selective nmr pulse sequences for the study of solid ... · 1 feature article for macromol. rapid....

12
1 Feature article for Macromol. Rapid. Comm. Selective NMR Pulse Sequences for the Study of Solid Hydrogen-containing Fluoropolymers Shinji ANDO 1 , Robin K. HARRIS 2* , Paul HAZENDONK 3 & Philip WORMARD 4 1 Department of Organic and Polymeric Materials, Tokyo Institute of Technology, Ookayama, Meguro-ku, Tokyo 152-8552, Japan. 2 Department of Chemistry, University of Durham, South Road, Durham, DH1 3LE. 3 Department of Chemistry & Biochemistry, University of Lethbridge, 4401 University Drive W., Lethbridge, Alberta, Canada, T1K 3M4. 4 School of Chemistry, University of St. Andrews, Purdie Building, St. Andrews, Fife KY16 9ST, U.K. *To whom correspondence should be addressed: FAX +44-(0)191-384-4737; email r [email protected] Key W ords: Fluoropolymers; domain structure; magic-angle spinning NMR; relaxation; pulse sequences Abstract Fluorine-19 NMR spectra of solids have some special features, which are discussed in this article. In particular, they generally contain two abundant spin baths (protons and fluorine nuclei). This situation throws up some special operational requirements, as does the study of heterogeneous samples. The relaxation characteristics of heterogeneous systems, which are briefly described herein, frequently permit the use of specific pulse sequences to obtain subspectra for individual components. Various possible selective sequences for use in fluorinated heterogeneous organic solids are listed and their actions rationalised on the basis of molecular mobility. Semi-crystalline hydrogen- containing fluoropolymers form especially suitable systems for such operations, and in order to understand their domain structures it is essential to obtain subspectra of the amorphous and crystalline domains. Examples are given of the use of selective pulse sequences for studying fluoropolymers, especially for PVDF and the copolymer P(VDF 75 /TrFE 25 ). Introduction Synthetic polymers form an extremely important class of materials which have been extensively studied by NMR methods, applied to both solutions and solids. The first experiments combining cross-polarisation (CP), magic-angle spinning (MAS) and high-powered proton decoupling (HPPD) were applied to obtain high-resolution 13 C spectra of three solid polymers. 1 Indeed, in the first decade of the use of the CP/MAS/ HPPD suite of techniques, 2 solid polymers formed a high proportion of the samples studied, and they continue to be highly investigated by MAS NMR. 3 There are a number of reasons for this situation, in particular (i) the ability of NMR to obtain detailed chemical information from amorphous as well as crystalline materials, and (ii) the remarkable versatility of NMR exemplified by the wide range of pulse sequences which can be chosen to produce specific results. 2 Most MAS work on polymers has, naturally, concentrated on the ubiquitous 13 C nucleus, 3,4 but of course a number of different NMR-active nuclides are present in particular polymeric systems, 3,5 for instance 15 N, 29 Si and 31 P. These nuclei are also amenable to the CP/MAS/HPPD combination of techniques. For 1 H high-resolution spectra of solids one must generally use either very fast MAS 6 or else multiple-pulse operation combined with MAS (“CRAMPS”) 7-9 in

Upload: others

Post on 30-Jan-2021

3 views

Category:

Documents


0 download

TRANSCRIPT

  • 1

    Feature article for Macromol. Rapid. Comm.

    Selective NMR Pulse Sequences for the Study of

    Solid Hydrogen-containing Fluoropolymers†

    Shinji ANDO1, Robin K. HARRIS2*, Paul HAZENDONK3 & Philip WORMARD4

    1Department of Organic and Polymeric Materials, Tokyo Institute of Technology, Ookayama, Meguro-ku, Tokyo

    152-8552, Japan.

    2Department of Chemistry, University of Durham, South Road, Durham, DH1 3LE.

    3Department of Chemistry & Biochemistry, University of Lethbridge, 4401 University Drive W., Lethbridge,

    Alberta, Canada, T1K 3M4.

    4School of Chemistry, University of St. Andrews, Purdie Building, St. Andrews, Fife KY16 9ST, U.K.

    *To whom correspondence should be addressed:

    FAX +44-(0)191-384-4737; email [email protected]

    Key Words: Fluoropolymers; domain structure; magic-angle spinning NMR; relaxation; pulse sequences

    Abstract

    Fluorine-19 NMR spectra of solids have some special features, which are discussed in this article.

    In particular, they generally contain two abundant spin baths (protons and fluorine nuclei). This

    situation throws up some special operational requirements, as does the study of heterogeneous

    samples. The relaxation characteristics of heterogeneous systems, which are briefly described herein,

    frequently permit the use of specific pulse sequences to obtain subspectra for individual components.

    Various possible selective sequences for use in fluorinated heterogeneous organic solids are listed

    and their actions rationalised on the basis of molecular mobility. Semi-crystalline hydrogen-

    containing fluoropolymers form especially suitable systems for such operations, and in order to

    understand their domain structures it is essential to obtain subspectra of the amorphous and crystalline

    domains. Examples are given of the use of selective pulse sequences for studying fluoropolymers,

    especially for PVDF and the copolymer P(VDF75/TrFE25).

    Introduction

    Synthetic polymers form an extremely important

    class of materials which have been extensively studied

    by NMR methods, applied to both solutions and solids.

    The first experiments combining cross-polarisation

    (CP), magic-angle spinning (MAS) and high-powered

    proton decoupling (HPPD) were applied to obtain

    high-resolution 13C spectra of three solid polymers.1

    Indeed, in the first decade of the use of the CP/MAS/

    HPPD suite of techniques,2 solid polymers formed a

    high proportion of the samples studied, and they

    continue to be highly investigated by MAS NMR.3

    There are a number of reasons for this situation, in

    particular (i) the ability of NMR to obtain detailed

    chemical information from amorphous as well as

    crystalline materials, and (ii) the remarkable versatility

    of NMR exemplified by the wide range of pulse

    sequences which can be chosen to produce specific

    results.2

    Most MAS work on polymers has, naturally,

    concentrated on the ubiquitous 13C nucleus,3,4 but of

    course a number of different NMR-active nuclides are

    present in particular polymeric systems,3,5 for instance

    15N, 29Si and 31P. These nuclei are also amenable to

    the CP/MAS/HPPD combination of techniques. For

    1H high-resolution spectra of solids one must generally

    use either very fast MAS6 or else multiple-pulse

    operation combined with MAS (“CRAMPS”)7-9 in

  • 2

    order to overcome the strong homonuclear (1H, 1H)

    dipolar interactions.The 19F nuclide is also a special

    case. In several ways it is very suitable for high-

    resolution NMR. Thus, because it exists in 100%

    natural abundance and has a high magnetic moment

    (in fact the third highest of spin-1/2 nuclides after 3H

    and 1H), it has a very high receptivity (0.834 of that of

    1H, 4.90 × 103 times that of 13C).10 Unlike 1H,

    however, it has a large chemical shift range and so the

    spectra can be highly resolved (and therefore

    chemically informative). Arguably, then, 19F NMR is

    to be preferred to eith 1H or 13C NMR for suitable

    cases, though clearly the three nuclides can be studied

    together. One would therefore expect high-resolution

    19F NMR of solids to be very popular. However, the

    same properties that give advantages also confer

    problems. For instance, for perfluorinated systems

    the strong (19F, 19F) dipolar interactions have (at least

    until recently) required the use of CRAMPS,11 which

    is technically demanding. Moreover, for fluorinated

    materials which also contain protons (to which the rest

    of this article is dedicated), high-power proton

    decoupling has been considered necessary (again, at

    least until very recently), and this is not entirely

    straightforward because of the proximity of 1H and

    19F resonance frequencies (differing by only ca. 6%).

    However, in the mid-1990s commercial probes capable

    of 19F-{19H} double resonance, involving high powers

    in the proton channel but with efficient filtering,

    became available, so that work in this area began.12-19

    There is a residual oddity in that, at relatively low

    applied magnetic fields, high-power proton decoupling

    results in the appearance of the Bloch-Siegert

    effect20,21, which causes an apparent chemical shift on

    19F resonances13,18, as shown in figure 1 and expressed

    in equation 1:

    δ(BS) = (γF B1H)2 / (ωF 2 – ωH2) [1]

    where γF is the magnetogyric ratio of 19F, B1H is

    the proton radiofrequency magnetic field strength, ωFis the 19F resonance frequency and ωH is the 1H

    “decoupling” frequency. However, once this is

    recognised, it poses no difficulties for chemical shift

    measurement. All that is required is the use of the

    same 1H RF power during 19F observation of the

    reference sample (e.g. liquid C6F6) as when the sample

    of interest is examined.22 The Bloch-Siegert effect is

    avoided by 19F CRAMPS operation with synchronised

    π pulses on the proton channel providing the

    heteronuclear decoupling.19 Moreover, the effect is not

    significant for spectrometers operating at 7.1 T and

    above. However, for observation of 19F spectra of

    proton-containing fluoropolymers, increasing B0

    provides no advantages in dispersion and requires

    higher spin rates to minimise the occurrence of

    spinning sidebands.

    with decoupling

    without decoupling

    Figure 1. 188 MHz 19F CPMAS spectra withoutand with high-power proton decoupling, showingthe Bloch-Siegert shift (ca 2.8 ppm) example.

    Figure 2. Fluorine-19 MAS spectra16 of a physicalmixture of PTFE (95%) and PVDF (5%). (a) Directpolarisation. (b) 1H→19F cross polarisation. ThePVDF is severely discriminated against in (b).

    SPE

    CP

    PTFE

    PVDF

    PTFE

  • 3

    There are some other residual problems. For

    instance, geminal (19F, 19F) isotropic indirect (i.e.

    scalar) couplings are significant (ca. 280 Hz) for CF2

    groups, e.g. in P(VDF/TrFE)23 and in principle cause

    splittings in spectra. Moreover, partly for this reason

    and partly not fully understood, linewidths in 19F MAS

    spectra of solids remain relatively high (e.g. hundreds

    of Hz) 15even for well-crystallised samples. In

    addition, the large 19F shielding anisotropies can make

    some specialised pulse sequences inefficient or

    complex.19

    With the relatively recent advent of high-speed

    (>20 kHz) MAS, HPPD appears to be no longer

    essential for obtaining high-resolution 19F spectra24-26

    of some fluorinated solids which also contain protons.

    However, such spin rates can cause substantially higher

    increases in sample temperature unless controlled.

    Moreover, there remains a number of advantages of

    using CP from protons, and this is generally more

    efficient at somewhat lower spin rates (e.g. ca.15 kHz),

    which then also requires HPPD during signal

    acquisition. Whereas the gain in sensitivity relative

    to direct polarisation is rather small (γH/γF = 1.062)10,

    CP discr iminates agains t compounds in a

    heterogeneous sample which are perfluorinated. This

    applies also to probe components, which frequently

    involve PTFE and consequently lead to background

    signals (albeit broad) for MAS-only operation for some

    spectrometer/probe combinations (see Figure 2).13,16

    More importantly, CP is involved in a number of

    specialised pulse sequences, as described below,

    including some two-dimensional experiments.

    All these considerations clearly apply to NMR

    studies of hydrogen-containing fluoropolymers12,13.

    Such synthetic macromolecules are important

    industrially because of their excellent stability against

    chemical degradation under a variety of conditions and

    because o f the i r spec ia l p roper t i e s (e .g .

    piezoelectricity, ferroelectricity, pyroelectricity etc.).27

    Moreover they are frequently complex in physical

    structure and are thus of intrinsic interest. They are

    generally semi-crystalline so their domain structures

    require study. Several of them exhibit polymorphism

    of their crystalline domains so that characterization

    methods are necessary and their phase transition

    behaviour becomes important in practical usage. Such

    polymorphism is frequently of the conformational

    type, and variable conformation also characterises

    amorphous domains. As with many polymers,

    especially as produced commercially, the nature of the

    end-groups is relevant, as are any chemical defects in

    the polymer chains. Such defects, arising from

    occasional reversal in the ordering of a monomer unit

    in a chain, are common28-31 in samples of

    poly(vinylidene fluoride), PVDF. Finally, mobility at

    the molecular level, as always for polymers, conveys

    distinctive properties which are temperature-

    dependent. These motions can be analysed by

    measurement of NMR relaxation times.

    Relaxation in Homogeneous and Heterogeneous

    Samples

    One of the most powerful attributes of solid-state

    NMR is its ability to address heterogeneous systems

    and, in particular, to obtain, separately, subspectra

    relating to different components via use of specialised

    pulse sequences. This ability is clearly of value not

    only in cases where heterogeneities correspond to

    different chemical components but also, as in the

    situations considered here, for samples which are

    chemically uniform but physically diverse, i.e. for

    domain structures of pure polymers. This property of

    NMR renders it very unusual, if not unique, among

    characterization methods, especially as it extends to

    amorphous as well as crystalline domains (in contrast

    to all diffraction tools). NMR discrimination methods

    rely on the versatility of NMR as expressed in the

    immense range of pulse sequences which are possible.

    These can be tailored to give specific results. Their

    effective operation in terms of selectivity must rely

    on differences in the properties of the various physical

    domains which are under investigation.

    The property which readily distinguishes solid

    samples even of the same material but in different

    physical form (or of domains of samples of chemically

    homogeneous materials) is molecular-level mobility.

  • 4

    This is, of course, a complex property. Its extent shows

    a strong dependence on the motional frequency

    considered and it varies greatly with temperature as

    well as with molecular environment in the solid state.

    In particular, amorphous domains, which are

    unordered and therefore generally less tightly packed

    than the ordered crystalline domains, are usually the

    more mobile. Their mobility normally increases

    significantly when a sample is heated through the glass

    transition temperature.

    Any molecular motion causes modulation (leading

    to partial averaging) of nuclear dipole-dipole

    interactions and hence affects relaxation times and

    related parameters. Relaxation can also be caused by

    re-orientation of the shielding tensor, the anisotropy

    of which is often substantial for 19F. A number of

    different relaxation times may be measured by NMR

    and influence the operation of NMR experiments in

    different ways. In order to provide a reasonable

    background to the present review article, the relevant

    parameters are described briefly below, but for further

    detail the reader is referred to references 2,32,33 The

    relaxation parameters in question include:

    (i) Spin-lattice (also known as longitudinal) relaxation.

    The characteristic time T1 is generally of the order of

    seconds for abundant spins such as 1H or 19F in solids

    (though it can be significantly longer). It governs the

    return of magnetization in the direction of the applied

    field (B0) to its equilibrium value following a

    perturbation and responds to motions in the region of

    the NMR frequency, i.e. hundreds of MHz. It therefore

    influences the delay between repetition of pulse cycles

    in the NMR experiment.

    (ii) Spin-lattice relaxation in the rotating frame. The

    characteristic time is designated T1ρ and usually lies

    in the tens of ms range for solids. It governs return of

    magnetization perpendicular to B0 but under the

    influence of a radiofrequency field B1, to equilibrium

    following establishment of a significant magnitude.

    Its value is influenced by motions at the nutation

    frequency related to B1 (i.e. to γB1/2π), which is usually

    in the tens of kHz range. It influences the operation

    of the CP experiment.

    (iii) Spin-spin relaxation, perhaps better denoted as

    transverse relaxation. The characteristic time T2 relates

    to the return of magnetization perpendicular to B0 to

    zero (in the absence of B1) following the establishment

    of a non-zero value. Its magnitude is influenced by

    low-frequency motions, and for relatively rigid solids

    (static samples) it is generally a few tens of µs. It

    governs the observed free induction decay and hence

    the linewidth of resonances. It may have a complicated

    dependence on MAS rate.

    (iv) Cross-polarization rate (characteristic time THF).

    In the CP experiment (Figure 3), the contact time, τ,

    is a variable. The magnetization in the observed

    nucleus first rises as a function of τ at a rate governed

    by THF then decays as it leaks to the lattice by T1ρ

    processes. The value of THF is generally in the region

    of tens or hundreds of µs.

    Whereas linewidths tend to decrease and transverse

    relaxation rates tend to increase monotonically as

    increasing temperature promotes more molecular

    mobility, T1 and T1ρ pass through one or more minima.

    Thus the effect of temperature change on T1 and T1ρ is

    sometimes difficult to predict and measure,ments at a

    single temperature can be misleading. Of course, fo

    amorphous polymer domains, passing through the

    Spin Lock(SL)

    Dipolar Decoupling

    Contact(CP)

    (DD)

    F

    H

    π/2

    tCP

    SL

    F

    H

    π/2

    tCP

    DD

    b)

    a)

    CP

    τDelay

    SL

    F

    H

    π/2

    tCP

    DD

    c)

    CP

    τ

    Delay

    Figure 3. (a) Standard CP sequence, (b) delayed-acquisition CP (dipolar dephasing/non-quaternarysuppression) sequence, (c) delayed CP sequence.

  • 5

    glass transition temperature induces substantial

    changes in molecular mobility and hence in relaxation

    times.

    The typical times mentioned above are those

    appropriate for abundant spins such as 1H and 19F. The

    behaviour of spin-lattice relaxation, both in the

    laboratory and the rotating frame, is generally single-

    exponential for a homogeneous sample, but transverse

    relaxation is a more complex phenomenon which may

    be difficult to fit mathematically.34,35 The time T2

    may therefore have a meaning which depends on the

    circumstances. For a heterogeneous sample with large

    domain sizes, spin-lattice relaxation may be treated

    independently for the various domains (i.e. the total

    magnetization will relax as the sum of two

    exponentials, if there are two domains, with

    coefficients appropriate to the domain concentrations),

    as may spin-lattice relaxation in the rotating frame and

    transverse relaxation. However, the phenomenon of

    spin diffusion spreads magnetization in a random-walk

    fashion such that the degree of spread is governed by

    time. Thus, in typical times, the magnetization of

    different domains can be averaged if the domains are

    small enough. In such cases single values of T1 (and

    of T1ρ) will be observed even for heterogeneous

    samples. The critical domain size for these events is

    of the order of nm. 36 Moreover, T1 is more readily

    averaged than T1ρ. Relaxation behaviour in the critical

    region is complex34,35 and, whilst sums of exponentials

    may be used, the various T1 values will be distorted

    from those intrinsic to the various domains and the

    coefficients will not correspond to the concentrations

    of the domains. Transverse relaxation is virtually

    unaffected by spin diffusion. The effects of spin-

    diffusion on T1ρ in heterogeneous systems can be

    minimised37,38 by spin-locking at the magic angle39,40

    Discriminating Experiments in 19F Solid-state NMR

    As stated above, pulse sequences can be devised

    to discriminate between various domains in a

    heterogeneous polymer. These are primarily based

    on differences in mobility which cause changes in the

    various relaxation times via both the dipolar and the

    shielding anisotropy mechanisms. Since the

    relationship between mobility and relaxation is not

    simple, a variety of responses to the NMR experiment

    in question can occur. Therefore, whereas in some

    cases T1 may differ greatly between two domains (and

    an experiment based on T1 differentiation will work

    well), in others it will prove to be better to discriminate

    via T1ρ or T2. Moreover, such a choice will be

    temperature-dependent. In addition, whereas

    sometimes the relaxation of 19F itself may provide a

    good opportunity for selectivity, in other cases it will

    prove to be better to use differentiation based on 1H

    relaxation, via CP to 19F. Thus there are many

    possibilities, and therefore a range of appropriate pulse

    sequences is described briefly here.

    Firstly, the cross-polarisation pulse sequence

    (Figure 3) is itself selective. This is because the 19F

    magnetization variation with contact time (Figure 4)

    has the functional form41 of equation 2:

    Contact time, tCP

    B undetectable

    detectability level

    SBmax

    SAmax

    log

    S

    slope T1111ρρρρ(A)

    slope T1111ρρρρ(B)

    Figure 4. Schematic plot of variable-contact-timeCP intensity, S, for two domains with verydifferent CP rates and effective T1ρ. The

    DD

    CP

    SL DD

    F

    H

    CP

    SL DD

    F

    H

    τ

    π/2

    π

    π/2π/2

    π/2 τ

    a)

    b)

    DDDD

    Figure 5. (a) Pre-CP inversion-recovery sequence,(b) post-CP inversion-recovery sequence.

  • 6

    MF(t) ∝ exp(−τ/T*HF) + exp(−τ/T*1ρ) [2]

    where T*HF and T*

    1ρ are themselves functions of THF

    (the cross-polarization time), T1ρ(H) and T1ρ(F). [The

    cross-polarisation dynamics will be more complex

    when there are resolved 19F resonances for a given

    domain such that there are several distinct fluorine spin

    baths.42 Also, dipolar oscillations are often seen at short

    contact times.43,44] The values of both and will depend

    on local molecular mobility. CP rates depend on the

    strength of (H,F) dipolar interactions which are

    weakened by molecular mobility. Therefore, CP with

    short contact times generally favours crystalline

    domains. However, the same molecular mobility also

    frequently causes T1ρ(H) for amorphous domains to

    be significantly shorter, which means that crystalline

    domains may be selected by long contact times also.

    Figure 4 shows schematically the contact-time

    dependence for a system containing two domains

    (labelled A and B) with equal amounts of the observed

    nuclide but with CP and relaxation characteristics of

    crystalline (A) and amorphous (B) material. It can be

    seen that, in general, CP favours crystalline domains,

    especially at short contact times (when CP to A is

    efficient and the signal is large) and at very long times

    (when domain B may be undetectable because of the

    noise level).

    Discriminating CP experiments, which depend on

    T1(H) or T1(F) involve an inversion-recovery

    component, of 1H magnetisation pre-contact or of 19F

    magnetisation post-contact, respectively (Figure 5). In

    each case, the recovery time can be set so as to null

    the magnetisation of one domain, resulting in only the

    signal of the other domain being observed. The null

    condition for a simple system with relaxation time T1

    requires the recovery time, τ, to be T1ln2 = 0.693 T1.

    For the inversion-recovery experiment on 19F, the

    magnetisation after the CP must first be placed in the

    –z direction by a 90x° pulse and then brought back to

    the y direction (to be measured) after the relevant

    nulling time. Success for this experiment depends on

    T1 for the two domains differing substantially.

    However, there is a complication with these

    experiments since in principle spin-lattice relaxation

    for a coupled heteronuclear system is not single

    exponential unless the two types of spin are decoupled

    during the time allowed for relaxation. Decoupling

    would therefore be desirable for significant periods

    of time, which could cause probe damage. The method

    will often work without decoupling but the appropriate

    conditions are not readily determined. However, spin

    diffusion is relatively efficient at averaging

    longitudinal relaxation rates32, so discrimination based

    on T1 is not often implemented for fluoropolymers.

    Similar, but somewhat simpler pulse sequences are

    available (Figure 6) for selectivity based on T1ρ(H) or

    T1ρ(F). In the former case a 90°(H) pulse is followed

    by a spin-lock period which lies between the values

    of T1ρ(H) for the two domains. A CP contact time

    Spin Lock

    CP

    SL DD

    F

    H

    tSL

    Spin LockCP

    SL DD

    F

    H

    tSL

    π/2

    π/2

    a)

    b)

    Figure 6. (a) Pre-CP spin-lock sequence, (b) post-CP spin-lock sequence.

    CP

    SL DD

    F

    H

    π/2

    DD

    F

    H

    π/2

    CP

    SL DD

    F

    H

    π/2

    x x x x x x x x x x x x

    x x x x x x x x x x x x

    x -y x x -y x -x y -x -x y -x

    a)

    b)

    c)

    Figure 7. (a) Dipolar filter sequence, (b) DIVAM/CP sequence (c) direct DIVAM sequence.

  • 7

    follows immediately the spin-lock period ends, no

    additional 90° pulse being required. For the sequence

    based on T1ρ(F) the spin-lock period (which must be

    between the values of T1ρ(F) for the two domains)

    follows the CP contact time, and acquisition of the

    signal occurs immediately the spin-locking ends. In

    these cases decoupling during the relaxation (spin-

    lock) periods is scarcely feasible since it is essential

    to avoid CP during these times, so the relaxation may,

    in practice, be complex.

    There are several pulse sequences based on

    discrimination via linewidths (i.e. transverse

    relaxation). The simplest is the dipolar dephasing

    sequence (Figure 3b),45 also known as non-quaternary

    suppression46,47 because of its use for selecting 13C

    signals for quaternary carbons while eliminating those

    from CH and CH2 carbons. This sequence involves a

    (non-observed) free induction decay period during a

    decoupling window immediately after CP. Under these

    conditions, 19F signals from crystalline domains

    correspond to broad lines (because the polymer chains

    are rigid, giving rise to strong heteronuclear and

    homonuclear dipolar interactions) and so they decay

    quickly, leaving signals from amorphous domains

    (corresponding to relatively sharp lines) to be detected

    following the end of the decoupling window. In fact a

    post-CP delayed-acquisition sequence with proton

    decoupling during the delay would also give some

    discrimination because of the differences in (F,F)

    homonuclear dipolar interactions. Another alternative,

    which often gives similar results for semi-crystalline

    polymers, is the delayed CP sequence (Figure 3c),

    which relies on differences in proton bandwidths. In

    both the delayed-acquisition and delayed-CP cases

    (especially the former), it may be advantageous to

    introduce a π pulse into the middle of the delay period,

    with rotor synchronisation, to refocus chemical shifts.

    A further selective method based on differentiation

    between strong and weak (i.e. partially averaged)

    homonuclear dipolar interactions is the so-called

    dipolar filter (DF) pulse sequence (Figure 7a). 48 This

    consists of a series of 90° pulses with phase cycling.

    Weak dipolar couplings are refocussed, but strong

    dipolar interactions are affected less efficiently and

    consequently magnetization from rigid regions tends

    to be eliminated by this pulse sequence.

    Recently, a new method for obtaining selective

    spectra based on proton transverse relaxation has been

    reported. This pulse sequence is called Discrimination

    Induced by Variable Amplitude Minipulses (DIVAM)

    (Figure 7b) 44,49,50 and consists of a series of (usually

    12) minipulses of constant phase but with pulse

    (nutation) angles chosen to eliminate the magnetization

    from either rigid or relatively mobile domains. During

    the minipulse intervals My decays relatively slowly

    for the latter, so that the net pulse angle gradually

    Figure 8. Explanation of DIVAM, showing howmagnetisations with short and long transverse re-laxation times behave during the operation of asuccession of minipulses of common phase. Toget the pure subspectra. CP is used following thesituations shown at the bottom. Under the condi-tions illustrated, the “pure mobile spectrum” (seebottom left) will usually contain a small contribu-tion from the rigid domains unless a small oppo-sitely-phased pulse is applied before the CP.

    –60 –80 –100 –120 δδδδF/ppm

    melt-crystallisedfrom the film

    (αααα−−−−form)

    biaxially-drawn9 µµµµm film

    (mostly ββββ−−−−form)residualαααα−−−−form

    Figure 9. Fluorine-19 spectra (centreband only)of two samples of PVDF, showing discriminationusing a T1ρ(H) filter. The pre-CP proton spin-lockduration was 40 ms and a short contact time (50ms) was used. The different polymorphic contentof the two samples is clearly shown.13

  • 8

    increases over the 12 pulses whereas My for rigid

    regions is essentially lost between the minipulses (but

    a substantial Mz is retained if the minipulse angle is

    small). When the minipulse angle and interval for 1H

    are set so as to give a net nutation of 90° for the mobile

    domain magnetization, a final 90° pulse followed by

    a CP contact time will give the 19F spectrum selectively

    for the rigid domain. Alternatively, a CP contact time

    immediately following the 12 minipulses (but without

    the extra 90° pulse) will give the 19F spectrum mainly

    for the mobile domain. This may also be obtained by

    use of larger minipulse angles leading to, say, a net

    nutation angle of 360° for M (mobile), by which time

    M (rigid) may have been lost (see Figure 8), so that

    CP (90° pulse plus contact time) will yield a spectrum

    of the mobile region only. It will be seen that this

    ability to obtain spectra of both rigid and mobile

    domains selectively is an advantage of this method.

    The DF method can only give selective spectra of

    mobile regions, as does dipolar dephasing, and T1ρ

    filter methods only yield spectra of the region with

    the longer T1ρ. It is true that subtraction of a single

    selected spectrum from the full spectrum can yield the

    complementary spectrum, but this is often

    unsatisfactory in practice.

    Several of the above discrimination methods can,

    alternatively, proceed using 19F direct polarisation (DP)

    instead of 1H→19F cross polarisation. Thus DP can

    involve inversion-recovery (use of T1(F)), spin-locking

    (use of T1ρ(F)), dipolar dephasing (which simply

    becomes delayed acquisition) and dipolar filter

    components in the pulse sequence. Recently, direct

    DIVAM (Figure 7c) has also been utilised.51 However,

    the direct DIVAM pulse sequence appears to be more

    complicated in operation.than CP/DIVAM. The effects

    depend on offset, and shielding anisotropy (both of

    which are much larger than the corresponding

    parameters for proton NMR) but not significantly on

    dipolar interactions. Gerstein et al. have shown52 that

    significantly-delayed acquisition proton spectra can

    reveal the existence of mobile moieties in a solid

    system even in very low concentration, though the

    cause of this effect is subject to controversy.53,54 DP

    methods are, naturally, available for perfluorinated

    systems also. For fluoropolymers containing protons,

    direct detection of 1H spectra is also possible (as well

    as 19F→1H CP) 19F spectra contain the big advantage

    of better resolution. Methods involving 1H→19F CP

    are often preferred because of the elimination of

    background signals from fully protonated or fully

    fluorinated components of the probe (and rotor caps).

    Of course, any of the selective pulse sequences

    described above may be used as a preliminary to

    50 0 –50 –100 –150 –200 –250 δδδδ////ppm

    STANDARD

    SPECTRUM

    SPECTRUM

    from POST-CP

    T1_(F) FILTER

    contact time 3 ms

    reverse units

    in crystalline ( _ rigid)

    domains?

    spinning sidebands spinning sidebands

    -130-120-110-100-90-80-70-60

    δF / ppm

    G

    F

    E

    D

    C AB

    -130-120-110-100-90-80-70-60

    δF / ppm

    G

    F

    E

    D

    C

    AB

    powder(αααα-form)

    annealed film

    (γγγγ-form)

    a) Direct Polarisation b) After spin-lock for 20 ms

    Figure 10. Fluorine-19 CPMAS spectra of asample of PVDF using the standard sequence (top)and a T1ρ(F) filter (bottom). In this case the fullspectrum, including the spinning sidebands, isshown.

    normal spectrum

    dephased spectrum

    _-phase

    amorphous

    _-phase

    reverse units

    – 60 –120 _F / ppm

    Figure 11. Fluorine-19 CPMAS spectra of PVDFshowing selection of the amorphous subspectrumby dipolar dephasing.

  • 9

    -250-200-150-100-500

    δF / ppm

    1

    3

    4

    56

    78

    a) Direct polarisation

    b) Delayed CP

    (delay time = 0.5 ms

    scans = 128)

    c) Short CP (0.2 ms)

    further manipulation of the magnetization of the

    selected domain. For instance, they can be used in a

    Goldman-Shen55 experiment to utilise spin diffusion

    in order to obtain information on the size of domains.56

    Combination with a wideline separation (WISE) pulse

    sequence57 enables heteronuclear correlation (two-

    dimensional) experiments to be carried out selectively

    for crystalline and amorphous domains17.

    All the pulse sequences described above involve

    observation of 19F subspectra. It ispossible to use

    analogous methods to obtain proton subspectra for

    amorphous and crystalline domains44 (assuming the

    existence of two abundant spin baths, 19F and 1H), but

    such procedures are less attractive because fluorine

    spectra have far superior chemical shift dispersion than

    proton spectra.

    Applications to Fluoropolymers

    As far as we are aware, domain-selective

    measurements based on T1 differences have not been

    implemented on fluoropolymers, largely because

    typical domain sizes imply substantial (usually

    complete) averaging caused by spin diffusion.

    However, circumstances may occur for which such

    methods are appropriate.

    The first-reported domain-selective 19F-{1H}

    experiments on a fluoropolymer (PVDF) involved

    using a T1ρ(H) filter (i.e. a spin-locked delayed-contact

    CPMAS pulse sequence) as an introduction to a WISE

    sequence. 17 Such a filter is effective (Figure 9) at

    ambient probe temperature because T1ρ(H) for the

    crystalline domains is significantly longer than for the

    amorphous domains. Figure 9 shows how this method

    can be used12 to study the polymorphic form of the

    crystalline domains in PVDF samples produced by

    – 75 – 100 – 125 δδδδF/ ppm

    POWDER(αααα−−−−PHASE)

    40

    10

    0.01

    ττττ /ms

    Figure 12. Domain selec-tion of PVDF in 19F direct-polarisation spectra60. Top:full spectrum. Middle: dipo-lar-filtered spectrum. Bot-tom: T1ρ(F)-filtered spec-trum. In this case, T1ρ(F) forthe principal amorphouspeak is 3.5 ms, whereas forthe high-frequency crystal-line peak it is measured tobe 9.5 ms – a sufficient dif-ference to give good selec-tivity, as shown in the bot-tom spectrum.

    Figure 13. CPMAS spectra of PVDF, obtained us-ing a Goldman-Shen pulse sequence following se-lection of the crystalline subspectrum56. The de-lay time allowed for spin diffusion is given at theright-hand side. The signal for the amorphousphase can be seen to grow in as the delay timeincreases, analysis of which allows domain sizesto be determined.

    Figure 14. 19F-{1H}M A S s p e c t r a o fP(VDF75/TrFE25) at68ºC, obtained by (a)direct polarisation, (b)delayed CP (delay time0.5 ms), and (c) short-contact CP (contacttime 0.1 ms).

    Figure 15. DIVAM/CP spectra of the VDF/TrFEcopolymer as a function of the minipulse angleused. Top: Unfiltered spectrum. Middle: Selectionof the amorphous domain. Bottom: Selection ofthe crystalline domain. The inter-pulse spacing was6 ms, and the minipulse nutations angles used areindicated.

    θθθθ= 0º

    θθθθ= 30º

    θθθθ= 7.3º

  • 10

    different processing methods. It should, however, be

    noted that the precise values of the relevant relaxation

    parameters are difficult to obtain accurately because

    of the effects of spin diffusion and of cross-relaxation

    between proton and fluorine spin baths. Moreover, it

    needs to be recognised that CP dynamics are

    complicated when two abundant spin nuclides are

    involved,41,58 rendering the variable contact time

    method for obtaining T1ρ difficult to apply. In any

    case, all relaxation parameters will depend on the

    nature of the sample (average molecular mass,

    dispersion, regio-irregularity etc.) and on temperature.

    However, the ratio of T1ρ(H) values for the crystalline

    and amorphous domains for PVDF is found to be

    generally in the range 2-3,44,59,60 which suffices to

    make the pre-CP proton spin-lock an efficient tool to

    select the spectrum of the crystalline domains. A

    similar situation applies to T1ρ(F), so that a post-CP

    19F spin-lock is equally effective in selecting such

    domains. Figure 10 60 illustrates this fact and also

    shows that the selectivity extends to the spinning

    sidebands The result suggests that some reverse units

    in PVDF occur in rigid regions (possibly at the

    interface of amorphous and crystalline domains). This

    has been confirmed by the use of a post-CP 19F spin-

    lock as a preparation phase to a RFDR experiment. 60

    Experiments conducted at ca. 100°C showed that

    T1ρ(F) values for both the crystalline and amorphous

    domains of PVDF were significantly lower than those

    at ca. 60°C but that the ratio remained approximately

    the same, so that the T1ρ(F) filter selection method was

    still viable at the higher temperature.60 The 19F DP/

    Figure 16. Direct DIVAM measurements onPVDF. Left: normalised experimental results.Right: simulations. The value of the inter-pulsespacing was 6 ms.

    spin-lock method has been shown61,62 to be effective

    for selecting the subspectrum of the γ crystalline form

    of PVDF.

    However, T1ρ filters only select for crystalline

    domains, so they must be matched with other methods.

    The dipolar dephasing pulse sequence was the first

    one used to select the spectrum of amorphous domains

    in PVDF (Figure 11),59 since it was found that the

    decay time of the magnetization for these regions

    during the decoupling window was about three times

    that for the crystalline domains. Figure 11,59 illustrates

    this situation; the lower spectrum should be compared

    with the upper one of Figure 9 for the crystalline

    domains. Later,56 the dipolar filter (DF) sequence was

    also used for this purpose (Figure 12, middle spectrum)

    and proved to be very effective. The T1ρ(H) and DF

    methods have been inserted before a Goldman-Shen

    pulse sequence in order to determine the lamellar sizes

    of the two domains, as is shown in Figure 13.56

    Delayed-CP and short-contact CP can be readily

    used to select for amorphous and crystalline regions

    respectively, as is illustrated for a sample of P(VDF75/

    TrFE25) in Figure 14.This clearly shows that the sample

    (as received from a commercial source) contains both

    mobile (amorphous) and rigid (crystalline) domains.

    The spectrum of the latter is clearly depicted by the

    short-contact CP experiment, whereas it is largely

    obscured in the DP expeiment because the resonances

    are broad and the crystallinity of the sample is

    relatively low.

    The DIVAM/CP method was first used49 to select

    for amorphous domains in the case of PVF and was

    DELAY

    ssb

    CF2CHF

    Figurs 17. Rotor-synchronised delayed-acquisitionspectra of P(VDF75/TrFE25)

    Direct DIVAM for PVDF: Simulated data τ=6µs

    Pulse Angle

    0 20 40 60 80

    No

    rma

    lize

    d In

    ten

    sity

    -1.0

    -0.8

    -0.6

    -0.4

    -0.2

    0.0

    0.2

    0.4

    0.6

    0.8

    1.0

    1.2

    Pulse Angle vs Amorpho

    Pulse Angle vs Crystaline

    Pulse Angle vs Defect

    Direct DIVAM for PVDF: Experimental data τ=6µs

    Pulse Angle

    0 20 40 60 80

    Nli

    dI

    tit

    -1.0

    -0.5

    0.0

    0.5

    1.0Pulse Angle vs Normalized Amorphous

    Pulse Angle vs Normalized Crystaline2

    Pulse Angle vs Normalized defect2

    (a)(b)

  • 11

    later applied44 to PVDF. Its action was explained in

    terms of a simple dephasing process later50 and it was

    shown to be also effective in selecting for crystalline

    domains in a VDF/TrFE copolymer.50 Figure 15

    illustrates its use for selecting both amorphous and

    crystalline domains merely by varying the minipulse

    nutation angle. Domain selectivity for PVDF by the

    direct DIVAM pulse sequence has been successfully

    simulated and appears to be largely dependent on the

    difference in shielding anisotropy between the fluorine

    nuclei in the crystalline and amorphous regions51

    (Figure 16) On the other hand, the signal for the

    reverse units seems to be governed primarily by the

    offset term. The effect of relaxation could not be

    simulated together with the spin dynamics, so one

    cannot rule out a role for transverse relaxation in the

    selection process. Neither the T1ρ(H) filter nor the

    dipolar dephasing method appeared to give43 any

    significant selection for p(TrFE).

    A combination of a T1ρ(F) filter with a short spin-

    diffusion time gave60 a DP spectrum of PVDF at 100°C

    which revealed the existence of a signal in the “reverse

    unit” region which arose from a highly mobile group.

    This signal was dramatically selected60 in a delayed-

    acquisition spectrum obtained at 60°C. It has been

    attributed to –CF2H end groups occurring in only ca.

    0.013% concentration, as attested by solution-state

    spectra of a telomer.63 The delayed acquisition

    experiment has similarly revealed the existence of very

    mobile groups in a VDF/TrFE copolymer (Figure

    17).64

    Conclusion

    It is shown herein that there are many ways to

    discriminate effectively between 19F spectra of

    crystalline and amorphous domains for semi-

    crystalline fluoropolymers. The pulse sequences

    involved all rely on differences in magnetisation

    relaxation between the domains. Since various

    relaxation properties (linewidth, T1, T1ρ) may be

    involved, and either 1H or 19F relaxation can be used,

    the optimum experiment must be carefully chosen in

    each case.

    AcknowledgementsWe thank Dr. Paolo Avalle for figure 16, Dr. David

    Apperley for much assistance with obtaining spectra,Dr. Keitaro Aimi for work on g-PVDF and the copoly-mer, and N. Andres and T. Montina for their assistancein simulating the direct DIVAM results.

    References1. Schaefer J, Stejskal EO 1976. J Am Chem Soc

    98:1031.

    2. Duer MJ. 2004. Introduction to Solid-State NMR

    Spec t ro scopy. ed . , Oxfo rd : B lackwe l l

    Publishing.1998

    3. In Ando I, Asakura T, editors. Solid State NMR of

    polymers, ed., New York: Elsevier.

    4. McBrierty VJ, Packer KJ. 1993. Nuclear Magnetic

    Resonance in Solid Polymers. ed.: Cambridge

    University Press. (1991)

    5. In Mathias LJ, editor Solid State NMR of Polymers,

    ed., New York: Plenum Press.

    6. Hafner S, Spiess HW 1998. Concepts Magn Reson

    10:99.

    7. Ryan LM, Taylor RE, Paff AJ, Gerstein BC 1980.

    J Chem Phys 72:508.

    8. Gerstein BC, Pembleton RG, Wilson RC, Ryan LM

    1977. J Chem Phys 66:361.

    9. Harris RK, Jackson P 1990. NATO ASI C 322:355.

    10.Harris RK, Becker ED, Cabral de Menezes SM,

    Goodfellow R, Granger P 2001. Pure Appl Chem

    73:1795.

    11. Harris RK, Jackson P 1991. Chem Rev 91:1427.

    12.Har r i s RKM, G.A. ; Hols te in , P. 1998 .

    Fluoropolymers. In Ando I, ; Asakura, T.;, editor

    Solid state NMR of polymers, ed., New York:

    Elsevier. p 667-712.

    13.Harris RK, Monti GA, Holstein P. 1998. Fluorine-

    19 NMR. In Ando I, ; Asakura, T.;, editor Solid

    state NMR of polymers, ed., New York: Elsevier.

    p 253-266.

    14.Ando S, Harris RK, Scheler U. 2002. In Grant DM,

    Harris RK, editors. Encyclopedia of NMR, ed.:

    Wiley. p 531-550.

    15.Carss SA, Harris RK, Holstein P, Say BJ, Fletton

    RA 1994. J Chem Soc, Chem Comm:2407.

    16.Harris RK, Carss SA, Chambers RD, Holstein P,

    Minoja AP, Scheler U 1995. Bull Magn Reson

    17:37.

    17.Scheler U, Harris RK 1996. Solid State NMR 7:11.

    18.Vierkotter S 1996. J Magn Reson A 118:84.

    19.Scheler U, Harris RK 1996. Chem Phys Lett

    262:137.

    20.Bloch F, Siegert A 1940. Phys Rev 57:552.

  • 12

    21.Freeman R. 1988. A Handbook of Nuclear

    Magnetic Resonance. ed., Harlow: Longman

    Group.

    22.Holstein P, Scheler U, Harris RK 1998. Polymer

    20:4937.

    23.Mabboux PY, Gleason KK 2002. J Fluorine Chem

    113:27.

    24.Scheler U 1999. Bull Magn Reson 19:52.

    25.Scheler U 1998. Solid State NMR 12:9.

    26.Sue T-W, Tzou D-LM 2000. Polymer 41:7289.

    27.1997. In Scheirs J, editor Modern Fluoropolymers,

    ed., Chichester: John Wiley and Sons.

    28.Ferguson RC, Brame EG 1979. J Phys Chem

    83:1397.

    29.Tonelli AE, Schilling FC, Cais RE 1982. Macromol

    15:849.

    30.Cais RE, Kometani JM 1985. Macromol 18:1357.

    31.Lovinger AJ, Davis DD, Cais RR, Kometani JM

    1987. Polymer 28:617.

    32.Kenwright AM, Say, B.J. 1993. Solid-state Proton

    NMR of Polymers. In Ibbett RN, editor NMR

    Spectroscopy of Polymers, ed., Glasgow: Blackie

    Academic & Professional. p 231-274.

    33.Harris RK. 1996. NMR Studies of Solid Polymers.

    In Fawcett AH, editor Polymer Spectroscopy, ed.,

    Chichester: John Wiley & Sons. p 117-133.

    34.Kenwright AM, Packer KJ, Say BJ 1986. J Magn

    Reson 69:426.

    35.Geppi M, Harris RK, Kenwright AM, Say BJ 1998.

    Solid State NMR 12:15.

    36.S c h m i d t - R o h r K , S p i e s s H W. 1 9 9 4 .

    Multidimensional Solid State NMR and Polymers.

    ed., London: Academic Press.

    37.Jones GP 1966. Phys Rev 148:332-335.

    38.Tse D, Hartmann SR 1968. Phys Rev Lett 21:511-

    514.

    39.Goldburg WI, Lee M 1963. Phys Rev Lett 11:255-

    258.

    40.Lee M, Goldburg WI 1965. Phys Rev A 140:1261-

    1271.

    41.Ando S, Harris RK, Reinsberg SA 1999. J Magn

    Reson 141:91.

    42.Hazendonk P, Harris RK, Galli G, Pizzanelli S

    2002. Phys Chem Chem Phys 4:507-513.

    43.Reinsberg SA, Ando S, Harris RK 2000. Polymer

    41:3729.

    44.Ando S, Harris RK, Reinsberg SA 2002. Magn

    Reson Chem 40:97.

    45.Opella SJ, Frey MH 1979. J Am Chem Soc

    101:5854.

    46.Fyfe CA, Rudin A, Tchir WJ 1980. Macromol

    12:1322.

    47.Balimann GE, Groombridge CJ, Harris RK, Packer

    KJ, Say BJ, Tanner SF 1981. Phil Trans R Soc Lond

    A299:643.

    48.Egger N, Schmidt-Rohr K, Blumich B, Domke

    WD, Stapp B 1992. J Appl Polym Sci 44:289.

    49.Ando S, Harris RK, Holstein P, Reinsberg SA,

    Yamauchi K 2001. Polymer 42:8137.

    50.Hazendonk P, Harris RK, Ando S, Avalle P 2003. J

    Magn Reson 162:206.

    51.Hazendonk P, Andres N, Nieboer J, Montina T in

    preparation.

    52.Gerstein BC, Hu JZ, Zhou J, Ye C, Solum M,

    Pugmire R, Grant DM 1996. Solid State NMR 6:63.

    53.Fung BM, Tong T-H, Dollase T, Magnuson ML

    1996. J Magn Reson A 123:56-63.

    54.Ding S, McDowell CA 1995. J Magn Reson A

    117:171-178.

    55.Goldman M, Shen L 1966. Phys Rev 144:322.

    56.Holstein P, Monti GA, Harris RK 1999. Phys Chem

    Chem Phys 1:3549.

    57.Schmidt-Rohr K, Clauss J, Spiess HW 1992.

    Macromol 25:3273.

    58.Ando S, Harris RK, Monti GA, Reinsberg SA 1999.

    Magn Reson Chem 37:709.

    59.Holstein P, Harris RK, Say BJ 1997. Solid State

    NMR 8:201.

    60.Wormald P, Apperley DC, Beaume F, Harris RK

    2003. Polymer 44:643.

    61.Park J-W, Seo Y-A, Kim I, Aimi K, Ando S, Ha C-

    S 2004. Macromol 37:429.

    62. Koseki Y, Aimi K, Ando S unpublished work.

    63. Wormald P. 2004. ed.: University of Durham.

    64. Ando S, Avalle P, Harris RK unpublished work.