1. IntroductionPeriodic modulations of electron density from electron–phonon interactions lead to a novel type of ground state called charge density wave (CDW).[1] A great deal of materials such as layered transition metal dichalcogenides (TMDCs) and transition metal trichalcogenides (TMTCs) have been found to possess the CDW states.[1,2] Recently, these interesting states show unexpected and exotic properties when approaching the 2D limit. For example, with the thickness decreasing,[3] a commensurate CDW state completely disappears in TaS
; and the transition temperature can be drastically enhanced in monolayer NbSe
and TiSe
.[4–7] More interestingly, the quantum critical behavior may exist in 2D TaS
and other correlated systems[8,9] as the reduced dimensionality can strengthen Peierls instability, electron–phonon interactions, and the fluctuation effect.[1,7] Therefore, it is indispensable to investigate the effect of the dimensionality on the CDW states and uncover the intrinsic mechanism.
Like TMDCs, many bulk TMTCs exhibit layered structures with interlayer van der Waals interactions.[10–12] As a typical TMTC,
-type crystals (M = Ti, Zr, Hf, Ta, Nb; X = S, Se, Te) are quasi-1D materials, some of which possess charge density waves in the bulk form. And these states emerge and slide along the direction of metal-atom chains unlike in TMDC materials where the various CDW states are developed by several-atoms-clusters in 2D layers.[13,14] For instance, two phase transitions at 144 K and 59 K in NbSe
are the consequence of three structurally different chains.[15] Monoclinic TaS
has two types of chains which induce CDW transitions at 240 K and 160 K.[16] Importantly, the low dimensionality may exert strong effects on those chain-like quasi-1D materials due to the different crystal structures. Each CDW transition can be variously affected with the change of device scale along different crystal orientations, thus giving rise to an evident anisotropy. However, limited works have been reported on the CDW states in quasi-1D TMTC materials when the dimensionality is reduced to two dimensions.
TiS
, as one of the semiconducting TMTCs, shows extremely excellent optical properties.[17–21] Strong anisotropy from electric transport measurements represents 30% of increase over the black phosphorus.[19,22] The bulk crystals have phase transitions at 120, 60, and 17 K through quasi-1D transport.[23–26] with the first two along the perpendicular chains. We anticipate that the nanoribbon structure can have structural fluctuations in horizontal and perpendicular directions that lead to anisotropic CDW behavior upon external electrical gating. However, to date such studies remain lacking.
Here, we report the properties of exfoliated few-layer TiS
nanoribbons and find only two phase transitions at 46 K (
and 85 K (
in stark contrast to the three in thick TiS
at 29, 53, and 103 K.
can be tuned by the back-gate voltage because of the variable carrier concentration whereas
does not show an evident change owing to its robustness to the structural fluctuations. Our transport experiments clearly exhibit a strong anisotropy of CDW states in TiS
. We also reveal the light-induced melting of charge density wave under 633-nm laser illumination. When the laser intensity exceeds 10 μW, both two transitions disappear. Moreover, a low illumination intensity of 0.02 μW by a 532-nm laser can readily destroy the CDW states.
3. ResultsBulk TiS
crystals were synthesized via chemical vapor transport (CVT) method (sulphuration of titanium powder).[17]
Figure 1(a) shows a layered structure of TiS
; parallel chains of triangular prisms which are held by van der Waals forces constitute sheets. The parallel chains are responsible for the quasi-1D nature of the material which leads to charge density waves. Each sheet is formed by two titanium layers and four sulfide layers. The as-grown TiS
naturally develops groups of black needle-like bulk crystals. Figure 1(b) shows one of the exfoliated TiS
samples transferred onto SiO
(285 nm)/Si substrate with a number of nanoribbons. Figure 1(c) shows a scanning electron microscopy (SEM) image, where both bulk crystals and thin nanoribbons can be identified. The purple one in the red rectangle is a thin nanoribbon with a thickness of
nm. To identify the composition of sample, energy dispersive x-ray spectroscopy (EDX) has been carried out. Figure 1(d) displays the EDX results of the as-grown samples. More than five positions have been randomly picked and measured, all of which show clear peaks of Ti and S with the total error less than 0.5%. The atomic ratio of Ti and S is nearly Ti:S=1:3, indicating a correct composition of TiS
. We further carried out Raman spectroscopy to explore the phonon characteristics of materials and identify the chemical bond. Figure 1(e) displays the Raman spectrum of an isolated nanoribbon thicker than 50 nm; the optical image of the sample is shown in the inset under 532-nm laser illumination. While the peak at
is due to Raman mode of silicon substrate, the other four peaks located at
,
,
, and
correspond to A
-type Raman modes of the TiS
which are in good agreement with the bulk crystals. Besides, a previous study reports the observation of the
peak which has the highest intensity along b axis in nanoribbons,[27] which is consistent with our results.
In order to study the CDW properties, four-terminal devices with Cr/Au (5 nm/100 nm) electrodes were fabricated along the nanoribbon chain direction, i.e., the b axis, by standard e-beam lithography, metal evaporation, and lift-off process. Temperature-dependent resistance (R–T) is measured from room temperature down to 10 K as shown in Fig. 2(a). The inset displays the device image by a conventional optical microscope. The length of the sample is around 20 μm and the width is 2 μm. The thickness of the nanoribbon is determined to be
nm, which is relatively thick. The resistivity measured at 300 K has a value of
ρ=0.4Ω⋅cm−1
in good agreement with the previously-reported bulk results.[28] The R–T curves in our experiments are all measured in the cooling process to eliminate the hysteresis induced by the system. As the temperature cools down, the resistance shows several broad anomaly features. In order to clearly identify the transition temperatures, we adopted the widely-used derivative method and plotted
against the temperature. It is known that near the transition temperature, the conductivity can be extracted by
| |
based on the mean-field theory, where
is the mobility and
is the density of states.
[16] When
, the temperature derivative of the resistivity is approximately given by
Consequently, the derivative
has a cusp singularity at transition temperature. Here, we use the maxima from the derivative curve.
[29] to estimate the transition temperature of 29, 53, and 103 K in the
nm-thick nanoribbon, which are similar to that of the bulk crystals.
[23] Instead of showing a cusp, one observes a broadened maximum, which is consequence of fluctuations caused by the 2D dimensionality. At the same time, quasi-1D systems always induce nonlinear conductivity due to the sliding or creeping of the CDW states.
[30] And the current–voltage (
I–
V) curves follow the power law behavior
in CDW systems (along the chains) at low temperatures. Indeed, the
I–
V curve at 10 K exhibits a voltage-dependent conductivity whereas at 130 K it becomes linear with a constant conductivity. Figure
2(b) summarizes all
I–
V curves in the temperature range of 10 K to 130 K where a nonlinear behavior is exclusively observed. However, owing to the possible contribution from the Schottky barrier which is observed in monolayer MoS
using four-point probe measurement, the exact onset of the CDW transition remains elusive.
[31] We note that the four-probe method is unsuitable if the Schottky behavior occurs between the metallic contact and the sample where the Schottky barrier induces a large contact resistance causing inaccuracy of sample resistance.
[32,33] Since the nanoribbon is thick enough, the drain–source current
can be defined by three-dimensional (3D) thermionic emission equation
[34]
where
A is the Richardson constant,
n is the ideality factor. The fitting results show a Schottky barrier
of 42.3 meV (see Fig.
S1 in the supplementary information in Appendix
A). Thus, our
I–
V curves consist of contributions from both quasi-1D CDW and Schottky barrier. When the voltage is larger than 0.4 V,
is proportional to
V, matching the thermionic emission model (Fig.
S1(a), and the current induced by Schottky barrier at low voltage is much smaller than that from the quasi-1D transport. Therefore, at low voltage, the quasi-1D transport dominates while the Schottky barrier plays important roles at the high-voltage regime. When the voltage is lower than the threshold, the electrons cannot be excited across the CDW gap in a two-fluid model.
[1] The current can be written as
, where the subscripts n and CDW refer to the current carried by the uncondensed and condensed electrons. The former has an Ohmic contribution and the latter has a nonlinear response. For CDW materials, the threshold field
is around few millivolt per micron,
[1] which is much lower than that in the fitting region of Fig.
2(b). Moreover, no threshold field is observed in nonlinear transport measurements in bulk TiS
in the high temperature region,
[23] unlike the characteristics in very thin NbSe
samples with disorders.
[35] At 4.2 K, the threshold field
is reported to be 6 V/cm.
[36] Therefore, at a specific temperature, the
I–
V curves have two linear parts as shown in the inset of Fig.
2(c); the blue region is dominated by the quasi-1D CDW transport and the red regime corresponds to the Schottky barrier. We fit both the two regions by using the power law
at different temperatures; the resultant temperature-dependent parameter
is illustrated in Fig.
2(c). It can be seen that
becomes larger than 1 as the temperature goes below 110 K and a sharp increase occurs at about 60 K.
The phase transitions at 53 K and 103 K can be explained by the formation of two charge density waves in disordered quasi-1D system. It was found that below 110 K, a CDW state develops in the direction across the chains, like ZrTe
, while the other CDW forms along the chains below 60 K.[23] Our narrow TiS
nanoribbon increases the fluctuation in the direction across the chains, thus weakening the CDW, which also explains the weak non-linear transport after the first CDW state at 110 K. When the temperature is below 25 K, the noise at low voltage becomes ignorable; correspondingly the fitting parameter
has large errors to be analyzed. As the voltage becomes larger, the Schottky barrier turns to be prominent (the red curve). Below 130 K, the I–V curve turns to be nonlinear, and
gets larger with the temperature reduced, but three peaks clearly show up with two of which corresponding to the aforementioned two CDW states. The transition at 25 K, however, is still unknown and needs further investigation. The transition temperatures from the quasi-1D transport analysis and
derivative curve are quite similar, which also confirms the three CDW phase transitions.
Next, we focus on the thinner nanoribbons. The fabrication method is the same as before, but the sample thickness is reduced to < 15nm as displayed in the inset of Fig. 2(d). Note that experimentally it is challenging to reach monolayer TiS
nanoribbon because of the 1D chain structure and the subsequent electric measurements are difficult owing to the extremely large resistance when approaching the atomic thickness. The R–T curve shows anomaly peaks at 46 K and 85 K (the blue curve), which correspond to the two peaks in the derivative curve (in red). The third transition cannot be observed in transport measurements for which the underlying mechanism is not clear yet. We note that the carrier density can strongly affect the CDW phase.[16,37] Compared with the thick device, in the thin nanoribbons we find a rapid diminishment of the two transition temperatures. We provide two reasons to explain this behavior as follows. First, the scattering mechanism in TiS
is dominated by charge impurities at low temperatures (see Fig. S2 in Appendix A). The phase transition is greatly suppressed by impurities in quasi-2D or 1D CDW materials.[16] Non-isoelectronic impurities can induce potentials which lead to changes of the phase of the condensate, destroy long-range CDW order and lead to a finite phase–phase correlation length
.[38,39] Second, lowering the dimensionality of a system usually tends to reduce the phase space available for the phase parameter
to adapt to imperfections, and consequently it suppresses the long-range order and lowers the transition temperature
.[16] Moreover, the Coulombic interaction can modulate transport properties in two dimensions.[40] With the thickness decreased, the interaction between the SiO
/Si substrate and TiS
nanoribbon gradually dominates which can also jeopardize the CDW coherence. The size effect of CDW transition temperature can be described by empirical formula[41]
| |
where
is the bulk CDW transition temperature,
is the critical thickness at which the CDW transition vanishes,
is a temperature correction term. Alternatively, we think that the second CDW transition is more stable in the bulk or nanosheet but not in nanoribbon (chains) because
is in the direction across the chains. In other words, the interlayer coupling is unfavorable for the second CDW state. The device in Fig.
3(b) has a phase transition at 40 K and 100 K, which is different from the device in Fig.
3(a). This may be due to the different thicknesses, impurities, and electron concentrations. Therefore, at large gate voltage the
decreases. The transition temperature in TiS
nanoribbon shows opposite thickness-dependence trend with quasi-2D TMDCs (NbSe
, TiSe
which will help us to better understand the relation between Peierls instabilities and fluctuation effects.
[7,42]In bulk materials, the doping and high pressure can modulate the carrier density, for instance, to enhance the superconductivity in Cu-doped TiSe
.[43] In quasi-2D materials, the electric-field effect is much more convenient than these two methods: a gate voltage can induce a large number of electrons (holes) into the system so that CDW can be easily tuned.[3,37] Here, we use a back-gate voltage to tune the phase transitions in 15-nm thick nanoribbon. Figure 3(a) illustrates a set of R–T curves at gate voltage ranging from
V to 70 V using a constant source–drain current of 1.5 nA. At 40 K, the resistance changes from 2.6 G
to 0.07 G
. From
V to
V, a peak can be clearly identified near the first transition temperature (Fig. 3(a)). The semiconducting phase converting to be metallic looks more evident when we use logarithmic coordinates as demonstrated in the inset. It is possible that the large CDW conduction modulation is due to changes in CDW pinning arising from gate–voltage-induced variations in the CDW gap which was also observed in semiconducting quasi-1D material TaS
.[44] Likewise, figure 3(b) displays the derivative curves to ensure the extraction of the transition temperature. Apparently, two peaks are shifted as a function of the gate voltage. The gate–voltage dependence of conductance at 160 K is displayed in Fig. S2(b). At
V, the conductance tends to be zero and the extracted field-effect mobility and on/off ratio are 0.8 cm
and
, respectively (More information is available in Figs. S2 and S3 in Appendix A). Systematic
changes under different gate voltages are summarized in Fig. 3(c). For the first CDW transition, the critical transition temperature does not vary much and it positions at around 40 K. On the contrary,
is identified to be 110 K when
V and reduced to 93 K at
V and in the whole gate region
has negative dependence with the value of gate voltage. It can be naturally explained by the fact that the increase of the electron density at positive voltages suppresses
, which is also supported by a similar behavior occurring in 1T-TiSe
and TaS
.[3,37] Specifically, a large electron density may enhance CDW phase fluctuations, strengthen disorders, and thus ultimately destroy a long-range CDW coherence. More electrons tend to be excited across the gap, the periodic structure of charge density becomes unstable and much lower temperature is needed to reduce the thermal excitation and keep the density wave. This is commonly observed in quasi-1D and quasi-2D systems.[16] In contrast,
decreases a little at larger
; this is presumably caused by its robustness to the fluctuations because of the ribbon structure in which the CDW forms along the chains. However, for the thick nanoribbon sample, the gate cannot tune it well in a sharp contrast to the thin one, as shown in Fig. S4 in the supplementary information (Appendix A).
Since TiS
is a light-sensitive semiconductor, a 633-nm laser (1.94 eV) is used to illuminate the TiS
surface as displayed in the inset of Fig. 4(a). The R–T curve shows two phase transitions at 40 K and 85 K without laser illumination (purple curve). The illuminating region of TiS
is about 20
. As the laser intensity gradually increases, the resistance reduces and the phase transition becomes weaker prior to the disappearance under 1.0 μW. Figure 4(c) displays the derivative curves where the threshold intensity is about 1.0 μW and the two peaks come to be completely invisible above this threshold. Though the resistance is monotonically decreased by increasing the laser power, the position of two peaks do not change below 1.0 μW, which can exclude the heating effect. Like the process of light-induced superconductivity suppression,[45,46] laser can destroy the condensate in the distorted region in the CDW state and cause a prompt loss of charge order and equilibration of the electron–lattice system when the absorbed optical energy is higher than the condensation energy (the energy difference between the free energy of the CDW state and the normal state).[47,48] After the excitation, the periodic electron density structure is destroyed and TiS
converts to the normal semiconducting state. With a low laser excitation, however, the two phase transitions still exist but are softened with the increasing of the laser intensity, which may be due to the insufficient photons absorbed per unit cell.[49] The low laser excitation can raise the fluctuation but not sufficient to destroy the condensation energy. Figure 4(b) shows the I–V curves under different incident laser intensities at 25 K. The resistance reduces after the laser excitation and the I–V curves become linear under 11-μW laser intensity which indicates that the CDW states and Schottky barrier both disappear. Under 1.0-μW laser intensity, the curve is nonlinear, indicating the Schottky barrier existence while CDW states are destroyed. The quasi-linear I–V curve also shows the suppression of CDW states under laser illumination. Figure 4(d) displays the 532-nm laser condition: under a laser intensity of 0.02 μW, the CDW transitions are substantially weakened and two peaks in the derivative curve disappear (see the inset in Fig. 4(d)). Lower critical intensity of 532-nm laser may attribute to the larger illumination energy. In short, we conclude that the CDW state strongly depends on the excitation energy, reminiscent of superconductivity, where the CDW pseudo gap state is easily broken by external factors. Below a critical energy, the system needs much more photons to excite the redundant electrons across the gap. Otherwise, the CDW states can still survive.