Tuning the spin Hamiltonian of NENP by external pressure: a neutron scattering study
Extracted source text ยท arXiv cond-mat/9709150
\documentstyle[preprint,aps]{revtex}
\tightenlines
\begin{document}
\draft
\title{Tuning the spin Hamiltonian of NENP by external pressure:
a neutron scattering study}
\author{ I.~A.~Zaliznyak$^{1,2}$\cite{Kapitza}, D. C. Dender$^1$\cite{NIST},
C. Broholm$^{1,2}$, and Daniel H. Reich$^1$ }
\address{
$^1$Department of Physics and Astronomy, The Johns Hopkins
University, Baltimore, Maryland 21218\\
$^2$National Institute of Standards and Technology, Gaithersburg, Maryland
20899}
\date{\today}
\maketitle
\widetext
\begin{abstract}
We report an inelastic neutron scattering study of antiferromagnetic spin
dynamics in the Haldane chain compound Ni(C$_2$H$_8$N$_2$)$_2$NO$_2$ClO$_4$
(NENP) under external hydrostatic pressure $P = $ 2.5~GPa. At ambient
pressure, the magnetic excitations in NENP are dominated by a long-lived triplet
mode with a gap which is split by orthorhombic crystalline
anisotropy into a lower doublet centered at $\Delta_\perp\approx$
1.2meV and a singlet at $\Delta_\|\approx$ 2.5meV. With pressure
we observe appreciable shifts in these levels, which move to
$\Delta_\perp{\rm( 2.5GPa)}\approx$ 1.45~meV and $\Delta_\|{\rm( 2.5GPa)}\approx$
2.2meV. The dispersion of these modes in the crystalline $\hat{c}$-direction
perpendicular to the chain was measured here for the first time, and can be
accounted for by an interchain exchange $J'_c \approx 3\times 10^{-4} J$
which changes only slightly with pressure. Since the average gap value
$\Delta_H\approx$ 1.64~meV remains almost unchanged with $P$, we conclude
that in NENP the application of external pressure does not affect the
intrachain coupling $J$ appreciably, but does produce a significant decrease
of the single-ion anisotropy constant from $D/J = 0.16(2)$ at ambient pressure
to $D/J = 0.09(7)$ at $P = 2.5 $~GPa.
\vspace{0.5in}
\pacs{PACS numbers: 75.10Jm, 75.40Gb, 75.50Ee.}
\end{abstract}
%\newpage
\section{INTRODUCTION}
The Haldane ground state \cite{Haldane} with short-range correlations and
a gap to magnetic excitations has now been found in several
quasi-one-dimensional S=1 antiferromagnets
\cite{NENP1,NENP2,YBaNO,AgVP,TMNIN}.
The magnetic properties of these materials are well-described by
the spin Hamiltonian
\begin{equation}
\label{H1}
{\cal H} = \sum_{i} \left\{ J {\bf S}_{i} \cdot {\bf S}_{i+1}
+ J' \sum_{\delta_\perp} {\bf S}_{i} \cdot
{\bf S}_{i+\delta_{\perp}} + D(S_{i}^{z})^2 \right\}\;\; ,
\end{equation}
where in addition to the intrachain and interchain couplings $J$ and $J'$
we have included a uniaxial single-site anisotropy, a common feature of these
systems. Because of its finite spin-spin correlation length, the Haldane state
is stable with respect to small perturbations by the interchain exchange and
single-ion anisotropy, and hence can be found not only in the pure 1D
Heisenberg limit of Eq.~(\ref{H1}), $J' = D = 0$, but for a finite range of
$J'$ and $D$ around this point. If, however, the interchain coupling $J'$
exceeds some critical value $J'_{crit.}$, quantum disorder is no longer
favored and the system enters a long-range ordered N\'{e}el ground state
with soft Goldstone excitations. Thus, a zero-temperature phase transition
is expected to occur in a quasi-one-dimensional S=1 antiferromagnet
at $J' = J'_{crit.}$ if the interchain coupling could be varied. Such quantum
phase transitions have recently attracted much theoretical attention
\cite{Affleck,Azzouz,Senechal,Botet,SchulzZiman,Sakai,Meshkov,Jolicoeur}.
Various estimates using different model approximations
\cite{Affleck,Azzouz,Senechal} as well as numerical studies on finite
systems \cite{Sakai} have been performed, giving values
for the critical interchain coupling in the broad range $1.3\times10^{-3}\le
J'_{crit}/J \le 2.6\times10^{-2}$. On the other hand, solid upper and lower
limits on $J'_{crit} /J$ are available from experiments. It is found that
the hexagonal antiferromagnet CsNiCl$_3$ with $J\approx$ 2.8~meV and
$J' /J \approx 2 \times 10 ^{-2}$ undergoes 3D ordering at $T_{\rm N}\approx
4.8$~K \cite{Yelon-Cox}, while orthorhombic Ni(C$_2$H$_8$N$_2$)$_2$NO$_2$ClO$_4$
(NENP) with $J \approx 4$~meV and $J' / J \approx 8 \times 10 ^{-4}$
demonstrates all features of the Haldane system and remains disordered down to
millikelvin temperatures \cite{Meisel}.
The effects of single-ion anisotropy are less drastic than those
of interchain coupling, but are also eventually critical.
As the presence of the single-ion term in Eq.~(\ref{H1}) does not change
the basic 1D nature of the model, its effects may be readily explored by
numerical simulations on finite systems of realistic size, and have been
addressed in a number of studies \cite{Botet,SchulzZiman,Sakai,Meshkov,Jolicoeur}.
These predict that strong Ising-like anisotropy ($D < 0$) favors N\'{e}el
ordering for $D < -0.25 J$, while ``easy-plane'' anisotropy ($D>0$) eventually
drives the system into a so-called ``planar'' phase at $D \approx J$ where all
spins have $S^z =0$ in the ground state.
The currently available experimental data on the effect of nonzero $J'$ and
$D$ on the Haldane spin chain have been obtained from the comparison of
results from different S=1 quasi-1D antiferromagnetic compounds. This restricts
the experimentally accessible region of the ($J',D$) phase diagram discussed
above to just a few points. However, the exchange and anisotropy parameters
of the spin Hamiltonian can be controlled not only by changing the chemical
composition of the magnetic compound, but also by tuning the properties of a
single material. The application of hydrostatic pressure can change the
interatomic separations and local atomic environments to which the spin
Hamiltonian is highly sensitive, and therefore provides an opportunity
for the controlled and continuous tuning of the parameters in the Hamiltonian,
albeit in a restricted range. Here we report a study of the pressure dependence
of the magnetic properties of NENP. We have used inelastic neutron scattering to
measure the changes in the Haldane gap modes at $T\approx$1.8 K when an
external pressure $P \approx$ 2.5 GPa is applied. We find that both the
intrachain and interchain coupling are only marginally affected, but
the single-ion anistropy is reduced substantially, and thus hydrostatic
pressure drives this material closer to the ideal Heisenberg model and further
away from any potential quantum phase transitions.
\section{PROPERTIES OF NENP AT AMBIENT PRESSURE}
NENP is orthorhombic with space group $Pnma$. It has Ni$^{2+}$ chains stretching
along the {\bf b} direction with two chains per unit cell staggered in the {\bf a}
direction \cite{Meyer}.
The ambient-pressure ($P \approx 0.1$~MPa) lattice parameters measured
in our experiment at $T=1.8$ K after the sample had been pressurized are
$a=15.28$ \AA, $b=10.23$ \AA, and $c=8.096$ \AA. In describing neutron
scattering experiments, we refer to wavevector transfer in the corresponding
reciprocal lattice ${\bf Q} = h{\bf a^{*}} + k{\bf b^{*}} + l{\bf c^{*}}
\equiv (h, k, l)$. Since the Ni$^{2+}$ ions are displaced by ${\bf b}/2$ along
the chain, it is convenient to refer to the component of wavevector
transfer along the chain as $\tilde{q} = {\bf Q} \cdot ({\bf b}/2) = k \pi$
when discussing the 1D behavior of NENP. At ambient pressure the magnetic
properties of NENP are well described by the Hamiltonian of Eq.~(\ref{H1})
with $J=4.0(2)$ meV, $D/J=0.16(2)$, and an interchain coupling
in the $\hat{a}$-direction $J'_a/J=8 \times 10 ^{-4}$. The symmetry axis
of the single-site anisotropy ($z$--axis) is parallel to the chain axis.
The spin dynamics of NENP under these conditions have been carefully studied
in a number of neutron-scattering experiments \cite{NENP1,NENP2}. It was shown
that the principal contribution to the spectral density of spin fluctuations
${\cal S}({\bf Q},\omega)$ comes from a triplet of long-lived excitations
above a gap that follow a dispersion relation with fundamental periodicity
in $\tilde{q}$ of 2$\pi$. Most of the spectral weight is concentrated in the
close vicinity of the antiferromagnetic point $|\tilde{q}-\pi |< 2\pi/\xi
\sim 0.3\pi$ (Here $\xi\sim $6--8 spins is the correlation length in
the Haldane state). Over the full range $0.3\pi\leq\tilde{q} \leq\pi$
where they have been observed \cite{NENP2}, these excitations are well-described
by the single mode approximation (SMA):
\begin{equation}
\label{SMA}
{\cal S}^{\alpha\alpha}(\tilde{q},\omega) = -\frac{2}{3}\left( \frac{\langle{\cal H}\rangle
}{L}\right) \frac{1-\cos{\tilde{q}}}{\hbar\omega_\alpha(\tilde{q})}
\delta\left[\hbar\omega - \hbar\omega_\alpha (\tilde{q})\right]\;\; ,
\end{equation}
where $\langle{\cal H}\rangle / L$ is the ground state energy per spin.
The simplest dispersion relation that has the correct periodicity and adequately
fits all the experimental data was found to be \cite{NENP2}
\begin{equation}
\label{hw1}
\hbar\omega_\alpha (\tilde{q}) = \sqrt{\Delta_\alpha^2 + v^2\sin^2
\tilde{q}+A\sin^2\frac{\tilde{q}}{2}}\;\;
\end{equation}
with $v = 9.7$~meV, and $A = 34$~meV$^2$.
Instead of a degenerate triplet excitation with a single gap
$\Delta\approx 0.41J\approx 1.64$~meV at $\tilde{q}=\pi$ as expected in the
isotropic case \cite{Meshkov,Jolicoeur}, the planar anisotropy in Eq.~(\ref{H1})
splits the triplet into two branches with $\Delta_\|\approx $ 2.5~meV and
$\Delta_\perp\approx $1.2~meV for fluctuations polarized parallel and
perpendicular to the chain axis, respectively. The lower mode is further split
by a small orthorhombic anisotropy $\tilde{E}[(S_i^x)^2-(S_i^y)^2]$ with
$\tilde{E} \approx 0.01 J~=~0.04$~meV giving $\Delta_x\approx$ 1.34~meV and
$\Delta_y\approx$ 1.16 meV \cite{NENP1}. This latter splitting is smaller than
the instrumental resolution of our current experiment, and its effects will
therefore be neglected in our subsequent discussion. The lower doublet also
shows measurable dispersion along the {\bf a}$^*$ direction, with an effective
bandwidth parameter $\Delta E\approx$ 0.65 meV \cite{NENP1}. This is the
basis of the above estimate of the interchain exchange $J'_a$. The dispersion
along {\bf c}$^*$ has not been measured prior to this work.
\section{EXPERIMENTAL PROCEDURE}
Our sample was a 99$\%$ deuterated NENP single crystal of mass 0.27~g, which
was a part of the composite sample used in Ref.~\cite{NENP2}. The experiments
were performed on the SPINS cold neutron triple-axis spectrometer at the
National Institute of Standards and Technology. The beam divergences
employed were 50$^{\prime}/k_i$(\AA$^{-1})-80^{\prime}$ around the
vertically focussing pyrolytic graphite PG(002) monochromator, and
$80^{\prime}-240^{\prime}$ around the PG(002) analyzer. A liquid nitrogen
cooled Be filter was placed in the scattered beam path. For constant-Q scans
we fixed the final neutron energy $E_f$ at 5.1~meV. The resulting full width
at half maximum (FWHM) energy resolution for incoherent elastic scattering
was $\Delta \hbar \omega =$ 0.27(1)~meV. Scans with constant energy transfer
$\hbar\omega$ = 2~meV were performed with $E_f = $ 4.24~meV to optimize the
pressure cell transmission for both incident and scattered neutrons.
To provide high pressures, we used a clamp cell for neutron scattering
supplied by Oval Co., Ltd., which is described in detail in Ref.~\cite{cell}.
The pressure was generated at room temperature by applying an external load
of 2$\times$10$^5$~N on the cap of the cylindrical sample chamber, or
``microcell." The microcell is contained inside a barrel-shaped cylinder made
of high-density polycrystalline Al$_2$O$_3$, which in turn is supported by
steel rings. The microcell accommodated the parallelepiped-shaped
NENP sample (dimensions $\approx$ 6$\times$6$\times$8 mm$^3$), as well
as a small NaCl platelet (dimensions $\approx 4\times 4\times 2$mm$^3$)
used for pressure calibration. The microcell was filled with the
pressure-transmitting fluid fluorinert FC-75 (available from 3M Chemicals, Inc.).
The neutron beam reaches the sample through a 10 mm-high, continuous central
window between the steel rings after passing through the Al$_2$O$_3$
pressure cylinder and the 5 mm-thick Al outer sleeve of the cell. The loaded
cell was attached to the cold finger of a pumped He$^4$ cryostat for neutron
scattering and cooled down to 1.8~K.
Fig.~\ref{cellTrans} shows the transmission through the center of the loaded
cell at $T=1.8$ K for neutrons in the energy range $2 \le E \le 14$ meV
(a reduced beam size $\approx 3\times 3$~mm$^2$ was used for this
measurement). The complicated structure of the curve indicates that the
transmission is limited by Bragg diffraction from the Al$_2$O$_3$ cylinder.
In the energy range $4 \le E \le 14$ meV, the measured cell transmission is
well approximated by the superposition of three Gaussians,
\begin{eqnarray}
\label{transmission}
T(E)=&0.085+0.031\cdot \exp\left\{-\left(\frac{E-4.07}{0.283}\right)^2\right\}
-0.035\cdot \exp\left\{-\left(\frac{E-5.3}{0.629}\right)^2\right\}\nonumber\\
&+0.021\cdot \exp\left\{-\left(\frac{E-10.2}{3.58}\right)^2\right\} \;\; ,
\end{eqnarray}
as may be seen in Fig.~\ref{cellTrans}. Scattering intensities measured at
$P=2.5$~GPa were subsequently corrected using the above expression for the
energy-dependent transmission of the pressure cell normalized to unity for
the constant-$\hbar\omega$ = 2 meV scan. Specifically, the actual count rate
was divided by
\begin{equation}
\tilde{T}(\hbar\omega)=
\frac{\sqrt{T(E_f+\hbar\omega)T(E_f)}}{\sqrt{T(6.24~{\rm meV})
T(4.24~{\rm meV})}}~
\approx~$10$\sqrt{T(E_f+\hbar\omega)T(E_f)} ,
\end{equation}
where $E_f$ is the final neutron energy, and $T(E)$ is given by
Eq.~(\ref{transmission}). The actual pressure applied to the sample at low
temperatures in our experiment was $P \approx$ 2.5~GPa, as determined from the
measured change $\Delta a/a$=3.2$\times $10$^{-2}$ in the lattice
constant of the NaCl sample \cite{cell}.
For the reference measurements at ambient pressure we extracted the microcell
with the sample from the pressure cell and placed it in a standard (ILL-type)
flow cryostat. Its transmission for 5 meV neutrons measured the same way as
above was 0.60(1), which can be attributed to incoherent scattering from the
pressure-transmitting fluid and the NENP sample itself.
The sample was mounted with its (0,1,0) and (1,0,3) reciprocal
lattice directions in the horizontal scattering plane. With this
orientation, variations of the transverse (with respect to the spin chains)
components of the wavevector transfer ${\bf Q}=(h,k,l )$ are coupled through
$h = l/3$. However, the non-zero $h$ had negligible effect in our experiment.
First, recalling that inelastic neutron scattering only probes spin fluctuations
polarized perpendicular to {\bf Q} \cite{Lovesey}, we note that the (1,0,0)
direction was $\approx$ 80$^\circ$ out of the scattering plane.
This means that at least 97$\%$ of the spin fluctuations polarized along
{\bf a}$^*$ were always probed. Second, since the dispersion of the excitations
along {\bf a}$^*$ has period 2 in $h$ \cite{NENP1}, it is quite negligible for
$h < 0.2$ which is the case in present study. All constant-Q scans were performed
at $k = 1$, i.e. with wavevector transfer along the chain $\tilde{q}=\pi$.
\section{RESULTS AND DISCUSSION}
Our experimental results are shown in Figs.~\ref{ConstQfig}--\ref{ConstEfig}.
The scattering intensity is shown normalized to 10 minutes counting time,
although due to the low transmission of the pressure cell the bulk of the
high-pressure data was counted three times this long to obtain adequate
statistics. In Fig.~\ref{ConstQfig} we present energy scans at
{\bf Q}=(5/6,1,5/2). The large component of {\bf Q} transverse to the chains
means that more than 90$\%$ of the intensity of the longitudinally polarized
(i.e. parallel to the chain axis) fluctuations is observed. At ambient pressure
(0.1MPa) the corresponding gap is $\Delta_\|$= 2.49~meV, in agreement with
previous studies \cite{NENP1,NENP2}. At $P = 2.5$~GPa this mode shifts to lower
energy, and is found at $\Delta_\|$( 2.5 GPa) = 2.2~meV. The numbers quoted here
are determined from the fits discussed below. In contrast, the energy of the
transversely polarized mode, which at this {\bf Q} primarily contains
fluctuations polarized along {\bf a}$^*$, increases under pressure from
$\Delta_\perp$ = 1.18~meV to $\Delta_\perp$(2.5GPa) = 1.43~meV. This increase
of the energy gap in the magnetic excitations spectrum qualitatively agrees with
the results of recent heat capacity measurements \cite{Ito} performed at pressures
up to $\approx$ 0.5~GPa.
Figure \ref{dispFig} shows scans that probe the dispersion of the transversely
polarized modes with wavevector transfer perpendicular to the chains
$q_\perp\approx 2 \pi l$. Presence of cell material in the beam cause much
higher inelastic background and elastic scattering intensity in the
high-pressure data. For $0\leq l \leq 0.5$, {\bf Q} makes angles between
0$^\circ$ and $33^\circ$ with the chain direction, so that between
100\% and 70$\%$ of the intensity of the {\bf a}$^*$-polarized transverse
fluctuations and up to 30$\%$ of the intensity of the longitudinal fluctuations
is measured. The latter is seen as a small rise after the main peak in the
high-pressure data where the corresponding mode is centered at $\approx$ 2.2~meV.
As shown in Fig.~\ref{c-disp}, a small shift of $\Delta_\perp$ to higher
energies as $l$ changes from 0.5 to 0 is clearly observed at ambient
pressure. At P = 2.5~GPa, however, this shift is much smaller, and barely exceeds
the experimental accuracy. Finally, one constant-$\hbar\omega$ = 2~meV
scan was performed at each pressure to probe the dispersion of the transverse
fluctuations with $\tilde{q}$ along the chains. These scans are shown in
Fig.~\ref{ConstEfig}.
The solid curves shown in Figs.~\ref{ConstQfig}--\ref{ConstEfig}
were obtained as a result of a global fit to all the data at each pressure
using the resolution-convoluted theoretical expression \cite{Lovesey}
for the neutron scattering cross-section based on the single-mode
approximation (Eq.~(\ref{SMA})) for ${\cal S}^{\alpha\alpha}({\bf Q},\omega)$.
To account for the dispersion perpendicular to the chains we used
the dispersion relation
\begin{equation}
\label{hw2}
\hbar\omega_\alpha ({\bf Q}) = \sqrt{\Delta_\alpha^2 + v^2\sin^2
\tilde{q}+A\sin^2\frac{\tilde{q}}{2} + (\Delta E_\perp)^2
\frac{1-\cos\tilde{q}\cos q_\perp}{2} } \;\; ,
\end{equation}
with $q_\perp = 2 \pi l$. This was derived by adding in quadrature the energy
of the Haldane modes given by Eq.~(\ref{hw1}) with the transverse dispersion
obtained for magnons in spin-wave theory. We note that for $\tilde{q}$ close to
$\pi$, Eq.~(\ref{hw2}) was obtained by Affleck \cite{Affleck} by
treating the array of weakly interacting Haldane chains on the basis of
the nonlinear sigma model. Except for the common parameters $\Delta_\|$,
$\Delta_\perp$, $v$, $\Delta E_\perp$ in Eq.~(\ref{hw2}), for each scan only
the individual constant background and gaussian intensity and width of the
diffuse tail of incoherent scattering were varied in the global fit. Due
to the limited range of $\tilde{q}$ probed, the parameter A was fixed at
the value 34 meV$^2$ determined previously \cite{NENP2}. The results of
the fits for the parameters in the dispersion relation are summarized in
Table~1. The values $\chi^2$=3.27 and $\chi^2$(2.5GPa)=0.95 once more indicate
the validity of the single mode approximation (\ref{SMA}).
It can easily be shown within the framework of perturbation theory \cite{NENP3}
that the single-ion anisotropy term $D \sum_i(S_i^z)^2$
gives a splitting of the initially isotropic Haldane triplet such
that the average gap value remains unchanged: $\frac{1}{3}\Delta_\| +
\frac{2}{3}\Delta_\perp = \Delta_H$. This fact can also be established
with the symmetry arguments presented in the numerical study
by Meshkov\cite{Meshkov}, where to $\sim 1\%$ accuracy the splitting was
found to obey
\begin{equation}
\label{splitting}
\Delta_\|=\Delta_H+\frac{4}{3}D\;\; ,\;\;\;\;\;\;
\Delta_\perp=\Delta_H-\frac{2}{3}D \;\; ,
\end{equation}
so that $\Delta_\|-\Delta_\perp=2D$.
While we found that the average gap value $\Delta_H$ essentially did not
change under pressure, the splitting of the gap decreased by more than 50$\%$.
This implies that only $D$ is affected by hydrostatic pressure, while $J$
remains unchanged. Using Eq.~(\ref{splitting}) we arrive at the value
$D$(2.5~GPa) = 0.09(7)$J$ for the anisotropy constant as compared to
0.16(2)$J$ under normal conditions. This result can be given a simple
interpretation: since the hydrostatic pressure is isotropic,
it tends to pack charge distributions inside the crystal more
symmetrically, thereby reducing the local anisotropy. From the transverse
dispersion parameter $\Delta E_\perp\sim 4S\sqrt{2J'J}$, we estimate
the exchange along the {\bf c}$^*$ direction to be $J'_c \approx 3\times
10^{-4} J$ and $J'_c$(2.5GPa)$\approx 2\times 10^{-4} J $. This is even
smaller than the value $J'_a \approx 8\times 10^{-4} J $ in the {\bf a}$^*$
direction measured previously \cite{NENP1}. Thus $J'_c$ appears to decrease
with pressure, but this change is at the limit of our experimental accuracy.
\section{CONCLUSIONS}
In conclusion, changes in the spin dynamics of the Haldane gap antiferromagnet
Ni(C$_2$H$_8$N$_2$)$_2$NO$_2$ClO$_4$ (NENP) under applied hydrostatic
pressure of 2.5~GPa were studied by inelastic neutron scattering, and another
point in the (D/J,J$'$/J) phase diagram of the S=1 spin system in the Haldane
state has been attained experimentally. The principal effect of pressure was
found to be a decrease by a factor 1.7 of the single-ion anisotropy constant
in the spin Hamiltonian of the Ni$^{2+}$ magnetic ions. This finding explains
the increase with pressure of the effective spin gap in NENP observed in the
recent heat capacity measurements \cite{Ito}, as being the result of the
increasing energy of the lower doublet component of the Haldane triplet.
The exchange constant in the $\hat{c}$-direction perpendicular to the chains
was also measured for the first time. It is smaller than the exchange along
$\hat{a}$, and seems to decrease slightly with pressure. Thus the overall
effect of applying hydrostatic pressure in NENP is not to move the system
towards the phase transition to the N\'{e}el ordered ground state or the
large-$D$ planar phase but to bring it closer to the isotropic 1D limit.
\section{ACKNOWLEDGEMENTS}
This work is based upon activities supported by National Science Foundation
under agreement No. DMR-9413101. CB acknowledges support from the National
Science Foundation through DMR-9453362. DHR acknowledges the support of the
David and Lucile Packard Foundation.
\begin{references}
\bibitem[\dag]{Kapitza} Permanent affiliation: P.~Kapitza Institute for
Physical Problems, ul. Kosygina 2, 117334
Moscow, Russia.
\bibitem[\ddag]{NIST} Present address: National Institute of Standards
and Technology, Gaithersburg, MD 20899.
\bibitem{Haldane} F.~D.~M.~Haldane, Phys. Lett. {\bf 93A}, 464 (1993);
Phys. Rev. Lett. {\bf 50}, 1153 (1983).
\bibitem{NENP1} L.-P.~Regnault, I.~Zaliznyak, J.~P.~Renard and
C.~Vettier, Phys. Rev. B {\bf 50}, 9174 (1994), and
references therein.
\bibitem{NENP2} S.~Ma, D.~Reich, C.~Broholm, B.~Sternlieb and
R.~W.~Erwin, Phys. Rev. Lett. {\bf 69}, 3571 (1992);
Phys. Rev. B {\bf 51}, 3289 (1995).
\bibitem{YBaNO} G.~Xu, J.~F.~DiTusa, T.~Ito, K.~Oka, H.~Takagi,
C.~Broholm and G.~Aeppli, Phys. Rev. B {\bf 54}, R6827
(1996), and references therein.
\bibitem{AgVP} H.~Mutka, C.~Payen, P.~Molinie, J.~L.~Soubeyroux,
P.~Colombet and A.~D.~Taylor, Phys. Rev. Lett. {\bf 67},
497 (1991); Physica B {\bf 180-181}, 197 (1992).
\bibitem{TMNIN}
V.~Gadet, M~Verdaguer, V.~Briois, A.~Gleizes,
Phys. Rev. B 44, {\bf 44}, 705 (1991);
L.-K. Chou, K.A. Abboud, D.R. Talham, W.W. Kim, and M.W.Meisel,
Chem. Mater. {\bf 6}, 2051 (1994).
\bibitem{Affleck} I.~Affleck, Phys. Rev. Lett. {\bf 62}, 474 (1989).
\bibitem{Azzouz} M.~Azzouz and B.~Dou\c{c}ot, Phys. Rev. B {\bf 47},
8660 (1993).
\bibitem{Senechal} D.~S\'{e}n\'{e}chal, Phys. Rev. B {\bf 48},15880
(1993).
\bibitem{Botet}
R. Botet, R. Jullien, and M. Kolb, Phys. Rev. B {\bf 28}, 3914 (1983).
\bibitem{SchulzZiman}
H. J. Schulz and T. Ziman, Phys. Rev. B {\bf 33}, 6545 (1986).
\bibitem{Sakai} T.~Sakai and M.~Takahashi, J.~Phys. Soc. Jpn.
{\bf 58}, 3131 (1989); Phys. Rev. B {\bf 42}, 4537 (1990).
\bibitem{Meshkov} S.~V.~Meshkov, Phys. Rev. B {\bf 48}, 6167 (1993).
\bibitem{Jolicoeur} O.~Golinelli, Th.~Jolicoeur and R.~Lacaze,
J.~Phys.: Condens. Matter {\bf 5}, 7847 (1993); Phys.
Rev. B {\bf 46}, 10854 (1992).
\bibitem{Yelon-Cox} W.~B.~Yelon and D.~E.~Cox, Phys. Rev.~B {\bf 6},
204, (1972); {\bf 7}, 2024 (1973).
\bibitem{Meisel} O. Avenel {\em et al.}, J. Low Temp. Phys.
{\bf 89}, 547 (1992).
\bibitem{Meyer}A. Meyer {\em et al}., Inorg.\ Chem.\ {\bf 21}, 1729 (1982).
\bibitem{cell} A.~Onodera, Y.~Nakai, N.~Kunitomi, O.~A.~Pringle,
H.~G.~Smith, R.~M.~Nicklow, R.~M.~Moon, F.~Amita,
N.~Yamamoto, S.~Kawano, N.~Achiwa and Y.~Endoh,
Japanese Journ. Appl. Phys. {\bf 26}, 152 (1987).
\bibitem{Lovesey} S.~W.~Lovesey, {\it Theory of Neutron Scattering
from Condensed Matter} (Clarendon, Oxford, 1984).
\bibitem{Ito} M.~Ito, H.~Yamashita, T.~Kawae and K.~Takeda,
J.~Phys. Soc. Jpn. {\bf 66}, 1265, (1997).
\bibitem{NENP3} L.-P.~Regnault, I.~A.~Zaliznyak and S.~V.~Meshkov,
J.~Phys.: Condens. Matter, {\bf 5} L677, (1993).
\end{references}
\begin{figure}
\caption{ Energy dependence of neutron transmission for the pressure cell
loaded with our NENP sample at $T$ = 1.8~K. The solid line is an approximate
fit given in Eq.~(\protect\ref{transmission}). }
\label{cellTrans}
\end{figure}
\begin{figure}
\caption{Constant-{\bf Q} scans at ${\bf Q}=(.833,1,2.5)$ picking up both
transverse and longitudinal fluctuations at $P=2.5$~GPa (top) and $P=0.1$~MPa
(bottom).}
\label{ConstQfig}
\end{figure}
\begin{figure}
\caption{Scans at different $\tilde{q}_\perp$ probing the dispersion of
the transverse fluctuations with wavevector transfer perpendicular to the
chains: (a)--(c) $P=2.5$~GPa; (d) and (e) $P=0.1$~MPa. }
\label{dispFig}
\end{figure}
\begin{figure}
\caption{Dispersion curves for the transverse polarized gap mode obtained
with Eq.~(\protect\ref{hw2}) using parameters from Table~1. Open symbols and
error bars were obtained from independent unconstrained fits of each
individual peak with the resolution convoluted lineshape. }
\label{c-disp}
\end{figure}
\begin{figure}
\caption{Constant-$\hbar\omega$ scans sensitive to the
velocity of excitations
along the chains at $P=2.5$~GPa (top) and $P=0.1$~MPa (bottom).}
\label{ConstEfig}
\end{figure}
\newpage
\begin{table}
\caption{ Summary of the dispersion parameters obtained from the global fits
to data at two pressures; $\Delta_H$ is the average gap defined with
Eq.~(\protect\ref{splitting}).}
\vspace{10mm}
\begin{tabular}{||c|c|c|c|c|c||}
%\hline
Pressure & $\Delta_\perp$ & $\Delta_\|$ & $\Delta_H$ & $v$ & $\Delta E_\perp $ \\
(Pa) & (meV) & (meV) & (meV) & (meV) & (meV)\\
\hline
10$^5$ & 1.180(4) & 2.49(1) & 1.62(1) & 9.34(20) & 0.43(3) \\
\hline
2.5$\cdot 10^9$ & 1.426(6) & 2.20(2) & 1.68(2) & 9.47(20) & 0.32(7) \\
%\hline
\end{tabular}
\vspace{3mm}
\end{table}
\end{document}