LAMOST-Kepler LK-MRS: Medium-Resolution Survey

• The paper demonstrates that repeated medium-resolution (R~7500) spectra for over 50,000 stars significantly enhance radial velocity precision and atmospheric parameter accuracy.
• The LK-MRS methodology combines dual-arm spectral coverage with advanced data pipelines like LASP and LAMA to enable time-series analysis for binaries, pulsators, and activity diagnostics.
• The survey’s calibrated data products and multi-epoch observations offer actionable insights for exoplanet host characterization, stellar variability studies, and Galactic archaeology.

Searching arXiv for recent LK-MRS papers and related LAMOST MRS methodology. The LAMOST-Kepler/K2 Medium-Resolution Spectroscopic Survey (LK-MRS) is the medium-resolution, time-domain spectroscopic component of the broader LAMOST–Kepler/K2 effort, designed to obtain repeated $R \sim 7500$ spectra for stars in the original Kepler field and selected K2 campaign fields. In the literature it is identified as “Phase II of the LAMOST-Kepler/K2 survey” and is explicitly framed as a program that combines LAMOST’s multiplexed spectroscopy with Kepler, K2, and, in later work, TESS photometry to deliver atmospheric parameters, radial velocities, rotational information, and variability diagnostics for stellar astrophysics, binary-star studies, exoplanet host characterization, and Galactic archaeology (Zong et al., 2020).

1. Definition, institutional setting, and survey scope

LK-MRS is embedded in the LAMOST Medium-Resolution Spectroscopic Survey (MRS), which began after testing and entered full survey operation in September 2018. LAMOST itself is a 4-m-class, 4000-fiber spectroscopic facility with a 5°-diameter field of view, and its MRS mode operates at resolving power $R \approx 7500$ with two simultaneous wavelength windows: 495–535 nm in the blue arm and 630–680 nm in the red arm (Yan et al., 2022). Within this framework, LK-MRS targets stars in the Kepler prime field and multiple K2 fields through repeated observations rather than the broad one-pass coverage characteristic of earlier low-resolution LK work.

The program is explicitly described as “Phase II / LK-MRS,” initiated in 2018 as a parallel project to the low-resolution LAMOST–Kepler survey. Its science drivers are binary stars, high-amplitude pulsators, active stars, and improved atmospheric parameters for selected Kepler targets. The survey plan in its first formal description aimed at collecting medium-resolution spectra for more than 50,000 stars with multiple visits, about 60 epochs, over the period from September 2018 to June 2023 (Fu et al., 2020).

The footprint design evolved across publications. One early survey description states that LK-MRS selected 20 footprints distributed across the Kepler field and six K2 campaigns, with each plate containing about 2,000–3,000 stars (Zong et al., 2020). A later five-year synthesis describes LK-MRS-I as covering 20 LAMOST plates in the Kepler and K2 fields from 2018 to 2023, with 4 plates in the original Kepler field and 16 in K2 fields; by June 2023, 16 of those 20 plates had been observed (Qin et al., 26 Jul 2025). This suggests a progression from initial design to realized execution rather than a contradiction in the survey concept.

Within the broader LAMOST context, LK-MRS sits alongside the earlier low-resolution LAMOST–Kepler project, which had already collected very large homogeneous spectroscopic samples in the Kepler field. That Phase I low-resolution effort provided the atmospheric-parameter and target-selection backbone that motivated a medium-resolution, time-domain extension focused on radial-velocity precision, multi-epoch monitoring, and improved chemical and rotational diagnostics (Fu et al., 2020).

2. Instrumental realization and observing strategy

LK-MRS uses the standard LAMOST MRS instrumental configuration: $R \sim 7500$, blue-arm coverage 495–535 nm, and red-arm coverage 630–680 nm. The blue interval contains the Mg b triplet region and many Fe lines, while the red interval contains H$\alpha$ and additional metal lines. These windows were selected precisely because they are well suited to radial velocities, atmospheric-parameter estimation, line-profile work, and selected abundance measurements (Yan et al., 2022).

The observing strategy is explicitly time-domain. In the Kepler field, four central LK plates—K1a1, K1a2, K1a3, and K1a4—were assigned to LK-MRS, containing about 12,000 stars down to $g \sim 15.5$ mag, with approximately 75% of those stars observed by Kepler photometrically (Fu et al., 2020). The observing cadence is tied to LAMOST bright-night scheduling. One survey description states that LK-MRS is scheduled during bright nights in each lunar month and that each central Kepler plate is intended to be observed about 60 times over five years, with an automatic Python scheduler choosing the highest-priority plate in a session and continuing observations until it leaves the observable window (Fu et al., 2020).

The same time-domain logic appears in the more general MRS planning. LAMOST-MRS as a whole was designed to provide 60-epoch observations for about 200,000 stars, and LK-MRS is a dedicated subset of that broader time-domain architecture (Yan et al., 2022). A practical implication is that LK-MRS inherits both the multiplexing efficiency of MRS and its exposure-level systematics, calibration conventions, and parameter pipelines.

By June 2019, the early Kepler-plate implementation had already accumulated 39 visits on K1a1 and 4 visits on K1a2, yielding 76,921 medium-resolution spectra for 4,578 stars, with 71,914 atmospheric/RV parameter sets for 3,981 stars (Fu et al., 2020). Over the full 2018–2023 interval, LK-MRS-I accumulated 3,531,785 spectra in total, including 2,941,929 single-exposure spectra and 589,856 co-added spectra, corresponding to 49,310 unique stars; 36,588 stars had at least one spectrum with derived parameters (Qin et al., 26 Jul 2025). This later result establishes LK-MRS as a large time-series spectroscopic archive rather than a small pilot program.

3. Parameter pipelines, reduction, and calibrated data products

LK-MRS data are reduced with the standard LAMOST 2D and 1D pipelines, adapted to the medium-resolution dual-arm setup. The reduction sequence includes bias subtraction, flat-fielding, fiber extraction, wavelength calibration, sky subtraction, and flux calibration to produce one-dimensional spectra (Fu et al., 2020). For the main survey products, parameter estimation is built around LASP, the LAMOST Stellar Parameter Pipeline, which derives $T_\mathrm{eff}$, $\log g$, $[\mathrm{Fe/H}]$, and RV from medium-resolution spectra through template matching and cross-correlation (Fu et al., 2020).

In later MRS work, a dedicated pipeline named LAMA was developed specifically for LAMOST medium-resolution spectral analysis. LAMA estimates $T_\mathrm{eff}$, $\log g$, $R \approx 7500$0, radial velocity, and $R \approx 7500$1 from MRS spectra using template matching against synthetic spectra based on ATLAS9 atmospheres and SPECTRUM. It uses spectra only when both arms satisfy S/N $R \approx 7500$2, removes clear double- and triple-line systems before parameter fitting, models per-fiber and per-night resolving power, and calibrates the resulting parameter scale against GALAH, with external comparisons to APOGEE and PASTEL (Li et al., 2024). Because LK-MRS uses the same MRS hardware and spectral format, this pipeline is directly applicable to LK-MRS spectra even when a paper does not isolate the Kepler/K2 subset.

The LK-MRS literature also includes value-added analyses based on ROTFIT, especially for late-type stars in the Kepler field. ROTFIT uses empirical ELODIE templates degraded to $R \approx 7500$3 to derive $R \approx 7500$4, $R \approx 7500$5, $R \approx 7500$6, RV, and $R \approx 7500$7, and then uses photospheric subtraction to measure H$R \approx 7500$8 and Li I 6708 equivalent widths. In one such medium-resolution Kepler-field analysis, 16,300 spectra were processed, producing atmospheric parameters and RVs for 14,300 spectra of 7,443 stars, together with activity and lithium diagnostics (Frasca et al., 2022). This suggests a layered data ecosystem in which LASP and LAMA provide survey-homogeneous parameter baselines, while specialized pipelines extract targeted diagnostics for subsets of the LK-MRS sample.

The principal LK-MRS-I star-level catalog contains, for each star, the number of spectra contributing to the solution, an identifier, KIC or EPIC cross-match where available, coordinates, and weighted-average values and uncertainties for $R \approx 7500$9, $R \sim 7500$0, $R \sim 7500$1, $R \sim 7500$2, RV, and $R \sim 7500$3, plus labels for metal-poor stars, high-velocity stars, RV-variable stars, and specific photometric variability classes (Qin et al., 26 Jul 2025). A plausible implication is that LK-MRS should be understood not only as an observing program but also as a multi-layer catalog framework in which time-domain spectroscopy is progressively converted into survey-grade stellar characterization.

4. Precision, validation, and radial-velocity calibration

The internal precision of LK-MRS parameters is strongly S/N-dependent. In the first-year medium-resolution release, the internal uncertainties at S/N = 10 were reported as about 100 K in $R \sim 7500$4, 0.15 dex in $R \sim 7500$5, 0.09 dex in $R \sim 7500$6, and 1.00 km s$R \sim 7500$7 in RV (Zong et al., 2020). At S/N $R \sim 7500$8, the early Kepler-field analysis reported internal uncertainties approaching $R \sim 7500$9 K, $\alpha$0 dex, $\alpha$1 dex, and $\alpha$2 km s$\alpha$3, respectively (Fu et al., 2020). Over the full 2018–2023 LK-MRS-I release, the survey quoted 120 K, 0.18 dex, 0.13 dex, 0.08 dex, 1.9 km s$\alpha$4, and 4.0 km s$\alpha$5 at S/N = 10 for $\alpha$6, $\alpha$7, $\alpha$8, $\alpha$9, RV, and $g \sim 15.5$0, respectively (Qin et al., 26 Jul 2025). The differences between these publications reflect changes in sample definition, fitting function, included parameters, and five-year rather than one-year data aggregation.

External validation has been carried out against LAMOST low-resolution data, APOGEE, GALAH, and Gaia. The first-year LK-MRS release found that parameters derived from LK-MRS spectra are in general consistent with those from LRS and APOGEE, though the scatter increases as surface gravity decreases in APOGEE comparisons, and a large discrepancy appears between LK-MRS and Gaia effective temperatures (Zong et al., 2020). The five-year LK-MRS-I paper reports that $g \sim 15.5$1 and $g \sim 15.5$2 show good agreement with APOGEE and GALAH, whereas $g \sim 15.5$3 and especially $g \sim 15.5$4 display more noticeable systematic compression relative to high-resolution optical and infrared scales (Qin et al., 26 Jul 2025). LAMA likewise reports no obvious bias in comparisons with APOGEE, GALAH, and PASTEL, and after calibration to GALAH it achieves uncertainty peaks around 75 K in $g \sim 15.5$5, 0.12 dex in $g \sim 15.5$6, and about 0.17 dex in $g \sim 15.5$7 for dwarfs when S/N $g \sim 15.5$8 (Li et al., 2024).

Radial-velocity calibration is a particularly important aspect of LK-MRS because the survey is explicitly time-domain. Pathfinder work on a Kepler-field MRS pointing showed that uncorrected medium-resolution RVs suffer from small but non-negligible spectrograph- and exposure-dependent zero-point shifts; after correcting those systematics, the precision reached about 1.3, 1.0, 0.5, and 0.3 km s$g \sim 15.5$9 at $T_\mathrm{eff}$0 and 100, respectively (Liu et al., 2019). Subsequent work on the relative calibration of LAMOST MRS RVs formalized an exposure-by-exposure correction based on multi-epoch constant stars observed with the same spectrographs. That method reduces the relative RV zero-point scatter to below 0.5 km s$T_\mathrm{eff}$1 and identifies 65,123 constant-star candidates that can function as secondary RV standards (Xiong et al., 2021).

These calibration results are directly relevant to LK-MRS. They imply that the single-epoch RV precision quoted in the survey products and the relative epoch-to-epoch stability needed for binary and pulsation work are not independent properties: both rely on a calibration framework that treats the spectrograph and exposure as explicit sources of RV systematics. This suggests that the scientific value of LK-MRS depends at least as much on calibration design as on nominal resolving power.

5. Scientific applications in exoplanets, asteroseismology, activity, and binaries

LK-MRS inherits and extends the scientific agenda of the LAMOST–Kepler low-resolution survey. In exoplanet work, the earlier LK project had already enabled studies of host-star metallicities, eccentricity architectures, and planetary-system occurrence statistics through homogeneous stellar parameters in the Kepler field. Examples summarized in the LAMOST overview include the metallicity dependence of close-in Neptune-size planets (“Hoptunes”), the distinction between single- and multiple-transiting systems in mean eccentricity, and estimates that about 30% of Sun-like stars host “Kepler-like” planetary systems (Yan et al., 2022). LK-MRS is designed to refine such work by providing more precise RVs, improved atmospheric parameters, and a direct handle on host-star binarity.

In asteroseismology, LK-MRS is particularly well aligned with Kepler’s original strengths. The LAMOST–Kepler literature had already used low-resolution spectra combined with Kepler seismology to calibrate $T_\mathrm{eff}$2, derive masses and ages, and distinguish evolutionary phases in giant stars (Yan et al., 2022). Medium-resolution follow-up strengthens this program through higher-precision $T_\mathrm{eff}$3, $T_\mathrm{eff}$4, and metallicity, and through time-series RVs for pulsators. The first LK-MRS overview explicitly cites RR Lyrae stars, classical and type II Cepheids, and high-amplitude $T_\mathrm{eff}$5 Scuti stars as prime targets for time-series spectroscopy intended to constrain pulsation modes and support seismic modeling (Fu et al., 2020).

For stellar activity, the red arm of LK-MRS provides H$T_\mathrm{eff}$6, which the survey papers identify as a key chromospheric indicator. Medium-resolution spectra enable measurement of the strength of H$T_\mathrm{eff}$7 chromospheric emission for stars with lower levels of magnetic activity, and the time-domain design allows that quantity to be monitored over repeated visits (Fu et al., 2020). In the ROTFIT-based medium-resolution Kepler analysis, 327 active stars were identified from H$T_\mathrm{eff}$8 fluxes, and the same study linked those spectroscopic diagnostics to Kepler rotation periods and lithium content (Frasca et al., 2022). This suggests that LK-MRS is not merely a parameter survey but also a framework for synoptic stellar magnetic diagnostics.

Binary-star science is perhaps the clearest native application of LK-MRS. The survey was explicitly created to discover non-eclipsing binaries through RV variability, derive orbital properties in combination with Kepler light curves, and characterize high-amplitude pulsators and heartbeat-like systems (Fu et al., 2020). A tentative simulation cited in the early overview, based on K1a1 observations, indicated that more than 10% of stars in one sample exhibit significant RV variability consistent with binarity, with about 200 binaries among about 1,900 stars—roughly an order of magnitude larger than the eclipsing-binary fraction in that subset (Fu et al., 2020). This makes clear that the time-domain spectroscopic layer reveals a large hidden binary population inaccessible to photometry alone.

6. Spectroscopic multiplicity, peculiar stars, and legacy value

A major downstream use of MRS data, directly relevant to LK-MRS, is systematic mining for double-line and triple-line spectroscopic systems. A convolutional-neural-network approach trained on synthetic MRS spectra identified 2,198 SB2 candidates in more than 5 million MRS spectra from DR8, with the method favoring FGK main-sequence binaries with mass ratio $T_\mathrm{eff}$9 and radial-velocity separation $\log g$0 (Zhang et al., 2021). A later human–AI hybrid method, combining CCF analysis and machine learning on MRS DR9, identified 7,096 SB2 candidates and 1,903 SB3 candidates from 6,565,721 selected blue-arm spectra, with 70.1% of SB2s and 89.6% of SB3s newly identified (Li et al., 2024). Those studies are not restricted to the Kepler/K2 fields, but because LK-MRS is part of the same MRS data stream, their catalogs implicitly include LK-MRS targets.

Peculiar-star discovery is also part of LK-MRS output. The five-year LK-MRS-I release identifies 764 metal-poor stars, 174 very metal-poor stars, and 30 high-velocity stars, and finds 2,333 stars whose radial velocity appears variable; 371 of those RV-variable stars are confirmed as periodic variables using Kepler/K2 or TESS photometry and classified by variability type (Qin et al., 26 Jul 2025). This extends an earlier low-resolution tradition in which the LK survey had already identified candidate metal-poor, very metal-poor, and high-velocity stars in the Kepler field (Ren et al., 2016).

The survey’s value for hot and peculiar stars should not be understated, even though many medium-resolution parameter pipelines are optimized for FGK stars. LAMOST–Kepler work on misclassified B stars had already demonstrated that spectroscopic follow-up in the Kepler field can correct severe KIC temperature and gravity biases, reclassifying stars and transforming their asteroseismic interpretation (Zhang et al., 2018). A plausible implication is that LK-MRS, despite wavelength-coverage and pipeline constraints, contributes to this same corrective role by providing repeated medium-resolution spectra that can be reanalyzed with specialized methods for hot stars, subdwarfs, and white dwarfs.

As a legacy resource, LK-MRS is best understood as a spectroscopic time-series infrastructure for the Kepler and K2 fields. Its final products are not limited to one catalog or one pipeline scale. Instead, the survey supports a layered ecosystem: survey-homogeneous parameters from LASP and LAMA, specialized re-analyses such as ROTFIT for late-type stars, RV-calibration frameworks for time-domain studies, and multiplicity catalogs derived with machine learning and CCF methods. This suggests that the enduring significance of LK-MRS lies not only in its own catalog releases but also in the way it enables heterogeneous, reusable spectroscopic products to be combined with Kepler, K2, Gaia, APOGEE, GALAH, and TESS across multiple subfields of stellar and Galactic astrophysics.