选取环境中含量最多的9种多溴联苯醚（PBDEs）及已知具有高/中等生物毒性的14种PBDEs作为目标辨识物, 首先基于振动归属将上述PBDEs红外光谱分为6个谱区, 利用相似度分析评价各谱区对23种同系物的辨识能力并筛选易于辨识的红外谱区, 其次根据各谱区特征建立辨识方法并进行方法评价, 最终通过主成分分析及判别分析法在所研究谱区提取影响高/中等生物毒性PBDEs的红外特征信息并建立生物毒性判别模型。 研究结果表明, 位于961～1 153 cm-1的三角呼吸振动谱区和1 153～1 346 cm-1的C—O伸缩振动谱区对所选PBDEs间光谱相似度最小, 可利用三角呼吸振动谱区内吸收峰的数量分布特征（辨识最小频差为3.19 cm-1, 但不易与多氯联苯醚等结构相似物质间的区分）及C—O伸缩振动谱区的主峰分布及波形特征（辨识最小频差为0.74 cm-1, 与多氯联苯醚等结构相似物质间辨识涉及最小频差为0.67 cm-1）实现对23种PBDEs间的辨识（红外光谱常用最小分辨率为0.5 cm-1）; 基于判别分析法所构建的高/中等生物毒性PBDEs判别模型的判别正确度为100%/88.9%, 表明所提取的红外光谱信息可实现对PBDEs生物毒性的准确分类。

In this paper, 9 most abundant PBDE congeners in environment and 14 congeners with high/medium biological toxicity were selected as target compounds. Firstly, IR spectra of 23 compounds were divided into 6 spectrum regions based on known vibrational assignment, and their identification abilities for PBDEs were evaluated by using similarity analysis. Then, the spectrum regions with high identification ability relatively have been analyzed to search identification method on the basis of each spectrum characteristic. At last, the principal component analysis and discriminant analysis method were used to extract characteristic spectrum information of congeners with high/medium biological toxicity and establish biological toxicity discriminant model. The triangle breathing vibration region (961～1 153 cm-1) and C—O stretching vibration region (1 153～1 346 cm-1) has been filtered as characteristic IR spectrum regions because of smaller similarity among PBDEs relatively. For the former, the information of vibration peaks distribution can be taken as tool to identification, with the smallest frequency difference of 3.19 cm-1 among 23 congeners and bad ability to distinguish PBDEs from dioxin-similar compounds. For the later, the frequency and shape of main peak is as mark and the smallest frequency difference among PBDEs and between PBDEs with dioxin-similar compounds are 0.74 and 0.67 cm-1 separately, both larger than the common minimum resolution of IR spectrum 0.5 cm-1, depicting good ability for both inner and outer identification of PBDEs. The established biological toxicity discriminant model via discriminant analysis also have shown well ability to biological toxicity prediction, with accuracy of 100% and 88.9% for high/medium and low biological toxicity relatively, representing the extracted IR spectrum information from congeners with high/medium toxicity can realize the accurate classification of biological toxicity for PBDEs.