1: \begin{abstract}

2: We investigate boundedness of the relative errors of derivatives of 

3: $\X=\X(\P) \in \R^n$

4: defined implicitly by the vanishing residual $\r=R(\X,\P)=0,$ $r \in \R^n,$ $\P \in \R^m$ in dependence of a given absolute error

5: $\DX \in \R^n$ in the primal solution $\X$. 

6: First-order error analysis is employed to prove conditions for boundedness.

7: 

8: 	Different types of implicit functions yield different directional derivatives (also: tangents) 

9: 	$\dot \X=\dot F(\X,\dot \P)$ and adjoint directional derivatives (also: adjoints) $\bar \P=\bar F(\X,\bar \X).$ 

10: 	For example,

11: the absolute error in tangents

12: $\|\Delta \dot \X\|=\|\dot F(\X+\Delta \X,\dot \P)-\dot F(\X,\dot \P)\|$ yields the corresponding relative error 

13: 	$$\|\delta \dot \X\|=\frac{\|\Delta \X\|}{\|\X\|}=\frac{\|\dot F(\X+\Delta \X,\dot \P)-\dot F(\X,\dot \P)\|}{\|\dot \X\|} \approx \frac{\|\dot F'(\X,\dot \P)\cdot \Delta \X\|}{\|\dot F(\X,\dot \P)\|} $$ 

14: assuming convergence of the Taylor series expansion of $F$ to the true value of

15: 	the tangent in a sufficiently large 

16: neighborhood of all points $\X$ of interest.

17: The following questions will be answered:

18: \begin{enumerate}

19: \item 

20: 	$\exists~\X, \dot \P : \; \lim_{\|\dot F(\X,\dot \P)\| \rightarrow 0} \|\delta \dot \X\|=\infty$ while $\|\delta \X\| < \infty$?

21: \item Under which conditions is $\|\delta \dot \X\|<\infty?$

22: \end{enumerate}

23: Similar results will be derived for adjoints. We focus on linear and nonlinear systems of equations including first-order optimality conditions in the context 

24: of smooth nonlinear optimization.

25: \end{abstract}

26: