Context : Fragmentation and feedback are two important processes during the early phases of star formation .

Aims : Massive clumps tend to fragment into clusters of cores and condensations , some of which form high-mass stars .
In this work , we study the structure of massive clumps at different scales , analyze the fragmentation process , and investigate the possibility that star formation is triggered by nearby H II regions .

Methods : We present a high angular resolution study of a sample of massive proto-cluster clumps G18.17 , G18.21 , G23.97N , G23.98 , G23.44 , G23.97S , G25.38 , and G25.71 .
Combining infrared data at 4.5 , 8.0 , 24 , and 70 \mu m , we use few-arcsecond resolution radio- and millimeter interferometric data taken at 1.3 cm , 3.5 mm , 1.3 mm , and 870 \mu m to study their fragmentation and evolution .
Our sample is unique in the sense that all the clumps have neighboring H II regions .
Taking advantage of that , we test triggered star formation using a novel method where we study the alignment of the centres of mass traced by dust emission at multiple scales .

Results : The eight massive clumps , identified based on single dish observations , have masses ranging from 228 to 2279 M _ { \sun } within an effective radius of R _ { eff } \sim 0.5 pc .
We detect compact structures towards six out of the eight clumps .
The brightest compact structures within infrared bright clumps are typically associated with embedded compact radio continuum sources .
The smaller scale structures of R _ { eff } \sim 0.02 pc observed within each clump are mostly gravitationally bound and massive enough to form at least a B3-B0 type star .
Many condensations have masses larger than 8 M _ { \sun } at small scale of R _ { eff } \sim 0.02 pc .
We find that the two lowest mass and lowest surface density infrared quiet clumps with < 300 M _ { \sun } do not host any compact sources , calling into question their ability to form high-mass stars .
Although the clumps are mostly infrared quiet , the dynamical movements are active at clump scale ( \sim 1 pc ) .

Conclusions : We studied the spatial distribution of the gas conditions detected at different scales .
For some sources we find hints of external triggering , whereas for others we find no significant pattern that indicates triggering is dynamically unimportant .
This probably indicates that the different clumps go through different evolutionary paths .
In this respect , studies with larger samples are highly desired .