H with Accent: Decode ĥ, ḥ and ẖ by the Scheme Behind Them

Which h with accent is a letter, and which one only romanizes another script?

An h with accent usually belongs to no alphabet. Only ĥ and ḥ appear in the CLDR alphabets behind the languages card above. The rest come from smaller orthographies and from romanization, where a mark on h stands for a letter of another script. Unicode's names list (opens in a new tab) notes the jobs in brief: Esperanto for ĥ, Finnish Romany for ȟ, Indic transliteration for ḥ, Semitic transliteration for ḫ and ẖ.

Unicode names each letter by its mark, which is also how people search for one. In those names, ĥ is the h with circumflex, ȟ the h with caron, ḥ the h with dot below, ḫ the h with breve below and ẖ the h with line below. ḣ, ḧ and ḩ carry a dot above, a diaeresis and a cedilla.

Wiktionary's entries fill in the rest. Its entries for ȟ (opens in a new tab), ḣ (opens in a new tab), ḥ (opens in a new tab) and ḧ (opens in a new tab) give these uses and sounds.

Letter	Who writes it	Sound, as Wiktionary gives it	Kind of use
ĥ	Esperanto	/x/, [ɣ] before a voiced obstruent	alphabet letter
ȟ	Lakota, Dakota	/χ/	alphabet letter
ȟ	Kalo Finnish Romani	/x/	alphabet letter
ȟ	Uralic Phonetic Alphabet	a voiceless laryngeal fricative	phonetic notation
ḣ	Assiniboine	/x/, an alternative form of ȟ	alphabet letter
ḥ	Asturian	/h/, said [h] or [x]	alphabet letter
ḥ	Kabyle, Tashelhit, Tuareg, Nootka	/ħ/	alphabet letter
ḥ	Arabic and Hebrew names; Sanskrit visarga	/ħ/ for the first two	romanization
ḧ	Northern Kurdish	/h/, for ح in Arabic loanwords	alphabet letter
ḧ	Arabic ة	none given	romanization
ḩ	Arabic ح	none given	romanization
ḫ	Arabic خ, Egyptian, Indic upadhmaniya	none given	romanization
ẖ	Arabic خ, Indic jihvamuliya	none given	romanization

The Kurdish row shows the whole idea in one letter. Arabic has two h letters, ه and ح. Wiktionary says Northern Kurdish uses ḧ in loanwords from Arabic to keep ح apart from ه.

One capital has a twist of its own. The names list says that in Nawdm, the lowercase of Ĥ is ɦ, not ĥ. The last section shows why software will not know that.

Why can one ẖ stand for ح in one text and خ in another?

Because the scheme, not the sound, picks the mark. The UN Working Group on Romanization Systems compares the Arabic schemes in its report on Arabic (opens in a new tab), version 5.0 of June 2018. Its table and notes give these spellings for the Arabic h letters.

Scheme	ح	خ	ة
UN system, approved 2017	ẖ	kh	h, or t in the construct form
UN system, 1972	ḩ	kh	h, or t in the construct form
BGN/PCGN, 1956	ḩ	kh	h, or t in the construct form
ISO 233, 1984	ḥ	ẖ	ẗ

The 1972 row follows from a note in the report: the old UN system used a cedilla wherever the 2017 one has a line below. BGN/PCGN 1956, it says, is almost identical to UN 1972 and differs only in how it writes the article and the -iyyah ending. The report adds that the French IGN system of 1973 and the Survey of Egypt system both write ḥ for ح.

So the same ẖ is ح on a map in the UN system and خ in a catalogue that follows ISO 233. Wiktionary's entry for ẖ (opens in a new tab) gives only the خ sense, and lists ḫ, ḵ, k͟h and x as other spellings of خ. A ḩ, with the hook, points to the older UN or BGN/PCGN style.

The kh column shows why the marks exist. The report warns that kh, dh, sh and th can be ambiguous, since each can also mean two Arabic letters in a row. The 2017 system may then use a middle dot, as in Ad·ham. ISO 233 avoids the problem by giving every Arabic character a unique equivalent, and ẖ is how it fits خ into a single letter.

Which h letters do Egyptologists use for the hieroglyphs?

Egyptian has its own set of h signs, and the most widely used scholarly schemes write them h, ḥ, ḫ and ẖ. Wiktionary's appendix of Egyptian transliteration schemes (opens in a new tab) lines up the scholars' choices from the decipherment to the present day. It highlights Erman and Grapow's and Gardiner's as schemes that gained wide use in their time.

Hieroglyph	Erman and Grapow, 1926–1931	Gardiner, 1927 and 1957	Budge, 1920
𓉔, U+13254	h	h	h
𓎛, U+1339B	ḥ	ḥ	ḥ
𓐍, U+1340D	ḫ	ḫ	kh
𓄡, U+13121	ẖ	ẖ	kha

The letters match the Arabic ones in shape only. Gardiner's ẖ is the last sign in the table, not ISO 233's خ and not the UN's ح. A text that mixes Arabic and Egyptian names needs its scheme named once, near the top. Wiktionary gives 𓐍 as the sign behind ḫ, and notes the Egyptian word ḫ, placenta or child, written with it.

Budge's earlier scheme, of 1898 and 1910, is harder to undo. The appendix shows it writing χ or kh for both of the last two signs. A kh in that scheme cannot be turned back into ḫ or ẖ without the hieroglyphs.

When may an Esperanto writer swap ĥ for k?

Only where a dictionary already lists the k form. In the Esperanto alphabet, ĥ is the one h with accent, and it follows h. Several ĥ words come from Greek χ, Wiktionary's etymologies show, and some of them gained a second spelling with k. Others did not, and a blind swap lands on a different word. Wiktionary's entries show both cases.

With ĥ	With k	What Wiktionary says
ĥemio	kemio	synonyms, both chemistry
teĥniko	tekniko	teĥniko is an alternative form of tekniko, technique
ĥaoso	kaoso	kaoso is an alternative form of ĥaoso, chaos
ĥoro	koro	different words: ĥoro is choir, koro is heart
eĥo	eko	different words: eĥo is echo, eko is a start

The headword moves between rows. For technique, the k form is the main entry; for chaos, the ĥ form is. Read ĥemio (opens in a new tab), teĥniko (opens in a new tab), ĥaoso (opens in a new tab), ĥoro (opens in a new tab) and eĥo (opens in a new tab) for the details.

A spell-checker that turns every ĥ into k would sing in the heart instead of the choir. When the letter itself cannot be typed, Wiktionary lists ehho and ehxo for eĥo, and hhoro and hxoro for ĥoro: the h-system and the x-system. The g with accent page covers where those systems come from.

What do lowercase, uppercase and accent stripping do to the h letters?

Each h with accent is a plain h plus one combining mark, so code handles them predictably, with a few traps. We checked them in Node.js v26.0.0 (ICU 78.3) on 2026-10-10. This is our own measurement.

• Nawdm's Ĥ. Lowercasing Ĥ gave ĥ, U+0125. The names list says Nawdm pairs Ĥ with ɦ instead. Unicode's default case mapping has no such rule, so a Nawdm text needs its own.
• ẖ in capitals. No capital ẖ exists. SpecialCasing.txt (opens in a new tab) maps it to H plus U+0331, two code points, and NFC left them as two. Lowercasing the pair and normalizing returned the single ẖ. The t with accent page shows the same pattern with ẗ.
• Accent stripping. Decomposed with NFD, ḥ, ḩ, ḫ and ẖ each split into h and one combining mark. Strip the marks, and an ISO 233 text writes ه, ح and خ all as h. A search folded that way cannot tell Ḥassān, from the names card below, from Hassan.

So keep the marked form as the stored spelling, and fold only a copy for search. In data from more than one scheme, store the scheme too: the same ẖ will not mean the same Arabic letter twice.