Silicon NMR: why the glass appears in the spectrum, why proton decoupling can erase a signal, and how M, D, T and Q units are counted
Share
Silicon NMR: why the glass appears in the spectrum, why proton decoupling can erase a signal, and how M, D, T and Q units are counted
A spin-½ nucleus with sharp lines, slow relaxation and a negative magnetogyric ratio, measured in a tube that itself contains silicon
A silicon-29 spectrum shows how many oxygen atoms each silicon atom carries and how many siloxane bonds it has formed, which is what a chemist needs to know about a silane, a silicone or a silsesquioxane resin. The measurement is made difficult by the glass of the tube and the probe, which gives a signal of its own, by proton decoupling, which can cancel or invert a signal, and by relaxation times of a minute or more, which make a quantitative spectrum slow.
Five terms used in this article. Receptivity tells how strong the signal of a nucleus is at natural abundance; we give it relative to 13C. T1, the spin–lattice relaxation time, describes how quickly the nuclei return to equilibrium after a pulse. The nuclear Overhauser effect (NOE) is the change in the signal of one nucleus when a nearby nucleus is irradiated. M, D, T and Q are the names of silicon atoms bonded to one, two, three and four oxygen atoms. Polarisation transfer hands the larger population difference of the protons to a nucleus that is coupled to them.
At a glance: the observation and its cause
| Observation | Cause | Comment |
|---|---|---|
| A broad signal lies near −110 ppm in every spectrum | Silicon in the glass of the tube and in the glass and quartz of the probe | A blank, polarisation transfer, or a tube and probe without silicon |
| A signal is weak, absent or inverted under proton decoupling | The NOE of 29Si is negative | Inverse-gated decoupling or a relaxation reagent |
| The spectrum takes hours | T1 of tens of seconds, and a receptivity twice that of 13C | A relaxation reagent or polarisation transfer |
| DEPT integrals do not match the composition | The transfer has a different efficiency for each group | Response factors from a sample of known composition |
| On hydrolysis one signal moves to higher and another to lower frequency | The effect of OH depends on the number of methyl groups on silicon | +9.4 ppm from Si(OEt)4 to Si(OH)4; −3.2 ppm from Me3SiOEt to Me3SiOH |
| A signal near −22 ppm does not belong to the compound | Silicone grease or oil | The shift of a poly(dimethylsiloxane) chain |
| Two small lines flank the 1H signal of a methyl group on silicon | The 4.7 % of molecules that contain 29Si | Satellites, 6.6 Hz apart |
| A silicon atom bonded to boron gives no signal | Relaxation of the quadrupolar boron nucleus | Reported for a silylborate; see Part 7 |
The nucleus in numbers
Silicon has a single magnetic isotope, 29Si, which has spin ½ and makes up 4.6832 % of natural silicon.[1] Because a spin-½ nucleus has no quadrupole moment, its lines are free of the broadening described in Part 7 of this series, and dilute tetramethylsilane in CDCl3 gives a line only 0.15 Hz wide.[2]
The magnetogyric ratio γ is −5.3190 × 107 rad s−1 T−1, about a fifth of the proton value and opposite in sign. Silicon therefore resonates at 19.867187 % of the proton frequency, which is 79.47 MHz on a 400 MHz spectrometer. Its receptivity is 2.16 times that of 13C but only 3.68 × 10−4 of that of 1H.[1] In absolute terms the signal is weak: at 11.7 T it comes from a population excess of one nucleus in 125 500.[3]
Silicon-29 beside three other spin-½ nuclei
| Nuclide | γ (107 rad s−1 T−1) | Frequency (MHz, 400 MHz instrument) | Intensity with full NOE (1 + γH/2γX) | Polarisation transfer (γH/|γX|) |
|---|---|---|---|---|
| 13C | 6.728 | 100.58 | 2.99 | 3.98 |
| 15N | −2.713 | 40.55 | −3.93 | 9.86 |
| 29Si | −5.319 | 79.47 | −1.51 | 5.03 |
| 119Sn | −10.032 | 149.16 | −0.33 | 2.67 |
γ and the frequency ratios are taken from the IUPAC Recommendations.[1] The last two columns are our arithmetic. The NOE column gives the intensity under the full NOE relative to a spectrum without it, for a nucleus that relaxes only through its dipolar interaction with protons;[4] a negative value means an inverted signal. The last column is the factor by which the polarisation of the protons exceeds that of the nucleus, and it is this excess that polarisation transfer makes available. The bars denote the magnitude of γ, whose sign plays no part here.
The negative sign of γ goes unnoticed in a simple spectrum and becomes important once the protons are decoupled. The full NOE triples a 13C signal, and it inverts a 15N signal at nearly four times its original size. For 29Si the factor is −1.51, so a partial NOE can bring the signal close to zero.
Chemical shifts
Shifts are quoted against tetramethylsilane at 0 ppm, with negative values at lower frequency. Most compounds are found between +50 and −200 ppm, and tetraiodosilane at −350 ppm marks the low-frequency end for tetravalent silicon.[5]
M, D, T and Q units
Nothing changes the shift as much as the number of oxygen atoms on silicon, and siloxane chemists name their building blocks accordingly.[5, 6] An M unit, R3SiO, has one oxygen atom and ends a chain. A D unit, R2SiO2, has two and extends the chain. A T unit, RSiO3, is a branch point, and a Q unit, SiO4, sits in a three-dimensional network. Each of these oxygen atoms is shared with the neighbouring silicon atom. When R is methyl, M units are found at 7 to 9 ppm, D units at −17 to −22 ppm, T units at −55 to −65 ppm and Q units at −105 to −115 ppm.[5] Phenyl groups shift a D unit to lower frequency, to −33 to −35 ppm for MePhSiO and to −46 to −49 ppm for Ph2SiO.[7]
Figure 1. Top: the four siloxane units; the dashed bonds lead to the next silicon atom. Schematic. Bottom: the shift ranges of methyl-substituted units and the shifts of nine reference compounds,[5, 8] with the region of the glass signal.[2] Our drawing of published values.
Reference points on the silicon-29 scale
| Compound | Class or unit | δ (ppm) |
|---|---|---|
| Trimethylsilyl triflate, Me3SiOSO2CF3 | silylating agent | 44 |
| Chlorotrimethylsilane, Me3SiCl | chlorosilane | 30 |
| Triethylsilanol, Et3SiOH | silanol | 19.3 |
| Hexamethyldisiloxane, (Me3Si)2O | M2 | 6.53 |
| Hexamethyldisilazane, (Me3Si)2NH | silazane | 2.2 |
| Tetramethylsilane, Me4Si | reference | 0.0 |
| Cyclic trimer D3, (Me2SiO)3 | D, strained ring | −9 |
| Trimethylsilane, Me3SiH | hydrosilane | −17 |
| Tetrachlorosilane, SiCl4 | chlorosilane | −18 |
| Hexamethyldisilane, Me3SiSiMe3 | disilane | −19.5 |
| Cyclic tetramer D4, (Me2SiO)4 | D | −19.86 |
| Silicone oil, (Me2SiO)x | D | −22.0 |
| Methyltriethoxysilane, MeSi(OEt)3 | T0 | −44.06 |
| Phenylsilane, PhSiH3 | hydrosilane | −60 |
| Octaphenyl cage, (PhSiO1.5)8 | T3 | −78.3 |
| Tetramethoxysilane, Si(OMe)4 | Q0 | −78.54 |
| Tetraethoxysilane, Si(OEt)4 | Q0 | −82.04 |
| Hydrido cage, H8Si8O12 | T3 | −84.5 |
| Tetrakis(trimethylsiloxy)silane, (Me3SiO)4Si | M4Q | 8.62 and −104.08 |
| Glass and fused quartz | Q4 region | about −110 |
The shifts are taken from the review of Uhlig and Marsmann,[5] apart from those of neat methyltriethoxysilane,[8] the octaphenyl cage in CDCl3[9] and the glass.[2] The superscript on D, T and Q is the number of siloxane bonds at that silicon atom. Where a value has no decimals, the source gives it to about ±2 ppm.
Ring strain is visible in the cyclic siloxanes. The strained trimer resonates at −9 ppm, whereas the tetramer at −19.86 ppm lies close to a linear silicone oil at −22.0 ppm. That position, near −22 ppm, is also where silicone grease from laboratory equipment turns up as an impurity.[5]
Substituent effects are not additive
A 29Si shift cannot be estimated by adding one increment for each substituent.[10] In the series Me4−nSiCln, the first chlorine atom raises the shift to 30 ppm and the second to 32 ppm, after which it falls to 12 ppm and then to −18 ppm in SiCl4.[5] The ethoxy series turns back after the first oxygen atom: the neat liquids from Me4Si to Si(OEt)4 resonate at 0, +14.66, −5.57, −44.06 and −82.22 ppm.[8] Nor do the tetrahalides follow a simple order, with SiF4 at −112, SiCl4 at −18, SiBr4 at −92 and SiI4 at −350 ppm.[5]
Figure 2. 29Si shifts of the chloro series[5] and of the ethoxy series as neat liquids.[8] Our drawing of published values.
The more electronegative groups silicon already carries, the less a further substituent changes the shift of an alkoxy- or chlorosilane. The quantitative theories available at the time did not predict this, and back-bonding to silicon was offered as the explanation.[11] A later treatment traces the trend to the paramagnetic part of the shielding, which depends on how unequally the silicon p orbitals are occupied.[8] Calculations now reproduce the solution shifts of tetracoordinate silanes to about 1 ppm, provided that spin–orbit corrections are included.[12]
Solvent and concentration
The medium can shift a signal further than a hydrolysis step does. Neat trimethylethoxysilane resonates at +14.66 ppm. At 2.24 mol/L in ethanol the signal moves to +16.71 ppm, and three equivalents of water take it to +17.78 ppm. Tetraethoxysilane changes by only 0.4 ppm across the same three conditions.[8] Chlorotrimethylsilane stays between 30.0 and 30.7 ppm in benzene, toluene and chloroform,[12] and the shift of tetramethylsilane itself varied by 0.7 ppm over six solvents.[13]
Five- and six-coordinate silicon
Silatranes resonate at lower frequency than the corresponding triethoxysilanes, because the Si←N bond makes silicon pentacoordinate.[14] In one family of chelates, the ionic pentacoordinate complexes lie between −59 and −75 ppm, and their signals do not move with temperature. In the same study, the signal of a compound that exists as an equilibrium between a penta- and a hexacoordinate form moved from −51.1 ppm at 360 K to −116.5 ppm at 170 K, where the hexacoordinate form is favoured.[15]
Hydrolysis and condensation of alkoxysilanes
Alkoxysilanes react with water in two stages: the alkoxy groups are hydrolysed to silanols, and the silanols then condense to form siloxane bonds. Each siloxane bond moves the signal to lower frequency. Starting from tetramethoxysilane, the species without OH groups are found at −78.25 ppm (Q0), −85.85 ppm (Q1) and −93.70 ppm (Q2). Q3 sites lie between −98.60 and −101.75 ppm, and Q4 sites near −110 ppm, where they overlap the signal of the glass.[16] The condensed units of methyltriethoxysilane, T1, T2 and T3, resonate at −51, −57 and −66 ppm.[17]
Hydrolysis is less uniform. Whether it moves the signal to higher or to lower frequency depends on how many methyl groups silicon carries.[8]
From alkoxysilane to silanol: measured shifts
| Hydrolysis | Alkoxysilane (ppm) | Silanol (ppm) | Change (ppm) |
|---|---|---|---|
| Me3SiOEt → Me3SiOH | 17.78 | 14.60 | −3.18 |
| Me2Si(OEt)2 → Me2Si(OH)2 | −3.85 | −4.17 | −0.32 |
| MeSi(OEt)3 → MeSi(OH)3 | −42.14 | −37.85 | +4.29 |
| Si(OEt)4 → Si(OH)4 | −81.82 | −72.43 | +9.39 |
| Me3SiOMe → Me3SiOH | 20.72 | 14.84 | −5.88 |
| Me2Si(OMe)2 → Me2Si(OH)2 | 1.25 | −4.05 | −5.30 |
| MeSi(OMe)3 → MeSi(OH)3 | −38.24 | −38.43 | −0.19 |
| Si(OMe)4 → Si(OH)4 | −78.29 | −72.21 | +6.08 |
Measured at 2.24 mol/L in the parent alcohol with three equivalents of water, without catalyst, at 298 K and 79.49 MHz.[8]
On a Q silicon atom, each OH group that replaces an alkoxy group moves the signal 1 to 3 ppm to higher frequency. One methyl group on silicon makes the steps smaller, and in MeSi(OMe)3 they almost disappear, so that all four species from the trimethoxysilane to the silanetriol fall within 0.3 ppm. With two or three methyl groups the signal moves the other way, to lower frequency.[8] The hydrolysis products of methyltrimethoxysilane, which lie so close together, were assigned with INEPT and DEPT, because these experiments tell the species apart by their couplings to protons.[18]
Figure 3. 29Si shifts of the condensation products of methyltriethoxysilane[17] and of the hydrolysis and condensation products of tetramethoxysilane.[16] In each row the dark number is the shift of the species without OH and the gold number that of the species with the most OH groups. Our drawing of published values; the two step sizes are our arithmetic.
Silicate anions in water follow the same order as the neutral species. The monomer gives a signal at −71 ppm, the dimer or the cyclic trimer at −81 ppm, the cyclic tetramer or the double three-ring at −89 ppm, and the cubic double four-ring cage at −99 ppm.[19]
A series of such spectra gives the rates of the individual steps. In acidified ethanol and water at 298 K, the first hydrolysis step of five organotriethoxysilanes RSi(OEt)3 slowed in the order methyl, vinyl, ethyl, phenyl, propyl, its rate constant falling from 4.51 × 10−3 to 7.71 × 10−4 L mol−1 min−1, and tetraethoxysilane was slower still. The second and third steps were 2.8 and 13.8 times as fast as the first, and condensation did not begin until the silanetriol had reached about 0.30 mol/L.[20] With tetramethoxysilane, hydrolysis remained incomplete at a water-to-alkoxide ratio of 2.8, whereas at a ratio of 8 it was fast and complete and was followed by a slower condensation to branched polymers.[16] In an aqueous emulsion of methyltriethoxysilane, methylsilanetriol and its dimer could still be detected after 40 days.[17]
Cages and resins
A silsesquioxane is built entirely of T units. In a closed cage every one of them is T3, so a cage with eight identical corners gives a single line, at −78.3 ppm for the octaphenyl cage in CDCl3,[9] at −79.4 ppm for the octavinyl cage and at −84.5 ppm for H8Si8O12.[5] When one corner differs from the other seven, four signals are expected, and often only three are seen because two of them coincide.[9]
A resin, unlike a cage, also contains T2 silicon atoms, each of which still carries an OH or alkoxy group. In a polysilsesquioxane with 2-(3,4-epoxycyclohexyl)ethyl groups, the T3 signal lies at −64 to −70 ppm and the T2 signal at −54 to −60 ppm. Patents specify such resins by the ratio of the two integrals, for example 5 to 500 for an adhesive used in making semiconductor devices.[21, 22] The same integrals give the degree of condensation, which is the fraction of all possible siloxane bonds that have actually formed:
(T1 + 2 T2 + 3 T3) / [3 (T0 + T1 + T2 + T3)]
In a resin that contains only T2 and T3 units, a T3/T2 ratio of 5 corresponds to a degree of condensation of 94 %, and a ratio of 20 to 98 %. MQ resins consist of M units together with Q4 and Q3 units, and each Q3 unit carries a silanol group.[23]
Couplings to silicon
Ranges of coupling constants
| Coupling | Range (Hz) | Example |
|---|---|---|
| 1J(29Si,1H) | 75–420 | 205 Hz in dimethylethoxysilane[3] |
| 2J(29Si,1H), SiCH3 | 3–10 | typically 6.6 Hz in alkylsilanes[2] |
| 3J(29Si,1H) | 1–8 | 3.5 Hz in Si(OCH3)4[16] |
| 1J(29Si,13C) | 37–113 | – |
| 1J(29Si,19F) | 108–488 | – |
| 1J(29Si,29Si) | 23–186 | – |
The ranges are magnitudes taken from the review of Uhlig and Marsmann.[5] Because γ is negative, most one-bond couplings to silicon are negative as well; those to fluorine and phosphorus are exceptions.[5]
Proton decoupling removes these couplings from the 29Si spectrum, but they can still be seen from the proton side. In the 4.7 % of molecules that contain 29Si, the 1H signal of a methyl group on silicon is split into a doublet. Its two lines show up as satellites, each about 2.3 % of the main signal, on either side of it.[2]
Couplings to quadrupolar nuclei obey the rules set out in Part 7. Triisopropylsilyllithium, made from the silylborane i-Pr3Si–B(pin) and methyllithium in THF, gives a 29Si signal at 11.8 ppm at −100 °C. It is a quartet with J(29Si,7Li) = 52 Hz, as expected for a single 7Li neighbour of spin 3/2. The intermediate silylborate gave no 29Si signal at all, and the authors attribute this to the quadrupolar boron atom.[24]
The signal of the glass
Both the glass of the tube and the glass and quartz parts of the probe contain silicon, and together they add a broad signal at about −110 ppm, in the Q4 region, to every spectrum.[2, 5] One study found it at −112 ppm with a width of 2.1 kHz, which amounts to 21 ppm at the 29Si frequency of 99.3 MHz.[25] For quartz tubes, one facility quotes a width of about 4.8 kHz at about −100 ppm.[26]
M, D and T signals are far enough away to be unaffected. The glass signal gets in the way of Q units, silica and silicates, and of any weak signal that sits on its flanks. A quartz tube, which rids an 11B spectrum of the tube’s share of the glass signal, does not help here, because fused quartz is itself silicon dioxide.
Figure 4. The glass signal and its removal by subtraction of a blank. Schematic; the positions follow the ranges of Figure 1.
Remedies
- A blank. A spectrum of the solvent alone is recorded under the same conditions and subtracted.[2, 5] This costs a second measurement, and it fails when the sample has broad signals of its own in the same region, as dilute silicate gels do.[2]
- Polarisation transfer. INEPT and DEPT detect only silicon that is coupled to protons or fluorine, so the glass stays invisible.[5]
- A composite pulse. The DEPTH sequence, a composite pulse of the kind described in Part 7 for 11B, suppresses the signals of the probe materials. The price is some sensitivity and, in a probe with an inhomogeneous radiofrequency field, altered relative integrals.[27, 28]
- A tube and probe without silicon. Patents on silicones and on spin-on-glass solutions call for Teflon tubes, or for tubes and probes free of silicon, with diameters of 10 and 16 mm.[29, 30, 31] One facility describes removing the quartz from the probe and using a Teflon tube, at the risk of damaging the probe. It also describes a probe and tubes made of sapphire, which are very expensive and rule out 27Al measurements on that probe.[2]
For 5 mm probes, Deutero makes a kit that combines a PTFE liner with a glass mount for the spinner, so that no glass lies in the measuring region. We supply it on request; see the PTFE liner. The kit removes the tube’s contribution to the glass signal, but not the probe’s.
Proton decoupling and the negative NOE
The protons are normally decoupled, because coupling to the organic groups splits each silicon signal into many lines; tetramethylsilane, with its twelve equivalent protons, gives thirteen.[2, 5] Decoupling also saturates the protons, however, and through the NOE this changes the intensity of the silicon signal.
How large the change is depends on how much of the silicon relaxation takes place through the dipolar interaction with protons. If all of it does, the intensity is multiplied by 1 + γH/2γSi = −1.51, which means that the signal is inverted and half as strong again.[4] If none of it does, the intensity stays as it was. Between these limits the factor is 1 − 2.51 f, where f is the dipolar share of the relaxation rate, and it passes through zero at f = 0.40. There the dipolar relaxation time is 1.5 times that of all the other paths together, which is the condition the literature gives for a signal that vanishes.[5]
Figure 5. Intensity of a 29Si signal under continuous proton decoupling, relative to a spectrum without NOE, against the share of the relaxation rate that is dipolar with protons. Our own calculation from the IUPAC magnetogyric ratios,[1] for small molecules in mobile solution; it is not measured data.
In organosilicon compounds the dipolar mechanism competes with spin–rotation, and the protons of neighbouring molecules can account for 15 to 20 % of the relaxation.[32] The dipolar share varies from one compound to the next and from one site to another, and it changes with temperature.[5] It has been worked out from T1 and the NOE for twelve trimethylsilyl compounds.[33] A single spectrum can therefore contain positive signals, negative signals and signals that are missing. Among the silicate anions of an alkaline solution, for example, the cubic octameric cage showed a large NOE and the smaller oligomers did not.[34]
Remedies
- Inverse-gated decoupling. The decoupler is switched on only while the signal is being acquired. That is long enough to collapse the multiplets and too short for the NOE to build up. The wait between scans has to be long, 3 to 5 times T1.[5]
- A relaxation reagent. Chromium(III) acetylacetonate, Cr(acac)3, at about 0.01 mol/L gives silicon a relaxation path that does not involve protons, without changing the shifts.[5] The NOE is suppressed and T1 is shortened at the same time.[32, 35]
- No decoupling. A patent on spin-on-glass solutions records its spectra without any proton decoupling, to keep the integrals quantitative.[29]
When a 29Si signal is missing from a proton-decoupled spectrum, the site may have a partial NOE that cancels it. Only a spectrum recorded with inverse-gated decoupling or with a relaxation reagent shows whether the signal is truly absent.
Slow relaxation
The magnetic moment of 29Si is only weakly coupled to its surroundings,[19] and as a result T1 ranges from a few seconds to minutes.
Published T1 values of silicon-29 in solution
| Sample | T1 (s) | Source |
|---|---|---|
| Tetramethylsilane, dilute in CDCl3 | 5 | [2] |
| Tetramethylsilane, test sample at 4.2 T | 16.2 | [19] |
| Cubic silicate cage, about 0.9 mol/L in alkaline solution | 5.3 | [19] |
| Dimethylethoxysilane | 38 | [3] |
| Four silanols | 21 to 103 | [3] |
| Tris(tert-butoxy)silanol | 138.4 | [3] |
The values of [3] come from a hyperpolarisation study at 11.7 T. The two values for tetramethylsilane were measured on different samples and at different fields.
T1 also depends on what else is in the tube. In aqueous silicate solutions it was markedly affected by paramagnetic species, some of which had been leached from the glass of the tubes.[36]
For a quantitative spectrum the delay between pulses must be about five times the longest T1,[27] which comes to 11.5 minutes per scan when T1 is 138 s. On a benchtop spectrometer, a quantitative spectrum of a silicone copolymer took 51 hours for 1024 scans, one scan every 3 minutes.[37]
A smaller pulse angle allows a shorter delay. The review of Uhlig and Marsmann gives 40° pulses with a recovery time of about 20 s under inverse-gated decoupling, and 30° pulses about every 20 s for compounds without protons.[5]
A relaxation reagent shortens the delay further. Industrial methods for siloxanes add 0.02 to 0.03 mol/L of Cr(acac)3 and then use 90° pulses, gated decoupling and delays of 6 to 13 s.[31, 38, 39] The method with the 6 s delay calls its own spectra semi-quantitative and compares them only with one another.[31] A kinetic study that used about 0.02 mol/L and no delay at all recorded each spectrum of 300 scans in 12 to 15 minutes.[20]
The reagent stays in the sample and can interfere with its chemistry. At neutral pH, both Cr(acac)3 and MnCl2 altered the rates of hydrolysis and condensation of organosilicates.[40]
Polarisation transfer: the signal of the protons
INEPT and DEPT hand the polarisation of the protons to silicon through the scalar coupling between the two nuclei. The silicon signal then owes its size and its repetition rate to the protons, and the NOE no longer enters.
- Population. The polarisation of the protons is larger than that of silicon by γH/|γSi| = 5.03. When silicon is coupled to n equivalent protons, the best DEPT signal is n (γH/|γSi|) sin θ cosn−1 θ, obtained with sin θ = n−1/2.[16] This works out at 5.0 for one proton, 5.8 for three, 7.8 for six, 9.4 for nine and 10.8 for twelve; the arithmetic is ours and ignores relaxation during the sequence. INEPT enhancements measured for a range of organosilicon compounds lay between 2.9 and 9.3.[41, 42]
- Repetition. The sequence can be repeated as fast as the protons relax, and their T1 is shorter than that of silicon.[16, 41]
- Sign. The transfer does not rely on the NOE, so the negative γ no longer matters. The enhancement also survives the addition of a paramagnetic relaxation reagent, which an NOE does not.[41]
Figure 6. 29Si signal per scan for a single pulse with and without NOE and for polarisation transfer from 1 to 12 equivalent protons, relative to a single pulse without NOE. Our own calculation from the IUPAC magnetogyric ratios[1] and the DEPT formula of [16], without relaxation losses; the measured range is that reported for INEPT.[41, 42]
The delay in the sequence is set from the coupling constant. For Si(OCH3)4, with its coupling of 3.5 Hz, the authors set it to 142 ms, close to 1/(2J).[16]
Measured gains
- Using INEPT with one standard set of parameters, the shifts of trimethylsilyl derivatives of unknown structure could be measured on 20 µmol of compound.[43]
- On a benchtop spectrometer, a DEPT spectrum of a silicone copolymer recorded in just over 20 minutes reached about the same signal-to-noise ratio as the 51-hour quantitative spectrum.[37]
- Detecting the signal on the protons gains more still. A 1H–29Si HMBC spectrum of such a copolymer reached a signal-to-noise ratio of 268 in 171 minutes, where DEPT reached 52 in 680 minutes. That is a tenfold gain per unit of time, paid for with a digital resolution of 7.7 Hz instead of 0.3 Hz.[44]
Limits
- No coupled proton, no signal. Si(OH)4 could not be observed by DEPT,[16] and in aqueous silicates fast proton exchange rules the method out.[19] Q units and silica, which have no coupled protons, are likewise invisible.
- The integrals need calibration. In a copolymer of dimethylsiloxane and methylhydrosiloxane units, the quantitative spectrum gave a unit ratio of 1.09 and the DEPT spectrum 0.79. Once response factors had been taken from a characterised sample of the same kind, the DEPT spectrum could be used quantitatively.[37]
- Species that stop tumbling disappear. When the polymerisation of a bridged alkoxysilane was followed by DEPT, signal was lost at certain concentrations, most probably because of microphase separation.[45]
Without a coupling, and beyond thermal polarisation
Where silicon has no proton to couple to, the DEFT sequence offers another way. It drives the magnetisation back to equilibrium after each acquisition, so that the pulses can follow one another quickly. In the same measuring time it gave about 30 times the signal of single-pulse acquisition for tetramethylsilane and about 100 times for a silicate cage. With a 27 cm3 sample at 4.2 T, an 800 µmol/L solution of 29Si-enriched silicic acid could be measured in 3 hours, and the detection limit was estimated at about 80 µmol/L.[19]
Hyperpolarisation with parahydrogen goes far beyond these gains. In a method that uses an iridium catalyst and benzylamine, the polarisation first reaches the OH proton of a silanol and from there passes to silicon. The 29Si signal of tris(tert-butoxy)silanol at 11.7 T was 82 times the thermal signal with the first catalyst and 2852 times after optimisation, which was enough to follow its reaction with triflic anhydride with one spectrum every 5 s. Combined with INEPT, the same approach gave a 772-fold gain for dimethylethoxysilane.[3]
Applications
Silicones: composition, chain length and sequence
In a silicone oil or copolymer each kind of unit gives its own signal, and the M signal of the chain ends serves as the yardstick for the rest. In a trimethylsilyl-terminated copolymer of dimethylsiloxane and methylhydrosiloxane units, the dimethyl units resonate at about −22 ppm and the methylhydro units at about −37 ppm. If the integral of the end groups is set to 2, the other two integrals are the average numbers of each unit per chain.[37]
Figure 7. End-group analysis of a trimethylsilyl-terminated copolymer of dimethylsiloxane (D) and methylhydrosiloxane (DH) units. Schematic; the integrals are published values,[44] and the positions of D and DH are those of [37].
One published set of quantitative integrals reads 2.00 for the end groups, 43.22 for the dimethyl units and 44.43 for the methylhydro units.[44] Multiplying each integral by the mass of its unit (81.19, 74.16 and 60.13 g/mol) and adding the products, we obtain a number-average molar mass of 6039 g/mol, which is the value the authors report.[44]
A 1H spectrum cannot give this result, because the methyl protons of the chain and those of the end groups are not resolved from each other. For one sample the 1H spectrum, evaluated without the end groups, gave a unit ratio of 0.91, where 29Si gave 1.09. With longer chains the end groups become the limiting signal. In two quantitative 29Si spectra of longer polymers, each recorded for more than 25 hours, they could not be seen, whereas calibrated DEPT spectra recorded in 11 hours showed them.[37]
The fine structure of each signal records the order in which the units follow one another. In copolymers of dimethylsiloxane with methylphenyl- or diphenylsiloxane units, every unit gives three groups of signals, one for each triad, and the run number calculated from their integrals agreed with that of a random copolymer.[7] In methylphenylsiloxane copolymers the splittings have been assigned as far as pentads.[46] For dimethyl–diphenyl copolymers used in electronic components, run lengths, sequence lengths and linkage probabilities were compared between production lots,[47] which matters because the microstructure affects the glass transition and melting temperatures.[48]
The same quantities appear in patents. One characterises polyether-modified siloxanes by their average numbers of trimethylsiloxane, dimethylsiloxane and branching units,[35] and another defines an aerogel by the ratio of its Q and T signal areas to its D signal area.[49]
Spin-on glass and dielectric films
A patent on spin-on-glass solutions for the insulating films of semiconductor devices describes its measurement in detail. The sample is 1.5 mL of the solution mixed with 1.5 mL of acetone-d6 and about 40 mg of Cr(acac)3, which makes the reagent about 0.04 mol/L, and the spectrum is the sum of 10 000 to 45 000 scans at 53.67 MHz. The integrals are grouped by the number of Si–C bonds on each silicon atom. In one solution, 49 % of the silicon had no Si–C bond, 40 % had one and 11 % had two. From these integrals the patent calculates a proportion X and relates it to the properties of the film. As X rose from 0 to 80 %, the shrinkage on curing fell from 10.0 to 0.8 % and the water absorption from 6.6 % by weight to zero.[29]
The deposited film is a solid and has to be measured with magic-angle spinning. For organosilicate films of low dielectric constant, 29Si NMR told the films apart more readily than 1H and 13C NMR, X-ray photoelectron spectroscopy or infrared spectroscopy, and it resolved up to twelve bonding environments.[50] In a film deposited from the vapour, it showed that UV curing had increased the Si–O cross-linking by about 11 % at the expense of Si–OH and Si–CH3 groups. The authors add that the measurement needed special sample preparation and extensive data collection.[51]
In a solid, a single-pulse spectrum counts all silicon sites quantitatively, whereas a cross-polarisation spectrum favours rigid sites and silicon close to protons.[6] The aerogel patent measures its unit ratios by the DD/MAS method, that is, without cross-polarisation.[49] Such an experiment takes time; on a silicone gasket it typically ran for about 12 hours.[6]
Silyl groups in organic synthesis
The common silylating agents and their by-products are easy to tell apart, with trimethylsilyl triflate at 44 ppm, chlorotrimethylsilane at 30 ppm, hexamethyldisiloxane at 6.5 ppm and hexamethyldisilazane at 2.2 ppm.[5] Reaction mixtures can be less clear. In a study of the base-catalysed silylation of alcohols, several silicon species had similar shifts and were assigned with the help of calculated values.[12]
In the tert-butyldimethylsilyl (TBS) derivatives of ten amino acids, the shift showed which functional group carried the silyl group, because each kind of group had its own region. The shifts correlated linearly (r = 0.999) with those of the corresponding trimethylsilyl derivatives and were larger by 1.4 to 1.7 ppm for Si–O groups and by 3.8 to 5.8 ppm for Si–N groups.[52] Our catalogue includes two TBS ethers, TBS-isopropylidene-D-ribonolactone and 3,5-bis-O-TBS-2-deoxy-D-ribonolactone.
Practical points for the measurement
- Amount. Published methods use concentrated samples, such as about 40 % by volume in CDCl3[39] or equal volumes of sample and deuterated solvent.[29] A 10 mm tube holds about four and a half times as much liquid as a 5 mm tube, but it needs a 10 mm probe.
- Decoupling. Integrals should come from a spectrum recorded with inverse-gated decoupling. A spectrum with continuous decoupling serves for chemical shifts.
- Reference. IUPAC defines the reference as tetramethylsilane at a volume fraction of 1 % in CDCl3.[1] Because its signal falls where many organosilicon compounds resonate, it is often left out of the sample, and the scale is set with a separate tetramethylsilane sample in the same solvent.[5]
- Alkaline samples. Strongly alkaline silicate solutions attack glass, so they are measured in a PTFE liner.
- Reactive samples. Silicon hydrides are very reactive,[2] and alkoxysilanes are hydrolysed by moisture. Both are best measured in a closed tube, for example a tube with a PTFE valve.
- Reporting. A report of a 29Si spectrum should state the solvent and concentration, the decoupling mode, the pulse angle, the delay between scans, the relaxation reagent and its concentration, and the kind of tube.
Part 6 describes which tube suits which experiment, and Part 4 the conditions for quantitative 1H spectra.
Summary
- Methyl-substituted M, D, T and Q units are found near +8, −20, −60 and −110 ppm, and each new siloxane bond moves a signal 6 to 9 ppm to lower frequency.
- Shifts are not additive. A first chlorine or alkoxy group moves the signal to higher frequency, and with further groups it turns back to lower frequency. Hydrolysis moves Q silicon and trimethylsilyl silicon in opposite directions.
- The glass of the tube and the probe gives a broad signal near −110 ppm, and a quartz tube does not remove it.
- Under proton decoupling the full NOE multiplies the signal by −1.51, and the signal vanishes when two fifths of the relaxation is dipolar.
- T1 ranges from seconds to more than two minutes. With 0.02 to 0.03 mol/L of Cr(acac)3, delays of 6 to 13 s are used.
- For silicon coupled to protons, INEPT and DEPT raise the signal per scan by a calculated factor of 5 to 11. Their integrals have to be calibrated, and they do not see silicon without coupled protons.
Silicon NMR from NorrChemica
Our NMR service measures 1H, 13C, 19F, 31P and 11B in packages per sample at 400 MHz, and we quote long acquisitions and further nuclei per sample. 29Si is not on the published list. If you need a silicon spectrum, please describe the sample and the question on the request form (a unit ratio, a chain length or a degree of condensation, for example), and we will reply with what can be measured and under which conditions. The service is available worldwide for chemical and pharmaceutical R&D. Every report is signed and comes with the native data. Research use only.
- PTFE liner for 5 mm NMR tubesglass-free kit on request
- 10 mm NMR tubesmore sample for an insensitive nucleus
- NMR tubes with PTFE valvehydrosilanes, alkoxysilanes
- TBS-isopropylidene-D-ribonolactoneTBS ether
- 3,5-Bis-O-TBS-2-deoxy-D-ribonolactoneTBS ether
NMR tubes made in Germany by Deutero and shipped from Finland. See the NMR tube guide →
References
- Harris RK, Becker ED, Cabral de Menezes SM, Goodfellow R, Granger P. NMR nomenclature. Nuclear spin properties and conventions for chemical shifts (IUPAC Recommendations 2001). Pure Appl. Chem. 2001, 73, 1795–1818. DOI
- NMR Laboratory, The Hebrew University of Jerusalem. (29Si) Silicon NMR. chem.ch.huji.ac.il/nmr
- Rayner PJ, Richardson PM, Duckett SB. The detection and reactivity of silanols and silanes using hyperpolarized 29Si nuclear magnetic resonance. Angew. Chem. Int. Ed. 2020, 59, 2710–2714 (accepted manuscript). DOI
- Nuclear Overhauser enhancement. Lecture slides, Chemistry 636, University of Wisconsin–Madison, autumn 2014. PDF
- Uhlig F, Marsmann HC. 29Si NMR: some practical aspects. Technical review, Gelest, Inc., pp. 208–222. PDF
- Process NMR Associates. Overview of solid-state 29Si NMR studies of silicone gasket materials. Page
- Zhu L, Cheng X, Su W, Zhao J, Zhou C. Molecular insights into sequence distributions and conformation-dependent properties of high-phenyl polysiloxanes. Polymers 2019, 11, 1989. DOI
- Alam TM, Henry M. Empirical calculations of 29Si NMR chemical shielding tensors: a partial charge model investigation of hydrolysis in organically modified alkoxy silanes. Manuscript for J. Chem. Soc., Faraday Trans., Sandia National Laboratories, 19 July 1999. Text
- Hunsicker et al. Interlinkage of a siliconoid with a silsesquioxane. Z. Anorg. Allg. Chem. 2022 (open access; passage on the NMR data). PDF
- Radeglia R. Eine theoretische Interpretation der gegenseitigen Wechselwirkungsparameter in den 29Si-NMR-Verschiebungen. Z. Naturforsch. B 1977, 32, 1091–1092 (first page). DOI
- Schraml J, Chvalovský V, Mägi M, Lippmaa E. The role of electronic and steric effects in 29Si-NMR spectra of compounds with Si–O–C group. Collect. Czech. Chem. Commun. 1981, 46, 377–390 (abstract). DOI
- Zhang C, Patschinski P, Stephenson DS, Panisch R, Wender JH, Holthausen MC, Zipse H. The calculation of 29Si NMR chemical shifts of tetracoordinated silicon compounds in the gas phase and in solution. Phys. Chem. Chem. Phys. 2014, 16, 16642–16650. DOI
- Silicon-29 chemical shifts for organosilicon reference compounds. J. Magn. Reson. 1973 (abstract). Record
- Sidorkin VF, Pestunovich VA, Voronkov MG. Origin of the shielding effect in the 29Si NMR spectra of silatranes. Magn. Reson. Chem. 1985, 23 (abstract). DOI
- Kalikhman I, Gostevskii B, Pestunovich V, Kocher N, Stalke D, Kost D. Penta- and hexacoordinate silicon mixed dichelates with the SiC2O2N(Cl) ligand environment. ARKIVOC 2006 (v), 63–77. PDF
- Brunet F, Cabane B, Dubois M, Perly B. Sol–gel polymerization studied through 29Si NMR with polarization transfer. J. Phys. Chem. 1991, 95, 945–951. Copy
- Kurjata J, Rozga-Wijas K, Stanczyk W. Investigation of hydrolysis and condensation of methyltriethoxysilane in aqueous systems. Eur. J. Chem. 2013, 4, 343–349. DOI
- Alam TM, Assink RA, Prabakar S, Loy DA. Identification and characterization of the hydrolysis products in TMOS and MTMS monomers using 29Si NMR and polarization transfer techniques. Magn. Reson. Chem. 1996, 34, 603–609 (abstract).
- Mann MD, Chmelka BF. Measurement of dilute 29Si species in solution using a large volume coil and DEFT NMR. Anal. Chem. 2000, 72, 5131–5135. DOI
- Moriones P, Arzamendi G, Cornejo A, Garrido JJ, Echeverria JC. Comprehensive kinetics of hydrolysis of organotriethoxysilanes by 29Si NMR. J. Phys. Chem. A 2019, 123, 10364–10371 (accepted manuscript). Repository
- Semiconductor device manufacturing method. US Patent 11,915,925 (passage on the 29Si NMR measurement). Text
- Polyorganosilsesquioxane, hard coat film, adhesive sheet, and laminate. US Patent 11,111,337 (passage on the 29Si NMR measurement). Text
- Vasil’ev SG, Volkov VI, Tatarinova EA, Muzafarov AM. A solid-state NMR investigation of MQ silicone copolymers. Appl. Magn. Reson. 2013 (abstract). DOI
- Shishido R, Uesugi M, Takahashi R, Mita T, Ishiyama T, Kubota K, Ito H. General synthesis of trialkyl- and dialkylarylsilylboranes: versatile silicon nucleophiles in organic synthesis. J. Am. Chem. Soc. 2020, 142, 14125–14133 (accepted manuscript). DOI
- Christensen B, Price JC. NMR lineshape of 29Si in single-crystal silicon. arXiv:1610.03511, 2016. Preprint
- IMSERC, Northwestern University. 29Si sensitivity test (facility guide). Page
- Paterson AL. DEPTH (protocol). National Magnetic Resonance Facility at Madison, protocols.io, 2025. Page
- Cory DG, Ritchey WM. Suppression of signals from the probe in Bloch decay spectra. J. Magn. Reson. 1988, 80, 128–132 (bibliographic record, as cited in the preceding protocol).
- Nakano T, Shimura M, Ohta T. Coating solution and method for preparing the coating solution, method for forming insulating films for semiconductor devices, and method for evaluating the coating solution. US Patent 5,998,522, Kawasaki Steel Corporation, 1999 (description). Text
- Silicone rubber composition and cured product thereof. US Patent 10,227,453 (passage on the 29Si NMR measurement). Text
- Clustered functional polyorganosiloxanes, processes for forming same and methods for their use. US Patent 9,944,758 (passage on the NMR methods). Text
- Levy GC, Cargioli JD, Juliano PC, Mitchell TD. Silicon-29 spin-lattice relaxation in organosilicon compounds. J. Am. Chem. Soc. 1973, 95, 3445 (abstract).
- Harris RK, Kimber BJ. 29Si and 13C NMR studies of organosilicon chemistry. I. Trimethylsilyl compounds. J. Magn. Reson. 1975, 17, 174–188 (abstract).
- Knight CTG, Harris RK. Silicon-29 nuclear magnetic resonance studies of aqueous silicate solutions. 8. Spin–lattice relaxation times and mechanisms. Magn. Reson. Chem. 1986, 24 (abstract). DOI
- Predictive method for polymers. US Patent 6,687,621 (passage on the NMR methods). Text
- Harris RK, Newman RH. 29Si N.M.R. studies of aqueous silicate solutions. J. Chem. Soc., Faraday Trans. 2 1977, 73, 1204–1215 (abstract). DOI
- Magritek. Determine copolymer composition and average molar mass of polysiloxanes without sample preparation by using 29Si-DEPT spectra. Application blog, 3 April 2025. Page
- Branched polymers from organohydrogensilicon compounds. US Patent 7,449,536 (passage on the 29Si NMR method). Text
- Polyaminosiloxane water tree repellant for electrical insulation. US Patent 12,630,697 (passage on the NMR methods). Text
- Delak KM, Farrar TC, Sahai N. 29Si NMR sensitivity enhancement methods for the quantitative study of organosilicate hydrolysis and condensation. J. Non-Cryst. Solids 2005, 351, 2244–2250 (abstract).
- Rothenberger KS. Inorganic applications of polarization transfer NMR. Literature seminar abstract, University of Illinois, 11 February 1983. PDF
- Helmer BJ, West R. Organometallics 1982, 1, 877–879 (bibliographic record, as reported in the preceding seminar abstract).
- Schraml J. Routine use of INEPT technique for measurements of 29Si NMR spectra of trimethylsilyl derivatives. Collect. Czech. Chem. Commun. 1983, 48, 3402–3406 (abstract). DOI
- Magritek. q-HMBC for sensitivity-enhanced 29Si quantitation. Application note, 2025. PDF
- Ambati J, Rankin SE. Comparison of the polymerization kinetics of bis(triethoxysilyl)ethane and methyltriethoxysilane under acidic conditions. AIChE Annual Meeting 2005, paper 517d (abstract). Page
- Engelhardt G, Jancke H. Structure investigation of organosilicon polymers by silicon-29 NMR. Polym. Bull. 1981, 5, 577–584 (abstract). DOI
- Alam TM. Quantitative analysis of microstructure in polysiloxanes using high resolution 29Si NMR spectroscopy: investigation of lot variability in the LVM97 and HVM97 PDMS/PDPS copolymers. Sandia report SAND2002-3785 (abstract). Record
- Alam TM. NMR spectroscopy characterization of silicone polymers and fluids used in electronic components. Presentation SAND2014-0717C, PANIC, Charlotte, 3–5 February 2014. Slides
- Aerogel. US Patent 10,563,016 (passage on solid-state 29Si NMR). Text
- Mabboux P-Y, Gleason KK. Chemical bonding structure of low dielectric constant Si:O:C:H films characterized by solid-state NMR. J. Electrochem. Soc. 2005, 152, F7–F13 (abstract). DOI
- Abell T, Houthoofd K, Iacopi F, Grobet P, Maex K. Solid state MAS NMR spectroscopic characterization of plasma damage and UV modification of low k dielectric films. MRS Online Proceedings Library 2005 (abstract). Page
- Kubec R, Velíšek J, Kvíčalová M, Čermák J, Schraml J. 29Si and 13C NMR spectra of tert-butyldimethylsilyl derivatives of amino acids. Magn. Reson. Chem. 1995, 33 (abstract). DOI
About this article. We read sources 2, 5, 6, 8, 15, 16, 17, 19, 20, 27 and 41 in full. Our copies of 8 and 16 are scans in which some exponents and table entries are illegible, and we use no number from those places. From source 1 we read the table of spin-½ nuclei and the recommendations on referencing. Sources 3 and 24 were read as accepted manuscripts without their Supporting Information, and source 12 without its structure drawings, so we use only the compounds named in its text. For sources 7, 37 and 44 we read the text; their tables and figures exist only as images, which we could not read, apart from one table of 44. Sources 4, 9, 10, 25, 26 and 48 were read in part: the slides, pages or passages we cite. Of the patents, we read the description of 29; in our copy the formula that defines its proportion X was not legible. Of 21, 22, 30, 31, 35, 38, 39 and 49 we read only the passages on the NMR methods. We cite patents because they state measurement conditions that journal articles seldom give. Sources 11, 13, 14, 18, 23, 32, 33, 34, 36, 40, 43, 45, 46, 47, 50, 51 and 52 were read as abstracts only, and 28 and 42 are bibliographic records. What we say about our own tubes rests on the product pages of our shop. The last two columns of the first table, the frequency at 400 MHz, the step sizes in Figure 3, the degrees of condensation, the width of the glass signal in ppm, the measuring times derived from delays and scan numbers, the reagent concentrations of the kinetic study and of the spin-on-glass patent, the DEPT enhancements, the intensity of the satellites and the molar mass in the silicone example are our own arithmetic on the published values. Figures 5 and 6 are our own calculations and not measured data. Figures 1, 2 and 3 are our drawings of published values. Figures 4 and 7 are schematic, as is the upper part of Figure 1. All sources were read on 13 November 2026.
NMR Series
- Part 1 — Reading 1H and 13C NMR
- Part 2 — Reading 19F, 31P and 11B NMR
- Part 3 — Reading 2D NMR: COSY, HSQC and HMBC
- Part 4 — Quantitative NMR: how a purity figure is made
- Part 5 — 15N NMR: which nitrogen is alkylated, protonated or coordinated?
- Part 6 — NMR tubes by experiment: which tube for which measurement
- Part 7 — Quadrupolar nuclei in NMR: broad lines and missing carbons
- Part 8 — Two sets of NMR signals: rotamers, diastereomers or an impurity
- Part 9 — NMR yield and isomer ratios: relaxation and the delay between pulses
- Part 10 — Silicon NMR: the glass signal, the negative NOE and M, D, T, Q units (this article)
- Part 11 — Solid-state NMR: magic-angle spinning, cross-polarisation and what a CP-MAS spectrum counts
- Part 12 — Enantiomeric excess by NMR: when NMR is the best method, the chiral auxiliaries and the detection limit for the minor enantiomer
- Part 13 — Predicted NMR shifts: how accurate ChemDraw, MestReNova and other predictors are
- Part 14 — Phosphorus NMR in biology: phospholipids, phosphorylated proteins, oligonucleotide drugs and phosphates in food
- Part 15 — Multiplets in ¹H NMR: coupling trees, second-order patterns and two signals from one CH₂
- Part 16 — NMR for cosmetics: adulterated essential oils, free formaldehyde, aloe vera, rancid lipsticks and mineral oil in lip products
- Part 17 — Polymers by NMR: composition, sequence, chain ends and branches, and what a spectrum of many chains can and cannot measure
- Part 18 — Benchtop NMR: what a permanent-magnet spectrometer does well, and when a sample needs high field
- Part 19 — NOESY, ROESY and TOCSY: regioisomers, E and Z isomers, anomers, and the signals of one compound in a mixture
- Part 20 — Rapid-injection NMR: starting a reaction inside the spectrometer, how fast the spectra can follow it, and the intermediates it has revealed
- Part 21 — Fluorine NMR for PFAS: total organofluorine, what targeted LC-MS/MS does not detect, and the detection limit
- Part 22 — Polysiloxanes by proton NMR: silicone oil in medicines, food and cosmetics, the reactive groups before curing, and the rubber network after it
© 2026 NorrChemica. All rights reserved.
Content may not be reproduced without written permission.