Disentangling techniques are often needed to obtain the spectra of the individual components of binary or multiple systems. A thorough analysis of the shift-and-add algorithm of Marchenko, Moffat, & Eenens (1998) reveals that in many cases the line fluxes are poorly reproduced and spurious wings appear. The causes of these discrepancies are discussed and a new disentangling package, QER20, is presented which significantly reduces these errors and vastly increases the performance. When applied to the massive binary 9 Sgr, our new code yields line fluxes which are notably different from those previously published and lead us to revise the spectral classification to slightly earlier subtypes: O3V((f +)) for the primary and O5V((f)) for the secondary. We show that with the MME98 algorithm the classification of massive stars in binaries can be off by several subtypes whilst there are no such errors when the QER20 package is used.

中文翻译：

---

重新审视大规模二进制系统 9 Sgr：对解开方法的新见解

通常需要解开技术来获得二元或多个系统的单个组件的光谱。对 Marchenko、Moffat 和 Eenens (1998) 的移位加法算法的彻底分析表明，在许多情况下，线通量的再现效果不佳，并且会出现伪翼。讨论了这些差异的原因，并提出了一个新的解开包 QER20，它显着减少了这些错误并大大提高了性能。当应用于大量二进制 9 Sgr 时，我们的新代码产生的线通量与之前发表的那些明显不同，并导致我们将光谱分类修改为稍早的亚型：O3V((f +)) 用于初级和 O5V(( f)) 为次要的。