I did a project where the microcontroller (SAMD51), radio (NRF24L01) with accelerometer (very low power) and mems microphone (ditto) were powered from a solar cell and a very bright LED floodlight. This was to bridge an air gap to allow movement.
Anyway, I looked at solar cells then and although we went with an IXYS model (only available from digikey AFAIK) I was disappointed with their performance when the illumination was pretty much anything less than full sunlight.
On a subsequent project I briefly tested flex solar cells from https://flexsolarcells.com/ and they did much better as a function of their spec in lower light levels (as in by a window in daylight but not full sun). May be worth a look for anyone keen to minimise their photovoltaic area.
Lastly, anyone who hasn’t used solar cells before should pay attention to the current/voltage curve on the cell, which usually shows the maximum power point. What these datasheets don’t show is that the current at which voltage drops drastically is variable, depending on illumination. For this reason, to get the best out of the cell you need to make sure that you don’t drag it into the condition of effectively open-circuit, pulling your voltage right down, even if you don’t implement maximum power point tracking. I have a solar powered night light in our bathroom which charges supercapacitors using an ATtiny402 that runs a charge pump to boost the peak voltage to 5.5V and a second ATtiny handles the PIR sensor, LDR and LED. A charge pump isn’t the most efficient way of using the solar power but it’s nice and simple. I had a fascinating chat with a guy in New Zealand who used to implement a battery charger from solar cells on agricultural electronics in the '90s using an inductor and the GPIO protection diodes on the PIC to charge the battery via the VCC connection to the PIC.
If I were going for this challenge, I think LCD display would be a good start for lower power and possibly a different chip or at least a slower clock rate.