Actually, in order to be most useful you probably want to keep the tabular data as you have it with a separate record (or attribute) for each depth. Combining the data from each depth into a single field will limit the use or make it much more difficult. With it separate you can look at each depth as a different surface in interpolation or perform any other statistical analysis in a much more discrete manner.
I assume you have coordinates for a given point? If so you'll need to create points out of those xy coordinates using an XY Event Layer (covered at many other questions here). If you already have points as a shapefile or other geometry and you just need to get the analysis data as attributes you can move on to the next step. The points do need a common ID attribute that matches up to the Excel spreadsheet.
Once you have points that represent where all of the samples are you need to join the attributes to them. At the moment you'd have a one-to-many relationship between points and records. If you did a Join, the resulting attribute table would only show you the first returned record of the match. However if you then Export the results of that join you should get three points stacked on top of each other for each original point, each with a different depth record. Note the points and table need to be in a geodatabase for this to work. This does make manual selection or display in 2D more difficult.
An alternative would be to create a one-to-one relationship between points and table by renaming your points with a different scheme - for example instead of just three records called 1, have 1-1, 1-2, and 1-3. With this method, the xy coordinates would be duplicated for each record and could be put in as attributes to create with an XY Event all at once (you still end up with stacked points). If you already have the points, you'd have to copy them twice and similarly adjust the IDs to match the scheme you come up with in Excel before doing the join.
Yet another alternative, which would allow you to have a single set of points, would be to follow Richard's idea in his comment. Each point has an attribute for the pH at a given depth (three total) - the depths become columns, but the pH values are the attribute for each point in those columns - not separate columns themselves:
PointID pH00-20 ph20-40 ph40-60
1 4 5 7
2 ... ... ...
Having a single, non-stacked set of points would be easier to work with in 2D, and you can still create different surfaces out of it (if you're going there). Of course since you have depth ranges you don't really have enough to make a 3D surface with accurate depths, more just a pH map at some representative depth in that range.