最后我使用了下面的代码。它看起来很有效：

private List<Integer> getChildDocIds(IndexSearcher indexSearcher, int parentDocId) throws IOException {
    //Use a query in QueryBitSetProducer constructor which identifies parent docs
    BitSetProducer parentsFilter = new QueryBitSetProducer(new TermQuery(new Term("child", "N")));
    IndexReader indexReader = indexSearcher.getIndexReader();
    List<LeafReaderContext> leaves = indexReader.leaves();
    int subIndex = ReaderUtil.subIndex(parentDocId, leaves);
    LeafReaderContext leaf = leaves.get(subIndex);
    int localParentDocId = parentDocId - leaf.docBase;
    List<Integer> childDocs = new ArrayList<>();
    if (localParentDocId == 0) { 
        //not a parent, or parent has no children
        return childDocs;
    }
    int prevParent = parentsFilter.getBitSet(leaf).prevSetBit(localParentDocId - 1);
    for(int childDocIndex = prevParent + 1; childDocIndex < localParentDocId; childDocIndex++) {
        childDocs.add(leaf.docBase + childDocIndex);
    }
    return childDocs;
}

根据你的回答我做了一个C#端口的函数，看起来也是可行的。
我只需要得到嵌套的文档没有任何过滤器或quires。

private List<Lucene.Net.Documents.Document> GetNestedDocuments(IndexSearcher searcher, int parentDocId)
{
    List<Lucene.Net.Documents.Document> documents = new List<Lucene.Net.Documents.Document>();

    int subIndex = ReaderUtil.SubIndex(parentDocId, searcher.IndexReader.Leaves);
    var leaf = searcher.IndexReader.Leaves[subIndex];

    if (parentDocId > leaf.DocBase)
    {
        for (var childDocIndex = leaf.DocBase; childDocIndex < parentDocId; childDocIndex++)
        {
            var childDoc = searcher.Doc(childDocIndex);
            documents.Add(childDoc);
        }
    }

    return documents;
}