1 Introduction
As an archetype for the study of phase transitions and critical phenomena, the Ising model and its variants have been extended beyond mathematics and physics to diverse fields of complex systems ranging from neuroscience to financial markets, all of which are characterized by interacting components leading to emergent collective behavior [
1]. The quantum transverse-field Ising chain (TFIC) shares the same universality class with the textbook classical two-dimensional Ising model, whose exact solution was first obtained by Onsager decades ago [
2]. When the TFIC is further perturbed by a longitudinal field, the exact solvability is lost. Fortunately, however, the scaling limit of the model around the quantum critical point (QCP) can still be described by an integrable field theory with an emergent
exceptional Lie algebra symmetry [
3,
4]. The most salient feature is the appearance of a series of discrete modes corresponding to eight stable quasiparticles, the
particles, with particular ratios of their mass gaps predicted by the field theory [
4−
7].
The quasi-one-dimensional (quasi-1D) spin-1/2 ferromagnetic Ising chain system CoNb
2O
6 has long been exploited for the demonstration of the
physics [
8−
16]. For perfect 1D chains (isolated chains), there is a quantum phase transition from a ferromagnetic to a paramagnetic state at a critical transverse field, i.e., 1D QCP [
17]. The quasi-1D nature of real materials, on the other hand, brings about weak but nonzero interchain interactions that result in a 3D long-range magnetic order, which terminates at the 3D QCP [
11,
18,
19]. Discrete modes corresponding to the
particles are expected to be evident around the 1D QCP, which is confirmed for CoNb
2O
6 by neutron and terahertz spectroscopy [
10,
12−
15,
20]. Moreover, recent studies suggest that CoNb
2O
6 may exhibit complexities beyond the
physics: instead of a model of longitudinal-field-perturbed TFIC, the frustration induced quantum motion of domain walls in CoNb
2O
6 was argued to be alternatively described by a model of twisted Kitaev chain, reminiscent of the honeycomb Kitaev spin liquid [
21]; On another front, gapless fermionic (GLF) excitations were identified around the closely-spaced QCPs by a linear-in-temperature term in the specific heat [
16], and were further found to be localized according to the lack of a corresponding contribution from these excitations to thermal conductivity [
22]. The existence of the GLF excitations is not expected by existing theory. However, the proximity of the 1D and 3D QCPs in CoNb
2O
6 hinders a clear look at the origin of these excitations: are they more related to the 1D or the 3D QCP? Do the low-lying GLF excitations always appear in conjunction with the higher-energy
particles?
A way out may lie in the study of BaCo
2V
2O
8, which is a quasi-1D spin-1/2 antiferromagnetic Heisenberg-Ising (
) chain with Ising-like anisotropy attracting recent attention due to its various novel quantum phases and excitations. In longitudinal fields, the research focus has been on the emergence of an incommensurate spin density wave phase and string excitations [
23−
29]. In transverse fields, the 3D QCP has been identified to correspond to a topological transition between two phases characterized by dual solitonic excitations [
30,
31]. For smaller transverse fields, the “1D QCP hidden inside 3D order” picture for CoNb
2O
6 applies also to BaCo
2V
2O
8 [
32−
34], despite the antiferromagnetic intrachain coupling in BaCo
2V
2O
8 compared to the ferromagnetic one in CoNb
2O
6. BaCo
2V
2O
8 crystallizes in a tetragonal structure, with the edge-sharing CoO
6 octahedra forming a screw chain along the
axis [Fig. 1(a)]. Due to the screw chain structure, the local easy axes of the magnetic Co
2+ ions tilt slightly away from the crystallographic
axis, leading to an anisotropic
tensor [
32,
35,
36]. The consequence is that when a uniform field is applied along, say, the tetragonal
axis, a staggered field perpendicular to both the Ising axis and the uniform field is generated, serving as the transverse field in the TFIC model. Inside the 3D order (with
at zero field), the finite interchain coupling further induces an effective staggered longitudinal field along the Ising axis. These characteristics combined provide all the prerequisites for the realization of a TFIC perturbed by a longitudinal field or, equivalently, a quantum Ising chain in both transverse and longitudinal fields.
Indeed, similar to CoNb
2O
6, the
spectrum in BaCo
2V
2O
8 in transverse fields around its 1D QCP has been identified by several experimental probes, showing quantitative agreement with theory [
34,
37,
38]. Reflecting such physics, the phase diagram of BaCo
2V
2O
8 is shown in Fig. 1(d), with the Néel temperatures obtained from the magnetic susceptibility data in Fig. 1(b). For comparison, the phase diagram of CoNb
2O
6 adapted from Ref. [
16] is shown in Fig. 1(c). Considering the remarkable similarities between the two systems, the low-lying quantum critical excitations in BaCo
2V
2O
8 are also expected to be similar to those in CoNb
2O
6 [
37,
38]. More importantly, compared to CoNb
2O
6, the critical field values for the QCPs in BaCo
2V
2O
8 are
and
, respectively [
34], creating a magnetic field range in which the influence of the 1D and 3D QCPs on the quantum critical excitations can be clearly separated.
Ultralow-temperature thermal conductivity measurement has proven to be a powerful means of studying magnetic excitations in low-dimensional magnets [
22,
39−
60] — a case in point being that it confirmed the localized nature of the GLF excitations in CoNb
2O
6 as mentioned above [
22]. Here, we utilize such measurements down to 50 mK to investigate the quantum critical excitations around the 1D QCP of BaCo
2V
2O
8, in a transverse field along the
axis up to 8 T, i.e., with the presence of the 3D order. At lower temperatures, the heat-carrying phonons in the boundary-scattering limit result in field-independent thermal conductivity
. With increasing temperature, the scattering of phonons by magnetic excitations thermally excited across their gap comes into play, leading to an ever more suppressed
with an increasing magnetic field. Crucially, the lack of a linear-in-temperature contribution to
, together with the monotonic evolution of
across the 1D critical field, clearly rules out the existence of GLF excitations around the 1D QCP. To account for the magnetic excitations that scatter phonons, one needs to invoke only the gapped spinon pair excitations, which span the whole 3D order and manifest as the
particles around the 1D QCP. The contrasting nature of the quantum critical excitations in the two Ising-like spin chains, i.e., the presence of GLF excitations in CoNb
2O
6 and their absence in BaCo
2V
2O
8, suggests that the GLF excitations might only emerge in systems in which the 1D and 3D QCPs are in close proximity.
2 Methods
Single crystals of BaCo
2V
2O
8 were grown by the self-nucleation method [
34]. The magnetic susceptibility was measured with a Quantum Design magnetic property measurement system. For thermal conductivity measurements, the samples were cut into a rectangular shape, with the longest dimension along the
axis. Four silver wires were attached to the sample with silver paint as the thermal contacts. The thermal conductivity was measured in a dilution refrigerator, using a standard four wire steady-state method with two RuO
2 chip thermometers, calibrated
in situ against a reference RuO
2 thermometer. The experimental setup for thermal conductivity measurements is schematically shown in Fig. 1(a), with a heat current along the chain direction
and transverse magnetic fields along
. For a typical sample, the distance between the
and
probes is 1.26 mm, with a width and thickness of 0.29 and 0.25 mm, respectively. We exploit specific procedures to account for heat leaks, thermal contact resistance, and other factors that may affect the data quality at ultralow temperatures. With these procedures, the experimental error bar can be as low as
.
3 Results and discussion
3.1 Absence of GLF excitations in BaCo2V2O8
The thermal conductivity
of BaCo
2V
2O
8 under various fields is shown in Fig. 2(d), with the
of CoNb
2O
6 extracted from Ref. [
22] also shown in Fig. 2(a) for comparison. The most salient feature of the
of BaCo
2V
2O
8 is that there are two temperature regions: the applied field barely affects
at lower temperatures, while it clearly suppresses
at higher temperatures. This is better visualized in Figs. 2(b) and 2(e), where
is plotted. The
of CoNb
2O
6 is clearly negative at finite temperatures, and the curves corresponding to different magnetic fields do not overlap, as a result of the gapless but localized nature of the GLF excitations therein; while for BaCo
2V
2O
8,
is negligible in region I but negative in region II, and the curves corresponding to different magnetic fields diverge from each other only in region II. The choice of the zero-field data as the baseline in the definition of
will be justified later. The normalized
curves of BaCo
2V
2O
8 displayed in Fig. 2(f) are almost flat at lower temperatures, while they show a monotonic decrease with the increasing field at higher temperatures, consistent with the behavior of
and in sharp contrast to the behavior of CoNb
2O
6 where the normalized
shows a minimum around the closely-spaced 1D and 3D QCPs at all temperatures [Fig. 2(c)].
In the temperature region I, the fact that
(for simplicity, we omit the independent variable
or
in the following) is insensitive to the magnetic field already indicates that phonons are the only heat carriers and the magnetic degrees of freedom are barely relevant. Nevertheless, we perform a fitting procedure for further analysis. The thermal conductivity of an insulating magnet can generally be decomposed as
, in which the two terms represent the contributions from magnetic excitations and phonons, respectively. If the GLF excitations observed in CoNb
2O
6 also existed in BaCo
2V
2O
8, they would be expected to give rise to a linear contribution to
, i.e.,
[
22]. At low temperatures, phonons can be scattered by the sample boundaries or by magnetic excitations. The latter scattering channel can be excluded by the field independence of
in region I. Therefore,
, where
is usually between 2 and 3 [
22,
42,
55]. Combining these two terms, one can fit the data to
, and tell whether itinerant GLF excitations contribute to the heat transport based on the value of the residual linear term
.
A typical fitting line for
around the 1D QCP of BaCo
2V
2O
8 is shown in the inset of Fig. 3, while the
values obtained at all fields are summarized in the main panel. The obtained
is weakly dependent on the magnetic field, with a typical value between 2.3 and 2.4, consistent with boundary-scattering limited phonon heat transport. The negligible
at all fields clearly rules out the existence of itinerant GLF excitations in BaCo
2V
2O
8. First hint of a negligible magnetic contribution to
was given based on the lack of difference between the
with a heat current along the spin chain and perpendicular to it [
32,
61] . Here, we determine
down to ultralow temperatures, which allows for an unambiguous extraction of the zero-temperature asymptotic behavior, with a focus on
in fields near the 1D QCP. For comparison, we note that in a typical antiferromagnetic Heisenberg chain copper benzoate, the spinons, which are gapless and fermionic, contribute to a
[
58], two orders of magnitude higher than the upper limit here.
Since there are no direct contributions from magnetic excitations, the field-induced suppression of in the temperature region II can only be attributed to a phonon-dominated thermal conductivity with the phonons being more strongly scattered as the field increases. Moreover, such scattering must be ineffective at lower temperatures to be in accord with the data in region I. A scenario that can reconcile these observations is that there is a set of gapped magnetic excitations scattering phonons. The gap can be thermally overcome in region II, and the gap size decreases with increasing field. Consequently, (i) a substantial number of magnetic excitations start to participate in the scattering of phonons in region II, leading to a downward trend of compared to a simple extrapolation of the phonon-dominated in region I to region II; (ii) more magnetic excitations participate in the scattering of phonons as the field increases, leading to a suppressed . Since in this scenario the gap size is largest at zero field, can be viewed as the closest to a pure boundary-limited-phonon-dominated . In this sense, the value of can act as a measure of the scattering strength of phonons by magnetic excitations thermally excited across their gap.
The above analysis delineates the restrictions on the magnetic excitations set by our data. In the following, we discuss which set of magnetic excitations specifically meets the requirements. The existence of itinerant GLF excitations in BaCo
2V
2O
8 has been ruled out by the negligible
at all fields shown in Fig. 3. What about the possibility of a band of localized GLF excitations, as in the case of CoNb
2O
6 [
16,
22]? If such localized GLF excitations exist around the 1D QCP of BaCo
2V
2O
8, a dip structure centering around
in Fig. 2(f), similar to the case of CoNb
2O
6 in Fig. 2(c), is expected, which is in sharp contrast to the smooth evolution observed here. Furthermore, since the GLF excitations are gapless, if present, they would scatter phonons at any finite temperature, resulting in a nonzero
for all temperatures measured. This is again inconsistent with our observations of a discernible
only in region II [Fig. 2(e)]. Therefore, we can safely rule out the existence of GLF excitations, itinerant or localized, around the 1D QCP of BaCo
2V
2O
8.
In the 3D long-range ordered state of BaCo
2V
2O
8, the established magnetic excitations are spinon pair excitations associated with the formation of quantized spinon bound states [
30,
34,
38]. The magnetic excitation gap corresponds to the confining potential of spinons. The gap size of the lowest-lying mode is determined by terahertz spectroscopy to be
at zero field and decreases by a factor of about two upon approaching the 3D QCP [
38], while neutron scattering and electron spin resonance measurements point to a steeper decrease to
around
[
30,
62]. In Fig. 2(e), the onset temperature for a nonzero
, i.e., the onset temperature for phonon scattering, does not vary strongly with the field, pointing to a slower decrease of the gap. The zero-field gap is sufficiently large so that there are barely any excited spinon pair excitations in the whole temperature range of our measurements; as the gap size decreases with increasing field, more spinon pair excitations are thermally excited in region II, leading to a decreased
due to the enhanced scattering inflicted by these excitations on phonons. Since even for the largest field of 8 T, the gap size remains large compared to our measurement temperature, the decrease of
with the field is only moderate.
The large gap size across the whole field range of our measurement explains the lack of a direct contribution from the magnetic excitations. On the other hand, the phonon thermal transport appears to be a more sensitive probe of the increasing density of the magnetic excitations with increasing field, although the gap may not show a strong decrease with field, and the density of the magnetic excitations remains low all along. In this aspect, we note a recent heat transport study on a ferromagnetic Ising spin
α−CoV
2O
6, where the phonon thermal conductivity below 1 K decreases noticeably with the application of a transverse field, while the magnetic excitation gap remains large (
) even for a 9 T field [
63]. The cases of BaCo
2V
2O
8 and
α−CoV
2O
6 show that the phonon thermal conductivity provides a surprisingly sensitive probe of magnetic degrees of freedom in Ising spin chain systems in a transverse field, whose microscopic origin calls for future studies but is beyond the scope of this work. We remark that although the picture of phonon-dominated heat transport was already invoked in previous studies [
32,
61], the different scattering conditions in regions I and II, and the identification of specific magnetic excitations (i.e., the gapped spinon pair excitations) as the phonon scatterers were not possible until now.
3.2 Why do the GLF excitations emerge only in CoNb2O6?
Finally, we discuss the implications of our results on the quantum critical excitations in various spin chain systems. The absence of the GLF excitations around the 1D QCP of BaCo2V2O8 immediately indicates that the GLF excitations observed in CoNb2O6 are not a feature common to spin chain systems that can be described, at least to some extent, by the longitudinal-field-perturbed TFIC model, and the origin of the GLF excitations in CoNb2O6 must lie beyond the physics.
Are the GLF excitations then purely a consequence of the 3D QCP? It is worth noting that spectroscopic studies all point to a finite magnetic excitation gap at the 3D QCP of BaCo
2V
2O
8 and its sister compound SrCo
2V
2O
8 [
30,
38,
62,
64,
65]. One may envisage a scenario in which gapless excitations may elude spectroscopic probes either due to a mismatch with the energy or momentum window that the measurements are sensitive to, and manifest themselves only in ultralow-temperature specific heat and heat transport measurements that treat all portions of the Brillouin zone on an equal footing [
22]. This scenario may explain why the GLF excitations in CoNb
2O
6 were not found until recently when the specific heat and heat transport measurements were performed [
16,
22]. Such a possibility cannot be excluded, leaving room for the GLF excitations to be present around the 3D QCP of BaCo
2V
2O
8. Previous heat transport studies showed that as a function of the applied transverse field,
exhibits a dip around
down to
, attributed to the scattering of phonons by quantum fluctuations [
32,
61]. One cannot tell from the existing data whether there are GLF excitations around
, since their effect on
is indistinguishable from quantum fluctuations if the potential GLF excitations are localized and can thus affect
only by scattering phonons. The picture is further complicated considering the case of SrCo
2V
2O
8: although still under debate [
66], the emergence of gapless excitations associated with an exposed 1D QCP outside the 3D order (
) was concluded from nuclear magnetic resonance measurements [
65].
Considering that (i) out of the various Ising-like spin chain systems, the existence of the GLF excitations is only conclusive in CoNb2O6 around its closely located 1D and 3D QCPs, and (ii) we have demonstrated through the study of BaCo2V2O8 that the realization of physics does not guarantee the emergence of the GLF excitations, deviations from the longitudinal-field-perturbed TFIC model must be considered in the effort to understand the origin of such novel quantum critical excitations.
In CoNb
2O
6, BaCo
2V
2O
8, and SrCo
2V
2O
8, the magnetic Co
2+ ion with a
electron configuration exhibits a total orbital angular momentum
and total spin
. The ground state manifold is further split by the local octahedral crystal field and spin-orbit coupling, resulting in an
[
7]. This is the common starting point for the effective (pseudo)spin-1/2
Hamiltonian governing the low-energy physics of all three systems. On the other hand, such a setting is reminiscent of the Kitaev physics in honeycomb cobalt-based systems. In fact, the proposal of Kitaev physics realized in cobalt compounds has attracted substantial attention [
67−
70], with prominent examples being the possible realization of quantum spin liquid in BaCo
2(AsO
4)
2 and Na
2Co
2TeO
6 [
56,
60,
71−
74]. For the zigzag chain systems, the glide symmetry breaking and the resultant additional terms in the Hamiltonian of CoNb
2O
6 have been scrutinized recently, leading to the proposal of CoNb
2O
6 as a realization of the twisted Kitaev chain, also known as the generalized quantum spin compass chain [
21,
75−
77], in which the bond-dependent interactions depend sensitively on the geometry of electron orbitals and hopping paths [
76].
In this sense, the emergence of the GLF excitations around the two closely-spaced QCPs in CoNb
2O
6 may not be a coincidence: the specific combination of microscopic parameters, including the level of distortion of the CoO
6 octahedra in the crystal structure, the exchange anisotropy, and the spin-orbit coupling strength, etc., may give rise to the adjacent QCPs and the GLF excitations around them in CoNb
2O
6 but not in related systems. In CoNb
2O
6, one chain is coupled to two other chains on a triangle with antiferromagnetic interchain interactions, leading to strong frustration and, hence, a relatively weak interchain interaction. In BaCo
2V
2O
8, one chain is antiferromagnetically coupled to three other chains on a square. The frustration between nearest-neighbor chains persists, but the interaction with the diagonal chain becomes effective, leading to stronger interchain coupling:
in BaCo
2V
2O
8 [
26,
31], as compared to
in CoNb
2O
6 [
75], where
is the coupling constant in the
Hamiltonian. Note that the precise nature of the interchain coupling is still largely unknown due to the screw nature of the chains, and the estimations are only on a phenomenological level.
It has been recently argued that in CoNb
2O
6, certain spectral features incompatible with the
description originate from excitations described by the
Lie algebra associated with the
integrable model [
78,
79]. Crucially, a key prerequisite for the realization of the
physics is the strong interchain spin fluctuations, whose enhancement can be attributed to the frustrated alignment of chains and the proximity of the 1D and 3D QCPs [
78,
79]. In CoNb
2O
6, the frustrated interchain alignment causes the cancellation of an effective longitudinal field
from neighboring chains. In addition, the ordered moment is minute at the 1D QCP since the long-range order is already close to its demise at the 3D QCP, leading to further suppression of
. Therefore, CoNb
2O
6 is closer to the small
limit where the
physics is expected [
78]. Though the
physics is invoked to address excitations with a higher energy scale than the GLF excitations, it highlights that closely spaced 1D and 3D QCPs indeed have a profound impact on the quantum critical excitations in quasi-1D magnets, consistent with our observations.
In our previous work on CoNb
2O
6, we proposed that the frustrated alignment of chains may be a key factor in the localized nature of the GLF excitations [
22]. The absence of GLF excitations in BaCo
2V
2O
8 demonstrated in this work leads us to further speculate that frustration may be pivotal in determining whether GLF excitations exist in the first place.
4 Summary
We measured the thermal conductivity of the Ising-like spin chain BaCo2V2O8 within its three-dimensional antiferromagnetic order and down to ultralow temperatures. The thermal conductivity is dominated by phonons, while the effect of magnetic degrees of freedom is evident only at higher temperatures. The magnetic excitations scatter phonons more severely as their gap size decreases with the increase of the applied transverse field, leading to a suppressed thermal conductivity. Attributing the magnetic excitations to the spinon pair excitations associated with the physics provides a scenario consistent with all the experimental observations. Combining the absence of a linear-in-temperature contribution, the smooth evolution of the thermal conductivity around the one-dimensional quantum critical point, and the negligible magnetic scattering of phonons at lower temperatures, the existence of gapless fermionic excitations can be safely ruled out. The contrasting cases of BaCo2V2O8 and CoNb2O6 suggest that intertwined quantum critical points of different dimensions may play a fundamental role in the realization of the gapless fermionic quantum critical excitations.