Chapter 3 is devoted to be investigation of and . The determination of the geatest non-representable number, , seems to be a very difficult task already in the case . The exact values can be given only under special conditions, for which we show several examples in Chapter 2. Maybe, just these difficulties induced that the evolution of the topic brought the various estimations into the center of interest. In fact, in connection with these estimations ERDŐS introduced and began to study the extremal function in 1971. Since our main result is related to this function, we give a complete survey of the previous investigations about .

Particularly we stress DIXMIER's estimate from 1990 [4]:

**THEOREM 3.1.3. **

DIXMIER gave an even sharper upper bound, which has
many consequences and implies nearly all previously known results.

**THEOREM 3.2.5. **

where and

The exact results on can be divided into two
larger groups. For the smaller values of , the case
has been known long since, and later
LEWIN gave the exact value for in the early 70-es.
The cases and follow from Dixmier's theorems
[4], except for the numbers
and . For larger values of one could
obtain exact values only if
was not much greater than .
The most general result of this type was the theorem of ERDŐS and GRAHAM
from 1972 [6]:

**THEOREM 3.2.3. ***
Let and be positive integers,
then
*

LEV gave a different proof for this theorem, with
the weaker
assumption
. Our main result is a
generalization of this theorem, determining the exact value
of
for two residue classes mod [16]:

**THEOREM 3.3.1. ***
Let be integers such that
If
or then
*

Using our theorem we can compute also in some further special cases, for instance

**COROLLARY 3.4.2. ***
Let and be integers such that ,
If
then
*