Special Symbols Reference is a character table grouped by purpose: math operators (±×÷√∞≈≤≥∑∫∈∀…), superscripts and subscripts (²³ⁿ₂…), currency (¥€£₩₹₽₿…), arrows (→⇒↔↻ plus key symbols such as ⌘⌥⇧), CJK punctuation (、。「」《》【】 along with the full-width dashes and ellipses), English and general punctuation (§¶†«»¿ plus the many kinds of space and the zero-width characters), Greek letters, units and technical symbols (℃℉µ№™®©⚠⏻), geometric shapes and bullets (■●★✓✗☑♥♂♀♪), enclosed characters and numbers (①–⑳❶–❿ⅠⅡⅢ), box drawing and line art, accented Latin letters (é ñ ü ß, plus pinyin tone marks ā á ǎ à and IPA characters such as ə θ ð), and a hand-picked set of kaomoji — 680+ characters altogether.

Every symbol copies on click. What sets this page apart from a plain symbol list is that it also answers “what is this character, exactly?”: click any symbol and the detail card below lists its Unicode code point, its HTML decimal and hexadecimal entities, the named entity where one exists, the JavaScript and CSS escape forms, the LaTeX command, and how to type it on Windows (Alt codes), on macOS (Option combinations) and with a pinyin IME. The lookup runs backwards too — put U+2603, ☃, ♥ or \u00B2 into the “code point / entity → character” box and the character comes straight back.

Three kinds of need bring people here: you are writing a document and want one symbol (“how do I type a squared sign”, “degree symbol”, “circled numbers”); you are writing a page or code and need the entity or the escape form (°, \u00B0); or you are chasing an encoding problem — why does this character show as a box on someone else's computer, why can a µ that looks identical not be found — and the Unicode block overview at the end of the page exists for that last case.

1. 01Search— type a purpose or a name (“squared”, “degree”, “arrow”, “check mark”, “euro”, “third tone”), or go straight to a code point such as U+00B2 or an English name such as alpha.
2. 02Click to copy— clicking any cell copies the character itself and adds it to “Recently copied” (the last 24 are kept in this browser); the star pins it to Favorites.
3. 03Read the detail card— after a click the card appears at the top and every attribute can be copied individually: code point, HTML entity, LaTeX, JS/CSS escape. The “how to type it” line gives the keyboard and IME method.
4. 04Look up a character from a code point— enter U+2192, 2192, → or → in the small box on the right and the character appears at once with all of its encoding forms.
5. 05Browse by category— the chips switch category, and each category heading names the Unicode block its characters come from.

Searching for “squared” returns:

Typing U+2603 into the “code point → character” box produces this detail card:

²   squared, superscript 2      U+00B2   ²  ²  ²   LaTeX ^2   Win Alt+0178 · Pinyin IME: pingfang
㎡  square metre (one character) U+33A1   ㎡                a compatibility character — use m² for real typesetting
√   square root, radical sign   U+221A   √ √              Win Alt+251 · Mac ⌥v

☃  snowman                        Miscellaneous Symbols
Unicode           U+2603
HTML decimal      ☃
HTML hexadecimal  ☃
JavaScript        \u{2603}
CSS content       "\2603"

Code point, character and glyph are three different layers

Unicode gives every abstract character a number, written as U+ plus hexadecimal digits: that is the code point. ² is U+00B2 and → is U+2192. How a code point is stored in a file is decided by the encoding — in UTF-8, U+00B2 takes 2 bytes, U+2192 takes 3 and an emoji takes 4; in UTF-16 (the internal representation of a JavaScript string) characters outside the basic plane take two code units, which is why '😀'.length === 2. How the code point is finally drawn depends on the font: if the font has no glyph for it you get a box (the “tofu” character) and the browser walks through its fallback fonts — that is the root cause of the same text looking different on two computers.

Why there are two µ, two Å and two Ω

When Unicode absorbed earlier character sets it kept the numbers they had already assigned, and compatibility characters were the result: the micro sign µ from Latin-1 (U+00B5) next to Greek mu μ (U+03BC), Latin-1 Å (U+00C5) next to the ångström sign Å (U+212B), Greek Ω (U+03A9) next to the ohm sign Ω (U+2126), and even the degree sign ° next to the masculine ordinal indicator º (U+00BA), which look almost the same. For searching, sorting and de-duplication they are different characters, so mixing them in one text produces “the word is obviously there but the search finds nothing”. Unicode normalisation (NFKC) folds compatibility characters to the preferred form, and this page marks which of the two is the recommended one.

HTML entities: named, decimal, hexadecimal

HTML can write the same character three ways: a named entity °, a decimal reference ° and a hexadecimal reference °. Named entities exist only for the two thousand or so characters in the WHATWG table (♥ has one, ☃ does not), while numeric references work for any code point at all. Web pages today are almost all UTF-8, so you can simply type the character itself; entities survive in three situations: characters that must be escaped (< > & "), invisible characters that are awkward to type in source (  ​), and characters an editor might rewrite. JavaScript strings use \u{2603} and the CSS content property uses "\2603" — the detail card lists all of these forms.

Invisible characters

The “English and general punctuation” category holds a row of characters that copy as nothing at all: no-break space (U+00A0, stops two words from being split across lines), zero-width space (U+200B, allows a line break without taking any width), zero-width joiner (U+200D, joins two emoji into one, as in 👨‍👩‍👧), soft hyphen (U+00AD, shown as a hyphen only when the word breaks) and the bidirectional marks LRM and RLM. They are useful in typesetting, but inside code, passwords or file names they cause the nasty “looks identical but is not equal” bug. This page shows them as ␣ as a placeholder, while the copy is the real character; to see what is hidden in a piece of text, use Zero-width steganography or Escape tools to make them visible.

CJK punctuation and the full-width forms

Most CJK punctuation sits in two blocks, CJK Symbols and Punctuation (U+3000–303F) and Halfwidth and Fullwidth Forms (U+FF00–FFEF): the full-width comma ， is U+FF0C and has nothing to do with the ASCII comma U+002C; the middle dot · (U+00B7) separates foreign names and is a different character from the bullet • (U+2022); the CJK dash — is two em dashes (U+2014) in a row, and the CJK ellipsis …… is two U+2026 characters. A Chinese IME in Chinese mode turns , . ? into their full-width forms automatically, and maps the backtick key to ·, ^ to ……, _ to — and < > to 《》 — that is the basis for the “Chinese IME, Shift+X” notes in the “how to type it” column.

Superscripts and subscripts: characters or typesetting?

² and ³ are standalone Latin-1 characters, ⁴–⁹ and ₀–₉ live in the Superscripts and Subscripts block, and font coverage for them is good. They are still only characters that look raised: size and position come from the font, unlike a typeset superscript in Word made with ++Ctrl+Shift+=++ (an ordinary character plus formatting). The character version survives in plain text — file names, chat, code comments — while the typeset version exists only where rich text does. For chemical formulas such as H₂O, units such as m² and footnote numbers such as ¹, the character is the safer choice; for real mathematical typesetting use LaTeX or an equation editor, and this page ships a LaTeX command with every maths symbol.

What gets included

— characters that appear often in everyday documents, web pages, code and chat and that have a glyph in the system fonts of Windows, macOS, Android and iOS. Names follow the Unicode Character Database with a plain-English alias added, and named entities follow the WHATWG table.

What is left out

— the full emoji set (see Symbol library); private-use characters that only a few specialist fonts carry (Nerd Font icons, for instance); the mathematical alphanumeric symbols block (𝐀𝔸𝕒 and friends — see Fancy text); and braille (see Braille converter).

Typing methods

cover the general case only: Windows Alt codes need a numeric keypad and the Latin-1 range, macOS combinations assume a US keyboard layout, and pinyin IME candidates were checked against Microsoft Pinyin and Sogou, so other IMEs may differ.

Display

— a handful of characters (⏻⏼, some Dingbats) may show as a box on older Windows versions; that is a font problem, not a missing character, and the copy still displays correctly on other devices.

What it does not do

— no normalisation and no bulk conversion (to escape a whole block of text, use Escape tools).

All of the character data is loaded statically with the page, and searching, copying and code point lookups happen entirely in the browser without any request being sent. Recently copied and Favorites live in your local localStorage. Copying goes through the browser's clipboard API and only writes the character you clicked.

• Unicode Character Database — UnicodeData.txt（访问日期：2026-09-09）— the official name and category of every code point.
• Unicode Blocks.txt（访问日期：2026-09-09）— block ranges and names, which the overview table at the end of the page is built from.
• WHATWG HTML Standard: Named character references（访问日期：2026-09-09）— the complete list of named entities.
• Microsoft: Insert ASCII or Unicode characters (Alt codes)（访问日期：2026-09-09）— official notes on Alt codes and the Alt+X method.
• Unicode Standard Annex #15: Normalization Forms（访问日期：2026-09-09）— the definition of compatibility characters and NFKC folding.