We present LOFAR Low Band observations of the Boötes and 3C 295 fields .
Our images made at 34 , 46 , and 62 MHz reach noise levels of 12 , 8 , and 5 mJy beam ^ { -1 } , making them the deepest images ever obtained in this frequency range .
In total , we detect between 300 and 400 sources in each of these images , covering an area of 17 to 52 deg ^ { 2 } .
From the observations we derive Euclidean-normalized differential source counts .
The 62 MHz source counts agree with previous GMRT 153 MHz and VLA 74 MHz differential source counts , scaling with a spectral index of -0.7 .
We find that a spectral index scaling of -0.5 is required to match up the LOFAR 34 MHz source counts .
This result is also in agreement with source counts from the 38 MHz 8C survey , indicating that the average spectral index of radio sources flattens towards lower frequencies .
We also find evidence for spectral flattening using the individual flux measurements of sources between 34 and 1400 MHz and by calculating the spectral index averaged over the source population .
To select ultra-steep spectrum ( \alpha < -1.1 ) radio sources , that could be associated with massive high redshift radio galaxies , we compute spectral indices between 62 MHz , 153 MHz and 1.4 GHz for sources in the Boötes field .
We cross-correlate these radio sources with optical and infrared catalogues and fit the spectral energy distribution to obtain photometric redshifts .
We find that most of these ultra-steep spectrum sources are located in the 0.7 \lesssim z \lesssim 2.5 range .