Kinetic inductance of few-layer NbSe2 in the two-dimensional limit Authors: Sameia Zaman, Joel Î-j Wang, Thomas Werkmeister, Miuko Tanaka, Thao Dinh, Max Hays, Daniel Rodan-Legrain, Aranya Goswami, Réouven Assouly, Ahmet Kemal Demir, David K Kim, Bethany M Niedzielski, Kyle Serniak, Mollie E Schwartz, Kenji Watanabe, Takashi Taniguchi, Philip Kim, Riccardo Comin, Jeffrey A Grover, Terry P Orlando, Pablo Jarillo-Herrero, William D Oliver Published: 2026 ABSTRACT AbstractVan der Waals (vdW) superconductors remain superconducting down to the monolayer limit, enabling the exploration of emergent physical phenomena and functionality driven by reduced dimensionality. Here, we report the characterization of the kinetic inductance of atomically thin NbSe2, a two-dimensional van der Waals superconductor, using superconducting coplanar waveguides and microwave measurement techniques familiar to circuit quantum electrodynamics (cQED). The kinetic inductance scales inversely with the number of NbSe2 layers, reaching 1.2 nH/□ in the monolayer limit. Furthermore, the measured kinetic inductance exhibits a thickness-dependent crossover from clean- to dirty-limit behavior, with enhanced dirty-limit contributions emerging in the ultra-thin regime. These effects are likely driven by increased surface scattering, multi-band superconductivity, and geometric confinement. Additionally, the self-Kerr nonlinearity of the NbSe2 films ranges from K/2π = −0.006 to −14.7 Hz/photon, indicating its strong potential in applications requiring compact, nearly linear, high-inductance superconducting quantum devices and detectors. The fabrication and characterization techniques demonstrated here are extensible to the investigation of other two-dimensional superconductors.Subject terms: Superconducting properties and materials, Superconducting devices, Two-dimensional materials Introduction Kinetic inductance (Lk) in a superconductor arises from the inertia of Cooper pair transport in response to a varying electromagnetic field. Unlike geometric inductance, which varies with the geometry of the electronic element, kinetic inductance depends on intrinsic properties of the material, e.g., superfluid density, carrier effective mass, and film thickness. For example, in a clean superconductor, the kinetic inductance per square is Lk,sq = meff/e2n0d, where meff is the effective mass of the charge carriers, n0 is the superfluid density (number of Cooper pairs per unit volume), and d is the thickness of the sample. This relation indicates that Lk,sq increases with decreasing film thickness or suppressed superfluid density. Such an enhancement is useful in quantum circuits, where superconductors with high-Lk are used to realize superinductors—inductive elements with a reactance that approaches and even exceeds the quantum unit RQ = h/(2e)2 = 6.4 kΩ for Cooper pairs, where h is the Planck constant, and e is the elementary charge. Furthermore, the kinetic inductance in a superconductor is directly related to the superfluid stiffness; probing it enables insight into several fundamental properties such as the superconducting gap structure and pairing symmetry in unconventional superconductors1–6. Superinductors are used in a variety of applications, including in superconducting qubit circuits such as fluxonium and 0–π qubits, which require large (and ideally linear) shunt inductance7,8. They have also been used to realize coherent quantum phase-slip circuits9,10, long-range coupling for spin qubits11, and tailored Kerr non-linearities in light–matter interactions12,13. Furthermore, these materials can be used for wide-band parametric amplifiers and high-impedance readout resonators14–18. For many of these applications, low-Lk aluminum is used, and thus, large inductance is realized primarily by an array of Al–AlOx–Al Josephson junctions19–21. However, these junctions exhibit strong nonlinearity and are therefore limited in the current they can support while remaining in the linear inductance regime. This motivates the search for alternative materials that can provide large inductance with minimal nonlinearity. Recent efforts toward this goal have focused on implementing superinductors using amorphous superconducting thin films, such as TiN22,23, NbN24,25, NbTiN26,27, InOx9,28, NbSi29, WSi30, AlOx31, and granular Al32,33. The inherent disorder of these materials suppresses the density of Cooper pairs, leading to high-Lk in the superconducting regime34. Typically, these amorphous superconductors operate in the dirty limit, where the mean free path of charge carriers is shorter than the superconducting coherence length35. This relationship implies that, in the BCS framework, the size of a Cooper pair is greater than the average travel distance between consecutive scattering sites. In contrast, the coherence length of a clean superconductor is shorter than its mean free path, and Cooper pair transport can be regarded as ballistic when the length of the device is shorter than the mean free path. In superconducting quantum devices, disorder and the resulting scattering of charge carriers are expected to affect quasiparticle dynamics and, consequently, device coherence. While the link between disorder, quasiparticle generation, and the associated noise remains an area of active investigation32,36–38, microwave characterization of superconductors with varied disorder levels may offer a way to probe their dynamic responses and help explore the role of scattering in decoherence38. We investigate 2H-NbSe2 (further referred to as NbSe2 for brevity), a transition-metal dichalcogenide and a type-II superconductor. The lattice structure of NbSe2 is hexagonal, with the Nb atoms in trigonal prismatic coordination surrounded by Se atoms (Fig. 1a). The layers are stacked in an ABAB sequence along the c-axis and are bonded by van der Waals forces between adjacent layers. NbSe2 possesses out-of-plane mirror symmetry and broken in-plane inversion symmetry. This broken in-plane symmetry gives rise to an unconventional Ising spin–orbit coupling, which maintains superconductivity in the presence of a strong in-plane magnetic field exceeding 30 T, as observed in the monolayer limit39,40. Both superconductivity and the charge–density wave transition exist from the bulk down to monolayer thickness. However, the superconducting transition temperature (Tc) decreases with reduced film thickness, attributable to a suppressed Cooper-pair density, the presence of a superconductor–vacuum interface, and modifications to the electronic band structure41,42. As the thickness is reduced, NbSe2 exhibits several unusual quantum phenomena, including thickness-dependent Higgs modes43, dissipationless phase diagrams44, unconventional finite momentum pairing45–47, and the emergence of a quantum metallic phase under a small perpendicular magnetic field48. Prior direct current (DC) transport studies indicate that the carrier mean free path in NbSe2 decreases with its thickness due to enhanced surface scattering41,44. When encapsulated by hBN, even bilayer and monolayer NbSe2 exhibit mean free paths (approximately 17 and 15 nm, respectively) that are larger than their corresponding superconducting coherence length (8 nm)40,41,45. These DC transport measurements are consistent with encapsulated NbSe2 exhibiting superconductivity in the clean limit41,49–51. Superconducting resonators and terminations to characterize the kinetic inductance of NbSe2. a Crystal structure of monolayer 2H-NbSe2, top-down and side views. b Measurement circuit schematic. CPW resonators are capacitively coupled to a common transmission line and terminated either directly to ground via aluminum (control) or through a hBN–NbSe2–hBN heterostructure (experiment). The resonant frequency fr of the aluminum λ/4-resonator is determined by the effective inductance and capacitance of the CPW. Termination by NbSe2 introduces additional kinetic inductance, shifting the resonant frequency to a new value fr′. (Zoom-in of inductor) hBN–NbSe2–hBN heterostructure contacted by Al electrodes with both RF and DC input signals applied across the sample and terminated at ground. c Simulated resonator spectra illustrating the effect of increased inductance on the resonant frequency. Lk0 denotes the inductance of the aluminum-terminated resonator, while Lk represents the additional inductance introduced by the NbSe2 sample. In this work, we measure the kinetic inductance to access the complex conductivity of few-layer NbSe2 in the two-dimensional limit. This method allows us to determine whether the material lies in the clean limit, the dirty limit, or an intermediate regime. Moreover, it is sensitive to fundamental superconducting properties such as superfluid stiffness, surface impedance, and loss mechanisms4,52,53, providing insight into the material suitability for superconducting quantum devices operating in the microwave regime. Results We study the kinetic inductance of hBN-encapsulated NbSe2 thin films using quarter-wavelength λ/4 coplanar waveguide (CPW) resonators made of thin-film Al. The resonant frequency of a resonator is fr=12πLeffCeff, where Leff and Ceff are the effective inductance and capacitance of the resonator, respectively (Fig. 1b, left panel). The Al inductance is primarily geometric, and its kinetic inductance per square is much smaller than that of NbSe2. When the λ/4-resonator is terminated to ground by an NbSe2 sample, the resonant frequency shifts to a lower frequency fr′=12πLeff′Ceff′ due to the additional kinetic inductance from the NbSe2 (Fig. 1b, right panel). We perform microwave simulations to determine the expected frequency shift of a λ/4-resonator due to the additional inductance Lk (Fig. 1c). The Al-only resonator has a frequency of 5.6 GHz. When terminated by NbSe2, the resonance frequency decreases by 200 MHz per 0.1 nH of added termination inductance. Figure 2a shows an optical micrograph of the superconducting circuit with λ/4 CPW resonators used for this work. The aluminum CPW resonators are fabricated from a 250 nm Al film deposited on a high-resistivity silicon substrate and patterned using a combination of photolithography and wet etching (see ref. 4 and the “Methods” section for details). Two λ/4-resonators, an aluminum spectator resonator, and an experiment resonator, are capacitively coupled to a common feedline. The spectator resonator is used as a witness to monitor the impact of the vdW materials fabrication process (without incorporating vdW materials), while the experiment resonator is terminated by a NbSe2 sample situated within a square hosting window of size 200 μm × 200 μm. To extract the kinetic inductance of each NbSe2 device, we analyze two λ/4-resonators: the NbSe2-terminated experiment resonator (Fig. 2b) and a control resonator of the same geometry that is terminated instead by a 250 nm-thick aluminum film (Fig. 2c). The NbSe2 thin film is encapsulated in hBN to maintain a nominally clean and atomically flat interface while ensuring a relatively low-microwave-loss environment53,54. This hBN–NbSe2–hBN heterostructure is assembled through mechanical exfoliation and standard dry-transfer techniques inside an argon-filled glove box, ensuring minimal oxidation of the NbSe2 before full encapsulation. Measurement configuration and device fabrication for the kinetic inductance measurement of NbSe2. a Optical micrograph of a 5 × 5 mm2 chip containing CPW resonators, a shared feedline, DC bias lines, on-chip LC filters, and a ground plane, all patterned from 250 nm-thick aluminum on a high-resistivity silicon substrate. Only one (of four) DC bias lines is used in this experiment. b and c Optical micrographs of the NbSe2-terminated and aluminum-terminated λ/4 CPW resonators, respectively. d–g Illustration of the edge-contact fabrication process. The NbSe2 (purple) flake is fully encapsulated by hBN (blue) and patterned via reactive-ion etching (RIE) to expose its edges. After in-situ argon ion milling, superconducting edge contacts are formed by angled aluminum evaporation with substrate rotation. The hBN–NbSe2–hBN heterostructure is galvanically connected to the microwave resonator and the ground plane. To achieve a highly transparent, superconducting contact, we make a one-dimensional edge-contact to the NbSe2 using reactive ion etching of the heterostructure, followed by in-situ argon ion milling and aluminum deposition (Fig. 2d)55. A low-loss superconducting contact is critical to maintaining a narrow resonance line (i.e., a high quality factor of the resonator), which translates to higher measurement resolution of the resonator frequency. Details of the fabrication process and the verification of the superconducting contact are provided in the Supplementary Note 1 and Supplementary Table 1. The samples are characterized in a dilution refrigerator with a base temperature of ~10 mK (see Supplementary Fig. 2 for wiring diagram). The microwave transmission coefficient S21 is measured as a function of frequency using a two-port vector network analyzer (VNA). Fig. 3a shows a representative ∣S21∣ measurement from a λ/4-resonator terminated by a 9 nm thick (determined by atomic force microscopy) NbSe2 film (Device ID D6, see Table 1). The resonant frequency fr,Al-NbSe2 and quality factor are determined from a Lorentzian fit using the standard circular fit procedure56 (Fig. 3a inset and Supplementary Note 2). In the single-photon limit, the NbSe2-terminated resonator exhibits a resonant frequency fr,Al-NbSe2 = 4.79 GHz and an internal quality factor Qi,Al-NbSe2=2.0×104, while the Al-terminated control resonator has a resonant frequency fr,Al = 5.34 GHz and an internal quality factor Qi,Al = 1.0 × 105. The lower internal quality factor observed in the NbSe2-terminated Al resonator is likely due to additional dissipation arising from imperfect electrical contacts between the Al and the NbSe2, as well as dielectric loss introduced by residual polymer contamination from the dry transfer process. Moreover, the frequency difference fr,Al−fr,Al-NbSe2 = 0.55 GHz highlights the impact of terminating the resonator with NbSe2 and enables us to extract a kinetic inductance Lk,sq = 124 pH/□ for this NbSe2 sample. Microwave measurements of NbSe2-terminated λ/4-resonator. a Transmission coefficient (∣S21∣2) of a NbSe2-terminated λ/4-resonator measured at 10 mK (Device ID D6). Inset: Complex I-Q plane showing the measured response (blue) and Lorentzian fit (red). b Transmission coefficient (∣S21∣2) of NbSe2-terminated λ/4-resonator as a function of input microwave power (Prf). The resonant frequency (fr,Al-NbSe2) shifts to the lower frequency with increasing microwave power. The onset of the frequency shift—defined as exceeding one standard deviation from the low-power baseline (fr,Al-NbSe2= 4.785 GHz)—is observed at an input power of −100 dBm (indicated by the green arrow). At −82 dBm, the resonant peak bifurcates (blue arrow), indicating the onset of a bistable regime. c Linear fit of the resonant frequency shift (Δf=fr,Al-NbSe2(nr)−fr,Al-NbSe2(nr=1)) as a function of resonator photon number (nr) in the NbSe2-terminated λ/4-resonator. d Optical image of the NbSe2-terminated λ/4-resonator with a DC bias line connected at the microwave input port. e DC bias dependence of the NbSe2-terminated λ/4-resonator measured at a fixed microwave power of Prf = −120 dBm. The resonant frequency shifts to the lower frequency with increasing bias current. The onset of the frequency shift—defined as exceeding one standard deviation from the baseline (fr,Al-NbSe2= 4.785 GHz)—is observed at a bias current of 1 μA (maroon arrow). f Extracted resonant frequency shift (Δf=fr,Al-NbSe2(IDC)−fr,Al-NbSe2(IDC=0)) and kinetic inductance shift (∣ΔLk,sq∣=Lk,sq(IDC)=0) Lk,sq(IDC)) of the NbSe2-terminated resonator as a function of DC bias current. Measured data (circle-green and sky-blue) and theoretical fit (line-green) are shown. Table 1Summary of kinetic microwave characterization of NbSe2 samples across different thicknessesDevice IDd (nm)W (μm)L (μm)fr,Al-NbSe2 (GHz)fr,Al (GHz)Lk (nH)Lk,sq (nH/sq)Qi,Al-NbSe2K/2π (Hz)D11.993.13.13.625.301.1951.1950.8 × 104−14.7D24.082.55.74.145.210.6430.2826.0 × 104−3.8D35.862.63.04.885.330.2200.1910.7 × 104−2.9D47.134.86.74.935.350.2000.1432.4 × 104−2.6D57.984.512.04.635.330.3600.1351.2 × 104−1.9D68.813.98.64.795.340.2740.1242.0 × 104−1.7D78.904.54.25.265.490.1000.1072.1 × 104−1.5D810.072.56.65.165.490.1490.0560.7 × 104−0.4D911.896.06.95.465.600.0570.0490.6 × 104−0.03D1024.734.17.55.345.370.0140.00818 × 104−0.006D1130.203.46.15.425.440.0090.0052.5 × 104−0.006D1252.122.34.65.65.620.0080.0045.8 × 104−0.008W and L represent the nominal width and nominal length of the NbSe2 samples with a thickness of d. Qi,Al-NbSe2 and Kerr coefficients (K/2π) are extracted from NbSe2-terminated Al resonators. The film thickness d is measured with atomic force microscopy. Summary of kinetic microwave characterization of NbSe2 samples across different thicknesses W and L represent the nominal width and nominal length of the NbSe2 samples with a thickness of d. Qi,Al-NbSe2 and Kerr coefficients (K/2π) are extracted from NbSe2-terminated Al resonators. The film thickness d is measured with atomic force microscopy. The resonant frequency of the NbSe2-terminated resonator decreases with increasing microwave probe power, e.g., with a noticeable downshift in fr,Al-NbSe2 at the estimated sample power of −90 dBm (Fig. 3b, green arrow). This frequency downshift indicates a contribution from kinetic inductance. As the microwave power increases, enhanced Cooper-pair breaking leads to a reduction in the superfluid density (n0), resulting in a larger kinetic inductance Lk and, consequently, a lower resonant frequency. In contrast, the aluminum resonator exhibits no observable power dependence up to an estimated input power of −60 dBm at the sample (see Supplementary Fig. 4), corresponding to the upper limit of our VNA output (VNA outputs a power of 10 dBm at room temperature, which is reduced to −60 dBm at the sample after passing through a total attenuation of 70 dB), confirming that its inductance is predominantly geometric. The nonlinear microwave power dependence exhibited by the NbSe2-terminated resonator, observed over the power range from −100 to −80 dBm, is well-described by a Kerr-type or Duffing-type nonlinearity57 (see Supplementary Note 3). When the microwave power in the resonator exceeds a critical threshold, the resonant frequency bifurcates and enters a bistability regime (Fig. 3b, blue arrow)58–60. From a linear fit of the resonant frequency shift (Δf=fr,Al-NbSe2(nr)−fr,Al-NbSe2(nr=1)) with the photon number of the resonator (nr), shown in Fig. 3c, we extract a self-Kerr coefficient K/2π = −1.7 Hz/photon at 10 mK. Typically, conventional superconducting inductors of similar dimensions, such as granular aluminum, with a self-Kerr coefficient of 4.5 MHz61, and Josephson junction arrays, with a self-Kerr of 8 MHz62–exhibit significantly higher nonlinearity. In contrast, NbSe2 films show a lower self-Kerr coefficient, indicating low nonlinearity. We also investigate the dependence of the kinetic inductance on the DC bias current. A DC bias current IDC is introduced via a contact to the microwave line just above the NbSe2 termination (Fig. 3d), while a microwave tone is maintained in the low-power linear regime (Prf = −120 dBm). The resonant frequency (fr,Al-NbSe2) decreases with an increasing bias current (up to IDC = 15 μA), indicating a corresponding increase in kinetic inductance (Fig. 3e). This behavior is similar to the microwave response due to the Kerr-type nonlinearity. In the Ginzburg–Landau framework, the dependence of kinetic inductance of the superconducting film with bias current is well-described by a quadratic function (ignoring the terms beyond 2nd order)35: where Lk(0) is the kinetic inductance of the resonator at zero current bias, and I* is a characteristic current, also known as the depairing current. I* is typically comparable to the superconducting critical current Ic, and sets the scale of non-linearity. By fitting to this relationship, I* is determined to be 0.4 mA for the 9 nm thick, 3.9 μm wide NbSe2 sample. We observe that both the resonant frequency and the kinetic inductance per square exhibit a quadratic dependence on the DC bias current (Fig. 3f), consistent with the fully gapped pairing of NbSe2 reported in previous studies46,63. Discussion VdW superconductors remain superconducting in the monolayer limit39,64, which enables one to increase kinetic inductance by reducing the thickness. Figure 4a shows the sheet kinetic inductance (Lk,sq) of NbSe2 films extracted from microwave measurements as a function of thickness (d) and the corresponding layer numbers. The measured Lk,sq (red circles) exhibits an approximate 1/d dependence, as expected for thin superconducting films, and reaches 1.2 nH for the monolayer sample (see Supplementary Note 6 for the determination of layer numbers). The values are compared to those of other intrinsic superinductors based on amorphous materials, including TiN22,65, NbTiN26,27, granular aluminum (gr-Al)61, and NbN24 across a range of film thicknesses up to 13 nm, as shown in Fig. 4b. Voltage-induced flat-band superconductors, such as magic-angle twisted bilayer4,66 and trilayer graphene5,67, exhibit a wide range of inductances depending on bias voltages and are therefore not included in Fig. 4b. Thickness dependence of kinetic inductance in NbSe2. a Sheet kinetic inductance (Lk,sq) from microwave measurements (circle-red) of nine NbSe2 devices with varying film thicknesses. The blue dashed line shows a 1/d scaling fit. b Comparison of measured Lk,sq in NbSe2 with values reported for other high-impedance intrinsic superconducting thin films22,24,26,27,61,65. c Sheet kinetic inductance (Lk,sq) from microwave measurements is plotted as a function of Rs/Tc, where Rs and Tc values are obtained from both our transport measurements (red circles) and previous studies (green and black circles)41,42. The blue dashed line is a linear fit with a slope of 16.3 and a y-intercept of 80.0 pH. The clean-limit contribution (α/d=meffn0e2d) of total kinetic inductance for each data point is plotted (purple hexagons) using an estimate of α. d The magnified view of the lower region for clarity. The error bars in the plots include contributions from both geometric uncertainty of the device and the uncertainty in the resonance frequencies extracted from circle fits to the complex S21 data. The observed layer dependence may provide insight into the underlying mechanisms responsible for the high Lk, including the role of scattering processes, and whether our NbSe2 devices operate in the dirty limit, the clean limit, or an intermediate regime. We consider a general expression for the sheet kinetic inductance Lk,sq that spans from the clean to the dirty limit (see Supplementary Note 4 for the derivation): Here, ρs denotes the normal-state resistivity, ℏ is the reduced Planck’s constant, kB is the Boltzmann constant, and Rs is the normal-state sheet resistance. If we assume that the term in parentheses varies weakly with film thickness, we expect the approximate 1/d dependence observed in Fig. 4a. This expression suggests that the total kinetic inductance arises from a combination of clean-limit (first term) and dirty-limit (second term) contributions. Figure 4c plots the measured sheet kinetic inductance Lk,sq and clean-limit contribution (the first term of Eq. (2)) as a function of Rs/Tc. The Rs and Tc values are obtained from both our transport measurements and previous studies41,42 (shown in green and black circles in Fig. 4c and summarized in Supplementary Table 3). We estimate α =meffn0e2 for the clean-limit contribution, assuming α is independent of disorder and film thickness (see Supplementary Fig. 5). We observe a linear dependence for the larger values of Lk,sq in Fig. 4c. A linear fit yields a slope β = 16.3, which captures the dirty-limit contribution. In contrast, the α/d term for each data point reflects the clean-limit component, which approaches the total Lk,sq in thicker samples. As the sample thickness approaches the monolayer limit, the ratio of the measured kinetic inductance to the clean-limit kinetic inductance (Lk,sq/Lk,sq,clean) increases from 0.7 to 15 monotonically, indicating that the dirty-limit contribution becomes increasingly significant in thinner NbSe2 samples. This model captures the crossover from clean to the dirty-limit behavior as the sample thickness decreases. We now discuss possible origins of the observed crossover behavior. In thicker NbSe2 films, the majority of charge carriers reside in the interior layers of the 2D crystal. The clean-limit-like behavior observed in this regime—corresponding to the lower-left corner of Fig. 4c—may be attributed to the high crystallinity of the bulk material and effective screening from carriers in the top and bottom NbSe2 layers. In contrast, thinner devices exhibit a greater influence from their surrounding environment. In the monolayer limit, all charge carriers are in close proximity to the encapsulating hBN on both sides, leading to enhanced surface scattering from impurities or fabrication residues. This increased scattering may result in more pronounced dirty-limit contributions to the kinetic inductance. Moreover, NbSe2 has been shown to host two superconducting bands with distinct gap sizes, with the larger gap decreasing monotonically as thickness is reduced63. In other multi-band superconductors68,69, such disparity across the Fermi surface has been linked to the coexistence of clean- and dirty-limit behaviors, arising from band-dependent coherence lengths and scattering rates. Our experiment does not explicitly resolve contributions from individual bands, but similar multi-band effects may contribute to the observed layer dependence of kinetic inductance in NbSe2 (Fig. 4c). Additionally, as the superconducting gap (or Tc) decreases with reduced thickness, the superconducting coherence length becomes longer. For a fixed mean free path, this leads to a larger coherence length to mean free path ratio, indicating a shift toward dirty-limit behavior in the thinner samples. We note that the combined contributions from both clean- and dirty-limit superconductivity observed in the microwave measurements here contrast with previous DC transport studies, which reported only clean-limit behavior. This discrepancy suggests that the microwave response may be sensitive to the multi-band nature of NbSe241,49–51. Finally, in thin superconducting films where the thickness d is much smaller than the penetration depth λ, the relevant length scale becomes the Pearl length Λ=2λ2d70. All of our NbSe2 samples operate in this Pearl regime (d ≪ λ ~ 230 nm71), with both thickness and width much smaller than the estimated Pearl length (10–70 μm; see Supplementary Note 5). Recent studies on high-Tc YBCO superconductors have shown that geometric confinement in this regime can strongly enhance the kinetic inductance as the device width approaches or falls below Λ72. While a detailed investigation of this effect in NbSe2 lies beyond the scope of this work, our observation of large Lk values—and a steeper-than-expected inverse thickness scaling, relative to the prediction of Eq. (2)—may be qualitatively consistent with such confinement-enhanced kinetic inductance in the Pearl regime. We characterize the kinetic inductance and nonlinear properties of NbSe2 using a coplanar waveguide resonator terminated by hBN-encapsulated NbSe2 thin films. The kinetic inductance shows an approximate 1/d-type scaling for the thinner samples and reaches 1.2nH/□ at monolayer thickness—among the highest reported for superconductors within this thickness regime. Under the microwave drive, the device exhibits a nonlinear frequency shift that is consistent with the nonlinear Kerr model. The extracted Kerr coefficients (K/2π = −0.006 to 14.7 Hz/photon) are relatively small, making NbSe2 well-suited for applications that require large linear inductance with minimal nonlinearity, such as the linear shunt-inductor used in fluxonium qubits and in photon detectors operating in the microwave regime73. The Lk exhibits a quadratic dependence on DC bias current, consistent with the Ginzburg-Landau framework. The measured kinetic inductance in NbSe2 exhibits a crossover from clean- to dirty-limit behavior as the thickness decreases. This transition can be qualitatively described using a generalized expression for Lk derived from BCS theory. In the ultra-thin regime, enhanced contributions from the dirty limit may arise due to increased surface scattering, multi-band superconductivity, and geometric confinement. We demonstrate that NbSe2 is a crystalline two-dimensional superconductor that exhibits large, linear inductance. The hBN–NbSe2–hBN heterostructure is well suited for compact lump-element devices, such as shunt inductors for fluxonium qubits or readout resonators, while minimizing parasitic coupling. These results also highlight the ability of microwave measurements to probe superconducting properties in low-dimensional materials, offering a valuable complement to DC transport and other techniques. Methods Device assembly and fabrication Aluminum CPW resonators are fabricated from a highly resistive 2-inch Si wafer coated with 250 nm MBE-grown aluminum film. The resonator circuit is patterned via photolithography and wet etched using Transene Aluminum etchant. Afterward, the 2-inch wafer is diced into 5 mm × 5 mm chips. The bulk crystal of NbSe2 is grown by HQ Graphene. The mechanical exfoliation and encapsulation of thin NbSe2 layers with hBN are carried out inside an Argon-filled glove box, maintaining oxygen and water levels below 0.1 ppm. This prevents oxidation of NbSe2, keeping the devices within the low-disorder regime. The hBN–NbSe2–hBN heterostructures are transferred on the chip using dry polymer-based techniques74–76. After transferring the hBN–NbSe2–hBN stacks onto the pre-patterned chips, the van der Waals heterostructures are taken out of the glove box. To process edge-contact, first, the stack goes through reactive-ion etching with O2/Ar/CHF3 to open edges of NbSe2. The samples are immediately loaded into an e-beam evaporator to do in-situ ion-milling to clean the exposed edges. Aluminum is then deposited onto the cleaned NbSe2 edge at a tilt angle of 30∘ with the substrate rotating at 60 rpm, ensuring uniform coverage55. The bridging between the contact aluminum and the pre-fabricated aluminum resonator and ground plane is done by another in-situ ion mill and a normal-incidence aluminum deposition77. Measurement setup The experiment was performed using a Bluefors XLD1000 dilution refrigerator with a base electron temperature of 10 mK. To send the microwave signals down to the package, we use semi-rigid coaxial cables with low thermal conductivity and low-loss superconducting niobium–titanium (NbTi) cables for the readout. Within our readout input lines, we have integrated attenuation mechanisms: a 20 dB attenuator at 4 K, a 10 dB attenuator at the still plate, and a 40 dB attenuator at the mixing chamber stage. The output signals pass through a 2 GHz high-pass filter and a 12 GHz low-pass filter. The readout chain incorporates a Josephson traveling-wave parametric amplifier (TWPA), whose pump tone is combined with the readout signal using a directional coupler. The output signal is subsequently amplified by a low-noise high-electron-mobility transistor (HEMT) amplifier at the 4 K stage, followed by room-temperature amplification. DC lines connected to the device are filtered at 4 K with an RC π-filter and an RF + RC filter with a cut-off frequency of 50 kHz at the mixing chamber stage. DC transport measurements were performed using standard low-frequency lock-in techniques. A typical AC excitation current of 1 nA at 17.77 Hz was applied to the sample, while the resulting AC voltage drop was measured simultaneously. A DC bias current was applied using a multichannel low-noise voltage source (QDAC from Qdevil). Microwave transmission measurements were performed using a Keysight N5232A VNA to send and receive the microwave signal. The control electronics used in this experiment are listed in Supplementary Table 2. Extraction of kinetic inductance from resonance frequency To extract the kinetic inductance (Lk) of NbSe2, we employ a comparative model using λ/4-resonators terminated with aluminum and NbSe2. For the aluminum-terminated resonator with a length of l and a resonant frequency of fr,Al, the input impedance ZAl is expressed as78 where Z0 is the characteristic impedance which is 50 Ω in our work, α and β are the attenuation constant and phase constant respectively. For a lossless line, the attenuation constant α is zero, and β(f) becomes For the λ/4-resonator terminated by NbSe2, the kinetic inductance (Lk) contributes an additional frequency-dependent term in the impedance, which is modeled as The physical length l of the resonator is determined from the aluminum reference resonator using Eq. (3), assuming identical geometry and effective permittivity. On resonance, the impedance of a λ/4-resonator diverges; for the NbSe2-terminated case, Eq. (5) reduces to We extract the kinetic inductance Lk of NbSe2 from Eq. (6), where the resonant frequency of NbSe2-terminated resonator is in the linear regime (in the single-photon limit). Supplementary information Supplementary Information Transparent Peer Review file