利用太赫兹时域光谱（THz-TDS）和傅里叶变换远红外光谱（FT-Far-IR）技术研究了大米中西维因在太赫兹频段的吸收光谱特征, 并结合化学计量学方法对大米中西维因进行了测定。 样品的制备采取待测农药西维因与大米粉末混合压片的方法模拟真实检测情景, 无需样品的分离富集。 分别将样品在1.8～6.3 THz特征波段内的Far-IR吸收光谱数据和在0.5～1.5 THz特征波段内的THz-TDS吸收光谱数据随机划分为训练集和验证集。 采用偏最小二乘回归（PLSR）方法建立定量分析模型, 将校正均方根误差（RMSECV）、 预测均方根误差（RMSEP）、 预测相关系数（Rv）作为模型性能评判的依据, RMSECV, RMSEP越小, Rv越大, 则所建立的模型越好。 两种检测技术均得到较好的结果。 其中, 运用Far-IR技术所得数据建立的定量分析模型预测相关系数(Rv)为0.99, 校正均方根误差（RMSECV）为0.007 7, 预测均方根误差（RMSEP）为0.008 6; 运用THz-TDS技术所得数据建立的定量分析模型预测相关系数为0.98, 校正均方根误差（RMSECV）为0.002 5、 预测均方根误差（RMSEP）为0.004 4。 该研究为定量检测粮食中的农残提供了一种新方法。

Determination of carbaryl in rice by using Fourier transform far-infrared (FT- Far-IR) and terahertz time-domain spectroscopy (THz-TDS) combined with chemometrics was studied and the spectral characteristics of carbaryl in terahertz region was investigated. Samples were prepared by mixing carbaryl at different amounts with rice powder, and then a 13 mm diameter, and about 1 mm thick pellet with polyethylene (PE) as matrix was compressed under the pressure of 5～7 tons. Terahertz time domain spectra of the pellets were measured at 0.5～1.5 THz, and the absorption spectra at 1.8～6.3 THz were acquired with Fourier transform far-IR spectroscopy. The method of sample preparation is so simple that it does not need separation and enrichment. The absorption peaks in the frequency range of 1.8～6.3 THz have been found at 3.2 and 5.2 THz by Far-IR. There are several weak absorption peaks in the range of 0.5～1.5 THz by THz-TDS. These two kinds of characteristic absorption spectra were randomly divided into calibration set and prediction set by leave-N-out cross-validation, respectively. Finally, the partial least squares regression (PLSR) method was used to establish two quantitative analysis models. The root mean square error (RMSECV), the root mean square errors of prediction (RMSEP) and the correlation coefficient of the prediction are used as a basis for the model of performance evaluation. For the Rv, a higher value is better; for the RMSEC and RMSEP, lower is better. The obtained results demonstrated that the predictive accuracy of the two models with PLSR method were satisfactory. For the FT-Far-IR model, the correlation between actual and predicted values of prediction samples (Rv) was 0.99. The root mean square error of prediction set (RMSEP) was 0.008 6, and for calibration set (RMSECV) was 0.007 7. For the THz-TDS model, Rv was 0.98, RMSEP was 0.004 4, and RMSECV was 0.002 5. Results proved that the technology of FT-Far-IR and THz-TDS can be a feasible tool for quantitative determination of carbaryl in rice. This paper provides a new method for the quantitative determination pesticide in other grain samples.