
□ Do not use the word “significantly” unless you mean it
in the true statistical sense and are prepared to back
it up quantitatively.
□ Do not use the word “successfully” – just say what
you did and let it speak for itself.
□ Remove redundancy, including redundancy between
main text and figure captions. When in doubt, the
information probably belongs in the caption but not
the main text.
□ Check that all equations are dimensionally correct.
□ For each equation, define all symbols in a previous
equation or in the surrounding text.
□ Report each quantity consistently throughout the text
(e.g. don’t exaggerate a quantity in one place, give an
exact version of the same quantity in another place,
and round to the nearest 100 in a third place).
□ Check that all numbers have units.
□ Use reasonable significant figures, report errors where
appropriate, and clearly explain the method of error
determination [11].
□ When in doubt, check examples (e.g. if you wonder
whether acronyms are appropriate in the abstract,
check a few recent published examples in your target
journal).
□ “Only” can be an adjective or an adverb, so its mean-
ing can be ambiguous. It should be placed immedi-
ately before the noun, adjective, or verb that it is
modifying. For example “I only bought groceries at
the store” means I didn’t dance or sing at the store,
I only bought. But “I bought only groceries at the
store” means I didn’t buy party balloons at the store.
□ “Its” is possessive;
“it’s” is a contraction of “it” and “is”.
□ “fewer” describes discrete items, whereas “less” de-
scribes continuous substances [12].
□ “which” vs. “that”: If the sentence doesn’t need the
clause that the word in question is connecting, use
which. If it does, use that [13].
□ “affect” vs. “effect”: Each word can be an adjective or
a noun, with very different meanings [14].
□ See additional tips from Prof. Margo Seltzer [15].
V. FIGURES
10 It is worth taking 2-3 hours to read the definitive
guide to “The Visual Display of Quantitative Informa-
tion” by Edward Tufte [16]. Tufte defines several metrics
for figure optimization:
Data-ink ratio =
data-ink
total ink used to print the graphic
= 1.0 − fraction of a graphic that can be
erased without loss of data-information
Data density =
number of entries in a data matrix
area of data graphic
A clear and visually pleasing figure should:
□ Maximize data-ink ratio.
□ Maximize data density.
□ Avoid “chart-junk”, i.e. hatching patterns that inter-
act with the natural motion of the eye to promote the
distracting perception of vibration in a static graphic.
□ Avoid excessive colors.
□ Avoid red-green combinations (5-10% of people are
red-green colorblind! [17])
4
□ Use concise but clear words (not inscrutable abbrevi-
ations) directly on the graphic, so the reader doesn’t
have to dig through a lengthy caption or text to un-
derstand the components of the figure.
□ Orient words horizontally whenever possible.
11 Figures should typically be made in Python, Adobe
Illustrator, or other program that allows vector format
export, so that all fonts, arrows, etc. will scale cleanly
when zoomed. Most journals prefer to stay away from
Microsoft Powerpoint (although it can be exported to eps
or pdf) because the fonts are often not transcribed cor-
rectly in publication format. A bigger problem with Mi-
crosoft is that it does not faithfully reproduce the pixela-
tion of data images. Microscope images are acquired with
a specific pixel resolution, and that pixelation should be
honestly communicated to the reader without interpola-
tion. Fig. 1 illustrates this point.
FIG. 1. Comparison between blurry pixels (dishonest interpo-
lation occurs when the image is processed in Microsoft Pow-
erpoint) vs. clean pixels (honest representation is preserved
when the image is processed in Python and Adobe Illustra-
tor). MFM images of vortices in NdFeAsO
1−x
F
x
[18].
12 Note that faithful representation of images in
vector format usually also results in a smaller figures
size. This can be important, because some journals and
preprint servers have strict size limits on components of
a submitted manuscript.
13 Sans-serif fonts are preferred by most journals, as
they are typically easier to read in smaller sizes. Arial
is the most robust choice, because it is a free font that
comes installed on all Mac and Windows computers [19];
whereas Helvetica is a licensed font [20], available only
to Mac users (only ∼15% of market share [21]). Use of
Helvetica may cause your figures to render incorrectly for
the majority of readers.
14 All figure fonts should be at least size 6 in the
final publication figure [22]. To achieve the appropriate
font size, please start by measuring the desired final fig-
ure size (e.g. one or two column width) in the desired
journal. Then make a box in Adobe Illustrator (or other
program) of exactly the final size, and build your figure
within it, using no fonts smaller than size 6. Although
some journals do prefer that you initially submit your fig-
ure at full-page size, you can easily scale up your figure
for this purpose. But if you start with a page-size figure
and arbitrary font sizes, it becomes harder to later scale
it down while maintaining adequate font size.
Figure checklist:
□ Use consistent font, at least size 6.
□ Label all axes, with units.
□ Each plot should have a legend that describes all sym-
bols and lines.
□ Each image (or set of same-scale images) should have
an accurate length scalebar, with numerical label.
(Note that some journals discourage or “forbid” super-
imposing the numerical length on the image. But our
goal is clarity: we want the reader to understand the
image at a glance, without digging through a lengthy
caption to find the necessary number. Journals will
generally accept this argument for keeping the num-
ber on the image.)
□ Each image (or set of same-palette images) should
have a colorbar. The colorbar should be labeled with
numerical values and units if possible.
□ If using a waterfall plot to display a set of spectra:
clearly state the offset of the waterfall plot, or use
small horizontal lines to denote the true zero reference
for each individual spectrum.
□ The caption should describes all figure sub-parts, in
order. Each and every mark on the figure should be
described; there should be no mysterious unexplained
arrows or other features.
□ If any analysis has been performed (i.e. if it’s not raw
data), then all analysis steps should be clearly di-
vulged, usually in the caption (rather than main text).
□ Clearly explain the origin of all error bars, usually in
the caption (rather than main text).
□ For STM images: give setup conditions in the caption
(V
sample
= 100 mV; R
J
= 1 GΩ).
□ For STM spectra: give sample bias modulation in the
caption (V
rms
= 2 mV).
□ For all data: clarify temperature and field conditions
in the caption.
□ Appropriately cite all copied figures or data, in the
caption of the figure.
□ Use \label{fig:name} within the caption, and use
\ref{fig:name} in the paper to refer to it.
VI. REFERENCES
Referencing should be done using BibTeX.
□ Consistently use reference tags that will be eas-
ily recognizable and editable from the tex file, e.g.
AuthorJournalYear. Suppose you will be citing
Huang et al, Nano. Lett. 16, 4224 (2016) [23]
and Huang et al, Phys. Rev. B 93, 125129 (2016)
[24]. Instead of Huang2016a and Huang2016b, use
HuangNanoLett2016 and HuangPRB2016. (Note: You
can make these bib tags within Mendeley.)
□ Spend 5 minutes to use find-replace with regular ex-
pressions to delete the abstracts, keywords, and other
5
useless information from your bib file, to make it easier
to read.
□ Alphabetize the bib entries by author last name, so
that it will be easy to notice if there are duplicates.
(Note: Mendeley can automatically export them in
alphabetical order.)
□ The hyperlink should be generated from the
doi, so make sure every bib entry has a valid
doi. You should delete the URL field (unless
you are specifically citing a website), because it
may contain a non-general URL such as https:
//journals-aps-org.ezp-prod1.hul.harvard.
edu/prb/abstract/10.1103/PhysRevB.93.125129.
□ Check the L
A
T
E
X formatting of any special characters
in the author’s names, e.g. S\'{a}nchez
□ Check the L
A
T
E
X formatting in the titles. You need to
use extra {} around any letters that should be cap-
italized, e.g. title = {Quantum Anomalous {H}all
Effect}
□ Chemical formulas should also be enclosed in {} so
element abbreviations will be capitalized, e.g. title
= {Experiments on {Bi$ 2$Se$ 3$}}. Don’t use ex-
cessively complicated formatting for chemicals. Just
put the subscripts in $$.
□ Use longbibliography in the documentclass at the
top of the tex file, so that the title of each paper ap-
pears in the references.
□ For most citations, you can just use
\cite{AuthorJournalYear}, e.g. Jeehoon used
conducting force microscopy to measure a metal-
insulator transition [25]. But if you’re using a
superscript citation style, and the citation comes
directly after a number or chemical formula, use
Ref.\ \onlinecite{AuthorJournalYear} instead
to avoid confusion, e.g. Jessie measured the pinning
force on vortices in NdFeAsO
1−x
F
x
(Ref. 18).
□ In the compiled PDF, check all the references carefully
to make sure they have correctly formatted authors,
titles, and hyperlinks.
VII. PROFESSIONAL INTEGRITY
□ Authorship – are all major contributors and collabo-
rators included? See the American Physical Society
guidelines for authorship [26].
□ Plagiarism – have you been careful to distinguish be-
tween your own work and ideas, as opposed to those
of others?
□ Citations – have you properly cited prior work, and
references that you used?
□ Data Integrity – have you clearly described the data
analysis methods, and justified any data points that
were excluded?
□ Image Processing – have you clearly described any pro-
cessing that was applied to images?
□ Acknowledgments – have you given appropriate credit
and thanks to collaborators and other individuals or
organizations who deserve recognition?
□ Clarity of collaborative structure – if this is a joint
effort, have you identified people who you worked with
on this project? Acknowledgments should clearly state
who did which parts of the experiment & analysis, and
who wrote the paper.
□ Conflicts of Interest – do you have any conflicts of
interest where you or someone close to you stands to
gain, financially or otherwise, from this work?
VIII. FINAL CHECK
Don’t write one of these: https://xkcd.com/2456/
Think critically about all of your own claims
and about
all of the claims made by your coauthors. If you do not
understand something that your coauthor has written
in the draft, push back until you do understand, then
suggest an alternative phrasing or labeling to clarify the
manuscript or figure.
Before you submit, pause to make sure you can clearly
articulate in 1-2 sentences, what is the actual new fact
about nature that you discovered? What exactly do you
know now, that you didn’t know before you completed
this project? Say those sentences out loud. Rehearse
until they are concise and clear. Write them down.
Here are some phrases to beware of:
Amplifiers:
• remarkably
• successfully
The amplifiers should just be deleted. Real facts will
stand on their own merit without additional subjective
descriptors. Such hyperbolic language is a red flag to
the editor or reviewer that you are not confident in the
importance of your own work.
Vague claims of non-specific accomplishment:
• pave the way
• shed light on
• provide insight into
• unveil
The vague claims should be replaced with 1-2 factual
sentences that state your new discovery. Such vague lan-
guage is a red flag to the editor or reviewer that you have
not yet figured out the conclusion of your own research.
If you haven’t figured out your own conclusion, how can
you expect the reader to figure it out?
ACKNOWLEDGMENTS
We acknowledge advice from Jessie Zhang and Harry
Pirie to produce Fig. 1. We acknowledge suggestions
for the professional integrity checklist from Chris Stubbs.
We also acknowledge a generation of students who have
made all of the errors that led to these checklists.
6
Appendix A: Mendeley
Mendeley provides a convenient (although not 100%
bug-free) database structure for storing, sorting, and an-
notating the papers you read. Mendeley also provides
an export function to automatically create your bib file.
Here are some tips to use Mendeley most effectively.
1. Download Mendeley: http://www.mendeley.com/
Launch it on your desktop.
2. Set Mendeley options: Make sure your Mendeley
database is set correctly to include the DOI (digital
object identifier) field. Go to Tools → Options →
Document Details. Scroll down and make sure the
DOI box is checked.
3. Import paper: In the upper left corner of Mendeley
Desktop, click the drop-down menu for “Add” and se-
lect the bottom option “Add Entry Manually”. In the
dialog box that pops up, scroll down until you find the
DOI field. Paste the DOI into the field, and click the
little magnifying glass icon to the right of the field.
This will auto-populate all of the relevant paper infor-
mation such as author names, title, etc., without risk
of typos due to manual copying.
Note 1: Mendeley also allows you to import directly
from a PDF file, and it tries to pull all of the meta-
data from the PDF, but the process is imperfect. So
it’s safest to use the DOI for an error-free import.
Note 2: Even if you use the DOI, some journal titles
will not import correctly with special characters, so
you may need to manually correct.
4. Add tags: It’s useful to add tags to help sort your
imported papers. For example, if you are going to
be writing a manuscript in 2019 on superconductivity,
you might add the tag “sc19” to all the relevant papers
that you will be citing in your manuscript.
5. Export bib file: Select all of the references that you
want to include, and go to File → Export. Name
your file, and it will add a citation key to each paper
(e.g. Whitesides2004) and automatically export to a
bib file.
6. Resolve redundant citation keys: At this point, you
may have several references with the same citation key,
e.g. Huang2016a and Huang2016b [23, 24]. For your
future convenience, you should manually change the
redundant citation keys to be more informative, e.g.
HuangNanoLett2016 and HuangPRB2016. Now re-
export the bib file.
7. Open the bib file in your tex file editor. By default,
Mendeley exports all fields, including long ones like the
abstract. To reduce clutter in your bib file, and make
it easier to debug any errors, it’s a good idea to remove
the abstracts and other unnecessary fields. For exam-
ple, in WinEdt go to Search → Replace, check the reg-
ular expressions box, search for “<abstract**\},>,”
and replace it with nothing.
Appendix B: REVTeX
How to install REVTeX 4-2 on Windows for MikTex:
1. Download REVTeX 4-2:
http://authors.aps.org/revtex4/
2. Unzip the downloaded folder revtex4-2-tds
3. If you don’t already have one, create a folder C:\TeX-
local
4. Copy the four subfolders (bibtex, doc, source, tex) into
C:\TeX-local
5. From the start menu, open MikTex 2.9 → Mainte-
nance (Admin) → Settings
6. On the root tab, add the path C:\TeX-local
7. On the general tab, click “Refresh FNDB” (You may
need to close WinEdt, or whatever is using your
MikTex installation, in order to properly “Refresh
FNDB”.)
8. Download natbib:
http://www.ctan.org/tex-archive/macros/
latex/contrib/natbib/
9. Unzip natbib.
10. Open natbib.ins, and run TeX on it (Shift+Ctrl+T in
WinEdt)
11. Open bibentry.ins, and run TeX on it (Shift+Ctrl+T
in WinEdt)
Appendix C: Vector Graphics
To create vectorized graphics from python, include
rasterized=True inside the imshow command. Unfor-
tunately, this preference doesn’t seem to be allowable as
a default in any style sheet, so you need to remember
to include it it every time. Then save the python fig-
ure as a PDF, which ensures that annotations (like scale
bars, axis labels, and lines) remain as vector format, but
images retain their proper pixelation. For example:
1 fig, ax = subplots(1, 1, figsize=[1.8,1.6])
2 ax.imshow(a.Z, cmap=stmpy.cm.blue2, aspect=1,
3 clim=(0,1.4), rasterized=True)
4 ...
5 tight_layout(pad=0)
6 savefig("topo.pdf")
Note that some PDF viewers apply their own smooth-
ing to images. For example, MacOS Preview may blur
images that look fine in Adobe Illustrator.
[1] G. M. Whitesides, Whitesides’ group: Writing a paper,
Advanced Materials 16, 1375 (2004).
[2] How to construct a Nature summary paragraph.
[3] S. M. Gardezi, Example outline of “Simulating twistron-
7
ics in acoustic metamaterials”.
[4] S. M. Gardezi, H. Pirie, S. Carr, W. Dorrell, and J. E.
Hoffman, Simulating twistronics in acoustic metamateri-
als, 2D Materials 8, 031002 (2021).
[5] S. Keshav, How to read a paper, ACM SIGCOMM Com-
puter Communication Review 37, 83 (2007).
[6] How experts communicate, Nature Neuroscience 3, 97
(2000).
[7] D. Kahneman, Thinking, fast and slow (Farrar, Straus
and Giroux, 2011) Chap. 5: Cognitive Ease.
[8] Physical Review journals web submission guidelines.
[9] J. E. Hoffman, How to write a paper and format it using
L
A
T
E
X, Source files.
[10] Create L
A
T
E
X tables online.
[11] C. Witkov and K. Zengel, Chi-Squared Data Analysis
and Model Testing for Beginners (Oxford Scholarship
Online, 2019) Chap. 2.5: Random versus Systematic Un-
certainties and Rounding Rules.
[12] M. Fogarty, Less vs. Fewer.
[13] B. A. Klems, Which vs. That (Grammar Rules).
[14] Grammarist, Affect vs. Effect.
[15] M. Seltzer, Pet Peeves.
[16] E. R. Tufte, The Visual Display of Quantitative Information,
2nd ed. (Graphics Press LLC, Cheshire, CT, 2001).
[17] F. Crameri, G. E. Shephard, and P. J. Heron, The misuse
of colour in science communication, Nature Communica-
tions 11, 5444 (2020).
[18] J. T. Zhang, J. Kim, M. Huefner, C. Ye, S. Kim, P. C.
Canfield, R. Prozorov, O. M. Auslaender, and J. E. Hoff-
man, Single-vortex pinning and penetration depth in su-
perconducting NdFeAsO
1−x
F
x
, Physical Review B 92,
134509 (2015).
[19] Arial Font.
[20] Helvetica Font.
[21] Usage share of operating systems: Desktop and laptop
computers.
[22] B. de Bivort, Figure style guide.
[23] D. Huang, T. A. Webb, C.-L. Song, C.-Z. Chang, J. S.
Moodera, E. Kaxiras, and J. E. Hoffman, Dumbbell de-
fects in FeSe films: A scanning tunneling microscopy
and first-principles investigation, Nano Letters 16, 4224
(2016).
[24] D. Huang, T. A. Webb, S. Fang, C.-L. Song, C.-Z. Chang,
J. S. Moodera, E. Kaxiras, and J. E. Hoffman, Bounds on
nanoscale nematicity in single-layer FeSe/SrTiO
3
, Phys-
ical Review B 93, 125129 (2016).
[25] J. Kim, C. Ko, A. Frenzel, S. Ramanathan, and J. E.
Hoffman, Nanoscale imaging and control of resistance
switching in VO
2
at room temperature, Applied Physics
Letters 96, 213106 (2010).
[26] APS Guidelines for Professional Conduct.
