# CLRS's Fibonacci Heap size(x) analysis has a flaw?

In CLRS's Introduction to Algorithms 3rd edition P.525, when it analyzes the size($x$)'s lower bound, there is a sentence that I quote as "because adding children to a node cannot decrease the node’s size, the value of $s_k$ increases monotonically with $k$". But in fact, I think I can give a counterexample to show that $s_k$ does not necessarily increase monotonically with k.

In the following graph, the degree of the node with key 1 is 2, and there is no other node with degree of 2. So $s_2=8$. Similarly, $s_3=6$. But now $s_3$ is less than $s_2$ which means $s_k$ is not monotonically increasing with $k$ at all.

2 - 0 - 4 - 2 - 5 - 8 - 7 -  1
|               /  \
8              2    9
/ | \
10 14 16
|  |
11 15


The heap can be derived from an unordered binomial subtree by executing a series of cuts and cascading-cuts.

I want to know whether the above structure is a valid Fibonacci heap. If so, then it is also a valid counterexample.

In CLRS, $s_k$ is defined to be "the minimum possible size of any node of degree $k$ in any Fibonacci heap." Assuming your example data structure is a valid Fibonacci heap, your Fibonacci heap has a node of degree 2 with size 8 and a node of degree 3 with size 6. Thus, these are possible sizes that nodes of degree 2 and 3 can have but not lower bounds on their sizes.
Using Lemma 19.1 (as in the proof of Lemma 19.4), $s_2 = 3$ and $s_3 = 5$, monotonically increasing as CLRS claim.