Writing is not sounds written down. The primary function of English spelling is to represent meaning. Sounds do matter, but not just any grapheme can spell any sound. The grapheme that spells a sound in a word is informed by meaning, relatives, and etymology.
If writing were sounds written down, why does English have so many ways to spell the same sound? If the primary function of English spelling were to represent phonology, why does English have so many so-called “silent letters”? Because written language is not sound writing.
Letters can spell sounds. But not all letters spell sounds. Within the English spelling system, letters can function as markers (etymological, phonological, lexical) or be zeroed. A letter or letters that spell a sound in one word but not in another related word spells the zero phone as in in the <b> in <crumble> versus <crumb>. A marker can mark the history of the word such as the <l> in <walk>, the phonology of other grapheme such as the <e> in <make>, or mark a lexical spelling such as the <e> in <are>.
I will update this post as I collect information on more words.
<a>
<aisle>: the <a> is a marker added on the model of the French cognate <aile>, from Old French <ele> and ultimately from Latin <ala> (see <s>)
<b>
<aplomb>: the <b> spells the zero phone, from French <aplomb> and ultimately from Latin <plumbum>, compare to relatives <plumbago> and <plumbic> in which the <b> spells [b] (see <plumb>)
<bomb>: the <b> spells the zero phone, compare the relative <bombard> in which the <b> spells [b]
<climb>: the <b> spells the zero phone, from Old English <climban>, compare the relative <clamber> in which the <b> spells [b]
<comb>: the <b> spells the zero phone, from Old English <camb>, compare the relative <unkempt> (<b> and <p> share an etymological relationship)
<crumb>: the <b> spells the zero phone, compare the relative <crumble> in which the <b> spells [b]
<debt>: the <b> is an etymological marker, from Latin <debitum>, compare the relative <debit>
<doubt>: the <b> is an etymological marker, from Latin <dubitare>, from duo meaning “two,” <doubt> and <double> are relatives
<dumb>: the <b> spells the zero phone, compare the complex relative <dumbo> in which the <b> spells [b]
<limb>: “edge of a quadrant or other instrument,” the <b> spells the zero phone, from Latin <limbus>, compare the complex relative <limbal> in which the <b> spells [b]
<limb>: “part, body part distinct from the head,” from Old English <lim>, the <b> is unetymological from the 16th century and perhaps influenced by <limb> meaning “edge of a quadrant or other instrument,” possibly related to <limber> although the origin of <limber> is uncertain
<numb>: the <b> is unetymological from the 17th century to conform with the spelling of other words that end in the letters <mb>, from <nome> and ultimately Old English <niman>
<plumb>: the <b> spells the zero phone, from Latin <plumbum> via Old French, compare to relatives <plumbago> and <plumbic> in which the <b> spells [b]
<thumb>: the <b> is unetymological from the 13th century and perhaps influenced by <dumb>, from Old English <þuma>
<c>
<muscle>: the <c> spells the zero phone, compare the relative <muscular> in which the <c> spells the <k>
<ch>
<yacht>: the <ch> is an etymological marker, from Middle Low German <jacht> (shortened form of <jachtschip>) in which the <ch> spelled [x]
<d>
<handkerchief>: the <d> spells the zero phone in complex word <hand + kerchief> (see <hands>)
<hands>: the <d> can spell the zero phone in complex word <hand + s>, compare the base <hand> in which the <d> spells [d]
<handsome>: the <d> can spell the zero phone in complex word <hand + some>, compare the base <hand> in which the <d> spells [d]
<grandfather>, <grandpa>, <grandmother>, <grandma>, <grandson>: the <d> can spell the zero phone in complex familial words <grand + X>
<sandwich>: the <d> can spell the zero phone in some Englishes
<Wednesday>: the <d> is an etymological marker, from Old English <wodnesdæg> meaning “Woden’s day” (see <e>)
<e>
<are>: the <e> is a lexical marker, <are> is a function word and could thus be spelled <*ar> but the form developed in the 17th century and therefore has a lexical spelling
<camera>: the <e> can spell the zero phone, compare the related word <cameral> in which the <e> spells [ɛ]
<eye>: the first <e> is a lexical marker and the <ye> is a digraph, compare bye, rye, lye
<gardener>: the first <e> can spell the zero phone, compare the related word <garden> in which the <e> spells [ə]
<vegetable>: the second <e> spells the zero phone, compare the base <vegete> and the relatives <vegetative> and <vegetarian> in which the <e> spells [ə]
<Wednesday>: the <e> is a marker, from Old English <wodnesdæg> meaning “Woden’s day” (see <d>)
<were>: the <e> is a lexical marker, <were> is a function word and could thus be spelled <*wer> but the form developed in the 18th century and therefore has a lexical spelling
<g>
Words such as <gnat>, <gnaw>, and <gnu>, among others, contain the digraph <gn>. The <gn> spells [n].
Words such as <gnocchi> and <lasagna> contain the digraph <gn>. The <gn> spells [nj].
<apothegm>: the <g> spells the zero phone, compare the relative <apothegmatic> in which the <g> spells [g]
<diaphragm>: the <g> spells the zero phone, compare the relative <diaphragmatic> in which the <g> spells [g]
<gnathal>: the <g> spells the zero phone, compare the relative <prognathodon> in which the <g> spells [g], the base <gnath> means “jaw, cheek, lower jaw” from Greek <gnathos>
<gnathic>: the <g> spells the zero phone, compare the relative <prognathodon> in which the <g> spells [g], the base <gnath> means “jaw, cheek, lower jaw” from Greek <gnathos>
<gnathite>: the <g> spells the zero phone, compare the relative <prognathodon> in which the <g> spells [g], the base <gnath> means “jaw, cheek, lower jaw” from Greek <gnathos>
<paradigm>: the <g> spells the zero phone, compare the relative <paradigmatic> in which the <g> spells [g]
<phlegm>: the <g> spells the zero phone, compare the relative <phlegmatic> in which the <g> spells [g]
<reign>: the <g> is a marker, from Old French <reigne> and ultimately Latin <regnum>, compare the relatives <regime>, <regiment>, and <regal>
<sign>: the <g> spells the zero phone, compare the complex relatives <signature> and <signal> in which the <g> spells [g] (also morphological relatives including, but not limited to, <assign>, <assignment>, <cosign>, <consign>, <countersign>, <design>, <ensign>, <resign>, and <undersign>)
<syntagm>: the <g> spells the zero phone, compare the relative <syntagmatic> in which the <g> spells [g]
<h>
Words such as <ache>, <character>, <chord>, <Christmas>, and <echo>, among others, contain the digraph <ch>. The <ch> spells [k].
Words such as <brochure>, <chef>, <machine>, <mustache>, and <parachute>, among others, contain the digraph <ch>. The <ch> spells [ʃ].
<ephebic>: the <h> spells the zero phone while the <p> spells [f], <ephebic> is a complex word that consists of <ep + hebe + ic> and graphemes cannot cross morphemic boundaries
<ephedra>: the <h> spells the zero phone while the <p> spells [f], <ephedra> is a complex word that consists of <ep + hedr + a> and graphemes cannot cross morphemic boundaries
<ephemera>: the <h> spells the zero phone while the <p> spells [f], <ephemera> is a complex word that consists of <ep + hemere + a> and graphemes cannot cross morphemic boundaries
<ephor>: the <h> spells the zero phone while the <p> spells [f], <ephor> is a complex word that consists of <ep + hor> and graphemes cannot cross morphemic boundaries
<hyphen>: the <h> spells the zero phone while the <p> spells [f], <ephemera> is a complex word that consists of <hyp + hen> and graphemes cannot cross morphemic boundaries
<shepherd>: the <h> spells the zero phone, compare the base <herd> in which the <h> spells [h]
<vehicle>: the <h> spells the zero phone, compare the complex relative <vehicular> in which the <h> spells [h], from French <véhicule> and Latin <vehiculum> and ultimately Latin <vehere> meaning “to bear, carry, convey”
<i>
<business>: the <i> spells the zero phone, compare the <y> in the base <busy> (<business> is a complex word that consists of <busy + ness>, <y> and <i> can toggle)
<conscience>: the <i> spells the zero phone, compare the relative <science> in which the <i> spells [ɑi]
<family>: the <i> spells the zero phone, compare the relative <familiar> in which the <i> spells [ɪ]
<friend>: the <i> spells the zero phone in the complex word that consists of <fri + end>, compare the relative <fri + day -> Friday>
<fruit>: the <i> spells the zero phone, compare the relative <fruition> in which the <i> spells [ɪ]
<nescience>: the <i> spells the zero phone, compare the relative <science> in which the <i> spells [ɑi]
<prescience>: the <i> spells the zero phone, compare the relative <science> in which the <i> spells [ɑi]
<igh>
The trigraph <igh> can spell [ɑi] in words such as <blight>, <bright>, <flight>, <high>, <knight>, <nigh>, <plight>, <sigh>, <thigh>, <fight>, <light>, <might>, <night>, <right>, and <tight>.
<neigh>: the <igh> is an etymological marker, from Middle English <neighen> and ultimately from Old English <hnægan>
<k>
Words such as <knight>, <knife>, <know>, <knew>, and <known>, among others, contain the digraph <kn>. The <kn> spells [n].
<l>
<balk>: the <l> is an etymological marker, from Old English <balca>, related to <balcony> in which the <l> spells [l]
<caulk>: the <l> is an etymological marker, from Old North French <cauquer> and Late Latin <calicare> meaning “to stop up chinks with lime” and ultimately from Latin <calx>, related to <chalk>
<chalk>: the <l> is an etymological marker, from Old English <cealc> and ultimately from Latin <calx>, related to <calcium> and <calculate> in which the <l> spells [l], related to <caulk>
<could>: the unetymological <l> is a marker added on the model of <should> and <would>, from Old English <cuðe> past tense of <cunnan> meaning “to be able”
<half>: the <l> is an etymological marker from the Old English <half, halb, healf>
<should>: the <l> is an etymological marker, from Old English <sceolde> the past tense of <sceal> meaning “shall,” related to <shall> (see <could> and <would>)
<stalk>: the <l> is an etymological marker, from Old English <-stealcian> and from <stale>, related to <steal>
<talk>: the <l> is an etymological marker, from Middle English <talken>, related to <tale> and <tell>
<walk>: the <l> is an etymological marker, from Old English <wealcan>, related to <well> and <welter>
<would>: the <l> is an etymological marker, from Old English <wolde> the past tense and past subjunctive of <willan> meaning “to will,” related to <will> (see <could> and <should>)
<m>
<mnemonic>: the <m> spells the zero phone, compare the relative <amnesia> in which the <m> spells [m]
<n>
<autumn>: the <n> spells the zero phone, compare the complex relative <autumnal> in which the <n> spells [n]
<column>: the <n> spells the zero phone, compare the complex relatives <columnist> and <columnar> in which the <n> spells [n]
<condemn>: the <n> spells the zero phone, compare the complex relatives <condemnation> in which the <n> spells [n]
<damn>: the <n> spells the zero phone, compare the complex relative <damnation> in which the <n> spells [n]
<hymn>: the <n> spells the zero phone, compare the complex relative <hymnal> in which the <n> spells [n]
<solemn>: the <n> spells the zero phone, compare the complex relative <solemnity> in which the <n> spells [n]
<o>
<favorite>: the <o> can spell the zero phone, compare the relative <favor> in which the <o> spells [ɔ]
<history>: the <o> can spell the zero phone, compare the complex relative <historic> in which the <o> spells [ɔ]
<leopard>: the <o> spells the zero phone, compare the relative <Leo> in which the <o> spells [oʊ]
<p>
Words such as <pfeffernusse> contain the digraph <pf>. The <pf> spells [f].
<cupboard>: the <p> spells the zero phone, from <cup + board>, the <p> in the base morpheme <cup> spells [p]
<pterodactyl>: the <p> spells the zero phone in the complex word that consists of <pter + o + dactyl>, compare the relative <helicopter> in which the <p> spells [p]
<s>
<aisle>: the <s> is a marker added because of the <s> in the etymologically unrelated <isle>, from Old French <ele> and ultimately from Latin <ala> (see <a>)
<Illinois>: the <s> is an etymological marker, from the French adaptation of the Algonquian <Inoca> and <Ilinouek> meaning “ordinary speaker”
<island>: the <s> is a marker added because of the <s> in the etymologically unrelated <isle>, from Old English <yland>, <igland>, and <iegland> and ultimately from <ieg> meaning “island”
<isle>: the <s> is an etymological marker, from Old French <ile> and <isle> from Latin <insula> “island” (unrelated to <island>)
<t>
<apostle>: the <t> is an etymological marker, from Old English <apostol> and ultimately from Late Latin <apostolus> and Greek <apostolos>, built from <apo + stellein> meaning “send away, send forth” (see <epistle>)
<Christmas>: the <t> is an etymological marker, from Old English <Crīstesmæsse> meaning “Christ’s mass”, compare <Christ> and <Christian>
<epistle>: the <t> is an etymological marker, from Old English <epistol> and ultimately from Late Latin <epistola> and Greek <epistole>, built from <epi + stellein> meaning “send to, send as a message or letter,” compare the relative <epistolary> in which the <t> spells [t] (see <apostle>)
<th>
<asthma>: the <th> is an etymological marker, from Greek <asthma>
<clothes>: the <th> spells the zero phone, compare the relatives <cloth> and <clothe>
<isthmus>: the <th> is an etymological marker, from Greek <isthmos>
<u>
<build>: the <u> is a marker, from Middle English <bilden> and ultimately from late Old English <byldan>, compare the related word <bower> meaning “a shady place or shaded shelter,” the graphemes <u> and <w> are related so the <u> in <build> marks the etymological relationship with <bower>
<built>: see <build>
<buy>: the <u> is a marker, from Old English <bycgan>, the simple past of <buy> is <bought> in which the <ugh> is an etymological marker that marks the history from the Old English <bohte>, the <u> in <buy> marks a relationship with the <ugh> in <bought> (see <bought>)
<fluoride>, <fluorine>, <fluorite>: the <u> spells the zero phone, compare the relative <fluor> in which the <u> spells [u]
<ugh>
<bought>: the <ugh> is an etymological marker that marks a relationship with the <u> in <buy>, from the Old English <bohte> (see <buy>)
<w>
Words such as <wren>, <wrench>, <wrinkle>, <wrist>, and <write>, among others, contain the digraph <wr>. The <wr> spells [ɹ].
<answer>: the <w> is an etymological marker, from Old English <andswaru> and <answarian>, related to <swear> in which the <w> spells [w]
<sword>: the <w> is an etymological marker, from Old English <sweord, swyrd> in which the <w> spelled [w]
<two>: the <w> is an etymological marker that marks a relationship with words such as <twelve>, <twenty>, <twin>, <twice>, <twist>, and <twilight>
For more information about English graphemes, I cannot recommend the LEX Grapheme Deck more highly.
References
Etymonline: https://www.etymonline.com/
Wiktionary: https://en.wiktionary.org/wiki/Wiktionary:Main_Page
