Transmission Line vs. Lumped Inductor Choise in Distributed Amplifier Design

It is mentioned in Razavi book that the choice between transmission lines and lumped inductors in distributed amplifier design depends on the frequency and environment, and the transition from one to the other takes place at 20-40 GHz.
I want to design a DA in the Ka band (27-40GHz) , and the first question coming to mind is whether to use actual lines or realize them using lumped inductors. I can't find anything in any reference books or papers regarding on what basis the choice can be made in the 20-40GHz frequency band. I know the pros and cons of each, but I still can't choose specifically. Any relevant information could be helpful.

Lumped Inductors are not used in this band ( 27-40 GHz) so you should implement inductive components with Transmission Line elements.

I have some papers that have actually used lumped inductors in the mentioned band. Also, the Razavi reference I quoted points out that either of the two choices could be used in 20-40 GHz, and you have to decide considering other things.

is there a difference? a sort piece of transmission line acts like an inductor, and vice versa. At 40 ghz, you might be better able to model a transmission line than an actual chip inductor.

It's possible to create Lumped-Integrated Passive Components on Silicon with a proper process for 40-50 GHz but it's not really possible to find Lumped-Discrete Components on the market due to their parasitic elements and the size/mechanical/manufacturing constraints.

I think we're discussing on-chip implementations here. To me, the choice of lumped inductors vs. transmission lines depends mostly on the required chip area. Wideband distributed amplifiers at low frequency, as shown below, are built with inductors because transmission lines at these frequencies would be too large. But above 20GHz, the lines become reasonably small.

../imgqa/eboard/Antenna/rf-zeow0gbsfyf.jpg

Yes, there is a difference though they have almost the same behaviour. Transmission lines become too long at frequencies below 20GHz, and inductors are difficult to design as a scalable model at frequencies above 40GHz, but in the 20-4-GHz band I have no idea which is better.

I think I had to mention that I'm talking about on-chip inductors. You're right.

Yes, that's exactly right. But I'm looking for a more specific merit to help me judge which is better at my band. I'm currently in the cadence spectre design stage, and I don't know how much chip area I need yet. I've seen papers designing the DA with inductors in the mentioned band, and I've seen TL circuits as well. Neither of them have explained why they've preferred one choice to another. Maybe it was chip area as you said, but how do I answer the question on paper long before I'm concerned with designing an actual chip?

read (foundations of interconnect and microstrip design) book, chapter 1, 1.4.
if t-rise < 2.5 t-fall : t-line
t-rise >5 t-fall : rc model.